Improving Disease Measurement in Pregnant Women: A Simulation Approach

Monitoring disease incidence in pregnant women is important because it can inform vaccine and other health policies during pregnancy, and can improve neonatal outcomes by increasing understanding of disease in this population. Incidence estimates, by definition, include implicit time-period components; however, surveillance systems report on pregnancy status at a single point in time (usually during case ascertainment). One method historically used for estimating incidence in pregnant women includes live births, which only accounts for about two-thirds of all pregnancies. This approach does not account for induced abortions (about one-sixth of all pregnancies) and other fetal losses which include early- to late- fetal mortality (an additional one-sixth of all pregnancies). More recent methods have incorporated richer algebraic approaches which better capture the person-time component of the denominators. Our Monte Carlo simulation takes this recent improvement on earlier methods one step further. We substantiate the assumptions that back the simulation by current data found in the literature. In the simulation, we then apply these assumptions and randomly compute conception date, assign a random outcome (live birth, induced abortion, or fetal loss), and calculate pregnant-days-at-risk (PDAR). After running a large series of trials (approximately 1,000), we were able to calculate the PDAR as the improved denominator for the incidence of streptococcal disease in pregnant women. We believe these new denominators, which account for the time-period component of risk, provide better estimates than naïve approaches historically used.