Towards Protecting Vital Healthcare Programs by Extracting Actionable Knowledge from Policy

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 3509–3521 August 1–6, 2021. ©2021 Association for Computational Linguistics 3509 Abstract In challenging economic times, obtaining value for money by ensuring financial integrity and fairer distribution of services are among the top priorities for social and health-care systems globally. However, healthcare billing policies are complex and identifying non-compliance is often narrow-scope, manual and expensive. Maintaining ‘integrity’ is a challenge - ensuring that scarce resources get to those in need and are not lost to fraud and waste. Our approach fuses recent advances in dependency parsing with a policy ontology to convert the content of regulatory healthcare policy into human-friendly policy rules, that are amenable to machine- execution, with human oversight. We describe the ontology-guided transformation of textual patterns into a semantically-meaningful knowledge graph of rules, outline our experiments and evaluate results against policy rules obtained from professional investigators. The aim is to make a policy-compliance ‘landscape’ visible to healthcare programs - helping them identify Fraud, Waste or Abuse. 1 Introduction The WHO (World Health Organization, 2010) lists fairness, financial integrity [“Program Integrity”] and access in healthcare among the top global healthcare priorities. In the U.S., an estimated annual amount of USD$20-30B is lost to Fraud, Waste and abuse (FWA) (Shrank et al., 2019). These vital funds never make it to the vulnerable citizens that they were intended to serve. To combat this, countries with insurance-based healthcare programs (e.g. Medicaid, Medicare), 1 The ontology and Benefit Rules benchmark are released as open source at: employ claim investigators to validate the integrity of reimbursement claims submitted by providers. Investigators verify these claims against the program’s policies, with the goal of reducing wasteful practices, identifying fraud or abuse and closing policy gaps. This is a labor-intensive task – claim volumes are high, policies are complex and investigative resources are limited. More broadly, governments regulate a wide range of sectors, with extensive rules and policies. These policies drive significant spending by the regulated organizations. e.g., a European Union Commission study (2019) found that the annual cost of complying with EU financial regulations is around EUR€11.3B. In Australia, Deloitte (2014) estimated the costs of administering and complying with public sector rules at AUD$94B. In this paper we propose a methodology for automatically extracting knowledge from healthcare policy documents, in the form of Benefit Rules (BRs) that are both human- understandable and machine consumable. These BRs can be applied automatically to flag discrepancies in claims, with limited human effort. This paper focuses on the extraction of these rules, and not their execution. Policy rules have a major impact both on the health of the citizens they serve, and the financial integrity of the Programs that pay the Service Providers. Human-understanding, oversight and control are first-class AI-design concerns for this domain. Model ‘explainability’ is not enough. Transparency is needed anywhere that provider or citizen coverage is at stake. Users need to see and influence which rules are being applied, and on what policy basis decisions are being …

Content truncated to 500 words