- Email: [email protected]
Correlates of Early Hospital Readmission or Death in Patients With Congestive Heart Failure
Correlates of Early Hospital Readmission or Death in Patients With Congestive Heart Failure Marshall H. Chin,
MD, MPH,
and Lee Goldman,
MD, MPH
Among patients with heart failure who survive an admission to the hospital, those who are readmitted or die soon after discharge may warrant special attention. Therefore, we prospectively followed 257 patients admitted nonelectively to an urban university hospital, with a complaint of shortness of breath or fatigue and evidence of congestive heart failure on admission chest radiograph, who were discharged alive. Through survey of patients and families, review of the hospital computer system, and a search of the National Death Index, we recorded death and hospital readmission. Within 60 days of discharge, 13 patients (5%) died and 82 (32%) died or were readmitted to the hospital. Using Cox proportional-hazards modeling, the multivariable correlates of readmission or death were single marital status (adjusted hazard ratio [HR] 2.1, 95% confidence interval [CI] 1.3 to 3.3), Charlson Comorbidity Index score (HR
1.3 per point to maximum 4 points, 95% CI 1.1 to 1.6), admission systolic blood pressure of °100 mm Hg (HR 2.8, 95% CI 1.6 to 5.0), and absence of new ST-T-wave changes on the initial electrocardiogram (HR 1.9, 95% CI 1.1 to 3.3). Self-reported patient compliance and clinical instability at discharge were not correlates. Almost all patients stratified by these factors had at least a 25% risk of readmission or death. Our independent correlates of readmission or death support the importance of both medical and social factors in the pathway to clinical decline. However, we could not reliably identify a truly low-risk group. Interventions to decrease early readmission or death among patients with heart failure should target both medical management and the adequacy of social support, and probably need to be applied to all admitted patients. Q 1997 by Excerpta Medica, Inc. (Am J Cardiol 1997;79:1640–1644)
atients with heart failure are frequently readmitted to the hospital, and have attracted the atP tention of clinicians and administrators seeking to
viewed the chart of every patient with a cardiac, pulmonary, or renal admission diagnosis. Patients were eligible for the study if they had both a complaint of shortness of breath or fatigue and had pulmonary edema, interstitial edema, or pulmonary vascular redistribution noted on initial chest x-ray by the emergency department physician, admitting physicians, or radiologist. Patients were excluded if they died during initial hospitalization or were transferred from another acute care hospital, and patients who were readmitted for heart failure during the study period were included only on the index admission. To eliminate patients who were volume overloaded solely secondary to end-stage renal disease, patients who had both an admission creatinine ú530.4 mmol/ L (6.0 mg/dl) and neither angina nor systolic blood pressure ¢180 mm Hg on presentation to the emergency department were excluded. The investigator or research assistant then obtained informed consent. If patients were cognitively impaired, too ill, or did not speak English, then the proxy was approached for consent and survey. Of the 436 patients presenting with radiographic evidence of heart failure, 84 (19%) were excluded because no proxy was available (12%), they died during hospitalization (5%), or they were discharged and could not be reached by telephone (2%). Of the remaining 352 patients (81%), 257 (73%) agreed to enroll and 95 (27%) refused. The study team surveyed participants about access to care, compliance, social support, primary insurance, and demographic data. Our compliance scale was composed of 7 items that asked the patient to rate his or her adherence to the treatment regimen over the preceding 4 weeks
1–6
decrease morbidity and costs.7 If health providers could identify patients at high risk for readmission, they could target programs to improve the quality of care and reduce costs.8 – 10 The few studies determining risk factors for early readmission among patients hospitalized with heart failure have focused on selected geriatric populations1,8,11 or have grouped heart failure patients with other cardiac subjects.12 Therefore, within a broad cohort of patients presenting to the hospital in heart failure, we sought to identify characteristics associated with a high risk for readmission or death within 60 days of discharge.
METHODS Study population: We prospectively enrolled patients admitted nonelectively with heart failure in 1993 and 1994 to the Brigham and Women’s Hospital, a 751-bed urban teaching hospital.13,14 With approval of the human research committee, each day a study investigator or a trained research assistant reFrom the Section for Clinical Epidemiology, Division of General Medicine and Primary Care, and the Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, and Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts. Dr. Chin is supported by Geriatric Academic Program Award 5-K12-AG-00488 from the National Institutes of Health/National Institute on Aging, Bethesda, Maryland. Manuscript received October 3, 1996; revised manuscript received and accepted February 26, 1997. Address for reprints: Marshall H. Chin, MD, MPH, University of Chicago Medical Center, Section of General Internal Medicine, 5841 South Maryland Avenue, MC 6098, Chicago, Illinois 60637.
1640
Q1997 by Excerpta Medica, Inc. All rights reserved.
