Skip to main content
. 2022 Jan 18;1(1):e0000004. doi: 10.1371/journal.pdig.0000004

Table 1. PBC is well powered for comorbidity discovery on demographically complex datasets, unlike stratification.

Atrial Fibrillation and Acute Cerebrovascular Disease
Features PBC p-value χ2 p-value
N = 1,538,059 N = 95,407 N = 9,525 N = 1,538,05 N = 95,407 N = 9,525
no features 1e-31020 1e-1715 1e-203 1e-31020 1e-1715 1e-203
+sex 1e-31017 1e-1955 1e-215 1e-16657 1e-1125 1e-147
+age 1e-25448 1e-1589 1e-200 1e-1304 1e-88.3 1e-13.1
+ancestry 1e-14381 1e-628 1e-73.1 1e-15.72 1 1
+ethnicity 1e-11357 1e-806 1e-110 1e-12.25 1 1
+insurance 1e-11533 1e-771 1e-83 1e-2.68 1 1
+span 1e-11325 1e-698 1e-84.1 1e-1.75 1 1

Progressively smaller random samples were drawn from the Utah EHR corpus, such that each cohort is a subset of this larger precursor. N = the number of subjects in each cohort under consideration. Cells in the table contain p-values for the association between Atrial Fibrillation and Acute Cerebrovascular Disease (stroke), as calculated by PBC or χ2 (for stratification). P-values less than the Bonferroni corrected alpha (1e-9.5) are shown in light blue, while cells that do not pass the significance threshold are red. Stratum filters apply to the features’ column, row by row as follows: no filters, female, 50–59 years of age, white, non-Hispanic, commercial insurance, minimum of 2 years of medical history.