A predictive model to identify Parkinson disease from administrative claims data

Area of Science:

• Neurology
• Data Science
• Epidemiology

Background:

• Parkinson disease (PD) diagnosis relies on clinical symptoms, often appearing after significant neurodegeneration.
• Early identification of PD is crucial for managing the disease and improving patient outcomes.
• Administrative claims data offer a vast resource for retrospective patient analysis.

Purpose of the Study:

• To develop and validate a predictive model for incident Parkinson disease (PD) using administrative medical claims data.
• To assess the feasibility of identifying individuals with a high probability of future PD diagnosis prior to clinical confirmation.

Main Methods:

• A population-based case-control study utilized Medicare claims data from beneficiaries aged 66-90 years.
• The elastic net algorithm was employed to build a predictive model using demographic data and historical claims (2004-2009).
• Model performance was evaluated using receiver operator characteristic area under the curve (AUC), comparing it to simpler models.

Main Results:

• A comprehensive model incorporating 536 diagnosis and procedure codes achieved an AUC of 0.857 (95% CI 0.855-0.859).
• This model demonstrated a sensitivity of 73.5% and specificity of 83.2% at the optimal threshold.
• Simpler models based on known PD risk factors had significantly lower predictive power (AUC 0.670).

Conclusions:

• Administrative claims data, including demographic information and specific diagnosis/procedure codes, can effectively identify individuals at high risk for Parkinson disease.
• This predictive modeling approach holds promise for earlier detection of PD.
• The findings support the use of claims data for proactive identification of neurodegenerative diseases.