External validation of the Probability of repeated admission (Pra) risk prediction tool in older community-dwelling people attending general practice: a prospective cohort study

OBJECTIVES Emergency admission is associated with the potential for adverse events in older people and risk prediction models are available to identify those at highest risk of admission. The aim of this study was to externally validate and compare the performance of the Probability of repeated admission (Pra) risk model and a modified version (incorporating a multimorbidity measure) in predicting emergency admission in older community-dwelling people. SETTING 15 general practices (GPs) in the Republic of Ireland. PARTICIPANTS n=862, ≥70 years, community-dwelling people prospectively followed up for 2 years (2010-2012). EXPOSURE Pra risk model (original and modified) calculated for baseline year where ≥0.5 denoted high risk (patient questionnaire, GP medical record review) of future emergency admission. PRIMARY OUTCOME Emergency admission over 1 year (GP medical record review). STATISTICAL ANALYSIS descriptive statistics, model discrimination (c-statistic) and calibration (Hosmer-Lemeshow statistic). RESULTS Of 862 patients, a total of 154 (18%) had ≥1 emergency admission(s) in the follow-up year. 63 patients (7%) were classified as high risk by the original Pra and of these 26 (41%) were admitted. The modified Pra classified 391 (45%) patients as high risk and 103 (26%) were subsequently admitted. Both models demonstrated only poor discrimination (original Pra: c-statistic 0.65 (95% CI 0.61 to 0.70); modified Pra: c-statistic 0.67 (95% CI 0.62 to 0.72)). When categorised according to risk-category model, specificity was highest for the original Pra at cut-point of ≥0.5 denoting high risk (95%), and for the modified Pra at cut-point of ≥0.7 (95%). Both models overestimated the number of admissions across all risk strata. CONCLUSIONS While the original Pra model demonstrated poor discrimination, model specificity was high and a small number of patients identified as high risk. Future validation studies should examine higher cut-points denoting high risk for the modified Pra, which has practical advantages in terms of application in GP. The original Pra tool may have a role in identifying higher-risk community-dwelling older people for inclusion in future trials aiming to reduce emergency admissions.