Unplanned hospital readmissions profoundly impact both patients and the healthcare system via increased costs, patient dissatisfaction, and increased mortality.1-3 Readmission to hospital within 30 days occurs in 13.9% and 9.1% of discharged patients in the United States and Canada, respectively, representing billions of dollars in health care expenditure.4,5 Regional variability of hospital readmission rates suggests that at least some readmissions are avoidable, leading to their tracking and use as a health care quality metric.6-8 This has fueled significant research and implementation of interventions to reduce readmissions at all levels, from clinicians to policymakers. A notable example is the Hospital Readmissions Reduction Program in the United States, where hospitals are financially penalized for above-expected 30-day readmission rates.
Despite a growing body of robust research, the evidence remains inconclusive as to whether any single- or multifaceted approach reliably reduces 30-day hospital readmission rates.9-12 The uncertain effect of published interventions likely reflects complexity and variability in the causes underlying readmissions. More importantly, recent literature suggests that only about 25% of hospital readmission are preventable.13,14 Therefore, to improve the success of a readmission reduction intervention, it may be advantageous to identify and target preventable readmissions.15 Tracking and analyzing preventable hospital readmissions would reflect quality of care that can be influenced by the health care system. In previous studies, preventable readmissions have largely been identified through physician assessment of data from manual chart reviews, or by use of automated software.13,16 Both methods have multiple limitations-computerized methods are less accurate and overestimate the number of preventable readmissions, while physician chart review data offer only a clinical perspective on preventability and risk bias.17,18
The purpose of our study was to identify preventable hospital readmissions and their common causes to inform future readmission reduction interventions. We performed a retrospective cohort analysis of 28-day hospital readmissions among patients discharged from the general internal medicine (GIM) teaching service. In contrast to most previous studies, a multidisciplinary panel of physicians, nurses, and hospital administrators adjudicated preventability and identified common causes of preventable hospital readmissions.
METHODS
Setting
This observational retrospective cohort study was conducted at Vancouver General Hospital (VGH), a teaching hospital and quaternary referral center with 110 GIM beds in Vancouver, British Columbia, Canada. British Columbia administers a publicly funded, nonprofit, health care system where citizens access medically necessary hospital and physician services without paying out of pocket.19
Patient selection
Inclusion criteria were 18 years and older, discharge from the GIM service, readmission to the GIM or family practice services within 28 days of discharge, and hospital readmission length of stay 24 hours or more. Elective readmissions were excluded. The study included readmissions from September 7, 2015, to January 7, 2016. Eligible patients were identified following readmission via the hospital electronic medical record (EMR), which was checked daily during weekdays. Informed consent was obtained for access to medical records and participation in an interview. If patients were unable to provide consent, it was obtained from substitute decision-makers who were then interviewed to understand their perspective on the hospital readmission. Ethics approval was obtained through the University of British Columbia Research Ethics Board.
Data collection
Data were collected from the hospital EMR, paper charts, and interview with patients and their families. Demographic information and characteristics of each hospital admission and readmission were collected from hospital paper charts and EMR by 2 physicians. The LACE index and HOSPITAL score, which are hospital readmission prediction tools, were calculated for each patient.20,21 To assess for the presence of contributing subjective variables, physicians reviewed the charts independently, and then discussed the case and reached a consensus decision. A trained nurse educator interviewed patients to explore their perspectives on factors that led to their readmission. Interview participation was voluntary and consisted of fixed-choice and open-ended questions (see Supplemental Digital Content Appendix 1, available at: http://links.lww.com/QMH/A12). The main domains assessed were symptom recurrence, follow-up care, medication-related issues, mobility/functional difficulties, and social concerns (eg, housing and finances).
Identification of preventable hospital 28-day readmissions
A panel consisting of 3 physicians, 1 clinical nurse educator, and the VGH Director for Strategic Initiatives reviewed each patient's hospital readmission and adjudicated whether it was preventable. The review process consisted of a collective assessment of available data followed by a 15- to 30-minute panel discussion, after which the panel adjudicated whether the hospital readmission was preventable or not by consensus. In contrast to other studies, which utilized 5- or 6-point scales to assess preventability, we employed consensus decision-making to avoid neutral responses and obtain results that would meaningfully inform intervention planning.22-24 A few studies have previously used automated algorithms to identify preventable hospital readmissions, but they were shown to be less accurate, have poor specificity, and overestimate the amount of preventable readmissions.18 For these reasons, we opted to collect and review data manually, as well as determine preventability by panel consensus. None of the reviewers were involved in the care of the patients included in this study.
