CAManim: Animating end-to-end network activation maps.

Deep neural networks have been widely adopted in numerous domains due to their high performance and accessibility to developers and application-specific end-users. Fundamental to image-based applications is the development of Convolutional Neural Networks (CNNs), which possess the ability to automatically extract features from data. However, comprehending these complex models and their learned representations, which typically comprise millions of parameters and numerous layers, remains a challenge for both developers and end-users. This challenge arises due to the absence of interpretable and transparent tools to make sense of black-box models. There exists a growing body of Explainable Artificial Intelligence (XAI) literature, including a collection of methods denoted Class Activation Maps (CAMs), that seek to demystify what representations the model learns from the data, how it informs a given prediction, and why it, at times, performs poorly in certain tasks. We propose a novel XAI visualization method denoted CAManim that seeks to simultaneously broaden and focus end-user understanding of CNN predictions by animating the CAM-based network activation maps through all layers, effectively depicting from end-to-end how a model progressively arrives at the final layer activation. Herein, we demonstrate that CAManim works with any CAM-based method and various CNN architectures. Beyond qualitative model assessments, we additionally propose a novel quantitative assessment that expands upon the Remove and Debias (ROAD) metric, pairing the qualitative end-to-end network visual explanations assessment with our novel quantitative "yellow brick ROAD" assessment (ybROAD). This builds upon prior research to address the increasing demand for interpretable, robust, and transparent model assessment methodology, ultimately improving an end-user's trust in a given model's predictions. Examples and source code can be found at: https://omni-ml.github.io/pytorch-grad-cam-anim/.

Temporal Trends in Gestational Diabetes Mellitus and Associated Risk Factors in Ontario, Canada, 2012-2020: A Population-Based Cohort Study.

OBJECTIVES: The prevalence of gestational diabetes mellitus (GDM) has been increasing globally over recent decades; however, underlying reasons for the increase remain unclear. We analyzed trends in GDM rates and evaluated risk factors associated with the observed trends in Ontario, Canada. METHODS: We conducted a retrospective population-based cohort study using the Better Outcomes Registry and Network Ontario, linked with the Canadian Institute for Health Information Discharge Abstract Database. All pregnant individuals who had a singleton hospital delivery from 1 April 2012 to 31 March 2020 were included. We calculated rates and 95% CIs for GDM by year of delivery and contrasted fiscal year 2019/20 with 2012/13. Temporal trends in GDM were quantified using crude and adjusted risk ratios by modified Poisson regression. We further quantified the temporal increase attributable to changes in maternal characteristics by decomposition analysis. RESULTS: Among 1 044 258 pregnant individuals, 82 896 (7.9%) were diagnosed with GDM over the 8 years. GDM rate rose from 6.1 to 10.4 per 100 deliveries between fiscal years 2012/13 and 2019/20. The risk of GDM in 2019/20 was 1.53 times (95% CI 1.50-1.56) higher compared with 2012/13. 27% of the increase in GDM was due to changes in maternal age, 8 BMI, and Asian ethnicity. CONCLUSIONS: The GDM rate has been consistently increasing in Ontario, Canada. The contribution of increasing maternal age, pre-pregnancy obesity, and Asian ethnicity to the recent increase in GDM is notable. Further investigation is required to better understand the contributors to increasing GDM.

Outlier analysis for accelerating clinical discovery: An augmented intelligence framework and a systematic review.

Clinical discoveries largely depend on dedicated clinicians and scientists to identify and pursue unique and unusual clinical encounters with patients and communicate these through case reports and case series. This process has remained essentially unchanged throughout the history of modern medicine. However, these traditional methods are inefficient, especially considering the modern-day availability of health-related data and the sophistication of computer processing. Outlier analysis has been used in various fields to uncover unique observations, including fraud detection in finance and quality control in manufacturing. We propose that clinical discovery can be formulated as an outlier problem within an augmented intelligence framework to be implemented on any health-related data. Such an augmented intelligence approach would accelerate the identification and pursuit of clinical discoveries, advancing our medical knowledge and uncovering new therapies and management approaches. We define clinical discoveries as contextual outliers measured through an information-based approach and with a novelty-based root cause. Our augmented intelligence framework has five steps: define a patient population with a desired clinical outcome, build a predictive model, identify outliers through appropriate measures, investigate outliers through domain content experts, and generate scientific hypotheses. Recognizing that the field of obstetrics can particularly benefit from this approach, as it is traditionally neglected in commercial research, we conducted a systematic review to explore how outlier analysis is implemented in obstetric research. We identified two obstetrics-related studies that assessed outliers at an aggregate level for purposes outside of clinical discovery. Our findings indicate that using outlier analysis in clinical research in obstetrics and clinical research, in general, requires further development.

