Maternal depression: Technology enabled self screening in real time

Abstract

The first step toward providing treatment, is getting the right diagnosis in real time; before it is too late. Without this, resource deployment may appear to be comparable to the scale and scope of the problem, while in reality it may just be a drop in the ocean. Maternal depression, during pregnancy is a debilitating risk to both the mother and the child, but the bigger problem is, it goes unnoticed, undetected, and therefore untreated. If mobile technology can be deployed to screen for depression in real time by the pregnant mother herself, it will go miles in creating a HOPE for health.

Antenatal depression (AND) is a form of clinical depression that can affect a woman during pregnancy, it is also the leading cause of postpartum depression if not treated in time (Wilson, 2013). Researchers have found AND prevalence to range between 15 and 65%, and the pooled prevalence has been found to be greater in low and middle income countries as compared to high income countries (Dadi et al., 2020). World Bank officials report that there is still a poor coverage of data driven research from low income countries (The World Bank Group, 2016).

However the problem is not just that prevalence of AND is not well and widely researched and documented, but that due to the stigma attached with depression in many countries, and also due to lack of resources for treating the women thus diagnosed, the treatment gap is very wide.

The authors of the current research have tried to contribute a methodological literature on a self screening tool for antenatal depression via a mobile app, which can be used by a participating pregnant mother in any country and in any given language. It also gives a theoretical plan on how technology can be leveraged for treatment for depressed mothers - by connecting with health care providers in real time, and through personalized Artificial Intelligence (AI) enabled treatment options.

Background

As per WHO, over 300 million people (4.4% of the world population) are suffering from depression (World Health Organization, 2017). The researchers of the World Mental Health Survey- an initiative of WHO, translated this figure as 1 in 20 people with atleast one reported episode of depression (Kessler et al., 2009).

However the spread of depression definitely has a gender bias! As per the WHO report – the “burden of depression is 50% higher in females than in males” (WHO, 2011), and the life time risk for major depressive disorder is 20% to 25%, approximately, which is twice the figure of 7% to 12% found in men. While the individual country statistics may differ yet the fact is that more women than men are depressed whether it is a high, middle or a low income country (Dadi et al., 2020; US Department of Health & Human Services, 1993). In 2010, the Global annual prevalence was found to be 3.2% for men, and 5.5% for women, which also means a 1.7 times greater incidence of depression among women (Baxter et al., 2014; Whiteford et al., 2013).

While women may be prone to general depression like men; there are two types of depression unique to women, which may add to the overall burden of the disease. These are Perinatal Depression and Premenstrual Dysphoric Disorder (PMDD) (Harvard Health, 2020). The latter is a severe form of pre menstrual syndrome or PMS, and has been categorized as a “depressive disorder not otherwise specified” in the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders), (American Psychiatric Association, 1994). It begins shortly after ovulation and ends at menstruation, and is characterized by the symptoms of anger or irritability, anxiety and panic attacks, depression and suicidal thoughts, difficulty in concentrating, fatigue and low energy, food cravings, headaches, insomnia and mood swings (Mayo Clinic, 1998).

