Machine Learning Methods for Predicting Postpartum Depression: Scoping Review

Revolutionizing Maternal Health: The Role of Artificial Intelligence in Enhancing Care and Accessibility

Maternal health remains a critical global health challenge, with disparities in access to care and quality of services contributing to high maternal mortality and morbidity rates. Artificial intelligence (AI) has emerged as a promising tool for addressing these challenges by enhancing diagnostic accuracy, improving patient monitoring, and expanding access to care. This review explores the transformative role of AI in maternal healthcare, focusing on its applications in the early detection of pregnancy complications, personalized care, and remote monitoring through AI-driven technologies. AI tools such as predictive analytics and machine learning can help identify at-risk pregnancies and guide timely interventions, reducing preventable maternal and neonatal complications. Additionally, AI-enabled telemedicine and virtual assistants are bridging healthcare gaps, particularly in underserved and rural areas, improving accessibility for women who might otherwise face barriers to quality maternal care. Despite the potential benefits, challenges such as data privacy, algorithmic bias, and the need for human oversight must be carefully addressed. The review also discusses future research directions, including expanding AI applications in maternal health globally and the need for ethical frameworks to guide its integration. AI holds the potential to revolutionize maternal healthcare by enhancing both care quality and accessibility, offering a pathway to safer, more equitable maternal outcomes.

Finding the best predictive model for hypertensive depression in older adults based on machine learning and metabolomics research

Objective

Depression is a common comorbidity in hypertensive older adults, yet depression is more difficult to diagnose correctly. Our goal is to find predictive models of depression in hypertensive patients using a combination of various machine learning (ML) methods and metabolomics.

Methods

Methods We recruited 379 elderly people aged ≥65 years from the Chinese community. Plasma samples were collected and assayed by gas chromatography/liquid chromatography-mass spectrometry (GC/LC-MS). Orthogonal partial least squares discriminant analysis (OPLS-DA), volcano diagrams and thermograms were used to distinguish metabolites. The attribute discriminators CfsSubsetEval combined with search method BestFirst in WEKA software was used to find the best predicted metabolite combinations, and then 24 classification methods with 10-fold cross-validation were used for prediction.

Results

34 individuals were considered hypertensive combined with depression according to our criteria, and 34 subjects with hypertension only were matched according to age and sex. 19 metabolites by GC-MS and 65 metabolites by LC-MS contributed significantly to the differentiation between the depressed and non-depressed cohorts, with a VIP value of more than 1 and a P value of less than 0.05. There were multiple metabolic pathway alterations. The metabolite combinations screened with WEKA for optimal diagnostic value included 12 metabolites. The machine learning methods with AUC values greater than 0.9 were bayesNet and random forests, and their other evaluation measures are also better.

Conclusion

Altered metabolites and metabolic pathways are present in older adults with hypertension combined with depression. Methods using metabolomics and machine learning performed quite well in predicting depression in hypertensive older adults, contributing to further clinical research.

Comparative analysis of machine learning versus traditional method for early detection of parental depression symptoms in the NICU

Introduction

Neonatal intensive care unit (NICU) admission is a stressful experience for parents. NICU parents are twice at risk of depression symptoms compared to the general birthing population. Parental mental health problems have harmful long-term effects on both parents and infants. Timely screening and treatment can reduce these negative consequences.

Objective

Our objective is to compare the performance of the traditional logistic regression with other machine learning (ML) models in identifying parents who are more likely to have depression symptoms to prioritize screening of at-risk parents. We used data obtained from parents of infants discharged from the NICU at Children's National Hospital (n = 300) from 2016 to 2017. This dataset includes a comprehensive list of demographic characteristics, depression and stress symptoms, social support, and parent/infant factors.

Study design

Our study design optimized eight ML algorithms - Logistic Regression, Support Vector Machine, Decision Tree, Random Forest, XGBoost, Naïve Bayes, K-Nearest Neighbor, and Artificial Neural Network - to identify the main risk factors associated with parental depression. We compared models based on the area under the receiver operating characteristic curve (AUC), positive predicted value (PPV), sensitivity, and F-score.

