Short- and longer-term predictive capacity of the Multidimensional Prognostic Index: The timing of the assessment is of no consequence

Several studies have tested the ability of the Multidimensional Prognostic Index (MPI) to predict mortality for acute elderly patients admitted to hospital. We compared the reliability of MPI scores obtained both on admission to, and at discharge from hospital. We tested MPI performance in predicting short- and longer-term mortality grouped into three risk groups and according to single MPI scores. A longitudinal prospective study was conducted on 699 elderly patients admitted to the Geriatric Unit at Padua General Hospital. MPI scores were obtained on admission and at discharge. In-hospital and one-year mortality was recorded. Adjusted Cox's regression models were used to assess the prognostic value of the MPI scores. 691 were included in the study: 459 (66.4%) women and 232 (33.6%) men, mean age=85.2 ± 7.0 years. Patients were grouped as: low risk MPI 12.5%; moderate risk MPI 28.6%; severe risk MPI 58.9%. The cumulative in-hospital mortality rate was 7.4%. In the adjusted model, only MPI score (not MPI risk group) was significantly associated with in-hospital death ([HR]=1.22, 95%CI 1.07-1.39). 1-Year crude mortality rate: 39.2%. The patients' MPI scores at admission and at discharge were equally predictive of death (adjusted HR of MPI on admission 1.20 [1.15-1.27], p<0.0001; at discharge 1.24 [1.18-1.30], p<0.0001). The performance (AUC) of the MPI score on admission and at discharge proved much the same. This study confirmed the value of the MPI in predicting mortality for acute elderly patients. Grouping MPI scores into risk levels may not be appropriate when applied to hospitalized acute geriatric patients. The prognostic value of MPI scores was confirmed only for MPI value ≥0.68. Judging from our study, the timing of the assessment during a patient's hospital stay (on admission or at discharge) may be irrelevant for longer-term prognostic purposes.