A Hospice-Hospital Partnership: Reducing Hospitalization Costs and 30-Day Readmissions among Seriously Ill Adults
John C. Tangeman, MD, FACP, 1 Carole B. Rudra, PhD, MPH,2 Christopher W. Kerr, MD, PhD, 1 and Pei C. Grant, PhD1
Abstract Background: Inpatient palliative care (IPC) has been associated with numerous clinical bene?ts. Observational and randomized studies of cost savings
associated with IPC provide con?icting results, and the association with readmission is not well understood. Objective: We aimed to estimate the in?uence of IPC on
hospitalization costs and readmission rates. Methods: We measured hospitalization costs and 30-day readmission rates among 1004 patients who received IPC at two
western New York hospitals in 2012. Using propensity score matching, we compared outcomes among patients receiving palliative care with those among 1004 similar adults
who were hospitalized during the same period and did not receive palliative care. Results: On average, cost per admission was $1,401 (13%) lower among patients
receiving palliative care than comparison patients (p<0.05). Cost reductions were evident within intensive care and laboratory services. Readmission rates were signi?
cantly lower among palliative care patients discharged with hospice care (1.1%) than comparison patients (6.6%), but signi?cantly higher among palliative care patients
discharged to other locations (12.1%). Conclusions: Receipt of IPC appears to reduce hospitalization costs among adult western New Yorkers. Furthermore, care
coordinated with postdischarge hospice services appears to substantially reduce the likeli- hood of readmission.
Introduction There is a growing evidence base supporting the ob- servation that inpatient palliative care (IPC) consultation services reduce length of stay (LOS) and
hospital costs1–7 while concurrently increasing patient and provider satisfac- tion,8,9 improving symptom control,10 and increasing ad- vance directive
completion.11,12 In addition, studies have shown that IPC services reduce the likelihood of intensive care unit (ICU) readmission.13–15 There is, however, scant
evidence supporting ongoing cost savings for those patients who survive to discharge. A recent retrospective study showed that receipt of hospice or home-based
palliative care was associated with a lower rate of readmission when com- pared with usual care.16 Another propensity-matched study
examining home-based palliative care compared with usual care showed a lower probability of hospital readmission for those patients receiving palliative care.17 There
are myriad new models of care emerging as health systems adapt to the Patient Protection and Affordable Care Act. Cost-effective postacute care models are becoming
increasingly important as accountable care organizations mature and hospitals face ?nancial penalties for 30-day readmissions. In many communities, hospice providers
are the local experts in providing palliative and end-of life care and, as pointed out by Meier and colleagues, hospitals and hospice providers should partner to
extend the care contin- uum into the home.18 The impact of palliative care programs administered by hospice providers has not been well exam- ined. Therefore, in this
study, a propensity-matched cohort
1Center for Hospice and Palliative Care, Cheektowaga, New York. 2Rudra Research, LLC, Buffalo, New York. Accepted March 9, 2014.
JOURNAL OF PALLIATIVE MEDICINE Volume 17, Number 9, 2014 ª Mary Ann Liebert, Inc. DOI: 10.1089/jpm.2013.0612
1005
was used to examine the hospitalization costs and 30-day readmission rate of patients seen by a hospice- and hospital- sponsored IPC service, regardless of payer
source.
Methods
Setting and study population
This was a retrospective cohort study of adult patients who were admitted to two New York State hospitals (Millard Fill- more Suburban Hospital, a 265-bed facility in
Williamsville, NY, or Buffalo General Medical Center, a 501-bed facility in Buffalo, NY) between January 1, 2012 and December 31, 2012 and who received IPC services
during their hospitalizations. Enrolled palliative care patients received inpatient consulta- tion from a palliative care team consisting of a palliative care
physician or nurse practitioner, a hospice nurse liaison, and a hospital-based social worker. Palliative care patients in the upper ?fth percentile of LOS were
excluded because these patients maynotbe typicalrecipientsofpalliativeservices.IPC patients were matched to nonparticipating adult patients who were hospitalized at
either facility during the same period. Matching was based on several demographic and admission characteristics routinely collected in hospital databases, de- scribed
in detail below. This study was approved by the Uni- versity at Buffalo Health Sciences Institutional Review Board.
Outcomes and patient characteristics
Outcomes of interest were hospitalization costs and 30-day hospital readmissions. Total cost per admission and average cost per day within each admission were
calculated overall and within subgroups according to service type, as captured by Universal Billing form 92 (UB-92) revenue codes in ad- ministrative databases. Cost
subgroups included intensive care (UB-92 200–209), diagnostic imaging (320–329, 341, 350–359, 400–409, 610–619), laboratory (300–319), and pharmacy including
intravenous therapy (250–269). Read- missions were de?ned as admissions to any of the following hospital units between 1 and 30 days after discharge from the ?rst
hospital stay: ICU, emergency room, telemetry cardiac, neuro-stepdown, medical stroke, intermediate care, cardiac care, and general nursing. Admissions to palliative
care and rehabilitation units were not considered readmissions. Several patient characteristics were examined. Demo- graphic characteristics at enrollment included
age, gender, marital status, race, and insurance status. Admission char- acteristics included the specialty of the attending physician, primary diagnosis, all patient
re?ned diagnosis related group (APR-DRG19) illness severity classi?cation, LOS, discharge status, and utilization of a hospice ‘‘swing bed.’’ These var- iables were
categorized as shown in Table 1. All information used in this analysis was extracted from hospital databases.
Propensity score matching
Propensity scores were used to identify a comparison group of patients who did not receive palliative care but were otherwise comparable with regard to measured
characteris- tics.20 In?uences of palliative care participation may differ between patients who were discharged alive and those who died while hospitalized. Therefore,
two propensity scores were estimated: one for patients discharged alive and another for patients who died in the hospital.4 Both scores were
estimated using a logistic regression model with receipt of palliative care services as the outcome and the following characteristics as predictors: age, gender,
marital status, race, insurance status, attending physician specialty, primary di- agnosis, illness severity score, and LOS. These characteris- tics were modeled using
sets of indicator variables de?ned by categories shown in Table 1. Within strata de?ned by vital status at discharge, palliative care patients were matched 1:1 to
comparison patients using the nearest neighbor method.20 The largest acceptable difference in propensity scores be- tween an IPC patient and his or her matched
comparison patient was 20% of the score’s standard deviation (SD).
Statistical analysis
Distributions of characteristics between IPC patients and propensity-matched non-IPC patients were compared. Dif- ferences in these distributions were tested using the
v2 sta- tistic. Averagecosts withinbothgroupsandtheaverage (95% con?dence interval [CI]) difference in costs between the two groups were calculated. Differences in
overall costs per ad- mission and per-day and subgroup-speci?c costs per admis- sion were also calculated. The proportions of readmissions within 30 days of discharge
were compared using the v2 statistic. Statistical signi?cance was de?ned as p<0.05 or a CI excluding the null difference of zero.
Cost differences within subgroups
In secondary analyses, cost differences were examined to determine ifdifferences were strongerafterrather thanbefore the initial palliative care consult, as would be
expected if IPC reduced inpatient costs. Differences in average cost per day between IPC patients and comparison patients were plotted according to day of initial
consult (IPC patients) or reference day(comparisonmembers). Thereference daywas de?nedas the median day of initial consult among patients receiving palliative care
within categorized LOS, ranging from day 6 among patients with 9- to 16-day stays to day 23 among patients with 30- to 43-day stays. A priori, this analysis was
restricted to costs from 4 days before the initial consult through 4 days afterward within the subgroup of patients who were admitted for 9 days or longer. Costs
differences were examined within subgroups de- ?ned by payer (Medicare, Medicare HMO, Medicaid, and commercial/other) for consistency. Finally, to examine whether
relationships between palliative care and cost per day differed according to duration of hospitalization, aver- age cost differences within subgroups de?ned by LOS
were calculated.
Sensitivity analyses
Inclusion of characteristics in the propensity score that are in?uencedbytheexposureofinterestmaycauseovermatching and biased estimates of the exposure’s effects. For
instance, if IPC consults in?uence the LOS, inclusion of LOS in the pro- pensity score is incorrect. However, the reverse relationship may exist: Patients with longer
hospital stays may be more likely to be approached for or be amenable to palliative care services. LOS was positively associated with likelihood of receipt of IPC in
this study population. LOS was therefore included as a covariate in the propensity score. In post hoc
1006 TANGEMAN ET AL.
analyses, the impact of this decision was evaluated by using a comparison group identi?ed from scores based on all predic- tors listed above, excluding LOS (using
separate scores for patients discharged alive and those who died in the hospital). Similarly, receiving IPC services may be associated with likelihood of death during
hospitalization. Therefore, the sensitivity of the results was also examined by using a com- parison group identi?ed froma singlepropensityscore thatdid not include
vital status at discharge as a stratifying character- istic or as a predictor.
Results
During the study period, 1116 individuals received pallia- tive care services during an inpatient stay. Thirty-?ve of these received IPC during multiple hospital
stays; only the ?rst stay was includedinthisanalysis.About one-third (356) ofpatients died during their hospitalizations. This analysis excluded 54 participantswho
were admitted for 45to 453 days. Another 58 participants, 48 of whom died while hospitalized, were also excluded due to a lack of well-matched comparison patients.
These 58 excluded participants had longer hospitalizations than the deceased participants included in the analysis (17.2
daysversus10.3days,onaverage),butwereotherwisesimilar. The?nalanalyticsampleincluded1004IPCpatients,ofwhom 288 died in the hospital, and their propensity-matched com-
parison group of equal size. Characteristics at admission were similar between IPC pa- tients and matched comparison patients (Table 1). The distri- butions of
categorized LOS were statistically signi?cantly
Table 1. Frequency and Percent Distributions of Characteristics at Admission among Inpatient Palliative Care (IPC) Patients and Comparison Patients
IPC patients (n=1004)
Comparison patients (n=1004)
Characteristic N % N %
Age (years) 23–29 10 1.0 3 0.3 30–39 6 0.6 7 0.7 40–49 32 3.2 33 3.3 50–59 92 9.2 98 9.8 60–69 130 13.0 121 12.1 70–79 249 24.8 246 24.5 80–89 333 33.2 327 32.6 90–102
152 15.1 169 16.8 Gender Female 591 58.9 538 53.6 Male 413 41.1 466 46.4 Marital status Married or with life partner 409 40.7 416 41.4 Widowed 344 34.3 342 34.1
Divorced or separated 96 9.6 86 8.6 Single 149 14.8 153 15.2 Unknown 6 0.6 7 0.7 Race White 851 84.8 875 87.2 Black 122 12.2 104 10.4 Other 28 2.8 25 2.5 Unknown 3 0.3
0 0.0 Insurance status Medicare 379 37.7 378 37.7 Medicare HMO 449 44.7 452 44.9 Medicaid 19 1.9 10 1.0 Medicaid HMO 31 3.1 37 3.7 Commercial 112 11.2 111 11.1 Self-
pay 3 0.3 5 0.5 Other 11 1.1 11 1.1 Attending physician specialty Family medicine or pediatrics 136 13.6 134 13.4 Internal medicine 478 47.6 455 45.3 Reference 329
32.8 354 35.3 Surgery 37 3.7 35 3.5 Other 24 2.4 26 2.6 Primary diagnosis Blood 14 1.4 13 1.3 Circulatory 212 21.1 231 23.0 Digestive 65 6.5 66 6.6 Endocrine 64 6.4 62
6.2 Genitourinary 54 5.4 45 4.5 Infection/parasite 84 8.4 85 8.5 Injury/poisoning 46 4.6 40 4.0 Musculoskeletal 32 3.2 25 2.5 Neoplasm 42 4.2 41 4.1 Respiratory 132
13.2 137 13.7 Symptoms, signs, conditions 224 22.3 230 22.9 Other 35 3.5 29 2.9 APR-DRG illness severity score
(continued)
Table 1. (Continued)
IPC patients (n=1004)
Comparison patients (n=1004)
Characteristic N % N %
1 (minor) 7 0.7 113 0.5 2 (moderate) 113 11.3 102 10.2 3 (major) 404 40.2 443 44.1 4 (extreme) 480 47.8 454 45.2 Length of stay (days)a 1–2 114 11.4 170 16.9 3–5 244
24.3 255 25.4 6–8 195 19.4 156 15.5 9–16 270 26.9 243 24.0 17–29 132 13.1 129 12.9 30–43 49 4.8 51 5.1 Discharge statusa Home, self-care 76 7.6 260 25.9 Inpatient
facility 8 0.8 36 3.6 Skilled care 289 28.8 409 40.7 Hospice at home 136 13.5 8 0.8 Hospice facility 207 20.6 3 0.3 Died in hospital 288 28.7 288 28.7 Utilization of
hospice ‘‘swing bed’’a 105 10.5 4 0.4 av2 p<0.05. APR-DRG, all patient re?ned diagnosis related group; HMO, health maintenance organization.
REDUCING COSTS AND 30-DAY READMISSIONS 1007
different between the two groups: IPC patientswere lesslikely to have stays <3 days than comparisonpatients (11.3%versus 16.9%, respectively). However, average LOS was
similar be- tween IPC patients (mean–SD: 10.3–0.3 days) and com- parison patients (9.8–0.3 days). Distributions of discharge status differed signi?cantly between the
two groups: IPC pa- tients were considerably more likely to be discharged to hos- pice care at home (13.5% versus 0.8%) or a hospice facility (20.6% versus 0.3%) than
comparison participants; distribu- tions of those who died while hospitalized were identical by design. IPC patients were also more likely to utilize a hospice swing
bed (10.5% versus 0.4%). Cost per admission was $1,401 lower, on average, among IPC patients than comparison patients (95% CI: $322 to $2,481 lower) (Table 2). The
cost reduction was somewhat more pronounced among patients who died in the hospital ($1,824 lower among IPC patients) than those discharged alive ($1,231 lower). IPC
was also associated with lower cost per day: $228 lower overall, $164 lower among those who died, and $387 lower among live discharges, on average. All differences
except admission costs for those who died in the hospital were statistically signi?cant. Costs for intensive care and laboratory services were lower among IPC patients
than matched comparison patients ($492 and $109 lower, respectively) (Table 3). Diagnostic imaging and pharmacy costs were not strongly associated with the re- ceipt
of IPC services ($16 and $20 lower among participants, respectively). Relationships between IPC and service-speci?c costs were generally similar between patients
discharged alive and those who died in the hospital. The difference in cost per day between IPC patients and comparison patients was most apparent on and after the day
of initial consult (Fig. 1). Average cost per day was $101 lower in the 1 to 4 days before the initial consult and $233 lower 0 to 4 days after the consult among IPC
patients versus comparison patients hospitalized for 8 to 43 days. A similar pattern was observed among patients hospitalized for 1 to 7 days (data not shown). This
pattern was more pronounced among patients who died in the hospital than those dis- charged alive (data not shown). Overall, the likelihood of readmission did not
differ be- tween IPC patients and comparison patients (6.7% versus 6.6%, p=0.92). However, IPC patients discharged to a hos- pice facility, hospice swing bed, or
hospice at home were signi?cantly less likely to experience a readmission than
comparison patients (1.1% versus 6.6%, p<0.01). IPC pa- tients discharged to home, skilled care, and other facilities had a signi?cantly higher risk of readmission
than compari- son patients (12.1% versus 6.6%, p<0.01). The relationship between IPC and cost per day was robust to our choice of propensity score estimator. Results
did not change when using either scores unadjusted for LOS or a single score estimated from both live discharges and those who died in the hospital (data not shown).
The relationship between cost per day and IPC was also generally consistent withinsubgroupsde?nedbyLOSandpayer(datanotshown).
Discussion
This study adds to the evidence that IPC results in lower hospitalization cost regardless of vital status at discharge. The average cost reduction was 13% in this
study, similar to the 11% reduction among New York State Medicaid recipients
Table 2. Average Hospitalization Cost and Average (95% CI) Difference in Cost between Inpatient Palliative Care (IPC) Patients and Comparison Patients
Cost
Palliative care participants
Comparison patients
Palliative versus comparison
Cost per admission ($) All patients 9713 11,114 -1,401 (-2,481 to -322) Discharged alive 8421 9653 -1,231 (-2,205 to -257) Died in hospital 12,924 14,748 -1,824 (-
4,636 to 988) Cost per day ($) All patients 991 1,219 -228 (-316 to -140) Discharged alive 844 1,009 -164 (-234 to -94) Died in hospital 1,742 1,355 -387 (-624 to –
151)
Both groups comprised 1004 individuals: 716 discharged alive and 288 who died in the hospital. CI, con?dence interval.
Table 3. Average Service-Speci?c Costs and Average (95% CI) Difference in Costs per Admission between Inpatient Palliative Care (IPC) Patients and Comparison Patients:
Overall and According to Discharge Status
Cost per admission ($)
IPC patients
Comparison patients
Palliative versus comparison
Intensive care All patients 2148 2640 -492 (-985 to 2) Discharged alive 1397 1885 -488 (-878 to -97) Died in hospital 4020 4520 -500 (-1879 to 879) Diagnostic imaging
All patients 288 304 -16 (-45 to 12) Discharged alive 283 253 -30 (-62 to 1) Died in hospital 376 357 20 (-41 to 80) Laboratory All patients 614 722 -109 (-175 to -42)
Discharged alive 520 616 -96 (-155 to -37) Died in hospital 847 987 -139 (-312 to -34) Pharmacy All patients 599 620 -20 (-109 to 68) Discharged alive 528 546 -18 (-98
to 62) Died in hospital 777 802 -25 (-259 to 208)
Both groups comprised 1004 individuals: 716 discharged alive and 288 who died in the hospital. CI, con?dence interval.
1008 TANGEMAN ET AL.
observed by Morrison and coworkers.4 Relative cost savings shown herewerealso similar tothe 14% reduction observedin a matched cohort study in a California academic
medical center.3 In a matched cohort analysis of patients in two aca- demic centers from 2005 to 2008, cost savings were 13% greater among IPC patients; however, cost
savings were not apparent among patients with hospitalizations of 30 days or more.1 In contrast, we found no difference in cost savings according to LOS. In two
separate observational studies of veterans receiving care at several hospitals from 2002 to 2006, Penrod et al. observed lower costs per day for patients re- ceiving
palliative care ($239 and $464 lower in their ?rst and second studies, respectively).14,15 In their more recent study,15 nursing, laboratory, and radiology costs were
lower among palliative care patients, although pharmacy costs were higher. There have been fewer examinations of the relationship between palliative care and 30-day
risk of readmission. En- guidanos and colleagues16 showed that 10% of patients who received an IPC consultation were readmitted within 30 days of discharge. The rate
of readmission was much lower within the subgroups that received outpatient hospice or palliative care follow-up after discharge (4.6% and 8.3%, respectively) than
those patients discharged without in-home care (25.7%). The study was undertaken at a managed care medical center with a well-supported home-based palliative care
team. Therefore, the results may not be applicable to many areas of the United States with a relative lack of home-based palliative care services outside of the
Medicare Hospice Bene?t. Ran- ganathan and coworkers,17 using a propensity matched cohort, recently showed a reduced risk of 30-day readmission via a home-based
palliative care program. Neither IPC consultation nor hospice referral rates were examined in this study. In this current study, the overall likelihood of readmission
did not differ between IPC patients and comparison patients
(6.7% versus6.6%).IPC patientsdischargedwithhospice care athomeortoahospicefacility,however,hadamuchlower30- day readmission rate (1.1%) than IPC patients discharged
to home, skilled care, or other facilities without hospice support (12.1%). Our study adds to the emerging evidence base sug- gesting that signi?cant decreases in 30-
day readmissions among IPC patients who survive to discharge can generally only be achieved via discharge with hospice care or home- based palliative care.16 In many
communities, hospice pro- grams are often the sole provider of home-based palliative care, usually under the Hospice Medicare Bene?t. This study
suggeststhatinpatientpalliativeprogramsshouldworkclosely with hospice programs when discharging eligible patients. This, and other studies,16,17,21,22 suggest the need
for novel funding sources to support robust home-based palliative care offerings, outside the Hospice Bene?t, to expand the palliative continuum to those patients
still receiving active treatments or who have a prognosis >6 months. Preliminary data (unpub- lished) from our center show signi?cant cost reductions, hos- pital and
emergency department avoidance, and more timely hospicereferralamongpatientsinahome-basedpalliativecare program ?nancially supported by a local commercial and Medicare
Advantage insurance company. Our study population is larger than that examined in many previous studies, affording us power to detect differences in service-speci?c
costs and to examine costs per day according to the timing of the initial consultation. The diversity of the analytic population, for example with regard to payer and
primary diagnosis, is also a strength of this study. Because of our exclusion criteria, these results may not be generalizable to patients with longer
hospitalizations. Furthermore, we were unable to identify appropriate comparison patients for about 5% of the patients. However, these exclusions were a small
proportion of the total population of IPC patients.