Results

The results showed that all eight models achieved an AUC above 0.8, suggesting that the logistic regression-based model's performance is comparable to other common ML models.

Conclusion

Logistic regression is effective in identifying parents at risk of depression for targeted screening with a performance comparable to common ML-based models.

Wife-Mother Role Conflict at the Critical Child-Rearing Stage: A Machine-Learning Approach to Identify What and How Matters in Maternal Depression Symptoms in China

Maternal depression (MD) was one of the most prevalent psychiatric problems worldwide. However, it easily remains untreated and misses the best time to prevent the emergence or worsening of major depressive symptoms due to under-observed stigma and the lack of effective screening tools. Thus, this study aims to develop and validate a machine learning-based MD symptoms prediction model integrating more observable and objective factors to early detect and monitor MD risk. A cross-sectional study was conducted in 10 community vaccination centers in Wenzhou, China, and a total of 1099 mothers were surveyed by using purposive sampling. A questionnaire containing questions regarding socio-demographic variables, psychophysiological variables, wife role-related variables, and mother role-related variables was used to collect data. A framework of data preprocessing, feature selection, and model evaluation was implemented to develop an optimal risk prediction model. Results demonstrated that the XG-Boost algorithm provided robust performance with the highest AUC and well-balanced sensitivity and specificity (AUC = 0.90, sensitivity = 0.74, specificity = 0.90). Furthermore, the causal mediation analysis indicated that wife-mother role conflict positively predicted MD symptoms, and it also exerted influence on mothers suffering through the mediation of anxiety and insomnia. Findings from the present study may help guide the development of MD screening tools to early detect and provide the modifiable risk factor information for timely tailored prevention.

Using machine learning to develop a five-item short form of the children's depression inventory

BACKGROUND

Many adolescents experience depression that often goes undetected and untreated. Identifying children and adolescents at a high risk of depression in a timely manner is an urgent concern. While the Children's Depression Inventory (CDI) is widely utilized in China, it lacks a localized revision or simplified version. With its 27 items requiring professional administration, the original CDI proves to be a time-consuming method for predicting children and adolescents with high depression risk. Hence, this study aimed to develop a shortened version of the CDI to predict high depression risk, thereby enhancing the efficiency of prediction and intervention.

METHODS

Initially, backward elimination is conducted to identify various version of the short-form scales (e.g., three-item and five-item versions). Subsequently, the performance of five machine learning (ML) algorithms on these versions is evaluated using the area under the ROC curve (AUC) to determine the best algorithm. The chosen algorithm is then utilized to model the short-form scales, facilitating the identification of the optimal short-form scale based on predefined evaluation metrics. Following this, evaluation metrics are computed for all potential decision thresholds of the optimal short-form scale, and the threshold value is determined. Finally, the reliability and validity of the optimal short-form scale are assessed using a new sample.

RESULTS

The study identified a five-item short-form CDI with a decision threshold of 4 as the most appropriate scale considering all assessment indicators. The scale had 81.48% fewer items than the original version, indicating good predictive performance (AUC = 0.81, Accuracy = 0.83, Recall = 0.76, Precision = 0.71). Based on the test of 315 middle school students, the results showed that the five-item CDI had good measurement indexes (Cronbach's alpha = 0.72, criterion-related validity = 0.77).

CONCLUSIONS

This five-item short-form CDI is the first shortened and revised version of the CDI in China based on large local data samples.

