Short Staffing, Provider Preferences, and Time Allocation: Evidence from Public Health Clinics

Monday, June 11, 2018: 6:10 PM
1051 - First Floor (Rollins School of Public Health)

Presenter: Matthew Harris

Co-Authors: Yinan Liu; Ian McCarthy

Discussant: Nicolas Robert Ziebarth


In this paper, we examine how public health clinicians respond to exogenous temporary reductions in capacity (short staffing) by adjusting the number of patients seen and the average length of each visit. We incorporate insights from queueing models and construct an expository theoretical model of providers' responses to these staff reductions. For our purposes, the queuing model posits a provider's utility as a function of the number of patients seen and the amount of time spent with each patient, relative to some threshold `sufficient' visit length. We derive comparative statics showing that the optimal amount of time that providers spend with patients is a function of the relative importance of visit length versus number of patients seen and the stochastic arrival rate of patients, among other parameters.

Our empirical analysis then aims to quantify a provider's trade-off between patients seen versus time with each patient, where an exogenous reduction in clinic staff serves as a shock to a given provider's arrival rate. Our data were provided by the Knox County Health Department (KCHD) in Tennessee and are comprised of time records for each patient visit in six public health clinics over eighteen months. The causal effects of binding time constraints are identified through a series of repeated, but not periodic, exogenous temporary reductions in the number of nurses working in a given clinic on a given day.

We consider two primary specifications in our empirical analysis. First, we run panel data regressions at the clinician (nurse)-day level to examine how the exogenous reductions in staff (or, equivalently, clinic capacity) affect the number of patients a nurse sees and the time spent with patients in a day. Second, we examine the effects of capacity reductions on the time spent in each component of individual visits. Our initial analysis finds that, when capacity is reduced, there were fewer patients seen and patients' total visit time significantly decreased by at least 6\% (or about 5 minutes). This primarily occurred through a reduction in check-in and check-out times, with a small (and insignificant) reduction in time with providers. Exogenous staff reductions also lead nurses to prioritize scheduled visits over walk-in patients, suggesting a reduction in access to care among walk-in patients. In the context of our queuing model, our results suggest that nurses prioritize time with patients over number of patients seen. Additional analysis allowing for differential effects along the support of daily visit volumes shows that these results are reflective of actual provider preferences rather than any excess capacity built into each clinic. We also find that greater numbers of patients in the queue decreases the likelihood that the next patient is a walk-in, which is consistent with patients balking from the queue. While we interpret the inelasticity of time spent with patients as indicative of provider preferences, we acknowledge that our findings may be partially driven by the structural constraints of the provider-patient interaction.