Prediction of gestational diabetes based on nationwide electronic health records

A machine learning model trained on a high-throughput antibacterial screen increases the hit rate of drug discovery

Screening for novel antibacterial compounds in small molecule libraries has a low success rate. We applied machine learning (ML)-based virtual screening for antibacterial activity and evaluated its predictive power by experimental validation. We first binarized 29,537 compounds according to their growth inhibitory activity (hit rate 0.87%) against the antibiotic-resistant bacterium Burkholderia cenocepacia and described their molecular features with a directed-message passing neural network (D-MPNN). Then, we used the data to train an ML model that achieved a receiver operating characteristic (ROC) score of 0.823 on the test set. Finally, we predicted antibacterial activity in virtual libraries corresponding to 1,614 compounds from the Food and Drug Administration (FDA)-approved list and 224,205 natural products. Hit rates of 26% and 12%, respectively, were obtained when we tested the top-ranked predicted compounds for growth inhibitory activity against B. cenocepacia, which represents at least a 14-fold increase from the previous hit rate. In addition, more than 51% of the predicted antibacterial natural compounds inhibited ESKAPE pathogens showing that predictions expand beyond the organism-specific dataset to a broad range of bacteria. Overall, the developed ML approach can be used for compound prioritization before screening, increasing the typical hit rate of drug discovery.

Predictive Modeling for the Diagnosis of Gestational Diabetes Mellitus Using Epidemiological Data in the United Arab Emirates

Gestational diabetes mellitus (GDM) is a common condition with repercussions for both the mother and her child. Machine learning (ML) modeling techniques were proposed to predict the risk of several medical outcomes. A systematic evaluation of the predictive capacity of maternal factors resulting in GDM in the UAE is warranted. Data on a total of 3858 women who gave birth and had information on their GDM status in a birth cohort were used to fit the GDM risk prediction model. Information used for the predictive modeling were from self-reported epidemiological data collected at early gestation. Three different ML models, random forest (RF), gradient boosting model (GBM), and extreme gradient boosting (XGBoost), were used to predict GDM. Furthermore, to provide local interpretation of each feature in GDM diagnosis, features were studied using Shapley additive explanations (SHAP). Results obtained using ML models show that XGBoost, which achieved an AUC of 0.77, performed better compared to RF and GBM. Individual feature importance using SHAP value and the XGBoost model show that previous GDM diagnosis, maternal age, body mass index, and gravidity play a vital role in GDM diagnosis. ML models using self-reported epidemiological data are useful and feasible in prediction models for GDM diagnosis amongst pregnant women. Such data should be periodically collected at early pregnancy for health professionals to intervene at earlier stages to prevent adverse outcomes in pregnancy and delivery. The XGBoost algorithm was the optimal model for identifying the features that predict GDM diagnosis.

Incidence, Characteristics, and Outcomes of Clinically Undetected Bacteremia in Children Discharged Home From the Emergency Department

BACKGROUND

Despite a recent decline in the rates of invasive infections, bacteremia in young children remains a significant challenge. We aimed to describe patient characteristics, microbial etiology, and outcomes of bacteremic, well-appearing children 3-36 months of age who were discharged home from the pediatric emergency department (PED) on their index visit.

METHODS

A retrospective cohort study in the PED of a tertiary children's hospital from 1 June 2015 until 30 June 2021. We included all well appearing, immunocompetent infants 3-36 months old evaluated for fever and discharged home from the PED after a blood culture was drawn. We extracted demographic, clinical and laboratory data from the patient's electronic medical records for the index visit and subsequent encounters.

RESULTS

During the study period, 17,114 children evaluated for fever met the inclusion criteria. Seventy-two patients (0.42%) had positive cultures for known pathogens. Thirty-six (50%) were male and 36 (50%) younger than 1 year. The most common isolates were S. pneumonia 26%. (n = 19), K. Kingae 25%. (n = 18) and Salmonella spp. 13.9% (n = 10). Sixty patients (85.7%) were recalled to the ED or had a scheduled appointment, 10 (14.3%) returned spontaneously and two were followed up by phone. The median time between visits was 28.7 hours (IQR 19.1-41.1). One patient was admitted to intensive care during the course of hospitalization. There were no deaths.

CONCLUSION

The rate of undetected true bacteremia in our study was low and our data suggest that significant clinical deterioration during the first 24 hours is rare.

Dense phenotyping from electronic health records enables machine learning-based prediction of preterm birth

BACKGROUND

Identifying pregnancies at risk for preterm birth, one of the leading causes of worldwide infant mortality, has the potential to improve prenatal care. However, we lack broadly applicable methods to accurately predict preterm birth risk. The dense longitudinal information present in electronic health records (EHRs) is enabling scalable and cost-efficient risk modeling of many diseases, but EHR resources have been largely untapped in the study of pregnancy.

METHODS

Here, we apply machine learning to diverse data from EHRs with 35,282 deliveries to predict singleton preterm birth.

RESULTS

We find that machine learning models based on billing codes alone can predict preterm birth risk at various gestational ages (e.g., ROC-AUC = 0.75, PR-AUC = 0.40 at 28 weeks of gestation) and outperform comparable models trained using known risk factors (e.g., ROC-AUC = 0.65, PR-AUC = 0.25 at 28 weeks). Examining the patterns learned by the model reveals it stratifies deliveries into interpretable groups, including high-risk preterm birth subtypes enriched for distinct comorbidities. Our machine learning approach also predicts preterm birth subtypes (spontaneous vs. indicated), mode of delivery, and recurrent preterm birth. Finally, we demonstrate the portability of our approach by showing that the prediction models maintain their accuracy on a large, independent cohort (5978 deliveries) from a different healthcare system.

CONCLUSIONS

By leveraging rich phenotypic and genetic features derived from EHRs, we suggest that machine learning algorithms have great potential to improve medical care during pregnancy. However, further work is needed before these models can be applied in clinical settings.

Development of early prediction model for pregnancy-associated hypertension with graph-based semi-supervised learning

Clinical guidelines recommend several risk factors to identify women in early pregnancy at high risk of developing pregnancy-associated hypertension. However, these variables result in low predictive accuracy. Here, we developed a prediction model for pregnancy-associated hypertension using graph-based semi-supervised learning. This is a secondary analysis of a prospective study of healthy pregnant women. To develop the prediction model, we compared the prediction performances across five machine learning methods (semi-supervised learning with both labeled and unlabeled data, semi-supervised learning with labeled data only, logistic regression, support vector machine, and random forest) using three different variable sets: [a] variables from clinical guidelines, [b] selected important variables from the feature selection, and [c] all routine variables. Additionally, the proposed prediction model was compared with placental growth factor, a predictive biomarker for pregnancy-associated hypertension. The study population consisted of 1404 women, including 1347 women with complete follow-up (labeled data) and 57 women with incomplete follow-up (unlabeled data). Among the 1347 with complete follow-up, 2.4% (33/1347) developed pregnancy-associated HTN. Graph-based semi-supervised learning using top 11 variables achieved the best average prediction performance (mean area under the curve (AUC) of 0.89 in training set and 0.81 in test set), with higher sensitivity (72.7% vs 45.5% in test set) and similar specificity (80.0% vs 80.5% in test set) compared to risk factors from clinical guidelines. In addition, our proposed model with graph-based SSL had a higher performance than that of placental growth factor for total study population (AUC, 0.71 vs. 0.80, p < 0.001). In conclusion, we could accurately predict the development pregnancy-associated hypertension in early pregnancy through the use of routine clinical variables with the help of graph-based SSL.

Association of Genetic Predisposition and Physical Activity With Risk of Gestational Diabetes in Nulliparous Women

Importance

Polygenic risk scores (PRS) for type 2 diabetes (T2D) can improve risk prediction for gestational diabetes (GD), yet the strength of the association between genetic and lifestyle risk factors has not been quantified.

Objective

To assess the association of PRS and physical activity in existing GD risk models and identify patient subgroups who may receive the most benefits from a PRS or physical activity intervention.

Design, Settings, and Participants

The Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be cohort was established to study individuals without previous pregnancy lasting at least 20 weeks (nulliparous) and to elucidate factors associated with adverse pregnancy outcomes. A subcohort of 3533 participants with European ancestry was used for risk assessment and performance evaluation. Participants were enrolled from October 5, 2010, to December 3, 2013, and underwent genotyping between February 19, 2019, and February 28, 2020. Data were analyzed from September 15, 2020, to November 10, 2021.

Exposures

Self-reported total physical activity in early pregnancy was quantified as metabolic equivalents of task (METs). Polygenic risk scores were calculated for T2D using contributions of 84 single nucleotide variants, weighted by their association in the Diabetes Genetics Replication and Meta-analysis Consortium data.

Main Outcomes and Measures

Estimation of the development of GD from clinical, genetic, and environmental variables collected in early pregnancy, assessed using measures of model discrimination. Odds ratios and positive likelihood ratios were used to evaluate the association of PRS and physical activity with GD risk.

