Developing methods to link patient records across data sets that preserve patient privacy
We also conducted structured interviews with PPRN representatives to gauge their interest in health plan member outreach and to assess patient understanding of the need for data linkage with health plans to close gaps in evidence for the PPRN disease conditions of interest. METHODS: Aim 1 developed...
| Main Author: | |
|---|---|
| Corporate Author: | |
| Format: | eBook |
| Language: | English |
| Published: |
[Washington, D.C.]
Patient-Centered Outcomes Research Institute (PCORI)
[2020], 2020
|
| Series: | Final research report
|
| Online Access: | |
| Collection: | National Center for Biotechnology Information - Collection details see MPG.ReNa |
| Summary: | We also conducted structured interviews with PPRN representatives to gauge their interest in health plan member outreach and to assess patient understanding of the need for data linkage with health plans to close gaps in evidence for the PPRN disease conditions of interest. METHODS: Aim 1 developed PPRL methods for the anonymous linkage of overlapping members using secure HIPAA-compliant, 1-way, cryptographic hash functions. A cryptographic hash function is a mathematical algorithm that converts a string of text into an irreversible hash text string. For the linked data from the 4 PPRNs--the American BRCA Outcomes and Utilization of Testing PPRN, Arthritis Patient Partnership With Comparative Effectiveness Researchers, Multiple Sclerosis (MS) PPRN, and Vasculitis PPRN--we compared self-reported diagnoses by PPRN members with claims-based computable phenotypes to validate agreement between the 2 data sources on the knowledge of disease status. We queried the resulting overlapping data to identify and validate claims-based computable phenotypes and tested different outreach approaches to engage health plan members in patient-centered registry participation. Our research advanced the methodology standards for data linkage within a distributed research network (DRN) as well as the PCORI methodology domains Describe data linkage plans and Requirements for the design and features of data networks. We developed methods that improved the capacity of DRNs by ensuring the appropriate privacy and confidentiality of network participants. In addition, we engaged a payer-stakeholder data repository, the HealthCore Integrated Research Environment, to conduct population-based patient identification of potential research participants. OBJECTIVE: This research aimed to develop and test linkage and validation methods to identify potential participants for PCOR opportunities. We assessed the feasibility of linking and using PPRN membership and health plan data to confirm self-reported diagnosis (aim 1) and to contact health plan members to participate in patient-centered registries (aim 2). Objectives in aim 1 were to test PPRL processes between PPRN members' and health plan enrollees' data and to measure disease-specific confirmation rates, a validation measure of health plan administrative data on patient-reported disease status, as indicated based on participation in a disease-specific patient-centered registry, for particular health conditions. Aim 2 sought to quantify health plan members' registration rates in any of 4 disease-specific PPRNs following 2 common payer-initiated outreach methods for inviting member participation: mail and email. The confirmation rates for breast or ovarian cancer, rheumatoid or psoriatic arthritis or psoriasis, MS, and vasculitis PPRNs were 72%, 50%, 75%, and 67%, respectively, which increased to 91%, 67%, 93%, and 80%, respectively, when limiting the cohort to those with continuous health plan enrollment ≥5 years. AIM 2: A total of 14 571 members were randomly assigned to each outreach method (email or regular mail). Invitations were sent to 13 834 (94.9%) mail group members and 10 205 (70.0%) email group members. A significantly larger proportion of the mail group (n = 78; 0.54%; 95% CI, 0.42%-0.67%) registered in PPRNs relative to the email group (n = 24; 0.16%; 95% CI, 0.11%-0.25%; P < 0.001). Registrants had more comorbidities and greater medical system use, especially emergency department visits (52.0% vs 42.5%; P = 0.053), compared with nonregistrants. PPRN STRUCTURED INTERVIEWS: We conducted 9 structured interviews with PPRN members to discuss health plan outreach and data linkage. BACKGROUND: Patient-centered outcomes research (PCOR) relies on access to researchable health care data for a broad spectrum of patients. Payer-stakeholders, such as Anthem, can use longitudinal patient-level claims data from their general membership to identify patients for engagement in PCOR opportunities. We hypothesized that the contributions of payer-stakeholder member engagement to PCOR initiatives could expand patient-centered registry participation as evaluated through linkage with health plan data. To design and test a model for improving PCOR engagement and data integration between patient-centered registry self-reported data and health plan administrative claims data, we developed privacy-preserving record linkage (PPRL) methods to determine overlapping membership in the National Patient-Centered Clinical Research Network (PCORnet), patient-powered research networks (PPRNs), and a health plan research network (HPRN). This qualitative approach explored perceptions about HPRN collaboration with PPRNs, HPRN outreach, data linkage, and patient privacy and sought to identify opportunities for HPRNs to engage patients. CONCLUSIONS: We demonstrated the feasibility of linking and using PPRN membership and health plan data to confirm self-reported diagnosis. Our outreach initiative to invite health plan members to register in PPRNs was modestly effective. Responses to regular mail slightly out-performed email outreach. Our structured interviews identified opportunities for HPRNs to engage in PCOR. LIMITATIONS: Our data were limited to a commercially insured population. We were constrained in overlap identification by the accuracy of recorded PPRN member-level identifiers. In addition, patient views reflected the perceptions of a population already heavily engaged in PPRN research Aim 2 identified enrolled members who met the diagnostic criteria for computable phenotypes but were not registered in a PPRN. We then randomly assigned members to 2 groups (outreach by mail or email) and quantified new PPRN registrants by employing PPRL methods. In a separate study, we conducted structured interviews with representatives of PPRNs to understand their interest in and willingness to participate in relevant research sponsored by health plans. RESULTS: AIM 1: Data identifiers for 21 616 PPRN members were converted through PPRL into hashed identifiers, and 4487 (21%) were linked to health plan data. Of the linked cohort, 3548 (16%) PPRN members were commercially insured, with health plan membership eligible for inclusion in the aim 1 analysis. Irrespective of enrollment duration, confirmation rates (the agreement between patient self-reported disease status as indicated through PPRN membership and health plan administrative record of disease status) were determined. |
|---|---|
| Physical Description: | 1 PDF file (78 pages) illustrations |