ABSTRACT
PURPOSE: We investigated time trends in validation performance characteristics for six sources of death data available within the Healthcare Integrated Research Database (HIRD) over 8 years. METHODS: We conducted a secondary analysis of a cohort of advanced cancer patients with linked National Death Index (NDI) data identified in the HIRD between 2010 and 2018. We calculated sensitivity, specificity, positive predictive value, and negative predictive value for six sources of death status data and an algorithm combining data from available sources using NDI data as the reference standard. Measures were calculated for each year of the study including all members in the cohort for at least 1 day in that year. RESULTS: We identified 27 396 deaths from any source among 40 692 cohort members. Between 2010 and 2018, the sensitivity of the Death Master File (DMF) decreased from 0.77 (95% CI = 0.76, 0.79) to 0.12 (95% CI = 0.11, 0.14). In contrast, the sensitivity of online obituary data increased from 0.43 (95% CI = 0.41, 0.45) in 2012 to 0.71 (95% CI = 0.68, 0.73) in 2018. The sensitivity of the composite algorithm remained above 0.83 throughout the study period. PPV was observed to be high from 2010 to 2016 and decrease thereafter for all sources. Specificity and NPV remained at high levels throughout the study. CONCLUSIONS: We observed that the sensitivity of mortality data sources compared with the NDI could change substantially between 2010 and 2018. Other validation characteristics were less variable. Combining multiple sources of mortality data may be necessary to achieve adequate performance particularly for multiyear studies.
Subject(s)
Databases, Factual , Humans , Algorithms , Neoplasms/mortality , Cohort Studies , Sensitivity and Specificity , Cause of Death , Male , Female , Reproducibility of Results , AgedABSTRACT
OBJECTIVE: Patient-powered research networks (PPRNs) are a valuable source of patient-generated information. Diagnosis code-based algorithms developed by PPRNs can be used to query health plans' claims data to identify patients for research opportunities. Our objective was to implement privacy-preserving record linkage processes between PPRN members' and health plan enrollees' data, compare linked and nonlinked members, and measure disease-specific confirmation rates for specific health conditions. MATERIALS AND METHODS: This descriptive study identified overlapping members from 4 PPRN registries and 14 health plans. Our methods for the anonymous linkage of overlapping members used secure Health Insurance Portability and Accountability Act-compliant, 1-way, cryptographic hash functions. Self-reported diagnoses by PPRN members were compared with claims-based computable phenotypes to calculate confirmation rates across varying durations of health plan coverage. RESULTS: Data for 21 616 PPRN members were hashed. Of these, 4487 (21%) members were linked, regardless of any expected overlap with the health plans. Linked members were more likely to be female and younger than nonlinked members were. Irrespective of duration of enrollment, the confirmation rates for the breast or ovarian cancer, rheumatoid or psoriatic arthritis or psoriasis, multiple sclerosis, or vasculitis PPRNs were 72%, 50%, 75%, and 67%, increasing to 91%, 67%, 93%, and 80%, respectively, for members with ≥5 years of continuous health plan enrollment. CONCLUSIONS: This study demonstrated that PPRN membership and health plan data can be successfully linked using privacy-preserving record linkage methodology, and used to confirm self-reported diagnosis. Identifying and confirming self-reported diagnosis of members can expedite patient selection for research opportunities, shorten study recruitment timelines, and optimize costs.