/ 2w24 5047 Mp
1640
Wednesday May 14 03:30 PM
0002-9149/97/$17.00 PII S0002-9149(97)00214-2
EL–AJC (v. 79, no. 12 ’97)
5047
TABLE I Patient Characteristics (n Å 257)
TABLE I (Continued) No. (%)
No. (%) Demographic Sex Female Male Race White Non–white Age (yr) °70 ú70 Social and economic Yearly income °$7,500 ú$7,500 Education õHigh school ¢High school Single* Yes No Someone at home who can take care of you Yes No Access to care Regular doctor or place of care Yes No History Congestive heart failure Yes No Myocardial infarction Yes No Hypertension Yes No Lowest ejection fraction during hospitalization or before admission °0.20 0.21–0.35 0.36–0.49 ¢0.50 Ventricular tachycardia or ventricular fibrillation Yes No Diabetes mellitus Yes No Charlson Comorbidity Index score 0 1 2 3 ¢4 Low compliance† Yes No Admission vital signs Systolic blood pressure (mm Hg) °100 ú100 Respiratory rate (breaths/min) °30 ú30 Admission laboratories Serum sodium, mmol/L °135 ú135
Creatinine, mmol/L (mg/dl) ú265.2 (3.0) °265.2 (3.0) Admission electrocardiogram Normal sinus rhythm Yes No New ST-T-wave changes‡ Yes No Admission chest radiograph Cardiomegaly Yes No Sickness at discharge RAND symptoms/signs§ Yes No RAND laboratoriesx Yes No
130 (51) 127 181 (71) 71 160 (62) 97
55 (25) 164 74 (29) 177 127 (50) 125 178 (69) 78
24 (9) 233
176 (68) 81 85 (33) 172
141 (55) 116
130 (51) 127 43 (17) 214
* Divorced, widowed, not married, or not living as if married. † Bottom 20% of compliance scale based on work of DiMatteo and Hays.15 § ST-T-wave changes on initial electrocardiogram known to be neither old nor attributable to digoxin. § Presence of ¢1 of the RAND sickness at discharge criteria.17 x Presence of ¢1 of the RAND abnormal laboratory criteria.17
243 (95) 14
171 (67) 86
TABLE II Multivariable Correlates of Readmission or Death Within 60 Days (n Å 257)
96 (37) 161 164 (64) 93
Outcome
35 53 39 110 (46)
Death Diabetes mellitus Systolic blood pressure °100 mm Hg Nonsinus rhythm Readmission or death Single marital status† Charlson Comorbidity Index score‡ Systolic blood pressure °100 mm Hg No ST-T-wave ECG changes§
17 (7) 239
47 (21) 174
31 (12) 226 230 (89) 27
72 (28) 185
95% Confidence Interval
4.2 4.8 3.1
1.3–13.4 1.4–15.9 1.04–9.4
2.1 1.3 2.8 1.9
1.3–3.3 1.1–1.6 1.6–5.0 1.1–3.3
* Cox proportional-hazards models adjusting for other independent correlates. † Divorced, widowed, not married, or not living as if married. ‡ Per point to maximum of 4 points. § No ST-T-wave changes on initial electrocardiogram neither known to be old nor attributable to digoxin.
88 (34) 169 48 61 53 43 52 (20)
Adjusted Hazard Ratio*
on a 5-point scale ranging from ‘‘all of the time’’ to ‘‘none of the time.’’15 For example, ‘‘I was unable to do what was necessary to follow my doctor’s treatment plans for my heart.’’ Patients were given the option of self-administration or interviewer administration. The study team administered the survey if the patient would have difficulty on his or her own. Two months after discharge from the hospital, the patient received a second mailed questionnaire that inquired whether he or she had been readmitted to a hospital overnight since the original hospitalization. The questionnaire also asked how many days after discharge the readmission occurred, and whether the CONGESTIVE HEART FAILURE/HEART FAILURE READMISSION
/ 2w24 5047 Mp
1641
Wednesday May 14 03:30 PM
EL–AJC (v. 79, no. 12 ’97)
5047
1641
also performed a search of the National Death Index through December 31, 1994. One of the study investigators performed chart review to obtain the history, initial physical examination, vital signs, and test findings. We defined new ST-Twave changes as ST-T-wave changes on the initial electrocardiogram neither known to be old nor attributable to digoxin. Ejection fraction was the lowest one recorded during or before hospitalization. The Charlson Comorbidity Index was used as a summary measure of coexisting diseases.16 We recorded presence or absence of the RAND sickness at discharge and abnormal laboratory criteria.17 We tested the reproducibility of the chart abstraction in a separate analysis of 81 patients whom 1 researcher thought had a major complication or death.13 The senior physician determined that 77 of these patients (95%) had a major complication. Statistical analysis: We categorized continuous variables such as age, ejection fraction, systolic blood pressure, heart rate, respiratory rate, serum sodium, serum potassium, creatinine, and hematocrit at clinically relevant cutpoints. We considered patients to be low compliers if their scores were in the bottom 20% of the compliance scale. We compared characteristics of enrolled to nonenrolled patients using Fisher’s exact test for binary variables, the chi-square test for categorical variables, and the Cochran-MantelHaenszel test for categorical variables with trend. The outcome variables were death, and readmission or death within 60 days of discharge. We examined the effects of covariates on these outcomes using Kaplan-Meier curves and the log rank test in bivariate FIGURE 1. Kaplan-Meier curves for 60-day survival without hospital readmisanalyses. Significant (p °0.10) bivariate sion. Patients are stratified by marital status (A), Charlson Comorbidity Index correlates available during the hospitalscore (B), admission systolic blood pressure (C), and ST-T-wave changes on iniization were then entered into stepwise tial electrocardiogram (D). These factors were the independent correlates (p °0.05) of readmission or death within 60 days in Cox proportional-hazards Cox proportional-hazards regression modeling. New ST-T-wave electrocardiogram changes are ST-T-wave changes models