HCEI Scale Development
The DSS Health Care Engagement Scale was developed using the Domain Sampling Model1:
- Specify domain of construct. Using past research, literature search and expert opinion, the primary and secondary domains of health care engagement were defined. The theoretical model of the engagement process includes four major sub-domains or constructs: Literacy, Knowledge, Attitudes and Behaviors. The theoretical model outlines a clear hierarchy between these four constructs. Health care literacy precedes knowledge, knowledge affects health care attitudes, and attitudes impact health care behaviors over time.
- Generate sample of items. Over 100 new and previously tested items expected to be associated with these constructs were generated for testing.
- Collect data. The sample items, along with cost data, plan information and respondent characteristics were organized into a questionnaire. The questionnaire was administered to a simple random sample of U. S. adults 18 and older.
- Purify measure. Cronbach’s Alpha, Guttman Split-Half Coefficient, Intraclass correlation and factor analysis were employed to test the consistency of the items in each sub-domain and the strength of their correlations. The lowest performing items on each sub-domain were removed and the statistical tests were rerun until a stable set of items remained that met the minimum criterion for inclusion.
- Collect data. A second wave of data collection was conducted using the revised scale items and a similar random sampling approach.
- Assess reliability. The same statistical measures used to purify the initial survey items were rerun to insure the consistency of the items in the second wave of data collection. Split-half reliability tests were run between the first and second wave of data to further test the reliability of these measures.
- Access validity. The hypotheses used to identify and define the initial constructs were validated by comparing correlations between the index and self-stated ratings of engagement, health care involvement, health insurance value, as well as, hypothesized correlations with characteristics such as education level, health plan selection and health status. Furthermore, the engagement index survey was administered to a targeted group of health plan members so that claims data could be appended to results for all survey respondents. The claims data validated the overall construct of engagement as it relates to expenditures on preventive care and adherence to recommended care and treatment amongst those chronic conditions.
- Final index. A weighted average of the four constructs is used to calculate the final DSS Health Care engagement Index. Actual positive behaviors are assigned more weight than attitudes, knowledge or health literacy. The DSS Health Care Engagement Index contains 41 response items in its detailed form and 16 in its reduced form (R squared of 0.91 predicting detailed form score).
- Develop norms. Norms have been developed for the DSS Health Care Engagement Index by conducting multiple waves of data collection with large samples in each wave. The index is scaled from 0 to 100 for easy interpretation, with a normal distribution and an initial mean around 50 through the first few waves of data collection.
- Categorizing the index. The DSS Health Care Engagement Index was subdivided into various contiguous groups to determine the optimal number of categories and the break-points for each category or segment. Analyzing the distribution of the index and employing ANOVA and CHAID, the optimal break-points were determined to be around the 30th, 60th and 90th percentiles. These boundaries were adjusted slightly to produce four categories based on: 0 – 39, 40 – 59, 60 – 79 and 80 – 100 points on the DSS Health Care Engagement Index.
1 Gilbert A. Churchill, Jr., “A Paradigm for Developing Better Measures for Marketing Constructs,” Journal of Marketing Research, 16:1. (Feb., 1979), pp. 64-73.