Population size and birth rate

Current population estimates of women of reproductive age were obtained from the US Census. Age- and race/ethnicity-specific live birth rates were obtained from the National Vital Statistics System. The live birth rate among women with diabetes was assumed to be similar to the general population.

Diabetes prevalence

An estimated 2.9% (95% confidence interval [CI], 2.7–3.2) of women of reproductive age (defined as 18-44 years in the source publication) have diagnosed diabetes. This estimate is based on self-report among study participants in the nationally representative 2009 Behavioral Risk Factor Surveillance System (BRFSS). Another study estimated a 0.5% prevalence (no measure of dispersion was reported because of limited sample size) of undiagnosed diabetes among US women of reproductive age (defined as 15-44 years in the source publication) using the nationally representative 1999-2010 National Health and Nutrition Examination Survey (NHANES). We used sex-, age-, and race/ethnicity-specific data on diagnosed diabetes from the BRFSS to estimate the number of women of reproductive age with diagnosed and undiagnosed diabetes by selected age and race/ethnicity categories ( Table 1 ).

Impact of preconception care

A systematic review and meta-analysis of 12 observational cohort studies (n = 2502 participants) examined the health impact of PCC interventions among women with PGDM. Interventions in original studies included a combination of some or all of the following: pharmacologic or dietary glycemic control, blood glucose monitoring, counseling or education on the risks of diabetes in pregnancy, screening and treatment of diabetic complications, and contraception until glycemic control was achieved. The metaanalysis reported the frequency of the following adverse birth outcomes among women with PGDM who did not receive PCC services: 41.4% delivered preterm (n = 155/374 women in 4 original studies), 7.3% had children with birth defects (n = 110/1512 women in 11 original studies), and 4.4% had children who died in the perinatal period (n = 28/634 women in 5 original studies). Preterm deliveries were infants born before 37 weeks’ gestation; birth defects were not defined in the metaanalysis and the contributing studies used a variety of definitions. Because of overlap among outcomes in source publications (eg, a child with birth defects could have been born preterm), the metaanalysis did not report rates of adverse birth outcomes on a per-newborn basis. We are not aware of similar US studies with which to compare these estimates. A 1996-2004 population-based UK study of women with known PGDM (n = 1258 pregnancies) reported 3.7% perinatal mortality and 9.0% of live born children had birth defects; the preterm birth rate was not reported. However, no information was provided on women’s PCC status or glycemic control in the UK study; therefore, these results are not directly comparable.

Results of the metaanalysis indicated PCC was associated with statistically significant reductions in preterm delivery, birth defects, and perinatal mortality ( Table 1 ). There was no significant association reported between PCC and cesarean delivery, preeclampsia, spontaneous abortion, macrosomia, neonatal hypoglycemia, respiratory distress syndrome, or newborns’ small for gestational age status. In the metaanalysis, PCC was associated with a significantly higher risk of maternal hypoglycemia, although we did not evaluate maternal outcomes in the present study.

Lifetime cost of adverse birth outcomes

We used published estimates of the lifetime costs of preterm birth and 17 selected birth defects. These estimates included separately reported medical, special education, developmental services, and lost productivity costs for affected children. Estimated costs of selected individual birth defects (eg, tetralogy of Fallot, spina bifida) were reported separately in the source publication, and estimates did not include all types of birth defects. To estimate a general lifetime cost of birth defects, we calculated the average cost of individual defects weighted by the general population frequency of those defects reported in the source publication, which was based on California data. We estimated the lifetime cost of perinatal death as the value of lifetime lost household and labor market productivity. The lifetime costs of preterm birth and perinatal mortality were calculated in the source publications as present values with the recommended 3% annual discount rate applied to costs in future years. To approximate this discount rate for the birth defects estimate, we obtained the lifetime medical-only cost of birth defects discounted at 3% from the source publication’s author (via personal communication, Dec. 2013), to which we applied the ratio of lifetime total costs to lifetime medical-only costs reported with a 5% discount rate in the source publication. Costs are reported as 2012 US dollars based on the Gross Domestic Product deflator and are reported separately as total, medical and other services, and lost productivity costs.

Model

Among the US population of women of reproductive age (62.7 million ), we estimated the number with diagnosed diabetes (2.9% of the population ) and with undiagnosed diabetes (0.5% of the population without a diabetes diagnosis ) by age group and race/ethnicity. Among that estimated number of women with diabetes, we applied age- and race/ethnicity-specific birth rates (eg, 109.7 births per 1000 white, nonHispanic women age 25-29, 197.1 per 1000 Hispanic women age 25-29 ). Among that estimated number of births to women with PGDM, we estimated the number of adverse birth outcomes in the absence of PCC (41.4% preterm deliveries, 7.3% children with birth defects, 4.4% perinatal deaths ). To estimate the PCC-preventable health burden of adverse birth outcomes associated with PGDM, we applied average rates of relative risk reduction associated with PCC for preterm birth, birth defects, and perinatal mortality estimated in the recent PCC metaanalysis. We estimated the preventable cost burden for each outcome by multiplying the estimated number of outcomes by their associated lifetime cost estimates.

We used probabilistic analysis to inform model inputs for which sufficient data were available in source publications, and we used only reported point estimates for other inputs. We included a Pert distribution for the prevalence of diagnosed diabetes among women of reproductive age based on the 95% CI that accompanied the 2.9% point estimate reported in the recent BRFSS study. We used the same distribution to test the 95% CI for relative risk reductions associated with PCC estimated in the recent metaanalysis. Using Latin Hypercube sampling, the number of model simulations was determined by convergence criteria. Simulations ceased when there was a 90% chance that the mean value of each simulation was within 3% of its actual value. We reported mean estimates and 90% prediction intervals (PIs)—or, the range of outcome results that included 90% of model simulations—for all outcome measures. We separately reported results for women with diagnosed and undiagnosed PGDM. We separately reported the estimated preventable cost burden for each adverse birth outcome in terms of direct and indirect costs, as well as the sum of those costs, not taking into account the overlap across outcomes. Analysis was conducted with @RISK software for Microsoft Excel (version 6.2; Palisade Corp, Ithaca, NY).