with entry criterion of p °0.10 on initial electrocardiogram neither known to be old nor attributable to diand stay criterion of p °0.05 (2-tailed). goxin; Single Å divorced, widowed, not married, or not living as if married. We used the adjusted hazard ratios of the independent correlates of readmission or patient needed extra help after leaving the hospital death to form a risk score. We also tested interaction that he or she could not get from family or friends. terms. Over the next month the team sent nonrespondents a reminder letter and another copy of the survey, and RESULTS finally a professional interviewer attempted to adDescription of patient population: Compared with minister the questionnaire by telephone. For the 70 the nonenrolled patients, patients in the study were patients (27%) who did not respond to the questions more likely to be aged °70 years (62% vs 53%, p on hospital readmission, we searched the hospital’s Å 0.06), to have an initial respiratory rate °30 computer system. Patients without any readmissions breaths/min (89% vs 81%, p Å 0.02), and to be white in the computer were assumed to have no such visits (72% vs 60%, p Å 0.02). Among the 257 enrolled anywhere. We obtained dates of death from the fam- patients, the average age was 67 years and 51% were ily, friends, and the hospital computer database. We women. Two thirds had a prior history of congestive 1642
THE AMERICAN JOURNAL OF CARDIOLOGYT
/ 2w24 5047 Mp
1642
VOL. 79
Wednesday May 14 03:30 PM
JUNE 15, 1997
EL–AJC (v. 79, no. 12 ’97)
5047
creatinine ú265.2 mmol/L (3.0 mg/dl), and nonsinus rhythm. Bivariate correlates of readmission or death were female gender, single marital status, diabetes, Charlson Comorbidity Index score, initial systolic blood pressure °100 mm Hg, and absence of new ST-T-wave changes on the initial electrocardiogram. The multivariable correlates (p °0.05) of death were diabetes, initial systolic blood pressure °100 mm Hg, and nonsinus rhythm. The independent correlates of readmission or death were single marital status, Charlson Comorbidity Index score, admission systolic blood pressure °100 mm Hg, and absence of new STT-wave changes (Table II). Kaplan-Meier curves for 60-day survival without readmission are shown in Figure 1. The risk index for readmission or death is described in Table III. None of the 17 patients (7%) with no risk factors, or, at most, 1 Charlson comorbidity point, were readmitted or died within 60 days of discharge. However, the 240 patients (93%) who had either ¢2 Charlson comorbidity points, or any of the other risk factors, had a ¢25% chance of readmission or death within 60 days. Single patients were more likely (p °0.01) than nonsingle subjects to be women (87 [69%] vs 40 [32%] patients) and to report that they needed extra help after leaving the hospital, help that they could not obtain from family or friends (66 [62%] vs 38 [35%] patients). None of the interaction terms of single marital status with any of the independent medical correlates of readmission or death was significant in the multivariable model. FIGURE 1. (Continued)
heart failure, 37% had had a myocardial infarction, and 64% had a history of hypertension. One quarter had a yearly income °$7,500, and 30% did not have someone at home who could take care of them (Table I). Rates of death and readmission: Within 60 days of discharge, 13 of 257 patients (5%) died, 80 (31%) were readmitted to the hospital, and 82 (32%) died or were readmitted. When the outcomes were limited to those within 14 days of discharge, 33 patients (13%) either died (0.4%) or were readmitted (13%). Within 30 days of discharge, 58 patients (23%) either died (2%) or were readmitted (22%). Correlates of death and readmission: Bivariate correlates (p °0.10) of death within 60 days of discharge were diabetes, initial systolic blood pressure °100 mm Hg, initial serum sodium °135 mmol/L,
DISCUSSION Hospital readmission within 60 days of discharge was frequent in our cohort, occurring in approximately one third of these patients with heart failure. We were able to find independent risk factors for particularly high rates of readmission or death, but conversely we were unable to identify a truly low-risk group of patients. Among patients with congestive heart failure, readmissions are common events, occurring in 20% to 50% of patients within 14 days to 6 months after discharge.1,3,6,8,9,11 Correlates of readmission have included male sex,1 prior admissions,1,11 prior heart failure,11 diabetes,8 comorbidity,1 lower systolic blood pressure,8 and lower serum sodium.8 In addition, acute heart failure caused by myocardial infarction or severe hypertension,11 and decreased physician adherence with readiness-for-discharge criteria have been correlated with readmission.10 Increasing comorbidity, greater severity of illness as indicated by an admission systolic blood pressure
CONGESTIVE HEART FAILURE/HEART FAILURE READMISSION
/ 2w24 5047 Mp
1643
Wednesday May 14 03:30 PM
EL–AJC (v. 79, no. 12 ’97)
5047
1643
TABLE III Risk Index for Readmission or Death Within 60 Days
Risk Score* 0–1 2–5 6–7 ú7
Readmission or Death/ No. of Patients 0/17 34/144 30/71 18/25
% Readmitted or Dead Within 60 Days (95% confidence interval) 0 24 42 72
(0–20) (17–31) (31–55) (51–88)
* Risk score is the sum of points based on the adjusted hazard ratios of the independent correlates of readmission or death: Single marital status Å 2 points; Charlson Comorbidity Index score Å 1 point per Charlson point to maximum of 4; initial systolic blood pressure °100 mm Hg Å 3 points; No ST-T-wave changes on initial electrocardiogram neither known to be old nor attributable to digoxin Å 2 points.
