Examining Predictive Modeling Based Approaches to Characterizing Healthcare Fraud

Background: Healthcare fraud can represent upwards of hundreds of billions of dollars in spending that could be better spent on patient care. There is often not sufficient detail on the underlying methodologies and data samples that lead to fraud estimates, which may be due to different purposes of these reports or the need to obscure the details of fraud detection methods to prevent fraudulent operators from responding to existing methods.

Objectives: The objective of this study was to provide a systematic evaluation and synthesis of the methodologies and data samples used in current peer-reviewed studies on characterizing healthcare fraud.

Data Sources: The academic databases searched were Academic Search Complete, Business Source Complete, EconLit, Medline (EBSCO), OneSearch, ProQuest Business Collection, ScienceDirect, and Web of Science. Governmental and commercial sources were also used for background research.

Synthesis of Methods: This examination was conducted using a systematic review methodology to identify relevant studies and determine their relevance. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was used to guide the performance of reviewing the literature. Study criteria for eligibility were collected by applying specific search terms: healthcare, health insurance, Medicare, Medicaid, Obamacare, Affordable Care Act, or health services; fraud, cheat, falsification, corruption, or kickback; detect, detection, prevent, prevention, deterrence, audit, or auditing. Results were restricted to scholarly journals, academic journals, working papers, and conference proceedings. Study selection occurred through two independent reviews of each study for inclusion or exclusion. Disagreements between reviewers were resolved through discussion by the entire research team.

Results: Our search terms resulted in 450 articles that were potentially appropriate for inclusion in our report. The results of independent reviews ended with twenty-seven studies considered as relevant to include after the application of our inclusion criteria. Variables are identified from the literature to synthesize each method of fraud detection used.

Limitations: One limitation of this study is that the strength of the evidence is reliant on the quality and number of studies previously performed on the topic. Another limitation is the quality of studies with regard to their applicability to different types of insurers. Finally, the majority of studies could not provide proof of intent to commit fraud.

Conclusions: A limited number of validated methods are used to detect healthcare fraud. The literature on this topic is spread among several academic fields. The majority of available studies utilize public or social health insurance systems such as Medicare or Medicaid in order to study fraud. The main gaps we identified are validation of existing methods and proof of intent to commit fraud in the studies analyzed.

Implication of Key Findings: Our insurer agnostic approach examines the availability and effectiveness of healthcare fraud analytic methods across different types of health insurers, posing great value for members of the health sectors.