This is not a post about impact evaluations during a time of global pandemic or what lessons can be learned in this current crisis from past events, such as the Ebola outbreak in 2014. Rather, it is about a paper on the burden of depression and the economic gains from a low-cost, community-based mental health intervention in a low-income setting. (…But perhaps it offers some ideas for prospective interventions after this time has passed.)
Victoria Baranov, Sonia Bhalotra, Pietro Biroli, and Joanna Maselko’s paper examines the impact of cognitive behavioral therapy (CBT) on women with perinatal depression in Pakistan.
Why study depression? It is not just a high-income country problem: Major Depressive Disorder is estimated to affect about one in eight people worldwide. Mental health has been linked to impulsivity and inconsistent intertemporal choices, lower aspirations, and lower productivity, with well-being implications for adults and for their children.
Why women? While we don’t have a lot of the data, estimates point to depression rates that may be twice as high for women than men. And why perinatal depression? Maternal depression (which is both pre and post birth) is increasing recognized as a public health issue, for both the mother and infant.
Why Pakistan? I am not sure why. The paper gives little background on the context of rural Punjab, mentioning only that women’s empowerment is low and limited by rigid gender norms, and most women “are confined to their homes, do housework for the extended family, and are excluded from decision-making”. Among the myriad of challenges facing women in rural Pakistan, the paper does not “make a case” for prioritizing addressing perinatal depression. Nonetheless, its insights on the impact of low-cost CBT delivered by female community health workers and being, as the paper states, “… one of the world’s largest psychotherapy interventions” with a 7-year follow-up put it at the top of my reading pile.
The Thinking Healthy Program was implemented across 40 communities in rural Pakistan – half assigned as treatment and half as controls. 903 pregnant women (26% of all pregnant women screened) who were clinically diagnosed as depressed were included in the trial. All of them received 16 home visits from community health workers for routine maternal and health services, starting in the third trimester of pregnancy and ending when the children were ten months old. In the treated communities, women additionally received a CBT component. This, in short, used techniques of active listening, collaboration with the family, and guided discovery of healthy thinking. An additional sample of women not suffering from prenatal depression were added. Control mothers received an equal number of visits by the health worker but without CBT and at a greater frequency (so presumably their visits ended earlier than the 9 months after birth for treatment women).
Details about treatment/control balance, sample attrition, and ITT & ATT interpretation are discussed in detail in the paper (and on-line appendices). The paper identifies the causal effect of the CBT intervention treating depression – rather than the causal effect of depression. The authors do not use the randomized assignment as an instrument for depression since the treatment could have had direct impacts on outcomes. Finally, power calculations for the 7-year follow-up focused on an IQ measure for children, as opposed to, for example, maternal outcomes. This choice and the implications are, unfortunately, not discussed in the paper.
At a cost of $10 per woman, what is the impact 7 years later on mothers and children? Depression rates were 5 percentage points lower, although smaller than the 1-year follow-up (39 percentage points). This is driven by “spontaneous” recovery in the control group, consistent with other evidence that major depression episodes can subside over time when untreated. The benefits are larger for women who were pregnant with girls at baseline. And those benefits grew from the 1-yearfollow-up to 7 years later.
The CBT intervention significantly increased mothers’ financial empowerment (measured as reported control over spending), monetary parental investments in the child, and time-intensive parental investments. Again, as with mental health, the benefits are larger and only significant for the mothers of girls – resulting in smaller gender gaps for daughters. There were no effects on parenting style (measured using the Parenting Practices Inventory) or fertility.
The estimated effects of the intervention on child outcomes (physical development based on 8 indicators, cognitive development with 9 components, socio-emotional based on 11 components, and sibling survival) are small and noisy. The authors conclude that the study is underpowered to detect reasonable effect sizes on child development (partly due to high attrition). But they are not able to rule out that the additional time and money investments made by treated mothers were not effective at promoting child development.
There are many possible mechanisms by which CBT might impact outcomes. Several intermediate outcomes are explored: husbands’ labor earnings (reducing time to care for a depressed wife or aid in child caregiving), mothers’ physical health, relationship quality with husband and mother-in-law, co-residence of child’s grandparent, and perceived social support. Though there were short-term (1-year follow-up) benefits on the quality of the relationship with the husband and the probability of co-residence with the grandmother of the child, the effects were small and were not sustained in the longer run. By the 7-year mark, only perceived social support was higher for treated women. The grandmother impacts were difficult for me to interpret. Treatment effects were larger for pregnant women without a grandmother presence at baseline. It was not clear which direction I’d expect this to go (especially for mothers-in-law!).
Three final thoughts on this paper, which covers a lot more ground than I can summarize in a blog. First, that results were concentrated among women who gave birth to daughters struck me as intriguing but woefully unexplored in the paper. I would have liked more discussion about the finding that the mothers of girls struggling more with depression (not recovering after 7 years), pulling in literature especially from Pakistan and from outside economics. Secondly, despite the claims in the paper, I do not think this paper brings evidence that additional resources in the hands of mothers (versus fathers) is more likely to be directed toward children. Fathers are absent in this study. Moreover, the previous research on this is, from what I can find, very scant. Here I point to the evidence from cash transfer programs where there are none that I have found that show a greater/improve impact for children from giving a transfer to the women versus the man (readers are welcome to send me studies within the past 15 years I have missed). Finally, I would have liked at least a small discussion about other intervention options for perinatal depression in low or middle income contexts besides CBT and not just passing mention of antidepressants (see this 2018 review in the Asian Journal of Psychiatry).