Development and testing of the diabetes self -management instrument.

Abstract

<italic>Purpose</italic>. The purpose of this study was to develop and psychometrically test a new instrument for measuring self-management behaviors among Taiwanese adults with type 2 diabetes. <italic>Method</italic>. The items generated for the instrument were drawn from a comprehensive review on existing instruments related to diabetes management and a previous focus group study conducted by the author in Taiwan. The instrument content was validated by a panel of experts. The diabetes self-management instrument (DSMI) was pilot tested with a sample of 97 Taiwanese adults with type 2 diabetes. To examine the psychometric properties of DSMI, data were collected from 634 adults with type 2 diabetes recruited from three teaching hospitals in Taiwan. Construct validity was determined by three methods: (1) factor analysis; (2) correlation with existing measures (i.e., Summary of Diabetes Self-Care Activities, SDSCA); and (3) hypothesis testing with known groups. Criterion-related validity was examined by correlating scores on the DSMI with a related criterion measure (i.e., A1C). The internal consistency and two-week interval retest were used to assess the reliability of the DSMI. <italic>Results</italic>. Five factors were extracted that explained 45.66% of the total variance: Self-Integration, Self-Regulation, Interacting With Health Care Provider and Significant Others, Self-Monitoring Blood Glucose, and Adherence to Recommended Regimen. Initial results of confirmatory factor analysis (CFA) did not provide full support for the proposed 5-factor model. After the model was modified by correlating four pairs of error terms and adding one indicator cross loading on another factor, the fit indices indicated that this modified correlated model fit the data best. This modified model was further cross validated in Sample 2. Divergent validity was supported. There was a moderate positive significant correlation (<italic>r</italic> = .545, <italic>p</italic> < .001) between the DSMI and SDSCA. Known-groups validity was supported. There was a significant difference (<italic>t</italic> = 15.213, <italic>p</italic> < .001) between a group of high empowerment and low empowerment and DSMI score. Criterion-related validity was also supported. Alc was negatively related to the total DSMI when controlling for the presence of co-morbidities (<italic>r</italic> = -.155, <italic> p</italic> < .001). Cronbach's alpha coefficient of the DSMI total scale was .94 and of subscales ranged from .77 to .90. The test-retest correlations for the DSMI total scale (<italic>r</italic> = 73, <italic>p</italic> < .01) and for the subscales (<italic>r</italic> = .59--.86, <italic>p </italic> < .01) were acceptable. <italic>Conclusions</italic>. Results supported the reliability and validity of the DSMI developed in this study, suggesting that this 35-item instrument can be used for measuring the self-management behaviors of adults with type 2 diabetes.