Over the past decade, many professional organizations have called for universal screening for depression in pregnant and postpartum women. The goal of screening is to identify women with depression and start treatment early. While this is certainly an important aspect of mental health care for pregnant and postpartum women, optimal screening of this population should include identification of women at increased risk for perinatal mental health conditions prior to symptom onset.
To date, epidemiologic studies have revealed that the strongest predictor of risk for perinatal depression is a history of mood or anxiety disorders before pregnancy. Other risk factors modulate risk, including a history of childhood adversity, recent life stressors, intimate partner violence, and general physical health. Although there are data to support the validity of these risk factors at the population level, it is difficult to use these factors when trying to estimate an individual’s risk of perinatal psychiatric illness. In addition, there are also risk mitigating factors (eg social support) and it is often difficult to take these beneficial factors into account in our risk calculations.
Clinical Risk Prediction: Consider Mrs. A
Take the example of Mrs. A. She is 35 years old, has been married for 6 years and works as an architect. She battled anxiety and depression at the age of 14, after her father died unexpectedly (heart attack). For about four years, she was in regular psychotherapy and took 50 mg of sertraline. She stopped treatment when she started college and has had no recurrence of her symptoms and has not taken any medication for depression or anxiety. She has been in psychotherapy for the past several years to help deal with daily stressors.
She is in good health. She reports drinking 1-2 glasses of wine over the weekend, with no recreational drugs. She has been taking oral contraceptives for about 10 years. She has no history of PMS or PMDD.
Her family history is notable for generalized anxiety in her mother and older sister; for both, the disorder arose after the death of Ms. A’s father. She is not aware of any mental health problems on her father’s side of the family.
Based on these factors, we might consider Ms. A to be at relatively low risk for perinatal depression and anxiety. She has a remote history of depression and anxiety, but that was situational and she has been symptom free and medication free for 17 years. On the other hand, she does have a history of childhood adversity (father’s death) and a family history of anxiety (arising after her father’s death). What weight should we give these risk factors?
Using Big Data to estimate risk
As physicians, our ability to estimate risk in an individual patient is fair and often subjective. This is where big data can come in handy. Every time a patient visits a health care provider, a large amount of data is collected: sociodemographic information such as marital status, employment, and educational level; medical history; data from standardized questionnaires; Lab tests; Vital signs; prescription drugs. It is impossible for doctors to take and analyze every piece of data to generate an accurate estimate of risk. But powerful computers can.
To effectively collect and use all of the potentially valuable information contained in the medical record, researchers are turning to machine learning to sift through vast amounts of data and determine which factors are most relevant to predicting risk for perinatal psychiatric illness. The goal is to create an algorithm that reliably predicts the risk of each individual. This approach allows us to understand which factors are the strongest predictors of risk, and may also help identify other factors that we do not yet know about.
In a recent review, Cellini and colleagues (2022) identified 11 studies focusing on the identification of predictors of postpartum depression (PPD) using machine learning techniques. In these studies, the investigators evaluated a wide range of potential predictors measured during pregnancy or at the time of delivery. Using machine learning, they used relevant data to generate an algorithm that could be used to predict an individual’s PPD risk.
These studies evaluated a wide range of potential risk factors, including psychiatric history (before and during pregnancy), sociodemographic variables (eg, age, marital status), obstetric variables (eg, pregnancy complications), and pediatric variables (eg, gestational age, birth). weight). Three studies used biological variables, in the form of blood, genetic, and epigenetic predictors. None of the studies used imaging techniques.
All studies achieved a precision or area under the curve (AUC) greater than 0.7. An ROC value greater than 0.7 is considered to be reasonable performance for a model to be used to predict a particular outcome, such as the occurrence of pPD. (AUC ranges in value from 0 to 1. A model whose predictions are 100% wrong has an AUC of 0; a model whose predictions are 100% correct has an AUC of 1.0.)
The most powerful risk predictors included a history of depression or anxiety before pregnancy and depressive symptoms or anxiety during pregnancy. Several studies indicated that the prescription of antidepressants at any time in a woman’s life is one of the strongest predictors of risk. Other important predictor variables included smoking, age (younger or older), pregnancy complications, increased use of health care services during pregnancy, increased number of emergency room visits during pregnancy, BMI pregestational, the lowest weight of the baby at birth, the shortest duration of gestation, the sex. of the child, and recent stressful life events.