Results

A total of 3533 women were included in this analysis (mean [SD] age, 28.6 [4.9] years). In high-risk population subgroups (body mass index ≥25 or aged ≥35 years), individuals with high PRS (top 25th percentile) or low activity levels (METs <450) had increased odds of a GD diagnosis of 25% to 75%. Compared with the general population, participants with both high PRS and low activity levels had higher odds of a GD diagnosis (odds ratio, 3.4 [95% CI, 2.3-5.3]), whereas participants with low PRS and high METs had significantly reduced risk of a GD diagnosis (odds ratio, 0.5 [95% CI, 0.3-0.9]; P = .01).

Conclusions and Relevance

In this cohort study, the addition of PRS was associated with the stratified risk of GD diagnosis among high-risk patient subgroups, suggesting the benefits of targeted PRS ascertainment to encourage early intervention.

A Pipeline for the Implementation and Visualization of Explainable Machine Learning for Medical Imaging Using Radiomics Features

Machine learning (ML) models have been shown to predict the presence of clinical factors from medical imaging with remarkable accuracy. However, these complex models can be difficult to interpret and are often criticized as “black boxes”. Prediction models that provide no insight into how their predictions are obtained are difficult to trust for making important clinical decisions, such as medical diagnoses or treatment. Explainable machine learning (XML) methods, such as Shapley values, have made it possible to explain the behavior of ML algorithms and to identify which predictors contribute most to a prediction. Incorporating XML methods into medical software tools has the potential to increase trust in ML-powered predictions and aid physicians in making medical decisions. Specifically, in the field of medical imaging analysis the most used methods for explaining deep learning-based model predictions are saliency maps that highlight important areas of an image. However, they do not provide a straightforward interpretation of which qualities of an image area are important. Here, we describe a novel pipeline for XML imaging that uses radiomics data and Shapley values as tools to explain outcome predictions from complex prediction models built with medical imaging with well-defined predictors. We present a visualization of XML imaging results in a clinician-focused dashboard that can be generalized to various settings. We demonstrate the use of this workflow for developing and explaining a prediction model using MRI data from glioma patients to predict a genetic mutation.

Simple method for identification of women at risk of gestational diabetes mellitus in Arusha urban, Tanzania

BACKGROUND

Screening for gestational diabetes mellitus in Tanzania is challenged by limited resources. Therefore, this study aimed to develop a simple method for identification of women at risk of gestational diabetes mellitus in Arusha urban, Tanzania.

METHODS

This study used data from a cross sectional study, that was conducted between March and December 2018 in Arusha District involving 468 pregnant women who were not known to have diabetes before pregnancy. Urine glucose was tested using urine multistics and blood glucose levels by Gluco-Plus™ and diagnosed in accordance with the World Health Organization's criteria. Anthropometrics were measured using standard procedures and maternal characteristics were collected through face-to-face interviews using a questionnaire with structured questions. Univariate analysis assessed individual variables association with gestational diabetes mellitus where variables with p-value of < 0.05 were included in multivariable analysis and predictors with p-value < 0.1 remained in the final model. Each variable was scored based on its estimated coefficients and risk scores were calculated by multiplying the corresponding coefficients by ten to get integers. The model's performance was assessed using c-statistic. Data were analyzed using Statistical Package for Social Science™.

RESULTS

The risk score included body fat ≥ 38%, delivery to macrosomic babies, mid-upper arm circumference ≥ 28 cm, and family history of type 2 diabetes mellitus. The score correctly identified 98% of women with gestational diabetes with an area under the receiver operating characteristic curve of 0.97 (95% CI 0.96-0.99, p < 0.001), sensitivity of 0.98, and specificity of 0.46.

CONCLUSION

The developed screening tool is highly sensitive and correctly differentiates women with and without gestational diabetes mellitus in a Tanzanian sub-population.

Prediction of type 2 diabetes mellitus onset using logistic regression-based scorecards

Background

Type 2 diabetes (T2D) accounts for ~90% of all cases of diabetes, resulting in an estimated 6.7 million deaths in 2021, according to the International Diabetes Federation. Early detection of patients with high risk of developing T2D can reduce the incidence of the disease through a change in lifestyle, diet, or medication. Since populations of lower socio-demographic status are more susceptible to T2D and might have limited resources or access to sophisticated computational resources, there is a need for accurate yet accessible prediction models.

Methods

In this study, we analyzed data from 44,709 nondiabetic UK Biobank participants aged 40-69, predicting the risk of T2D onset within a selected time frame (mean of 7.3 years with an SD of 2.3 years). We started with 798 features that we identified as potential predictors for T2D onset. We first analyzed the data using gradient boosting decision trees, survival analysis, and logistic regression methods. We devised one nonlaboratory model accessible to the general population and one more precise yet simple model that utilizes laboratory tests. We simplified both models to an accessible scorecard form, tested the models on normoglycemic and prediabetes subcohorts, and compared the results to the results of the general cohort. We established the nonlaboratory model using the following covariates: sex, age, weight, height, waist size, hip circumference, waist-to-hip ratio, and body mass index. For the laboratory model, we used age and sex together with four common blood tests: high-density lipoprotein (HDL), gamma-glutamyl transferase, glycated hemoglobin, and triglycerides. As an external validation dataset, we used the electronic medical record database of Clalit Health Services.

Results

The nonlaboratory scorecard model achieved an area under the receiver operating curve (auROC) of 0.81 (95% confidence interval [CI] 0.77-0.84) and an odds ratio (OR) between the upper and fifth prevalence deciles of 17.2 (95% CI 5-66). Using this model, we classified three risk groups, a group with 1% (0.8-1%), 5% (3-6%), and the third group with a 9% (7-12%) risk of developing T2D. We further analyzed the contribution of the laboratory-based model and devised a blood test model based on age, sex, and the four common blood tests noted above. In this scorecard model, we included age, sex, glycated hemoglobin (HbA1c%), gamma glutamyl-transferase, triglycerides, and HDL cholesterol. Using this model, we achieved an auROC of 0.87 (95% CI 0.85-0.90) and a deciles' OR of ×48 (95% CI 12-109). Using this model, we classified the cohort into four risk groups with the following risks: 0.5% (0.4-7%); 3% (2-4%); 10% (8-12%); and a high-risk group of 23% (10-37%) of developing T2D. When applying the blood tests model using the external validation cohort (Clalit), we achieved an auROC of 0.75 (95% CI 0.74-0.75). We analyzed several additional comprehensive models, which included genotyping data and other environmental factors. We found that these models did not provide cost-efficient benefits over the four blood test model. The commonly used German Diabetes Risk Score (GDRS) and Finnish Diabetes Risk Score (FINDRISC) models, trained using our data, achieved an auROC of 0.73 (0.69-0.76) and 0.66 (0.62-0.70), respectively, inferior to the results achieved by the four blood test model and by the anthropometry models.

Conclusions

The four blood test and anthropometric models outperformed the commonly used nonlaboratory models, the FINDRISC and the GDRS. We suggest that our models be used as tools for decision-makers to assess populations at elevated T2D risk and thus improve medical strategies. These models might also provide a personal catalyst for changing lifestyle, diet, or medication modifications to lower the risk of T2D onset.

Funding

The funders had no role in study design, data collection, interpretation, or the decision to submit the work for publication.

Risk assessment with gut microbiome and metabolite markers in NAFLD development

A growing body of evidence suggests interplay between the gut microbiota and the pathogenesis of nonalcoholic fatty liver disease (NAFLD). However, the role of the gut microbiome in early detection of NAFLD is unclear. Prospective studies are necessary for identifying reliable, microbiome markers for early NAFLD. We evaluated 2487 individuals in a community-based cohort who were followed up 4.6 years after initial clinical examination and biospecimen sampling. Metagenomic and metabolomic characterizations using stool and serum samples taken at baseline were performed for 90 participants who progressed to NAFLD and 90 controls who remained NAFLD free at the follow-up visit. Cases and controls were matched for gender, age, body mass index (BMI) at baseline and follow-up, and 4-year BMI change. Machine learning models integrating baseline microbial signatures (14 features) correctly classified participants (auROCs of 0.72 to 0.80) based on their NAFLD status and liver fat accumulation at the 4-year follow up, outperforming other prognostic clinical models (auROCs of 0.58 to 0.60). We confirmed the biological relevance of the microbiome features by testing their diagnostic ability in four external NAFLD case-control cohorts examined by biopsy or magnetic resonance spectroscopy, from Asia, Europe, and the United States. Our findings raise the possibility of using gut microbiota for early clinical warning of NAFLD development.