Prevalence and risk factors analysis of postpartum depression at early stage using hybrid deep learning model

Postpartum Depression Disorder (PPDD) is a prevalent mental health condition and results in severe depression and suicide attempts in the social community. Prompt actions are crucial in tackling PPDD, which requires a quick recognition and accurate analysis of the probability factors associated with this condition. This concern requires attention. The primary aim of our research is to investigate the feasibility of anticipating an individual's mental state by categorizing individuals with depression from those without depression using a dataset consisting of text along with audio recordings from patients diagnosed with PPDD. This research proposes a hybrid PPDD framework that combines Improved Bi-directional Long Short-Term Memory (IBi-LSTM) with Transfer Learning (TL) based on two Convolutional Neural Network (CNN) architectures, respectively CNN-text and CNN audio. In the proposed model, the CNN section efficiently utilizes TL to obtain crucial knowledge from text and audio characteristics, whereas the improved Bi-LSTM module combines written material and sound data to obtain intricate chronological interpersonal relationships. The proposed model incorporates an attention technique to augment the effectiveness of the Bi-LSTM scheme. An experimental analysis is conducted on the PPDD online textual and speech audio dataset collected from UCI. It includes textual features such as age, women's health tracks, medical histories, demographic information, daily life metrics, psychological evaluations, and 'speech records' of PPDD patients. Data pre-processing is applied to maintain the data integrity and achieve reliable model performance. The proposed model demonstrates a great performance in better precision, recall, accuracy, and F1-score over existing deep learning models, including VGG-16, Base-CNN, and CNN-LSTM. These metrics indicate the model's ability to differentiate among women at risk of PPDD vs. non-PPDD. In addition, the feature importance analysis demonstrates that specific risk factors substantially impact the prediction of PPDD. The findings of this research establish a basis for improved precision and promptness in assessing the risk of PPDD, which may ultimately result in earlier implementation of interventions and the establishment of support networks for women who are susceptible to PPDD.

Recentering responsible and explainable artificial intelligence research on patients: implications in perinatal psychiatry

In the setting of underdiagnosed and undertreated perinatal depression (PD), Artificial intelligence (AI) solutions are poised to help predict and treat PD. In the near future, perinatal patients may interact with AI during clinical decision-making, in their patient portals, or through AI-powered chatbots delivering psychotherapy. The increase in potential AI applications has led to discussions regarding responsible AI and explainable AI (XAI). Current discussions of RAI, however, are limited in their consideration of the patient as an active participant with AI. Therefore, we propose a patient-centered, rather than a patient-adjacent, approach to RAI and XAI, that identifies autonomy, beneficence, justice, trust, privacy, and transparency as core concepts to uphold for health professionals and patients. We present empirical evidence that these principles are strongly valued by patients. We further suggest possible design solutions that uphold these principles and acknowledge the pressing need for further research about practical applications to uphold these principles.

Predicting internalizing symptoms with machine learning: identifying individuals that need care

Objective The current project aims to identify individuals in urgent need of mental health care, using a machine learning algorithm (random forest). Comparison/contrast with conventional regression analyses is discussed. Participants: A total of 2,409 participants were recruited from an anonymous university, including undergraduate and graduate students, faculty, and staff. Methods: Answers to a COVID-19 impact survey, the Patient Health Questionnaire-9 (PHQ-9), and the Generalized Anxiety Disorder-7 (GAD-7) were collected. The total scores of PHQ-9 and GAD-7 were regressed on six composites that were created from the questionnaire items, based on their topics. A random forest was trained and validated. Results: Results indicate that the random forest model was able to make accurate, prospective predictions (R2 = .429 on average) and we review variables that were deemed predictively relevant. Conclusions: Overall, the study suggests that predictive models can be clinically useful in identifying individuals with internalizing symptoms based on daily life disruption experiences.

Identification of cuproptosis-related molecular subtypes as a biomarker for differentiating active from latent tuberculosis in children

BACKGROUND

Cell death plays a crucial role in the progression of active tuberculosis (ATB) from latent infection (LTBI). Cuproptosis, a novel programmed cell death, has been reported to be associated with the pathology of various diseases. We aimed to identify cuproptosis-related molecular subtypes as biomarkers for distinguishing ATB from LTBI in pediatric patients.