FIG. 1. Average cost per day for inpatient palliative care (IPC) patients (black bars) and comparison patients (white bars) according to day from initial consult or
reference day. Reference day within comparison patients is de?ned as the median day
betweenadmissionandinitialconsultamongIPCpatientswithsimilarlengthofstay;seetextfordetails.Figureisbasedondatafrom 451 palliative care participants and 423 comparison
patients who were hospitalized for 9 to 43 days and discharged alive or dead.
REDUCING COSTS AND 30-DAY READMISSIONS 1009
Therefore, tothe extentthat the patients at these twohospitals during the study period are representative of western New Yorkers, these results are likely applicable
to the regional and possibly national population of adults who would choose to participate in IPC. Similarities in characteristics at admission suggest that palliative
care participants were well matched to comparison patients. Additionally, our ?ndings were robust to covariates included in the propensity score, suggesting that cost
savings associated with palliative care were not mediated by LOS or vital status at discharge. Finally, cost differences between palliative care participants and
comparison patients were strongest at and after the day of initial consult. This temporal speci?city supports a causal relationship. However, as with any observational
study, we cannot discount the possibility that unmeasured characteristics may have in?uenced both the likelihood of participation in palliative care and outcomes of
interest. Such characteristics could account for part or all of the cost reductions and differences in readmission rates ob- served here. Furthermore, the higher rate
of readmission among IPC patients who were not discharged to hospice services mightre?ectaselectionbias amongtheIPCpatients; the fact that a palliative care consult
was requested, and hospice services declined, might indicate a patient, or family, with more aggressive, unrealistic goals. These limitations notwithstanding, these
results suggest that participation in IPC reduces hospitalization costs and, combined with post- discharge hospice coordination, reduces the likelihood of readmission
among adult western New Yorkers.
Acknowledgments
The authors gratefully acknowledge Joseph A. Bach and William D. Riemer of the Center for Hospice & Palliative Care for the coordination and acquisition of data needed
to complete the study. The authors also wish to thank Carol Kopacz and Jennifer Graff of Kaleida Health (Buffalo, NY) for the provision of data, without which the
study could not have been accomplished. Further, the authors thank Debra L. Luczkiewicz, MD, and Elizabeth Marks, MS, for editorial review of the manuscript.
Author Disclosure Statement
No competing ?nancial interests exist.
References
1. Albanese TH, Radwany SM, Mason H, Gayomali C, Dieter K: Assessing the ?nancial impact of an inpatient acute palliative care unit in a tertiary care teaching
hospital. J Palliat Med 2013;16:289–294. 2. Armstrong B, Jenigiri B, Hutson SP, Wachs PM, Lambe CE: The impact of a palliative care program in a rural Appalachian
community hospital: A quality improvement process. Am J Hospice Palliat Care 2013;30:380–387. 3. Ciemins EL, Blum L, Nunley M, Lasher A, Newman JM: The economic and
clinical impact of an inpatient palliative care consultation service: A multifaceted approach. J Pal- liat Med 2007;10:1347–1355. 4. Morrison RS, Dietrich J, Ladwig S,
et al.: Palliative care consultation teams cut hospital costs for Medicaid bene?- ciaries. Health Aff 2011;30:454–463. 5. Smith TJ, Coyne P, Cassel B, Penberthy L,
Hopson A, Hager MA: A high-volume specialist palliative care unit
and team may reduce in-hospital end-of-life care costs. J Palliat Med 2003;6:699–705. 6. Starks H, Wang S, Farber S, Owens DA, Curtis JR: Cost
savingsvarybylengthofstayforinpatientsreceivingpalliative care consultation services. J Palliat Med 2013;16:1215–1220. 7. Wu FM, Newman JM, Lasher A, Brody AA: Effects
of initiating palliative care consultation in the emergency department on inpatient length of stay. J Palliat Med 2013. 8. Meo N, Hwang U, Morrison RS: Resident
perceptions of palliative care training in the emergency department. J Palliat Med 2011;14:548–555. 9. O’Mahony S, Blank AE, Zallman L, Selwyn PA: The bene?ts of a
hospital-based inpatient palliative care con- sultation service: Preliminary outcome data. J Palliat Med 2005;8:1033–1039. 10. Laguna J, Goldstein R, Allen J, Braun W,
Enguidanos S: Inpatient palliative care and patient pain: Pre- and post- outcomes. J Pain Symptom Manage 2012;43:1051–1059. 11. Sacco J, Deravin Carr DR, Viola D: The
effects of the Palliative Medicine Consultation on the DNR status of African Americans in a safety-net hospital. American J Hospice Palliat Care 2013;30:363–369. 12.
Zaide GB, Pekmezaris R, Nouryan CN, et al.: Ethnicity, race, and advance directives in an inpatient palliative care consultation service. Palliat Support Care
2013;11:5–11. 13. Gade G, Venohr I, Conner D, et al.: Impact of an inpatient palliative care team: A randomized control trial. J Palliat Med 2008;11:180–190. 14.
Penrod JD, Deb P, Dellenbaugh C, et al.: Hospital-based palliative care consultation: Effects on hospital cost. J Palliat Med 2010;13:973–979. 15. Penrod JD, Deb P,
Luhrs C, et al.: Cost and utilization outcomes of patients receiving hospital-based palliative care consultation. J Palliat Med 2006;9:855–860. 16. Enguidanos S,
Vesper E, Lorenz K: 30-day readmissions among seriously ill older adults. J Palliat Med 2012;15: 1356–1361. 17. Ranganathan A, Dougherty M, Waite D, Casarett D: Can
palliative home care reduce 30-day readmissions? Results of a propensity score matched cohort study. J Palliat Med 2013;16:1290–1293. 18. Meier DE, Beresford L:
Hospitals and hospices partner to extend the continuum of palliative care. J Palliat Med 2007;10:1231–1235. 19. Pilotto A, Scarcelli C, D’Ambrosio LP, et al.: All
patient re?ned diagnosis related groups: a new administrative tool for identifying elderly patients at risk of high resource consumption. J Am Geriatr Soc
2005;53:167–168. 20. RubinDB:Estimatingcausaleffectsfromlargedatasetsusing propensityscores.AnnInternMed1997;127(8Pt2):757–763. 21. Brumley R, Enguidanos S, Jamison P,
et al.: Increased satisfaction with care and lower costs: Results of a ran- domized trial of in-home palliative care. J Am Geriatr Soc 2007;55:993–1000. 22. Brumley
RD, Enguidanos S, Cherin DA: Effectiveness of a home-based palliative care program for end-of-life. J Palliat Med 2003;6:715–724.
Address correspondence to: Pei C. Grant, PhD Center for Hospice and Palliative Care 225 Como Park Boulevard Cheektowaga, NY 14227
E-mail: pgrant@palliativecare.org
1010 TANGEMAN ET AL.
Copyright of Journal of Palliative Medicine is the property of Mary Ann Liebert, Inc. and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 1
O R I G I N A L R E S E A R C H Effect of post-discharge follow-up care on re- admissions among US veterans with congestive heart failure: a rural-urban comparison
KJ Muus1, A Knudson1, MG Klug1, J Gokun2, M Sarrazin2, P Kaboli2 1University of North Dakota Center for Rural Health, Grand Forks, North Dakota, USA 2VA Office of
Rural Health, Midwest Rural Health Resource Center, Iowa City VA Medical Center, Iowa City, Iowa, USA
Submitted: 4 July 2009; Revised: 7 April 2010; Published: 8 June 2010
Muus KJ, Knudson A, Klug MG, Gokun J, Sarrazin M, Kaboli P
Effect of post-discharge follow-up care on re-admissions among US veterans with congestive heart failure: a rural-urban comparison Rural and Remote Health 10: 1447.
(Online), 2010
Available from: http://www.rrh.org.au
A B S T R A C T
Introduction: Hospital re-admissions for patients with congestive heart failure (CHF) are relatively common and costly occurrences within the US health
infrastructure, including the Veterans Affairs (VA) healthcare system. Little is known about CHF re-admissions among rural veteran patients, including the effects of
socio-demographics and follow-up outpatient visits on these re- admissions. Purpose: To examine socio-demographics of US veterans with CHF who had 30 day potentially
preventable re- admissions and compare the effect of 30 day VA post-discharge service use on these re-admissions for rural- and urban-dwelling veterans. Methods: The
2005-2007 VA data were analyzed to examine patient characteristics and hospital admissions for 36 566 veterans with CHF. The CHF patients who were and were not re-
admitted to a VA hospital within 30 days of discharge were identified. Logistic regression was used to examine and compare the effect of VA post-acute service use on
re-admissions between rural- and urban-dwelling veterans. Results: Re-admitted veterans tended to be older (p=.002), had disability status (p=.024) and had longer
hospital stays (p<.001). Veterans Affairs follow-up visits were negatively associated with re-admissions for both rural and urban veterans with CHF (ORs
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 2
0.16–0.76). Rural veterans aged 65 years and older who had VA emergency room visits following discharge were at high risk for re-admission (OR=2.66). Conclusions:
Post-acute follow-up care is an important factor for promoting recovery and good health among hospitalized veterans with CHF, regardless of their rural or urban
residence. Older, rural veterans with CHF are in need of special attention for VA discharge planning and follow up with primary care providers.
Key words: heart failure, rural populations, veterans.
Introduction
Congestive heart failure (CHF) affects an estimated 5 million Americans and approximately 550 000 people are diagnosed with CHF annually1. Congestive heart failure is
the principal cause of death for approximately 400 000 persons annually in the USA2. The prevalence of CHF is expected to rise in future years due to several factors,
including higher rates of cardiovascular disease and increased life expectancy resulting from advances in medical treatment and technology. Major clinical risk factors
for CHF include advancing age, male sex, hypertension, myocardial infarction, diabetes mellitus, valvular disease and obesity3-13.
Congestive heart failure is the most common diagnosis among hospitalized US Medicare patients14 and is associated with six-month hospital re-admission rates of more
than 40%15. Hospital re-admissions may indicate one factor or a combination of factors including: poor in-hospital care; insufficient discharge planning; uncoordinated
transition care; and inadequate post-discharge and follow-up care16-19. The Medicare Payment Advisory Commission (MedPAC) has recommended public reporting of
hospital-specific re- admission rates, with CHF as a priority condition19. In response to this recommendation, the Centers for Medicare and Medicaid (CMS) developed a
30 day risk-standardized re-admission measure for CHF, designed to measure and improve patient care quality and decrease costs20.
Several studies have addressed socio-demographic and/or health factors related to re-admissions among CHF
patients21-28; however, these studies had one or more of the following limitations: use of data from small geographic areas; incorporating definitions of re-admission
that did not exclude re-hospitalizations due to other health conditions or planned stays (eg hospitalizations that were a part of the treatment regimen associated with
the initial hospitalization); or use of variant, non-standard and extended post-discharge timeframes (eg 6 months; 1 year) for defining re- admission29.
Some of these studies have examined veterans, a group that has been found to be at higher risk for hospital re- admissions30,31. Regarding patients’ rurality of
residence, one study involving older (ie =65 years) veterans found that rural-based patients were slightly more likely than urban- based patients to have unplanned 30
day re-admissions for all conditions combined and for several diagnostic categories, including circulatory disorders32. No studies were found which address demographic
and health predictors of 30 day hospital re-admissions for CHF among veterans. Also, no studies were found which addressed the relationship between veterans’ rurality
of residence, use of post-acute care and CHF re-admissions. The objectives of this study were to derive 30 day CHF potentially preventable re- admission prevalence,
delineate a socio-demographic profile of re-admitted CHF patients and compare the effect of post- acute VA physician service use on CHF re-admissions among rural- and
urban-dwelling veterans.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 3
Methods
Data sources
Subjects were identified using the VA Patient Treatment File (PTF) from the Department of VA’s Austin Automation Center, a file that contains information on inpatient
encounters in all VA hospitals, including demographics, income level, VA eligibility status (determined by several factors including the nature of a veteran’s
discharge from military service, length of service and VA adjudicated disabilities [commonly referred to as service-connected disabilities]), dates of admission and
discharge, primary/secondary diagnosis and procedure codes (ie International Classification of Diseases, Ninth Revision- Clinical Modification [ICD-9-CM]; current
procedural terminology [CPT]), admission source (eg transfer from another hospital, emergency room [ER], direct admission), hospital units where care was provided (eg
medicine, surgery, intensive care unit) and discharge disposition. To complement the identification of deaths that occurred in the hospital, the VA Vital Status File
was used to determine deaths that occurred after discharge. Veterans Affairs’ geo- coded enrollment files were used to determine veteran rural residence and travel
times to Veterans Affairs Medical Center (VAMC) facilities. Outpatient services received after discharge were identified in VA outpatient care files, which include
information pertaining to all outpatient encounters in VA facilities. Variables include visit date, clinic type (eg ER, primary care clinic), provider type (eg nurse,
physician), diagnosis codes and procedure codes.
Study cohorts
Creation of the study cohort involved three steps. First, admissions to acute care VA hospitals during the period October 2005 to September 2007 (ie Federal fiscal
years 2006–2007) were identified, and categorized as a clinically related re-admission or other admission using the 3M Potentially Preventable Re-admissions (PPR)
grouping software (3M company; Wallingford, CT, USA)33. The PPR
software identifies rehospitalizations that may result from deficiencies in the process of care or treatment, rather than unrelated events that occur post-discharge.
This software uses primary and secondary diagnosis codes, procedure codes and all-patient refined diagnosis related group (APR- DRG) codes to determine if two
admissions are clinically related. If the admissions are clinically related and occur within the set time period (eg 30 days), the second admission is classified as a
PPR.
Second, 36 566 admissions were defined as initial admissions with a primary diagnosis of CHF (ICD-9-CM codes 402.x1, 404.x1, 404.x3 or 428.xx). A total of 3568 initial
admissions were excluded from the study due to having any one of the following factors: died within 30 days of their CHF admission; transferred to another hospital;
admitted for trauma or malignancies; or had a missing or invalid home address (note: preventing the determination of rural/urban location). The remaining 32 998
admissions were divided into two cohorts: CHF admissions with a clinically- related re-admission (n=5698) within 30 days; and CHF admissions with no clinically-related
re-admission (n=27 300) within 30 days.
The admissions were further split into two groups based on rural and urban residence through use of the VA’s standard definition of rurality34. Level of rurality is
designated based on a veteran’s geo-coded address of primary residence at the end of FY2008. Veterans were classified as urban if they lived in a US Census Bureau
defined urbanized area, which consists of contiguous densely settled block groups that along with adjacent densely settled census blocks that together encompass a
population of at least 50 000 people. Veterans were categorized as rural if they did not reside in an urbanized area. Using this method, 21 664 urban veterans were
classified with CHF, 3792 (17.5%) which had clinically related re-admissions to VA hospitals. Of the 11 334 rural veterans who were identified with CHF, 1906 (16.8%)
had clinically related re-admissions to VA hospitals.
Additional patient characteristics were identified using the PTF discharge records, VA enrollment files, and outpatient
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 4
care files. Demographic characteristics included age, sex, and race (white, black, Hispanic and other), marital status, and income (=$5,000, $5,001–$10k, $10,001–$15k,
$15,001–$25k and >$25k). Travel times to the nearest VA primary care site were obtained from VA enrollment files and were derived using a methodology that incorporates
information on road networks from the US Department of Transportation, population density from the US Census, and average travel times from the 2002 Urban Mobility
Report35. Four types of follow-up VA outpatient visits were identified including: any physician or physician extender; primary care clinic; cardiology clinic; and
ER/department. These visits were identified as having occurred within a 30 day period from the first day after discharge up to but not including the day of re-
admission or the 13th day for those not re- admitted.
Statistical analysis
The prevalence of all variables for each cohort was described and compared using ?2 tests. Also, logistic regression was used to estimate the relative odds of re-
admission associated with each patient characteristic. Logistic regression models were generated separately for rural and urban patients and controlled for nine
socio-demographic and healthcare factors, including: age; gender; marital status; annual income; VA health service eligibility; military service era; travel time to
the nearest primary care source (minutes); and hospital length of stay (LOS). Additionally, the number of days until a specific outpatient visit type was included in
the multivariable models and the coefficient associated with each outpatient visit type was used to estimate the relationship between time to first outpatient visit
and the odds of re-admission. Significant interactions between the four VA outpatient visit variables and nine control variables were found using logistic regression.
Results from models for rural and urban veterans were compared.
Results
Most of the 32 998 veterans hospitalized for CHF were male (Table 1). Approximately two- thirds of veterans were white,
three-fourths had an income of less than $25,000 per year, half were married and approximately one-third were from rural communities. It was found that 17.3% of CHF
patients had a 30 day PPR; by residential location, it was found that urban patients (17.5%) had higher re-admission prevalence than rural patients (16.8%), but the
association was statistically non-significant (p=.121).
A number of socio-demographic factors were associated with CHF re-admission status (Table 1). Veterans under age 60 years were least likely to have re-admission, while
those over 80 years were most likely (p=.002). Veterans Affairs health service eligibility status was also associated with re- admissions because veterans who received
services due to low income or a disability were more likely to have a re- admission (p=.024). Although travel time to the nearest primary care source approached
significance (p=.056), having a LOS in the hospital of more than 1 week was a possible indication of health condition severity influencing re-admission (p< .001).
The 30 day post-discharge VA outpatient visits for veterans by CHF re-admission status are described (Table 2). Although ER visits were more common among veterans who
were re-admitted (p<.001), over 95% of veterans did not use ER visits for outpatient services. In contrast, visits to cardiology clinics were more frequent overall,
and the proportion of re-admitted veterans who had a cardiology clinic visit was significantly lower compared with the proportion for non-readmitted veterans (14% vs
27%; p<.001). Similarly, veterans who were re-admitted were significantly less likely than non-readmitted veterans to have had a physician/extender visit (50% vs 75%;
p<.001) or primary care clinic visit (30% vs 50%; p<.001).
There was also significant variation between re-admission status and time until the outpatient visits (Table 2). The percentage of veterans having a physician or
physician extender visit earlier versus later in the month decreased more rapidly for the re-admission cohort (30% to 8%) compared with the no re-admission cohort (29%
to 25%), which resembled the results for primary care and cardiology
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 5
clinic visits. For VA outpatient visits in the last 2 weeks of the 30 day period, the percentage of non-readmitted veterans was three to seven times higher than for
re-admitted veterans. However, the percentage of veterans who saw a VA physician/extender in the first 6 days was essentially the same for both cohorts (29% and 30%).
Separate logistic regressions were performed for rural and urban veterans (Table 3). Both cohorts had decreases in re- admission risk when having had a VA cardiology
clinic, primary care clinic or physician/extender visit in the 30 day post-discharge period. Also, longer time intervals between veterans’ (both rural and urban alike)
hospital discharges and follow-up visits were associated with greater reductions in re-admission risk. The odds of re-admission when having any physician or physician
extender visit within the first 6 days was not significantly different from no visit at all, but those odds decreased to 0.52 (rural) and 0.73 (urban) during the next
6 days, and then to 0.58 (rural) and 0.40 (urban) if the person waited for 13 days or more. A similar pattern was found for primary care clinic visit where ORs ranged
from 0.25 (13 to 30 days) to 0.57 (1 to 12 days) (rural and urban were almost identical), and cardiology clinic visits with ORs ranging from 0.16 (20 to 30 days) to
0.76 (1-10 days), with urban having greater variance.