Improving preeclampsia risk prediction by modeling pregnancy trajectories from routinely collected electronic medical record data

Preeclampsia is a heterogeneous and complex disease associated with rising morbidity and mortality in pregnant women and newborns in the US. Early recognition of patients at risk is a pressing clinical need to reduce the risk of adverse outcomes. We assessed whether information routinely collected in electronic medical records (EMR) could enhance the prediction of preeclampsia risk beyond what is achieved in standard of care assessments. We developed a digital phenotyping algorithm to curate 108,557 pregnancies from EMRs across the Mount Sinai Health System, accurately reconstructing pregnancy journeys and normalizing these journeys across different hospital EMR systems. We then applied machine learning approaches to a training dataset (N = 60,879) to construct predictive models of preeclampsia across three major pregnancy time periods (ante-, intra-, and postpartum). The resulting models predicted preeclampsia with high accuracy across the different pregnancy periods, with areas under the receiver operating characteristic curves (AUC) of 0.92, 0.82, and 0.89 at 37 gestational weeks, intrapartum and postpartum, respectively. We observed comparable performance in two independent patient cohorts. While our machine learning approach identified known risk factors of preeclampsia (such as blood pressure, weight, and maternal age), it also identified other potential risk factors, such as complete blood count related characteristics for the antepartum period. Our model not only has utility for earlier identification of patients at risk for preeclampsia, but given the prediction accuracy exceeds what is currently achieved in clinical practice, our model provides a path for promoting personalized precision therapeutic strategies for patients at risk.

Automated Machine Learning (AutoML)-Derived Preconception Predictive Risk Model to Guide Early Intervention for Gestational Diabetes Mellitus

The increasing prevalence of gestational diabetes mellitus (GDM) is contributing to the rising global burden of type 2 diabetes (T2D) and intergenerational cycle of chronic metabolic disorders. Primary lifestyle interventions to manage GDM, including second trimester dietary and exercise guidance, have met with limited success due to late implementation, poor adherence and generic guidelines. In this study, we aimed to build a preconception-based GDM predictor to enable early intervention. We also assessed the associations of top predictors with GDM and adverse birth outcomes. Our evolutionary algorithm-based automated machine learning (AutoML) model was implemented with data from 222 Asian multi-ethnic women in a preconception cohort study, Singapore Preconception Study of Long-Term Maternal and Child Outcomes (S-PRESTO). A stacked ensemble model with a gradient boosting classifier and linear support vector machine classifier (stochastic gradient descent training) was derived using genetic programming, achieving an excellent AUC of 0.93 based on four features (glycated hemoglobin A1c (HbA1c), mean arterial blood pressure, fasting insulin, triglycerides/HDL ratio). The results of multivariate logistic regression model showed that each 1 mmol/mol increase in preconception HbA1c was positively associated with increased risks of GDM (p = 0.001, odds ratio (95% CI) 1.34 (1.13-1.60)) and preterm birth (p = 0.011, odds ratio 1.63 (1.12-2.38)). Optimal control of preconception HbA1c may aid in preventing GDM and reducing the incidence of preterm birth. Our trained predictor has been deployed as a web application that can be easily employed in GDM intervention programs, prior to conception.

Evolution of Mitochondrially Derived Peptides Humanin and MOTSc, and Changes in Insulin Sensitivity during Early Gestation in Women with and without Gestational Diabetes

Our purpose is to study the evolution of mitochondrially derived peptides (MDPs) and their relationship with changes in insulin sensitivity from the early stages of pregnancy in a cohort of pregnant women with and without gestational diabetes (GDM). MDPs (humanin and MOTSc) were assessed in the first and second trimesters of gestation in 28 pregnant women with gestational diabetes mellitus (GDM) and a subgroup of 45 pregnant women without GDM matched by BMI, age, previous gestations, and time of sampling. Insulin resistance (IR) was defined as a HOMA-IR index ≥70th percentile. We observed a significant reduction in both humanin and MOTSc levels from the first to the second trimesters of pregnancy. After adjusting for predefined variables, including BMI, statistically nonsignificant associations between lower levels of humanin and the occurrence of a high HOMA-IR index were obtained (adjusted OR = 2.63 and 3.14 for the first and second trimesters, linear p-trend 0.260 and 0.175, respectively). Regarding MOTSc, an association was found only for the second trimester: adjusted OR = 7.68 (95% CI 1.49–39.67), linear p-trend = 0.012. No significant associations were observed in humanin change with insulin resistance throughout pregnancy, but changes in MOTSc levels were significantly associated with HOMA-IR index: adjusted OR 3.73 (95% CI 1.03–13.50). In conclusion, MOTSc levels, especially a strong decrease from the first to second trimester of gestation, may be involved in increasing insulin resistance during early gestation.

Circulating Proteome and Progression of Type 2 Diabetes

CONTEXT

Circulating proteomes may provide intervention targets for type 2 diabetes (T2D).

OBJECTIVE

We aimed to identify proteomic biomarkers associated with incident T2D and assess its joint effect with dietary or lifestyle factors on the T2D risk.

METHODS

We established 2 nested case-control studies for incident T2D: discovery cohort (median 6.5 years of follow-up, 285 case-control pairs) and validation cohort (median 2.8 years of follow-up, 38 case-control pairs). We integrated untargeted mass spectrometry-based proteomics and interpretable machine learning to identify T2D-related proteomic biomarkers. We constructed a protein risk score (PRS) with the identified proteomic biomarkers and used a generalized estimating equation to evaluate PRS-T2D relationship with repeated profiled proteome. We evaluated association of PRS with trajectory of glycemic traits in another non-T2D cohort (n = 376). Multiplicative interactions of dietary or lifestyle factors with PRS were evaluated using logistic regression.

RESULTS

Seven proteins (SHBG, CAND1, APOF, SELL, MIA3, CFH, IGHV1-2) were retained as the proteomic biomarkers for incident T2D. PRS (per SD change) was positively associated with incident T2D across 2 cohorts, with an odds ratio 1.29 (95% CI, 1.08-1.54) and 1.84 (1.19-2.84), respectively. Participants with a higher PRS had a higher probability showing unfavored glycemic trait trajectory in the non-T2D cohort. Red meat intake and PRS showed a multiplicative interaction on T2D risk in the discovery (P = 0.003) and validation cohort (P = 0.017).

CONCLUSION

This study identified proteomic biomarkers for incident T2D among the Chinese populations. The higher intake of red meat may synergistically interact with the proteomic biomarkers to exaggerate the T2D risk.

Development and evaluation of an online questionnaire to identify women at high and low risk of developing gestational diabetes mellitus

BACKGROUND

Established risk factors for Gestational Diabetes Mellitus (GDM) include age, ethnicity, family history of diabetes and previous GDM. Additional significant influences have recently been demonstrated in the literature. The oral glucose tolerance test (OGTT) used for GDM diagnosis has sub-optimal sensitivity and specificity, thus often results in GDM misdiagnoses. Comprehensive screening of risk factors may allow more targeted monitoring and more accurate diagnoses, preventing the devastating consequences of untreated or misdiagnosed GDM. We aimed to develop a comprehensive online questionnaire of GDM risk factors and triangulate it with the OGTT and continuous glucose monitoring (CGM) parameters to better evaluate GDM risk and diagnosis.

METHODS

Pregnant women participating in two studies on the use of CGM for GDM were invited to complete the online questionnaire. A risk score, based on published literature, was calculated for each participant response and compared with the OGTT result. A total risk score (TRS) was then calculated as a normalised sum of all risk factors. Triangulation of OGTT, TRS and CGM score of variability (CGMSV) was analysed to expand evaluation of OGTT results.

RESULTS

Fifty one women completed the questionnaire; 29 were identified as 'high-risk' for GDM. High-risk ethnic background (p < 0.01), advanced age, a family diabetic history (p < 0.05) were associated with a positive OGTT result. The triangulation analysis (n = 45) revealed six (13%) probable misdiagnoses (both TRS and CGMSV discordant with OGTT), consisting of one probable false positive and five probable false negative by OGTT results.

CONCLUSIONS

This study identified pregnant women at high risk of developing GDM based on an extended evaluation of risk factors. Triangulation of TRS, OGTT and CGMSV suggested potential misdiagnoses of the OGTT. Future studies to explore the correlation between TRS, CGMSV and pregnancy outcomes as well as additional GDM pregnancy biomarkers and outcomes to efficiently evaluate OGTT results are needed.

Cohort design and natural language processing to reduce bias in electronic health records research

Electronic health record (EHR) datasets are statistically powerful but are subject to ascertainment bias and missingness. Using the Mass General Brigham multi-institutional EHR, we approximated a community-based cohort by sampling patients receiving longitudinal primary care between 2001-2018 (Community Care Cohort Project [C3PO], n = 520,868). We utilized natural language processing (NLP) to recover vital signs from unstructured notes. We assessed the validity of C3PO by deploying established risk models for myocardial infarction/stroke and atrial fibrillation. We then compared C3PO to Convenience Samples including all individuals from the same EHR with complete data, but without a longitudinal primary care requirement. NLP reduced the missingness of vital signs by 31%. NLP-recovered vital signs were highly correlated with values derived from structured fields (Pearson r range 0.95-0.99). Atrial fibrillation and myocardial infarction/stroke incidence were lower and risk models were better calibrated in C3PO as opposed to the Convenience Samples (calibration error range for myocardial infarction/stroke: 0.012-0.030 in C3PO vs. 0.028-0.046 in Convenience Samples; calibration error for atrial fibrillation 0.028 in C3PO vs. 0.036 in Convenience Samples). Sampling patients receiving regular primary care and using NLP to recover missing data may reduce bias and maximize generalizability of EHR research.

