Using Machine Learning to Predict Patterns of Employment and Day Program Participation

Area of Science:

• Health Services Research
• Artificial Intelligence in Healthcare
• Disability Studies

Background:

• Evidence-based policy making for intellectual and developmental disabilities (IDD) requires robust analytic tools.
• Current data analytics may not fully capture the complexity of factors influencing outcomes for people with IDD.

Purpose of the Study:

• To explore the utility of machine learning (ML) as an inductive analytic tool for IDD policy and practice.
• To predict employment status and day activity participation for individuals with IDD using ML models.

Main Methods:

• Utilized data from the National Core Indicators In-Person Survey (NCI-IPS), a large-scale survey of individuals with IDD.
• Developed and compared classification tree and random forest models to predict key outcomes.
• Employed a nationally representative sample of over 20,000 individuals with IDD.

Main Results:

• A random forest classifier achieved high accuracy, predicting employment outcomes at 89% on the testing sample and 80% on the holdout sample.
• The most critical predictor identified was whether community employment was a stated goal in the individual's service plan.
• Models analyzed relationships between employment/activity participation and other survey responses.

Conclusions:

• Machine learning demonstrates significant potential for enhancing the evidence base in IDD research and policy.
• ML tools can effectively analyze complex datasets to identify key drivers of valued outcomes for people with IDD.
• Future applications could extend ML to other critical outcomes for informed policy and practice development.