Over one-third (39%) of mothers were covered by public insurance, about one-third (33%) were covered by private insurance, and the other 28% had no insurance at the time of the one-year follow-up interview (). Results from the χ2 and F-tests indicate that all characteristics other than multiple birth, child health status, and percent below poverty in the census tract differed significantly across the insurance groups (at P <.05).
Sample characteristics (proportions unless indicated otherwise)
While maternal mental and physical health conditions that predate the birth vary by insurance status at follow-up, the difference is most striking for mental illness. Twice as many mothers with public insurance have a history of diagnosed mental illness (14%) as do those with private insurance (7%).
Foreign-born mothers, who account for about 15% of the sample, comprise a sizable fraction of the uninsured (24%) and much smaller shares of the privately (14%) and publicly (9%) insured. The percentage of households in the mother’s census tract with income under the poverty line is lowest in the privately insured group (13% for the privately insured group compared to 24% for the publicly insured group and 20% for the uninsured group). First births are more common among privately insured than among publicly insured and uninsured mothers.
Among mothers with public insurance at follow-up, 93% were unmarried at the time of the birth and about 40% of those who were unmarried lived with the child’s father. Among mothers with private insurance at follow-up, half (51%) were unmarried when the child was born and 57% of those who were unmarried lived with the child’s father. About 15% of mothers with no insurance were married and over half (55%) of the unmarried mothers with no insurance resided with the child’s father at the time of the birth.
One-fifth of the children in the sample were coded as having been diagnosed with a health condition (20% of the publicly insured group, 19% of the privately insured group). The prevalence of self-reported suboptimal health status and the mean depression scores were higher for fathers in the “mothers with public insurance” and “mothers are uninsured” groups than for those in the “mothers with private insurance” group.
The fathers associated with mothers in the public and uninsured groups were less likely than those associated with mothers in the private insurance group to have served in the military, to have been employed, and to have received health insurance through an employer. Other than for employment, the associations are quite large. For 13% of the mothers who were privately insured at follow-up, the father had served in the military; the corresponding figures for publicly insured and uninsured mothers were 6% and 8%, respectively. For two-thirds (64%) of mothers with private insurance at follow-up, the father had insurance through an employer at the time of the birth; the corresponding figures for mothers with public and no insurance at follow-up were 31% and 35%, respectively.
Both mental and physical health status at follow-up differed by insurance status. Maternal depression or anxiety after the birth of the child (this could include postpartum depression) was much more prevalent among mothers who were covered by public insurance (15%) or uninsured (17%) than among mothers with private insurance (9%). Similarly, mothers who reported poor physical health at the follow-up were more likely to have public insurance (41%) or to be uninsured (46%) than to have private insurance (29%). Mothers who reported in the follow-up interview that they had a disability that limits their ability to work were more likely to have public insurance coverage than to have private or no insurance. Mothers with public insurance at follow-up were the least likely to have gotten married, the least likely to be employed, and the most likely to have had a subsequent pregnancy between the baseline and follow-up interviews.
When controlling for prenatal physical health status and other baseline covariates in the basic model (), mothers with a prenatal history of mental illness were half as likely as those without a history of mental illness to have private insurance coverage (vs. no insurance) at follow-up (RRR =.51), but prenatal mental illness was not significantly associated with having public versus no insurance.
Multinomial logit estimates of health insurance status of mothers with one-year-old children (N = 2956)
Mothers who were high school graduates and those with private insurance at baseline were more likely than those with less than a high school education and those without private insurance at baseline to have private insurance at follow-up. In fact, mothers with private insurance at baseline were about six times more likely to have private insurance at follow-up than to be uninsured. The poorer the mother’s census tract at the time of the birth, the less likely that she had private insurance and the more likely that she had public insurance 12–18 months later. If the birth was the mother’s first, she was significantly less likely to have public insurance coverage than to have no insurance (RRR =.69). The parents having been married at the time of the birth more than doubled the probability that the mother had private insurance, and decreased the probability that she had public insurance, at follow-up. Finally, mothers who are foreign-born and who are Hispanic were significantly less likely than their native-born and non-Hispanic counterparts to have public or private insurance coverage at follow-up.
