Association of Shared Physician Training with Specialty Referral Networks and Patient Outcomes

Abstract

Research ObjectiveSpecialty referrals are a key determinant of health care quality and a quintessential manifestation of physicians' agency on behalf of patients based on their unique knowledge as professionals and their intrinsic motivation to help. Yet we know surprisingly little about why physicians refer to the specialists they choose. Ideally, referrals by primary care physicians (PCPs) would be fully informed and made according to the expected benefits to patients of specialist attributes. However, referrals may also be influenced by professional or personal relationships among physicians, the implications of which are unclear. On one hand, such relationships may lead physicians to disproportionately refer to whom they know instead of who is best. On the other, PCPs may accumulate more information about the quality of specialists with whom they are more familiar, allowing them to better match their patients to specialists. With their actions subject to observation by friends or respected peers, referring PCPs may transmit more useful information and specialists may devote more effort to the patient's care. We leverage detailed EHR data to investigate whether training together influences referral decisions and the implications for patient care. In this abstract, we focus on preliminary analyses of the relationship between co‐training and referral preferences.Study DesignUsing the universe of EHR data from a large Boston‐area academic health system, we identified all referrals to five specialties (orthopedic surgery, general surgery, obstetrics, urology, and cardiology) initiated by system‐affiliated PCPs in 2018 and 2019. For each referral, we determined whether PCPs and specialists trained at the same medical school or post‐graduate institution at the same time using medical licensing data. We characterized how referral decisions and associated outcomes differed between PCP‐specialist pairs that trained together compared to pairs that did not.Population Studied40,495 specialty referrals of 37,672 patients to 1273 specialists they had not seen before, initiated by 4890 primary care physicians in 2018 and 2019. Of these, 13,063 referrals were to orthopedic surgery, 8788 to general surgery, 3728 to obstetrics, 4205 to urology, and 10,846 to cardiology.Principal FindingsOverall, 22.5% (n = 9095) of referrals could have potentially been directed to co‐trainees. However, only 8.0% (n = 724) of these potential referrals (or 1.8% of all referrals) were directed to co‐trainees, ranging from 7.0% (n = 188) of general surgery referrals to 16.0% (n = 47) of obstetrics referrals. Among PCPs who co‐trained with specialists, a higher proportion of their referrals, on average, went to specialists they trained with compared to specialists they did not train with (26.2% vs. 21.4%, p < 0.001). This effect was almost entirely driven by shared post‐graduate (27.0% vs. 21.4%, p < 0.001) versus shared medical school training (21.6% vs. 21.4%, p = 0.965).ConclusionsCo‐training among physicians influenced referral preferences with implications for care delivery networks.Implications for Policy or PracticeOur findings suggest that PCPs had private information about co‐trainee attributes. This has implications for delivery system design. If outcomes of co‐trainees are improved, low‐cost interventions that foster professional interactions among physicians may improve the value of care. If outcomes are worse, EHRs could suggest alternative, peer‐preferred specialists at the time of referral.Primary Funding SourceNational Institutes of Health.