The former, also called as “Maternal Depression” or “Perinatal Depression” (means – about or around pregnancy) occurs amongst pregnant and delivered mothers. This affects mothers (up to 12 months postpartum) and mothers to be and encompasses prenatal/antenatal (AD)/antepartum depression, postnatal (PD)/postpartum depression, and postnatal/postpartum psychosis. The DSM 5 defines perinatal depression as a “major depressive episode (MDD), with peri partum onset,” where “peri partum” onset means, the onset of symptoms during pregnancy or in the 4 weeks following delivery (American Psychiatric Association, 2013). Sometimes peri partum depression gets confused with “baby blues,” whereby symptoms in both appear to be the same (Marcus & Heringhausen, 2009; Wisner et al., 2002). Almost 70% new mothers may experience baby blues, but it generally resolves within two weeks of delivery (Marcus, 2009). Unlike the PMDD which is limited to a defined period only, both the forms of maternal depression- whether they be antenatal or post natal are a chronic form of depression lasting for more than two weeks and for some individuals, as long as two years. The conditions associated with the women suffering from Perinatal depression can range from a persistently sad, anxious, or “empty” mood, irritability, feelings of guilt, worthlessness, hopelessness, or helplessness, loss of interest or pleasure in hobbies and activities, fatigue or abnormal decrease in energy, feeling restless or having trouble sitting still, difficulty concentrating, remembering, or making decisions, difficulty sleeping (even when the baby is sleeping), awakening early in the morning, or oversleeping, abnormal appetite, weight changes, or both, aches or pains, headaches, cramps, or digestive problems that do not have a clear physical cause or do not ease even with treatment, trouble bonding or forming an emotional attachment with the new baby, persistent doubts about the ability to care for the new baby, thoughts about death, suicide, or harming oneself or the baby (NIH, 2022). It also affects the new born with developmental, emotional and attachment problems, poor academic performance, malnutrition, respiratory disorders and a higher risk of the infant developing mental health disorders later in life (Marcus, 2009; Raisanen et al., 2014).

Perinatal depression is an important phenomenon for healthcare policymakers to focus on. It is the least studied and most untreated disorders amongst all disorders of depression (Gelaye et al., 2016). It is estimated that more than two billion babies will be born over the next fifteen years - signaling a 2% increase over the babies born in the last 15 years (United Nations DESA. Commission on Population and Development, 2015). With the reproductive age of a woman being the phase of highest possible risk for developing depression (Dadi et al., 2020), it is strongly advised that policymakers understand the prevalence, symptoms and causes of maternal depression since it’s a disorder that has a huge significance for public health (Chisholm et al., 2019). This underlying emphasis is because the stakeholders of Maternal and Child Health (MCH) programs have still not included preventive and treatment oriented initiatives to address maternal mental health even when it’s negative impact is now evidence based (Rahman et al., 2008). Worse still Governments of the majority of the countries have not allocated more than 1% of their total health expenditure on mental health initiatives and implementation (The World Health Report, 2001).

While Postnatal Depression has been extensively studied all over the world (WHO, 2009), by comparison, very little research has been targeted at Antenatal Depression (Heron et al., 2004). Scholars have shown that anxiety and depressive symptoms are actually higher during pregnancy than they are after childbirth (postpartum period), thereby signifying the importance of antenatal mental health problems (Andersson et al., 2003; Bennett et al., 2004; Evans et al., 2001; Gavin et al., 2005). Some of the known causes leading to Antenatal depression are life stress, lifetime depression history, lack of social support, domestic violence, unintended pregnancy, history of drinking and substance abuse, low self esteem, inadequate nutrition, poor education status of mother, unemployement, pressure of delivering a male child, advancing age of the pregnant mother, and candida infection (Bergner et al., 2008; Blaney et al., 2004; Da Costa et al., 2000; Glazier et al., 2004; Grant et al., 2008; Holzman et al., 2006; Jesse et al., 2003; 2005; Morse et al., 2000; Records & Rice, 2007; Ritter et al., 2000; Rubertsson et al., 2003; Zayas et al., 2003; Zelkowitz et al., 2004).

With the status quo on resources, non mainstreaming of mental health, stigma associated with maternal depression, and lack of implementation of mental health policy – researchers should explore on how technology can be leveraged in the new era of personalized medicine, such that technology may be used as a tool for connecting healthcare providers with the individuals who require personalized preventive and curative care.