PREDICTION OF GESTATIONAL DIABETES AND HYPERTENSION BASED ON PREGNANCY EXAMINATION DATA

Gestational diabetes mellitus and hypertension are two common pregnancy complications, which seriously threaten the life safety of pregnant women and adversely affect the growth and development of the fetus. Therefore, it is of great significance to detect and prevent hypertension and diabetes at an early stage of pregnancy. Each pregnant woman will undergo multiple tests at different gestational weeks. This progress produces lots of pregnancy examination data. These data can reflect the dynamic changes of pregnant women’s health indicators during pregnancy. This study aims to establish gestational diabetes and hypertension prediction model with a machine learning method based on real pregnancy examination data from the hospital. We use Logistic Regression, XGBoost, LightGBM, and Neural Network Model based on LSTM to do the prediction, respectively, and compare the performance. We check the prediction accuracy at different stages of pregnancy. We found that with pregnancy examination data at all gestational weeks, the predictive AUCs for diabetes and hypertension can reach 0.92 and 0.87, respectively. At 16th gestational week, the AUCs are 0.68 for diabetes and 0.70 for hypertension. We extract the checking items which are most important and get a simplified model with a modest reduction in predictive accuracy. This study demonstrates that based on several routine pregnancy examination items we can establish a machine learning model to detect and predict gestational diabetes and hypertension. This can be used as a diagnostic aid and is conducive to early prevention and treatment.

Sulfated Progesterone Metabolites That Enhance Insulin Secretion via TRPM3 Are Reduced in Serum From Women With Gestational Diabetes Mellitus

Serum progesterone sulfates were evaluated in the etiology of gestational diabetes mellitus (GDM). Serum progesterone sulfates were measured using ultra-performance liquid chromatography-tandem mass spectrometry in four patient cohorts: 1) the Hyperglycemia and Adverse Pregnancy Outcomes study; 2) London-based women of mixed ancestry and 3) U.K.-based women of European ancestry with or without GDM; and 4) 11-13 weeks pregnant women with BMI ≤25 or BMI ≥35 kg/m2 with subsequent uncomplicated pregnancies or GDM. Glucose-stimulated insulin secretion (GSIS) was evaluated in response to progesterone sulfates in mouse islets and human islets. Calcium fluorescence was measured in HEK293 cells expressing transient receptor potential cation channel subfamily M member 3 (TRPM3). Computer modeling using Molecular Operating Environment generated three-dimensional structures of TRPM3. Epiallopregnanolone sulfate (PM5S) concentrations were reduced in GDM (P < 0.05), in women with higher fasting plasma glucose (P < 0.010), and in early pregnancy samples from women who subsequently developed GDM with BMI ≥35 kg/m2 (P < 0.05). In islets, 50 µmol/L PM5S increased GSIS by at least twofold (P < 0.001); isosakuranetin (TRPM3 inhibitor) abolished this effect. PM5S increased calcium influx in TRPM3-expressing HEK293 cells. Computer modeling and docking showed identical positioning of PM5S to the natural ligand in TRPM3. PM5S increases GSIS and is reduced in GDM serum. The activation of GSIS by PM5S is mediated by TRPM3 in both mouse and human islets.

Population-centric risk prediction modeling for gestational diabetes mellitus: A machine learning approach

AIMS

The heterogeneity in Gestational Diabetes Mellitus (GDM) risk factors among different populations impose challenges in developing a generic prediction model. This study evaluates the predictive ability of existing UK NICE guidelines for assessing GDM risk in Singaporean women, and used machine learning to develop a non-invasive predictive model.

METHODS

Data from 909 pregnancies in Singapore's most deeply phenotyped mother-offspring cohort study, Growing Up in Singapore Towards healthy Outcomes (GUSTO), was used for predictive modeling. We used a CatBoost gradient boosting algorithm, and the Shapley feature attribution framework for model building and interpretation of GDM risk attributes.

RESULTS

UK NICE guidelines showed poor predictability in Singaporean women [AUC:0.60 (95% CI 0.51, 0.70)]. The non-invasive predictive model comprising of 4 non-invasive factors: mean arterial blood pressure in first trimester, age, ethnicity and previous history of GDM, greatly outperformed [AUC:0.82 (95% CI 0.71, 0.93)] the UK NICE guidelines.

CONCLUSIONS

The UK NICE guidelines may be insufficient to assess GDM risk in Asian women. Our non-invasive predictive model outperforms the current state-of-the-art machine learning models to predict GDM, is easily accessible and can be an effective approach to minimize the economic burden of universal testing & GDM associated healthcare in Asian populations.

The Contribution of Data-Driven Technologies in Achieving the Sustainable Development Goals

The United Nations’ Sustainable Development Goals (SDGs) set out to improve the quality of life of people in developed, emerging, and developing countries by covering social and economic aspects, with a focus on environmental sustainability. At the same time, data-driven technologies influence our lives in all areas and have caused fundamental economical and societal changes. This study presents a comprehensive literature review on how data-driven approaches have enabled or inhibited the successful achievement of the 17 SDGs to date. Our findings show that data-driven analytics and tools contribute to achieving the 17 SDGs, e.g., by making information more reliable, supporting better-informed decision-making, implementing data-based policies, prioritizing actions, and optimizing the allocation of resources. Based on a qualitative content analysis, results were aggregated into a conceptual framework, including the following categories: (1) uses of data-driven methods (e.g., monitoring, measurement, mapping or modeling, forecasting, risk assessment, and planning purposes), (2) resulting positive effects, (3) arising challenges, and (4) recommendations for action to overcome these challenges. Despite positive effects and versatile applications, problems such as data gaps, data biases, high energy consumption of computational resources, ethical concerns, privacy, ownership, and security issues stand in the way of achieving the 17 SDGs.

An explainable machine learning-based clinical decision support system for prediction of gestational diabetes mellitus

Gestational Diabetes Mellitus (GDM), a common pregnancy complication associated with many maternal and neonatal consequences, is increased in mothers with overweight and obesity. Interventions initiated early in pregnancy can reduce the rate of GDM in these women, however, untargeted interventions can be costly and time-consuming. We have developed an explainable machine learning-based clinical decision support system (CDSS) to identify at-risk women in need of targeted pregnancy intervention. Maternal characteristics and blood biomarkers at baseline from the PEARS study were used. After appropriate data preparation, synthetic minority oversampling technique and feature selection, five machine learning algorithms were applied with five-fold cross-validated grid search optimising the balanced accuracy. Our models were explained with Shapley additive explanations to increase the trustworthiness and acceptability of the system. We developed multiple models for different use cases: theoretical (AUC-PR 0.485, AUC-ROC 0.792), GDM screening during a normal antenatal visit (AUC-PR 0.208, AUC-ROC 0.659), and remote GDM risk assessment (AUC-PR 0.199, AUC-ROC 0.656). Our models have been implemented as a web server that is publicly available for academic use. Our explainable CDSS demonstrates the potential to assist clinicians in screening at risk patients who may benefit from early pregnancy GDM prevention strategies.

Using Machine Learning to Predict Complications in Pregnancy: A Systematic Review

Introduction: Artificial intelligence is widely used in medical field, and machine learning has been increasingly used in health care, prediction, and diagnosis and as a method of determining priority. Machine learning methods have been features of several tools in the fields of obstetrics and childcare. This present review aims to summarize the machine learning techniques to predict perinatal complications.

Objective: To identify the applicability and performance of machine learning methods used to identify pregnancy complications.

Methods: A total of 98 articles were obtained with the keywords “machine learning,” “deep learning,” “artificial intelligence,” and accordingly as they related to perinatal complications (“complications in pregnancy,” “pregnancy complications”) from three scientific databases: PubMed, Scopus, and Web of Science. These were managed on the Mendeley platform and classified using the PRISMA method.

Results: A total of 31 articles were selected after elimination according to inclusion and exclusion criteria. The features used to predict perinatal complications were primarily electronic medical records (48%), medical images (29%), and biological markers (19%), while 4% were based on other types of features, such as sensors and fetal heart rate. The main perinatal complications considered in the application of machine learning thus far are pre-eclampsia and prematurity. In the 31 studies, a total of sixteen complications were predicted. The main precision metric used is the AUC. The machine learning methods with the best results were the prediction of prematurity from medical images using the support vector machine technique, with an accuracy of 95.7%, and the prediction of neonatal mortality with the XGBoost technique, with 99.7% accuracy.

Conclusion: It is important to continue promoting this area of research and promote solutions with multicenter clinical applicability through machine learning to reduce perinatal complications. This systematic review contributes significantly to the specialized literature on artificial intelligence and women’s health.

