reports

An economic analysis of secure messaging between patients and providers

Seattle, WA, USA: University of Washington (doctoral dissertation).

University of Washington, Seattle, WA, USA

This dissertation presents an economic analysis of patient-provider communication through secure messaging. Section 1 presents the evolution of secure messaging. Section 2 aims to determine whether secure messages can substitute for face-to-face office visits. To answer this question, two demand equations were estimated and outpatient copayment was used as a key factor. A quasi-conditional maximum likelihood estimator was used as it is robust when there are distributional misspecifications and accounts for individual unobserved heterogeneity. Controlling for individual characteristics such as age, gender, insurance source, health status, and individual unobserved heterogeneity (fixed effects), we found that an increase by $1 in the price of office visits, measured by outpatient copayment, leads to 0.45% fewer office visits and 4.4% more secure messages. This is a strong indication that secure messages and office visits are substitutes. Section 3 aims to determine whether secure messaging reduces provider costs. To answer this question, a cost minimizing model is estimated. An instrumental variable approach was used to correct for the potential correlation of health and secure messaging user measures with the error term. A potential problem of unobserved individual heterogeneity was also taken into account. While we did not find evidence that secure messaging users impact the primary, specialty, inpatient, and emergency care costs, we did find evidence that individual costs on drugs are reduced by 66.4% per quarter. A lack of evidence on the impact of secure messaging on most costs may be due to limitations in the way costs are measured. Finally, to overcome a limitation of the quasi estimator used in Section 2, Section 4 seeks to introduce a new estimator ZI QCMLE aimed at estimating count and continuous models with an excess number of zeros and unobserved individual heterogeneity. This approach is similar to a zero inflated model in the way it deals with excess zeros, but unlike a zero inflated model it is robust to distributional misspecifications. Also, its ability to deal with fixed effects makes this estimator unique.