The Bystander Effect in Medical Consult Provision: A Game Theoretical and Empirical Analysis

The complex interactions of generalist physicians and specialists are of interest to health services researchers and economists because these physicians effectively determine patterns of care and resource use and because of unclear expectations for each group’s role in care delivery in many contexts.  In the setting of perioperative care for complex surgical inpatients—where the costs of physician visits are rising rapidly (MedPAC 2013)—such uncertainty can lead directly to adverse patient outcomes.  Typically, specialists render consults per the attending physician’s request and then decide whether or not to render additional consults before “signing off.”  For particularly complex patients with greater needs for care management, it may be optimal for one or more consulting physicians to render additional consults and provide this care management support.  Owing to a dearth of work identifying factors that affect physician decisions to render consults and the substantial variation in consult use (Stevens et al., 2013; Wijeysundera et al., 2012), it is not known whether these patients’ needs for effective care management are being met.

In this paper, I apply and extend a well-known game theoretical framework—Osborne’s (2000) “Reporting a Crime,” which examined whether larger groups of bystanders intervene less—to generate several relevant and empirically testable hypotheses.  My principal hypothesis indicates (perhaps counter-intuitively) that the probability that one or more consulting physicians render multiple consults—theoretically, to aid in the patient’s care management—will decrease as the number of the patient’s consulting physicians increases.  This prediction runs counter to the health status-based argument that sicker patients may require more consulting physicians and motivate them to provide additional consults.  Moreover, my model suggests that this unexpected effect should be stronger (i) for more complex patients, (ii) in hospitals with less consistent patterns of consult use historically, and (iii) when the collection of consulting physicians has comparable and high marginal costs of providing consults.  I explore these hypotheses using multiple alternative frameworks: for example, I characterize consulting physician’s marginal costs separately in terms of the opportunity costs of high fees for other services or academic physicians’ research efforts and the risks of violating institutional and professional norms.

My empirical tests of these hypotheses—relying on two large administrative claims data samples for Medicare beneficiaries undergoing either coronary artery bypass graft or colectomy procedures between 2007 and 2010 (319,000 and 98,000 patients, respectively)—are ongoing.  I use patient-level and patient-physician dyad-level logit models to estimate the marginal effect of an extra consulting physician involved in the patient’s care on the probability of at least one consulting physician providing more than one consult (yes / no).  To clarify whether extra consults may be motivated by the patient’s care management needs, I retest these effects in the context of a subset of relatively uncomplicated patients, for whom this need is less significant.  Associations between my outcome of interest and patient health outcomes—30-day mortality and select post-operative complications—are also assessed.

These analyses will inform hospital and medical staff decisions regarding optimal care organization and delivery structures in the treatment of vulnerable patients undergoing major surgical procedures.