The object of this article was to systematically review available methods to measure comorbidity and to assess their validity and reliability. A search was made in Medline and Embase, with the keywords comorbidity and multi-morbidity, to identify articles in which a method to measure comorbidity was described. The references of these articles were also checked, and using a standardized checklist the relevant data were extracted from these articles. An assessment was made of the content, concurrent, predictive and construct validity, and the reliability. Thirteen different methods to measure comorbidity were identified: one disease count and 12 indexes. Data on content and predictive validity were available for all measures, while data on construct validity were available for nine methods, data on concurrent validity, and interrater reliability for eight methods, and data on intrarater reliability for three methods. The Charlson Index is the most extensively studied comorbidity index for predicting mortality. The Cumulative Illness Rating Scale (CIRS) addresses all relevant body systems without using specific diagnoses. The Index of Coexisting Disease (ICED) has a two-dimensional structure, measuring disease severity and disability, which can be useful when mortality and disability are the outcomes of interest. The Kaplan Index was specifically developed for use in diabetes research. The Charlson Index, the CIRS, the ICED and the Kaplan Index are valid and reliable methods to measure comorbidity that can be used in clinical research. For the other indexes, insufficient data on the clinimetric properties are available.

Changes in scores on health status questionnaires are difficult to interpret. Several methods to determine minimally important changes (MICs) have been proposed which can broadly be divided in distribution-based and anchor-based methods.

The object of this article was to systematically review available methods to measure comorbidity and to assess their validity and reliability. A search was made in Medline and Embase, with the keywords comorbidity and multi-morbidity, to identify articles in which a method to measure comorbidity was described. The references of these articles were also checked, and using a standardized checklist the relevant data were extracted from these articles. An assessment was made of the content, concurrent, predictive and construct validity, and the reliability. Thirteen different methods to measure comorbidity were identified: one disease count and 12 indexes. Data on content and predictive validity were available for all measures, while data on construct validity were available for nine methods, data on concurrent validity, and interrater reliability for eight methods, and data on intrarater reliability for three methods. The Charlson Index is the most extensively studied comorbidity index for predicting mortality. The Cumulative Illness Rating Scale (CIRS) addresses all relevant body systems without using specific diagnoses. The Index of Coexisting Disease (ICED) has a two-dimensional structure, measuring disease severity and disability, which can be useful when mortality and disability are the outcomes of interest. The Kaplan Index was specifically developed for use in diabetes research. The Charlson Index, the CIRS, the ICED and the Kaplan Index are valid and reliable methods to measure comorbidity that can be used in clinical research. For the other indexes, insufficient data on the clinimetric properties are available.

Background: Minimally important changes (MIC) in scores help interpret results from health status instruments. Various distribution-based and anchor-based approaches have been proposed to assess MIC. Objectives: To describe and apply a visual method, called the anchor-based MIC distribution method, which integrates both approaches. Method: Using an anchor, patients are categorized as persons with an important improvement, an important deterioration, or without important change. For these three groups the distribution of the change scores on the health status instrument are depicted in a graph. We present two cut-off points for an MIC: the ROC cut-off point and the 95% limit cut-off point. Results: We illustrate our anchor-based MIC distribution method determining the MIC for the Pain Intensity Numerical Rating Scale in patients with low back pain, using two conceivable definitions of minimal important change on the anchor. The graph shows the distribution of the scores of the health status instrument for the relevant categories on the anchor, and also the consequences of choosing the ROC cut-off point or the 95% limit cut-off point. Discussion: The anchor-based MIC distribution method provides a general framework, applicable to all kind of anchors. This method forces researchers to choose and justify their choice of an appropriate anchor and to define minimal importance on that anchor. The MIC is not an invariable characteristic of a measurement instrument, but may depend, among other things, on the perspective from which minimal importance is considered and the baseline values on the measurement instrument under study. A balance needs to be struck between the practicality of a single MIC value and the validity of a range of MIC values.