Urban and rural differences in re-admissions were found primarily in demographics and use of the ER (Table 3). Unique predictors for rural veterans included use of the
ER at least once for those aged 65 years or older, which increased re-admission risk by 166%. Unique urban aspects included visiting the ER at least once increasing
risk of re- admission by 52% (for all ages), having low income or disability status increasing risk by 13% and having an initial hospital stay of 8 days or more
increasing re-admission risk by 11%.
Discussion
The 2005-2007 VA patient data were used to derive the prevalence of 30 day potentially preventable hospital re-
admissions among US veterans with CHF, examine socio- demographic traits of re-admitted veterans with CHF, and compare the effect of VA physician follow-up visits on
re- admission for rural- and urban-based veterans. It was found that approximately one-sixth (17.3%) of discharged CHF patients incurred a 30 day PPR within the VA
healthcare system. Re-admitted veterans, compared with non- readmitted veterans, tended to be older, had disability status, and had longer stays (ie 8 days or more)
during their initial hospitalization. Urban veterans had a slightly higher prevalence of CHF re-admissions than rural veterans (17.5% vs 16.8%); therefore, no
increased risk for CHF re- admissions was found for rural veterans.
Having a VA post-acute visit within 30 days of discharge at a primary care or cardiology clinic were each strongly negatively associated with CHF re-admissions for
both rural and urban veterans. This finding is consistent with the results of several studies involving non-veteran patient populations36-40. Timely post-discharge
follow-up care can promote positive health outcomes for the patient by allowing the healthcare provider to address any emerging health exacerbations, check for patient
compliance with home care instructions and adjust (as needed) medication regimen/dosages. Coleman et al. found that using ‘transition coaches’ to assist chronically
ill older patients and their caregivers by providing them with tools and skills that empower them to take a more active role in their care reduced re-admission
rates41. This approach may be especially beneficial for veterans in rural communities who may experience access barriers due to greater distances to VA healthcare
services and challenging terrain. However, a study by Weinberger et al. found that an intensive primary care intervention for severely chronically ill veterans
increased the rate of re-admissions42. Clearly, additional research is needed to develop an evidence-base to identify what interventions are most effective in
decreasing PPRs for veteran populations.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 6
Table 1: Association of socio-demographics, healthcare factors and re-admission for 32 998 congestive heart failure admissions
30-Day re-admission n (%)
Factor
No Yes
p
Rurality of residence Urban 17 872 (65.5) 3792 (66.6) .121 Rural 9428 (34.5) 1906 (33.5) Age 18–59 6299 (23.1) 1202 (21.1) .002 60–64 3529 (12.9) 742 (13.0) 65–74
6157 (22.6) 1263 (22.2) 75–79 3931 (14.4) 827 (14.5) =80 7384 (27.1) 1664 (29.2) Sex Male 26 822 (98.3) 5611 (98.5) .259 Female 478 (1.8) 87 (1.5) Marital status
Married 12 276 (45.0) 2608 (45.8) .188 Divorced/separated 8487 (31.1) 1684 (29.6) Widowed 4001 (14.7) 877 (15.4) Never married 2476 (9.1) 518 (9.1) Race White 18
005 (68.4) 3794 (68.1) .962 Black 6796 (25.8) 1453 (26.1) Hispanic 1231 (4.7) 265 (4.8) Other 284 (1.1) 62 (1.1) Income (annual) = $5,000 5316 (19.5) 1084 (19.0)
.765 $5,001–10k 3765 (13.8) 780 (13.7) $10,001–15k 5638 (20.7) 1203 (21.1) $15,001–25k 4858 (17.8) 991 (17.4) = $25,001 7723 (28.3) 1640 (28.8) VA Health Service
eligibility Low Income 14 415 (52.8) 3015 (52.9) .024 Disability 9752 (35.7) 2097 (36.8) Other 3133 (11.5) 586 (10.3) Travel time to primary healthcare source (min)
0–15 12 131 (44.4) 2630 (46.2) .056 16–30 7767 (28.5) 1612 (28.3) 31–60 5623 (20.6) 1126 (19.8) 61–90 1490 (5.5) 282 (5.0) = 91 289 (1.1) 48 (0.8) Length of
initial hospital stay (days) 1–2 1701 (6.2) 258 (4.5) <.001 3–7 18 191 (66.6) 3628 (63.7) = 8 7408 (27.1) 1812 (31.8) VA, Veterans Affairs.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 7
Table 2: Association of time to follow-up Veterans Affairs outpatient clinic visits and re-admissions for 32 998 CHF admissions
30-Day re-admission n (%) p Outpatient visit No Yes VA emergency room Yes 662 (2.4) 197 (3.5) <.001 No 26 638 (97.6) 5501 (96.5) Any VA physician or physician
extender (days) None 6852 (25.1) 2723 (47.8) <.001 1–6 7805 (28.6) 1704 (29.9) 7–12 5779 (21.2) 817 (14.3) 13–30 6864 (25.1) 454 (7.97) VA primary care clinic
(days) None 12 530 (45.9) 4088 (71.7) <.001 1–12 7809 (28.6) 1251 (22.0) 13–30 6961 (25.5) 359 (6.3) VA Cardiology Clinic (days) None 19 934 (73.0) 4926 (86.5)
<.001 1–10 2899 (10.6) 498 (8.7) 11–19 2222 (8.1) 211 (3.7) 20–30 2245 (8.2) 63 (1.1) VA, Veterans Affairs.
It was found that the greater the time between patients’ VA hospital discharge and use of VA follow-up outpatient care during the 30 day post-discharge period, the
less likely patients were to be re-admitted. This finding may be due, in part, to veterans with shorter discharge-to-outpatient care intervals may have more serious
health problems than those veterans who are able to wait longer for outpatient care. It may also indicate that veterans who receive comprehensive discharge planning
and/or coordination of home care are better able to maintain their health after discharge41 and, as a result, may have a lower need for immediate follow up. It may
also reflect other factors not captured in the data such as home support (ie a family member or neighbor that monitors medication) or telehealth (ie telephone contacts
with VA providers to monitor veterans’ blood pressure and weight). Healthcare providers within the VA may want to consider care alternatives, such as telehealth
follow-ups, for rural- dwelling veterans who may face transportation challenges in accessing VA healthcare services. Additional research is needed to determine if
telehealth interventions for veterans
decrease CHF re-admission rates, particularly for rural veterans.
A limitation of this study is the lack of information pertaining to health care provided outside of the VA system, such as an admission to a non-VA hospital following
a discharge from a VA hospital. Rural veterans who seek care from VA and non-VA providers may utilize health care differently than veterans who rely solely on VA
providers; thus, the co-management of care may impact the PPR risk for rural veterans. Also, other factors that are potentially predictive of re-admission risk were
not accounted for in this study, including specific indicators of inpatient care quality, discharge planning, care coordination, home support, patient compliance and
patient self-care. The inclusion of such factors in future studies would assist in determining the extent to which the type/quality of received care and patient health
behaviors are associated with veterans’ re-admission risk.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 8
Table 3: Logistic regression analyses of predictors of 30 day potentially preventable re-admissions for 11 334 rural veterans and 21 664 urban veterans with CHF,
controlling for demographic and healthcare factors
Veteran location OR (95% CI)
Predictor
Rural n = 11 334
Urban n = 21 664
Days to any VA physician or physician extender visit post-discharge No visit 1.00 (reference) 1.00 (reference) 1–6 0.92 (0.79–1.08) 0.98 (0.88–1.09) 7–12 0.52 (0.44–
0.63) 0.73 (0.64–0.82) 13–30 0.38 (0.31–0.46) 0.40 (0.35–0.47) Days to VA primary care clinic visit post-discharge No visit 1.00 (reference) 1.00 (reference) 1–12 0.57
(0.49–0.66) 0.56 (0.51–0.62) 13–30 0.26 (0.21–0.32) 0.25 (0.22–0.30) Days to VA cardiology clinic visit post-discharge No Visit 1.00 (reference) 1.00 (reference) 1–10
0.66 (0.54–0.81) 0.76 (0.66–0.86) 11–19 0.58 (0.44–0.77) 0.56 (0.46–0.66) 20–30 0.23 (0.15–0.35) 0.16 (0.11–0.22) VA Emergency room visits (30 days post-discharge)
None – 1.00 (reference) = 1 – 1.52 (1.23–1.87) Age and VA emergency room visit <65 or no visit 1.00 (reference) – =65 and ER visit 2.66 (1.87–3.78) – LOS =8 days No –
1.00 (reference) Yes – 1.11 (1.02–1.20) VA Health service eligibility category Other – 1.00 (reference) Low income/disability – 1.13 (1.01–1.27) LOS,
Length of stay; VA, Veterans Affairs.
Some rural/urban differences were noted regarding indicators of CHF re-admission, particularly among older veterans and the use of ERs. Differences among rural and
urban veterans’ likelihood of re-admission following use of the ER for outpatient care may be an indication of a healthcare access issue. Using an ER within 30 days
post- discharge and having disability or low income VA eligibility status each independently increased the likelihood of CHF re-admission for urban veterans only.
These factors may denote or contribute to veterans’ higher condition severity and/or presence of CHF exacerbations, both of which would increase the likelihood of re-
admission.
For rural veterans only, use of the ER by those aged 65 years or older was a significant, independent indicator of re- admission (OR=2.66, CI=1.87-3.78). Elderly
veterans residing in rural areas may be more susceptible to delayed treatment of CHF complications due to transportation challenges and, hence, poorer access to VA
clinics. Thus, rural-dwelling veterans with CHF may warrant special attention by VA healthcare providers for comprehensive discharge planning and follow-up care
provided by healthcare providers who are familiar with their patients’ unique situations, including home support, medical histories, health conditions, access to
transportation and current treatment regimens. Further research is needed to examine
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 9
the care that is provided to rural veterans in their home communities.
The strength of this study is its in-depth examination of inpatient and outpatient care for rural and urban CHF patients, regardless of age, within the VA healthcare
system. However, study results may not be generalizable to non-VA patients or co-managed veteran patients. Future studies on rural VA patients are needed which address
these gaps, incorporate additional datasets, such as Medicare, and delineate care models which promote patient-centered discharge planning and coordinated, timely
follow-up care for veterans in both rural and urban areas of the country.
Conclusion
Post-acute follow-up care is an important factor for promoting recovery and good health among hospitalized veterans with CHF, regardless of their rural or urban
residence. Older, rural veterans with CHF are in need of special attention for VA discharge planning and follow up with primary care providers.
Acknowledgments
The work reported here was supported by the Department of Veterans Affairs, Veterans Health Administration, Office of Rural Health. The authors also recognize the
contributions of Mary K Wakefield, PhD.
References
1. Rosamond W, Flegal K, Friday G, Furie K, Go A, Greenlund K
et al. Heart disease and stroke statistics: 2007 update: a report from
the American Heart Association Statistics Committee and Stroke
Statistics Subcommittee. Circulation 2007; 115: e69-e171.
2. Kannel WB, Belanger AJ. Epidemiology of heart failure.
American Heart Journal 1991; 121: 951-957.
3. Kenchaiah S, Narula J, Vasan RS. Risk factors for heart failure.
Medical Clinics of North America 2004; 88: 1145-1172.
4. Schocken DD, Benjamin EJ, Fonarow GC, Krumholz HM, Levy
D, Mensah GA et al. Prevention of heart failure: a scientific
statement from the American Heart Association Councils on
Epidemiology and Prevention, Clinical Cardiology, Cardiovascular
Nursing, and High Blood Pressure Research; Quality of Care and
Outcomes Research Interdisciplinary Working Group; and
Functional Genomics and Translational Biology Interdisciplinary
Working Group. Circulation 2008; 117: 2544-2565.
5. Levy D, Kenchaiah S, Larson MG, Benjamin EJ, Kupka MJ, Ho
KK et al. Long-term trends in the incidence of and survival with
heart failure. New England Journal of Medicine 2002; 347: 397-
1402.
6. Gottdiener JS, Arnold AM, Aurigemma GP, Polak JF, Tracy RP,
Kitzman DW et al. Predictors of congestive heart failure in the
elderly: the Cardiovascular Health Study. Journal of the American
College of Cardiology 2000; 35: 1628-1637.
7. He J, Ogden LG, Bazzano LA, Vupputuri S, Loria C, Whelton
PK. Risk factors for congestive heart failure in US men and
women: NHANES I epidemiologic follow-up study. Archives of
Internal Medicine 2001; 161: 996-1002.
8. Chen YT, Vaccarino V, Williams CS, Butler J, Berkman LF,
Krumholz HM. Risk factors for heart failure in the elderly: a
prospective community-based study. American Journal of Medicine
1999; 106: 605-612.
9. Levy D, Larson MG, Vasan RS, Kannel WB, Ho KK. The
progression from hypertension to congestive heart failure. Journal
of the American Medical Association 1996; 275: 1557-1562.
10. Kannel WB, D’Agostino RB, Silbershatz H, Belanger AJ,
Wilson PW, Levy D. Profile for estimating risk of heart failure.
Archives of Internal Medicine 1999; 159: 1197-1204.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 10
11. Chae CU, Pfeffer MA, Glynn RJ, Mitchell GF, Taylor JO,
Hennekens CH. Increased pulse pressure and risk of heart failure in
the elderly. Journal of the American Medical Association 1999;
281: 634-639.
12. Ho KK, Pinsky JL, Kannel WB, Levy D. The epidemiology of
heart failure: the Framingham Study. Journal of the American
College of Cardiology 1993; 22: 6A-13A.
13. Kenchaiah S, Evans JC, Levy D, Wilson PW, Benjamin EJ,
Larson MG et al. Obesity and the risk of heart failure. New England
Journal of Medicine 2002; 347: 305-313.
14. Centers for Medicare and Medicaid Services. Medicare ranking
for all short-stay hospitals by discharges fiscal year 2005 versus
2004. Baltimore, MD: CMS, 2008.
15. Krumholz HM, Parent EM, Tu N, Vaccarino V, Wang Y,
Radford MJ et al. Re-admission after hospitalization for congestive
heart failure among Medicare beneficiaries. Archives of Internal
Medicine 1997; 157: 99-104.
16. Marcantonio ER, McKean S, Goldfinger M, Kleefield S,
Yurkofsky M, Brennan TA. Factors associated with unplanned
hospital re-admission among patients 65 years of age and older in a
Medicare managed care plan. American Journal of Medicine 1999;
107: 13-17.
17. Hunt SA, Baker DW, Chin MA, Cinquegrani M, Feldman A,
Francis G et al. ACC/AHA guidelines for the evaluation and
management of chronic heart failure in the adult: executive
summary. Journal of Heart and Lung Transplantation 2002;
21: 189-203.
18. McAlister FA, Lawson FM, Teo KK, Armstrong PW. A
systematic review of randomized trials of disease management
programs in heart failure. American Journal of Medicine 2001;
110: 378-384.
19. Medicare Payment Advisory Commission. Promoting greater
efficiency in Medicare. Washington, DC: MedPAC, 2007.
20. Qualitynet.org. Re-admission measures overview: CMS 30 day
risk-standardized heart failure re-admission measure. (Online)
2008. http://www.qualitynet.org/dcs/ContentServer?cid=1219069
855273&pagename=QnetPublic%2FPage%2FQnetTier2&c=Page
(Accessed 18 November 2009).
21. Krumholz HM, Parent EM, Tu N, Vaccarino V, Wang Y,
Radford MJ et al. Re-admission after hospitalization for congestive
heart failure among Medicare beneficiaries. Archives of Internal
Medicine 1997; 157: 99-104.
22. Kosiborod M, Curtis JP, Wang Y, Smith GL, Masoudi FA,
Foody JM et al. Anemia and outcomes in patients with heart failure:
a study from the National Heart Care Project. Archives of Internal
Medicine 2005; 165: 2237-2244.
23. Krumholz HM, Chen YT, Wang Y, Vaccarino V, Radford MJ,
Horwitz RI. Predictors of re-admission among elderly survivors of
admission with heart failure. American Heart Journal 2000;
139: 72-77.
24. Philbin EF, Dec GW, Jenkins PL, DiSalvo TG. Socioeconomic
status as an independent risk factor for hospital re-admission for
heart failure. American Journal of Cardiology 2001; 87: 1367-
1371.
25. Philbin EF, DiSalvo TG. Influence of race and gender on care
process, resource use, and hospital-based outcomes in congestive
heart failure. American Journal of Cardiology 1998; 82: 76-81.
26. Philbin EF, DiSalvo TG. Prediction of hospital re-admission for
heart failure: development of a simple risk score based on
administrative data. Journal of the American College of Cardiology
1999; 33: 1560-1566.
27. Rathore SS, Foody JM, Wang Y, Smith GL, Herrin J, Masoudi
FA et al. Race, quality of care, and outcomes of elderly patients
hospitalized with heart failure. Journal of the American Medical
Association 2003; 289: 2517-2524.
© KJ Muus, A Knudson, MG Klug, J Gokun, M Sarrazin, P Kaboli, 2010. A licence to publish this material has been given to ARHEN http://www.rrh.org.au 11
28. Rathore SS, Masoudi FA, Wang Y, Curtis JP, Foody JP,
Havranek EP et al. Socioeconomic status, treatment, and outcomes
among elderly patients hospitalized with heart failure: findings
from the National Heart Failure Project. American Heart Journal
2006; 152: 371-378.
29. Ross JS, Mulvey GK, Stauffer B, Patlolla V, Bernheim SM,
Keenan PS et al. Statistical models and patient indicators of re-
admission for heart failure. Archives of Internal Medicine 2008;
168: 1371-1386.
30. Holloway JJ, Medendorp SV, Browning J. Risk factors for early
re-admission among veterans. Health Services Research 1990;
25: 213-237.
31. Ludke RL, MacDowell NM, Booth BM, Hunter SA.
Appropriateness of admissions and discharges among readmitted
patients. Health Services Research 1990; 25: 501-525.
32. Weeks WB, Lee RE, Wallace AE, West AN, Bagian JP. Do
older rural and urban veterans experience different rates of
unplanned re-admission to VA and non-VA hospitals? Journal of
Rural Health 2008; 25: 62-69.
33. Goldfield NI, McCullough EC, Hughes JC, Tang AM, Eastman
B, Rawlins LK et al. Identifying potentially preventable re-
admissions. Health Care Financing Review 2008; 30: 75-91.
34. Weeks WB, Wallace AE, West AN, Heady HR, Hawthorne K.
Research on rural veterans: an analysis of the literature. Journal of
Rural Health 2008; 24: 337-344.
35. Shrank D, Lomax T. The urban mobility report. College
Station, TX: Texas A&M University System, 2002.
36. Lappe JM, Muhlestein JB, Lappe DL, Badger RS, Bair TL,
Brockman R et al. Improvements in 1-year cardiovascular clinical
outcomes associated with a hospital-based discharge medication
program. Annals of Internal Medicine 2004; 141: 446-453.
37. Phillips CO, Wright SM, Kern DE, Singa RM, Shepperd S,
Rubin HR. Comprehensive discharge planning with postdischarge
support for older patients with congestive heart failure: a
metaanalysis. Journal of the American Medical Association 2004;
291: 1358-1367.
38. Jovicic A, Holroyd-Leduc JM, Straus SE. Effects of self-
management intervention on health outcomes of patients with heart
failure: a systematic review of randomized controlled trials. BMC
Cardiovascular Disorders 6: 43. (Online) 2006. Available: http://
www.biomedcentral.com/content/pdf/1471-2261-6-43.pdf
(Accessed 26 May 2009).
39. McAlister FA, Lawson FM, Teo KK, Armstrong PW. A
systematic review of randomized trials of disease management
programs in heart failure. American Journal of Medicine 2001;
110: 378-384.
40. Gonseth J, Guallar-Castillon P, Banegas JR, Rodriguez-Artalejo
F. The effectiveness of disease management programmes in
reducing hospital re-admission in older patients with heart failure: a
systematic review and meta-analysis of published reports.
European Heart Journal 2004; 25: 1570-1595.
