Minimal important difference thresholds and the standard error of measurement: is there a connection?

J Biopharm Stat. 2004 Feb;14(1):97-110. doi: 10.1081/BIP-120028508.

Abstract

Several recently published investigations have examined the relationship between the magnitude of the standard error of measurement (SEM) and established thresholds for a minimal clinically important difference (MCID) or a minimal important difference (MID) for change scores on health-related quality of life (HRQOL) or health status measures. These investigations, however, have resulted in differing SEM criteria for the MCID or MID. This study reviews and compares two sets of studies: (1) three investigations using a disease-specific HRQOL measure among patient samples with the chronic disease (heart disease, chronic obstructive pulmonary disease, or asthma) that have consistently demonstrated a 1 SEM correspondence with the established MCIDs or MIDs and (2) three investigations among patients referred to physical therapists with back, lower extremity, and neck pain showing that approximately 2.3 SEMs estimated the established MCID standards for three different measures of health status. Chronic disease patients were classified to have a MCID or MID if their global change ratings for the better or the worse were 1, 2, or 3 on a Likert scale ranging from 1 (almost the same, hardly any better, or worse at all) to 7 (a very great deal better or worse). Back pain patients, however, needed average global transition scores of 5, 6, or 7 (a good, a great, or a very great deal better) on the same 7-point Likert scale in order to experience an MCID in their condition. Charting these change levels against their respective SEM-MID criteria provides insight and promise for linking SEM-based criteria to MCID standards for other HRQOL and health status measures.

Publication types

  • Research Support, U.S. Gov't, P.H.S.

MeSH terms

  • Data Interpretation, Statistical*
  • Health Status Indicators*
  • Health Status*
  • Humans
  • Research Design / statistics & numerical data