Objective To introduce the appropriate application of statistical analysis method in medical science and technology articles. Methods Expatiated the correct application of statistical theories in statistical package, statistical analysis methods and test criterion which compose the basic statistical content in an article.Results If the distribution of numerical variable is normal, mean and standard deviance can be used to describe this variable. In the same way, t test and analysis of variance (ANOVA) can be used to test the difference of mean in each group. If it is not normal, median and range can be used to describe the variable and rank sum test can be used to test the difference of distribution in each group. Categorical variable can be described by rate, proportion and ratio. There are chi-square test, fisher’s exact test and ranksum test to test the difference of rates. Conclusion It is the key of choosing rational statistical methods to distinguish the type of design and variable.
Objective To assess the frequency and the proportion of correct use of statistical analytic methods in five Chinese otorhinolaryngological journals from 2000 to 2002. Methods The statistical methods used in all original articles (n=1 331) published in these journals in three years were evaluated. Results Only 52.0 % of the articles were used statistical analytic methods. And the frequency was higher in basic research (63.5%) than that in clinical research (48.7%) (P＜0.01). The proportions of correct use of statistical analytic methods in the five journals varied from 48.7% to 72.7%, with an average rate of 56.5%. The most frequently used statistical methods were t tests (37.9%), contingency tables (chi-square test) (28.2%) and ANOVAs (14.3%). The most common errors were on the presentation of P values without specifying the test used, using t tests instead of ANOVAs in the comparison among three and more groups, and using unpaired t tests when paired tests were required. Conclusions The rate of application statistical analytic methods is rather high, but incorrect or inappropriate use remain a serious problem.
Objective To evaluate the quality of clinical studies on dentistry from the Chinese Journals. Methods Clinical studies in Chinese Journal of Conservative Dentistry of 2002 were searched. The quality of the clinical studies on assessment of treatments’ efficacy was evaluated. Results Among 204 related studies from 12 issues, there were 93 (45.58%) restrospective intervention studies, 6 randomized controlled blinded trials (2.94%), 42 randomized trials without blindness (20.58%), 20 controlled trials without randomization (9.80%) and 25 clinical observational studies (12.25%). The statistical analysis showed that 20 studies were with inappropriate methods. Conclusions It is necessary to improve the design and statistical analysis of clinical studies on stomatology in China to produce high-quality research evidence.
In the study of real-world data, the pragmatic randomized controlled trial can provide the optimal evidence for clinical decisions. Although randomization protects against confounding, post-randomization confounding may still arise due to non-compliance. Traditional intention-to-treat analysis will drift apart from true estimation and lead to deviation of clinical decisions. Meanwhile, the alternative traditional methods would subject to bias and confounding. Thus, new methods are required for revolution, i.e., instrument variable method and modern per-protocol analysis. Our study reviews the defects of traditional methods in pragmatic randomized controlled trials, and then refers to two new methods with a detailed discussion of strengths and weaknesses. We aim to provide researches with insights on choosing the statistical methods for pragmatic trial.
Pragmatic randomized controlled trials can provide high-quality evidence. However, pragmatic trials need to frequently encounter the missing outcome data due to the challenges of quality assurance and control. The missing outcome could lead to bias which may misguide the conclusions. Thus, it is crucial to handle the missing outcome data appropriately. Our study initially summarized the bias structures and missingness mechanisms, and then reviewed important methods based on the assumption of missing at random. We referred to the multiple imputations and inverse probability of censoring weighting for dealing with missing outcomes. This paper aimed to provide insights on how to choose the statistical methods on missing outcome data.
Randomized controlled trials (RCTs) are often limited because of ethical or operational reasons. Quasi-experimental studies could be an alternative to RCTs to make causal inferences without randomization by controlling the confounding effects of the study. This paper introduced the general statistical analysis methods of quasi-experimental design through basic ideas, characteristics, limitations and applications in medicine, including difference-in-difference models, instrumental variables, regression discontinuity design, interrupted time series, and so on, and to provide references for future research.
ObjectiveTo summarize and explore the application of machine learning models to survival data with non-proportional hazards (NPH), and to provide a methodological reference for large-scale, high-dimensional survival data. MethodsFirst, the concept of NPH and related testing methods were outlined. Then the advantages and disadvantages of machine learning algorithm-based NPH survival analysis methods were summarized based on the relevant literature. Finally, using real-world clinical data, a case study was conducted with two ensemble machine learning models and two deep learning models in survival data with NPH: a study of the risk of death within 30 days in stroke patients in the ICU. ResultsEight commonly used machine learning model-based NPH survival analyses were identified, including five traditional machine learning models such as random survival forest and three deep learning models based on artificial neural networks (e.g., DeepHit). The case study found that the random survival forest model performed the best (C-index=0.773, IBS=0.151), and the permutation importance-based algorithm found that age was the most important characteristic affecting the risk of death in stroke patients. ConclusionSurvival big data in the era of precision medicine presenting NPH are common, and machine learning model-based survival analysis can be used when faced with more complex survival data and higher survival analysis needs.