Table 2.
Multilevel mixed-effects multivariable linear regression model of maternal mortality – United States, 1997–2012
| Predictive Variable | Coefficient | 95% CI | P value | Share of change in maternal mortality attributable to factora |
|---|---|---|---|---|
| Adoption of 2003 Revision of the U.S. Standard Certificate of Death by 2011 | + 6.26 | [5.41–7.11] | <.001 | 31.1% |
| Proportion of women of childbearing age with BMI ≥ 30 | + 0.25 | [0.15–0.34] | <.001 | 31.0% |
| Proportion of births to women with diabetes | + 0.39 | [0.04–0.75] | .03 | 17.0% |
| Proportion of women of childbearing age not having completed high school/GED | + 0.17 | [0.05–0.30] | .005 | 5.3% |
| Proportion of births to women who attended fewer than 10 prenatal visits | + 0.07 | [0.01–0.14] | .03 | 4.9% |
| Proportion of births to African American women | + 0.20 | [0.14–0.27] | <.001 | 2.0% |
aPercentages sum to 91.3% of the time trend in maternal mortality attributable to the factors retained in the most parsimonious final model. Candidate predictor variables for the multilevel mixed-effects multivariable model that had P > 0.05 were removed from the most parsimonious final model shown here. BMI: Body Mass Index; GED: General Equivalency Diploma