Changes associated with the COVID-19 pandemic on postpartum screening results in Ontario, Canada: The healthy babies healthy children screening tool.

BACKGROUND: Research on the impact of the COVID-19 pandemic on mothers/childbearing parents has mainly been cross-sectional and focused on psychological symptoms. This study examined the impact on function using ongoing, systematic screening of a representative Ontario sample. METHODS: An interrupted time series analysis of repeated cross-sectional data from a province-wide screening program using the Healthy Babies Healthy Children (HBHC) tool assessed changes associated with the pandemic at the time of postpartum discharge from hospital. Postal codes were used to link to neighborhood-level data. The ability to parent or care for the baby/child and other psychosocial and behavioral outcomes were assessed. RESULTS: The co-primary outcomes of inability to parent or care for the baby/child were infrequently observed in the pre-pandemic (March 9, 2019-March 15, 2020) and initial pandemic periods (March 16, 2020-March 23, 2021) (parent 209/63,006 (0.33%)-177/56,117 (0.32%), care 537/62,955 (0.85%)-324/56,086 (0.58%)). Changes after pandemic onset were not observed for either outcome although a significant (p = 0.02) increase in slope was observed for inability to parent (with questionable clinical significance). For secondary outcomes, worsening was only seen for reported complications during labor/delivery. Significant improvements were observed in the likelihood of being unable to identify a support person to assist with care, need of newcomer support, and concerns about money over time. CONCLUSIONS: There were no substantive changes in concerns about ability to parent or care for children. Adverse impacts of the pandemic may have been mitigated by accommodations for remote work and social safety net policies.

Deep learning prediction of renal anomalies for prenatal ultrasound diagnosis.

Deep learning algorithms have demonstrated remarkable potential in clinical diagnostics, particularly in the field of medical imaging. In this study, we investigated the application of deep learning models in early detection of fetal kidney anomalies. To provide an enhanced interpretation of those models' predictions, we proposed an adapted two-class representation and developed a multi-class model interpretation approach for problems with more than two labels and variable hierarchical grouping of labels. Additionally, we employed the explainable AI (XAI) visualization tools Grad-CAM and HiResCAM, to gain insights into model predictions and identify reasons for misclassifications. The study dataset consisted of 969 ultrasound images from unique patients; 646 control images and 323 cases of kidney anomalies, including 259 cases of unilateral urinary tract dilation and 64 cases of unilateral multicystic dysplastic kidney. The best performing model achieved a cross-validated area under the ROC curve of 91.28% ± 0.52%, with an overall accuracy of 84.03% ± 0.76%, sensitivity of 77.39% ± 1.99%, and specificity of 87.35% ± 1.28%. Our findings emphasize the potential of deep learning models in predicting kidney anomalies from limited prenatal ultrasound imagery. The proposed adaptations in model representation and interpretation represent a novel solution to multi-class prediction problems.

Infant Infections Following Cesarean Delivery on Maternal Request: A Population-Based Cohort Study.

OBJECTIVES: Investigations about cesarean delivery (CD) on maternal request (CDMR) and infant infection risk frequently rely on administrative data with poorly defined indications for CD. We sought to determine the association between CDMR and infant infection using an intent-to-treat approach. METHODS: This was a population-based cohort study of low-risk singleton pregnancies with a term live birth in Ontario, Canada between April 2012 and March 2018. Subjects with prior CD were excluded. Outcomes included upper and lower respiratory tract infections, gastrointestinal infections, otitis media, and a composite of these 4. Relative risk and 95% CI were calculated for component and composite outcomes up to 1 year following planned CDMR versus planned vaginal deliveries (VDs). Subgroup and sensitivity analyses included age at infection (≤28 vs. >28 days), type of care (ambulatory vs. hospitalisation), restricting the cohort to nulliparous pregnancies, and including individuals with previous CD. Last, we re-examined outcome risk on an as-treated basis (actual CD vs. actual VD). RESULTS: Of 422 134 pregnancies, 0.4% (1827) resulted in a planned CDMR. After adjusting for covariates, planned CDMR was not associated with a risk of composite infant infections (adjusted relative risk 1.02; 95% CI 0.92-1.11). Findings for component infection outcomes, subgroup, and sensitivity analyses were similar. However, the as-treated analysis of the role of delivery mode on infant risk for infection demonstrated that actual CD (planned and unplanned) was associated with an increased risk for infant infections compared to actual VD. CONCLUSIONS: Planned CDMR is not associated with increased risk for neonatal or infant infections compared with planned VD. Study design must be carefully considered when investigating the impact of CDMR on infant infection outcomes.