°100 mm Hg, and absence of new ST-T-wave changes were the medical correlates of readmission or death in our study. Absence of new ST-T-wave changes may be a marker for patients who presented with decompensated chronic heart failure, as opposed to transient ischemic pulmonary edema. Single marital status may be a proxy for poor social support, which can worsen clinical condition through lack of needed help and services, or potentially through direct physiologic means.18,19 Ejection fraction, patient compliance, and sickness at discharge were not correlated with readmission or death. Possibly, the scales for compliance and sickness at discharge were not sufficiently sensitive or discriminating in our study population, or perhaps our independent correlates are of fundamental importance. Whereas our model may identify some patients at especially high risk for readmission or death, we urge caution in applying it when the goal is to target most patients who will have these outcomes. None of the 17 patients in our lowest risk category were readmitted or died within 60 days, but the number of patients in this subgroup were low. Other subgroups had a ¢25% chance of readmission or death. The high rates of hospital readmission for patients with heart failure in this study and others raise the question of whether it is possible to define a subgroup with a clinically meaningful low rate of readmission. Our study has limitations. First, it was performed at 1 hospital, thus limiting generalizability. Second, we relied primarily on patient self-report for the outcome measures because we wanted to record readmissions to any hospital. However, we were able to determine through our hospital’s administrative database whether a patient had a readmission or emergency department visit to our hospital. Whereas medical factors were the correlates of death, both medical and social factors were associated with readmission or death in patients with congestive heart failure: Therefore, a broad systems ap-
1644
THE AMERICAN JOURNAL OF CARDIOLOGYT
/ 2w24 5047 Mp
1644
proach that attempts to coordinate care in the inpatient, outpatient, and home settings is likely to be most effective in improving the health of patients with heart failure while decreasing costs.8,20,21 Acknowledgment: We would like to thank Cynthia Crespin, EdM, Jeffrey Geller, MD, Julie Newton, and Alexander Pedan, PhD, for their technical assistance, and Nicholas Christakis, MD, PhD, MPH, Peter Friedmann, MD, MPH, and Theodore Karrison, PhD, for their helpful reviews of the manuscript. 1. Krumholz HM, Parent EM, Tu N, Vaccarino V, Wang Y, Radford MJ, Hen-
nen J. Readmission after hospitalization for congestive heart failure among Medicare beneficiaries. Arch Intern Med 1997;157:99–104. 2. Wray NP, Ashton CM, Kuykendall DH, Petersen NJ, Souchek J, Hollingsworth JC. Selecting disease-outcome pairs for monitoring the quality of hospital care. Med Care 1995;33:75–89. 3. Gooding J, Jette AM. Hospital readmissions among the elderly. J Am Geriatr Soc 1985;33:595–601. 4. Welch HG, Larson EH. Patients requiring at least five admissions in 1 year: data from Washington state. Med Care 1991;29:578–582. 5. Burns R, Nichols LO. Factors predicting readmission of older general medicine patients. J Gen Intern Med 1991;6:389–393. 6. Rich MW, Freedland KE. Effect of DRGs on three-month readmission rate of geriatric patients with congestive heart failure. Am J Public Health 1988;78:680–682. 7. The National Managed Health Care Congress. Management Strategies in Heart Failure: Optimizing Practice, Outcome, and Cost. Boston, MA: The National Managed Health Care Congress, 1996. 8. Rich MW, Beckham V, Wittenberg C, Leven CL, Freedland KE, Carney RM. A multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure. N Engl J Med 1995;333:1190–1195. 9. Rich MW, Vinson JM, Sperry JC, Shah AS, Spinner LR, Chung MK, DavilaRoman V. Prevention of readmission in elderly patients with congestive heart failure: results of a prospective, randomized pilot study. J Gen Intern Med 1993;8:585–590. 10. 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;122:415–421. 11. Vinson JM, Rich MW, Sperry JC, Shah AS, McNamara T. Early readmission of elderly patients with congestive heart failure. J Am Geriatr Soc 1990;38:1290–1295. 12. Berkman B, Dumas S, Gastfriend J, Poplawski J, Southworthe M. Predicting hospital readmission of elderly cardiac patients. Health and Social Work 1987;221–227. 13. Chin MH, Goldman L. Correlates of major complications or death in patients admitted to the hospital with congestive heart failure. Arch Intern Med 1996;156:1814–1820. 14. Chin MH, Goldman L. Factors contributing to the hospitalization of patients with congestive heart failure. Am J Public Health 1997;87:645–650. 15. DiMatteo MR, Hays RD. Adherence to cancer regimens: implications for treating the older patient. Oncology 1992;6 (suppl):50–57. 16. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40:373–383. 17. Kosecoff J, Kahn KL, Rogers WH, Reinisch EJ, Sherwood MJ, Rubenstein LV, Draper D, Roth CP, Chew C, Brook RH. Prospective payment system and impairment at discharge: the ‘‘quicker-and-sicker’’ story revisited. JAMA 1990;264:1980–1983. 18. Berkman LF. Assessing the physical health effects of social networks and social support. Ann Rev Public Health 1984;5:413–432. 19. Berkman LF, Leo-Summers L, Horwitz RI. Emotional support and survival after myocardial infarction: a prospective, population-based study of the elderly. Ann Intern Med 1992;117:1003–1009. 20. West JA, Miller NH, Parker KM, Senneca D, Ghandour G, Clark M, Greenwald G, Heller RS, Fowler MB, DeBusk RF. A comprehensive management system for heart failure improves clinical outcomes and reduces medical resource utilization. Am J Cardiol 1997;79:58–63. 21. Kornowski R, Zeeli D, Averbuch M, Finkelstein A, Schwartz D, Moshkovitz M, Weinreb B, Hershkovitz R, Eyal D, Miller M, Levo Y, Pines A. Intensive home-care surveillance prevents hospitalization and improves morbidity rates among elderly patients with severe congestive heart failure. Am Heart J 1995;129:762–766.