41. Coleman EA, Parry C, Chalmers S, Min SJ. The care transitions
intervention: results of a randomized controlled trial. Archives of
Internal Medicine 2006; 166: 1822-1828.
42. Weinberger M, Oddone EZ, Henderson WG. Does increased
access to primary care reduce hospital re-admissions? New England
Journal of Medicine 1996; 334(22): 1441-1447.
Copyright of Rural & Remote Health is the property of Australian Rural Health Education Network and its content may not be copied or emailed to multiple sites or
posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.
Impact of a standardized heart failure order set on mortality, readmission, and quality and costs of care
DAVID J. BALLARD1, GERALD OGOLA1, NEIL S. FLEMING1, BRETT D. STAUFFER1, BRADLEY M. LEONARD2, RAINER KHETAN3 AND CLYDE W. YANCY4
1Institute for Health Care Research and Improvement, Baylor Health Care System, Dallas, TX, USA, 2The Heart Hospital at Baylor Plano, Plano, TX, USA, 3Baylor
University Medical Center, Dallas, TX, USA, and 4Baylor Heart and Vascular Institute, Dallas, TX, USA Address reprint requests to: David J. Ballard, Baylor Health Care
System, Institute for Health Care Research and Improvement, 8080 N. Central Expressway, Suite 500, Lockbox 81, Dallas, TX 75206, USA. Tel: þ1-214-265-3670; Fax: þ1-
214-265-3640; E-mail: dj.ballard@baylorhealth.edu
Accepted for publication 6 September 2010
Abstract
Objective. To determine the impact of a standardized heart failure order set on mortality, readmission, and quality and costs of care. Design. Observational study.
Setting. Eight acute care hospitals and two specialty heart hospitals. Participants. All adults (.18 years) discharged from one of the included hospitals between
December 2007 and March 2009 with a diagnosis of heart failure, who had not undergone heart transplant, did not have a left ventricular assistive device, and with a
length of stay of 120 or less days. Interventions. A standardized heart failure order set was developed internally, with content driven by the prevailing American
College of Cardiology/American Heart Association clinical practice guidelines, and deployed systemwide via an intranet physician portal. Main Outcome Measures.
Publicly reported process of care measures, in-patient mortality, 30-day mortality, 30-day readmis- sion, length of stay, and direct cost of care were compared for
heart failure patients treated with and without the order set. Results. Order set used reached 73.1% in March 2009. After propensity score adjustment, order set use
was associated with signi?cantly increased core measures compliance [odds ratio (95% con?dence interval)¼1.51(1.08; 2.12)] and reduced in- patient mortality [odds
ratio (95% con?dence interval)¼0.49(0.28; 0.88)]. Reductions in 30-day mortality and readmission approached signi?cance. Direct cost for initial admissions alone and
in combination with readmissions were signi?cantly lower with order set use. Conclusions. Implementing an evidence-based standardized order set may help improve
outcomes, reduce costs of care and increase adherence to evidence-based processes of care. Keywords: quality improvement, quality indicators, mortality, readmissions,
cardiovascular diseases, hospital care
Introduction
The estimated prevalence of heart failure in adults age 20 and older in the USA was 5.7 million in 2006, with an annual incidence of 10 per 1000 persons over age 65
[1]. Hospital discharges for heart failure rose from 877 000 in 1996 to 1.1 million in 2006—an increase of 26%—and the estimated cost of heart failure in 2009 is $37.2
billion [1].
Substantial evidence has demonstrated reduced morbidity and mortality with angiotensin-converting enzyme inhibitor (ACEI) or angiotensin receptor blocker (ARBs), b-
blocker therapy and other evidence-based treatment modalities in patients with chronic heart failure caused by left ventricular systolic dysfunction, with clinical
trials showing improved survival, reduced risk of hospitalization or death and improved New York Heart Association functional class
International Journal for Quality in Health Care vol. 22 no. 6 # The Author 2010. Published by Oxford University Press in association with the International Society
for Quality in Health Care; all rights reserved 437
International Journal for Quality in Health Care 2010; Volume 22, Number 6: pp. 437–444 10.1093/intqhc/mzq051 Advance Access Publication: 8 October 2010
[2–6]. Despite this evidence and clinical practice guidelines that strongly recommend evidence-based therapies as part of standard care for heart failure, both
ACEIs/ARBs and b-blockers (as well as other evidence-based therapies) are underutilized in heart failure patients without contraindica- tions or documented intolerance
[7]. Studies of hospitalized heart failure patients show that only 60–80% of eligible patients are prescribed ACEIs and b-blockers at discharge [6, 8–12], 24–55%
receive discharge instructions [11, 12], 86–88% receive left ventricular function assessment [11, 12] and 43–72% receive smoking cessation coun- seling [11, 12].
Hospital Quality Alliance data for patients receiving care in non-federal US hospitals in 2005 showed only 57.3% of heart failure patients received all process of care
measures for which they were eligible [13]. Given the high mortality and morbidity rates for heart failure, as well as the fact that it is the primary or secondary
cause of approximately 3.6 million hospitalizations annually in the USA [1, 14], inpatient heart failure care is an important target for quality improvement. Baylor
Health Care System (BHCS), an integrated health- care delivery system in North Texas, developed and implemented standardized order sets for heart failure care across
10 hospitals, 8 acute care and 2 specialty hospitals, as part of a system-wide approach focused on heart failure [15]. Previously, we have shown successful improvement
in com- pliance with treatment recommendations and in mortality and costs of care for adult pneumonia patients following widespread adoption of a standardized
pneumonia order set [15, 16]. We sought to determine whether a similar effect was achieved following the implementation of a standardized order set for heart failure.
Methods
Study setting BHCS is a not-for-pro?t, multi-hospital system in Dallas-Fort Worth, Texas incorporating 20 owned, leased, af?liated and short-stay hospitals with more
than 103 000 admissions per year. This study included patients treated in the eight BHCS acute care hospitals, as well as the two speci- alty heart hospitals.
Intervention The standardized heart failure order set was developed intern- ally [15], with content driven by the prevailing American College of Cardiology/American
Heart Association clinical practice guidelines, and deployed systemwide via the intranet physician portal in December 2007. For a copy of the standar- dized order set,
see Appendix. ‘Order set used’ was made a required ?eld in MIDASþTM—the integrated outcomes, resource and case management system used at all BHCS hos- pitals—for heart
failure patients to facilitate tracking of order set usage. Increasing use of the order set is one of four action items in the BHCS Heart Failure Initiative—along with
medication reconciliation focused on heart failure; establishing a continuum of care for inpatient heart failure patients; and standardizing end-of-life care,
palliative care and advance directives practices for heart failure. The initial goal was for 80% of health failure admissions systemwide to use the order set by March
2009. For hospitalists—who treat more heart failure patients than any other single specialty—this goal was set to the higher standard of 95%. To spearhead implemen-
tation of the Heart Failure Initiative, BHCS established a Heart Failure Task Force and designated a heart failure physician champion for each hospital. Monthly
reports, containing order set usage and missed opportunities for use, were provided to physician champions and health-care improvement directors at each facility. The
reports included physician-speci?c use to aid in academic detailing and allow physicians to track individual performance. With increased acceptance of the order set,
the health system promoted further adoption by mandating use of the order set, with consequences for omission established my each hospital’s Medical Executive
Committee.
Patients All adults (.18 years) discharged from one of the included hospitals between December 2007 and March 2009 with a diagnosis of heart failure, who had not
undergone heart transplant, did not have a left ventricular assistive device, and with a length of stay 120 or less days (inclusion criteria de?ned in Appendix A) were
eligible for this study. Patients with physician orders for ‘comfort measures only’ recorded in the admitting physician orders or note, consultation notes, emergency
department record, history and physical, physician orders or progress notes were excluded from the study, as were patients for whom the heart failure order set use
data were missing or recorded as ‘not applicable’ or for whom a heart failure order set other than the BHCS standardized order set was used.
Outcome measures Primary outcome measures investigated were differences in in-hospital mortality, 30-day mortality, 30-day readmission and compliance with an all-or-
none bundle of process of care measures. Compliance of the all-or-none bundle is calculated as the proportion of heart failure patients eligible for the four heart
failure measures—discharge instructions, evaluation of left ventricular function, ACEI or ARB for left ventricular systolic dysfunction and smoking cessation
counseling—who receive all the measures for which they are eligible [17]. In addition, we examined length of stay and direct cost of treatment.
Data collection Age, gender, race/ethnicity, admitting BHCS hospital, deliv- ery of core measures for heart failure and information on order set use was collected from
MIDAS for each patient. In-patient mortality, readmissions, length of stay and direct costs of care were determined from BHCS administrative
Ballard et al.
438
data; 30-day mortality was determined from the Social Security Death Master File [18]. Order set use is recorded as a required ?eld in MIDASþ for heart failure
patients, with the options ‘BHCS standard order set used’, ‘no standard order set used’, ‘other standard order set used’ and ‘BHCS order set not applicable.’ The order
set was regarded as ‘not applicable’ if: (i) the diagnosis of heart failure was made after the admission orders were written by the attending phys- ician (not the
Emergency Department physician); (ii) both heart failure and pneumonia were present on admission and the physician used the BHCS pneumo- nia order set; or (iii) the
patient was admitted for an elective implantable cardioverter de?brillator placement and other orders sets were in place for the speci?c procedure.
Statistical analysis Some patients included in the study population had multiple admissions during the study period. For these patients a single admission record was
randomly selected as initial admission in the analysis. For the unadjusted analyses, binomial tests, t-tests, chi-square tests and robust analysis of variance tests
were used to deter- mine association between patient characteristics and order set use, or differences seen in outcomes with order set use. This being an observational
study, patients were not ran- domized into the intervention group. A propensity score adjustment approach was used to reduce the impact of selec- tion bias on the
association between order set use and out- comes of interest. The propensity score, the conditional probability of a patient being treated with the order set given the
patient characteristics [age, gender, race, risk of mortality/ severity of illness, type of physician (hospitalists vs. non- hospitalists), facility and month of
discharge], was determined from a multivariable logistic regression model, and then used as a covariate in the adjusted effect model. The risk of mor- tality and
severity of illness were used independently depend- ing on the outcome evaluated. For clinical outcomes, for example mortality, the risk of mortality variable was
included in the model while for resource utilization, for example the cost of care, severity of illness variable was included. Propensity score adjusted effects of the
order set on mor- tality and core measures compliance were modeled using multivariable logistic regression model. Length of stay was modeled using a multivariable
robust regression due to high skewness in its distribution. Readmission was modeled using competing risk regression analysis [19] implemented in R software [20]. All
patients who died during their hospitaliz- ation were excluded from the readmission analysis. Death after hospitalization was considered a competing risk event for
readmission. Our approach addresses two important aspects of cost data that are non-normal: (i) they are typically right skewed (with a thick tail) as some patients
(beyond simply outliers) incur substantial health-care costs compared with the mean); and (ii) a substantial number of patients have no readmis- sions and associated
costs. This two part-model with a
logarithmic distribution on the conditional (re-admitted patient) costs appropriately addresses the lack of normality and speci?c distribution of the cost data. Direct
costs were modeled using a two-step estimation method. In the ?rst step, the probability of readmission was modeled based on a propensity score adjusted logistic
regression model; in the second step, direct costs were modeled based on the log-link function with the gamma dis- tribution, given that the readmission had occurred
[19, 21]. The product of these two steps creates the estimate of the readmission costs. Subsequently, the method of recycled pre- dictions [21] was employed to
appropriately consider under- lying differences between the order set and non-order set groups and essentially create a balanced design. Initial admis- sion direct
costs and readmission direct costs were predicted from the modeled equations based on two scenarios: (i) every patient received the heart failure order set; and (ii)
every patient did not receive the heart failure order set. The differ- ence between these two predictions constitutes the predicted mean differences in initial
admission and readmission costs. Total heart failure direct costs were estimated as part of a three-step process: we generated 1000 bootstrap samples and created
recycled predictions to estimate differences with and without order set use in direct costs during initial admission (Step 1); probability of all-cause readmission
(30-day and 1-year) (Step 2); and, for each bootstrap sample where a readmission actually occurred for the patient, direct costs for the hospital readmission (Step 3).
For this last estimate, we created recycled predictions for the entire bootstrap sample, not just for the individuals who had readmissions. Total direct heart failure
costs were then estimated as: DirectindexcostsðStep1ÞþðreadmissionprobabilityðStep2Þ directcostsof readmissionðStep3ÞÞ; where mean direct costs estimated in Step 1
represented the order set effect for index admissions; the mean of the read- mission probabilities estimated in Step 2, the order set effect for readmissions and the
mean direct costs of readmission estimated in Step 3 to estimate the order set effect for total direct costs including readmissions. The standard error for each of
these statistics was calculated as the standard deviation of the combined 1000 bootstrap samples. This methodology accounts for two potential effects use of the order
set during initial admissions may have on readmissions: the potential impact on probability of readmission and the potential impact on the direct cost of the
readmissions themselves. Analyses were done using SAS 9.2 (Cary, NC) and R version 2.9.0 (2009–04-17) statistical software. P-values less than 5% were considered
statistically signi?cant.
Results
A total of 2633 eligible patients were discharged from the 10 included BHCS hospitals between December 2007 and March 2009. Age ranged from 22 to 103 years with a mean
age (standard deviation) of 69.1 (15.3) years. Approximately
Impact of a heart failure order set
439
half (50.3% or 1324 of 2633) were females; and 37.5% were non-white (988 of 2633; see Table 1). There was signi?cant variation in order set use at all hos- pitals
combined by month (P, 0.01), ranging from 15 to 80%, with use generally increasing over time (see Fig. 1). Table 2 shows the results of the unadjusted comparison of
order set by patient demographics and all Patient Re?ned Diagnosis Related Group (APR-DRG) risk of mortality and severity of illness. The use of the order set differed
signi?- cantly by severity of illness class (P¼0.02)—with patients in the highest severity class being least likely to receive the order set—and approached signi?cance
by risk of mortality class (P¼0.07) (Fig. 2). Table 3 shows the unadjusted results of comparisons between patients who did and did not receive the BHCS heart failure
order set on in-hospital and 30-day mortality, 30-day readmission, core measures compliance, length of stay and direct cost. There were 183 (7.0%) deaths within 30-
days of admission (either in-hospital of following discharge) and 60 (2.3%) in-hospital deaths. Statistically signi?cant differ- ences were observed for in-hospital
mortality, core measures compliance and direct costs, with patients receiving the order set having lower mortality rates, smaller direct costs and greater core
measures compliance. Following risk adjustment, the effect of the standardized heart failure order set on in-hospital mortality and core measures compliance was
retained (see Table 4). The reductions in 30-day mortality and 30-day readmission bordered on signi?cance [odds ratio (95% con?dence interval)¼0.81(0.58; 1.13)]. The
decrease in length of stay with order set use approached but did not achieve signi?cance. All measures of direct costs examined except 30-day and 1-year readmission
direct costs showed signi?cant decreases with order set use.
Discussion
Summary of results During the ?rst year of implementation, the use of the BHCS heart failure order set reached 60% of eligible cases systemwide and increased to .70%
by March 2009. In unadjusted comparisons, the use of the BHCS order set was associated with signi?cantly lower in-hospital mortality, smaller direct costs and signi?
cantly higher core measures compliance. Importantly, following risk adjustment, differ- ences in in-hospital mortality and core measures compliance remained
statistically signi?cant. These ?ndings demonstrate that wide deployment of a standardized evidence-based heart failure order set is associated with improved in-
hospital out- comes. Based on the observed mortality rates with and without order set use, for every 85 heart failure patient encounters in which the BHCS order set is
used, one in-hospital death is prevented. Signi?cant direct cost savings, overall (initial admissionþre-admissions) and when initial admissions were examined in
isolation, were also observed with order set use. Decreases in 30-day mortality and 30-day readmission among patients receiving the order set bordered on signi?cance.
Limitations As this was an observational study rather than a randomized trial, it is possible that order set use was in?uenced by unmea- sured patient or hospital
characteristics, potentially under- or over-estimating the impact of order set use. Counterbalancing this limitation is the fact that observational studies based on
real-world clinical data provide a realistic estimate of the effect that may be expected in similar real-world settings. To account for the differences between the
patients that did and did not receive the order set, we performed risk-adjusted analyses based on a propensity score approach, which also accounted for the variation
in order set use by facility and time observed in this study. However, risk adjustment is limited in that it can only account for measured confounders and does not
ensure a balanced distribution of all covariates among the study sub- jects such as randomization provides. Since only administrative data were available, we were
unable to adjust for patient’s clini- cal characteristics and we cannot eliminate the possibility of selection bias, with ‘healthier’ patients who ?t most easily
within standardized treatment guidelines being more likely to receive the order set. Such selection bias leads to overestima- tion of the intervention’s effects. On
the other hand, since all clinicians had access to the order set and may have used it for some heart failure patients but not others, it is also possible that
treatment of the ‘no order set’ group was in?uenced or contaminated by the clinicians’ exposure to and incorporation of the items included in the order set. If so, the
results demonstrated here may underestimate the full bene?t of order set implementation. Our study population represents the subset of heart failure patients eligible
for the CMS heart failure core measures and excludes any patient who was admitted with both heart
…………………………………………………………………………
Table 1 Descriptive summary statistics of heart failure patients discharged from Baylor health care system (BHCS) hospitals from December 2007 to March 2009
Summary statistics (n¼2633) Age (years), mean (SD) 69.1 (15.3) Gender (female), n (%) 1324 (50) Race (white), n (%) 1645 (63) Physician (hospitalists), n (%) 685 (26)
Risk of mortality, n (%) 1 348 (13) 2 1126 (43) 3 854 (33) 4 305(11) Severity of illness, n (%) 1 183 (7) 2 1027 (39) 3 1147 (44) 4 276 (10)
SD, standard deviation.
Ballard et al.
440
failure and another diagnosis for which a BHCS standardized order set was available (pneumonia or sepsis) and the physician used the order set for that diagnosis.
Additionally, all data were from a single hospital system: data regarding readmissions to other facilities were not available, and the
estimates of mortality and cost savings may not be broadly applicable to other systems, hospitals or geographical regions.
Comparison to the literature Adoption of the BHCS heart failure order set is comparable to that reported for order sets implemented in the Grey Bruce Health Network in
Canada (40% systemwide after 6
Figure 2 Percentage of heart failure patients treated with the Baylor health care system-heart failure (BHCS-HF) Standardized Order Set by facility from December 2007
to March 2009 (sorted by facility total number of cases).
Figure 1 Proportion of heart failure patients who received the Baylor health care system (BHCS) Standardized Order Set by month among patients discharged from BHCS’
eight acute care hospitals and two specialty heart hospitals from December 2007 to March 2009.
…………………………….
…………………………………………………………………………
Table 2 Unadjusted comparison of BHCS order set use by APR-DRG risk of mortality and severity of illness for heart failure patients discharged from BHCS hospitals from
December 2007 to March 2009
Order set BHCS-HF (n¼1363)
None (n¼1270)
P-value
Age, mean (SD) 69.4 (15.3) 68.8 (15.3) 0.34a Gender (female), n (%) 703 (52) 621 (49) 0.17b Race (white), n (%) 815 (60) 830 (65) 0.01b Physician (hospitalists), n (%)
481 (35) 204 (16) ,0.01b Risk of mortality, n (%) 1 200 (15) 148 (11) 0.07b 2 558 (41) 568 (45) 3 449 (33) 405 (32) 4 156 (11) 149 (12) Severity of lllness, n (%) 1
101 (7) 82 (6) 0.02b 2 540 (40) 487 (38) 3 604 (44) 543 (43) 4 118 (9) 158 (13)
BHCS-HF, Baylor health care system-heart failure; SD, standard deviation; aBased on t-test; bBased on chi-square test.
