To be able to answer the first and second research question, we will perform simple and multiple Cox regression. Although the aim of our study did not include investigating the covariates and their own association to the outcome, it is still relevant to conduct these analyses to further understand the analytical model and how the covariates relate to the outcome. Therefore, these are also included below.

Step 10: Simple Cox regression

First, we will do a simple Cox regression with depression as the x variable.

stcox i.depression if pop==1, noshow

The hazard ratio (HR) is 1.91. This means that for each unit increase in depression, the hazard of seeking BUP care increases. In other words, those who have had mothers with postpartum depression have higher risk of later seeking BUP care. The association is statistically significant, as reflected in the p-value (0.000) and the 95% confidence intervals (1.54–2.37).

Step 11: Simple Cox regression with all covariates

Next, we will do a simple Cox regression with mothereduc as the x variable.

stcox i.mothereduc if pop==1, noshow

Compared with those whose mothers have finished only compulsory education, those whose mothers have finished high school (HR=0.61), higher education (HR=0.43), and a postgraduate degree (HR=0.15), have increasingly lower risk of seeking BUP care. All hazard ratios are statistically significant (p=0.000).

Next, we will do a simple Cox regression with famtype as the x variable.

stcox i.famtype if pop==1, noshow

The output shows that children with absent fathers (aka single mothers) have higher risk of seeking BUP care. Even these results are statistically significant (p=0.024).

Next, let us do a simple Cox regression with sex as the x variable.

stcox i.sex if pop==1, noshow

Girls appear to have a somewhat lower risk (HR= 0.93) than boys of seeking BUP care. This association is, however, not statistically significant (p=0.351).

Finally, we will do a simple Cox regression with support as the x variable.

stcox i.support if pop==1, noshow

As reflected by the hazard ratios, the more social support mothers had during the child’s first year, the lower the risk for the child later seeking BUP care. Compared with those who had low support, the risk was 0.77 for those with medium support, and even lower (HR=0.49) for those with high support. Both associations are statistically significant (p=0.023; p=0.000).

Note For more review, re-visit Simple Cox regression.

Step 12: Multiple Cox regression with control variables

Now, we will perform a multiple Cox regression with depression, mothereduc, famtype, sex, and support as independent variables.

stcox i.depression i.mothereduc i.famtype i.sex i.support if pop==1, noshow

When we control for all variables, the relationship between postpartum depression in mothers and the risk of psychiatric problems in children changes from HR=1.91 to HR=1.56. The association is still statistically significant (p=0.000).

Note For more review, re-visit Multiple Cox regression.

Note Some things to think about after performing the multiple Cox regression with control variables: – Is it reasonable to view all other variables (mothereduc, famtype, sex, and support) as confounding variables, or can some of them be mediators? – Is there a confounding variable that you believe is missing in the model? – Is there any other mediator variable that you think could explain the relationship between postpartum depression in mothers and the risk of psychiatric problems in children?

Step 13: Kaplan-Meier curves based on multiple regression

Next, we will produce Kaplan-Meier curves (with failure function and confidence intervals) for each category of depression, based on the multiple regression model. Adjust the graph to make it more readable.

sts graph, failure ci noorigin ylabel(0(0.01)0.13) xlabel(1(1)17) by(depression)

In children with mothers with postpartum depression, the hazard increases.

Summary After performing simple and multiple Cox regression, we should now be able to answer the first and second research question: 1. Is there a relationship between postpartum depression in mothers and psychiatric problems in children? Yes. There is an association between postpartum depression in mothers and psychiatric problems in children. 2. Is this relationship explained by the mother’s level of education, family type, child’s gender, and/or social support? Controlling for maternal educational level, family type, the sex of the child, and the mothers social support during the child’s first year led to a reduction of the estimate, but the conclusion from the simple model remained the same.