VOL. 79
Wednesday May 14 03:30 PM
JUNE 15, 1997
EL–AJC (v. 79, no. 12 ’97)
5047
MD, MPH,
and Lee Goldman,
MD, MPH
Among patients with heart failure who survive an admission to the hospital, those who are readmitted or die soon after discharge may warrant special attention. Therefore, we prospectively followed 257 patients admitted nonelectively to an urban university hospital, with a complaint of shortness of breath or fatigue and evidence of congestive heart failure on admission chest radiograph, who were discharged alive. Through survey of patients and families, review of the hospital computer system, and a search of the National Death Index, we recorded death and hospital readmission. Within 60 days of discharge, 13 patients (5%) died and 82 (32%) died or were readmitted to the hospital. Using Cox proportional-hazards modeling, the multivariable correlates of readmission or death were single marital status (adjusted hazard ratio [HR] 2.1, 95% confidence interval [CI] 1.3 to 3.3), Charlson Comorbidity Index score (HR
1.3 per point to maximum 4 points, 95% CI 1.1 to 1.6), admission systolic blood pressure of °100 mm Hg (HR 2.8, 95% CI 1.6 to 5.0), and absence of new ST-T-wave changes on the initial electrocardiogram (HR 1.9, 95% CI 1.1 to 3.3). Self-reported patient compliance and clinical instability at discharge were not correlates. Almost all patients stratified by these factors had at least a 25% risk of readmission or death. Our independent correlates of readmission or death support the importance of both medical and social factors in the pathway to clinical decline. However, we could not reliably identify a truly low-risk group. Interventions to decrease early readmission or death among patients with heart failure should target both medical management and the adequacy of social support, and probably need to be applied to all admitted patients. Q 1997 by Excerpta Medica, Inc. (Am J Cardiol 1997;79:1640–1644)
atients with heart failure are frequently readmitted to the hospital, and have attracted the atP tention of clinicians and administrators seeking to
viewed the chart of every patient with a cardiac, pulmonary, or renal admission diagnosis. Patients were eligible for the study if they had both a complaint of shortness of breath or fatigue and had pulmonary edema, interstitial edema, or pulmonary vascular redistribution noted on initial chest x-ray by the emergency department physician, admitting physicians, or radiologist. Patients were excluded if they died during initial hospitalization or were transferred from another acute care hospital, and patients who were readmitted for heart failure during the study period were included only on the index admission. To eliminate patients who were volume overloaded solely secondary to end-stage renal disease, patients who had both an admission creatinine ú530.4 mmol/ L (6.0 mg/dl) and neither angina nor systolic blood pressure ¢180 mm Hg on presentation to the emergency department were excluded. The investigator or research assistant then obtained informed consent. If patients were cognitively impaired, too ill, or did not speak English, then the proxy was approached for consent and survey. Of the 436 patients presenting with radiographic evidence of heart failure, 84 (19%) were excluded because no proxy was available (12%), they died during hospitalization (5%), or they were discharged and could not be reached by telephone (2%). Of the remaining 352 patients (81%), 257 (73%) agreed to enroll and 95 (27%) refused. The study team surveyed participants about access to care, compliance, social support, primary insurance, and demographic data. Our compliance scale was composed of 7 items that asked the patient to rate his or her adherence to the treatment regimen over the preceding 4 weeks
1–6
decrease morbidity and costs.7 If health providers could identify patients at high risk for readmission, they could target programs to improve the quality of care and reduce costs.8 – 10 The few studies determining risk factors for early readmission among patients hospitalized with heart failure have focused on selected geriatric populations1,8,11 or have grouped heart failure patients with other cardiac subjects.12 Therefore, within a broad cohort of patients presenting to the hospital in heart failure, we sought to identify characteristics associated with a high risk for readmission or death within 60 days of discharge.
METHODS Study population: We prospectively enrolled patients admitted nonelectively with heart failure in 1993 and 1994 to the Brigham and Women’s Hospital, a 751-bed urban teaching hospital.13,14 With approval of the human research committee, each day a study investigator or a trained research assistant reFrom the Section for Clinical Epidemiology, Division of General Medicine and Primary Care, and the Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, and Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts. Dr. Chin is supported by Geriatric Academic Program Award 5-K12-AG-00488 from the National Institutes of Health/National Institute on Aging, Bethesda, Maryland. Manuscript received October 3, 1996; revised manuscript received and accepted February 26, 1997. Address for reprints: Marshall H. Chin, MD, MPH, University of Chicago Medical Center, Section of General Internal Medicine, 5841 South Maryland Avenue, MC 6098, Chicago, Illinois 60637.
1640
Q1997 by Excerpta Medica, Inc. All rights reserved.