METHOD

The expression profiles of cuproptosis regulators and immune characteristics in pediatric patients with ATB and LTBI were analyzed based on GSE39939 downloaded from the Gene Expression Omnibus. From the 52 ATB samples, we investigated the molecular subtypes based on differentially expressed cuproptosis-related genes (DE-CRGs) via consensus clustering and related immune cell infiltration. Subtype-specific differentially expressed genes (DEGs) were found using the weighted gene co-expression network analysis. The optimum machine model was then determined by comparing the performance of the eXtreme Gradient Boost (XGB), the random forest model (RF), the general linear model (GLM), and the support vector machine model (SVM). Nomogram and test datasets (GSE39940) were used to verify the prediction accuracy.

RESULTS

Nine DE-CRGs (NFE2L2, NLRP3, FDX1, LIPT1, PDHB, MTF1, GLS, DBT, and DLST) associated with active immune responses were ascertained between ATB and LTBI patients. Two cuproptosis-related molecular subtypes were defined in ATB pediatrics. Single sample gene set enrichment analysis suggested that compared with Subtype 2, Subtype 1 was characterized by decreased lymphocytes and increased inflammatory activation. Gene set variation analysis showed that cluster-specific DEGs in Subtype 1 were closely associated with immune and inflammation responses and energy and amino acids metabolism. The SVM model exhibited the best discriminative performance with a higher area under the curve (AUC = 0.983) and relatively lower root mean square and residual error. A final 5-gene-based (MAN1C1, DKFZP434N035, SIRT4, BPGM, and APBA2) SVM model was created, demonstrating satisfactory performance in the test datasets (AUC = 0.905). The decision curve analysis and nomogram calibration curve also revealed the accuracy of differentiating ATB from LTBI in children.

CONCLUSION

Our study suggested that cuproptosis might be associated with the immunopathology of Mycobacterium tuberculosis infection in children. Additionally, we built a satisfactory prediction model to assess the cuproptosis subtype risk in ATB, which can be used as a reliable biomarker for the distinguishment between pediatric ATB and LTBI.

Genetic analysis of cuproptosis subtypes and immunological features in severe influenza

The mechanisms regulating cuproptosis in severe influenza are still unknown. We aimed to identify the molecular subtypes of cuproptosis and immunological characteristics associated with severe influenza in patients requiring invasive mechanical ventilation (IMV). The expression of cuproptosis modulatory factors and immunological characteristics of these patients were analyzed using the public datasets (GSE101702, GSE21802, and GSE111368) from the Gene Expression Omnibus (GEO). Seven cuproptotic-associated genes (ATP7B, ATP7A, FDX1, LIAS, DLD, MTF1, DBT) related to active immune responses were identified in patients suffering from severe and non-severe influenza and two cuproptosis-associated molecular subtypes were discovered in severe influenza patients. Singe-set gene set expression analysis (SsGSEA) indicated that compared with subtype 2, subtype 1 was characterized by reduced adaptive cellular immune responses and increased neutrophil activation. Gene set variation assessment revealed that cluster-specific differentially expressed genes (DEGs) in subtype 1 were involved in autophagy, apoptosis, oxidative phosphorylation, and T cell, immune, and inflammatory responses, amongst others. The random forest (RF) model revealed the most differentiating efficiency with relatively small residual and root mean square error and an increased area under the curve value (AUC = 0.857). Lastly, a five-gene-based RF model (CD247, GADD45A, KIF1B, LIN7A, HLA_DPA1) was established, which showed satisfactory efficiency in the test datasets GSE111368 (AUC = 0.819). Nomogram calibration and decision curve analysis demonstrated its accuracy for the prediction of severe influenza. This study suggests that cuproptosis might be associated with the immunopathology of severe influenza. Additionally, an efficient model for the prediction of cuproptosis subtypes was developed which will contribute to the prevention and treatment of severe influenza patients needing IMV.

An optimization for postpartum depression risk assessment and preventive intervention strategy based machine learning approaches

BACKGROUND

Postpartum depression (PPD) is one of the most common psychiatric disorders for women after delivery. The establishment of an effective PPD prediction model helps to distinguish high-risk groups, and verifying whether such high-risk groups can benefit from drug intervention is very important for clinical guidance.