A hospital readmission was defined as preventable if the patient's care could have been reasonably altered at any point between initial hospitalization and readmission in such a way that the readmission would have been avoided. In line with previous studies, 4 aspects of a patient's care were considered: (1) quality of care during index admission, (2) discharge planning, (3) postdischarge follow-up, and (4) coordination between inpatient and outpatient health care providers.25 Opportunity to reasonably alter care was judged in context of locally available resources and clinical care standards. The hospital admission and readmission were not required to be clinically related to be considered preventable. This was done to avoid bias within the cohort toward only related clinical events, as preventable readmissions can be due to nonclinical factors, such as social supports. In addition, the clinical cause of the index hospitalization and readmission may be different and hence potentially missed.15
Thematic analysis of preventable 28-day hospital readmissions
All hospital readmissions identified as preventable were reviewed by the panel and categorized into major themes, which were derived from previously reported common causes of preventable readmissions.22,23 Major themes were parsed into subthemes if any were identified by the panel. Other themes identified in the review, but not initially derived from the literature, were also noted. Each preventable hospital readmission was permitted to be classified under multiple themes to reflect that a readmission may have multiple causes.
Statistical analysis
Data were expressed as percentages for categorical variables and mean values for continuous variables, unless otherwise specified. GraphPad Prism 7 was used for statistical analysis (GraphPad Software Inc, La Jolla, California). Continuous variables were tested for normality using the Shapiro-Wilk test, and if they followed a normal distribution then unpaired t tests were used to assess the results; otherwise Mann-Whitney U tests were used. The Fisher exact test was used for categorical variables to assess study results. P values were calculated as 2-tailed and a value of <.05 was interpreted as significant.
RESULTS
Patient characteristics
Out of 1607 patients discharged from the GIM service at VGH during the study period, 117 (7.3%) were readmitted to VGH within 28 days. Our study captured 55 of these readmissions, while consent could not be obtained in 62 patients. Table 1 outlines participants' demographics as well as the characteristics of their initial hospitalization and hospital readmission. There were no significant differences in any of the variables between readmissions deemed preventable and nonpreventable. The diagnosis on readmission was concordant with the index admission diagnosis in 53% of patients. The average HOSPITAL score was 5.4 +/- 2.0, which correlates to an intermediate risk (9.8%) of a potentially avoidable 30-day hospital readmission.21 The average LACE index was 9.5 +/- 2.7, which correlates to a 10.3% to 12.2% risk of unplanned 30-day hospital readmission or death.20 The LACE index and HOSPITAL score did not significantly differ between patients with preventable and nonpreventable hospital readmissions.
Patients' interviews
Table 2 summarizes the results of patient interview responses. There were no significant differences in any answer between patients with preventable and nonpreventable hospital readmissions. The most frequent reason for hospital readmission was "symptom recurrence" followed by "new medical issue." No patients attributed their hospital readmission to difficulty accessing primary care, medication-related issues, or functional concerns. Of the 6 patients who reported barriers to taking their medications as prescribed, 2 were unable to fill their prescription, 2 patients stopped their medications due to side effects, and 2 labeled "other" as their barrier. The most common reason reported for not following up with their primary care provider after discharge was rehospitalization. This was followed by lack of awareness of appointment, or not having a primary care provider.
Analysis of preventable hospital readmissions
Twenty-nine patients (53%) were determined by the review panel to have a preventable hospital readmission, with an average of 2.1 major causal themes per patient. Results are displayed in the Figure. "Inadequate coordination with community services on discharge" was the most common cause of a preventable hospital readmission (18/29 patients). This major theme was found to have 2 subthemes, which are outlined in Table 3. The second most common major theme was "insufficient postdischarge follow-up with physician" (13/29 patients).
Among preventable readmissions attributed to "suboptimal end-of-life care," 72% (8/11) were due to lack of referral to community palliative care services. The remainder were due to inadequate communication between the discharging health care team and preexisting palliative care services. These subthemes are shown in Table 3.
The review panel identified 2 additional themes among preventable hospital readmissions. These were poor communication with primary care and inadequate multidisciplinary communication on discharge, also displayed in Table 3. They were a causal factor in 38% (11/29) and 14% (4/29) of preventable hospital readmissions, respectively.