/ 2w24 5047 Mp
1640
Wednesday May 14 03:30 PM
0002-9149/97/$17.00 PII S0002-9149(97)00214-2
EL–AJC (v. 79, no. 12 ’97)
5047
TABLE I Patient Characteristics (n Å 257)
TABLE I (Continued) No. (%)
No. (%) Demographic Sex Female Male Race White Non–white Age (yr) °70 ú70 Social and economic Yearly income °$7,500 ú$7,500 Education õHigh school ¢High school Single* Yes No Someone at home who can take care of you Yes No Access to care Regular doctor or place of care Yes No History Congestive heart failure Yes No Myocardial infarction Yes No Hypertension Yes No Lowest ejection fraction during hospitalization or before admission °0.20 0.21–0.35 0.36–0.49 ¢0.50 Ventricular tachycardia or ventricular fibrillation Yes No Diabetes mellitus Yes No Charlson Comorbidity Index score 0 1 2 3 ¢4 Low compliance† Yes No Admission vital signs Systolic blood pressure (mm Hg) °100 ú100 Respiratory rate (breaths/min) °30 ú30 Admission laboratories Serum sodium, mmol/L °135 ú135
Creatinine, mmol/L (mg/dl) ú265.2 (3.0) °265.2 (3.0) Admission electrocardiogram Normal sinus rhythm Yes No New ST-T-wave changes‡ Yes No Admission chest radiograph Cardiomegaly Yes No Sickness at discharge RAND symptoms/signs§ Yes No RAND laboratoriesx Yes No
130 (51) 127 181 (71) 71 160 (62) 97
55 (25) 164 74 (29) 177 127 (50) 125 178 (69) 78
24 (9) 233
176 (68) 81 85 (33) 172
141 (55) 116
130 (51) 127 43 (17) 214
* Divorced, widowed, not married, or not living as if married. † Bottom 20% of compliance scale based on work of DiMatteo and Hays.15 § ST-T-wave changes on initial electrocardiogram known to be neither old nor attributable to digoxin. § Presence of ¢1 of the RAND sickness at discharge criteria.17 x Presence of ¢1 of the RAND abnormal laboratory criteria.17
243 (95) 14
171 (67) 86
TABLE II Multivariable Correlates of Readmission or Death Within 60 Days (n Å 257)
96 (37) 161 164 (64) 93
Outcome
35 53 39 110 (46)
Death Diabetes mellitus Systolic blood pressure °100 mm Hg Nonsinus rhythm Readmission or death Single marital status† Charlson Comorbidity Index score‡ Systolic blood pressure °100 mm Hg No ST-T-wave ECG changes§
17 (7) 239
47 (21) 174
31 (12) 226 230 (89) 27
72 (28) 185
95% Confidence Interval
4.2 4.8 3.1
1.3–13.4 1.4–15.9 1.04–9.4
2.1 1.3 2.8 1.9
1.3–3.3 1.1–1.6 1.6–5.0 1.1–3.3
* Cox proportional-hazards models adjusting for other independent correlates. † Divorced, widowed, not married, or not living as if married. ‡ Per point to maximum of 4 points. § No ST-T-wave changes on initial electrocardiogram neither known to be old nor attributable to digoxin.
88 (34) 169 48 61 53 43 52 (20)
Adjusted Hazard Ratio*
on a 5-point scale ranging from ‘‘all of the time’’ to ‘‘none of the time.’’15 For example, ‘‘I was unable to do what was necessary to follow my doctor’s treatment plans for my heart.’’ Patients were given the option of self-administration or interviewer administration. The study team administered the survey if the patient would have difficulty on his or her own. Two months after discharge from the hospital, the patient received a second mailed questionnaire that inquired whether he or she had been readmitted to a hospital overnight since the original hospitalization. The questionnaire also asked how many days after discharge the readmission occurred, and whether the CONGESTIVE HEART FAILURE/HEART FAILURE READMISSION
/ 2w24 5047 Mp
1641
Wednesday May 14 03:30 PM
EL–AJC (v. 79, no. 12 ’97)
5047
1641
also performed a search of the National Death Index through December 31, 1994. One of the study investigators performed chart review to obtain the history, initial physical examination, vital signs, and test findings. We defined new ST-Twave changes as ST-T-wave changes on the initial electrocardiogram neither known to be old nor attributable to digoxin. Ejection fraction was the lowest one recorded during or before hospitalization. The Charlson Comorbidity Index was used as a summary measure of coexisting diseases.16 We recorded presence or absence of the RAND sickness at discharge and abnormal laboratory criteria.17 We tested the reproducibility of the chart abstraction in a separate analysis of 81 patients whom 1 researcher thought had a major complication or death.13 The senior physician determined that 77 of these patients (95%) had a major complication. Statistical analysis: We categorized continuous variables such as age, ejection fraction, systolic blood pressure, heart rate, respiratory rate, serum sodium, serum potassium, creatinine, and hematocrit at clinically relevant cutpoints. We considered patients to be low compliers if their scores were in the bottom 20% of the compliance scale. We compared characteristics of enrolled to nonenrolled patients using Fisher’s exact test for binary variables, the chi-square test for categorical variables, and the Cochran-MantelHaenszel test for categorical variables with trend. The outcome variables were death, and readmission or death within 60 days of discharge. We examined the effects of covariates on these outcomes using Kaplan-Meier curves and the log rank test in bivariate FIGURE 1. Kaplan-Meier curves for 60-day survival without hospital readmisanalyses. Significant (p °0.10) bivariate sion. Patients are stratified by marital status (A), Charlson Comorbidity Index correlates available during the hospitalscore (B), admission systolic blood pressure (C), and ST-T-wave changes on iniization were then entered into stepwise tial electrocardiogram (D). These factors were the independent correlates (p °0.05) of readmission or death within 60 days in Cox proportional-hazards Cox proportional-hazards regression modeling. New ST-T-wave electrocardiogram changes are ST-T-wave changes models