Keywords

artificial intelligence, Bayesian networks, diary, lifestyle management, machine learning, personal improvement, relapse prevention

 

Authors

  1. Alemi, Farrokh PhD
  2. Moore, Shirley PhD
  3. Baghi, Heibatollah PhD

Abstract

Patients give many reasons for why they have not kept up with their resolutions; research shows that many of these causal attributions are wrong. This article provides a tool to help patients sort out causes of and constraints on their behavior, in general, and exercise, in particular. Patient's diary data can be analyzed to flag erroneous causal attributions, and thus assist patients to understand their behavior. To start the diary, the clinician works with the patient to assemble a list of possible causes. Using the list, a diary is organized that tracks the occurrences of various causes and the target behavior. At the end of 2 to 3 weeks, the diary data is analyzed using conditional probability models, causal Bayesian networks or logistic regression. A key issue in the analysis of diary data is to separate out the effect of various causes. Typically, causes co-occur, making it difficult to understand their independent effects. Another problem with analysis of diary data is the small size of the data. This article shows how small longitudinal data from patient diaries can be analyzed. The analysis may refute or support causes hypothesized by the client. The patient uses the insights gained from the diary analysis to prevent relapse to unhealthy behaviors. The process is continued for several cycles of organizing, keeping, and analyzing the diary data. In each cycle, the patient gains new insights and makes additional attempts to create a positive environment that allows him or her to succeed even if his or her motivation waivers. This article provides details of how diary data can be analyzed to help patients make correct causal attributions.