Impact of a heart failure order set
441
months) [22] and mirrored the 72% adoption of an emergency department heart failure clinical pathway 14 months post- implementation at a single suburban tertiary care
hospital [23]. Similarly, previous studies have shown similar improvements in heart failure outcomes following implementation of hospital-based tools to increase
standardization. The Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF) provided 259 participating hospitals
with evidence-based practice algorithms, critical pathways, standardized orders, dis- charge checklists and other tools to assist with the manage- ment of heart
failure. The OPTIMIZE-HF intervention
resulted in substantial increases in the provision of complete discharge instructions, assessment of left ventricular function, adherence to both ACEI/ARB and b-
blocker therapy, and smoking cessation instructions. This process of care improve- ment strategy was associated with a decrease in risk-adjusted in-hospital deaths,
90-day mortality and a decrease in the com- bined endpoint of readmission and mortality [10]. Clinical pathways for heart failure have also been associated with an
70% increase in the eligible patients receiving ACEIs [24]; and signi?cant increases in delivery of heart failure clinical care measures, with concomitant decreases
in total admissions, readmissions and inpatient mortality [25].
……………………………………….
…………………………………………………………………………………………………………………………………………………..
……
Table 3 Unadjusted comparison of BHCS order set vs. no order set on mortality, readmission, core measures compliance and length of stay for heart failure patients
discharged from BHCS hospitals from December 2007 to March 2009
Order set All BHCS-HF None P-value
Safety and effectiveness indicators In-hospital mortality, n (%) 60 (2.3) 23 (1.7) 37 (2.9) 0.04a 30-day post-admission mortality, n (%) 183 (7.0) 84 (6.2) 99 (7.8)
0.10a 30-day readmission, n (%) 338 (13.1) 166 (12.4) 172 (13.9) 0.50b Core measure compliance, n (%) 2139 (91.9) 1120 (93.5) 1019 (90.3) ,0.01a Financial and ef?
ciency indicators Length of stay (days), mean (SD) 5.2 (4.5) 5.0 (4.0) 5.5 (5.0) 0.21c Cost ($), mean (SD) Initial admission direct cost 6212 (7527) 5493 (5588) 6981
(9098) ,0.01d 30-day readmission direct cost 1126 (8203) 725 (2964) 1551 (11360) 0.02d One-year readmission direct cost 3347 (12110) 2611 (5202) 4121 (16468) ,0.01d
Total direct cost (initialþ30-day readmission) 7337 (11366) 6220 (6434) 8522 (14823) ,0.01d Total direct cost (initialþ1-year readmission) 9556 (14307) 8122 (7703)
11062 (18790) ,0.01d BHCS-HF, Baylor health care system-heart failure; SD, standard deviation; aBased on chi-square test; bBased on a competing risk model; cBased on
Robust ANOVA; dBased on t-test.
…………………………………………………………………………………………………………………………………………………..
……
Table 4 Unadjusted and adjusted effect of BHCS-HF order set on mortality, readmission, core measures compliance, length of stay and direct cost for heart failure
patients discharged from BHCS hospitals from December 2007 to March 2009
Safety and effectiveness indicators Unadjusted Propensity score adjusteda
In-hospital mortality, OR (95% CI) 0.57 (0.34;0.97) 0.49 (0.28;0.88) 30-day mortality, OR (95% CI) 0.78 (0.58;1.05) 0.81 (0.58;1.13) 30-day readmission, RR (95% CI)
0.93 (0.77;1.14) 0.91 (0.73;1.14) Core-measure compliance, OR (95% CI) 1.55 (1.15;2.10) 1.51 (1.08;2.12) Financial and ef?ciency indicators Unadjusted Propensity score
adjustedb Length of stay, difference (95% CI) (days) 20.15 (20.38;0.08) 20.07 (20.34;0.17) Cost ($), difference (95% CI)c Initial admission direct cost 21408
(22011;2806) 2685 (21287;287) 30-day readmission direct cost 2820 (21500;2141) 2665 (21379;49) One-year readmission direct cost 21614 (22676;2552) 21224 (22276;2171)
Total direct cost (initialþ30-day readmission) 22229 (23150;21308) 21350 (22804;2396) Total direct cost (initialþ1-year readmission) 23022 (24240;21806) 21909
(23143;2676) OR, odds ratio; CI, con?dence interval; RR, risk ratio. All signi?cant results are in boid. aPropensity score covariates: age, gender, race, type of
physician (hospitalists vs. non-hospitalists), APR DRG risk of mortality, facility, payor type and quarter of discharge; bPropensity score covariates: age, gender,
race, type of physician (hospitalists vs. non-hospitalists), APR DRG severity of illness, facility, payor type and quarter of discharge; cCost difference obtained via
bootstrap recycled prediction algorithm discussed in the statistical analysis section.
Ballard et al.
442
Our study adds to the strength of the evidence favoring implementation of standardized order sets to increase com- pliance with clinical guidelines for heart failure
and positively impact clinical outcomes. Improved heart failure outcomes have previously been demonstrated with tools to increase the use of evidence-based orders [10,
26]. Narrowing treatment choices to reduce variation in care through the use of evidence-based medicine simultaneously supports smooth work?ow and safe practices, and
reduces risk of errors [27, 28]. A recent publication emphasizes that outcome improve- ments can be achieved when clinicians practice in accordance with clinical
guidelines [29]. The study results are also consistent with the observation that collection of performance data and feedback to phys- icians and other decision-makers
is an essential part of quality improvement [30]. Monitoring of order set utilization and feedback of usage using academic detailing and monthly reporting to hospital
leadership was an essential component of promoting order set adoption in our study. The decrease seen in the direct cost measures re?ects obser- vations in previous
reports. For example, implementation of a heart failure clinical pathway in a 376-bed community hospital lowered median costs of hospitalization from $4500 to $2,250
following strati?cation of patients into moderate and severe comorbidity groups [25]. Given that inpatient hospitalization costs account for 50–70% of the health costs
of patients with heart failure in Western industrialized nations [1, 31, 32], tools that decrease hospitalization costs likely have substantial impact on the overall
cost of heart failure care.
Conclusion Improving heart failure outcomes and decreasing costs associated with heart failure care are of pressing importance as the aging of the population and the
decreasing mortality rates for ischemic cardiac events are combining to increase both incidence and prevalence of this condition [33]. In our study, analysis of
administrative data showed improved clini- cal and ?nancial outcomes in a large integrated health system associated with the deployment of a standardized heart failure
order set. In addition to possible clinical bene?ts of reduced inpatient mortality for heart failure patients, the potential cost savings demonstrated are of timely
importance given the current market challenges, and the growing demands to control escalations in the cost of care. Given the recently demonstrated disconnection
between performance on the CMS core measures for heart failure and patient outcome measures [34], hospitals should consider broader- reaching quality improvement tools
that span the full spec- trum of care during the patient’s hospitalization. Our results suggest that implementation of evidence-based standardized order sets may be an
accessible tool whereby hospitals can improve performance in a relatively short period of time.
Implications Our study provides evidence that health systems with com- mitted leadership and strategically dedicated resources can
successfully deploy and foster the rapid adoption of a stan- dardized order set across multiple facilities. Order sets should be based on evidence and developed to be
compliant with international guidelines for the management of heart failure. Engagement of front-line providers, including phys- ician champions and other local
leaders at each facility, is essential to ensuring application of the order set in the majority of appropriate cases. While recognizing the limitations of our
observational study, we believe that the use of a standardized heart failure order set for all of the 1 million heart failure admissions in the USA annually [14] could
result in a signi?cant improve- ment in heart failure care while reducing costs. Based upon an average in-hospital mortality rate of 2.97% estimated from a large
sample of university hospitals (data from the University HealthSystem Consortium) [35], the 51% decrease in risk of in-hospital mortality and $1909 decrease in total
direct cost (initial hospitalization plus 1-year all-cause read- mission costs) associated with use of a standardized heart failure order set use within BHCS translate
into annual savings of 15 147 in-hospital deaths and $1.9 billion dollars nationally. Further research using clinical data is needed to con?rm our results, and
ideally, this intervention should be tested in a multicenter randomized controlled trial.
References
1. Lloyd-Jones D, Adams R, Carnethon M et al. Heart disease and stroke statistics—2009 update: a report from the American Heart Association Statistics Committee and
Stroke Statistics Subcommittee. Circulation 2009;119:e21–181. 2. The Cardiac Insuf?ciency Bisoprolol Study II (CIBIS-II): a ran- domised trial. Lancet 1999;353:9–13.
3. Hjalmarson A, Goldstein S, Fagerberg B et al. Effects of controlled-release metoprolol on total mortality, hospitalizations, and well-being in patients with heart
failure: the Metoprolol CR/XL Randomized Intervention Trial in congestive heart failure (MERIT-HF). MERIT-HF Study Group. JAMA 2000;283:1295–302. 4. Packer MCA,
Charlon V, Cohen-Solal A et al. for the CONSENSUS Trial Study Group. Effects of enalapril on mortality in severe congestive heart failure. Results of the Cooperative
North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med 1987;316:1429–35. 5. Yusef S, Pitt B, Davis CE et al. for the SOLVD Investigators. Effect of
enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 1991;325:293–302. 6. Fonarow GC, Abraham
WT, Albert NM et al. Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF): rationale and design. Am Heart J
2004;148:43–51. 7. Fonarow GC. The role of in-hospital initiation of cardioprotec- tive therapies to improve treatment rates and clinical outcomes. Rev Cardiovasc Med
2002;3 (Suppl. 3):S2–S10.
Impact of a heart failure order set
443
8. Ahmed A, Allman RM, DeLong JF et al. Age-related underutili- zation of angiotensin-converting enzyme inhibitors in older hospitalized heart failure patients. South
Med J 2002;95:703–10. 9. Jencks SF, Huff ED, Cuerdon T. Change in the quality of care delivered to Medicare bene?ciaries, 1998–1999 to 2000–2001. JAMA 2003;289:305–12.
10. Fonarow GC, Abraham WT, Albert NM et al. In?uence of a performance-improvement initiative on quality of care for patients hospitalized with heart failure: results
of the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure (OPTIMIZE-HF). Arch Intern Med 2007;167:1493–502. 11. Williams
SC, Schmaltz SP, Morton DJ et al. Quality of care in U.S. hospitals as re?ected by standardized measures, 2002– 2004. N Engl J Med 2005;353:255–64. 12. Fonarow GC,
Yancy CW, Heywood JT. Adherence to heart failure quality-of-care indicators in US hospitals: analysis of the ADHERE Registry. Arch Intern Med 2005;165:1469–77. 13.
Vogeli C, Kang R, Landrum MB et al. Quality of care provided to individual patients in US hospitals: results from an analysis of national Hospital Quality Alliance
data. Med Care 2009;47:591–9. 14. Hunt SA, Abraham WT, Chin MH et al. ACC /AHA 2005 Guideline Update for the Diagnosis and Management of Chronic Heart Failure in the
Adult: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Update the 2001 Guidelines for
the Evaluation and Management of Heart Failure): developed in collaboration with the American College of Chest Physicians and the International Society for Heart and
Lung Transplantation: endorsed by the Heart Rhythm Society. Circulation 2005;112:e154–235. 15. Ballard DJ, Ogola G, Fleming NS et al. The Impact of standardized order
sets on quality and ?nancial outcomes. In: Rockville (ed). Advances in Patient Safety: New Directions and Alternative Approaches. MD: Agency for Healthcare Research
and Quality, 2008. 16. Fleming NS, Ogola G, Ballard DJ. Implementing a standardized order set for community acquired pneumonia: impact on mor- tality and cost. Jt Comm
J Qual Saf 2009;35:414–21. 17. Joint Commission. National Hospital Quality Measures Speci?cations Manual. Available at http://www.jointcommission.
org/PerformanceMeasurement/PerformanceMeasurement/ Current+NHQM+Manual.htm (24 September 2010, date last accessed). 18. National Technical Information Service. Social
Security Administration’s Death Master File. Available at http://www. ntis.gov/products/ssa-dmf.aspx (24 September 2010, date last accessed). 19. Manning WG, Basu A,
Mullahy J. Generalized modeling approaches to risk adjustment of skewed outcomes data. J Health Econ 2005;24:465–88.
20. R Development Core Team. R: A language and environment for stat- istical computing. Vienna, Austria: R Foundation for Statistical Computing, 2009. 21. Glick HA,
Doshi JA, Sonnad SS et al. Economic Evaluation in Clinical Trials. Oxford: Oxford University Press, 2007. 22. Meleskie J, Eby D. Adaptation and implementation of
standar- dized order sets in a network of multi-hospital corporations in rural Ontario. Healthc Q 2009;12:78–83. 23. Reingold S, Kulstad E. Impact of human factor
design on the use of order sets in the treatment of congestive heart failure. Acad Emerg Med 2007;14:1097–105. 24. Ranjan A, Tarigopula L, Srivastava RK et al.
Effectiveness of the clinical pathway in the management of congestive heart failure. South Med J 2003;96:661–3. 25. Panella M, Marchisio S, Di Stanislao F. Reducing
clinical vari- ations with clinical pathways: do pathways work? Int J Qual Health Care 2003;15:509–21. 26. Panella M, Marchisio S, Gardini A et al. A cluster
randomized controlled trial of a clinical pathway for hospital treatment of heart failure: study design and population. BMC Health Serv Res 2007;7:179. 27. Pettker CM,
Funai EF. Clinical practice guidelines and scienti?c evidence. JAMA 2009;302:143; author reply 5–7. 28. Antman EM, Gibbons RJ. Clinical practice guidelines and
scienti?c evidence. JAMA 2009;302:143–4; author reply 5–7. 29. Tricoci P, Allen JM, Kramer JM et al. Scienti?c evidence under- lying the ACC/AHA clinical practice
guidelines. JAMA 2009;301: 831–41. 30. Howlett JG. Performance feedback: a common thread in the process to provide optimal heart failure care. J Am Coll Cardiol
2009;53:423–5. 31. Neumann T, Biermann J, Erbel R et al. Heart failure: the commonest reason for hospital admission in Germany: medical and economic perspectives.
Dtsch Arztebl Int 2009; 106:269–75. 32. Stewart S, Jenkins A, Buchan S et al. The current cost of heart failure to the National Health Service in the UK. Eur J Heart
Fail 2002;4:361–71. 33. Lloyd-Jones DM, Larson MG, Leip EP et al. Lifetime risk for developing congestive heart failure: the Framingham Heart Study. Circulation
2002;106:3068–72. 34. Fonarow GC, Peterson ED. Heart failure performance measures and outcomes: real or illusory gains. JAMA 2009;302:792–4. 35. Brigham and Women’s
Hospital. Quality of Patient Care: Cardiovascular Center of Excellence – Heart Failure Quality Measures. Available at http://www.brighamandwomens.org/
Quality/Cardiovascular/HF.aspx#6 (17 July 2009, date last accessed).
Ballard et al.
444
Copyright of International Journal for Quality in Health Care is the property of Oxford University Press / UK and its content may not be copied or emailed to multiple
sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.
RESEARCH ARTICLE Open Access Predicting readmission of heart failure patients using automated follow-up calls Shelby Inouye1, Vasileios Bouras2, Eric Shouldis3, Adam
Johnstone3, Zachary Silverzweig2 and Pallav Kosuri4*
Abstract Background: Readmission rates for patients with heart failure (HF) remain high. Many efforts to identify patients at high risk for readmission focus on
patient demographics or on measures taken in the hospital. We evaluated a method for risk assessment that depends on patient self-report following discharge from the
hospital. Methods: In this study, we investigated whether automated calls could be used to identify patients who are at a higher risk of readmission within 30 days. An
automated multi-call follow-up program was deployed with 1095 discharged HF patients. During each call, the patient reported his or her general health status. Patients
were grouped by the trend of their responses over the two calls, and their unadjusted 30-day readmission rates were compared. Pearson’s chi-square test was used to
evaluate whether readmission risk was independent of response trend. Results: Of the 1095 patients participating in the program, 837 (76%) responded to the general
status question in at least one of the calls and 515 (47%) patients responded to the general status question in both calls. Out of the 89 patients exhibiting a
negative response trend, 37% were readmitted. By contrast, the 97 patients showing a positive trend and the 329 patients showing a neutral trend were readmitted at
rates of 16% and 14% respectively. The dependence of readmission on trend group was statistically significant (P<0.0001). Conclusions: Patients at an elevated risk of
readmission can be identified based on the trend of their responses to automated follow-up calls. This presents a simple method for risk stratification based on
patient self-assessment. Keywords: Heart failure, Risk assessment, Post-discharge follow-up, Readmission prediction
Background An estimated 5.1 million people in the United States suffer from heart failure, and approximately 550,000 new diagnoses are made each year [1,2]. Although
notable improvements have been made in the treatment of pa- tients diagnosed with heart failure (HF), the national average readmission rate remains stagnant, with
approxi- mately one in four patients readmitted within 30 days of discharge [3]. In addition to the excessive trauma this may cause for the patient, readmissions can
also place a large financial burden on the hospital. In FY2013 alone, Centers for Medicare and Medicaid Services penalized 2,200 U.S. hospitals a combined $280 million
[4]. While it may not be possible to determine the exact proportion of preventable readmissions, evidence shows
that comprehensive discharge planning and early follow- up can reduce the likelihood of readmission in HF patients [5,6]. The American Heart Association has ad-
vocated for post-discharge follow-up and has published a set of guidelines for post-discharge telephone calls [7]. However, due to the high volume of discharged
patients, it is likely necessary to perform targeted interventions based on risk stratification. Prior studies have identified demographic and clinical patient data,
such as marital status, insurance status, and comorbidities, as predictive factors for readmission [8-15]. While these models may provide considerable value, they tend
to omit the poten- tially significant component of the patient’s self-reported general condition. One efficient method for obtaining patient information post-discharge
is through the use of automated calling. Automated calls have been used in many studies to monitor patients and attempt to minimize readmission [16,17]. However,
despite the often unique
* Correspondence: pallavkosuri@fas.harvard.edu 4Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, MA 02138, USA Full list
of author information is available at the end of the article
© 2015 Inouye et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,
unless otherwise stated.
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 DOI 10.1186/s12911-015-0144-8
insight provided into a patient’s condition, patient re- sponse data have yet to be used as an effective means for risk stratification. In this study, we investigated
whether automated calls could be used to identify patients with HF who were at a higher risk of readmission within 30 days of hospital dis- charge. Our analysis showed
that for this category of pa- tients, self-assessment could provide a simple and efficient means for risk stratification.
Methods Study population and eligibility The study population was comprised of individuals enrolled in an automated post-discharge follow-up call program. The program
was initiated at Charleston Area Medical Center (CAMC) in December of 2010 with the purpose of improving quality of care and patient out- comes. All enrolled patients
for this study were discharged from CAMC in Charleston, West Virginia between December 2010 and September 2012. Individuals eli- gible for the call program were over
18 years of age, English-speaking, had a valid phone number, and had been admitted with a diagnosis of HF. The automated call program was used to deliver infor- mation
to CAMC clinicians regarding the patient’s condi- tion following discharge. A third party executed CAMC’s Business Association Agreement prior to providing any
automated call services. The third party was in full com- pliance with all HIPAA standards, rules and regulations. The study was performed using data acquired from a
pro- gram that was implemented for the purpose of improving care management services at CAMC. Therefore, there was no requirement for external institutional review
board approval [18].