with entry criterion of p °0.10 on initial electrocardiogram neither known to be old nor attributable to diand stay criterion of p °0.05 (2-tailed). goxin; Single Å divorced, widowed, not married, or not living as if married. We used the adjusted hazard ratios of the independent correlates of readmission or patient needed extra help after leaving the hospital death to form a risk score. We also tested interaction that he or she could not get from family or friends. terms. Over the next month the team sent nonrespondents a reminder letter and another copy of the survey, and RESULTS finally a professional interviewer attempted to adDescription of patient population: Compared with minister the questionnaire by telephone. For the 70 the nonenrolled patients, patients in the study were patients (27%) who did not respond to the questions more likely to be aged °70 years (62% vs 53%, p on hospital readmission, we searched the hospital’s Å 0.06), to have an initial respiratory rate °30 computer system. Patients without any readmissions breaths/min (89% vs 81%, p Å 0.02), and to be white in the computer were assumed to have no such visits (72% vs 60%, p Å 0.02). Among the 257 enrolled anywhere. We obtained dates of death from the fam- patients, the average age was 67 years and 51% were ily, friends, and the hospital computer database. We women. Two thirds had a prior history of congestive 1642
THE AMERICAN JOURNAL OF CARDIOLOGYT
/ 2w24 5047 Mp
1642
VOL. 79
Wednesday May 14 03:30 PM
JUNE 15, 1997
EL–AJC (v. 79, no. 12 ’97)
5047
creatinine ú265.2 mmol/L (3.0 mg/dl), and nonsinus rhythm. Bivariate correlates of readmission or death were female gender, single marital status, diabetes, Charlson Comorbidity Index score, initial systolic blood pressure °100 mm Hg, and absence of new ST-T-wave changes on the initial electrocardiogram. The multivariable correlates (p °0.05) of death were diabetes, initial systolic blood pressure °100 mm Hg, and nonsinus rhythm. The independent correlates of readmission or death were single marital status, Charlson Comorbidity Index score, admission systolic blood pressure °100 mm Hg, and absence of new STT-wave changes (Table II). Kaplan-Meier curves for 60-day survival without readmission are shown in Figure 1. The risk index for readmission or death is described in Table III. None of the 17 patients (7%) with no risk factors, or, at most, 1 Charlson comorbidity point, were readmitted or died within 60 days of discharge. However, the 240 patients (93%) who had either ¢2 Charlson comorbidity points, or any of the other risk factors, had a ¢25% chance of readmission or death within 60 days. Single patients were more likely (p °0.01) than nonsingle subjects to be women (87 [69%] vs 40 [32%] patients) and to report that they needed extra help after leaving the hospital, help that they could not obtain from family or friends (66 [62%] vs 38 [35%] patients). None of the interaction terms of single marital status with any of the independent medical correlates of readmission or death was significant in the multivariable model. FIGURE 1. (Continued)
heart failure, 37% had had a myocardial infarction, and 64% had a history of hypertension. One quarter had a yearly income °$7,500, and 30% did not have someone at home who could take care of them (Table I). Rates of death and readmission: Within 60 days of discharge, 13 of 257 patients (5%) died, 80 (31%) were readmitted to the hospital, and 82 (32%) died or were readmitted. When the outcomes were limited to those within 14 days of discharge, 33 patients (13%) either died (0.4%) or were readmitted (13%). Within 30 days of discharge, 58 patients (23%) either died (2%) or were readmitted (22%). Correlates of death and readmission: Bivariate correlates (p °0.10) of death within 60 days of discharge were diabetes, initial systolic blood pressure °100 mm Hg, initial serum sodium °135 mmol/L,
DISCUSSION Hospital readmission within 60 days of discharge was frequent in our cohort, occurring in approximately one third of these patients with heart failure. We were able to find independent risk factors for particularly high rates of readmission or death, but conversely we were unable to identify a truly low-risk group of patients. Among patients with congestive heart failure, readmissions are common events, occurring in 20% to 50% of patients within 14 days to 6 months after discharge.1,3,6,8,9,11 Correlates of readmission have included male sex,1 prior admissions,1,11 prior heart failure,11 diabetes,8 comorbidity,1 lower systolic blood pressure,8 and lower serum sodium.8 In addition, acute heart failure caused by myocardial infarction or severe hypertension,11 and decreased physician adherence with readiness-for-discharge criteria have been correlated with readmission.10 Increasing comorbidity, greater severity of illness as indicated by an admission systolic blood pressure
CONGESTIVE HEART FAILURE/HEART FAILURE READMISSION
/ 2w24 5047 Mp
1643
Wednesday May 14 03:30 PM
EL–AJC (v. 79, no. 12 ’97)
5047
1643
TABLE III Risk Index for Readmission or Death Within 60 Days
Risk Score* 0–1 2–5 6–7 ú7
Readmission or Death/ No. of Patients 0/17 34/144 30/71 18/25
% Readmitted or Dead Within 60 Days (95% confidence interval) 0 24 42 72
(0–20) (17–31) (31–55) (51–88)
* Risk score is the sum of points based on the adjusted hazard ratios of the independent correlates of readmission or death: Single marital status Å 2 points; Charlson Comorbidity Index score Å 1 point per Charlson point to maximum of 4; initial systolic blood pressure °100 mm Hg Å 3 points; No ST-T-wave changes on initial electrocardiogram neither known to be old nor attributable to digoxin Å 2 points.