All these variables have been identified as risk factors in previous epidemiological studies. What the machine learning studies add is a more nuanced estimate of the weight to be given to each of these variables. For example, a higher BMI may increase the risk of PPD, but it does not appear to be as strong a predictor of postpartum risk as having a history of depression. Because all of these calculations are done by a computer, we can use multiple variables simultaneously to estimate risk.
Thinking in the future
Let’s go back to Mrs. A. Although her mood was relatively stable during the pregnancy, the pregnancy itself was difficult. She had fairly severe nausea for most of her pregnancy, which made it difficult for her to gain weight. Her activity was limited and she was unable to exercise regularly. Although she felt much better physically after the birth of her daughter, her anxiety, especially related to the well-being of the baby, was very high. Breastfeeding was difficult, and her anxiety about her feeding made it difficult for her to sleep at night. At her 6-week postpartum visit, she scored 24 on the EPDS, a score consistent with severe postpartum depression.
Is this something we could have predicted?
Maybe, but probably not. Based on what we have learned from the machine learning studies described above, it appears that the use of an antidepressant medication at any time in a woman’s life, even if used many years ago, may be a predictor of depression risk. postpartum more potent than other factors. But the risk estimate is subjective, and given the situational and remote nature of her prior episode of depression, her consistently high level of functioning, and the fact that she has been well and off medication for the past 17 years, Ms. A did not consider her to be at significant risk of postpartum depression.
As mental health providers, we see many patients who are at high risk for recurrent psychiatric illnesses and follow them more closely. But in the general population, there are many women who are at increased risk for PPD, even though they may not currently be receiving treatment for a psychiatric illness. Although the 6-week assessment identified Ms. A as having postpartum depression, perhaps we could have made the diagnosis earlier, or perhaps we could have provided additional support to minimize her risk of depression.
Imagine if we could use machine learning to improve the care we provide and make risk estimates more reliable. Maybe when Mrs A visits her OB during her next pregnancy, she answers a few questions, a magical risk prediction algorithm does its thing, and we get some sort of number or score that quantifies Mrs A’s risk of depression. perinatal. or anxiety. Then she and her obstetrician will discuss what options are available to lower her risk: for example, perhaps a course of mindfulness-based cognitive therapy or the introduction of an antidepressant after delivery. (Okay, this may sound like a stretch, but this kind of precision medicine approach is already being used to make treatment decisions for patients with breast cancer.)
These studies are preliminary but very exciting. In the future, we will need to test these predictive models in a variety of environments. A model designed to predict risk in Iowa City may not work as well in Beijing. Similarly, a model trained with publicly insured people delivering in an urban hospital setting may not work the same way in a mostly privately insured suburban obstetric practice. Factors including race, ethnicity, socioeconomic status, community support, and access to health care are likely to vary from site to site. There may also be cultural differences to consider when generating these models, such as a preference for a child of a particular gender. However, it will be exciting to see how precision medicine develops within the field of perinatal psychiatry.
Ruta Nonacs, MD, PhD
Cellini P, Pigoni A, Delvecchio G, Moltrasio C, Brambilla P. Machine learning in the prediction of postpartum depression: a review. J Affects Disorder. July 15, 2022; 309:350-357.
Hochman E, Feldman B, Weizman A, Krivoy A, Gur S, Barzilay E, Gabay H, Levy J, Levikron O, Lawrence G. Development and validation of a machine learning-based postpartum depression prediction model: a nationwide cohort study. depressive anxiety April 2021; 38(4):400-411. doi: 10.1002/da.23123. Epub 2020 Dec 7.
Yang ST, Yang SQ, Duan KM, Tang YZ, Ping AQ, Bai ZH, Gao K, Shen Y, Chen MH, Yu RL, Wang SY. The development and application of a prediction model for postpartum depression: optimizing risk assessment and prevention in the clinic. J Affects Disorder. 2022 January 1; 296: 434-442.