APPLICATION OF ITEM RESPONSE THEORY IN VALIDATING THE PATIENT HEALTH QUESTIONNARE (PHQ-9) IN THE AMERICAN GENERAL POPULATION
Author(s)
Gupta S, Witt EA
Kantar Health, Princeton, NJ, USA
Presentation Documents
OBJECTIVES: The Patient Health Questionnaire (PHQ-9) is a multipurpose instrument for screening, diagnosing and measuring the severity of depression. Item Response Theory (IRT) methods are an increasingly popular, sophisticated means of validation in health outcomes research. The current study examined psychometric properties of the nine-item PHQ-9 using IRT. METHODS: Data were used from the U.S. 2013 National Health and Wellness Survey (NHWS; N=75,000), a web-based survey of the general population which collects data on patients’ demographics, health care attitudes and health outcomes. All 75,000 respondents answered the PHQ-9 instrument and were included in this study. Reliability and validity of the PHQ-9 were examined using both Classic Test Theory (Cronbach’s alpha) and IRT (graded response model) analyses. RESULTS: Among 75,000 respondents, 51.6% were female, mean age was 48.3 (SD = 16.7). Severity of depression varied across respondents; minimal (66.6%), mild (18.4%), moderate (8.0%), moderately severe (4.5%), severe (2.5%). The PHQ-9 was highly reliable (Cronbach’s alpha=0.92); with the removal of any item providing less reliability (range: 0.63-0.81). The nine items varied in their discrimination (range: 1.8–3.6), with item “Trouble falling or staying asleep, or sleeping too much” showing the weakest discrimination. Item location parameters reflected a large, high range of depression severity (range: -0.17–3.0). Item “Feeling down, depressed, or hopeless “provided the highest level of item information and item “Moving or speaking so slowly that other people could have noticed? Or the opposite — being so fidgety or restless that you have been moving around a lot more than usual” provided good item information at high levels of depression. CONCLUSIONS: The PHQ-9 has adequate psychometric properties for evaluating depression in the general population. Each item provided a good and unique contribution to the overall instrument. IRT is a helpful means for evaluating the reliability and validity of the PHQ-9.
Conference/Value in Health Info
2016-05, ISPOR 2016, Washington DC, USA
Value in Health, Vol. 19, No. 3 (May 2016)
Code
PRM116
Topic
Methodological & Statistical Research
Topic Subcategory
PRO & Related Methods
Disease
Mental Health