°100 mm Hg, and absence of new ST-T-wave changes were the medical correlates of readmission or death in our study. Absence of new ST-T-wave changes may be a marker for patients who presented with decompensated chronic heart failure, as opposed to transient ischemic pulmonary edema. Single marital status may be a proxy for poor social support, which can worsen clinical condition through lack of needed help and services, or potentially through direct physiologic means.18,19 Ejection fraction, patient compliance, and sickness at discharge were not correlated with readmission or death. Possibly, the scales for compliance and sickness at discharge were not sufficiently sensitive or discriminating in our study population, or perhaps our independent correlates are of fundamental importance. Whereas our model may identify some patients at especially high risk for readmission or death, we urge caution in applying it when the goal is to target most patients who will have these outcomes. None of the 17 patients in our lowest risk category were readmitted or died within 60 days, but the number of patients in this subgroup were low. Other subgroups had a ¢25% chance of readmission or death. The high rates of hospital readmission for patients with heart failure in this study and others raise the question of whether it is possible to define a subgroup with a clinically meaningful low rate of readmission. Our study has limitations. First, it was performed at 1 hospital, thus limiting generalizability. Second, we relied primarily on patient self-report for the outcome measures because we wanted to record readmissions to any hospital. However, we were able to determine through our hospital’s administrative database whether a patient had a readmission or emergency department visit to our hospital. Whereas medical factors were the correlates of death, both medical and social factors were associated with readmission or death in patients with congestive heart failure: Therefore, a broad systems ap-
1644
THE AMERICAN JOURNAL OF CARDIOLOGYT
/ 2w24 5047 Mp
1644
proach that attempts to coordinate care in the inpatient, outpatient, and home settings is likely to be most effective in improving the health of patients with heart failure while decreasing costs.8,20,21 Acknowledgment: We would like to thank Cynthia Crespin, EdM, Jeffrey Geller, MD, Julie Newton, and Alexander Pedan, PhD, for their technical assistance, and Nicholas Christakis, MD, PhD, MPH, Peter Friedmann, MD, MPH, and Theodore Karrison, PhD, for their helpful reviews of the manuscript. 1. Krumholz HM, Parent EM, Tu N, Vaccarino V, Wang Y, Radford MJ, Hen-
nen J. Readmission after hospitalization for congestive heart failure among Medicare beneficiaries. Arch Intern Med 1997;157:99–104. 2. Wray NP, Ashton CM, Kuykendall DH, Petersen NJ, Souchek J, Hollingsworth JC. Selecting disease-outcome pairs for monitoring the quality of hospital care. Med Care 1995;33:75–89. 3. Gooding J, Jette AM. Hospital readmissions among the elderly. J Am Geriatr Soc 1985;33:595–601. 4. Welch HG, Larson EH. Patients requiring at least five admissions in 1 year: data from Washington state. Med Care 1991;29:578–582. 5. Burns R, Nichols LO. Factors predicting readmission of older general medicine patients. J Gen Intern Med 1991;6:389–393. 6. Rich MW, Freedland KE. Effect of DRGs on three-month readmission rate of geriatric patients with congestive heart failure. Am J Public Health 1988;78:680–682. 7. The National Managed Health Care Congress. Management Strategies in Heart Failure: Optimizing Practice, Outcome, and Cost. Boston, MA: The National Managed Health Care Congress, 1996. 8. Rich MW, Beckham V, Wittenberg C, Leven CL, Freedland KE, Carney RM. A multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure. N Engl J Med 1995;333:1190–1195. 9. Rich MW, Vinson JM, Sperry JC, Shah AS, Spinner LR, Chung MK, DavilaRoman V. Prevention of readmission in elderly patients with congestive heart failure: results of a prospective, randomized pilot study. J Gen Intern Med 1993;8:585–590. 10. 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;122:415–421. 11. Vinson JM, Rich MW, Sperry JC, Shah AS, McNamara T. Early readmission of elderly patients with congestive heart failure. J Am Geriatr Soc 1990;38:1290–1295. 12. Berkman B, Dumas S, Gastfriend J, Poplawski J, Southworthe M. Predicting hospital readmission of elderly cardiac patients. Health and Social Work 1987;221–227. 13. Chin MH, Goldman L. Correlates of major complications or death in patients admitted to the hospital with congestive heart failure. Arch Intern Med 1996;156:1814–1820. 14. Chin MH, Goldman L. Factors contributing to the hospitalization of patients with congestive heart failure. Am J Public Health 1997;87:645–650. 15. DiMatteo MR, Hays RD. Adherence to cancer regimens: implications for treating the older patient. Oncology 1992;6 (suppl):50–57. 16. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40:373–383. 17. Kosecoff J, Kahn KL, Rogers WH, Reinisch EJ, Sherwood MJ, Rubenstein LV, Draper D, Roth CP, Chew C, Brook RH. Prospective payment system and impairment at discharge: the ‘‘quicker-and-sicker’’ story revisited. JAMA 1990;264:1980–1983. 18. Berkman LF. Assessing the physical health effects of social networks and social support. Ann Rev Public Health 1984;5:413–432. 19. Berkman LF, Leo-Summers L, Horwitz RI. Emotional support and survival after myocardial infarction: a prospective, population-based study of the elderly. Ann Intern Med 1992;117:1003–1009. 20. West JA, Miller NH, Parker KM, Senneca D, Ghandour G, Clark M, Greenwald G, Heller RS, Fowler MB, DeBusk RF. A comprehensive management system for heart failure improves clinical outcomes and reduces medical resource utilization. Am J Cardiol 1997;79:58–63. 21. Kornowski R, Zeeli D, Averbuch M, Finkelstein A, Schwartz D, Moshkovitz M, Weinreb B, Hershkovitz R, Eyal D, Miller M, Levo Y, Pines A. Intensive home-care surveillance prevents hospitalization and improves morbidity rates among elderly patients with severe congestive heart failure. Am Heart J 1995;129:762–766.
VOL. 79
Wednesday May 14 03:30 PM
JUNE 15, 1997
EL–AJC (v. 79, no. 12 ’97)
5047