Abstract
Background/Aims We evaluated the effect of online messaging between patients and providers on other healthcare utilization. Messaging may increase the administrative burden on providers but may also substitute for utilization such as calls or visits. To identify the effect of messaging, we exploited a natural experiment, which occurred in 3/2011, when a large multispecialty practice in California lowered the price of messaging from $60/year to free for patients and initiated incentives for doctors ($3/thread), resulting in increased rates of patients messaging.
Methods We conducted a longitudinal study of 65,332 active primary care patients from 3/2009–3/2012. All patients in the study had activated their online portal before 3/2011 and none had paid to use messaging previously; 38,438 of these patients initiated a message following 3/2011. Messaging was measured by the number of threads initiated by patients annually. Outcome measures included annual rates of office visits and phone calls to primary and specialty care. The effects of messaging were identified through difference-indifferences analyses using generalized least squares, comparing the utilization of those initiating any messages with those not messaging, before and after messaging became free. Controls included patient fixed effects along with patient age, insurance type, Charlson score, continuity of care, and relevant characteristics of their primary care provider.
Results Comparing the rates of service use before and after 3/2011, between messaging and non- messaging users, we found no difference in the annual number of primary care phone calls (0.002 per patient-year, P = 0.394), and a small increase in primary care office visits (0.05 per patient-year, P <0.001) for each thread initiated. For specialty care, we found an increase in 0.16 phone calls (P <0.001) and 0.21 (P <0.001) office visits per patient-year for each thread initiated.
Conclusions Among patients who were already using the online portal, use of secure messaging was associated with little difference in primary care utilization, but with a moderate increase in specialty utilization. Although current analyses controlled for patient and provider characteristics, patient self-selection may still threaten validity; we will address this through propensity score stratification in future analyses.
