Background Biostatistics is an increasing focus in current medical school curricula, even appearing as a topic on the United States Medical Licensing Exam Steps 1, 2, and 3 (USMLE). Physicians already in practice who did not have the benefit of biostatistics training in medical school likely can benefit from clarification and instruction of statistical methods. The goal of the study is to document and evaluate statistical methods utilized in a high-impact factor Journal, then to develop a guide to those statistics that can be applied to facilitate the interpretation of statistics for practicing physicians, fellows, residents, and medical students.

Methods In Part I of our tandem manuscripts, the 100 most recent primary articles, dating from February 2021 to December 2021, were extracted and analyzed from the Journal of Intensive Care Medicine. The evaluation consisted of study temporality, study design, types of descriptor variables, and types of statistical tests. Statistical analyses was conducted using the Fisher’s Exact test with p-values <0.05 considered statistically significant.

Results Retrospective studies were most common (75/100, 75%), followed by prospective (23/100, 23%), and cross-sectional studies (2/100, 2%). The most popular designs were cohort (82/100, 82%), followed by case series (9/100, 9%), randomized control trials (4/100, 4%), and case-control (3/100, 3%). The most commonly utilized descriptor variables were frequency and proportion (100/100, 100%), followed by both median (74/100, 74%), average (71/100, 71%), and interquartile range (68/100, 68%). The chi-square test was the most frequently used statistical test (59/100, 59%), while logistic regression (48/100, 48%), Mann-Whitney-U (46/100, 46%), and two-sample independent t-test (40/100, 40%) also were popular. A subgroup analysis of cohort studies showed higher use of the chi-square test (44/63, 70% vs 8/19, 42%; p = 0.03) and logistic regression (40/63, 63% vs 5/19, 26%; p = 0.008) in retrospective compared to prospective studies.

Conclusion This review of statistics used in 100 primary journal articles revealed that retrospective and cohort studies were utilized most frequently. The chi-square test was used in the majority of studies, particularly in retrospective cohort studies. This information can help determine areas in which supplemental training will be most beneficial to improve the understanding of statistical methods in medical journals by practicing physicians, fellows, residents, and students. As an outgrowth of this study, we have developed a practical guide to relevant statistical methods serving as part 2 of this double manuscript, and plan to conduct a larger study based on this methodology.