The health of the child at birth was not significantly associated with the mother’s insurance status one year after the birth. However, the father’s suboptimal physical health at the birth had a strong negative association with the mother’s private insurance coverage one year later (RRR =.73). The father’s CES-D depression score was associated with public, but not private, health insurance of the mother; the mother’s likelihood of having public (vs. no) health insurance increased as the father’s risk for depression increased. The father having had health insurance through an employer at baseline significantly increased the likelihood that the mother had private insurance at the time of the follow-up interview (RRR = 1.71).
The follow-up characteristics, which are potential mediating variables, are strongly associated with both private and public insurance in the expected directions (augmented model). However, there are some differences depending on whether the outcome is private or public insurance. Mothers who married between the two survey waves and those who were employed were far more likely than those who did not marry and were not employed to have private health insurance. Being disabled and becoming pregnant again, both of which are related to eligibility, nearly doubled the likelihood of public insurance. Working, which should reduce eligibility by increasing earnings, decreased the likelihood of public insurance by about 50%. Mothers who screened positive for depression and/or anxiety at follow-up were less likely than those without symptoms of depression or anxiety to have either kind of insurance at follow-up, controlling for prenatal mental illness and the other covariates and potential mediators.
It is clear from the above discussion that the mediating factors are strongly related to the mother’s private health insurance status at follow-up. Including them in the model somewhat reduces the magnitude of the effect of prenatal mental illness on private insurance (RRR increases from 0.51 to 0.64), but that estimate remains statistically sig-nificant at the 5% level. The finding of no significant association between prenatal mental illness and public insurance coverage one year later is not sensitive to the inclusion of the follow-up measures.
We estimated auxiliary models with sample restrictions, different variable constructions, or alternative model specifications to further explore the findings and assess the sensitivity of the results (not shown in tables). Because pregnancy is an important route to eligibility for public insurance, we estimated the basic and augmented models for the subset of mothers who did not have a pregnancy between the baseline and follow-up interviews. For this restricted sample, the associations between prenatal mental illness and public (vs. no) coverage were negative and almost statistically significant in the basic model (RRR =.78, P =.11), and were of similar magnitude and statistically significant in the augmented model (RRR =.73, P =.07). Models that excluded fathers who did not complete baseline interviews (instead of using indicators for missing data) produced findings virtually identical to those in . The estimates were insensitive to recoding of the physical health variable to capture additional pre-existing conditions (which were rare) or using alternative measures of poor maternal physical health (e.g., number of pre-existing conditions or other subsets of conditions). Finally, unadjusted multinomial logistic regressions for each independent variable revealed, almost without exception, associations very similar to or larger than the corresponding multivariate estimates in the basic model of . For two of the follow-up (mediating) characteristics (got married and new pregnancy), the estimated effects were larger in the augmented multivariate model.
We also explored whether specific independent variables of interest modify the effects of prenatal mental illness (results not shown in tables). First, we estimated a variant of the augmented model that included the following instead of the prenatal and follow-up mental illness variables: prenatal mental illness but no mental health problem at follow-up, mental health problem at follow-up but no prenatal mental illness, and both prenatal mental illness and follow-up mental health problem. Having a mental health problem at either wave reduced the likelihood of having either private or public insurance. However, the effects were strongest for women who had a mental health problem at both waves; those mothers were about 50 percent less likely to have public insurance (RRR =.47, P <.001) and 72 percent less likely to have private insurance (RRR =.28, P <.01) than to be uninsured at follow-up.
Second, we explored potential modifying effects of maternal baseline characteristics that are known barriers to health insurance and that were found to be significant factors in our analyses reported in —having less than a high school education, being foreign-born, having a Medicaid-financed birth, being unmarried, and living in a census tract with 25% or more of the families below the poverty line. For each characteristic, we estimated (basic) models as described above (i.e., including variables for mental illness but not having the characteristic, having the characteristic but not having mental illness, and having both mental illness and the characteristic). We found that for all characteristics other than being Hispanic, having the barrier compounded the adverse effects of mental illness (results not shown).