Follow-up call design and protocol The call script was generated via a collaborative effort between the third party (CipherHealth LLC, New York) and physicians at CAMC
and JFK Medical Center in Edison, New Jersey. Follow-up questions were formu- lated based on best practice guidelines published by the American Heart Association, and
the American College of Cardiology [7,19]. Upon discharge from the hospital, patient information was stored in the third party database. The call program consisted of
two automated phone calls: patients re- ceived the first call within 48 hours of discharge and the second call seven days later. No calls were made on weekends, so if
a patient was scheduled to be called dur- ing the weekend, his/her call was transferred to Monday. Therefore, patients who were discharged on Thursday or Friday,
received their first call (two days after discharge) on Monday instead of Saturday or Sunday, respectively. Patients who were discharged during the weekend
received their second call (nine days after discharge) on Monday after the following weekend instead of Saturday or Sunday respectively. Patients input their responses
using a touch-tone phone. On the first call, four questions related to general health status, medica- tions, follow-up appointments, and weight gain were asked. On the
second call, the same inquiries were made, and a fifth question regarding maintenance of a low-sodium diet was included. We hypothesized that responses to the general
health status would provide predictive information about a pa- tient’s readmission risk. The general status question reads, “How are you feeling compared to when you
were discharged from the hospital?” Possible responses were 1-better, 2-about the same, 3-worse, or 4-much worse. A trend was generated based on patients’ responses to
the general status question on the second call as compared to the first. Patients who responded more positively on the second call than on the first were considered to
have a positive trend. Patients who responded more negatively on the second call than on the first were considered to have a negative trend. We hypothesized that
patients showing a negative trend would be more likely to be re- admitted than those with a positive or neutral trend. Several measures were taken to maximize
compliance. Patients were notified by the nursing staff of the ap- proximate time and date of the calls. In addition, the au- tomated calls were made using a phone
number from the hospital, and the voice talent reflected the accent of the region. In the event of a missed call, the patient re- ceived a voicemail explaining that
another attempt would be made in the near future. Up to four call at- tempts would be made on the scheduled call day.
Data collection and statistical analysis Patient response data were delivered via automatic re- ports. Prior to analysis, all identifying data such as name, date of
birth, and medical record number were removed. For the trend analysis, patients were assigned a value of “positive”, “neutral”, or“negative” based on the trend of
their answers. Readmissions within 30 days of discharge were recorded and unadjusted readmission rates were calculated for each trend group. Pearson’s chi-square test
of independence was used to assess whether the un- adjusted readmission rates were independent of response trend group.
Results Out of the 1095 HF patients selected for the study, 837 patients (76%) responded to the general status question in at least one call, and 515 patients (47%)
responded to the general status question in both calls (Figure 1). A total of 244 patients (22%) were readmitted within 30 days of discharge from the hospital, which
is
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 Page 2 of 6
consistent with the nationwide average rate of readmis- sion [20]. The outcomes for different patient groups are summarized in Table 1. The rate of readmissions among
patients who answered the general status question at least once was 21% as compared to 27% for those who did not answer the question. This difference was found to be
statistically significant with P=0.03.
Of the 515 patients who completed both follow-up calls, 89 exhibited a negative response trend, 329 exhibited a neutral trend, and 97 exhibited a positive trend. Among
patients with a negative trend, the readmission rate was 37%. Among patients with positive or neutral trends, the readmission rates were 16% and 14%, respectively.
With a P value less than 0.0001, trend group was found to be a significant predictor of readmission rate. Further analysis revealed a relationship between re-
admission probability and the patient’s self-assessed sta- tus in the second follow-up call (P<0.0001). 622 patients answered the general health status question in the
second call. Of those patients, 412 responded feeling better, 152 responded feeling the same, and 58 responded feeling worse or much worse. The readmission rates for
patients feeling better, same, and worse/much worse were 13%, 24%, and 43%, respectively. Results from the first call also revealed a difference in readmission rate
among the groups, however with less significance (P=0.03).
Discussion Predicting readmission Our study found that patients who responded to two automated follow-up calls could be stratified by readmis- sion risk based on their
self-assessments of health in two
Figure 1 Patient inclusion flow chart.
Table 1 Readmission rates by patient group Readmitted patients Not readmitted patients Total patients Risk index P-value All patients 244 851 1095 22% General status
question Responded to general status question 0.0323 Yes 174 663 837 21% No 70 188 258 27% Total 244 851 1095 22% Response trend <0.0001 Neutral trend 46 283 329 14%
Positive trend 16 81 97 16% Negative trend 33 56 89 37% Total 95 420 515 18% First call response 0.0324 Better 80 363 443 18% Same 60 164 224 27% Worse or much worse
14 49 63 22% Total 154 576 730 21% Second call response <0.0001 Better 53 359 412 13% Same 37 115 152 24% Worse or much worse 25 33 58 43% Total 115 507 622 18%
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 Page 3 of 6
automated phone calls. Patients who displayed a self- reported decline in condition following discharge were more than twice as likely to be readmitted as those who
reported a neutral or improved condition over the two phone calls. Useful risk information was also obtained from the second call alone. Patients who responded
negatively were readmitted almost three times as frequently as pa- tients who responded positively or neutrally. These two methods could be complementary, since these
latter high-risk patients were not necessarily represented in the trend group analysis (these patients may not have answered the first call). Furthermore, these
patients may have responded negatively on both calls and thus have been included in the neutral trend group. This study takes a new approach to risk stratification.
Many preceding efforts have been made to stratify patients based on medical records and data obtained during the hospital stay. For instance, Krumholz et al. reviewed
2,176 patients in 18 hospitals and derived a model comprising four independent predictors, which included hospitalization in the prior year, medical history of HF,
medical history of diabetes mellitus, and serum creatinine levels at discharge [9]. Philbin and DiSalvo accessed a data set including 42,731 patients in 236 hospitals
and derived a risk score based on 11 variables, including black race/ethnicity, primary insurance of Medicare or Medicaid, medical history of ischemic heart disease,
the use of telemetry during hospitalization, etc [10]. Although successful within the scope of their stud- ies, models such as these have been shown to lack
consistency when compared to other studies. In a review of statistical models for predicting HF readmission, 117 studies were examined and it was discovered that few
characteristics were consistently associated with readmis- sion [8]. With regard to risk stratification, no studies to date have demonstrated strong model
discrimination for readmission [14,21]. We speculate that a risk model that includes dynamic patient-reported data, such as the data recorded in this study, may help
strengthen discrimination.
Preventing readmission With the objective of improving healthcare quality and reducing costs, the United States government has in- creasingly encouraged hospitals to
reduce preventable readmissions. In 2009, Medicare began publicly report- ing 30-day risk-standardized readmission rates for HF, acute myocardial infarction, and
pneumonia [22]. Al- though quality of inpatient care is an important factor associated with early readmission, there is also evidence for the efficacy of reaching out
to high-risk patients fol- lowing discharge [16,23]. A randomized controlled study showed that patients with HF who received telephone
care post discharge had 84% lower HF-related readmis- sion charges (P<0.05) than the usual care group. This result suggests that follow-up care may lead to reduc-
tions in readmissions, emergency visits, and cost of care [24]. More recently, a Cochrane review of 30 peer- reviewed randomized controlled trials found that telemo-
nitoring and structured telephone support decreased the rate of hospitalization in patients with HF [16]. Studies have also presented contrary evidence. Follow- ing
the Cochrane review, a 2012 review of studies in- volving remote monitoring of patients with HF showed inconsistent results with regard to outcome improve- ment [17].
Of importance, however, were the inclusion criteria for the studies analyzed in the referenced re- views. In the Cochrane review, a program was classified as being
“structured telephone support” if the remote care were delivered using simply a telephone, and a pro- gram was considered “telemonitoring” if there were digital
transmission of physiologic or other non-invasive data [16]. This is contrasted with the 2012 review in which data from more invasive means such as implanted devices
were included [17]. It is possible that certain interventional studies reported high readmission rates due to increased anxiety or even increased complications in
patients using self-monitoring devices. The mixed re- sults may also reflect the dependence on quality of follow-up. Two large trials were recently published showing
no significant difference in readmission for patients with heart failure participating in a telemonitoring program as compared to the control group [25,26]. In a study
by Chaudhry, telemonitoring was performed using a telephone-based interactive-response system. Informa- tion was collected from discharged heart failure pa- tients
regarding symptoms and weight, and responses were reviewed by clinicians. Patients who triggered variances in their responses received follow-up care. While it was
found that telemonitoring plus clinician intervention did not significantly improve readmission rates, the authors of that study pointed out that other telemonitoring
studies may have yielded positive results due to an especially motivated follow-up staff [25,27]. Based on the varied success of such studies, it appears that the
specific type of intervention plays a significant role in successfully preventing readmission. By identifying high-risk patients, the method presented in this study
can be used to more efficiently direct re- sources for follow-up care. In addition, such targeting could be of value when comparing the efficacy of various follow-up
strategies.
Limitations We observed a difference in readmission rates between patients who adhered to the call program and patients
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 Page 4 of 6
who did not. Readmission risk was higher for non- adherent patients, which raises a concern about behav- ioral differences between these two patient groups. For
instance, it is possible that patients who complied with the program were more inclined to follow their dis- charge instructions and therefore had a lower likelihood
of readmission. It is also possible, however, that some of the non-adherent patients experienced very early re- admission. Since dates of readmission were unavailable,
we were not able to investigate this further. Regardless, it may be useful to identify characteristics associated with non-adherent patients since they displayed a
higher rate of readmission. Our data showed that non-adherent patients tended to be younger than the patients who responded to calls. This might indicate that older
patients consider post-discharge instructions more seriously. On the other hand, we found that the readmission rate did not differ significantly between different age
groups, indi- cating that age difference alone could not explain the higher readmission rate of non-adherent patients. Another important limitation is the lack of
response to the first call. Of the 1095 patients called, 365 (33%) pa- tients either failed to answer the first call or did not re- spond to the general status
question on the first call. Studies show that 32% of 30-day readmissions in heart failure patients occur within the first seven days post- discharge [28]. Since the
first call went out within 48 hours of discharge and the second call followed seven days later, a potentially significant number of patients at risk for readmission
could not be identified. However, 107 of the 365 who failed to respond to the question on the first call responded on the second call, bringing the total number of
completely non-adherent patients down to 258. Addressing this limitation, it is likely that stron- ger predictability could be attained by increasing the overall
compliance of the program. This could be achieved by providing better information to the patients or by decreasing the number of questions asked in the phone calls. It
is important to note that no tangible incentives were offered to patients to complete the auto- mated calls, so the adherence rate should be representa- tive of the
general patient population. This study focused on predicting readmissions post- discharge, and so we did not consider other factors that might contribute to risk of
readmission. A future study could incorporate demographic information, health his- tories, and other relevant characteristics of automated call respondents to
investigate which factors contribute to readmission risk. The approach of dynamic self-report may present a potential advantage over risk stratification based on
demographic or hospital data alone. Since it captures those who report a decline in condition regardless of their demographic profiles and health histories, self-
assessment
may identify individuals who are overlooked with alterna- tive strategies. Furthermore, the method presented here can with advantage be used in situations when a
patient’s health history and clinical information is unavailable or unreliable. Nevertheless, in most cases, an integrated ap- proach that uses several sources of data
would likely prove advantageous.
Conclusions Hospitals and patients both stand to benefit from effi- cient post-discharge care. In particular, readmission rates can potentially be lowered through
targeted interven- tions. An important step towards this goal is finding methods that can reliably stratify patients according to their readmission risk. In this
study, we have described a simple method, using self-assessment through auto- mated phone calls, to identify patients with HF who are at a high risk of readmission.
Future studies may lead to a more integrated approach, whereby patient history and demographic data are utilized to further improve the ac- curacy of risk
stratification. We conclude that automated phone calls present an effective initial means for hospi- tals to identify and engage in targeted follow-up of high- risk HF
patients.
Competing interests Vasileios Bouras, Zachary Silverzweig, and Shelby Inouye are currently or were formerly employed by CipherHealth, LLC. The other authors declare no
competing interest. The authors declare that their affiliation did not influence the interpretation of the results.
Authors’ contributions SI wrote the manuscript and assisted with data analysis and interpretation. VB led data analysis and interpretation and assisted with writing of
the manuscript. ES coordinated data collection, implemented the automated call program in the hospital, and provided revisions to the manuscript. AJ assisted with data
collection and implementation of the automated call program. ZS conceived of the study and participated in interpretation of the results. PK directed the team with
regard to data analysis and writing the manuscript. All authors read and approved the final manuscript.
Acknowledgements We thank Alex Hejnosz for his helpful guidance and feedback.
Author details 1Keck School of Medicine of USC, Los Angeles, CA, USA. 2CipherHealth, New York, NY, USA. 3Department of Internal Medicine, Charleston Area Medical
Center, Charleston, WV, USA. 4Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, MA 02138, USA.
Received: 5 June 2013 Accepted: 4 March 2015
References 1. Hospital Outcomes of Care Measures – National Average. 2013. 2. Levy D, Kenchaiah S, Larson MG, Benjamin EJ, Kupka MJ, Ho KKL, et al. Long-Term Trends in
the Incidence of and Survival with Heart Failure. N Engl J Med. 2002;347:1397–402. 3. Epstein AM, Jha AK, Orav JE. The Relationship between Hospital Admission Rates
and Rehospitalizations. N Engl J Med. 2011;365:2287–95. 4. Rau J. Medicare Revises Readmissions Penalties. Kaiser Health News 14 March 2013.
http://kaiserhealthnews.org/news/revised-readmissions- statistics-hospitals-medicare/. Accessed 20 March 2015.
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 Page 5 of 6
5. Hernandez AF, Greiner MA, Fonarow GC, Hammill BG, Heidenreich PA, Yancy CW, et al. Relationship between early physician follow-up and 30-day readmission among
Medicare beneficiaries hospitalized for heart failure. JAMA. 2010;303:1716–22. 6. Phillips CO, Wright SM, Kern DE, Singa RM, Shepperd S, Rubin HR. Comprehensive
discharge planning with postdischarge support for older patients with congestive heart failure: a meta-analysis. JAMA. 2004;291:1358–67. 7. American Heart Association
Telephone Follow-Up Fact Sheet. http://www. heart.org/idc/groups/heart-public/@wcm/@hcm/@gwtg/documents/down- loadable/ucm_429608.pdf. Accessed 20 March 2015. 8. Ross
JS, Mulvey GK, Stauffer B, Patlolla V, Bernheim SM, Keenan PSP, et al. Statistical Models and Patient Predictors of Readmission for Heart Failure. Arch Intern Med.
2008;168:1371–86. 9. Krumholz HM, Chen YT, Wang Y, Vaccarino V, Radford MJ, Horwitz RI. Predictors of readmission among elderly survivors of admission with heart
failure. Am Heart J. 2000;139:72–7. 10. Philbin EF, DiSalvo TG. Prediction of hospital readmission for heart failure: development of a simple risk score based on
administrative data. J Am Coll Cardiol. 1999;18:855–6. 11. Volz A, Schmid J-P, Zwahlen M, Kohls S, Saner H, Barth J. Predictors of readmission and health related
quality of life in patients with chronic heart failure: a comparison of different psychosocial aspects. J Behav Med. 2011;34:13–22. 12. Hasan O, Meltzer DO, Shaykevich
SA, Bell CM, Kaboli PJ, Auerbach AD, et al. Hospital readmission in general medicine patients: a prediction model. J Gen Intern Med. 2010;25:211–9. 13. Hamner JB,
Ellison KJ. Predictors of hospital readmission after discharge in patients with congestive heart failure. Hear Lung. 2005;34:231–9. 14. Amarasingham R, Moore BJ, Tabak
YP, Drazner MH, Clark CA, Zhang S, et al. An automated model to identify heart failure patients at risk for 30-day readmission or death using electronic medical record
data. Med Care. 2010;48:981–8. 15. Shulan M, Gao K, Moore CD. Predicting 30-day all-cause hospital readmissions. Health Care Manag Sci. 2013;16:167. 16. Inglis SC,
Clark RA, McAlister FA, Ball J, Lewinter C, Cullington D, et al. Structured telephone support or telemonitoring programmes for patients with chronic heart failure.
Cochrane Database Syst Rev. 2010;8, CD007228. 17. Palaniswamy C, Mishkin A, Aronow WS, Kalra A, Frishman WH. Remote Patient Monitoring in Chronic Heart Failure.
Cardiol Rev. 2013;3:141–50. 18. Casarett D. Determining When Quality Improvement Initiatives Should Be Considered Research: Proposed Criteria and Potential
Implications. JAMA, J Am Med Assoc. 2000;283:2275–80. 19. Jessup M, Abraham WT, Casey DE, Feldman AM, Francis GS, Ganiats TG, et al. Focused Update: ACCF/AHA
Guidelines for the Diagnosis and Management of Heart Failure in Adults: A Report of the American College of Cardiology Foundation/American Heart Association Task Force
on Practice Guidelines: Developed in Collaboration with the International Society for Heart and Lung Transplantation. Circulation. 2009;2009(119):1977–2016. 20. Ross
JS, Chen J, Lin Z, Bueno H, Curtis JP, Keenan PS, et al. Recent national trends in readmission rates after heart failure hospitalization. Circ Hear Fail. 2010;3:97–
103. 21. Dharmarajan K, Krumholz HM. Strategies to Reduce 30-Day Readmission in Older Patients Hospitalized with Heart Failure and Acute Myocardial Infarction. Curr
Geriatr Reports. 2014;3:306–15. 22. Kocher RP, Adashi EY. Hospital readmissions and the Affordable Care Act: paying for coordinated quality care. JAMA. 2011;306:1794–
5. 23. Ashton CM, Kuykendall DH, Johnson ML, Wray NP, Wu L. The association between the quality of inpatient care and early readmission. Ann Intern Med. 1995;122:415–
21. 24. Jerant A, Azari R, Nesbitt T. Reducing the cost of frequent hospital admissions for congestive heart failure: a randomized trial of a home telecare
intervention. Med Care. 2001;39:1234–45. 25. Chaudhry SI, Mattera JA, Curtis JP, Spertus JA, Herrin J, Lin Z, et al. Telemonitoring in Patients with Heart Failure. N
Engl J Med. 2010;24:2301–9. 26. Koehler F, Winkler S, Schieber M, Sechtem U, Stangl K, Böhm M, et al. Impact of remote telemedical management on mortality and
hospitalizations in ambulatory patients with chronic heart failure: The telemedical interventional monitoring in heart failure study. Circulation. 2011;123:1873–80.
27. Krumholz HM, Amatruda J, Smith GL, Mattera JA, Roumanis SA, Radford MJ, et al. Randomized trial of an education and support intervention to prevent readmission of
patients with heart failure. J Am Coll Cardiol. 2002;39:83–9. 28. Dharmarajan K, Hsieh AF, Lin Z, Bueno H, Ross JS, Horwitz LI, et al. Diagnoses and timing of 30-day
readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia. JAMA. 2013;309:355–63.
Submit your next manuscript to BioMed Central and take full advantage of:
• Convenient online submission • Thorough peer review • No space constraints or color ?gure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS,
Scopus and Google Scholar • Research which is freely available for redistribution
Submit your manuscript at www.biomedcentral.com/submit
Inouye et al. BMC Medical Informatics and Decision Making (2015) 15:22 Page 6 of 6
BioMed Central publishes under the Creative Commons Attribution License (CCAL). Under the CCAL, authors retain copyright to the article but users are allowed to
download, reprint, distribute and /or copy articles in BioMed Central journals, as long as the original work is properly cited.
ORIGINAL CONTRIBUTION
Relationship Between Early Physician Follow-up and 30-Day Readmission Among Medicare Beneficiaries Hospitalized for Heart Failure
Context Readmission after hospitalization for heart failure is common. Early outpa tient follow-up after hospitalization has been proposed as a means of reducing
read mission rates. However, there are limited data describing patterns of follow-up after heart failure hospitalization and its association with readmission rates.
Objective To examine associations between outpatient follow-up within 7 days af ter discharge from a heart failure hospitalization and readmission within 30 days.
Design, Setting, and Patients Observational analysis of patients 65 years or older with heart failure and discharged to home from hospitals participating in the Orga
nized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Fail ure and the Get With the Guidelines-Heart Failure quality improvement program
from January 1,2003, through December 31, 2006.
Main Outcome Measure All-cause readmission within 30 days after discharge.
Results The study population included 30136 patients from 225 hospitals. Median length of stay was 4 days (interquartile range, 2-6) and 21.3% of patients were read
mitted within 30 days. At the hospital level, the median percentage of patients who had early follow-up after discharge from the index hospitalization was 38.3%
(interquartile range, 32.4%-44.5%). Compared with patients whose index admission was in a hos pital in the lowest quartile of early follow-up (30-day readmission
rate, 23.3%), the rates of 30-day readmission were 20.5% among patients in the second quartile (risk- adjusted hazard ratio [HR], 0.85; 95% confidence interval [Cl],
0.78-0.93), 20.5% among patients in the third quartile (risk-adjusted HR, 0.87; 95% Cl, 0.78-0.96), and 20.9% among patients in the fourth quartile (risk-adjusted HR,
0.91; 95% Cl, 0.83-1.00).