Circulating exosomal hsa_circRNA_0039480 is highly expressed in gestational diabetes mellitus and may be served as a biomarker for early diagnosis of GDM

Background

Gestational diabetes mellitus (GDM) seriously affects the health of mothers and babies, and there are still no effective early diagnostic markers. Therefore, it is necessary to find diagnostic biomarkers for screening GDM in early pregnancy. Circular RNA (circRNA) is more stable than linear RNA, and can be encapsulated in exosomes and participate in the pathological process of various diseases, which makes it a better candidate biomarker for various diseases. In this study, we attempted to identify the exosomal circRNA biomarkers for detecting early GDM.

Methods

We performed microarray analysis to compare the plasma exosomal circRNA expression profiles of three GDM patients 48 h before and 48 h after delivery. The repeatability of the expression of circRNAs were randomly validated by RT-PCR analysis. Pearson correlation analysis was applied to evaluate the correlation between circRNAs and OGTT level. ROC curve was established to assess the diagnostic value of circRNAs for GDM at different stages.

Results

Plasma exosomal hsa_circRNA_0039480 and hsa_circRNA_0026497 were highly expressed in GDM patients before delivery (P < 0.05). The hsa_circRNA_0039480 expression was higher for GDM group than NGT group at different stages, and was also positively correlated with OGTT during the second trimester (P < 0.05). The expression of hsa_circRNA_0026497 was higher for GDM group during the third, and second trimesters. And there was a strong correlation between two circRNAs in GDM patients during the first-trimester (r = 0.496, P = 0.014). Hsa_circRNA_0039480 showed significant diagnostic value in the first, second, and third trimesters of pregnancy (AUC = 0.704, P = 0.005; AUC = 0.898, P < 0.001 and AUC = 0.698, P = 0.001, respectively). Notably, the combination of hsa_circRNA_0039480 and hsa_circRNA_0026497 exhibited promising discriminative effect on GDM in the first trimesters (AUC = 0.754, P < 0.001).

Conclusion

Plasma exosomal hsa_cirRNA_0039480 is highly expressed in GDM patients at different stages and may be served as a candidate biomarker for early detection of GDM.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-021-03195-5.

Gestational diabetes mellitus in women born small or preterm: Systematic review and meta-analysis

PURPOSE

There is some evidence that women born preterm or with low birth weight (LBW) have an increased future risk of gestational diabetes mellitus (GDM) during pregnancy; however, a quantitative summary of evidence is lacking. In this systematic review and meta-analysis, we examined the published data to investigate whether being born preterm, with LBW or small for gestational age (SGA) are associated with GDM risk.

METHODS

We searched the MEDLINE, Embase, and CINAHL databases and study registries, including ClinicalTrials.gov and ICTRP, from launch until 29 October 2020. Observational studies examining the association between birth weight or gestational age and GDM were eligible. We pooled the odds ratios and 95% confidence intervals using the DerSimonian and Laird random-effects model.

RESULTS

Eighteen studies were included (N = 827,382). The meta-analysis showed that being born preterm, with LBW or SGA was associated with increased risk of GDM (pooled odds ratio = 1.84; 95% confidence interval: 1.54-2.20; I² = 78.3%; τ² = 0.07). Given a GDM prevalence of 2.0, 10, and 20%, the absolute risk differences were 1.6%, 7.0%, and 11.5%, respectively. The certainty of the evidence was low due to serious concerns of risk of bias and publication bias.

CONCLUSIONS

Women born prematurely, with LBW or SGA status, may be at increased risk for GDM. However, whether this should be considered in clinical decision-making depends on the prevalence of GDM.

Association of maternal triglyceride responses to thyroid function in early pregnancy with gestational diabetes mellitus

INTRODUCTION

The prevalence of Gestational Diabetes Mellitus (GDM) is increasing globally, and high levels of triglyceride (TG) and low levels of free thyroxine (FT4) in early pregnancy are associated with an increased risk of GDM; however, the interaction and mediation effects remain unknown. The aim of the present study is to examine the impact of FT4 and TG combined effects on the prevalence of GDM and the corresponding casual paths among women in early pregnancy.

MATERIALS AND METHODS

This study comprised 40,156 pregnant women for whom early pregnancy thyroid hormones, fasting blood glucose as well as triglyceride were available. GDM was diagnosed using a 2-hour 75-g oral glucose tolerance test (OGTT) according to the American Diabetes Association guidelines, and the pregnant women were grouped and compared according to the results.

RESULTS

An L-shaped association between FT4 and GDM was observed. The prevalence of GDM increased with increasing TG levels. After accounting for multiple covariables, the highest risk for GDM was found among pregnant women of lower FT4 with the highest TG concentrations (odds ratio, 2.44, 95% CI, 2.14 to 2.80; P<0.001) compared with mothers of higher FT4 with the TG levels in the lowest quartile (Q1). There was a significant interaction effect of maternal FT4 and TG levels on the risk for GDM (P for interaction = 0.036). The estimated proportion of the mediating effect of maternal TG levels was 21.3% (95% CI, 15.6% to 36.0%; P < 0.001). In the sensitivity analysis, the mediating effect of TG levels was stable across subgroups.

CONCLUSION

This study demonstrated an L-shaped association between maternal FT4 levels and GDM and the benefit of low TG levels, in which maternal TG levels act as an important mediator in this association. Our findings suggested that pregnant women who treat hypothyroidism should also reduce triglycerides levels in early pregnancy to prevent GDM development.

mHealth apps for gestational diabetes mellitus that provide clinical decision support or artificial intelligence: A scoping review

AIMS

Gestational diabetes (GDM) is the most common metabolic disorder of pregnancy, requiring complex management and empowerment of those affected. Mobile health (mHealth) applications (apps) are proposed for streamlining healthcare service delivery, extending care relationships into the community, and empowering those affected by prolonged medical disorders to be equal collaborators in their healthcare. This review investigates mHealth apps intended for use with GDM; specifically those powered by artificial intelligence (AI) or providing decision support.

METHODS

A scoping review using the novel Survey Tool approach for collaborative literature Reviews (STaR) process was performed.

RESULTS

From 18 papers, 11 discrete GDM-based mHealth apps were identified, but only 3 were reasonably mature with only one currently in use in a clinical setting. Two-thirds of the apps provided condition-relevant contextual user feedback that could aid in patient self care. However, although each app targeted one or more components of the GDM clinical pathway, no app addressed the entirety from diagnosis to postpartum.

CONCLUSIONS

There are limited mHealth apps for GDM that incorporate AI or AI-based decision support. Many exist only to record patient information like blood glucose readings or diet, provide generic patient education or advice, or to reduce adverse events by providing medication or appointment alerts. Significant barriers remain that continue to limit the adoption of mHealth apps in clinical care settings. Further research and development are needed to deliver intelligent holistic mHealth apps using AI that can truly reduce healthcare resource use and improve outcomes by enabling patient self care in the community.

Development and validation of a machine learning-augmented algorithm for diabetes screening in community and primary care settings: A population-based study

BACKGROUND

Opportunely screening for diabetes is crucial to reduce its related morbidity, mortality, and socioeconomic burden. Machine learning (ML) has excellent capability to maximize predictive accuracy. We aim to develop ML-augmented models for diabetes screening in community and primary care settings.

METHODS

8425 participants were involved from a population-based study in Hubei, China since 2011. The dataset was split into a development set and a testing set. Seven different ML algorithms were compared to generate predictive models. Non-laboratory features were employed in the ML model for community settings, and laboratory test features were further introduced in the ML+lab models for primary care. The area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (auPR), and the average detection costs per participant of these models were compared with their counterparts based on the New China Diabetes Risk Score (NCDRS) currently recommended for diabetes screening.

RESULTS

The AUC and auPR of the ML model were 0·697and 0·303 in the testing set, seemingly outperforming those of NCDRS by 10·99% and 64·67%, respectively. The average detection cost of the ML model was 12·81% lower than that of NCDRS with the same sensitivity (0·72). Moreover, the average detection cost of the ML+FPG model is the lowest among the ML+lab models and less than that of the ML model and NCDRS+FPG model.

CONCLUSION

The ML model and the ML+FPG model achieved higher predictive accuracy and lower detection costs than their counterpart based on NCDRS. Thus, the ML-augmented algorithm is potential to be employed for diabetes screening in community and primary care settings.

Characterization of Symptoms and Symptom Clusters for Type 2 Diabetes Using a Large Nationwide Electronic Health Record Database

OBJECTIVE

A variety of symptoms may be associated with type 2 diabetes and its complications. Symptoms in chronic diseases may be described in terms of prevalence, severity, and trajectory and often co-occur in groups, known as symptom clusters, which may be representative of a common etiology. The purpose of this study was to characterize type 2 diabetes-related symptoms using a large nationwide electronic health record (EHR) database.

METHODS

We acquired the Cerner Health Facts, a nationwide EHR database. The type 2 diabetes cohort (n = 1,136,301 patients) was identified using a rule-based phenotype method. A multistep procedure was then used to identify type 2 diabetes-related symptoms based on International Classification of Diseases, 9th and 10th revisions, diagnosis codes. Type 2 diabetes-related symptoms and co-occurring symptom clusters, including their temporal patterns, were characterized based the longitudinal EHR data.