DISCUSSION
We found that 53% of 28-day hospital readmissions among GIM patients were potentially preventable, which was the primary objective of our study. Our secondary aim was to determine their common causes. On average, patients had more than 2 causes attributed to their hospital readmission. These findings highlight potential opportunities to reduce unplanned readmissions and emphasize the need for interventions that extend beyond the walls of the hospital.
Our study identified a higher proportion of potentially preventable hospital readmissions compared with prior research. A recent US multicenter observational cohort study of general medicine patients found 26.9% of hospital readmissions to be potentially preventable.23 A large Canadian multicenter prospective cohort study of both medical and surgical patients found the incidence of potentially avoidable hospital readmissions within 6 months to be 16%, which is a notably different timescale.24 Moreover, a systematic review and a meta-analysis both showed that on average one quarter of hospital readmissions are preventable.13,14 We believe our finding is explained by our broader inclusion criteria; the inclusion of readmissions where diagnoses were different than on index admission, and the inclusion of patients who left their initial hospitalization against medical advice. Furthermore, consideration of nonclinical factors-such as communication between health care professionals and the social determinants of health-may have led to increased identification of preventable readmissions as compared with other studies in which preventability was mainly judged by clinical factors.
Our methodology in adjudicating preventable hospital readmissions is unique due to its 5-member multidisciplinary review panel. Review of prior literature identified that 54.8% of similar studies had 1 reviewer, while 90.3% of prior studies were reviewed only by physicians.13 Moreover, a recent multicenter study of 1000 general medicine patients used a 2-physician review board.23
The most common cause of preventable hospital readmissions was "inadequate coordination with community services on discharge." This is similar to a prior study of 250 potentially preventable readmissions where the most common factor contributing to the readmission was termed "transition care planning and care coordination."22
In 11 of the 29 preventable readmissions, lack of liaison with palliative care on discharge was identified as a potential cause for readmission. We found this theme to be uniquely specific and actionable, thus being a good target for intervention. Eight of these patients had terminal conditions, functional decline, and were not referred to palliative care at the time of discharge. The remainder had inadequate arrangement of follow-up with community palliative care resources that were previously in place. Appropriate involvement of palliative care during hospitalization has been shown to reduce 30-day hospital readmission rates.26-28
In our cohort, neither the LACE index or HOSPITAL score showed a relationship with preventable hospital readmissions. Given the LACE index was designed to predict the risk of overall hospital readmission or death, and not preventable readmissions in specific, this could be an expected result. In contrast, the HOSPITAL score has been validated to predict potentially avoidable hospital readmissions.29 However, the international validation study of the HOSPITAL score did not show a statistically significant difference in patients with a preventable readmission in their "high" risk category.29 The inability to predict preventability using the HOSPITAL score in our cohort may be attributed to the fact that all of the patients were discharged from a GIM service. At our center there are few dedicated oncology beds, and many patients with active malignancy are admitted to the GIM service. These patients would not be eligible for 2 of the 11 points in the HOSPITAL score designated to discharges from an oncology ward, even if they had an active malignancy. Another reason may be that our process of assessing preventability involved investigation of nonclinical factors, not captured by the HOSPITAL score or LACE index.
LIMITATIONS
A significant limitation to our study is the lack of any available validated method for manual identification of preventable hospital readmissions. Larger studies have borrowed approaches used in identifying preventable adverse drug events; these use a continuous Likert scale to rank the relative degree of preventability.23,24 To optimize findings in our relatively small cohort, we employed consensus decision-making to more clearly delineate preventable from nonpreventable readmissions. Another limitation is that patients who were discharged from VGH and readmitted to other hospitals within 28 days were not captured. Close to 20% of hospital readmissions occur at other hospitals.30 While this limits generalizability, our study still captures the majority of unplanned hospital readmissions. We were not able to capture all readmissions that occurred during the study period (53% of readmission were not captured), yet our sample is large enough to suggest potential areas of focus for policymakers and hospital administrators.
CONCLUSION
This study analyzed 55 28-day hospital readmissions, including the perspectives of patients and varied health care professionals. Fifty-three percent of these readmissions were preventable. The most common contributing factor was "inadequate coordination with community services on discharge," which is concordant with prior studies. Suboptimal end-of-life care was identified as a major causal theme among preventable hospital readmissions. Future quality improvement efforts would benefit from focusing on these themes.
REFERENCES