Identifying specific subgroups of older patients at risk of repeated hospital readmissions and death after discharge in a prospective multicentre cohort in France

Objective To identify specific subgroups of older patients at risk of repeated hospital readmissions and death. Design Prospective, multicentre, DAMAGE (Patient Outcomes After Hospitalization in Acute Geriatric Unit) cohort of adults aged 75 and over, discharged from an acute geriatric unit (AGU) and followed up for 12 months. Setting Six recruiting hospital centres in the Hauts-de-France and Normandie regions of France. Main outcome measures We performed a latent class analysis to identify subgroups at risk of repeated hospital readmissions and death, followed by a logistic regression analysis to determine the characteristics associated with the identified subgroups. Results 3081 patients were included (mean (SD) age: 86.4 (5.5)) and two subgroups were identified. In subgroup 1 (n=2169, 70.4%), only 619 (28.5%) patients were readmitted to hospital once during the follow-up, and 495 (22.5%) died. In subgroup 2 (n=912, 29.6%), all patients were readmitted to hospital at least twice, and 523 (57.8%) died. Subgroup 2 accounted for 29.6% of patients but 74.4% of hospital readmissions, with longer lengths of stay, and 51.6% of deaths. A multivariate logistic regression analysis identified only four characteristics weakly associated with the risk of being in subgroup 2 (at least one hospital admission in the 6 months preceding the index hospital admission, cancer, polymedication and weight changes (gain or loss) during the index hospital admission). The area under the receiver operating characteristic curve was 63%. Conclusion A latent class analysis showed that a population of older adults hospitalised in an AGU is divided into two subgroups with regard to the postdischarge outcomes: one subgroup (70% of the individuals) will have a low rate of hospital readmission and a moderate death rate, whereas the other will have a high rate of hospital readmission and a very high death rate. There is a need for predictive scores for both events, with a view to better targeting at-risk patients.