RESULTS

Patients had a mean age of 61.4 years, 51.2% were female, and 70.0% were White. Among 1,136,301 patients, there were 8,008,276 occurrences of 59 symptoms. The most frequently reported symptoms included pain, heartburn, shortness of breath, fatigue, and swelling, which occurred in 21-60% of the patients. We also observed over-represented type 2 diabetes symptoms, including difficulty speaking, feeling confused, trouble remembering, weakness, and drowsiness/sleepiness. Some of these are rare and difficult to detect by traditional patient-reported outcomes studies.

CONCLUSION

To the best of our knowledge, this is the first study to use a nationwide EHR database to characterize type 2 diabetes-related symptoms and their temporal patterns. Fifty-nine symptoms, including both over-represented and rare diabetes-related symptoms, were identified.

Physician Knowledge Base: Clinical Decision Support Systems

With the introduction of electronic medical records (EMRs), it has become possible to accumulate massive amounts of qualitative medical data. As such, EMRs have become increasingly used in clinical decision support systems (CDSSs). While CDSSs aim to reduce medical errors normally occurring in the process of treating patients by physicians, technical maturity and the completeness of CDSSs do not meet standards for medical use yet. As data further accumulates, CDSS algorithms must be continuously updated to allow CDSSs to perform their core functions. Doing so, however, requires extensive time and manpower investments. In current practice, computational systems already perform a wide variety of functions in medical settings to allow medical staff to focus on other tasks. However, no prior research has evaluated the potential effectiveness of future CDSSs nor analyzed possibilities for their further development. In this article, we evaluate CDSS technology with the consideration that medical staff also understand the core functions of such systems.

Machine learning approaches for the prediction of postoperative complication risk in liver resection patients

Background

For liver cancer patients, the occurrence of postoperative complications increases the difficulty of perioperative nursing, prolongs the hospitalization time of patients, and leads to large increases in hospitalization costs. The ability to identify influencing factors and to predict the risk of complications in patients with liver cancer after surgery could assist doctors to make better clinical decisions.

Objective

The aim of the study was to develop a postoperative complication risk prediction model based on machine learning algorithms, which utilizes variables obtained before or during the liver cancer surgery, to predict when complications present with clinical symptoms and the ways of reducing the risk of complications.

Methods

The study subjects were liver cancer patients who had undergone liver resection. There were 175 individuals, and 13 variables were recorded. 70% of the data were used for the training set, and 30% for the test set. The performance of five machine learning models, logistic regression, decision trees-C5.0, decision trees-CART, support vector machines, and random forests, for predicting postoperative complication risk in liver resection patients were compared. The significant influencing factors were selected by combining results of multiple methods, based on which the prediction model of postoperative complications risk was created. The results were analyzed to give suggestions of how to reduce the risk of complications.

Results

Random Forest gave the best performance from the decision curves analysis. The decision tree-C5.0 algorithm had the best performance of the five machine learning algorithms if ACC and AUC were used as evaluation indicators, producing an area under the receiver operating characteristic curve value of 0.91 (95% CI 0.77–1), with an accuracy of 92.45% (95% CI 85–100%), the sensitivity of 87.5%, and specificity of 94.59%. The duration of operation, patient’s BMI, and length of incision were significant influencing factors of postoperative complication risk in liver resection patients.

Conclusions

To reduce the risk of complications, it appears to be important that the patient's BMI should be above 22.96 before the operation, and the duration of the operation should be minimized.

TiO2 NPs induce the reproductive toxicity in mice with gestational diabetes mellitus through the effects on the endoplasmic reticulum stress signaling pathway

The effect of one of the most widely studied nanomaterials at present, TiO₂ nanoparticles (NPs), on pregnancy-related diseases is not clear. In this study, the adverse effects of TiO₂ NPs on mice with gestational diabetes mellitus (GDM) and their possible mechanism were investigated. GDM mice were orally administered 0, 10, 50 and 250 mg/kg TiO₂ NPs for 14 days. GDM reduced the weight of pregnant mice, destroyed the placental structure and caused abnormal fetal development. After exposure to increasing doses of TiO₂ NPs, blood glucose levels increased significantly and body weight further decreased in GDM mice. The accumulation of the Ti content was detected in the placenta and fetus, which may further damage the placental structure in GDM mice, thereby exacerbating abnormal fetal development. In addition, the MDA and SOD activities were obviously increased, and the expression of genes associated with endoplasmic reticulum stress (ERS) (PERK, eIF2α, AFT4, IRE1α, and XBP1s) and apoptosis (CHOP, JNK, Bax/Bcl-2, Caspase-12, Caspase-9, and Caspase-3) were also obviously increased in the placenta, which reflected the possible activation of apoptosis. It could be speculated that the reproductive toxicity of TiO₂ NPs in GDM mice triggered oxidative stress that subsequently activated ERS pathways to induce cell apoptosis.

Estimating heritability of glycaemic response to metformin using nationwide electronic health records and population-sized pedigree

Background

Variability of response to medication is a well-known phenomenon, determined by both environmental and genetic factors. Understanding the heritable component of the response to medication is of great interest but challenging due to several reasons, including small study cohorts and computational limitations.

Methods

Here, we study the heritability of variation in the glycaemic response to metformin, first-line therapeutic agent for type 2 diabetes (T2D), by leveraging 18 years of electronic health records (EHR) data from Israel’s largest healthcare service provider, consisting of over five million patients of diverse ethnicities and socio-economic background. Our cohort consists of 80,788 T2D patients treated with metformin, with an accumulated number of 1,611,591 HbA1C measurements and 4,581,097 metformin prescriptions. We estimate the explained variance of glycated hemoglobin (HbA1c%) reduction due to inheritance by constructing a six-generation population-size pedigree from national registries and linking it to medical health records.

Results

Using Linear Mixed Model-based framework, a common-practice method for heritability estimation, we calculate a heritability measure of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}^{2}=12.6 \%$$\end{document}h2=12.6% (95% CI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$6.1 \%\! -\!19.1 \%$$\end{document}6.1%−19.1%) for absolute reduction of HbA1c% after metformin treatment in the entire cohort, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}^{2}=21.0 \%$$\end{document}h2=21.0% (95% CI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.8 \%\! -\!34.4 \%$$\end{document}7.8%−34.4%) for males and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}^{2}=22.9 \%$$\end{document}h2=22.9% (95% CI, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$10.0 \%\! -\!35.7 \%$$\end{document}10.0%−35.7%) in females. Results remain unchanged after adjusting for pre-treatment HbA1c%, and in proportional reduction of HbA1c%.

Conclusions

To the best of our knowledge, our work is the first to estimate heritability of drug response using solely EHR data combining a pedigree-based kinship matrix. We demonstrate that while response to metformin treatment has a heritable component, most of the variation is likely due to other factors, further motivating non-genetic analyses aimed at unraveling metformin’s action mechanism.

Individuals in a population might respond differently to the same medication and this phenomenon is commonly attributed to either genes or the environment. Here, we studied the familial aspects of the response to metformin, a medication used in the treatment of type 2 diabetes. We combined information from 18 years of medical records identifying newly treated patients with type 2 diabetes with information about how the trait was inherited within their families. We calculated a metric that tells us how well differences in people’s genes account for differences in their traits, and demonstrate that although the difference in response to metformin is in part explained by the genes people with type 2 diabetes inherit, most of it is not explained by genes. This finding contributes to a better understanding of differences in metformin response and might help inform treatment in future.

Kalka and Gavrieli et al. assessed the heritability of variation in the glycaemic response to metformin by leveraging electronic health records data gathered from a large cohort of patients with diabetes and combining it with pedigree information. The authors show that although the variability in this response has a heritable component, most of it is likely non-genetic.

Predicting next-day discharge via electronic health record access logs

OBJECTIVE

Hospital capacity management depends on accurate real-time estimates of hospital-wide discharges. Estimation by a clinician requires an excessively large amount of effort and, even when attempted, accuracy in forecasting next-day patient-level discharge is poor. This study aims to support next-day discharge predictions with machine learning by incorporating electronic health record (EHR) audit log data, a resource that captures EHR users' granular interactions with patients' records by communicating various semantics and has been neglected in outcome predictions.

MATERIALS AND METHODS

This study focused on the EHR data for all adults admitted to Vanderbilt University Medical Center in 2019. We learned multiple advanced models to assess the value that EHR audit log data adds to the daily prediction of discharge likelihood within 24 h and to compare different representation strategies. We applied Shapley additive explanations to identify the most influential types of user-EHR interactions for discharge prediction.

RESULTS

The data include 26 283 inpatient stays, 133 398 patient-day observations, and 819 types of user-EHR interactions. The model using the count of each type of interaction in the recent 24 h and other commonly used features, including demographics and admission diagnoses, achieved the highest area under the receiver operating characteristics (AUROC) curve of 0.921 (95% CI: 0.919-0.923). By contrast, the model lacking user-EHR interactions achieved a worse AUROC of 0.862 (0.860-0.865). In addition, 10 of the 20 (50%) most influential factors were user-EHR interaction features.

