How Can Public Health Use Claims Data? an Evaluation of All-Payer Claims Data for Chronic Disease Surveillance — New Hampshire, 2011

BACKGROUND: Nationally, chronic diseases like hypertension and diabetes contribute to morbidity and cause thousands of deaths annually. Public health traditionally monitors chronic diseases from data sources including self-reported surveys, hospital discharges, and mortality statistics. These sources have limitations in timeliness, comprehensiveness, or self-reporting biases. In 2005, by state statute, New Hampshire (NH) created the Comprehensive Healthcare Information System (CHIS), an all-payer claims database (APCD), to increase health cost transparency. NH law requires commercial and third-party payers with ≥200 subscribers to submit claims data within 30 days of the paid date. The NH Division of Public Health Services sought to evaluate NH-CHIS to understand how it might complement chronic disease surveillance. 

METHODS: We evaluated NH-CHIS using CDC surveillance system evaluation guidelines. We identified and interviewed key stakeholders using a semi-structured questionnaire to assess NH-CHIS attributes (timeliness, stability, simplicity, flexibility, representativeness, acceptability, and data quality). There was no gold standard from which to assess sensitivity; however, we compared 2011 data from NH-CHIS (commercial payers and Medicaid) to the NH Behavioral Risk Factor Surveillance System (BRFSS) and calculated hypertension prevalence among non-institutionalized adults ages ≥18 years. Hypertension was defined by any mention of hypertension as a claim diagnosis (ICD-9 codes 401–405 and 437.2) or by self-report.

RESULTS: We identified and interviewed 13 key stakeholders (data analysts [n = 2], policy makers [n = 2], project managers [n = 2], medical coders [n = 3], an APCD subject-matter expert [n = 1], and epidemiologists [n = 3]).  NH-CHIS contains Medicaid data through 2011, incomplete Medicare data, and no data from patients who are self-pay or uninsured. Approximately 60% of NH residents 0–64 years old are covered by commercial insurance; representativeness of the insured population is considered to be good. Stakeholders reported that NH-CHIS was timely, stable, and flexible, and had improved data quality since 2010. Complex claims data structure resulted in low simplicity and decreased acceptability among epidemiologists. In 2011, hypertension prevalence was 21.6% in NH-CHIS and 30.5% (95% CI 29.1–32.0) in BRFSS.

CONCLUSIONS: NH-CHIS provides data on clinical diagnoses that are otherwise unavailable. Overcoming the challenges of complex data structure might improve acceptability and use of claims data in public health. Exploration of the demographic representativeness and comparison of NH-CHIS to mortality statistics and hospital discharge data are needed to understand how NH-CHIS might complement current chronic disease surveillance.