Related Experiment Video
Updated: May 22, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Prediction of functional decline in older hospitalized patients: a comparative multicenter study of three screening
Mieke Deschodt1, Nathalie I H Wellens, Tom Braes
1Center for Health Services and Nursing Research, Katholieke Universiteit Leuven, Kapucijnenvoer 35/4, B-3000 Leuven, Belgium.
Background And Aims:
Older hospitalized patients are at risk of functional decline, which is associated with several negative outcomes. The aim of this study was to compare the predictive accuracy of the Identification of Seniors At Risk (ISAR), Variable Indicative for Placement risk (VIP) and the Flemish version of the Triage Risk Screening Tool (TRST) in predicting functional decline.
Methods:
A prospective cohort study with 30 days follow-up in geriatric, medical and surgical wards in 25 hospitals was conducted. 752 participants aged 75 years or older were eligible for inclusion. Baseline data were gathered within 72 hours of admission. Functional decline was defined as an increase of one point or more from the premorbid Katz score to the score 30 days post-discharge. Positive predictive value (PPV) and negative predictive value (NPV) were calculated on 2 x 2 tables as well as by Bayes' theorem.
Results:
Functional decline at 30 days postdischarge was observed in 279 participants (39%). ISAR and Flemish TRST showed high sensitivity (88% - 78%) and fair NPV (62% - 67%), but low specificity (19% - 30%) and low PPV (47% - 48%) using the original cut-off of ≥ 2. The sensitivity of VIP with cut-off ≥ 2 was too low (62%), but could be optimized with cut-off ≥ 1, showing sensitivity, specificity, PPV and NPV of 88%, 21%, 48% and 68%, respectively. Accuracy varied between 40% and 61% for all instruments on all calculated cut-offs.
Conclusions:
All three instruments performed similarly well, showing good sensitivity and fair NPV, the two major characteristics for good screening tools. False positives could be filtered out according to the clinical expert opinion of a care team.