CONCLUSION

EHR audit log data contain rich information such that it can improve hospital-wide discharge predictions.

Explainable machine learning model for predicting the occurrence of postoperative malnutrition in children with congenital heart disease

BACKGROUND & AIMS

Malnutrition is persistent in 50%-75% of children with congenital heart disease (CHD) after surgery, and early prediction is crucial for nutritional intervention. The aim of this study was to develop and validate machine learning (ML) models to predict the malnutrition status of children with CHD. We used explainable ML methods to provide insight into the model's predictions and outcomes.

METHODS

This prospective cohort study included consecutive children with CHD admitted to the hospital from December 2017 to May 2020. The cohort data were divided into the training and test data sets based on the follow-up time. The outcome of the study was CHD child malnutrition 1 year after surgery, the primary outcome was an underweight status, and the secondary outcomes were stunted and wasting status. We used five ML algorithms with multiple features to construct prediction models, and the performance of these ML models was measured by an area under the receiver operating characteristic curve (AUC) analysis. We also used the permutation importance and SHapley Additive exPlanations (SHAP) to determine the importance of the selected features and interpret the ML models.

RESULTS

We enrolled 536 children with CHD who underwent complete repair. The proportions of children with an underweight, stunted, or wasting status 1 year after surgery were 18.1% (97/536), 12.1% (65/536), and 17.5% (94/536), respectively. All patients contributed to the generation of 115 useable features, which allowed us to build models to predict malnutrition. Five prediction algorithms were used, and the XGBoost model achieved the greatest AUC in all outcomes. The results obtained from the permutation importance and SHAP analyses showed that the 1-month postoperative WAZ-score, discharge WAZ score and preoperative WAZ score were the top 3 important features in predicting an underweight status in the XGBoost algorithm. Regarding the stunted status, the top 3 important features were the 1-month postoperative HAZ score, discharge HAZ score, and aortic clamping time. Regarding the wasting status, the top 3 important features were the hospital length of stay, formula intake, and discharge WHZ-score. We also used a narrative case report as an example to describe the clinical manifestations and predicted the primary outcomes of two children.

CONCLUSIONS

We developed an ML model (XGBoost) that provides accurate early predictions of malnutrition 1-year postoperatively in children with CHD. Because the ML model is explainable, it may better enable clinicians to better understand the reasoning underlying the outcome. Our study could aid in determining individual treatment and nutritional follow-up strategies for children with CHD.

Novelelectronic health records applied for prediction of pre-eclampsia: Machine-learning algorithms

OBJECTIVE

To predict risk of pre-eclampsia (PE) in women using machine learning (ML) algorithms, based on electronic health records (EHR) collected at the early second trimester.

STUDY DESIGN

A total of 3759 cases of pregnancy who received antenatal care at Xinhua hospital Chongming branch Affiliated to Shanghai Jiaotong University were included in this retrospective EHR-based study. Thirty-eight candidate clinical parameters routinely available at the first visit in antenatal care were collected by manual chart review. Logistic regression (LR), random forest (RF), support vector machine (SVM) and extreme gradient boosting (XGBoost) were used to construct the prediction model. Features that contributed to the model predictions were identified using XGBoost.

OUTCOME MEASURES

The performance of ML models to predict women at risk of PE was quantified in terms of accuracy, precision, recall, false negative score, f1_score, brier score and the area under the receiver operating curve (auROC).

RESULTS

The XGboost model had the best prediction performance (accuracy = 0.920, precision = 0.447, recall = 0.789, f1_score = 0.571, auROC = 0.955). The most predictive feature of PE development was fasting plasma glucose, followed by mean blood pressure and body mass index. An easy-to-use model that a patient could answer independently still enabled accurate prediction, with auROC of 0.83.

CONCLUSION

risk of PE development can be predicted with excellent discriminative ability using ML algorithms based on EHR collected at the early second trimester. Future studies are needed to assess the real-world clinical utility of the model.

Artificial intelligence and diabetes technology: A review

Artificial intelligence (AI) is widely discussed in the popular literature and is portrayed as impacting many aspects of human life, both in and out of the workplace. The potential for revolutionizing healthcare is significant because of the availability of increasingly powerful computational platforms and methods, along with increasingly informative sources of patient data, both in and out of clinical settings. This review aims to provide a realistic assessment of the potential for AI in understanding and managing diabetes, accounting for the state of the art in the methodology and medical devices that collect data, process data, and act accordingly. Acknowledging that many conflicting definitions of AI have been put forth, this article attempts to characterize the main elements of the field as they relate to diabetes, identifying the main perspectives and methods that can (i) affect basic understanding of the disease, (ii) affect understanding of risk factors (genetic, clinical, and behavioral) of diabetes development, (iii) improve diagnosis, (iv) improve understanding of the arc of disease (progression and personal/societal impact), and finally (v) improve treatment.

Increased risk of non-hematological cancer in young patients with aortic stenosis: a retrospective cohort study

Background

We have previously reported an increased risk for non-hematological malignancies in young patients with moderate or severe aortic stenosis (AS). These findings were the result of a post-hoc analysis from a large echocardiography database and needed verification. Our aim was to determine, using a different study population, whether young patients with AS are at increased risk for cancer.

Methods

A large echocardiographic database was used to identify patients (age ≥ 20 years) with moderate or severe AS (study group) and patients without aortic stenosis (comparative group). The new occurrence of non-hematological malignancies was determined after the index date (first echo with moderate or severe AS or first recorded echo in the control group).

Results

The final study group included 7013 patients with AS and 98,884 without AS. During a median follow-up of 6.9 years (3.0–11.1) there were 10,705 new cases of non-hematological cancer. The crude incidence rate of cancer was higher in AS compared to non-AS patients (22.3 vs. 13.7 per 1000 patient-year, crude HR 1.58 (95%CI 1.46–1.71). After adjustment for relevant covariates, there was no difference between groups (HR 0.93, 95% CI 0.86–1.01). Only patients in the lowest age quartile (20–49.7 years), had an increased adjusted risk of cancer (HR 1.91, 95%CI 1.08–3.39). The HR for the risk of cancer associated with AS was inversely proportional to age (P < 0.001 for the interaction between AS and age).

Conclusions

Young patients with moderate or severe AS may have an increased risk for cancer. Cancer surveillance should be considered for young patients with AS.

Nonalcoholic fatty liver disease and early prediction of gestational diabetes using machine learning methods

Background/Aims

To develop an early prediction model for gestational diabetes mellitus (GDM) using machine learning and to evaluate whether the inclusion of nonalcoholic fatty liver disease (NAFLD)-associated variables increases the performance of model.

Methods

This prospective cohort study evaluated pregnant women for NAFLD using ultrasound at 10–14 weeks and screened them for GDM at 24–28 weeks of gestation. The clinical variables before 14 weeks were used to develop prediction models for GDM (setting 1, conventional risk factors; setting 2, addition of new risk factors in recent guidelines; setting 3, addition of routine clinical variables; setting 4, addition of NALFD-associated variables, including the presence of NAFLD and laboratory results; and setting 5, top 11 variables identified from a stepwise variable selection method). The predictive models were constructed using machine learning methods, including logistic regression, random forest, support vector machine, and deep neural networks.

Results

Among 1,443 women, 86 (6.0%) were diagnosed with GDM. The highest performing prediction model among settings 1–4 was setting 4, which included both clinical and NAFLD-associated variables (area under the receiver operating characteristic curve [AUC] 0.563–0.697 in settings 1–3 vs. 0.740–0.781 in setting 4). Setting 5, with top 11 variables (which included NAFLD and hepatic steatosis index), showed similar predictive power to setting 4 (AUC 0.719–0.819 in setting 5, P=not significant between settings 4 and 5).

Conclusions

We developed an early prediction model for GDM using machine learning. The inclusion of NAFLD-associated variables significantly improved the performance of GDM prediction. (ClinicalTrials.gov Identifier: NCT02276144)

Predicting bloodstream infection outcome using machine learning

Bloodstream infections (BSI) are a main cause of infectious disease morbidity and mortality worldwide. Early prediction of BSI patients at high risk of poor outcomes is important for earlier decision making and effective patient stratification. We developed electronic medical record-based machine learning models that predict patient outcomes of BSI. The area under the receiver-operating characteristics curve was 0.82 for a full featured inclusive model, and 0.81 for a compact model using only 25 features. Our models were trained using electronic medical records that include demographics, blood tests, and the medical and diagnosis history of 7889 hospitalized patients diagnosed with BSI. Among the implications of this work is implementation of the models as a basis for selective rapid microbiological identification, toward earlier administration of appropriate antibiotic therapy. Additionally, our models may help reduce the development of BSI and its associated adverse health outcomes and complications.

