Objective: This project aimed to develop an advanced claims-based algorithm to assign severity level to people with Parkinson’s (PwP) to enhance research on disease progression. Specifically, the research team plans to evaluate the impact of environmental factors on the progression of Parkinson’s Disease (PD) in Texas but devised an algorithm to identify severity levels first.

Background: Effective management of PD requires an accurate and dynamic assessment of disease severity to guide both pharmacological and non-pharmacological treatments. As the

severity of the disease symptoms increases, individualized therapeutic interventions become important to maintain optimal patient outcomes. Disease staging is useful for disease management but can also strengthen epidemiologic research measuring disease burden, patterns in progression, and treatment outcomes at a population level. In clinical practice and research, the severity of Parkinson’s disease is commonly measured with the Hoehn and Yahr scale and the Unified Parkinson’s Disease Rating Scale. Dahodwala et al also presented a pharmacy claims-based algorithm to identify advanced PD.

Method: The Texas All-Payor-Claims-Database (TX-APCD) contains claims for 76,768 PwP (~9% of the nation’s estimated prevalence). Applying prior strategies, we created an algorithm that abstracts diagnoses, symptoms, procedures, and levodopa equivalent dosage (LED) to assign severity level (Early, Mid, Late) to PwP in claims data. Testing of the algorithm left 11% of the PwP too ambiguous to assign a level. We then turned to artificial intelligence to improve the model. Random pairs of PwP were selected from the claims database and their demographics, medical and drug history were summarized. Neurologists who specialize in PD then applied their clinical judgment to assign severity levels. AI used reinforcement learning with text prompt to build an improved algorithm. The algorithm relies on a calculation table that assigns weighted points to factors derived from claims, resulting in a summed classification of early to late stage.

Results: A complex algorithm that incorporates key data from claims has been shown to perform well in severity assignment with minimal mismatches to clinical assessment.

Conclusion: The refined claims-based severity assignment algorithm will be utilized by our research team to assess the impact of environmental factors on the progression of Parkinson’s

Disease in Texas.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/refining-a-claims-based-severity-rating-scale-for-parkinsons-disease-using-machine-learning/