Administrative and Claims Records as Sources of Health Care Cost Data

Abstract

BACKGROUND: Many economic studies of disease require cost data at the person level to identify diagnosed cases and to capture the type and timing of specific services. One source of cost data is claims and other administrative records associated with health insurance programs and health care providers. OBJECTIVE: To describe and compare strengths and limitations of various administrative and claims databases. DATA AND METHODS: Data sources included claims and enrollment records from Medicare, Medicaid, and private insurers; Veterans' Health Administration records; state hospital discharge datasets; Healthcare Cost and Utilization Project hospital databases; managed care plan data systems; and provider cost reports. Claims provide information on payments, whereas cost reports yield resource costs incurred to produce services. Administrative data may be significantly augmented by linkage to disease registries and surveys. RESULTS: Administrative data are often available for large, enrolled populations, have detailed information on individual service use, and can be aggregated by service type, episode, and patient. Service use and costs can often be tracked longitudinally. Because they are not collected for research purposes, administrative data can be difficult to access and use. Limitations include generalizability, complexity, coverage and benefit restrictions, and lack of coverage continuity. Linked datasets permit identification of incident cases of disease, and analyses of health care costs by stage at diagnosis, phase of care, comorbidity status, income, and insurance status. CONCLUSIONS: Administrative data are an essential source of information for studies of the financial burden of disease. Cost estimates can vary substantially by specific measures (payments, charges, cost to charge ratios) and across data sources.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 252 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.