PROMIS Alcohol Use item banks: Construction and calibration

Colditz J.B., Yu L., Johnston K.L., Dodds N.E., Stover A.M., Maihoefer C., & Pilkonis P.A. (June 2011). PROMIS Alcohol Use item banks: Construction and calibration. Poster presentation in Pittsburgh, PA: Western Psychiatric Institute and Clinic 11th Annual Research Day. doi: 10.6084/m9.figshare.879704

 
Background: Many alcohol use measures tend to focus on severe symptoms associated with heavy use, and may not provide adequate information at lower thresholds of alcohol use. A National Institutes of Health Roadmap initiative, the Patient-Reported Outcomes Measurement Information System (PROMIS) develops self-report measures of physical and mental health status that are generalizable across diverse populations. The present project developed and calibrated item banks for alcohol use, utilizing Item Response Theory (IRT), to study clinical treatment and social drinker populations.

Methods: Consistent with previous PROMIS projects at the University of Pittsburgh, comprehensive categorization and revision processes guided item pool development. Patients in treatment for substance use disorders (n = 407) and social drinkers recruited from the YouGov® Polimetrix internet panel (n = 1,000) responded to the new items, extant measures of alcohol use, and PROMIS health status profiles. Exploratory and confirmatory factor analyses yielded five factors, and a unidimensional Graded Response Model was used to calibrate each of the resulting factors.

Results: 124 items performed well in calibration and contributed useful information to the five item banks, providing coverage along broad ranges of the alcohol use constructs. Wide bandwidths of information illustrated potentials for the measurement of constructs with greater inclusiveness than possible using many contemporary instruments. Both Expectancies banks demonstrated a broad range of information +/- 2-3 SD from the mean. Consumption and Negative Consequences banks showed a positive correlation with total scores on the Alcohol Use Disorders Identification Test (AUDIT), r = 0.89 and 0.78, respectively. Notable differences were found among Consumption items, with “activated drinking” items being more correlated with Negative Consequences. These banks show promise in measuring clinically significant alcohol use patterns.

Future Directions: The PROMIS Alcohol Use Item Banks may be validated in samples of individuals starting treatment for alcohol use disorders as well as social drinkers. By implementing health status profiles in conjunction with the calibrated alcohol banks, it is possible to investigate alcohol use across the lifespan and in conjunction with chronic health conditions. Utilizing these banks within an IRT framework, it becomes possible to use Computerized Adaptive Testing or pen-and-paper short forms to assess constructs in as few as 4-5 items per bank, reducing participant burden while providing versatile measures of alcohol-related constructs that may be used in epidemiological studies, clinical trials, and as potential screening tools in clinical applications.

View on FigShare