Utilizing administrative data for provider recruitment in qualitative research: insights from a pharmacy claims-based case study

Researchers often recruit providers through institutional affiliations, clinic leadership, and snowball sampling. However, these methods can restrict sample diversity, depend on self-reported eligibility, and introduce significant selection bias. Administrative datasets (e.g., Medicaid, commercial claims) offer a more systematic method for identifying providers, reducing reliance on peer networks. This paper demonstrates how administrative claims data can be used for recruiting healthcare providers, with this approach exemplified through a case example using New York State (NYS) Medicaid claims to identify high-volume medication for opioid use disorder (MOUD) providers for transition-age (TA) adults (18–25). NYS Medicaid pharmacy claims data from 2022 were used to identify MOUD providers to TA adults. From February to April 2024, 104 high-volume providers (treating n ≥ 5 TA adults) were identified and contacted using a multi-modal outreach strategy including phone, mail, and email. Recruitment strategies were iteratively adapted to provider responses to improve engagement. Of 104 providers contacted, 22 (21.2%) responded and 18 (17.3%) participated in qualitative interviews. Mail outreach produced a 9% recruitment rate, email 12%, and phone calls 3%. An average of 2.11 contact attempts (SD = 1.51) was made for participants recruited, with no responses beyond the fifth outreach attempt. The most significant recruitment barrier was the lack of contact information in the Medicaid dataset which required manual verification through online searches and calling clinics. This administrative data-based recruitment approach (exemplified with our Medicaid claims case study) offered multiple advantages over traditional methods: (1) systematic and quantifiable selection based on verified patient-level activity, making bias more transparent and measurable; (2) capacity to identify providers across unaffiliated institutions and regions, enabling broader and more diverse recruitment; and (3) precision in recruiting providers with specific clinical metrics, allowing access to narrowly defined or outcome-based subgroups. Limitations included the contact verification process and inherent exclusion of providers not accepting Medicaid. Our findings show the feasibility and methodological value of administrative data for the recruitment of providers in qualitative research. Although demonstrated with Medicaid claims, this framework may be adaptable to other fields and datasets, including Medicare, commercial insurance claims, and state prescription monitoring programs.