In a machine-learning analysis of 543 people with Parkinson’s disease in the Parkinson’s Progression Markers Initiative (PPMI), a random forest model predicted 1-year falls with 82.6% accuracy (MCC 0.456; NPV 91.8%). Beyond prior fall history, glaucoma ranked among the most influential predictors of future falls — alongside arising-from-chair difficulty and gait.

Why It Matters To Your Practice

• Falls drive morbidity, loss of independence, and hospital utilization in Parkinson’s disease (a Disease of the nervous system), and risk stratification remains imperfect.

• This study suggests an ophthalmic diagnosis (glaucoma) may meaningfully contribute to fall-risk prediction — not just classic motor features.

Clinical Implications

• Add vision and eye-disease status to fall-risk conversations: ask about glaucoma diagnosis/treatment and functional visual limitations during routine PD visits.

• Consider tighter coordination with ophthalmology/optometry for PD patients with glaucoma (or suspected glaucoma), as part of a multifactorial fall-prevention plan.

• Use risk signals pragmatically: the model’s high NPV (91.8%) may help reassure when fall risk appears low, while its modest PPV (50%) argues for confirming risk with clinical assessment rather than acting on prediction alone.

Insights

• Methods: CATDAP-based feature selection + random forest; 80/20 train-test split; outcome was falls at 1 year.

• Performance in the test set: sensitivity 63.2%, specificity 86.7%, F1 0.558 — suggesting better “rule-out” than “rule-in.”

• Feature importance highlights a practical triad beyond fall history: glaucoma, arising from chair, and gait — pointing to a combined visual + mobility pathway to falls.

The Bottom Line

• In PD, glaucoma may be more than a comorbidity: it emerged as a high-impact predictor of 1-year falls in an ML model, supporting more intentional vision-focused screening and interdisciplinary fall prevention.