What technology really can do is provide solutions to a very large audience. But specifically in the domain of healthcare and in particular mental health, technology is the boon which can offer solutions in the home of the individual actually suffering the pain of not only the disease but also the social stigma attached with seeking care publicly. The authors of this article explore this concept. They propose that a simple self screening tool for antenatal depression be made available on a mobile app, for use in real time by pregnant mothers all over the world. This tool will have to be translated and validated for different languages for wider reach. While translation and validation of some languages is already available (Bunevicius et al., 2009), the same is required for other spoken languages in the world because different cut off scores are recommended for each language due to timing of administration of the EDS, as well as the differences in expression in different cultures and languages (Cox & Holden, 2003; Gaynes et al., 2005). Similar validations have been done for Dutch (Bergink et al., 2011), French (Adouard et al., 2005), Maltese (Felice et al., 2006), Spanish languages (Garcia-Esteve et al., 2003; Vázquez & Míguez, 2019), which give a clear cut off score that helps in screening for depression, and provides impetus to the mental health care providers on the next step for treatment.

The authors have in their current research validated the full 10 item tool in Hindi language – which is the third most spoken language globally and is spoken by 637 million people (Eberhard et al., 2020); as well as a smaller 2 questionnaire tool for capturing the prevalence of antenatal depression for Hindi speaking populations. This work by the authors showcases how other language tools can also be validated, and how this repository of tools, can be used for Global achievement of the objectives set out by the authors in writing this article.

Method

A study was conducted with Ethical clearance from the Institutional Review Board, IIHMR University. All ethical guidelines were adhered to while conducting in-person interviews. An informed consent was obtained from all research participants.

The authors ensured that the study design was facility based, observational, quantitative, cross – sectional, to determine the point prevalence, of Antenatal Depression (AND), by a depression screening tool which is filled by the pregnant mother herself, and which takes less than 5 minutes to be answered. All pregnant women who had an Ante Natal Clinic (ANC) visit or a follow up visit at public and/or private health facilities, for ANC were the “source population”, whereas those pregnant women who came to ANC visit or follow up during the defined study period and who were randomly selected were the research participants. The study was conducted in Jaipur city, in Jaipur district in the state of Rajasthan- which is the largest State of India, and is primarily a Hindi speaking belt (Census of India, 2011).

The authors calculated the Sample size by the formula to determine samples for a cross sectional study (Pourhoseingholi et al., 2013). Researchers, in a study in 2019, (Mahendran et al., 2020) conducted a systematic review and meta-analysis of the prevalence of AND in South Asia – where a meta-analysis refers to the statistical analysis of the data from independent primary studies focused on the same question, which aims to generate a quantitative estimate of the studied phenomenon (Mikolajewicz & Komarova, 2019). Through this study, the researchers showcased the pooled prevalence of India at 17.74% (95% C.I. 11.19 TO 26.96). For this reason, prevalence was assumed at 17.74%, or p = 0.1774, and q was thus 1-.1774 = 0.8226. Some authors also recommend that the best precision for prevalence around 20% is 0.04 (Pourhoseingholi et al., 2013). The authors thus arrived at a sample size of 350 research participants. Non response rate was assumed at 10%, thus taking the desired sample size to 385 research participants.

The screening tool used for identifying depression was the EPDS (Edinburgh Postnatal Depression Scale). To understand how technology can impact maternal depression, we will focus on the evolution, value and impact of this tool – the EPDS and see how this can be built into mobile technology to provide choice based access for treatment of maternal depression, to the pregnant mothers.

The Edinburgh Postnatal Depression Scale (EPDS) is a 10-item questionnaire developed for identifying women who have post partum depression. It was developed in 1987, by Scottish Health Centers (Cox et al., 1987). The original EPDS was designed and researched and validated for English language on a sample of 84 mothers, by using the Research Diagnostic Criteria for depressive illness obtained from Goldberg’s Standardized Psychiatric Interview. The scale can be completed in about 5 minutes and can be scored by a simple method of adding all scores. Responses to the 10 items/questions on the EPDS are scored 0, 1, 2 and 3 based on the seriousness of the problem, questions numbered 3, 5, to10 are reverse scored. A higher score indicates more depressive symptoms.