Associations of neighborhood greenspace, and active living environments with autism spectrum disorders: A matched case-control study in Ontario, Canada.

BACKGROUND: Increasing evidence links early life residential exposure to natural urban environmental attributes and positive health outcomes in children. However, few studies have focused on their protective effects on the risk of autism spectrum disorder (ASD). The aim of this study was to investigate the associations of neighborhood greenspace, and active living environments during pregnancy with ASD in young children (≤6 years). METHODS: We conducted a population-based matched case-control study of singleton term births in Ontario, Canada for 2012-2016. The ASD and environmental data was generated using the Ontario Autism Spectrum Profile, the Better Outcomes Registry & Network Ontario, and Canadian Urban Environmental Health Research Consortium. We employed conditional logistic regressions to estimate the odds ratio (OR) between ASD and environmental factors characterizing selected greenspace metrics and neighborhoods conducive to active living (i.e., green view index (GVI), normalized difference vegetation index (NDVI), tree canopy, park proximity and active living environments index (ALE)). RESULTS: We linked 8643 mother-child pairs, including 1554 cases (18%). NDVI (OR 1.034, 0.944-1.024, per Inter Quartile Range [IQR] = 0.08), GVI (OR 1.025, 95% CI 0.953-1.087, per IQR = 9.45%), tree canopy (OR 0.992, 95% CI 0.903-1.089, per IQR = 6.24%) and the different categories of ALE were not associated with ASD in adjusted models for air pollution. In contrast, living closer to a park was protective (OR 0.888, 0.833-0.948, per 0.06 increase in park proximity index), when adjusted for air pollution. CONCLUSIONS: This study reported mixed findings showing both null and beneficial effects of green spaces and active living environments on ASD. Further investigations are warranted to elucidate the role of exposure to greenspaces and active living environments on the development of ASD.

The impact of isolated obesity compared with obesity and other risk factors on risk of stillbirth: a retrospective cohort study.

BACKGROUND: Maternal obesity is associated with stillbirth, but uncertainty persists around the effects of higher obesity classes. We sought to compare the risk of stillbirth associated with maternal obesity alone versus maternal obesity and additional or undiagnosed factors contributing to high-risk pregnancy. METHODS: We conducted a retrospective cohort study using the Better Outcomes Registry and Network (BORN) for singleton hospital births in Ontario between 2012 and 2018. We used multivariable Cox proportional hazard regression and logistic regression to evaluate the relationship between prepregnancy maternal body mass index (BMI) class and stillbirth (reference was normal BMI). We treated maternal characteristics and obstetrical complications as independent covariates. We performed mediator analyses to measure the direct and indirect effects of BMI on stillbirth through major common-pathway complications. We used fully adjusted and partially adjusted models, representing the impact of maternal obesity alone and maternal obesity with other risk factors on stillbirth, respectively. RESULTS: We analyzed data on 681 178 births between 2012 and 2018, of which 1956 were stillbirths. Class I obesity was associated with an increased incidence of stillbirth (adjusted hazard ratio [HR] 1.55, 95% confidence interval [CI] 1.35-1.78). This association was stronger for class III obesity (adjusted HR 1.80, 95% CI 1.44-2.24), and strongest for class II obesity (adjusted HR 2.17, 95% CI 1.83-2.57). Plotting point estimates for odds ratios, stratified by gestational age, showed a marked increase in the relative odds for stillbirth beyond 37 weeks' gestation for those with obesity with and without other risk factors, compared with those with normal BMI. The impact of potential mediators was minimal. INTERPRETATION: Maternal obesity alone and obesity with other risk factors are associated with an increased risk of stillbirth. This risk increases with gestational age, especially at term.

Ultrasonographic Fetal Nuchal Translucency Measurements and Cytogenetic Outcomes.

