Health Policy Research

Start with the decision, then test the data

Health-policy screening systems allocate investigative attention, create burdens for providers, and can affect access to care. The economic question is whether a score improves a real decision after label error, missing clinical context, false positives, and enforcement selection are counted.

Medicaid and Program Integrity

The named case is the HHS Medicaid Provider Spending by HCPCS dataset, released on February 9, 2026. It aggregates 2018 to 2024 billing-provider, servicing-provider, procedure, and month records. HHS warns that state variation may reflect policy, coding, or submission differences and that low-volume rows are suppressed.

The four-part series turns that data release into a competency demonstration: inventory the available fields, define the label, inspect billing patterns, then compare a prospective classifier with simple baselines. Exact label shares, provider comparisons, and classifier metrics are currently marked under reconciliation. They remain visible as validation targets, not established findings.

Why Health Policy Needs Better Methods

Health policy analysis sits at the intersection of several challenges: large administrative datasets with limited variables, strong political incentives to find dramatic results, and real consequences when analysis goes wrong. Providers flagged as suspicious may lose their ability to serve vulnerable populations. Genuine fraud may go undetected when screening methods optimize for volume rather than behavioral anomalies.

The working rule is simple: document the decision, population, label, observation window, threshold, and cost of error before interpreting model performance. Each article carries its own evidence status so a public-source fact cannot be confused with an article-reported result.

Research Methodology

Our health policy research draws on:

• Administrative data: CMS Medicaid billing records, LEIE exclusion database, NPI provider registry
• Statistical methods: Logistic regression, classification validation, honest train-test splitting
• Policy analysis: Cost-effectiveness frameworks, program evaluation design

Frequently asked questions

What is Medicaid fraud detection?

Medicaid fraud detection uses billing and provider records to prioritize claims or providers for review. Billing volume alone cannot establish fraud. This site reports a 40% non-fraud share among matched LEIE exclusions, but that estimate is article-reported and not yet publicly reproduced.

Can public Medicaid data identify fraud?

The HHS provider-spending dataset aggregates billing-provider, servicing-provider, procedure, and month records from 2018 through 2024. It can support utilization and outlier analysis, but it cannot establish fraud without richer clinical and investigative context.

How does this research relate to health economics?

The research treats fraud screening as an economic decision problem: define the label and threshold, measure false-positive and false-negative costs, compare simple baselines, test out of time, and document who bears each error.