CLASSIFYING PATIENTS ON THE PROBLEM SOLVING DECISION MAKING SCALE WITH LATENT CLASS ANALYSIS

Sunday, October 19, 2014
Poster Board # PS1-43

Osvaldo F. Morera, PhD1, Luis Rangel, BS2, Adon Neria, BS2, Roger E. Millsap, PhD3 and James G. Dolan, MD4, (1)University of Texas at El Paso, El Paso, TX, (2)Department of Psychology, El Paso, TX, (3)Arizona State University, Tempe, AZ, (4)University of Rochester, Rochester, NY
Purpose: The Problem Solving Decision Making (PSDM) scale measures a patient’s desire to participate in their medical decision making.  Patients are classified based on their mean scores on the Problem Solving (PS) and Decision Making (DM) subscales.  An individual is said to be “passive” if their mean scores on the PS and DM subscales are both less than 3.  A decision maker is said to be “autonomous” if their mean score on the PS exceeds 4 and if their mean score on the DM subscale equals 3 or higher.  Finally, a decision maker is said to be a “shared” decision maker if their mean score on the PS subscale does not exceed 4 and their mean score on the DM subscale exceeds 3.  The purpose of this study was to use latent class analysis to identify the underlying number latent classes to identify patients.

Method: Patients were recruited in the lobby of the Texas Tech University Health Science Center (TTUHSC) - Northeast Family Practice Center in El Paso, Texas, through direct person-to-person solicitation. A total of 141 patients were recruited with 85 females in the sample. All participants identified themselves as Hispanic and 72.3% stated that Spanish was their primary language. Participants’ age ranged from 50 to 88 years old with a mean of 63 years, 87.2% had an income lower than $20,000 a year. Participants completed the morbidity vignette (urinary tract infection) and the mortality vignette (chest pain) of the PSDM.

Result: A latent class analysis was conducted on each vignette, where a 2 class, 3 class and 4 class model was estimated.  Using the Bayes Information Criteria, the 4 class model was the best model for both vignettes.  In the morbidity vignette, two of the four latent classes represented various levels of “passive” decision making, accounting for almost 50% of the sample. In the mortality vignette, three of the four latent classes represented varying levels of “passive” decision making, accounting for 57% of the sample.

Conclusion: Latent class analysis is a powerful technique that can identify subgroupings in a non-arbitrary fashion.  While the majority of this Hispanic sample would be classified as “passive”, these results suggest that there are varying degrees of “passive” decision making that is not captured by the PSDM classifications.