Colleen A. McHorney

The MOS 36-Item Short-Form Health Survey (SF-36): II. Psychometric and clinical tests of validity in measuring physical and mental health constructs

Cross-sectional data from the Medical Outcomes Study (MOS) were analyzed to test the validity of the MOS 36-Item Short-Form Health Survey (SF-36) scales as measures of physical and mental health constructs. Results from traditional psychometric and clinical tests of validity were compared. Principal components analysis was used to test for hypothesized physical and mental health dimensions. For purposes of clinical tests of validity, clinical criteria defined mutually exclusive adult patient groups differing in severity of medical and psychiatric conditions. Scales shown in the components analysis to primarily measure physical health (physical functioning and role limitations-physical) best distinguished groups differing in severity of chronic medical condition and had the most pure physical health interpretation. Scales shown to primarily measure mental health (mental health and role limitations-emotional) best distinguished groups differing in the presence and severity of psychiatric disorders and had the most pure mental health interpretation. The social functioning, vitality, and general health perceptions scales measured both physical and mental health components and, thus, had the most complex interpretation. These results are useful in establishing guidelines for the interpretation of each scale and in documenting the size of differences between clinical groups that should be considered very large.

The MOS 36-item Short-Form Health Survey (SF-36): III. Tests of data quality, scaling assumptions, and reliability across diverse patient groups

The widespread use of standardized health surveys is predicated on the largely untested assumption that scales constructed from those surveys will satisfy minimum psychometric requirements across diverse population groups. Data from the Medical Outcomes Study (MOS) were used to evaluate data completeness and quality, test scaling assumptions, and estimate internal-consistency reliability for the eight scales constructed from the MOS SF-36 Health Survey. Analyses were conducted among 3,445 patients and were replicated across 24 subgroups differing in sociodemographic characteristics, diagnosis, and disease severity. For each scale, item-completion rates were high across all groups (88% to 95%), but tended to be somewhat lower among the elderly, those with less than a high school education, and those in poverty. On average, surveys were complete enough to compute scale scores for more than 96% of the sample. Across patient groups, all scales passed tests for item-internal consistency (97% passed) and item-discriminant validity (92% passed). Reliability coefficients ranged from a low of 0.65 to a high of 0.94 across scales (median=0.85) and varied somewhat across patient subgroups. Floor effects were negligible except for the two role disability scales. Noteworthy ceiling effects were observed for both role disability scales and the social functioning scale. These findings support the use of the SF-36 survey across the diverse populations studied and identify population groups in which use of standardized health status measures may or may not be problematic.

The validity and relative precision of MOS short- and long-form health status scales and Dartmouth COOP charts. Results from the Medical Outcomes Study

This study estimated the validity and relative precision (RP) of four methods (MOS long- and short-form scales, global items, and COOP Poster Charts) in measuring six general health concepts. The authors also tested whether and how precisely each method discriminated relatively well adult patients (N = 638) from those with only severe chronic medical (N = 168) and only psychiatric conditions (N = 163), as clinically defined. For comparisons between the well group and both medical and psychiatric groups, RP estimates favored long-form over short-form, multi-item scales, and favored multi-item scales over single-item global measures and poster charts. In relation to long forms, short-form multi-item scales achieved a median RP of .93; RP estimates for global items and poster charts were .81 and .67, respectively. Variations in RP across methods and concepts were linked to differences in the coarseness of measurement scales, reliability, and content (including the effects of chart illustrations). These variations in RP have implications for the interpretation of scores, the statistical power of comparisons between clinical groups, and the size of confidence intervals around individual patient scores.

Comparisons of the costs and quality of norms for the SF-36 health survey collected by mail versus telephone interview: results from a national survey

Many health status surveys have been designed for mail, telephone, or inperson administration. However, with rare exception, investigators have not studied the effect the survey mode of administration has on the way respondents assess their health and other important parameters (such as response rates, nonresponse bias, and data quality), which can affect the generalizability of results. Using a national sampling frame of noninstitutionalized adults from the General Social Survey, we randomly assigned adults to a mail survey (80%) or a computer-assisted telephone survey (20%). The surveys were designed to provide national norms for the SF-36 Health Survey. Total data collection costs per case for the telephone survey (

Patient preferences for medical decision making: who really wants to participate?

47.86) were 77% higher than that for the mail survey (

Evaluation of the MOS SF-36 physical functioning scale (PF-10) : I. Unidimensionality and reproducibility of the Rasch item scale

27.07). A significantly higher response rate was achieved among respondents randomly assigned to the mail (79.2%) than telephone survey (68.9%). Nonresponse bias was evident in both modes but, with the exception of age, was not differential between modes. The rate of missing responses was higher for mail than telephone respondents (1.59 vs. 0.49 missing items). Health ratings based on the SF-36 scales were less favorable, and reports of chronic conditions were more frequent, for mail than telephone respondents. Results are discussed in light of the trade-offs involved in choosing a survey methodology for health status assessment applications. Norms for mail and telephone versions of the SF-36 survey are provided for use in interpreting individual and group scores.

