Categorizing the severity of pain using questions from the 2012 National Health Interview Survey

BACKGROUND: Based on qualitative and mixed-method approaches, Miller and Loeb have proposed a coding system that combines questions on pain persistence and bothersomeness to create discrete categories of increasing pain severity for use in large population-based surveys. In the current analyses, using data from the 2012 National Health Interview Survey, we quantitatively assess the pain category definitions proposed by Miller and Loeb and compare this original definition to ten alternative definitions. METHODS: Using multivariate analysis of variance, each definition was related simultaneously to four dependent measures – the Kessler 6 score for measuring psychological distress, the number of health-related bed-disability days, the number of visits to a health professional, and the number of emergency room visits. Following the protocol of Serlin et al, the definition yielding the largest F score was considered the optimal definition. RESULTS: The Miller and Loeb definition produced the largest F value (185.87), followed consecutively by several alternative definitions #5 (184.17), #10 (180.95), and #9 (179.5). A nearly identical ordering was found when looking at the mean F value generated from 100 random samples. We also examined the frequencies with which each alternative definition achieved the optimal F value over the 100 random samples. Only two definitions had achieved the optimal F value >5% of the time: the Miller and Loeb definition was optimal 46% of the time, while alternative definition #5 was optimal 41% of the time. Similar results were seen in subpopulations with back pain and joint pain. CONCLUSION: Additional support was provided for the Miller and Loeb coding of pain persistence and bothersomeness to produce discrete categories of increasing pain severity. This two-question coding scheme may prove to be a viable option for assessing pain severity in clinical settings where clinician time and patient burden are limiting factors.