Conclusions Among patients who are hospitalized for heart failure, substantial variation exists in hospital-level rates of early outpatient follow-up after discharge.
Patients who are discharged from hospitals that have higher early follow-up rates have a lower risk of 30-day readmission.
Trial Registration clinicaltrials.gov Identifier: NCT00344513
JAMA. 2010;303(17):1716-7722 www.jama.com
\ririan F. Hernandez, Ml). MILS Melissa A. Greiner, MS Gregg C. Fonarow, Ml) Bradley G. Hammill. MS Paul A. Heidenreich, MD Clyde W. Yaiiey^NII) ~ Eric D. Peterson,
MD, MPH Lesley II. Curtis, PhD C l i n i c i a n s , p a y e r s, a n d policy makers seeking to promote efficiency and quality in health care are targeting hospital
readmission rales.’ One-fifth of Medicare beneficiaries are rehospitalized within 30 days and more than one-third within 90 days.’ Nearly 90% of these readmissions are
unplanned and potentially prevent able, which translates into $ l 7 billion or nearly 20% of Medicare’s hospital payments.2 As the most common diagnosis associated
with 30-day readmission among Medicare beneficiaries, heart failure is a prime example of the challenges in transi tional care.1 Studies suggest that care coordina
tion is important in preventing read mission.’ The days immediately fol lowing discharge are a vulnerable period due to the addition of therapies or changes to
existing medical therapy that may worsen a patient’s heart failure clinical status or other conditions such as chronic kidney disease.4 Early phy sician follow-up may
have the poten tial to reduce readmission rates. How ever, few studies have examined follow-up care after heart failure hos pitalization or its association with re
admission. In this study, we examined hospital- level variation in postdischarge physi cian follow-up and relationships be tween rates of early follow-up and patient
outcomes.
Author Affiliations: Duke Clinical Research Institute (Drs Hernandez, Peterson, and Curtis, Ms Greiner, and Mr Hammill) and Department of Medicine (Drs Hernandez,
Peterson, and Curtis), Duke University School of Medicine, Durham, North Carolina; Ahmanson-UCLA Cardiomyopathy Center, Los Angeles (Dr Fonarow), and Palo Alto VA
Medical Center, Palo Alto (Dr Heidenreich), California; and Baylor Heart and Vascular Institute, Dallas, Texas (Dr Yancy). Corresponding Author: Adrian F. Hernandez,
MD, MHS, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27715 (adrian.hernandez@duke .edu).
1716 JAMA, May 5, 2010—Vol 303, No. 17
PHYSICIAN FOLLOW-UP AND 30-DAY READMISION
M ETH O D S Data Sources We linked Medicare inpatient claims data from January 1,2003, through De cember 31,2006, with data from the Or ganized Program to Initiate
Lifesaving Treatment in Hospitalized Patients With Heart Failure (OPTIMIZE-HF) and the Get With the Guidelines-Heart Failure (GWTG-HF) registries. In 2005, the former
program transitioned to the lat ter under the sponsorship of the Ameri can Heart Association. The registries had the same design, inclusion criteria, and data
collection methods.5″ Patients were eligible for inclusion in the registries if they were admitted for an episode of worsening heart failure or had devel oped
significant heart failure symp toms during a hospitalization for which heart failure was the primary discharge diagnosis. Hospital teams used heart fail ure case-
ascertainment methods simi lar to those used by the Joint Commis sion and submitted data on medical history, signs and symptoms, medica tions, contraindications for
or intoler ance of medications, and diagnostic test results via a Web-based registry. The rep- resentativeness and validity of the OPTIMIZE-HF registry have been de
scribed previously.8 For each patient in this study, we ob tained Medicare inpatient, outpatient, carrier, and denominator files from 2003 through 2007. We used the
in patient files to examine readmission rates and the denominator files to ex amine mortality rates. We used the car rier files to examine outpatient transi tional
care after discharge from the initial heart failure hospitalization. The carrier liles contain claims from non- institutional providers for services cov ered under
Medicare Part B, including Healthcare Common Procedure Cod ing System (HCPCS) codes, physician specialty codes, and service dates. We used data from 2007 for patients
whose index discharge occurred in Decem ber 2006. The final date for 30-day pa tient follow-up was January 30, 2007. The institutional review board of the Duke
University Health System ap proved this study.
Study Population We used indirect identifiers to link data for patients 65 years or older from both registries with inpatient Medicare claims files, a method described
previ ously by Hammill et al.6 Using this method, we linked 62 311 (78%) of the 79 837 program-eligible hospitaliza tions to Medicare inpatient claims. Eli gible
patients were enrolled in fee-for service Medicare for al least 30 days af ter the index hospitalization and were discharged from a hospital participat ing fully in
either program. If a patient had multiple hospitalizations, we se lected the first as the index hospital ization. We excluded 9166 patients who were discharged to a
skilled nursing fa cility and 804 discharged to hospice care. We excluded 1390 patients from 143 hospitals that had fewer than 25 pa tients rem aining after prior
exclu sions.
Early Follow-up
At the patient level, the association be tween time from hospital discharge to outpatient follow-up with a physician and risk of readmission is confounded by severity
of illness. Patients who have more severe heart failure, have greater comorbid illness, or are medically less stable are commonly seen sooner after hospital discharge
but also are at greater risk of readmission. To avoid this con founding, we examined associations be tween hospital patterns of early fol low-up and patient-level
outcomes. We defined early follow-up as an outpa tient evaluation and management visit with a physician (HCPCS codes 992.xx- 994.xx) within 7 days after discharge
from the index hospitalization. We se lected 7 days to be consistent with cur rent efforts to improve transitional care.9 We excluded emergency department vis its
from calculations of early follow-up rates because they were unplanned vis its not reflective of a system of care. We used the physician specialty code recorded on
the claims to classify phy sicians as cardiologists or general in ternists. We considered patients who visited the same physician during the index hospitalization
and at follow-up
to have experienced continuity of care. Using Medicare hospital identifiers from the index hospitalization claims, we ag gregated patient-level follow-up at the
hospital level and calculated the pro portion of discharged patients who re ceived early follow-up by hospital. We used the same approach to summarize hospital-level
follow-up at 14, 21, and 28 days after discharge. We used the hospital rate of early physician follow up— grouped in quartile rankings— as the exposure of interest.
Patient and Hospital Characteristics
From the registry data, we obtained pa tient demographic characteristics, medi cal history, results of admission labo ratory tests and examinations, discharge
pharmacy records, and procedural in formation from the index hospitaliza tion. Patients were assigned to race/ ethnicity categories using options available on the
case report form. We used the reported category “black” and combined all others as “nonblack.” Variables in this analysis had low rates of missingness (ie, <5% of
records), with the exception of evaluation of left ventricular function (12.1%). For con tinuous variables and the variable for evaluation of left ventricular
function, we created categorical variables that in cluded a category for missing values. For other dichotomous variables (ie, sm oker w ithin the past year, dis
charge processes and performance mea sures, and index hospitalization pro cedures), we imputed missing values to “no.” From the American Llospital As sociation
annual survey,10 we ob tained information on membership in the Council of Teaching Hospitals, pres ence of a cardiac intensive care unit, availability of adult
diagnostic or in terventional catheterization and heart transplantation services, total number of beds, annual number of Medicare and Medicaid discharges, and rural
and geo graphic location.
Outcomes
We examined Medicare claims for up to 30 days after discharge from the in-
JAMA, May 5, 2010—Vol 30.3. No. 17 1717
PHYSICIAN FOLLOW-UP AND 30-DAY READM1S10N
Table 1. Characteristics of the Study Population by Quartile of Hospital-Level Rates of Early Follow-up
Hospital-Level Percentage Rate of Early Follow-up by Quartile, No. (%)
Characteristics
1
1 (<32.4)
2 (32.4-37.9)
3 (38.3-44.5)
I 4 P (>44.5) Value
No. of patients 7081 8662 7812 6581
Age. ya 65-69 1000(14.1) 1207(13.9) 1143 (14.6) 769(11.7) <.001 70-74 1312(18.5) 1526 (17.6) 1409(18.0) 1046(15.9) <.001 75-79 1630(23.0) 1873(21.6) 1653(21.2) 1436
(21.8) .04 2=80 3139(44.3) 4056 (46.8) 3607 (46.2) 3330(50.6) <.001 Female sex 3847 (54.3) 4576 (52.8) 4135(52.9) 3511(53.4) .24 Black race 1337 (18.9) 715(8.3) 772
(9.9) 390 (5.9) <.001
Medical history Anemia 1289(18.2) 1492 (17.2) 1253(16.0) 990(15.0) <.001 Atrial arrhythmia 2475 (35.0) 3239 (37.4) 2675 (34.2) 2303(35.0) <.001 COPD 2022 (28.6) 2589
(29.9) 2023 (25.9) 1636(24.9) <.001 Chronic renal insufficiency 1355(19.1) 1657(19.1) 1375 (17.6) 1068(16.2) <.001 Coronary artery disease 3839 (54.2) 4807 (55.5)
4032(51.6) 3430(52.1) <.001 Depression 672 (9.5) 912(10.5) 565 (7.2) 531 (8.1) <.001 Diabetes mellitus 2900(41.0) 3446 (39.8) 3118(39.9) 2428(36.9) <.001
Hyperlipidemia 2545 (35.9) 3563(41.1) 2759 (35.3) 2246(34.1) <.001 Hypertension 5278 (74.5) 6358 (73.4) 5437 (69.6) 4612(70.1) <.001 Peripheral vascular disease 976
(13.8) 1422 (16.4) 1013(13.0) 764(11.6) <.001 PriorCVAorTIA 1098(15.5) 1489(17.2) 1188(15.2) 1022(15.5) .002 Smoker within the past year 687 (9.7) 945(10.9) 771 (9.9)
512(7.8) <.001
Findings on admission, median (IQR) Hemoglobin, g/dL 12.0 (10.7-13.4)
12.0 (10.7-13.4)
12.1 (10.8-13.4)
12.1 .30 (10.8-13.4)
Serum creatinine, mg/dL 1.4 (1.0-1.9) 1.3 (1.0-1.8) 1.3 (1.0-1.8) 1.3 (1.0-1.8) <.001
Serum sodium. mEq/L 139.0 138.0 (136.0-141.0) (135.0-141.0)
138.0 (135.0-141.0)
138.0 <.001 (135.0-141.0)
Systolic blood pressure, mm Hg
140.0 140.0 140.0 140.0 .08 (122.0-161.0) (120.0-161.0) (120.0-160.0) (120.0-160.0)
Left ventricular function LVSD 2841 (40.1) 3404 (39.3) 2947 (37.7) 2300(34.9) <.001 Preserved systolic function 3441 (48.6) 4311 (49.8) 3930 (50.3) 3301 (50.2) .16
Discharge processes or performance measures Referral to outpatient disease management program
697 (9.8) 1296 (15.0) 948(12.1) 678(10.3) <.001
Discharge instructions completed
4355(61.5) 5508 (63.6) 4633 (59.3) 3642(55.3) <.001
(3-Blocker for patients with LVSDb
2210(77.8) 2579 (75.8) 2256 (76.6) 1721 (74.8) .08
ACE inhibitor or ARB for patients with LVSDb
2023 (71.2) 2425(71.2) 2073 (70.3) 1621 (70.5) .81
Characteristics of index hospitalization Coronary artery bypass graft surgery
39 (0.6) 47 (0.5) 48 (0.6) 45 (0.7) .67
Implantable cardioverter-defibrillator
203 (2.9) 362 (4.2) 206 (2.6) 163(2.5) <.001
Percutaneous coronary intervention
110(1.6) 114(1.3) 89 (1.1) 91(1.4) .18
Right cardiac catheterization 177(2.5) 259 (3.0) 233 (3.0) 172(2.6) .15
Length of stay, median (IQR), d
4.0 (3.0-7.0)
4.0 (2.0-6.0)
4.0 (2.0-6.0)
4.0 <.001 (2.0-6.0)
Abbreviations: ACE, angiotensin-converting enzyme; ARB, angiotensin II receptor blocker; COPD, chronic obstructive pul monary disease; CVA, cerebrovascular accident;
IQR, interquartile range; LVSD, left ventricular systolic dysfunction; TIA, transient ischemic attack. SI conversion factor: To convert creatinine from mg/dL to
pmol/L. multiply by 88.4 a Percentages may not sum to 100 due to rounding. bOf the patients with LVSD, 2841 were in quartile 1, 3404 were in quartile 2. 2947 were in
quartile 3, and 2300 were in quartile 4.
dex hospitalization. We calculated die time to first readmission as the num ber of days between the index dis charge date and the readmission date. Transfers to or
from another hospital and admissions for rehabilitation (di agnosis related group 462 or an admis sion diagnosis code ofV57.xx) did not count as readmissions.
Emergency de partment visits that did not result in ad missions were not considered as a re admission outcome. We obtained mortality information from the Medi care
denominator files and calculated time to death as the number of days be tween the index discharge date and the death date.
Statistical Analysis We present hospital-level rates of early follow-up as medians with interquar tile ranges. For baseline patient and hos pital characteristics, we
present cat egorical variables as frequencies with percentages and continuous variables as medians with interquartile ranges. We grouped patients and hospitals by
quartiles of hospital rates of early fol low-up. To test for differences by quar- tile, we used tests for categorical vari ables and Kruskal-Wallis tests for
continuous variables. We used the cu mulative incidence function, which ac counts for the competing risk of death, to calculate unadjusted 30-day read mission
rates. We used Gray tests to compare 30-day readmission rates by quartile of hospital-level early follow up. We calculated unadjusted 30-day mortality rates and 30-
day mortality or readmission rates as proportions and used x2 tests to lest for differences by quartile. We used Cox proportional hazards models to examine unadjusted
and ad justed relationships between hospital- level early follow-up and 30-day read mission. In multivariable analysis, we modeled 30-day readmission as a func tion
of hospital-level early follow-up rate, age, sex, black race, medical his tory, results of admission laboratory tests and examinations, completion of discharge
instructions, referral to a heart failure disease management program.
1718 JAMA, May 5. 2010—Vol 303, No. 17
PHYSICIAN FOLLOW-UP AND 50-DAY READMISION
length of stay for the index hospital ization more than 7 days, and year of the index hospitalization. In second ary analyses, we examined associa tions between
early follow-up with a cardiologist or continuity of care and 30-day readmission. In sensitivity analy ses, we allowed early follow-up to oc cur within 14 days after
discharge. In all readmission models, we cen sored patients’ data from the analysis on the date of death as obtained from the Medicare denominator file. We used
robust standard errors to account for clustering of patients within hospi tals. We also used Cox proportional hazards models to examine relation ships between
hospital-level early fol low-up and risk-adjusted 30-day mor tality and risk-adjusted 30-day mortality or readmission. We used a signifi cance level of .05 and 2-
sided tests for all hypotheses. In all analyses, we tested the proportionality assumption for the hospital-level exposure of interest. We did not conduct a prospective
power analysis. Flowever, for the primary analysis, with the observed rate of 30- day readmission and the observed sizes of each exposure group by hospital quartile of
early follow-up, we had greater than 80% power to detect a haz ard ratio of 0.90 between any 2 expo sure groups. We used SAS version 9.2 (SAS Institute Inc, Cary,
North Caro lina) for all analyses.
RESULTS The study included 30 136 patients from 225 hospitals. Median age was 79 years, 53.3% of patients were women, and 10.7% were black. The median number of study
patients per hospital was 229, ranging from 25 to 693. Me dian length of stay was 4 days. T a b l e 1 shows characteristics of the study population by quartile of
hospital- level rates of early follow-up. The pro portion of black patients was mark edly higher among hospitals with the lowest rates of early follow-up. The rate
of coronary artery disease was similar across quartiles, whereas the propor tion of documented left ventricular sys tolic dysfunction was highest among pa
tients in the lowest quartile. Chronic obstructive pulmonary disease, depres sion, and chronic renal insufficiency were more common among patients in the lower 2
quartiles compared with pa tients in the upper 2 quartiles. In 93.7% (28 229 of 30 136) of pa tients, it was documented that an out patient follow-up was scheduled
be fore hospital discharge. However, information on the date scheduled for the follow-up visit was not available. The FIGURE shows a histogram of hos pital-level
rates of early follow-up. At the hospital level, the median rate of early follow-up was 38.3% (interquar tile range, 32.4%-44.5%) and the maxi mum was 63.7% (T a b
le 2). A median of 18.1% of patients saw the same phy sician during the index hospitaliza tion and during early follow-up. By 28 days after discharge, this rate in
creased to 50.0%. The median rate of inpatient evaluation by a cardiologist was 68.7%, but the rate of early fol low-up with a cardiologist was 7.5%. Across
quartiles, hospitals were simi lar with respect to teaching status, avail- ability of cardiac services such as cardiac catheterization and heart trans plantation,
and number of beds (eTable available at http://www.jama.com). There were higher rates of early fol low-up in the West than in other geo graphic regions.
Predictors o f 30-Day Readmission In the first 30 days after discharge, 6428 patients (21.3%) were readmitted. As shown in T a ble 3, unadjusted 30-day readmission
rates were highest among patients in hospitals in the lowest quar tile of early follow-up (23.3% readmis sion). T ab le 4 shows unadjusted and adjusted associations
between early physician follow-up and all-cause re admission. The proportionality assump tion was met for the hospital-level ex posure of interest (x3=4.66; P=.20).
After adjustment for baseline pa tient characteristics of the index hos pitalization, there was an inverse rela tionship between early follow-up and the hazard of
30-day readmission. Com pared with patients whose index hos-
Figure. Variation in Physician Follow-up Within 7 Days After Discharge
0 10 20 30 40 50 60 70 80 90 100 Hospital-Level Rate of Early Follow-up. %
Table 2. Hospital-Level Rates of Early Follow-up During the Transitional Period
Physician Visited, by Days
Rate of Early Follow-up, Median (IQR)
Any physician <7 38.3 (32.4-44.5) £14 64.6(56.6-70.0) <21 76.5(70.7-81.2) <28 81.5 (76.7-85.7) Same physician during index hospitalization and transitional period £7
18.1 (13.5-24.1) £14 34.9 (26.3-42.4) £21 44.3 (34.4-53.3) £28 50.0 (38.8-58.6) Cardiologist £7 7.5 (4.1-13.8) £14 17.1 (11.6-26.7) £21 25.2(17.3-35.7) £28 31.3(22.6-
42.3) Same cardiologist during index hospitalization and transitional period £7 3.4 (1.2-7.0) £14 8.4 (4.1-15.8) £21 13.7(7.4-22.7) <28 17.2(9.1-26.9)
pitalization occurred in a hospital in the lowest quartile of early follow-up, the risk-adjusted hazard of 30-day read- mission was significantly lower in the second
quartile. We found no signifi cant difference in readmission risk when we compared the second, third, and fourth quartiles (x2=2.20; P=.33). Neither early follow-up
with a cardi ologist nor continuity of care from the same physician during the index hos pitalization and during early fol low-up was a significant predictor of 30-
day readmission.
JAMA. May 5, 2010— Vol 303, No. 17 1719
PHYSICIAN FOLLOW-UP AND 30-DAY READMISION
In sensitivity analyses, we varied the definition of early follow-up. The me dian frequency of follow-up within 2 weeks was 64.6% (interquartile range, 56.6%-70.0%).
Results were similar when we increased the transitional pe riod from 7 days to 14 days (Table 4).
