Can we predict functional decline in hospitalized older people admitted through the emergency department? Reanalysis of a predictive tool ten years after its conception

Isabelle De Brauwer, Pascale Cornette, Benoît Boland, Franck Verschuren, William D’Hoore

In the Emergency Department (ED), early and rapid identification of older people at risk of adverse outcomes, who could best benefit from complex geriatric intervention, would avoid wasting time, especially in terms of prevention of adverse outcomes, and ensure optimal orientation of vulnerable patients. We wanted to test the predictive ability of a screening tool assessing risk of functional decline (FD), named SHERPA, 10 years after its conception, and to assess the added value of other clinical or biological factors associated with FD. A prospective cohort study of older patients (n = 305, ≥ 75 years) admitted through the emergency department, for at least 48 h in non-geriatric wards (mean age 82.5 ± 4.9, 55% women). SHERPA variables (i.e. age, pre-admission instrumental Activity of Daily Living (ADL) status, falls within a year, self-rated health and 21-point MMSE) were collected within 48 h of admission, along with socio-demographic, medical and biological data. Functional status was followed at 3 months by phone. FD was defined as a decrease at 3 months of at least one point in the pre-admission basic ADL score. Predictive ability of SHERPA was assessed using c-statistic, predictive values and likelihood ratios. Measures of discrimination improvement were Net Reclassification Improvement and Integrated Discrimination Improvement. One hundred and five patients (34%) developed 3-month FD. Predictive ability of SHERPA decreased dramatically over 10 years (c = 0.73 vs. 0.64). Only two of its constitutive variables, i.e. falls and instrumental ADL, were significant in logistic regression analysis for functional decline, while 21-point MMSE was kept in the model for clinical relevance. Demographic, comorbidity or laboratory data available upon admission did not improve the SHERPA predictive yield. Prediction of FD with SHERPA is difficult, but predictive factors, i.e. falls, pre-existing functional limitation and cognitive impairment, stay consistent across time and with literature. As accuracy of SHERPA and others existing screening tools for FD is moderate, using these predictors as flags instead of using composite scales can be a way to screen for high-risk patients.