护理期刊论文中统计推断应用错误分析

目的 了解护理期刊论文中统计推断的应用情况,分析其中存在的统计错误,以提高护理人员及期刊编辑对统计学方法 的识别与正确应用能力.方法 以3种护理期刊为对象,手工检索其2007年1～12期刊登的文献.对每篇文章进行阅读并逐项记录,内容包括是否使用统计推断方法,统计推断方法 类型以及存在的统计错误类型.结果 3种期刊平均统计推断应用率32.62%,其中以X2>检验(占39.67%)及t检验(占32.23%)应用较多.平均统计推断错误率为40.10%,统计错误中以计量资料的统计推断中存在的错误比例最高,占39.69%;其次为等级资料的统计推断中存在的错误,占21.79%.结论 护理期刊论文中统计推断的使用率较高,但方法 的滥用及统计推断错误也较多.护理科研人员正确使用统计学方法 的意识亟待强化,应加强统计学知识学习,以提高统计学方法 的正确应用水平.同时,期刊编辑应提高对护理论文中统计学错误的识别能力,以促进护理科技期刊学术质量的提高.     

Development and Evaluation of a Graphical Anesthesia Drug Display

Background: Usable real-time displays of intravenous anesthetic concentrations and effects could significantly enhance intraoperative clinical decision-making. Pharmacokinetic models are available to estimate past, present, and future drug effect-site concentrations, and pharmacodynamic models are available to predict the drug's associated physiologic effects.

Methods: An interdisciplinary research team (bioengineering, architecture, anesthesiology, computer engineering, and cognitive psychology) developed a graphic display that presents the real-time effect-site concentrations, normalized to the drugs' EC95, of intravenous drugs. Graphical metaphors were created to show the drugs' pharmacodynamics. To evaluate the effect of the display on the management of total intravenous anesthesia, 15 anesthesiologists participated in a computer-based simulation study. The participants cared for patients during two experimental conditions: with and without the drug display.

Results: With the drug display, clinicians administered more bolus doses of remifentanil during anesthesia maintenance. There was a significantly lower variation in the predicted effect-site concentrations for remifentanil and propofol, and effect-site concentrations were maintained closer to the drugs' EC95. There was no significant difference in the simulated patient heart rate and blood pressure with respect to experimental condition. The perceived performance for the participants was increased with the drug display, whereas mental demand, effort, and frustration level were reduced. In a postsimulation questionnaire, participants rated the display to be a useful addition to anesthesia monitoring.     

Assumptions of Statistical Tests: What Lies Beneath

We have discussed many statistical tests and tools in this series of commentaries, and while we have mentioned the underlying assumptions of the tests, we have not explored them in detail. We stop to look at some of the assumptions of the t-test and linear regression, justify and explain them, mention what can go wrong when the assumptions are not met, and suggest some solutions in this case.     

A Statistical Method to Minimize Magnification Errors in Serial Vertebral Radiographs

Incident vertebral deformities are commonly defined by observed changes in height between measurements on two consecutive radiographs. However, conventional radiographs are subject to magnification, and this magnification may differ between films, leading to artifactual changes in height. In order to minimize this effect, it is common practice to record the spine–film and film–focus distances, and from this to calculate a magnification factor for each film. We present a simple statistical method for correcting for differences in magnification between two films if the spine–film and film–focus distances are unknown. This method is shown to reduce the variance of the magnification differences in vertebral heights by 14%, considerably more than is possible using the spine–film distance. Using the statistical method, the number of vertebrae that showed not only a reduction in one or more height of 15%, but were also judged clinically to be free from any incident deformity by an expert radiologist, was reduced from 100 to 46. The number showing a reduction of 20% that were judged fracture-free was reduced from 15 to 9. In the subset of subjects for whom the spine–film distance was known, the reduction in false positives was similar, whichever method was used to correct for magnification. There was no difference in the number of confirmed incident fractures detected when magnification correction by either method was employed. It is concluded that correcting for magnification differences using the statistical method outlined here reduces the number of false positive deformities very substantially and by a similar extent as correcting the magnification using reliable, measured spine–film and film–focus distances. A further advantage of this method is that it can be used retrospectively. Received: 26 May 1999 / Accepted: 9 May 2001     

