Predictors of 30-day post-discharge unplanned readmission in a subacute geriatric ward in Singapore

Abstract

Background. Unplanned readmission to hospital is common among older adults and contributes to considerable healthcare costs and hospital-associated complications. We aimed to identify predictors of 30-day post-discharge unplanned readmission among older adults in our subacute geriatric ward, and to determine the prevalence of geriatric syndromes and develop a new predictive model for readmission of subacute geriatric patients. Methods: Consecutive patients admitted to our subacute geriatric ward between June 2018 and June 2019 were invited to participate. Data collected included patient age, sex, weight, height, race, type of housing, destination upon discharge, functional and frailty status, presence of conduits (urinary catheters and nasogastric tubes), polypharmacy, high-risk medications, healthcare utilisation 6 months prior, laboratory test results, length of hospital stay, Charlson Comorbidity Index, and LACE index. Patients were assessed using the Mini Nutritional Assessment -Short Form, Geriatric Depression Scale, Mini-Mental State Examination, Clinical Frailty Scale, FRAIL scale, modified Barthel Index, hand grip strength, and gait speed. Patients with or without 30-day post-discharge unplanned readmission were compared. Multivariate logistic regression was used to identify independent predictors. Results: Of 284 patients followed up at 30 days post-discharge, 63 (22.2%) had unplanned hospital readmission within 30 days of discharge, with associated factors being history of myocardial infarction, moderate or severe liver or renal disease, low albumin level, history of emergency department visits, hospitalisation in the preceding 6 months, and discharge to a destination other than home. The prevalence of geriatric syndromes of falls, frailty, and immobility was 62.3%, 64.7%, and 86.6%, respectively. Independent predictors of 30-day post-discharge unplanned readmission were history of hospitalisation in the preceding 6 months (odds ratio=2.62, p=0.045) and discharge destination other than home (odds ratio=3.10, p=0.006). The area under the receiver operating characteristics curve for the predictive models was between 0.6 and 0.7, and Brier score was around 0.16. The discrimination ability of the models was weak. Conclusion: History of hospitalisation in the preceding 6 months and not being discharged to home were independent predictors for 30-day post-discharge unplanned readmission.