Estimates of Minimally Important Differences (MIDs) for Two Patient‐Reported Outcomes Measurement Information System (PROMIS) Computer‐Adaptive Tests in Chronic Pain Patients. Issue 4 (December 2014)

• Record Type: Journal Article
• Authors: Swanholm, Eric; McDonald, Wade; Makris, Una; Noe, Carl; Gatchel, Robert
• Is Part Of: Journal of applied biobehavioral research. Volume 19:Issue 4(2014)
• Page Start: 217
• Page End: 232
• Publication Date: 2014-12
• Subjects: Clinical health psychology -- Periodicals; Psychophysiology -- Periodicals; Behaviorism (Psychology) -- Periodicals; 616.0019
• DOI: 10.1111/jabr.12026
• Languages: English
• ISSNs: 1071-2089

Abstract

<abstract abstract-type="main"> <title> <x xml:space="preserve">Abstract</x> </title> <p>Anchor‐ and distribution‐based methods were combined to evaluate and establish minimally important differences (MIDs) for two Patient‐Reported Outcomes Measurement Information System (PROMIS) measures in an outpatient chronic pain population. These included the computer‐adaptive test (CAT) versions of two PROMIS measures: Depressive Symptoms and Anxiety‐Related Symptoms (PROMIS; Cella, Gershon, Lai & Choi). Participants (<italic>n</italic> = 170) undergoing a behavioral medicine evaluation in an interdisciplinary pain management clinic completed two PROMIS CATs and multiple clinical anchor measures/questions. Modeled after similar analyses (Yost, Eton, Garcia, & Cella), three a priori criteria were used to select usable cross‐sectional anchor‐based MID estimates; these included a minimum Spearman correlation of .3 between the PROMIS measure and anchor item/categories, a minimum comparison group sample size of 10 within each anchor, and an effect size between .2 and .8 for each anchor‐based estimate. For each PROMIS measure, the mean standard error of measurement was calculated and incorporated into MID analyses. Using a large sample (<italic>n</italic> = 170), a number of the cross‐sectional T‐score anchor‐based MID estimates (57%) were not included due to failure to meet a priori criteria. Based on the analyses, the following T‐score MID ranges are recommended: Depression CAT (3.5–5.5) and Anxiety CAT (3.0–5.5). The average effect sizes for MID estimates ranged from .32 to .57. This study is among the first to address MIDs for PROMIS measures; it is the first study to establish usable MIDs for psychological symptoms on outpatients with chronic/persistent pain. The results may be used to gauge minimally important clinical difference and/or treatment response for individuals within this patient population. MIDs for PROs are particularly useful when treatment responses are significant to the patient but are difficult to evaluate during the clinical visit.</p> </abstract>