Importance: Ultrasonographic measurement of fetal nuchal translucency is used in prenatal screening for trisomies 21 and 18 and other conditions. A cutoff of 3.5 mm or greater is commonly used to offer follow-up investigations, such as prenatal cell-free DNA (cfDNA) screening or cytogenetic testing. Recent studies showed a possible association with chromosomal anomalies for levels less than 3.5 mm, but extant evidence has limitations. Objective: To evaluate the association between different nuchal translucency measurements and cytogenetic outcomes on a population level. Design, Setting, and Participants: This population-based retrospective cohort study used data from the Better Outcomes Registry & Network, the perinatal registry for Ontario, Canada. All singleton pregnancies with an estimated date of delivery from September 1, 2016, to March 31, 2021, were included. Data were analyzed from March 17 to August 14, 2023. Exposures: Nuchal translucency measurements were identified through multiple-marker screening results. Main Outcomes and Measures: Chromosomal anomalies were identified through all Ontario laboratory-generated prenatal and postnatal cytogenetic tests. Cytogenetic testing results, supplemented with information from cfDNA screening and clinical examination at birth, were used to identify pregnancies without chromosomal anomalies. Multivariable modified Poisson regression with robust variance estimation and adjustment for gestational age was used to compare cytogenetic outcomes for pregnancies with varying nuchal translucency measurement categories and a reference group with nuchal translucency less than 2.0 mm. Results: Of 414 268 pregnancies included in the study (mean [SD] maternal age at estimated delivery date, 31.5 [4.7] years), 359 807 (86.9%) had a nuchal translucency less than 2.0 mm; the prevalence of chromosomal anomalies in this group was 0.5%. An increased risk of chromosomal anomalies was associated with increasing nuchal translucency measurements, with an adjusted risk ratio (ARR) of 20.33 (95% CI, 17.58-23.52) and adjusted risk difference (ARD) of 9.94% (95% CI, 8.49%-11.39%) for pregnancies with measurements of 3.0 to less than 3.5 mm. The ARR was 4.97 (95% CI, 3.45-7.17) and the ARD was 1.40% (95% CI, 0.77%-2.04%) when restricted to chromosomal anomalies beyond the commonly screened aneuploidies (excluding trisomies 21, 18, and 13 and sex chromosome aneuploidies). Conclusions and Relevance: In this cohort study of 414 268 singleton pregnancies, those with nuchal translucency measurements less than 2.0 mm were at the lowest risk of chromosomal anomalies. Risk increased with increasing measurements, including measurements less than 3.5 mm and anomalies not routinely screened by many prenatal genetic screening programs.

The Transformative Potential of AI in Obstetrics and Gynaecology.

The transformative power of artificial intelligence (AI) is reshaping diverse domains of medicine. Recent progress, catalyzed by computing advancements, has seen commensurate adoption of AI technologies within obstetrics and gynaecology. We explore the use and potential of AI in three focus areas: predictive modelling for pregnancy complications, Deep learning-based image interpretation for precise diagnoses, and large language models enabling intelligent health care assistants. We also provide recommendations for the ethical implementation, governance of AI, and promote research into AI explainability, which are crucial for responsible AI integration and deployment. AI promises a revolutionary era of personalized health care in obstetrics and gynaecology.

Gestational weight loss and fetal growth in uncomplicated pregnancies among women with obesity: a population-based retrospective cohort study.

OBJECTIVE: The impact of gestational weight loss (GWL) on fetal growth among women with obesity remains unclear. This study aimed to examine the association between weight loss during pregnancy among women with body mass index (BMI) ≥ 30 kg/m2 and the risk of small-for-gestational-age (SGA) and large-for-gestational-age (LGA) neonates. METHODS: We conducted a retrospective, population-based cohort study of women with pre-pregnancy obesity that resulted in a singleton live birth in 2012-2017, using birth registry data in Ontario, Canada. Women with pregnancy complications or health conditions which could cause weight loss were excluded. GWL is defined as negative gestational weight change (≤0 kg). The association between GWL and fetal growth was estimated using generalized estimating equation models and restricted cubic spline regression analysis. Stratified analysis was conducted by obesity class (I:30-34.9 kg/m2, II:35-39.9 kg/m2, and III + : ≥40 kg/m2). RESULTS: Of the 52,153 eligible women who entered pregnancy with a BMI ≥ 30 kg/m2, 5.3% had GWL. Compared to adequate gestational weight gain, GWL was associated with an increased risk of SGA neonates (aRR:1.45, 95% CI: 1.30-1.60) and a decreased risk of LGA neonates (aRR: 0.81, 95% CI:0.73-0.93). Non-linear L-shaped associations were observed between gestational weight change and SGA neonates, with an increased risk of SGA observed with increased GWL. On the contrary, non-linear S-shaped associations were observed between gestational weight change and LGA neonates, with a decreased risk of LGA observed with increased GWL. Similar findings were observed from the stratified analysis by obesity class. CONCLUSION: These findings highlight that GWL in women with obesity may increase the risk of SGA neonates but reduce the risk of LGA neonates. Recommendations of GWL for women with obesity should be interpreted with caution.

