From the Health Policy Research Program, JDC Brookdale Institute, Jerusalem, Israel.
We reviewed the recent literature on hospital readmissions and found that most of them are believed to be caused by patient frailty and progression of chronic disease. However, from 9% to 48% of all readmissions have been judged to be preventable because they were associated with indicators of substandard care during the index hospitalization, such as poor resolution of the main problem, unstable therapy at discharge, and inadequate postdischarge care. Furthermore, randomized prospective trials have shown that 12% to 75% of all readmissions can be prevented by patient education, predischarge assessment, and domiciliary aftercare. We conclude that most readmissions seem to be caused by unmodifiable causes, and that, pending an agreed-on method to adjust for confounders, global readmission rates are not a useful indicator of quality of care. However, high readmission rates of patients with defined conditions, such as diabetes and bronchial asthma, may identify quality-of-care problems. A focus on the specific needs of such patients may lead to the creation of more responsive health care systems for the chronically ill.
Hospitalizations account for about half of all health care expenses, and it has been estimated that 13% of the inpatients in the United States use more than half of all hospital resources through repeated admissions.1 During past decades, hospital readmissions have been the subject of retrospective surveys and prospective trials with a view to their prevention. Our objective is to review these studies and focus on the frequency of readmissions, their causes and validity as a measure of quality of care, and the attempts for their prevention.
Hospital readmissions cluster shortly after discharge and decline thereafter. About one third of them occur within a month of discharge, half of them within 90 days, and 80% within a year.1- 3 The term readmission has been defined variously as a repeated hospitalization within 1,4- 6 2,7,8 4,9 or 1210 months of discharge. Most preventable readmissions have been reported to occur early, within 1 month of discharge, and it has been suggested to adopt this time interval in comparative studies.9,11 Still, early readmissions may also result from a nonpreventable progression of the disease or from a different diagnosis, whereas even late readmissions of diabetic or ashmatic patients may be preventable by appropriate ambulatory care. Therefore, although readily available from hospitals or health maintenance organizations,1,6,8,12 data on global (all-cause) readmissions have a limited value in the assessment of quality of care. Indeed, after analyzing the data of the entire Medicare population, Gornick et al13 concluded that
"the development of the re-hospitalization data was the most complex part of the project . . . [R]e-hospitalizations after medical stays often indicated . . . the progression of disease, rather than discrete outcomes of care. . . . Therefore, [their] analysis . . . would require additional information not available from the Medicare data system."
The additional information needed to analyze readmissions can be acquired by a prospective follow-up of patients discharged from hospitals3,14,15 or retrospective chart audits of patients admitted to hospitals.7,10,16 Although the yield of such studies in terms of detected preventable readmissions may be relatively low,2 they may identify prototypic errors and suggest ways to better practice. The cornerstone of the continuous quality improvement theory is that system adjustments yield high reward.17
We searched the literature (using "patient readmission" and "quality assurance, health care" as key terms) for articles published from January 1, 1991, through December 31, 1998, and the reference sections of the identified articles were further searched for additional sources on unscheduled readmissions. We excluded articles dealing with readmission to psychiatric and pediatric wards and restricted the survey to internal medicine and surgical departments, with a focus on the frequency of preventable readmissions, efficacy of interventions aiming at their prevention, and directions of future research.
The causes of readmissions may be inferred from differences in their rates among various patient populations. Of all discharges from general acute care hospitals, the proportion of readmissions has been reported to be 5% to 14% after 1 month2,5,18,19 and 32% to 49% after 1 year.2,20 Somewhat higher rates have been reported for geriatric patients, ie, 12% to 16% after 1 month,6,16,21- 23 60% to 64% after 6 months,24,25 and 34% to 67% after 1 year.10,16,23,26 The highest readmission rates have been observed in "high-risk" or severely ill geriatric patients, mostly with heart failure and chronic obstructive pulmonary disease,27 ie, 35% after 1 month,28 26% to 44% after 4 to 6 months,29,30 and 70% after 1 year31 (Table 1).
Readmission rates have also varied according to demographic, social, and disease-related characteristics.32 A meta-analysis of 44 studies published before 1990 revealed that age, length of stay during the index hospitalization, and previous use of hospital resources were among the main independent predictors of readmissions.33 Other authors have identified as predictors of readmissions male sex,8,34 white race,8 supplemental Medicaid coverage,8 low socioeconomic status,35- 37 single marital status,38 psychiatric comorbidity,39 behavioral problems,35 diagnosis34,40- 44(Table 2), the severity of the illness,45 nutritional status,29,46,47 comorbidity,34 and length of stay during the index hospitalization.8,34,48
These findings indicate that patient-specific factors predict readmissions. Patient-specific factors could be independent and nonmodifiable risk indicators of readmissions; on the other hand, they could be markers of other, modifiable factors. For example, readmissions could be related to advancing age because of noncompliance with medication regimen or inappropriate home care. In other words, an apparently patient-specific factor may reflect a failure to provide adequate health care.
