Fig. 2

Observational outcome-wide analysis results by theme. Statistical significance and relative effect sizes from the observational OWA for maltreatment count 4–5 vs. 0 (diamonds), 3 vs. 0 (triangles), 2 vs. 0 (circles), and 1 vs. 0 (squares), grouped by theme. Statistical significance is assessed after false discovery rate correction at α = 0.05. Results are further categorized as robust according to the E-value metric; see the “Methods” section for details. The size of each shape reflects the strength of the association, with the largest shapes reflecting associations in the top decile and the smallest shapes reflecting the lowest decile of effect sizes across beta and OR/RR estimates. Themes are sorted (top to bottom) by the proportion of results within each theme that were considered significant and robust, according to the test of trend. Within each theme, results are sorted (left to right) by the relative effect size from the maltreatment count 4–5 vs. 0. Number of total results per theme is listed in parentheses on the y-axis. Proportion of results that are significant and robust per theme and maltreatment count comparison are listed in gray text