AlphaFold Accurately Predicts the Structure of Ribosomally Synthesized and Post-Translationally Modified Peptide Biosynthetic Enzymes.

Ribosomally synthesized and post-translationally modified peptides (RiPPs) are a growing class of natural products biosynthesized from a genetically encoded precursor peptide. The enzymes that install the post-translational modifications on these peptides have the potential to be useful catalysts in the production of natural-product-like compounds and can install non-proteogenic amino acids in peptides and proteins. However, engineering these enzymes has been somewhat limited, due in part to limited structural information on enzymes in the same families that nonetheless exhibit different substrate selectivities. Despite AlphaFold2's superior performance in single-chain protein structure prediction, its multimer version lacks accuracy and requires high-end GPUs, which are not typically available to most research groups. Additionally, the default parameters of AlphaFold2 may not be optimal for predicting complex structures like RiPP biosynthetic enzymes, due to their dynamic binding and substrate-modifying mechanisms. This study assessed the efficacy of the structure prediction program ColabFold (a variant of AlphaFold2) in modeling RiPP biosynthetic enzymes in both monomeric and dimeric forms. After extensive benchmarking, it was found that there were no statistically significant differences in the accuracy of the predicted structures, regardless of the various possible prediction parameters that were examined, and that with the default parameters, ColabFold was able to produce accurate models. We then generated additional structural predictions for select RiPP biosynthetic enzymes from multiple protein families and biosynthetic pathways. Our findings can serve as a reference for future enzyme engineering complemented by AlphaFold-related tools.

Pregnancy, fetal, and neonatal outcomes after a first booster dose of covid-19 vaccine during pregnancy in Ontario, Canada: population based, retrospective cohort study.

Objective: To assess risk of adverse pregnancy, fetal, and neonatal outcomes after a third dose (first booster dose) of covid-19 vaccine during pregnancy among individuals who had completed both doses of primary covid-19 vaccine series before pregnancy. Design: Population based, retrospective cohort study. Setting: Ontario, Canada, from 20 December 2021 to 31 August 2022. Participants: Individuals were included if they were pregnant with an expected date of delivery from 20 December 2021 (start date of third dose eligibility for everyone ≥18 years) to 31 August 2022, who had completed the two doses of primary covid-19 messenger RNA vaccine series before pregnancy, and became eligible for a third dose (≥six months since dose two) before the end of pregnancy. Main outcome measures: Pregnancy outcomes included hypertensive disorders of pregnancy, placental abruption, caesarean delivery, chorioamnionitis, and postpartum hemorrhage. Fetal and neonatal outcomes included stillbirth, preterm birth, admission to neonatal intensive care unit for >24 h, newborn 5 min Apgar score <7, and small-for-gestational age infant (<10th percentile). We estimated hazard ratios and 95% confidence intervals for study outcomes, treating dose three as a time varying exposure and adjusting for confounding using inverse probability weighting. Results: Among 32 689 births, 18 491 (56.6%) were born to individuals who received a third covid-19 dose during pregnancy. Compared with eligible individuals who did not receive a third dose during pregnancy, no increased risks were associated with receiving a third covid-19 vaccine dose during pregnancy for placental abruption (adjusted hazard ratio 0.84 (95% confidence interval 0.70 to 1.02)), chorioamnionitis (0.67 (0.49 to 0.90)), postpartum haemorrhage (1.01 (0.89 to 1.16)), caesarean delivery (0.90 (0.87 to 0.94)), stillbirth (0.56 (0.39 to 0.81)), preterm birth (0.91 (0.84 to 0.99)), neonatal intensive care unit admission (0.96 (0.90 to 1.03)), 5 min Apgar score<7 (0.96 (0.82 to 1.14)), or small-for-gestational age infant (0.86 (0.79 to 0.93)). Conclusion: Receipt of a third covid-19 vaccine dose during pregnancy was not associated with an increased risk of adverse pregnancy, fetal, or neonatal outcomes. These findings can help to inform evidence based decision making about the risks and benefits of covid-19 booster doses during pregnancy.

Comparison of adverse maternal and perinatal outcomes between induction and expectant management among women with gestational diabetes mellitus at term pregnancy: a systematic review and meta-analysis.