Closed-Loop Neural Prostheses With On-Chip Intelligence: A Review and a Low-Latency Machine Learning Model for Brain State Detection

The application of closed-loop approaches in systems neuroscience and therapeutic stimulation holds great promise for revolutionizing our understanding of the brain and for developing novel neuromodulation therapies to restore lost functions. Neural prostheses capable of multi-channel neural recording, on-site signal processing, rapid symptom detection, and closed-loop stimulation are critical to enabling such novel treatments. However, the existing closed-loop neuromodulation devices are too simplistic and lack sufficient on-chip processing and intelligence. In this paper, we first discuss both commercial and investigational closed-loop neuromodulation devices for brain disorders. Next, we review state-of-the-art neural prostheses with on-chip machine learning, focusing on application-specific integrated circuits (ASIC). System requirements, performance and hardware comparisons, design trade-offs, and hardware optimization techniques are discussed. To facilitate a fair comparison and guide design choices among various on-chip classifiers, we propose a new energy-area (E-A) efficiency figure of merit that evaluates hardware efficiency and multi-channel scalability. Finally, we present several techniques to improve the key design metrics of tree-based on-chip classifiers, both in the context of ensemble methods and oblique structures. A novel Depth-Variant Tree Ensemble (DVTE) is proposed to reduce processing latency (e.g., by 2.5× on seizure detection task). We further develop a cost-aware learning approach to jointly optimize the power and latency metrics. We show that algorithm-hardware co-design enables the energy- and memory-optimized design of tree-based models, while preserving a high accuracy and low latency. Furthermore, we show that our proposed tree-based models feature a highly interpretable decision process that is essential for safety-critical applications such as closed-loop stimulation.

Postpartum hemorrhage risk is driven by changes in blood composition through pregnancy

The extent to which women differ in the course of blood cell counts throughout pregnancy, and the importance of these changes to pregnancy outcomes has not been well defined. Here, we develop a series of statistical analyses of repeated measures data to reveal the degree to which women differ in the course of pregnancy, predict the changes that occur, and determine the importance of these changes for post-partum hemorrhage (PPH) which is one of the leading causes of maternal mortality. We present a prospective cohort of 4082 births recorded at the University Hospital, Lausanne, Switzerland between 2009 and 2014 where full labour records could be obtained, along with complete blood count data taken at hospital admission. We find significant differences, at a \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p<0.001$$\end{document}p<0.001 level, among women in how blood count values change through pregnancy for mean corpuscular hemoglobin, mean corpuscular volume, mean platelet volume, platelet count and red cell distribution width. We find evidence that almost all complete blood count values show trimester-specific associations with PPH. For example, high platelet count (OR 1.20, 95% CI 1.01–1.53), high mean platelet volume (OR 1.58, 95% CI 1.04–2.08), and high erythrocyte levels (OR 1.36, 95% CI 1.01–1.57) in trimester 1 increased PPH, but high values in trimester 3 decreased PPH risk (OR 0.85, 0.79, 0.67 respectively). We show that differences among women in the course of blood cell counts throughout pregnancy have an important role in shaping pregnancy outcome and tracking blood count value changes through pregnancy improves identification of women at increased risk of postpartum hemorrhage. This study provides greater understanding of the complex changes in blood count values that occur through pregnancy and provides indicators to guide the stratification of patients into risk groups.

A Smart Healthcare Recommendation System for Multidisciplinary Diabetes Patients with Data Fusion Based on Deep Ensemble Learning

The prediction of human diseases precisely is still an uphill battle task for better and timely treatment. A multidisciplinary diabetic disease is a life-threatening disease all over the world. It attacks different vital parts of the human body, like Neuropathy, Retinopathy, Nephropathy, and ultimately Heart. A smart healthcare recommendation system predicts and recommends the diabetic disease accurately using optimal machine learning models with the data fusion technique on healthcare datasets. Various machine learning models and methods have been proposed in the recent past to predict diabetes disease. Still, these systems cannot handle the massive number of multifeatures datasets on diabetes disease properly. A smart healthcare recommendation system is proposed for diabetes disease based on deep machine learning and data fusion perspectives. Using data fusion, we can eliminate the irrelevant burden of system computational capabilities and increase the proposed system's performance to predict and recommend this life-threatening disease more accurately. Finally, the ensemble machine learning model is trained for diabetes prediction. This intelligent recommendation system is evaluated based on a well-known diabetes dataset, and its performance is compared with the most recent developments from the literature. The proposed system achieved 99.6% accuracy, which is higher compared to the existing deep machine learning methods. Therefore, our proposed system is better for multidisciplinary diabetes disease prediction and recommendation. Our proposed system's improved disease diagnosis performance advocates for its employment in the automated diagnostic and recommendation systems for diabetic patients.

Identification of diagnostic cytosine-phosphate-guanine biomarkers in patients with gestational diabetes mellitus via epigenome-wide association study and machine learning

OBJECTIVE

To explore gestational diabetes mellitus (GDM) diagnostic markers and establish the predictive model of GDM.

METHODS

We downloaded the DNA methylation data of GSE70453 and GSE102177 from the Gene Expression Omnibus database. Epigenome-wide association study (EWAS) was performed to analyze the relationship between cytosine-phosphate-guanine (CpG) methylation and GDM. And then the logistic regression models were constructed, with the β-values of CpG sites as predictor variable and the GDM occurrence as binary outcome variable. Data from GSE70453 served as training sets and data from GSE102177 served as verification sets.

RESULTS

The EWAS and overlap analysis identified nine-shared significant CpGs in the two DNA methylation data sets. Remarkably, these nine CpGs were differently methylated in GDM samples compared to their matched normal specimens, among which five fully methylated CpGs were finally selected. Importantly, we established a binary logistic regression model based on the above five CpGs, in which cg11169102, cg21179618 and cg21620107 were critical. Hence, we further built a logistic regression model by using the three CpGs and found that the area under the curve was 0.8209. The validation of the model by using the verification sets indicated the area under the curve was 0.8519.

CONCLUSIONS

We identified potential CpG biomarkers for the diagnosis of gestational diabetes mellitus patients through using EWAS and Logistic regression models in combination.

Predicting brain function status changes in critically ill patients via Machine learning

OBJECTIVE

In intensive care units (ICUs), a patient's brain function status can shift from a state of acute brain dysfunction (ABD) to one that is ABD-free and vice versa, which is challenging to forecast and, in turn, hampers the allocation of hospital resources. We aim to develop a machine learning model to predict next-day brain function status changes.

MATERIALS AND METHODS

Using multicenter prospective adult cohorts involving medical and surgical ICU patients from 2 civilian and 3 Veteran Affairs hospitals, we trained and externally validated a light gradient boosting machine to predict brain function status changes. We compared the performances of the boosting model against state-of-the-art models-an ABD predictive model and its variants. We applied Shapley additive explanations to identify influential factors to develop a compact model.

RESULTS

There were 1026 critically ill patients without evidence of prior major dementia, or structural brain diseases, from whom 12 295 daily transitions (ABD: 5847 days; ABD-free: 6448 days) were observed. The boosting model achieved an area under the receiver-operating characteristic curve (AUROC) of 0.824 (95% confidence interval [CI], 0.821-0.827), compared with the state-of-the-art models of 0.697 (95% CI, 0.693-0.701) with P < .001. Using 13 identified top influential factors, the compact model achieved 99.4% of the boosting model on AUROC. The boosting and the compact models demonstrated high generalizability in external validation by achieving an AUROC of 0.812 (95% CI, 0.812-0.813).

CONCLUSION

The inputs of the compact model are based on several simple questions that clinicians can quickly answer in practice, which demonstrates the model has direct prospective deployment potential into clinical practice, aiding in critical hospital resource allocation.

Associations between Human Chorionic Gonadotropin, Maternal Free Thyroxine, and Gestational Diabetes Mellitus

Background: Human chorionic gonadotropin (hCG) is a marker of placental function, which also stimulates the maternal thyroid gland. Maternal thyroid function can be associated with the pathophysiology of gestational diabetes mellitus (GDM). We aimed to study whether there is an association of hCG concentrations in early pregnancy with GDM and whether it is mediated through maternal thyroid hormones. Methods: This study included 18,683 pregnant women presenting at a tertiary hospital in Shanghai, China, between January 2015 and December 2016. GDM was diagnosed using a 2-hour, 75-g, oral glucose tolerance test (OGTT) according to the American Diabetes Association guidelines. Multivariable logistic or linear regression models were used to identify associations, adjusting for maternal age, education level, family history of diabetes, parity, fetal sex, thyroperoxidase antibody (TPOAb) status, and prepregnancy body-mass index. Results: Higher hCG concentrations were associated with a lower plasma glucose level during the OGTT, but not with fasting plasma glucose or hemoglobin A1c concentrations tested during early pregnancy. hCG in early pregnancy was negatively associated with GDM risk (p = 0.027). Mediation analysis identified that an estimated 21.4% of the association of hCG-associated GDM risk was mediated through changes in free thyroxine (fT4) concentrations (p < 0.05). In the sensitivity analysis restricted to TPOAb-positive women, hCG was not associated with GDM (p = 0.452). Conclusions: Higher hCG levels in early pregnancy are associated with a lower risk of GDM. Maternal fT4 may act as an important mediator in this association.