The existence of modifiable factors of readmissions is suggested by their geographic variability. Regardless of the initial cause of the admission and its severity, Medicare beneficiaries had consistently higher rates of readmission in Boston, Mass, than did Medicare beneficiaries in New Haven, Conn,49 possibly because of variability in practice habits due to hospital-bed availability. Modifiable factors of readmissions are suggested also by their variability according to discharge destination. A study of a national sample of patients with chronic obstructive pulmonary disease or dementia revealed that after adjusting for severity and clinical and demographic characteristics, patients discharged to nursing homes were less likely to be readmitted within 30 days after discharge than those discharged to personal homes.44 Finally, some studies have found an association between readmission rates and inappropriate care during the index hospitalization. A case-control study revealed that 5 criteria of inpatient care (resolution of main problem, adequacy of the postdischarge destination, stability of doses of therapy, and appropriate timing of the first follow-up visit) predicted readmissions within 30 days.50 Another case-control study found that a set of disease-specific, explicit criteria of appropriateness of care51,52 predicted readmissions.53 It has been suggested that 1 of 7 readmissions in patients with diabetes, 1 of 5 readmissions in patients with heart failure, and 1 of 12 readmissions in patients with obstructive lung disease were attributable to substandard care.53 Absence of documentation of discharge planning, increased temperature, intravenous fluids on the day of discharge, or unaddressed abnormal test results at discharge were related to an increased subsequent mortality.6 A meta-analysis of 29 studies published from 1975 through 1993 confirmed that low-quality inpatient care during the index hospitalization increased the risk of subsequent readmissions.4 At least some readmissions, therefore, are associated with modifiable factors.
Of all readmissions, the proportion of those judged on retrospective chart audits to be preventable has varied from 9% to 50%9,10,21,22,54- 56(Table 3), and the meta-analysis by Ashton et al4 revealed that as many as 55% of the readmissions could be due to poor-quality and theoretically modifiable care during the index hospitalization. In contrast, the percentage of preventable hospital admissions in general (whether first or recurrent), has been estimated to be 9%,57 20%,58 21%,59 or 23%.60
The variability in estimated proportions of preventable readmissions could be caused by the limited reliability of the identification of a readmission as preventable. A study of 713 discharges revealed 109 (15.3%) unscheduled readmissions within 28 days. A review by 2 evaluators identified 34 (31%) of these readmissions as preventable. A second audit by another team of physicians identified only 16 (15%) of the same readmissions as preventable.21 Another study similarly found that the agreement between evaluators of readmissions was moderate (κ=0.43).56 A second possible explanation of the variability in the proportion of readmissions judged to be preventable is differences in the quality of care provided in various hospitals. Thus, an audit of 811 readmissions judged 277 (34%) of them as preventable. Hospital system factors accounted for 37%; clinician factors, 38%; and patient factors, 21%. Nine hospitals differed markedly in their profile of reasons for preventable readmissions,56 and hospitals with higher-than-expected readmission rates of patients with a given diagnostic entity were also more likely to have a high readmission rate in the next year.61 It may therefore be concluded that from 9% to 55% of the readmissions are due to inappropriate care during or after the index admission.
Readmission rates have been reported to decline after the implementation of predischarge reviews and improved follow-up after discharge.18,62,63 However, nonexperimental, before-after study designs are subject to confounding and to regression toward the mean. Confounding refers to changes beyond the planned intervention that occurred over time and that in and of themselves may have reduced readmission rates. Regression to the mean is the tendency of above-average rates to fall toward average over time. Since programs aiming to reduce readmission rates are likely to be implemented in institutions with high readmission rates, their favorable results may reflect a decline that would have occurred on subsequent determinations even without any specific interventions.