BACKGROUND: Induction at 38-40 weeks of gestation has been broadly suggested for women with gestational diabetes mellitus (GDM), yet its benefits and risks remain unclear. This study aimed to systematically review and meta-analyze existing evidence on the effect of induction at term gestation among women with GDM. METHODS: We searched MEDLINE, EMBASE, Cochrane Libraries, and Web of Science from inception to June 2021. We included randomized controlled trials (RCTs) and observational studies comparing induction with expectant management among GDM term pregnancies. Primary outcomes included caesarean section (CS) and macrosomia. All screening and extraction were conducted independently and in duplicates. Meta-analyses with random-effects models were conducted to generate the pooled odds ratios (ORs) and 95% confidence intervals (CIs) using the Mantel-Haenszel method. Methodological quality was assessed independently by two reviewers using the Cochrane Risk of Bias Tool for RCTs and the Newcastle-Ottawa Scale for observational studies. RESULTS: Of the 4,791 citations, 11 studies were included (3 RCTs and 8 observational studies). Compared to expectant management, GDM women with induction had a significantly lower odds for macrosomia (RCTs 0.49 [0.30-0.81]); observational studies 0.64 [0.54-0.77]), but not for CS (RCTs 0.95 [0.64-1.43]); observational studies 1.03 [0.79-1.34]). Induction was associated with a lower odds of severe perineal lacerations in observational studies (0.59 [0.39-0.88]). No significant difference was observed for other maternal or neonatal morbidities, or perinatal mortality between groups. CONCLUSIONS: For GDM women, induction may reduce the risk of macrosomia and severe perineal lacerations compared to expectant management. Further rigorous studies with large sample sizes are warranted to better inform clinical implications.

Temporal trends and determinants of COVID-19 vaccine series initiation after recent pregnancy.

During the rapid deployment of COVID-19 vaccines in 2021, safety concerns may have led some pregnant individuals to postpone vaccination until after giving birth. This study aimed to describe temporal patterns and factors associated with COVID-19 vaccine series initiation after recent pregnancy in Ontario, Canada. Using the provincial birth registry linked with the COVID-19 vaccine database, we identified all individuals who gave birth between January 1 and December 31, 2021, and had not yet been vaccinated by the end of pregnancy, and followed them to June 30, 2022 (follow-up ranged from 6 to 18 months). We used cumulative incidence curves to describe COVID-19 vaccine initiation after pregnancy and assessed associations with sociodemographic, pregnancy-related, and health behavioral factors using Cox proportional hazards regression to estimate adjusted hazard ratios (aHR) and 95% confidence intervals (CI). Among 137,198 individuals who gave birth in 2021, 87,376 (63.7%) remained unvaccinated at the end of pregnancy; of these, 65.0% initiated COVID-19 vaccination by June 30, 2022. Lower maternal age (<25 vs. 30-34 y aHR: 0.73, 95%CI: 0.70-0.77), smoking during pregnancy (vs. nonsmoking aHR: 0.68, 95%CI: 0.65-0.72), lower neighborhood income (lowest quintile vs. highest aHR: 0.79, 95%CI: 0.76-0.83), higher material deprivation (highest quintile vs. lowest aHR: 0.74, 95%CI: 0.70-0.79), and exclusive breastfeeding (vs. other feeding aHR: 0.81, 95%CI: 0.79-0.84) were associated with lower likelihood of vaccine initiation. Among unvaccinated individuals who gave birth in 2021, COVID-19 vaccine initiation after pregnancy reached 65% by June 30, 2022, suggesting persistent issues with vaccine hesitancy and/or access to vaccination in this population.

Augmented Intelligence for Clinical Discovery in Hypertensive Disorders of Pregnancy Using Outlier Analysis.

