Feasibility and Validity of Using Item Response Theory to Assess Cumulative Social Risk

Social risks are patient-level effects of social determinants of health. Cumulative social risk, the total of multiple risks, predicts health outcomes such as cancer mortality. Current measures of cumulative social risk count the number of domains in which a patient has risks. We tested whether an item response theory (IRT) model could be an alternative method for measuring cumulative social risk. A population-based sample (n = 1118) was recruited from 13 counties of western Washington State. Participants completed a one-time survey online. Questions on social risk included two on housing instability, two on food insecurity, one on affording medical care and one on general financial health. Additional measures to test the construct validity of the cumulative social risk score included perceived discrimination, perceived stress, smoking status and having a usual place for healthcare. The IRT model fit the six social risk items. Social risk items ranged in severity from just above the mean (0.18 standard deviations) to substantially above the mean (4.18 standard deviations). IRT scores were positively and significantly associated with perceived discrimination (p < 0.001), perceived stress (p < 0.001), being a current smoker vs. a never (p < 0.001) or former smoker (p = 0.006) and not having a usual place for healthcare vs. having a usual place of care (p = 0.033). Cumulative social risk can be measured and assessed using IRT models. Weighted IRT cumulative social risk scores were associated with multiple health markers, indicating promise for this type of measure. Future research is needed to compare IRT-based and count-based cumulative social risk scores.