Confounding and regression toward the mean can be averted by randomized controlled studies. Soeken et al33 reviewed 12 controlled studies of the efficacy of planned interventions in reducing readmissions, published from 1980 through 1990. All but 4 of them found lower readmission rates in the intervention group. Table 4 and Table 5 summarize 19 studies published since 1991,5,16,19,20,24- 26,28,30,31,40- 42,64- 69 all but 3 of them randomized. A 12% to 75% reduction in readmissions or in emergency visits was found in 14 of these 19 studies. Another study compared 2 control and 2 intervention hospitals to evaluate a utilization management program. The results indicated that both intervention hospitals and 1 control hospital had lower 30-day readmission rates after the intervention than before.5 The remaining 4 studies detected either no differences in readmissions between control and intervention patients19,20,64 or even higher readmission rates in the intervention group.30
The effect on mortality of interventions aiming to reduce readmission rates was reported in 7 studies.16,20,24,25,42,64,68 Three of them found that the intervention was associated with a 25%,42 70%,25 and 80%68 decline in mortality. Costs and length of hospital stay were reported to have been reduced in 5 of a total of 7 studies. None, however, studied cost-effectiveness. Most intervention studies reviewed by Eggert and Friedman70 were not cost-effective. Safran and Phillips71 used decision analysis to examine the cost-effectiveness of interventions aiming to prevent readmissions. The authors considered the following 3 strategies: no intervention, intervention for all patients, and intervention for patients at high risk for readmission. They found that an intervention that costs $250 per patient would reduce overall costs for high-risk patients if its success rate exceeded 9% and for all patients if its success rate exceeded 17%. Predischarge reviews and improved postdischarge care, therefore, may prevent readmissions, although their cost-effectiveness is uncertain.
The findings concerning the effect of interventions indicate that improved hospital and postdischarge care are associated with fewer readmissions. Still, there is evidence that global readmission rates have a limited value as indicators of quality of care. For example, about half of the studies reviewed by Ashton et al4 failed to uncover any relationship between quality of care and readmissions. In all clinical conditions studied by Thomas,43 readmission rates of patients who received poor-quality care were similar to those of patients whose care was judged acceptable. Similarly, Roe et al72 assessed risk-adjusted outcomes after renal failure, gastrointestinal tract hemorrhage, stroke, myocardial infarction, and heart failure and concluded that length of stay, death, and unplanned readmission were predicted mainly by age, severity, and comorbidity. Hayward et al73 reviewed 675 general medicine hospital admissions and found that care of 30% of the patients who died in hospital, but only 10% of those discharged alive, could be rated as substandard; on the other hand, patients who had subsequent early readmissions did not have poorer quality of care ratings than those without early readmissions. DesHarnais et al74 developed and validated 3 risk-adjusted indices of hospital quality: mortality, readmissions, and complications. They ranked 300 hospitals on each index, and found no relationship between a hospital's ranking on any one of these indices and its ranking on the other two. Finally, a Monte Carlo simulation indicated that readmission rates were a poor measure of quality.75
As with all diagnostic tests, measures of quality of health care may identify incorrectly some medical interventions as inappropriate when they are actually appropriate, and vice versa.76 Therefore, it appears that this is true also for readmissions, and punitive measures based on high readmission rates may penalize hospitals without ample reason.77 However, the uncertain validity of global readmission rates as an indicator of quality of care does not preclude efforts for their reduction. Hospital readmissions raise concern among health care providers, and therefore efforts for their reduction are likely to be endorsed by clinicians and administrators.
The survey of controlled studies of the efficacy of planned interventions in reducing readmissions demonstrated a significant improvement in outcome of care in all 6 studies, which focused on patients with defined disorders (Table 5), but only in 6 of 10 studies of patients with various chronic disorders (Table 4). This finding suggests that a focus on patients with defined disorders may yield a higher reward in terms of improved patient care than attempts to reduce readmissions in the general inpatient population.
We believe that future research will focus on readmissions of inpatients with specific conditions, such as labor and child birth,78 coronary artery bypass grafting,79 uncontrolled pain,80 traumatic spinal cord injury,81 or acute coronary disease.82 A scrutiny of the causes of these readmissions may lead to an identification of unmet clinical, educational, and psychosocial needs. Once defined, research will focus on possible ways to meet these needs. There is already evidence of the benefit of interventions combining clinical expertise with coordinated care of patients with specific chronic disorders, such as bronchial asthma,65,67 heart failure,40- 42,68 and diabetes,69 for which there are processes of care known to affect outcomes. The specific features of these interventions are patient education, close follow-up, home monitoring, medication adjustment, and regular communication with clinical experts.83 Approaches that ensure closer adherence to evidence-based guidelines and meet patient self-management needs may improve clinical outcomes and reduce health care expenditures.
Accepted for publication July 12, 1999.
Corresponding author: Jochanan Benbassat, MD, JDC Brookdale Institute, PO Box 13087, Jerusalem 91130, Israel (e-mail: email@example.com).
Thank you for submitting a comment on this article. It will be reviewed by JAMA Internal Medicine editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 100
Customize your page view by dragging & repositioning the boxes below.
Users' Guides to the Medical Literature
Chapter e23. How to Use an Article About Quality Improvement
All results at
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.