Closing the Gate Before the Horse is out of the Barn: A Model to Effectively Predict the First Fall in Patients with Parkinson's Disease.

BACKGROUND

Most fall-prediction models in Parkinson's disease (PD) rely on prior falls, limiting prevention. The Waiting Room of the Future (WROTF) platform collects standardized motor and cognitive measures during routine visits, integrated into electronic health records (EHR).

OBJECTIVE

To develop a first-fall prediction model combining WROTF-EHR data and compare its performance with established cognitive- and motor-based models.

METHODS

Previously published motor (Paul) and cognitive (CYCLE-II) models were refit using logistic regression. A new regularized (LASSO) model was developed. All models were evaluated using all participants and only those without prior falls.

RESULTS

Among 997 PD patients (mean age 69.6 ± 8.8; 62.8% male; disease duration 3.6 ± 4.3 years), AUCs were 0.77 (Paul), 0.70 (CYCLE-II), and 0.81 (LASSO). In fall-naïve patients, AUCs were 0.62, 0.68, and 0.70. Key predictors included processing speed, walking speed, disease duration, and age.

CONCLUSIONS

The model enables early fall-risk detection using routinely collected data, supporting proactive and personalized preventive care.