Interpretation of patient-reported outcome measures: an inventory of over 3000 minimally important difference estimates and an assessment of their credibility

Article type
Authors
Carrasco-Labra A1, Devji T1, Qasim A1, Phillips M1, Devasenapathy N2, Zeraatkar D1, Bhatt M1, Jin X3, Brignardello-Petersen R1, Urqhart O4, Faroutan F1, Schandelmaier S1, Pardo-Hernandez H5, Vernooij RW1, Huang W6, Rizwan Y1, Lytvyn L1, Siemieniuk R1, Johnston BC7, Ebrahim S1, Furukawa TA8, Patrick DL9, Schünemann HJ1, Nesrallah G10, Guyatt G1
1McMaster University
2 Indian Institute of Public Health-Delhi
3University of Alberta
4American Dental Association
5Iberoamerican Cochrane Centre
6University of Michigan
7Dalhousie University
8Department of Health Promotion and Human Behavior, School of Public Health, Kyoto University Graduate School of Medicine
9University of Washington
10Humber River Regional Hospital
Abstract
Background:
Patient-reported outcome measures (PROMs) capture patients' perspectives on treatment benefits and harms. Interpreting results of PROMs requires understanding the extent of improvement or deterioration that patients consider important: the minimal important difference (MID), the smallest change in a PROM that patients perceive as an important benefit or harm. No inventory of MIDs for PROMs is currently available, requiring clinicians and researchers to navigate a vast literature to retrieve a specific MID. Even if they find a MID, there is no guidance to ascertain its credibility.

Objective:
To create an inventory of published anchor-based MIDs associated with PROMs and to determine their credibility.

Methods:
We searched MEDLINE, Embase, PsycINFO and CINAHL to identify studies estimating anchor-based MIDs of PROMs. Teams of two authors independently screened citations, and identified and extracted relevant data. We collected information on study design, disease or condition, population demographics and characteristics of the PROMs and anchor, and created and applied a new instrument to assess the credibility of MID estimates.

Results:
Of 5656 citations retrieved for title and abstract screening, we selected 1716 for full-text screening of which 338 proved eligible. We summarised over 3000 estimates, including MIDs for PROMs across different populations, conditions and interventions, obtained using different anchors and statistical methods. Mean change methods and receiver operating characteristics curve analysis were the most common methods to estimate MIDs. MIDs were largely calculated using patient-reported, as opposed to proxy or clinician-reported anchors. However, most studies failed to report the correlation between the anchor and the PROM; when they did the correlation was lower than 0.5. Thus, credibility of MIDs was often limited.

Conclusion:
Our inventory of available MIDs in the medical literature and their associated credibility will be of great use for anyone using PROMs to inform healthcare decisions, including clinical trialists, systematic review authors, patients and clinicians. We will illustrate how our inventory can facilitate the interpretation of results by promoting the use of credible anchor-based MID estimates.

Patient, healthcare consumer involvement:
PROMs and MIDs directly reflect patient opinion.