Evaluation & the health professions
-
Many applied researchers analyzing longitudinal data share a common misconception: that specialized statistical software is necessary to fit hierarchical linear models (also known as linear mixed models [LMMs], or multilevel models) to longitudinal data sets. Although several specialized statistical software programs of high quality are available that allow researchers to fit these models to longitudinal data sets (e.g., HLM), rapid advances in general purpose statistical software packages have recently enabled analysts to fit these same models when using preferred packages that also enable other more common analyses. One of these general purpose statistical packages is SPSS, which includes a very flexible and powerful procedure for fitting LMMs to longitudinal data sets with continuous outcomes. This article aims to present readers with a practical discussion of how to analyze longitudinal data using the LMMs procedure in the SPSS statistical software package.
-
Assessments of temporal bone dissection performance among otolaryngology residents have not been adequately developed. At the Ohio State College of Medicine, an instrument (Welling Scale, Version 1 [WS1]) is used to evaluate residents' end-product performance after drilling a temporal bone. In this study, the authors evaluate the components that contribute to measurement error using this scale. ⋯ The largest source of measurement error was caused by residents' inconsistent performance across bones. Rater disagreement introduced only small error into scores. The WS1 provides small measurement error, with two raters and two bones for each participant.
-
Peer-assessment processes with chart review have been used for many years to assess the clinical performance of physicians. The Quebec medical licensing authority has been required by provincial law to assess the practicing Quebec physicians on a nonvoluntary basis. During the period from January 2001 to November 2004, 25 family physicians in active practice were randomly selected from a pool of about 300. ⋯ The concordance between chart review alone and that of chart review with chart-stimulated recall interview was 75% for chart keeping, 69% for clinical investigation, 81% for diagnostic accuracy, and 74% for treatment plan. Ratings based on chart review alone achieve moderate levels of reliability (Kappa = 0.44 to 0.56). It appears that some important information about quality of care is missed when only chart review is used.
-
Despite advances in the storage and retrieval of information within health care systems, health researchers conducting surveys for evaluations still face technical barriers that may lead to sampling bias. The authors describe their experience in administering a Web-based, international survey to English-speaking countries. ⋯ Sampling bias arose from (a) system incompatibility, which did not allow potential respondents to open the survey, (b) varying institutional gate-keeping policies that "recognized" the unsolicited survey as spam, (c) culturally unique program terminology, which confused some respondents, and (d) incomplete sampling frames. Solutions are offered to the first three problems, and the authors note that sampling bias remains a crucial problem.
-
Telephone triage programs have been shown to be cost-effective and favorably utilized by insured populations. However, there are 45 million Americans who are uninsured and who do not have access to telephone nursing. A telephone triage service was piloted for local uninsured residents. ⋯ Most callers (98%) believed that their health care concern was understood. Moreover, 98% agreed with the advice given, and 90% reported following up on the advice given. Overall satisfaction by the uninsured population with the telephone-based nurse triage service was positive and appears to be an effective and acceptable tool by those uninsured individuals who utilized its services.