One of the crucial aspects of using this form is to determine the cut –off score at which “depression” will be identified. A cut off of ≥ 12 was indicated by researchers on the original tool, to identify depression in English speaking countries as per the validation done for English language. For all languages the cut off score has to be calculated afresh.

While the original EPDS was applied for assessing post partum depression, researchers can also apply it for depression screening during pregnancy (Bergink et al., 2011; Felice et al., 2006; Gibson et al., 2009; Ryan et al., 2005). In that case it may be referred to as EDS or Edinburgh Depression Scale. The EDS is the most widely validated tool for Antenatal and Postnatal Depression (Murray & Cox, 1990; Ryan et al., 2005), and it’s estimates are found to be statistically equivalent to structured clinical interviews (Bennett et al., 2004).

EPDS also includes questions on symptoms of anxiety, (which are crucial to perinatal mood disorders), while excluding constitutional symptoms of depression such as, sleeping pattern changes, – which are common to pregnancy (part of PHQ-9, BDI, and CES), and therefore can reduce the specificity for perinatal depression (ACOG, 2018).

A comparison between EPDS and some other tools highlights the benefits of EPDS (ACOG, 2018), is shown in Table 1

Table 1. Comparison of different depression screening tools.

Screening Tool	Number of Items	Time to Complete – in minutes
EPDS – Edinburgh Post Natal Depression Scale	10	Less than 5 minutes
PHQ-9 – Patient Health Questionnaire 9	9	Less than 5 minutes
BDI – Beck Depression Inventory	21
BDI-II – Beck Depression Inventory II	21	5–10 minutes
CES – Centre for Epidemiologic Studies Depression Scale	20	5–10 minutes
ZSDS – Zung Self Rating Depression Scale	20	5–10 minutes

EPDS is not only the best tool for detecting Antenatal Depression, and found to be clinically effective, but it is also a simple tool which takes less than 5 minutes for filling.

For the purpose of this research a backward and forward – Hindi language translated version of EPDS was validated by the authors to determine the cut off points for screening for Antenatal depression for overall and trimester specific depression screening.

Data analysis

Data was analyzed on the software package SPSS (earlier known as Statistical Package for the Social Sciences, but changed to Statistical Product and Service Solutions) version 22. SPSS is used for interactive, or batched statistical analysis, where the key features include basic hypothesis testing, bootstrapping, cluster analysis, data access and management, data preparation, graphs and charts, linear regression. Version 22 has been known to have better predictive analytics techniques through improved tools, output, and ease-of-use features (Hejase & Hejase, 2013).

Data cleaning was done prior to analysis. Cronbach’s alpha co-efficient was used to determine the reliability of the EDS data. Cronbach’s alpha is a measure of internal consistency, that is, how closely related a set of items are as a group. It is considered to be a measure of scale reliability. As the average inter-item correlation increases, Cronbach’s alpha increases as well (holding the number of items constant). Cohen’s kappa co-efficient test was done to test the intra-rater reliability between EDS Hindi language tool with the 2-item gold standard of EDS. In statistics, intra-rater reliability is the degree of agreement among repeated administrations of a diagnostic test performed by a single rater. Receiver Operator Characteristic (ROC) curve was plotted to depicts the tradeoff between the sensitivity and (1-specificity) across a series of cutoff points. The curve can be plotted on SPSS Vs. 22, by taking the values of the state variable –(Gold standard), and the test variable (full 10-item questionnaire) to measure the overall predictive value with sensitivity and specificity, with the Area Under the Curve (AUC) corresponding to 95% C.I. (Confidence Interval). As a thumb rule, the interpretation of AUC of 0.5 means no discrimination (i.e., ability to diagnose patients with and without the disease or condition based on the test), 0.7 to 0.8 is considered acceptable, 0.8 to 0.9 is considered excellent, and more than 0.9 is considered outstanding. And, classifiers that give curves closer to the top-left corner indicate a better performance to the researchers. The ROC was plotted for determining the cutoff scores for overall as well as trimester specific screening for antenatal depression.