METHODS

We collected data of parturients that underwent a cesarean delivery. The Control group was divided into a training cohort and a testing cohort. Six different ML models were constructed and we compared their prediction performance in the testing cohort. For model interpretation, we introduced SHapley Additive exPlanations (SHAP). Then, training cohort, ketamine group and dexmedetomidine (DEX) group were classified as high or low risk for PPD by the model. A 1:1 propensity score matching (PSM) was performed to compare the incidence of PPD between two groups in different risk cohorts.

RESULTS

Extreme gradient enhancement (XGB) had the best recognition effect, with an area under the receiver operating characteristic curve (AUROC) of 0.789 (95 % CI 0.742-0.836) in the training cohort and 0.744 (95 % CI 0.655-0.823) in the testing cohort, respectively. A threshold of 21.5 % PPD risk probability was determined. After PSM, the results showed that the incidence of PPD in the two intervention groups was significantly different from the control group in the high-risk cohort (P < 0.001) but not in the low-risk cohort (P > 0.001).

CONCLUSION

Our study demonstrated that the XGB algorithm provided a more accurate in prediction of PPD risk, and it was beneficial to receive early intervention for the high-risk groups distinguished by the model.

Identifying women with postdelivery posttraumatic stress disorder using natural language processing of personal childbirth narratives

BACKGROUND

Maternal mental disorders are considered a leading complication of childbirth and a common contributor to maternal death. In addition to undermining maternal welfare, untreated postpartum psychopathology can result in child emotional and physical neglect and associated significant pediatric health costs. Some women may experience traumatic childbirth and develop posttraumatic stress disorder symptoms after delivery (childbirth-related posttraumatic stress disorder). Although women are routinely screened for postpartum depression in the United States, there is no recommended protocol to inform the identification of women who are likely to experience childbirth-related posttraumatic stress disorder. Advancements in computational methods of free text have shown promise in informing the diagnosis of psychiatric conditions. Although the language in narratives of stressful events has been associated with posttrauma outcomes, whether the narratives of childbirth processed via machine learning can be useful for childbirth-related posttraumatic stress disorder screening is unknown.

OBJECTIVE

This study aimed to examine the use of written narrative accounts of personal childbirth experiences for the identification of women with childbirth-related posttraumatic stress disorder. To this end, we developed a model based on natural language processing and machine learning algorithms to identify childbirth-related posttraumatic stress disorder via the classification of birth narratives.

STUDY DESIGN

Overall, 1127 eligible postpartum women who enrolled in a study survey during the COVID-19 pandemic provided short written childbirth narrative accounts in which they were instructed to focus on the most distressing aspects of their childbirth experience. They also completed a posttraumatic stress disorder symptom screen to determine childbirth-related posttraumatic stress disorder. After the exclusion criteria were applied, data from 995 participants were analyzed. A machine learning-based Sentence-Transformers natural language processing model was used to represent narratives as vectors that served as inputs for a neural network machine learning model developed in this study to identify participants with childbirth-related posttraumatic stress disorder.

RESULTS

The machine learning model derived from natural language processing of childbirth narratives achieved good performance (area under the curve, 0.75; F1 score, 0.76; sensitivity, 0.8; specificity, 0.70). Moreover, women with childbirth-related posttraumatic stress disorder generated longer narratives (t test results: t=2.30; p=.02) and used more negative emotional expressions (Wilcoxon test: sadness: p=8.90e-04; W=31,017; anger: p=1.32e-02; W=35,005.50) and death-related words (Wilcoxon test: p=3.48e-05; W=34,538) in describing their childbirth experience than those with no childbirth-related posttraumatic stress disorder.

CONCLUSION

This study provided proof of concept that personal childbirth narrative accounts generated in the early postpartum period and analyzed via advanced computational methods can detect with relatively high accuracy women who are likely to endorse childbirth-related posttraumatic stress disorder and those at low risk. This suggests that birth narratives could be promising for informing low-cost, noninvasive tools for maternal mental health screening, and more research that used machine learning to predict early signs of maternal psychiatric morbidity is warranted.

