Fig. 1. Graphical representation of methods used in the population-level characterization (top) and the phenotype-specific characterization (bottom).

In the population-level characterization, we analyze diseases and conditions that occur in patients with at least one year of continuous observation. We calculate condition prevalence in women and men and compute risk ratios. We also calculate age of onset for each condition among all women and men, then estimate the mean difference in age of onset across genders. In the phenotype-specific characterization, across 112 phenotype definitions (using Crohn’s disease as an example), we aggregate the presenting symptoms prior to diagnosis in 1-, 3-, and 10-year increments for acute, mid-length, and long-term chronic conditions, respectively. We assess time to diagnosis for each presenting symptom across genders, computing the mean differences of the distributions. We also assess the diagnostic delay from first relevant symptom to diagnosis.