Analysis of Statistical Tests to Compare Visual Analog Scale Measurements among Groups

Background: A common type of study performed by anesthesiologists determines the effect of an intervention on pain reported by groups of patients. The goal of this study was to evaluate the effectiveness of t, analysis of variance (ANOVA), Mann-Whitney, and Kruskal-Wallis tests to compare visual analog scale (VAS) measurements between two or among three groups of patients. These results may be particularly helpful during the design of studies that measure pain with a VAS.

Methods: One VAS measurement was obtained from each of 480 nulliparous women in labor who were receiving oxytocin (149), nalbuphine (159), or epidural bupivacaine (172). Multiple simulated samples were then drawn from these data. These simulated samples were used in computer simulations of clinical trials comparing VAS measurements among groups. t and ANOVA tests were performed before and after an arcsin transformation was used, to make the data closer to a normal distribution. VAS measurements were also compared after they were divided into five ranked categories.

Results: The statistical distributions of VAS measurements were not normal (P < 10 sup -7). Arcsin transformation made the distributions closer to normal distributions. Nevertheless, no statistical test incorrectly suggested that a difference existed among groups, when there was no difference, more often than the expected rate, t or ANOVA tests had a slightly greater statistical power than the other tests to detect differences among groups. Because arcsin transformation both decreased differences among means and reduced the variance to a lesser extent, it decreased power to detect differences among groups. Statistical power to detect differences among groups was not less for a five-category VAS than for a continuous VAS.     

Statistical Analysis of Notational AFL Data Using Continuous Time Markov Chains

Animal biologists commonly use continuous time Markov chain models to describe patterns of animal behaviour. In this paper we consider the use of these models for describing AFL football. In particular we test the assumptions for continuous time Markov chain models (CTMCs), with time, distance and speed values associated with each transition. Using a simple event categorisation it is found that a semi-Markov chain model is appropriate for this data. This validates the use of Markov Chains for future studies in which the outcomes of AFL matches are simulated.

Key Points

• A comparison of four AFL matches suggests similarity in terms of transition probabilities for events and the mean times, distances and speeds associated with each transition.
• The Markov assumption appears to be valid.
• However, the speed, time and distance distributions associated with each transition are not exponential suggesting that semi-Markov model can be used to model and simulate play.
• Team identified events and directions associated with transitions are required to develop the model into a tool for the prediction of match outcomes.

Key words: Homogeneity in time, sequential dependency, semi-Markov process, football     

Analysis of Data in Nephrology I. Choosing the Statistical Test: Dichotomous Variables

This is the first of six articles describing how to choose thecorrect statistical test to look for relationships between twovariables. All examples relate to clinical or laboratory aspectsof nephrology. In this article the overall strategy is outlined,and examples are given of analyses of data where each variablecan take only two values.     

Performance Data for Traumatic Brain-Injured Subjects on the Gordon Diagnostic System (GDS) Tests of Attention

Thirty adults with traumatic brain injury (TBI) (20 males and 10 females, mean age 40 years) and a non-injured control group (12 males and 13 females, mean age 41 years) were tested on 16 tests of attention including three tasks of the Gordon Diagnostic System (GDS), a relatively new set of attention tasks. No differences between groups were found on age or education. Both groups had estimated IQs in the average range. Performance data for the GDS are presented for the TBI and CON groups. Mild to moderate deficits of attention were seen in the TBI group relative to controls on the Vigilance and Distractibility tasks. No differences between groups were seen on the Standard Delay groups. Pearson product moment correlations suggested different patterns of relationships between the GDS tasks and other tests of attention for the TBI and CON groups. These results support the utility of the Vigilance and Distractibility tasks for assessment of attention in a mild to moderately injured population.