[Use and comprehensibility of statistical analysis in Archivos de Bronconeumología (1970-1999)]

To describe the type of statistical analysis used most often in original research articles published in Archivos de Bronconeumología, and the evolution of statistical analysis over time in terms of complexity. To determine comprehensibility, taking bivariate analysis as the reference threshold.

All articles published in the original research section of Archivos de Bronconeumología from 1970 through 1999 were reviewed manually. For each article we recorded the category or categories of statistical analysis and its comprehensibility (with a reference threshold set at category 7). We studied the following factors: year of publication, type of analysis, comprehensibility, maximum category achieved, subject area and number of authors.

Eight hundred sixty original articles, with a mean 5 2 authors per article were examined. The maximum category reached was a mean 4.15 4.61. The three types of analysis used most often in all articles were category 1 (descriptive only) at 49.4%, category 2 (t and z tests) at 26.4% and category 3 (bivariate tables) at 19.1%. Among the more complex analytical categories, the most often used were analysis of variance (category 8) at 9%, survival analysis (category 16) at 6.2%, and non-parametric correlations at 3.4%. Comparing results by decade, the proportion of articles with only descriptive analysis fell from 74% in the seventies to 63.9% in the eighties and to 36.1% in the nineties (90s vs. 80s, p < 0.05; 90s vs. 70s, p < 0.05; 80s vs. 70s, p < 0.05). The use of statistics from categories 3 and 6 increased (p < 0.05), and the use of the more complex categories 8, 10, 11 and 16 increased (p < 0.05). Comprehensibility decreased significantly to 68.1% in the 90s in relation to increased complexity of analysis.

The complexity of statistical analysis in articles published in Archivos de Bronconeumología increased over time, while comprehensibility decreased.