Item Response Patterns on the Patient Health Questionnaire-8 in a Nationally Representative Sample of US Adults

Abstract

Recent studies have shown that item responses on the Center for Epidemiologic Studies Depression Scale (CES-D) and Kessler Screening Scale for Psychological Distress (K6) exhibit the same characteristic item response patterns among the general population. However, the distributional patterns of responses on the Patient Health Questionnaire-8 (PHQ-8) among the general population have not been adequately studied. Thus, we conducted a pattern analysis of PHQ-8 item responses among US adults. Data (18,446 individuals) were obtained from the 2015 Behavioral Risk Factor Surveillance Survey (BRFSS). Item responses on the BRFSS version of the PHQ-8 were scored using the number of days response set and then converted to the original 4-point scale. The patterns of item responses were analyzed through graphical analysis. Lines of item responses scored using the number of days response set showed the same pattern among the eight items, characterized by crossing at a single point between “0 days” and “1 day,” and parallel fluctuation from “1 day” to “14 days” on a semi-logarithmic scale. Lines of item responses converted to the 4-point scale also showed the same characteristic pattern among the eight items. The present results demonstrate that the item responses on the PHQ-8 show the same characteristic patterns among items, consistent with the CES-D and the K6.

Distribution of item responses and total item scores for the Center for Epidemiologic Studies Depression Scale (CES-D): Data from the Irish Longitudinal Study on Ageing (TILDA)

“…As reported previously, the degree of parallelism of these 16 lines reflects how decreasing ratios of “occasionally” to “some of the time” and of “all of the time” to “occasionally” are similar among the 16 depressive item scores after logarithmic transformation [ 8 ]. Because log-normal scales represent these decreasing ratios after logarithmic transformation, small differences in the ratios between two consecutive options do not have a great effect on the parallelism of the 16 lines [ 10 ]. Unlike the lines between “some of the time” and “occasionally,” some of the lines (“insomnia,” “failure,” and “dislike”) between “occasionally” and “all of the time” were further from parallel ( Fig 4B ).…”

“…Of note, previous studies have demonstrated that the ratios between two consecutive response options were similar among all depressive symptom items, with the exception of response options at the lower end of the curve [ 8 , 10 , 12 ]. In addition, these similar ratios between two consecutive response options have been shown to result in the same characteristic pattern of item responses described above: response-curve lines cross at a single point between the option at the lower end and the adjacent option, with a parallel pattern across the remaining options on a log-normal scale.…”

“…The characteristic pattern of item responses has been replicated in previous studies using other depression screening scales. Analyses of Patient Health Questionnaire-8 (PHQ-8) data from the Behavioral Risk Factor Surveillance Survey and PHQ-9 data from the National Health and Nutrition Examination Survey (NHANES) in the United States have also revealed this characteristic pattern [ 10 , 11 ]. Moreover, analyses of four subsamples from the Nationally Survey of Midlife Development in the United States (MIDUS) confirmed that item responses to the Kessler Screening Scale for Psychological Distress (K6) followed this same characteristic pattern in a population of US adults [ 12 ].…”

Age-Related Changes in Item Responses to the Patient Health Questionnaire-9: Evidence From the National Health and Nutrition Examination Survey

“…Specifically, lines of item responses crossed at a single point between “not at all” and “several days” and showed a parallel pattern for the remaining response options on a logarithmic scale ( 20 ). This common characteristic pattern has been replicated for PHQ-8 data from the Behavioral Risk Factor Surveillance Survey in the USA ( 21 ), data for the Center for Epidemiologic Studies Depression Scale (CES-D) from the Irish Longitudinal Study on Ageing ( 22 ) and the Japanese Active Survey of Health and Welfare ( 19 ), and data for the six-item Kessler Psychological Distress Scale (K6) from the National Survey of Midlife Development in the USA ( 23 ).…”

“…In the general population, item responses in self-rated depression scales have recently been reported to exhibit a characteristic distribution ( 19 – 21 ). In an analysis of PHQ-9 data from the NHANES in the USA, we found that item responses for the PHQ-9 exhibited a common pattern across all nine items ( 20 ).…”