Readmissions in internal medicine: Epidemiology and determinants

Abstract

Introduction: Readmissions are frequent in Internal Medicine wards and can be considered a score for quality and efficiency in health care. The readmission rate (RR) is commonly used as a hospital quality indicator and is frequently included in the evaluation of hospital performance. Penalties based on RR are being introduced in several models for hospital financing. The effective value of RR and the time frame to be considered are not consensual. Objective: Quantify the RR, characterize the readmission episodes and identify its determinants. Methods: Retrospective study of all the admissions in an Internal Medicine ward of a university central hospital in 2011. Data source: Diagnostic Related Categories (DRG) electronic administrative database. Readmission was defined as any new admission following an index admission during the year 2011. Variables: patient demographics, diagnosis (ICD-9 and Clinical Classification Software — AHRQ, U.S.A.) and DRG, length of stay (LoS), comorbidity indexes (CoI) (Charlson and Elixhauser) and chronic conditions (Chronic Condition Indicator — AHRQ, U.S.A.). Results: From 3826 admissions, 663 (17.3%) were readmissions (5.8% below 7 days and 11.9% below 30 days). There were no differences in gender distribution. The average age of the readmission episodes was greater than the non-readmissions (74.2 vs 72.7 years, p = 0.001). The LoS of the index episodes preceding a readmission was higher than those without a subsequent readmission (9.3 vs 7.3 days, p < 0.001). Only 22.5% of the readmissions were coded in the same DRG and only 24.1% in the same main diagnosis of the index preceding admission. The cases associated with a readmission had a higher number of coded diagnosis (7.5 vs 6.8) and chronic conditions (4.7 vs 4.1). The CoI were also greater in the readmission index cases (Elixhauser: 2.8 vs 2.3 and Charlson: 2.9 vs 2.2) (p < 0.001). In the logistic regression multivariate model for the occurrence of readmission after an initial episode, age, LoS and CoI were identified as significant variables. Conclusions: Readmissions have relevant weight in inpatient hospital activity in Internal Medicine. In our study the early RR was low and 3/4 of the readmissions occurred in clinical contexts distinct from the initial index episode. We stress the inverse relation between the shorter LoS and the probability of readmission and emphasize the higher complexity associated with readmission episodes. Extra-hospitalar facts, namely individual patient characteristics, are relevant determinants for readmission. In Internal Medicine the raw RR should not be used as a quality, performance or efficiency indicator, as RR should be adjusted to the patient clinical complexity, gaged accordingly to the nosological profile of the institution and weighted taking into account the efficiency of community health care.