Predictors of 30-Day Mortality In the first 30 days after discharge, 1419 patients (4.7%) died. There was no sig
nificant difference in unadjusted mor tality rates by quartile of early fol low-up (Table 3). After adjustment for baseline characteristics, no significant
difference existed in the 30-day mor tality by quartile of early follow-up (quartile 1: reference; quartile 2: HR, 0.95; 95% confidence interval [Cl], 0.80-1.14;
quartile 3: HR, 0.88; 95% Cl, 0.74-1.04; quartile 4: HR, 0.84; 95% Cl, 0.69-1.03). The risk-adjusted hazard of
30-day mortality was significantly lower among patients admitted to hospitals in the highest quartile of early fol low-up with a cardiologist (quartile 4: HR, 0.75;
95% Cl, 0.62-0.90) com pared with the lowest quartile (quar tile 1: reference) but there was no sig nificant difference in quartile 2 (HR, 0.88; 95% Cl, 0.74-1.06)
or quartile 3 (HR, 0.85; 95% Cl, 0.71-1.02) com pared with the lowest quartile. Com pared with patients admitted to hospi tals in the lowest quartile of early
follow-up, the risk-adjusted hazard of 30-day mortality or readmission was 10% to 14% lower among patients ad mitted to hospitals with higher fre quency of early
follow-up (quartile I: reference; quartile 2: HR, 0.86; 95% Cl, 0.79-0.94; quartile 3: HR, 0.88; 95% Cl, 0.80-0.96; quartile4: HR, 0.90;95% Cl, 0.83-0.98).
COMMENT Despite the high risk of readmission among patients hospitalized for heart failure, most patients in this study did not visit a physician within a week of
discharge. Rates of early follow-up var ied substantially across hospitals. Most early follow-up care was handled by general internists rather than cardiolo gists
and usually not by the same phy sician who evaluated the patient dur ing the index hospitalization. Discharge from hospitals in which a greater pro portion of
patients received early fol low-up was independently associated with lower rates of all-cause 30-day re admission. Transitional care is designed to en sure
coordination and continuity in health care as patients transfer be tween locations.” Important elements of transitional care include communi cation between sending
and receiving clinicians, preparation of the patient and caregiver for what to expect at the next site of care, reconciliation of medica tions, follow-up plans for
outstanding tests, and discussions about monitor ing signs and symptoms of worsening conditions.”12 For patients with heart failure, the transition from inpatient to
outpatient care can be an especially vul-
Table 3. Rates of Mortality, Readmission, and Mortality or Readmission at 30 Days by Quartile of Hospital Rate of Early Follow-up Percentage Rate of Early Follow-up by
Quartile, No. (%) I —————————— 1
Variable
1 (<32.4)
2 (32.4-37.9)
3 (38.3-44.5)
4 (>44.5)
P Value
No. of patients 7081 8662 7812 6581 Event, 30 d Mortality3 353 (5.0) 417(4.8) 352 (4.5) 297 (4.5) .44 Readmission b 1658 (23.3) 1787(20.5) 1606(20.5) 1377(20.9) <.001
Mortality or readmission3 1849(26.1) 2015(23.3) 1813(23.2) 1544(23.5) <.001
3 Based on proportion of events. bBased on cumulative incidence function.
Table 4. Unadjusted and Adjusted Relationships Between Early Physician Follow-up by Quartile and 30-Day All-Cause ReadmissiorH___________________________________
Quartile (% of Follow-up)
Unadjusted HR (95% Cl)
P Value
Adjusted HR (95% Cl)
P Value
Model 1: early follow-up with a physician 1 (<32.4) 1 [Reference] 1 [Reference] 2 (32.4-37.9) 0.86 (0.78-0.94) .001 0.85 (0.78-0.93) <.001 3 (38.3-44.5) 0.85 (0.76-
0.94) .002 0.87 (0.78-0.96) .005 4 (>44.5) 0.87 (0.79-0.95) .002 0.91 (0.83-1.00) .05 Model 2: early follow-up with a cardiologist 1 (<4.1) 1 [Reference] 1 [Reference]
2 (4.1-7.4) 0.91 (0.82-1.02) .09 0.92(0.83-1.02) .09 3(7.5-13.8) 0.91 (0.82-1.00) .05 0.91 (0.82-1.00) .05 4 (>13.8) 0.91 (0.82-1.00) .06 0.95(0.85-1.05) .30 Model 3:
early follow-up with the same physician 1 (<13.5) 1 [Reference] 1 [Reference] 2(13.5-17.5) 0.93 (0.84-1.04) .20 0.96(0.86-1.05) .36 3(18.1-24.1) 0.91 (0.81-1.02) .11
0.94(0.84-1.04) .23 4 (>24.1) 0.93(0.83-1.03) .16 0.97 (0.87-1.08) .54 Model 4; 14-d follow-up with a physician 1 (<56.6) 1 [Reference] 1 [Reference] 2 (56.6-64.5)
0.88 (0.80-0.97) .01 0.89(0.81-0.97) .01 3 (64.6-70.0) 0.87 (0.78-0.97) .009 0.90(0.81-1.00) .04 4 (>70.0) 0.87 (0.79-0.96) .004 0.93(0.84-1.02) .13
Abbreviations: Cl, confidence interval; HR, hazard ratio. 3All models included age, sex, black race, medical history (ie, anemia, atrial arrhythmia, chronic
obstructive pulmonary disease, chronic renal insufficiency, coronary artery disease, depression, diabetes mellitus, hyperlipidemia, hyperten sion, peripheral vascular
disease, prior cerebrovascular accident or transient ischemic attack, and smoker within the past year), results of admission laboratory tests and examinations (ie,
serum creatinine level, serum sodium level, systolic blood pressure, and hemoglobin level, and left ventricular function), completion of discharge instructions,
referral to a heart failure disease management program, length of stay for the index hospitalization more than 7 days, and year of the index hospitalization.
1720 JAMA, May 5, 2010— Vol 303, No. 17
PHYSICIAN FOLLOW-UP AND 30-DAY READMISION
nerable period because of the age of the patients, complex medical regimens, the large number of comorhid conditions, and the multiple clinicians who may he involved.
As we found in this study, a central element of transitional care, outpa tient follow-up, varies significantly across hospitals and, for most pa tients, does not
occur in a timely man ner. It is common for different physi cians to care for patients in the hospital setting and outpatient settings. Early evaluation after
discharge is critical. This evaluation should include a re view of therapeutic changes and a thor ough assessment of the patient’s clinical status outside of the
highly structured hospital setting. Our findings highlight a need for im provement and greater uniformity in co ordination of care from inpatient to out patient
settings. Most follow-up during the transitional period, especially the first week, is handled by general inter nists. More than two-thirds of pa tients hospitalized
for heart failure are evaluated by a cardiologist during the inpatient stay, but less than 10% visit a cardiologist within 7 days after dis charge. As clinicians
narrow the scope of their practices to a single setting (eg, hospital) or subspecialty, coordina tion of care will become increasingly challenging.1314 Barriers to
coordina tion of care include overextended pri mary’ care, lack of interoperable com puterized records, lack of financial incentives, and lack of integrated sys
tems of care.15 Early postdischarge fol low-up may help to minimize gaps in understanding of changes to the care plan or knowledge of test results.16 Among hospitals
with higher rates of early follow-up, risk for readmission is lower. After adjustment for case mix, admission laboratory results, provi sion of discharge
instructions, and length of stay, the risk-adjusted haz ard of 30-day readmission was 15% lower among patients in hospitals in the second quartile of early follow-up
than among patients in hospitals in the low est quartile. There was no additional risk reduction associated with the third
and fourth quartiles of early follow up. All hough we found that patients discharged from hospitals with high rates of early follow-up have a lower risk of
readmission, even hospitals with the highest early follow-up rates had read mission rales of 20%. Documentation of discharge instructions is widely pre sumed to be a
process of care that helps to ensure early follow-up and better out comes, hut this measure is inconsis tent with hospital-level rates of early fol low-up and is
not associated with lower readmission rates.17 This finding raises the possibility that discharge instruc tions are becoming rote processes that do not adequately
address elements of care that ensure a safe transition.18 Hospitals and clinicians are also in terested in processes that improve 30-day mortality. We did not observe
statistically significant improvement in 30-day mortality by hospital quartiles of early follow-up by any physician. We did find that patients discharged from
hospitals with the highest rates of early follow-up by a cardiologist had lower risk of 30-day mortality, consistent with other studies of cardiology care for heart
failure.19 20 Discharge from hospitals in the higher quartiles of early follow-up was associated with a 10% to 14% lower risk of mortality or readmission com pared
with the lowest quartile of early follow-up. Validation of these find ings, and the potential for early fol low-up to improve 30-day mortality or readmission, would
he useful topics of investigation. This study also provides evidence in support of guidelines recommending the use of postdischarge systems of care.2122 Initiatives to
encourage early follow-up are ongoing.9 Achieving early follow-up may be difficult for some phy sician practices, but models of care that include nurse practitioners
or physi cian assistants under physician super vision may result in increased access to and timeliness of care. Given the low rate of early follow-up and variability
across hospitals, early follow-up is a potential measure of quality that could be integrated into heart failure performance measure sets and tar
geted for improvement by national initiatives. Our study has several limitations. First, this was an observational study and patients were not randomly as signed to
early follow-up. We cannot rule out the possibility of unmeasured confounding. Second, the analysis was restricted to fee-for-service Medicare beneficiaries enrolled
in heart failure clinical registries, and hospitals that par ticipated in the registry differed from nonparticipating hospitals.8 However, fee-for-service Medicare
beneficiaries in the OPTIMIZE-HF registry (the pre cursor to the GWTG-HF registry) were similar to all fee-for-service Medicare beneficiaries with heart failure.8
Third, we did not have access to data on home health visits and disease management programs, such as telephone or other remote monitoring, which may be im portant for
preventing 30-day reaclmis- sion. If a significant proportion of pa tients received early home health visits or monitoring but not early physi cians visits and if
these programs were effective in reducing readmissions, we would have a lower likelihood of de tecting associations between early phy sician follow-up and improved
out comes. In addition, we were unable to determine whether early follow-up vis its were with physician assistants or nurse practitioners with or without di rect
physician supervision but filed un der the physician. Fourth, data on so cioeconomic status were not available. Fifth, information about the types and extent of
discharge protocols in use by participating hospitals was not avail able. Moreover, we did not explore the relationships between early follow-up and other clinical
outcomes such as health status, quality of life, func tional status, patient satisfaction, and cause-specific readmission. Finally, the potential mechanisms by which
hos pital-level early physician follow-up rates are associated with lower 30-day rehospitalization rates could not he de termined. In conclusion, among patients hos
pitalized for heart failure, rates of phy sician follow-up within 1 week of dis-
JAMA, May 5, 2010—Vol 303, No. 17 1721
PHYSICIAN FOLLOW-UP AND 30-DAY READMISION
charge were low and varied substantially across hospitals. Among hospitals with higher rales of early follow-up, the risk of 30-day readmission was lower and
hospital-level early follow-up was inde pendently associated with 30-day read mission. Prospective studies should be performed to evaluate the effects of early
follow-up on readmission.
Author Contributions: Dr Hernandez had full access to all o f the data in the study and takes responsibility fo r the integrity o f the data and the accuracy o f the
data analysis. Study concept and design: H ernandez, G reiner, Fonarow, Heidenreich, Yancy, Curtis. Acquisition of data: Hernandez, Greiner, Fonarow, Peterson.
Analysis and interpretation of data: Hernandez, Greiner, Fonarow, Hammill, Heidenreich, Curtis. Drafting of the manuscript: Hernandez, Greiner, Fonarow. Critical
revision of the manuscript for important in tellectual content: G reiner, Fonarow , H am m ill, Heidenreich, Yancy, Peterson, Curtis. Statistical analysis: Greiner,
Hammill. Obtained funding: Hernandez, Fonarow. Administrative, technical, or material support: Fonarow, Yancy, Curtis. Study supervision: Hernandez, Fonarow, Peterson.
Financial Disclosures: Dr Hernandez reported receiv in g research su p p o rt fro m Johnson & Johnson, M edtronic, and M erck & Co; serving on the speak ers’ bureau
fo r Novartis; and receiving honoraria from AstraZeneca and M edtronic. Dr Fonarow reported re ceiving research grants or other research support from GlaxoSmithKline,
Pfizer, and the National Institutes of Health; receiving honoraria from Amgen, AstraZen eca, Boston S cientific/G uidant, G laxoSm ithKline, M edtronic, M erck,
Novartis, Pfizer, Schering Plough, and St Jude Medical; serving as a consultant to Bos ton Scientific/Guidant, GlaxoSmithKline, M edtronic, M erck, Novartis, Pfizer,
Scios, and St Jude Medical; and serving on the Am erican Heart Association’s Get W ith the Guidelines Steering Committee. He holds the Eliot Corday Chair o f
Cardiovascular M edicine and is also supported by the Ahmanson Foundation (Los A n geles, California). Dr Yancy terminated all financial re lationships in June
2008. Before 2008, D r Yancy re ported receiving research grants from Cardiodynamics. GlaxoSmithKline, Scios, M edtronic, and N itroM ed; serving as a c o n su lta n
t to o r on th e speakers’ bureaus o f AstraZeneca, Cardiodynam ics, G laxo SmithKline, M edtronic, NitroM ed, Novartis, and Scios; serving on advisory boards o f CHF
Solutions; and re ceiving honoraria from AstraZeneca, Cardiodynam ics, GlaxoSmithKline, M edtronic, Novartis, and Scios. Dr Peterson reported receiving research
support from Bristol-Myers Squibb, Sanofi-Aventis, M erck Scher ing, and Eli Lilly; and serving as the principal investi gator o f the analytic center fo r the
American Heart Association Get W ith the Guidelines Program. Dr Cur tis reported receiving research support from Johnson & Johnson and M edtronic. Drs Hernandez,
Peterson,
and Curtis have made available online detailed list ings o f financial disclosures (h ttp ://w w w .d cri.d u ke .edu/research/coi.jsp). F unding/S upport: G et W ith
the G uidelines-H eart Failure is a program of the American Heart Associa tio n and is supported in part by an unrestricted e d u c a tio n a l g ra n t fro m G la x
o S m ith K lin e and M e d tro n ic . O P T IM IZ E -H F w a s s u p p o rte d by G la xoS m ith K line. This s tu d y was sup p o rte d by Am erican Heart
Association Pharmaceutical Round table O utcom es Center grant 087 512N. Dr Hern andez was supported by Am erican Heart Associa tion Pharm aceutical Roundtable
grant 0675060N . Drs Peterson and Curtis were supported by grant U 18H S 016964 fro m the A g ency fo r H ealthcare Research and Q uality. Role o f the Sponsor: The Am
erican Heart Associa tion, GlaxoSmithKline, and M edtronic had no role in the design and conduct o f the study; collection, man agement, analysis, and interpretation
of the data; or preparation o f the manuscript. Representatives from the Am erican Heart Association reviewed and ap proved a draft of the manuscript. D isclaim er:
Dr Peterson, a con trib u tin g e dito r for JAMA, was not involved in the editorial review o f or the decision to publish this article. O nline-O nly M aterial: An
eTable is available at http: / / w w w .jam a.com . A dditional Contributions: W e thank Damon M Seils, M A , Duke University, fo r editorial assistance and m
anuscript preparation. M r Seils did n o t receive com pensation fo r his assistance apart from his em ploym ent at the institution where the study was con ducted.
REFERENCES
1. JencksSF, W illiam s M V , Coleman EA Rehospital izations am ong patients in the M edicare fee -fo r- service program . N Engl J Med. 200 9;36 0(1 4): 1418-1428.
2. Centers fo r M edicare & M edicaid Services M edi care & M edicaid Statistical Supplement; 2007. http: / / w w w .cm s.hhs.gov/M edicareM edicaidStatS upp/.
Accessed December 21,2009 . 3. Peikes D, Chen A, Schore J, Brown R. Effects of care coordination on hospitalization, quality o f care, and health care expenditures am
ong Medicare beneficia ries: 15 random ized trials. JAMA. 200 9;30 1(6 ): 603-618 4. Leendertse AJ, Egberts AC, Stoker LJ, van den Bemt PM; HARM Study Group.
Frequency o f and risk fac tors fo r preventable m edication-related hospital ad missions in the Netherlands. Arch Intern Med. 2008; 168(17): 1890-1896. 5. Fonarow
GC, Abraham W T, Albert NM , et al. Orga nized Program to Initiate Lifesaving Treatment in Hos pitalized Patients with Heart Failure (OPTIMIZE-HF): ration ale and
design. Am Heart J. 2004;148(1):43-51 6. Hammill BG, Hernandez AF, Peterson ED, Fonarow GC, Schulman KA, Curtis LH. Linking inpatient clinical registry data to M
edicare claims data using indirect identifiers. Am Heart J. 2009;157(6):995-1000. 7. Joint Commission Specifications M anual fo r Na tional Hospital Q uality
Measures; 2006. h ttp ://w w w .jointco m m ission.org /P e rform an ceM e asurem e nt /P e rfo rm a n c e M e a s u re m e n t/H is to ric a l+ N H Q M +manuals.htm.
Accessed December 21.2009.
8. Curtis LH, Greiner M A , Hamm ill BG, et al. Repre sentativeness o f a national heart failure quality-of- care registry: comparison o f OPTIMIZE-HF and non-
OPTIMIZE-HF Medicare patients. Circ Cardiovasc Qua! Outcomes. 2009;2(4):377-384. 9. American College o f Cardiology and the Institute for Healthcare Improvement.
Hospital to Home (H2H) W eb site, h ttp ://h 2 h q u a lity.o rg /. Accessed Novem ber 21, 2009. 10. American Hospital Association. AH A annual sur vey database,
2008 [database]. Chicago, IL: Am eri can Hospital Association; 2008. 11. Coleman EA, Berenson RA. Lost in transition: chal lenges and opportunities for im proving
the quality of transitional care. Ann Intern Med. 200 4;14 1(7 ): 533-536. 12. Colem an EA. Falling through the cracks: chal lenges and opportunities for improving
transitional care for persons w ith continuous complex care needs. J Am GeriatrSoc. 2003;51(4):549-555. 13. Pham HH, Grossman JM, Cohen G, Bodenheimer T. Hospitalists
and care transitions: the divorce o f in patient and outpatient care. Health Aff (Millwood). 2008;27(5):1315-1327. 14. Kuo YF, Sharma G, Freeman JL, G oodw in JS. G
row th in the care o f older patients by hospitalists in the U nited States. N Engl J Med. 2009;360(11): 1102-1112. 15. Bodenheimer T. Coordinating care: a perilous
jour ney through the health care system. N Engl J Med. 2008;358(10):1064-1071
16. Sharma G, Fletcher KE, Zhang D, Kuo YF, Freeman JL, G oodw in JS. C ontinuity o f outpatient and inpa tient care by prim ary care physicians for hospitalized
older adults. JAMA. 2009;301(16):1671-1680. 17. Jha AK, O rav EJ, Epstein A M . Public reporting o f discharge planning and rates o f readmissions. N Engl J Med.
2009;361 (27):2637-2645. 18. Romano PS. Im proving the quality o f hospital care in Am erica. N Engl J Med. 2005;353(3):302- 304. 19. Jong P, Gong Y, Liu PP, Austin
PC, Lee DS, Tu JV. Care and outcom es o f patients new ly hospital ized fo r heart failure in the com m unity treated by car diologists compared w ith other
specialists. Circulation. 2003;108(2):184-191. 20. Ezekowitz JA, van W alraven C, M cA lister FA, A rm strong PW , Kaul P. Im pact o f specialist fo l low -up in
outpatients w ith congestive heart failure. CMAJ. 2005,172(2):1 89-194. 21. Jessup M , Abraham W T, Casey DE, et al. 2009 focused update: ACCF/AHA Guidelines fo r the
Diag nosis and M anagem ent o f Heart Failure in Adults: a report o f the American College o f Cardiology Foun- dation/American Heart Association Task Force on Prac
tice Guidelines: developed in collaboration w ith the In te r n a tio n a l S o c ie ty fo r H e a rt a n d L u n g Transplantation. Circulation. 2009;119(14):1977-
2016 22. Heart Failure Society of America. Executive sum mary: HFSA 2006 Comprehensive Heart Failure Prac tice Guideline. J Card Fail 2006;12(1):10-38
1722 JAMA, May 5. 2010— V ol 303, No. 17
Copyright of JAMA: Journal of the American Medical Association is the property of American Medical Association and its content may not be copied or emailed to multiple
sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.