Results

The Hindi version of EDS showed high reliability with the Cronbach’s alpha co-efficient of 0.829. The ROC was plotted, and that gave an AUC of 0.90.

The results confirmed that the Hindi language EDS used in the study was a valid screening tool for screening for depression during pregnancy.

Discussion

The authors through this study indicate that screening for depression can be done by a simple tool, which can be filled by the pregnant mother herself, as long as the tool is in a language which she understands, and which has been pre validated to give the relevant cut off score. Similar validations have been done for many other languages, and give language specific cut off scores.

Now this knowledge can be used, to build an AI (Artificial Intelligence) enabled, mobile based application, which can be downloaded by the pregnant mother. After filling her trimester and other basic socio demographic details, she can fill out the EDS tool on the mobile app, as and when she feels like it. Thus providing the choice to a mother for “expression” of her real time feelings, without the impediments posed by – ANC (Antenatal Care) clinic timings, non availability of a doctor, travel time, dependency of a family member to take her to a doctor, long waiting queue, and specially in those cases where the pregnant mothers are carrying a high risk pregnancy and are unable to do all the above, they in particular can access help where needed from the palm of their hand.

The AI built into the app, would calculate the scores, and sound an alert to the Gynecologist treating the pregnant mother. Depending on the healthcare system as found in different countries this first point of contact may be not a Gyneacologist but another healthcare personnel. Irrespective of the cadre, the next step would be to provide medical help to the pregnant mother.

This could be done via technology driven telemedicine itself. A study (Naslund et al., 2017), on “Health Behavior Models” for informing Digital Technology interventions for individuals with mental illness was conducted. The first premise of this study was based on an evidenced based fact that mobile device ownership of people suffering from mental illness is comparable with the general population (Firth et al., 2016; Glick et al., 2016). The second premise was that interventions informed by theories of human behavior are found to be more effective than interventions not informed by theory (Glanz & Bishop, 2010). There have been a few technology interventions informed by behavior theories (Bull & Ezeanochie, 2016; Riley et al., 2011), and the theories recommended by the researchers in the study as being most effective for developing technology interventions are– (1) Health belief model, (2) Theory of planned behavior (3) Transtheoritical model (4) Social cognitive theory.

1. The Health Belief Model – Researchers through this model propound that the confidence or reliance in one’s own ability to take action for themselves, can be a key predictor of initiating behavior. So if a person becomes aware that they may be at a risk, it can trigger their ability to take an action. In this context, the self rating on the mobile app, and clarity on results in real time, can initiate the behavior of seeking help. It could also be a cue to taking medication.

2. Theory of Planned Behavior – Researchers through this model emphasize the role of “intention” as a determinant of behaving in a certain manner. For example, if an awareness is created on the various causes leading to antenatal depression, eg. substance abuse, on the mobile app, and if it is clear to the user, that non adherence of substance abuse can lead to reduced illness, then as per the theory of planned behavior the individual’s behavior will be influenced by this awareness.

   The Health Belief model and Theory of Planned Behavior can work together in affecting an individual’s behavior toward the self.

3. Transtheoritical model – Through this model, researchers show that an individual goes through six stages of behavior change to be able to make a decision. Precontemplation: In this stage due to limited or no awareness, there is no intention to change behavior. After this comes, Contemplation: where the individual may be aware of the harms of their behavior, and also understands the benefits of changing their behavior, however there is inertia, to take any action and this may exist for long periods. Preparation: is the stage of an intention to change behavior. This is an important stage that can be achieved by customized AI driven tools via the mobile app technology to trigger the “intention” of changing behavior, once the subtle awareness sessions/programs are highlighted to the individual basis his profile and depression score. The next stage is Action: and the objective is to keep sustaining the action taken by the individual. The stage of Maintenance: is to ensure sustainance of behavior over a long period. Termination: stage is to ensure that relapse due to unhealthy behavior does not happen.

