Standard error of measurement as a valid alternative to minimally important difference for evaluating the magnitude of changes in patient-reported outcomes measures Javier Rejas a, * , Antonio Pardo b , Miguel A ´ ngel Ruiz b a Department of Pharmacoeconomics and Health Outcomes Research, Medical Unit, Pfizer Spain, Parque Empresarial de La Moraleja, Avda. Europa, 20-B, 28108 Alcobendas, Madrid, Spain b Department of Social Psychology and Methodology, Madrid Autonomous University, Madrid, Spain Accepted 25 May 2007 Abstract Objective: To assess the degree of agreement between standard error of measurement (SEM) and minimally important difference (MID) criteria to evaluate the magnitude of the change caused by a medical intervention. Study Design and Setting: Data were obtained from a cohort of 603 patients with neuropathic pain undergoing analgesic treatment with gabapentin who completed four health scales: Medical Outcomes Study Sleep Scale, Sheehan Disability Scale, Covi Anxiety Scale, and Raskin Depression scale. After calculating MID and SEM values for all scales, patients were classified into three categories: improve- ment, no change, and worsening. Agreement between the two criteria was assessed using Cohen’s kappa index of agreement and Kendall’s tau-b linear correlation coefficient. Results: The 1 SEM criterion showed the highest agreement (kappa 5 0.68e1.00) and correlation (tau-b 5 0.75e1.00) with the MID criterion. Sensitivity analysis performed in gabapentin responders and nonresponders confirmed the results of the main analysis. Conclusion: The 1 SEM criterion is a valid alternative to the MID criterion to evaluate the magnitude of the change produced in patient-reported health outcomes measures. Ó 2008 Elsevier Inc. All rights reserved. Keywords: Standard error of measurement; Minimally important difference; Patient-reported outcomes; Clinical significance; Effect size; Questionnaires 1. Introduction Health-related quality-of-life (HRQOL) questionnaires have been used for several decades. However, genuine in- terest in use and application of these measurement instru- ments in medical practice has emerged only a few years ago. Although it is now difficult to find a study or clinical trial that does not include some patient-reported health measure (whether it is a HRQOL questionnaire, a treatment satisfaction scale, or a scale to determine the value given by patients to their health status), use of health questionnaires in daily medical practice (to support decision making aimed at improving patients’ health) is still far from being a routine practice. Reasons that may explain this insuffi- cient clinical application include clinician overload, short- age of time during clinical visits, complexity of scoring of some scales, and the difficulty in interpreting the scores obtained (due to either the lack of standardized guidelines or the difficulty involved in clinical interpretation of the magnitude of the changes seen in scale scores) [1,2]. Interpretation of the score obtained by a patient in a HRQOL questionnaire initially depends on whether the measurement is performed once (cross-sectional) or repeated over time (longitudinal). In the first case, the pa- tient’s response is interpreted by comparing it to the values found in the reference population, which requires standard- ized values that allow for assigning the patient a relative lo- cation on the scale using percentiles or standard scores [3e5]. Direct interpretation is also possible, but is less straightforward. When repeated measurements over time are available, in addition to comparison to the reference group, it is of inter- est to assess whether the questionnaire is able to detect a change in the patient’s health status to quantify the magnitude of the change (effect size) [6,7]. * Corresponding author. Tel.: þ91-490-9957. E-mail address: javier.rejas@pfizer.com (J. Rejas). 0895-4356/08/$ e see front matter Ó 2008 Elsevier Inc. All rights reserved. doi: 10.1016/j.jclinepi.2007.05.011 Journal of Clinical Epidemiology 61 (2008) 350e356