Objectives Clinical discoveries are heralded by observing unique and unusual clinical cases. The effort of identifying such cases rests on the shoulders of busy clinicians. We assess the feasibility and applicability of an augmented intelligence framework to accelerate the rate of clinical discovery in preeclampsia and hypertensive disorders of pregnancy-an area that has seen little change in its clinical management. Methods We conducted a retrospective exploratory outlier analysis of participants enrolled in the folic acid clinical trial (FACT, N=2,301) and the Ottawa and Kingston birth cohort (OaK, N=8,085). We applied two outlier analysis methods: extreme misclassification contextual outlier and isolation forest point outlier. The extreme misclassification contextual outlier is based on a random forest predictive model for the outcome of preeclampsia in FACT and hypertensive disorder of pregnancy in OaK. We defined outliers in the extreme misclassification approach as mislabelled observations with a confidence level of more than 90%. Within the isolation forest approach, we defined outliers as observations with an average path length z score less or equal to -3, or more or equal to 3. Content experts reviewed the identified outliers and determined if they represented a potential novelty that could conceivably lead to a clinical discovery. Results In the FACT study, we identified 19 outliers using the isolation forest algorithm and 13 outliers using the random forest extreme misclassification approach. We determined that three (15.8%) and 10 (76.9%) were potential novelties, respectively. Out of 8,085 participants in the OaK study, we identified 172 outliers using the isolation forest algorithm and 98 outliers using the random forest extreme misclassification approach; four (2.3%) and 32 (32.7%), respectively, were potential novelties. Overall, the outlier analysis part of the augmented intelligence framework identified a total of 302 outliers. These were subsequently reviewed by content experts, representing the human part of the augmented intelligence framework. The clinical review determined that 49 of the 302 outliers represented potential novelties.  Conclusions Augmented intelligence using extreme misclassification outlier analysis is a feasible and applicable approach for accelerating the rate of clinical discoveries. The use of an extreme misclassification contextual outlier analysis approach has resulted in a higher proportion of potential novelties than using the more traditional point outlier isolation forest approach. This finding was consistent in both the clinical trial and real-world cohort study data. Using augmented intelligence through outlier analysis has the potential to speed up the process of identifying potential clinical discoveries. This approach can be replicated across clinical disciplines and could exist within electronic medical records systems to automatically identify outliers within clinical notes to clinical experts.

Persisting risk factors for preeclampsia among high-risk pregnancies already using prophylactic aspirin: a multi-country retrospective investigation.

BACKGROUND: Low-dose aspirin is recommended for pregnant individuals at high-risk of developing preeclampsia, but less is known about those that develop preeclampsia even while using prophylactic aspirin for preeclampsia prevention as the best course of treatment. OBJECTIVES: The objective of this study is to investigate the risk factors with the highest risk of developing preeclampsia among pregnant individuals already using aspirin from high-risk obstetrical centers across five countries. DESIGN: This is a secondary analysis of pregnant individuals from the Folic Acid Clinical Trial (FACT) who were using prophylactic aspirin before 16 weeks gestation. The FACT randomized control trial took place in 70 high risk obstetrical centers in Canada, United Kingdom, Australia, Jamaica, and Argentina between 2011-2015. Participants were included if they had any of the risk factors for preeclampsia: diabetes, chronic hypertension, twin pregnancy, history of preeclampsia, and/or obesity (Body Mass Index ≥35). The outcomes of interest were preeclampsia and preterm preeclampsia (<37 weeks). Log binomial regressions assessed factors significantly associated with any preeclampsia or preterm-preeclampsia (<37 weeks) using adjusted risk ratios (ARR) and 95% confidence intervals (CI). RESULTS: There were 2296 pregnant individuals with complete information on aspirin included in this study. At baseline, all patients were at high risk of preeclampsia and were eligible for aspirin prophylaxis, however, only 660 (28.7%) were taking aspirin. Among the 660 pregnant individuals taking aspirin, 132 (20%) developed preeclampsia and 60 (9.09%) preterm preeclampsia. Among pregnant individuals using aspirin, the risks of preeclampsia were highest for twins (ARR:2.62, 95% CI: 1.68-4.11), history of preeclampsia (ARR: 2.42, 95% CI: 1.74-3.38), and hypertension (ARR:1.92, 95% CI: 1.37-2.69). Similar trends were found for preterm-preeclampsia for twins (ARR:4.10, 95% CI:2.15-7.82), history of preeclampsia (ARR:2.75, 95% CI:1.62-4.67), and hypertension (ARR:2.18, 95% CI:1.28-3.72). No significant differences were found for obesity or diabetes. CONCLUSION: These findings suggest that individuals with twin pregnancies, a history of preeclampsia, or hypertension may not benefit from aspirin to the same extent as those with other complications such as obesity or diabetes. Careful clinical monitoring for these risks factors is recommended and future research into the effectiveness in these populations would increase our understanding of the current best practice of prophylactic aspirin use to prevent preeclampsia. TRIAL REGISTRATION: Current Controlled Trials ISRCTN23781770 and ClinicalTrials.gov NCT01355159.

Do Case Reports and Case Series Generate Clinical Discoveries About Preeclampsia? A Systematic Review.