4. Transtheoritical model – Through this model, researchers show that an individual goes through six stages of behavior change to be able to make a decision. Precontemplation: In this stage due to limited or no awareness, there is no intention to change behavior. After this comes, Contemplation: where the individual may be aware of the harms of their behavior, and also understands the benefits of changing their behavior, however there is inertia, to take any action and this may exist for long periods. Preparation: is the stage of an intention to change behavior. This is an important stage that can be achieved by customized AI driven tools via the mobile app technology to trigger the “intention” of changing behavior, once the subtle awareness sessions/programs are highlighted to the individual basis his profile and depression score. The next stage is Action: and the objective is to keep sustaining the action taken by the individual. The stage of Maintenance: is to ensure sustainance of behavior over a long period. Termination: stage is to ensure that relapse due to unhealthy behavior does not happen.

5. Social Cognitive Theory – Researchers explain how an individual’s behavior change is affected by a mix of factors like their past experiences, and how they respond to the external social context.

Many researchers have explored ideas of intervention based on behavior change theories – a pilot feasibility study (Danaher et al., 2013) was done to a sample of pregnant mothers, from Iowa, USA, and Greater Melbourne, Australia, for web based “mood booster” intervention for post partum depression after being screened by the EPDS tool, and the Patient Health Questionnaires at select times, over a 9 month duration. The study concluded that while at pre test 55% of the participants were found to be depressed, by post test, 90% of the participants no longer were found to be depressed.

A systematic review of research on whether mental health apps for mobile devices have an efficacy for all ages was conducted by researchers in 2013 (Donker et al., 2013). Eight articles that described 5 apps, met the eligibility criteria; of these, through four apps support from a mental health professional was provided. In four studies, researchers had specifically targeted depression using apps. Although this did not specifically target pregnant mothers, yet the results show that there was a significant reduction in depression “caseness.” The review admitted that one of the limitations was a small number of participants that were studied, yet it clearly opined need for future research. Another review done in 2014, of studies reviewing the “usability, acceptability, feasibility, safety or efficacy” of user led, internet or mobile based interventions, concluded, that not only are such interventions acceptable, but they are also feasible and can improve clinical outcomes. However the review concludes, that further research is indeed indicated.

Conclusion

Various models have been tested in the past for providing interventions to address depression, but the issue is, are we really able to capture the state of depression at all? So far researchers have neither ideated nor conducted a research for the self enabled real time depression screening during pregnancy- the time of highest risk of all. The purpose of the authors of this research is to showcase, without an impediment of geography or language, and with a simple tool which can be filled in under 5 minutes, depression during pregnancy in real time, can be captured. This can provide an “opportunity” to any pregnant lady straight “from” and possibly “to,” the palm of her hand. Capturing the depression is the first step of providing HOPE for treatment.

The authors thus provide a technology driven, personalized option, for ensuring, that no pregnant mother the world over can be left undetected for depression during pregnancy. Policymakers, administrators and healthcare providers will have to gear up, to ensure that once this score sounds an alert, either mobile led interventions, or physical interventions, can alleviate and address Maternal Depression.

Author contribution

Mrs. Sheenu Jhawar – Conceptualisation, Methodology, Software, Validation, Formal Analysis, Investigation, Resources, Data Curation, Writing – Original Draft, Visualization, Project Administration.

Dr. Shiv Dutt Gupta – Methodology, Formal Analysis, Writing – Review and Editing, Supervision.

Dr. Arindam Basu – Methodology, Formal Analysis, Writing – Review and Editing, Validation.

Acknowledgement

The authors thank library at IIHMR University for availability and access to knowledge database and learning resources.

Funding

The author(s) reported there is no funding associated with the work featured in this article.