Methodological and statistical techniques: what do residents really need to know about statistics?

The purpose of this study was to catalog the statistical methods used in six journals two each from the fields of Family Practice, Emergency Medicine, and Obstetrics and Gynecology. We reviewed the quantitative articles from January 1998 through December 2000 from the Journal of Family Practice, the Journal of Family Medicine, the Annals of Emergency Medicine, the Journal of Academic Emergency Medicine. Articles from January 2000 through December 2000 of Obstetrics and Gynecology and the American Journal of Obstetrics and Gynecology were also included. Case reports and editorials were not included in this analysis. There were a total of 1828 articles reviewed (666 from Emergency Medicine articles, 380 from Family Practice, and 782 from Obstetrics and Gynecology). The distribution of study types (cross-sectional or survey, retrospective, or prospective) did not differ between the selected journals within Emergency Medicine, Family Practice, or Obstetrics and Gynecology. Pearson's chi-square/Fisher's Exact test was the statistic of choice overall (47.5%) followed by Student's t-test (33.1%). Analysis-of-variance was used in 23.3% of the studies, nonparametric methods (8.1%), linear regression (17.6%), and odds ratios/logistic regression (17.4%). Other statistical procedures were used less than 10% of the time. These results show that a physician who comfortably comprehends the appropriate use of descriptive statistics Student's t-test, Pearson's chi-square/Fisher's Exact test will be able to read and interpret at least 70% of the published medical literature. Educational efforts should focus on appropriate study design and analysis.