Machine Learning Based on Event-Related EEG of Sustained Attention Differentiates Adults with Chronic High-Altitude Exposure from Healthy Controls

(1) Objective: The aim of this study was to examine the effect of high altitude on inhibitory control processes that underlie sustained attention in the neural correlates of EEG data, and explore whether the EEG data reflecting inhibitory control contain valuable information to classify high-altitude chronic hypoxia and plain controls. (2) Methods: 35 chronic high-altitude hypoxic adults and 32 matched controls were recruited. They were required to perform the go/no-go sustained attention task (GSAT) using event-related potentials. Three machine learning algorithms, namely a support vector machine (SVM), logistic regression (LR), and a decision tree (DT), were trained based on the related ERP components and neural oscillations to build a dichotomous classification model. (3) Results: Behaviorally, we found that the high altitude (HA) group had lower omission error rates during all observation periods than the low altitude (LA) group. Meanwhile, the ERP results showed that the HA participants had significantly shorter latency than the LAs for sustained potential (SP), indicating vigilance to response-related conflict. Meanwhile, event-related spectral perturbation (ERSP) analysis suggested that lowlander immigrants exposed to high altitudes may have compensatory activated prefrontal cortexes (PFC), as reflected by slow alpha, beta, and theta frequency-band neural oscillations. Finally, the machine learning results showed that the SVM achieved the optimal classification F1 score in the later stage of sustained attention, with an F1 score of 0.93, accuracy of 92.54%, sensitivity of 91.43%, specificity of 93.75%, and area under ROC curve (AUC) of 0.97. The results proved that SVM classification algorithms could be applied to identify chronic high-altitude hypoxia. (4) Conclusions: Compared with other methods, the SVM leads to a good overall performance that increases with the time spent on task, illustrating that the ERPs and neural oscillations may provide neuroelectrophysiological markers for identifying chronic plateau hypoxia.

Identifying Women with Post-Delivery Posttraumatic Stress Disorder using Natural Language Processing of Personal Childbirth Narratives

Background

Maternal mental disorders are considered a leading complication of childbirth and a common contributor to maternal death. In addition to undermining maternal welfare, untreated postpartum psychopathology can result in child emotional and physical neglect, and associated significant pediatric health costs. Some women may experience a traumatic childbirth and develop posttraumatic stress disorder (PTSD) symptoms following delivery (CB-PTSD). Although women are routinely screened for postpartum depression in the U.S., there is no recommended protocol to inform the identification of women who are likely to experience CB-PTSD. Advancements in computational methods of free text has shown promise in informing diagnosis of psychiatric conditions. Although the language in narratives of stressful events has been associated with post-trauma outcomes, whether the narratives of childbirth processed via machine learning can be useful for CB-PTSD screening is unknown.

Objective

This study examined the utility of written narrative accounts of personal childbirth experience for the identification of women with provisional CB-PTSD. To this end, we developed a model based on natural language processing (NLP) and machine learning (ML) algorithms to identify CB-PTSD via classification of birth narratives.

Study Design

A total of 1,127 eligible postpartum women who enrolled in a study survey during the COVID-19 era provided short written childbirth narrative accounts in which they were instructed to focus on the most distressing aspects of their childbirth experience. They also completed a PTSD symptom screen to determine provisional CB-PTSD. After exclusion criteria were applied, data from 995 participants was analyzed. An ML-based Sentence-Transformer NLP model was used to represent narratives as vectors that served as inputs for a neural network ML model developed in this study to identify participants with provisional CB-PTSD.

