S. A. Castellanos1, G. Buentello1, J. W. Suliburk1 1Baylor College Of Medicine,Houston, TX, USA
Introduction: Patient engagement is frequently discussed as a goal of patient education/discharge planning, yet remains hard to model and therefore challenging to optimize. Game theory is a field of study in which different models (games) analyze a player’s decision making process when his/her decision is contingent on what another player is going to do. This study seeks to develop a model to characterize patient decision making in engagement and then sought to correlate the model with qualitative analysis of semi-structured interview transcripts.
Methods: Over a 6-month period, interviews were conducted within 6 weeks of discharge in patients undergoing thyroid, parathyroid or colorectal surgery. Interviews were recorded, transcribed, anonymized and then analyzed using Nvivo software platform. Blinded to transcript coding and results, a signaling game model was developed as follows: two players—Doctor (D) and Patient (P)—and two scenarios—one in which P is “engaged” [probability, α] and another in which P is “unengaged” [probability, (1-α)]. “Engaged” P’s represent patients who will call their doctor with problems at home post-hospitalization, and “unengaged” P’s are patients who will call no one or seek care elsewhere. Transcripts were reviewed for “Discharge Instructions,” “Discharge Process” and “Discharge Education” themes.
Results: As the model (game) evolves only P knows the starting state (engaged vs. unengaged). The game is played anytime during P’s clinical care episode both pre- and post-operatively. P moves first by deciding to Ask (a) or Refrain (r) from questions. “Engaged” P’s prefer choosing a, but will choose r in certain situations, and the converse is true for “unengaged” P’s. In response to P’s behavior, D moves by deciding to Invest (i) resources in care for P or Maintain (m) the care at normal levels. If D chooses i, then P becomes “engaged” P. Otherwise, P will act according to baseline. Unless they believe P to be “unengaged”, D prefers choosing m over i. The optimal outcome for D and P results if P ends the game as “engaged”. Review of transcripts determined the levels of questioning exhibited by the patients only partly reflected activation towards proficiency in self-care. Across surgeries, there was poor demonstration that the clinical care team altered education efforts based on signaling from the patient.
Conclusion: A game theoretic “signaling model” is able to adequately characterize interactions between the care team and the patient. If the care team cannot perceive a patient’s engagement status via these signals, it must look for other ways to bring clarity into assessment of engagement. Further work will be done to refine the model in order to optimize strategies to facilitate patient engagement.