Evaluation of the MOS SF-36 Physical Functioning Scale (PF-10): II. Comparison of relative precision using Likert and Rasch scoring methods.

OBJECTIVESnTo identify the determinants of patient preferences for participation in medical decision making.nnnMETHODSnData were analyzed for 2,197 patients from the Medical Outcomes Study, a 4-year observational study of patients with chronic disease (hypertension, diabetes, myocardial infarction, congestive heart failure, and depression). Multivariate logistic regression models estimated the effects of patients sociodemographic, clinical, psychosocial, and lifestyle characteristics on their decision-making preferences.nnnRESULTSnA majority of the patients (69%) preferred to leave their medical decisions to their physicians. The odds for preferring an active role significantly decreased with age and increased with education. Women were more likely to be active than men (odds ratio [OR] = 1.44, P < 0.001). Compared with patients who only suffered with unsevere hypertension, those with severe diabetes (OR = 0.62, P = 0.04) and unsevere heart disease (OR = 0.45, P = 0.02) were less likely to prefer an active role. Patients with clinical depression were more likely to be active (OR = 1.64, P = 0.01). Patients pursuing active coping strategies had higher odds for an active role than passive copers, while those who placed higher value on their health were less likely to be active than those with low health value (OR = 0.59, P < 0.001).nnnCONCLUSIONSnAlthough a majority of patients prefer to delegate decision making to physicians, preferences vary significantly by patient characteristics. Approaches to enhancing patient involvement will need to be flexible and accommodating to individual preferences in order to maximize the benefits of patient participation on health outcomes.

Construction and Validation of an Alternate Form General Mental Health Scale for the Medical Outcomes Study Short-Form 36-Item Health Survey

Indexes developed to measure physical functioning as an essential component of general health status are often based on sets of hierarchically-structured items intended to represent a broad underlying concept. Rasch Item Response Theory (IRT) provides a methodology to examine the hierarchical structure, unidimensionality, and reproducibility of item positions (calibrations) along a scale. Data gathered on the 10-item Physical Functioning Scale (PF-10) from a large sample of Medical Outcomes Study patients (N = 3445) were used to examine the hierarchical order, unidimensionality, and reproducibility of item calibrations. Rasch-IRT analyses generated an empirical item hierarchy, confirmed the unidimensionality of the PF-10 for most patients, and established the reproducibility of item calibrations across patient populations and repeated tests. These findings support the content validity of the PF-10 as a measure of physical functioning and suggest that valid Rasch-IRT summary scores could be generated as an alternative to the current Likert summative scores. Unidimensionality and reproducibility of the item scale are essential prerequisites for the development of Rasch-based person measures of physical functioning that can be used across populations and over repeated tests.

Equating health status measures with item response theory: illustrations with functional status items.

This study examined the relative precision (RP) of two methods of scoring the 10-item Physical Functioning Scale (PF-10) from a large sample of patients (n = 3445) of the Medical Outcomes Study. Based on a Likert scaling model, the PF-10 summated scoring method was compared with a Rasch Item Response Theory (IRT) scaling model in which raw scores were transformed into a latent trait variable of physical functioning. Potential differences between scoring methods were hypothesized to be attributed to: (1) the logarithmic nature of the Rasch transformation; (2) the unevenness of the PF-10 item distributions; and (3) reduction of within-group variance. RP ratios favored the Rasch model in discriminating between patients who differed in disease severity. The Rasch and Likert scoring models performed similarly for tests involving sensitivity to change over a two-year follow-up period. In all comparisons, differences between methods were most apparent in clinical groups whose scores most approximated the extremes of the score distribution. Further research is necessary to test for differences between scoring models in discrimination and sensitivity to change among clinical groups whose scores are sufficiently spread across the continuum of physical functioning, in particular patients with either very high or low physical functioning. The Rasch model of scoring may have important implications for the clinical interpretation of individual scores at all ranges of the scale.

Gender differences in medical treatment: The case of physician-prescribed activity restrictions

Alternate-form health measures are useful for clinical trials or health services research requiring repeated administrations over a short interval of time. Further, by using alternate-form methodology, they can be utilized to estimate score reliability. Data from the Medical Outcomes Study were used to evaluate five alternate forms of the Short-Form 36-Item Health Survey (SF-36) general mental health scale (MHI-5). Well-established psychometric criteria were used to select the best alternate form and to estimate the reliability of the MHI-5 using the alternate-form methodology. Although a considerable degree of comparability across the five alternate forms was observed for criteria pertaining to estimates of item-internal consistency and reliability, distributional characteristics of scales, tests of empirical validity, and score equivalence at the individual level, we recommend one alternate form that satisfied all evaluation criteria and did so better than any other alternate form. Using the alternate-form methodology of estimating reliability, results suggest that the internal-consistency method underestimates the reliability of the MHI-5 by 3%. The methodology presented here should prove useful to others interested in constructing and evaluating alternate forms, and the alternate form recommended here (MHI-5AF) should prove useful across many health status assessment applications.