Results

The ML model derived from NLP of childbirth narratives achieved good performance: AUC 0.75, F1-score 0.76, sensitivity 0.8, and specificity 0.70. Moreover, women with provisional CB-PTSD generated longer narratives (t-test results: t=2 . 30, p=0 . 02 ) and used more negative emotional expressions (Wilcoxon test: 'sadness': p=8 . 90e- ⁰⁴ , W=31,017 ; 'anger': p=1 . 32e- ⁰² , W=35,005 . 50 ) and death-related words (Wilcoxon test: p=3 . 48e- ⁰⁵ , W=34,538 ) in describing their childbirth experience than those with no CB-PTSD.

Conclusions

This study provides proof of concept that personal childbirth narrative accounts generated in the early postpartum period and analyzed via advanced computational methods can detect with relatively high accuracy women who are likely to endorse CB-PTSD and those at low risk. This suggests that birth narratives could be promising for informing low-cost, non-invasive tools for maternal mental health screening, and more research that utilizes ML to predict early signs of maternal psychiatric morbidity is warranted.

Prediction Model between Serum Vitamin D and Neurological Deficit in Cerebral Infarction Patients Based on Machine Learning

Objective

Vitamin D is associated with neurological deficits in patients with cerebral infarction. This study uses machine learning to evaluate the prediction model's efficacy of the correlation between vitamin D and neurological deficit in patients with cerebral infarction.

Methods

A total of 200 patients with cerebral infarction admitted to the Department of Neurology of our hospital from July 2018 to June 2019 were selected. The patients were randomly divided into a training set (n = 140) and a test set (n = 60) in a 7 : 3 ratio. The prediction model is constructed from the training set's data, and the model's prediction effect was evaluated by test set data. The area under the receiver operator characteristic curve was used to assess the prediction efficiency of models.

Results

In the training set, the area under the curve (AUC) of the logistic regression model and XGBoost algorithm model was 0.727 (95% CI: 0.601~0.854) and 0.818 (95% CI: 0.734~0.934), respectively. While in the test set, the AUC of the logistic regression model and XGBoost algorithm model was 0.761 (95% CI: 0.640~0.882) and 0.786 (95% CI: 0.670~0.902), respectively.

Conclusion

The prediction model of the correlation between vitamin D and neurological deficit in patients with cerebral infarction based on machine learning has a good prediction efficiency.

Predicting perinatal health outcomes using smartphone-based digital phenotyping and machine learning in a prospective Swedish cohort (Mom2B): study protocol

INTRODUCTION

Perinatal complications, such as perinatal depression and preterm birth, are major causes of morbidity and mortality for the mother and the child. Prediction of high risk can allow for early delivery of existing interventions for prevention. This ongoing study aims to use digital phenotyping data from the Mom2B smartphone application to develop models to predict women at high risk for mental and somatic complications.

METHODS AND ANALYSIS

All Swedish-speaking women over 18 years, who are either pregnant or within 3 months postpartum are eligible to participate by downloading the Mom2B smartphone app. We aim to recruit at least 5000 participants with completed outcome measures. Throughout the pregnancy and within the first year postpartum, both active and passive data are collected via the app in an effort to establish a participant's digital phenotype. Active data collection consists of surveys related to participant background information, mental and physical health, lifestyle, and social circumstances, as well as voice recordings. Participants' general smartphone activity, geographical movement patterns, social media activity and cognitive patterns can be estimated through passive data collection from smartphone sensors and activity logs. The outcomes will be measured using surveys, such as the Edinburgh Postnatal Depression Scale, and through linkage to national registers, from where information on registered clinical diagnoses and received care, including prescribed medication, can be obtained. Advanced machine learning and deep learning techniques will be applied to these multimodal data in order to develop accurate algorithms for the prediction of perinatal depression and preterm birth. In this way, earlier intervention may be possible.

ETHICS AND DISSEMINATION

Ethical approval has been obtained from the Swedish Ethical Review Authority (dnr: 2019/01170, with amendments), and the project fully fulfils the General Data Protection Regulation (GDPR) requirements. All participants provide consent to participate and can withdraw their participation at any time. Results from this project will be disseminated in international peer-reviewed journals and presented in relevant conferences.