Patient Activation: Prediction of Behavioral Outcomes During a Controlled Telemonitoring Intervention

Abstract

Patients’ skills, knowledge, and motivation to actively engage in their healthcare are assessed with the Patient Activation Measure (PAM) – a metric associated with positive healthcare outcomes. The literature on predicting PAM, when patient counseling is coupled with a telemonitoring intervention, is scant. This proof-of-concept study employs a two-phase framework to (i) posit a causal relationship between a telemonitoring intervention and enhanced patient activation; and (ii) link healthcare providers’ operating decisions and patients’ willingness to change with prediction of PAM. We test the framework using data from a randomized, controlled field experiment and find a causal relationship between telemonitoring and increased PAM. Based on this causal result the PAM levels are predicted as a function of the strength of the information signals using a machine learning methodology. We show that these predictions are subject to under-over-estimation biases, consistent with the behavioral concept of system neglect in signal detection theory.