ABSTRACT. van der Lee JH, de Groot V, Beckerman H, Wagenaar RC, Lankhorst GJ, Bouter LM. The intra-and interrater reliability of the Action Research Arm test: a practical test of upper extremity function in patients with stroke. Arch Phys Med Rehabil 2001;82:14-9.Objectives: To determine the intra-and interrater reliability of the Action Research Arm (ARA) test, to assess its ability to detect a minimal clinically important difference (MCID) of 5.7 points, and to identify less reliable test items.Design: Intrarater reliability of the sum scores and of individual items was assessed by comparing (1) the ratings of the laboratory measurements of 20 patients with the ratings of the same measurements recorded on videotape by the original rater, and (2) the repeated ratings of videotaped measurements by the same rater. Interrater reliability was assessed by comparing the ratings of the videotaped measurements of 2 raters. The resulting limits of agreement were compared with the MCID.Patients: Stratified sample, based on the intake ARA score, of 20 chronic stroke patients (median age, 62yr; median time since stroke onset, 3.6yr; mean intake ARA score, 29.2).Main Outcome Measures: Spearman's rank-order correlation coefficient (Spearman's rho); intraclass correlation coefficient (ICC); mean difference and limits of agreement, based on ARA sum scores; and weighted kappa, based on individual items.Results: All intra-and interrater Spearman's rho and ICC values were higher than .98. The mean difference between ratings was highest for the interrater pair (.75; 95% confidence interval, .02-1.48), suggesting a small systematic difference between raters. Intrarater limits of agreement were Ϫ1.66 to 2.26; interrater limits of agreement were Ϫ2.35 to 3.85. Median weighted kappas exceeded .92. Conclusion:The high intra-and interrater reliability of the ARA test was confirmed, as was its ability to detect a clinically relevant difference of 5.7 points.

Background and Purpose-The Motor Activity Log (MAL) is a semistructured interview for hemiparetic stroke patients to assess the use of their paretic arm and hand (amount of use [AOU]) and quality of movement [QOM]) during activities of daily living. Scores range from 0 to 5. The following clinimetric properties of the MAL were quantified: internal consistency (Cronbach ␣), test-retest agreement (Bland and Altman method), cross-sectional construct validity (correlation between AOU and QOM and with the Action Research Arm [ARA] test), longitudinal construct validity (correlation of change on the MAL during the intervention with a global change rating [GCR] and with change on the ARA), and responsiveness (effect size). Methods-Two baseline measurements 2 weeks apart and 1 follow-up measurement immediately after 2 weeks of intensive exercise therapy either with or without immobilization of the unimpaired arm (forced use) were performed in 56 chronic stroke patients. Results-Internal consistency was high (AOU: ␣ϭ0.88; QOM: ␣ϭ0.91). The limits of agreement were Ϫ0.70 to 0.85 and Ϫ0.61 to 0.71 for AOU and QOM, respectively. The correlation with the ARA score (Spearman ) was 0.63 (AOU and QOM). However, the improvement on the MAL during the intervention was only weakly related to the GCR and to the improvement on the ARA, Spearman was between 0.16 and 0.22. The responsiveness ratio was 1.9 (AOU) and 2.0 (QOM). Conclusion-The MAL is internally consistent and relatively stable in chronic stroke patients not undergoing an intervention. The cross-sectional construct validity of the MAL is reasonable, but the results raise doubt about its longitudinal construct validity.

The aim of this study is to show the relationship between test-retest reproducibility and responsiveness and to introduce the smallest real difference (SRD) approach, using the sickness impact profile (SIP) in chronic stroke patients as an example. Forty chronic stroke patients were interviewed twice by the same examiner, with a 1-week interval. All patients were interviewed during the qualification period preceding a randomized clinical trial. Test-retest reproducibility has been quantified by the intraclass correlation coefficient (ICC). the standard error of measurement (SEM) and the related smallest real difference (SRD). Responsiveness was defined as the ratio of the clinically relevant change to the SD of the within-stable-subject test-retest differences. The ICC for the total SIP was 0.92, whereas the ICCs for the specified SIP categories varied from 0.63 for the category 'recreation and pastime' to 0.88 for the category 'work'. However, both the SEM and the SRD far more capture the essence of the reproducibility of a measurement instrument. For instance, a total SIP score of an individual patient of 28.3% (which is taken as an example, being the mean score in the study population) should decrease by at least 9.26% or approximately 13 items, before any improvement beyond reproducibility noise can be detected. The responsiveness to change of a health status measurement instrument is closely related to its test-retest reproducibility. This relationship becomes more evident when the SEM and the SRD are used to quantify reproducibility, than when ICC or other correlation coefficients are used.

population-based Copenhagen Stroke Study, it was found that on admission 32% of the stroke patients had severe arm paresis and 37% had mild arm paresis. 1 In 64 (13%) out of 491 surviving patients, the affected arm remained entirely a-functional, despite the efforts of a comprehensive rehabilitation programme. These patients accounted for 25% of the total number of bed-days for all 491 patients. 2 In recent Objective: Assessment of the available evidence for the effectiveness of exercise therapy to improve arm function in patients who have suffered from a stroke. Methods: A systematic search of bibliographical databases and reference checking were performed to identify publications on randomized controlled trials (RCTs) which evaluated the effect of exercise therapy on arm function in stroke patients. The methodological quality was assessed systematically by two raters, based on a standardized list of methodological criteria. Study characteristics, such as the chronicity and severity of impairment of the patient population, the amount and duration of interventions, and specific methodological criteria, were related to reported effects.