Background: Preeclampsia is a leading cause of maternal and perinatal mortality and morbidity. The management of preeclampsia has not changed much in more than two decades, and its aetiology is still not fully understood. Case reports and case series have traditionally been used to communicate new knowledge about existing conditions. Whether this is true for preeclampsia is not known. Objective: To determine whether recent case reports or case series have generated new knowledge and clinical discoveries about preeclampsia. Methods: A detailed search strategy was developed in consultation with a medical librarian. Two bibliographic databases were searched through Ovid: Embase and MEDLINE. We selected case reports or case series published between 2015 and 2020, comprising pregnant persons diagnosed with hypertensive disorders of pregnancy, including preeclampsia. Two reviewers independently screened all publications. One reviewer extracted data from included studies, while another conducted a quality check of extracted data. We developed a codebook to guide our data extraction and outcomes assessment. The quality of each report was determined based on Joanna Briggs Institute (JBI) critical appraisal checklist for case reports and case series. Results: We included 104 case reports and three case series, together comprising 118 pregnancies. A severe presentation or complication of preeclampsia was reported in 81% of pregnancies, and 84% had a positive maternal outcome, free of death or persistent complications. Only 8% of the case reports were deemed to be of high quality, and 53.8% of moderate quality; none of the case series were of high quality. A total of 26 of the 107 publications (24.3%) included a novel clinical discovery as a central theme. Conclusion: Over two-thirds of recent case reports and case series about preeclampsia do not appear to present new knowledge or discoveries about preeclampsia, and most are of low quality.

Temporal trends and determinants of COVID-19 vaccine coverage and series initiation during pregnancy in Ontario, Canada, December 2020 to December 2021: A population-based retrospective cohort study.

BACKGROUND: Population-based COVID-19 vaccine coverage estimates among pregnant individuals are limited. We assessed temporal patterns in vaccine coverage (≥1 dose before or during pregnancy) and evaluated factors associated with vaccine series initiation (receiving dose 1 during pregnancy) in Ontario, Canada. METHODS: We linked the provincial birth registry with COVID-19 vaccination records from December 14, 2020 to December 31, 2021 and assessed coverage rates among all pregnant individuals by month, age, and neighborhood sociodemographic characteristics. Among individuals who gave birth since April 2021-when pregnant people were prioritized for vaccination-we assessed associations between sociodemographic, behavioral, and pregnancy-related factors with vaccine series initiation using multivariable regression to estimate adjusted risk ratios (aRR) and risk differences (aRD) with 95% confidence intervals (CI). RESULTS: Among 221,190 pregnant individuals, vaccine coverage increased to 71.2% by December 2021. Gaps in coverage across categories of age and sociodemographic characteristics decreased over time, but did not disappear. Lower vaccine series initiation was associated with lower age (<25 vs. 30-34 years: aRR 0.53, 95%CI 0.51-0.56), smoking (vs. non-smoking: 0.64, 0.61-0.67), no first trimester prenatal care visit (vs. visit: 0.80, 0.77-0.84), and residing in neighborhoods with the lowest income (vs. highest: 0.69, 0.67-0.71). Vaccine series initiation was marginally higher among individuals with pre-existing medical conditions (vs. no conditions: 1.07, 1.04-1.10). CONCLUSIONS: COVID-19 vaccine coverage among pregnant individuals remained lower than in the general population, and there was lower vaccine initiation by multiple characteristics.

Cytogenetic outcomes following a failed cell-free DNA screen: a population-based retrospective cohort study of 35,146 singleton pregnancies.

BACKGROUND: Cell-free fetal DNA screening is routinely offered to pregnant individuals to screen for aneuploidies. Although cell-free DNA screening is consistently more accurate than multiple-marker screening, it sometimes fails to yield a result. These test failures and their clinical implications are poorly described in the literature. Some studies suggest that a failed cell-free DNA screening result is associated with increased likelihood of cytogenetic abnormalities. OBJECTIVE: This study aimed to assess the association between a failed cell-free DNA test and common aneuploidies. The objectives were to determine: (1) the proportion of test failures on first and subsequent attempts, and (2) whether a failed cell-free DNA screen on first attempt is associated with increased likelihood of common aneuploidies (trisomies 21, 18, and 13, and sex chromosome aneuploidies). STUDY DESIGN: This was a population-based retrospective cohort study using data from Ontario's prescribed maternal and child registry, Better Outcomes Registry and Network Ontario. The study included all singleton pregnancies in Ontario with an estimated date of delivery from September 1, 2016 to March 31, 2019 that had a cell-free DNA screening record in the registry. Specific outcomes (trisomies 21, 18, and 13, and sex chromosome aneuploidies) of pregnancies with a failed cell-free DNA screen on first attempt were compared with those of pregnancies with low-risk cell-free DNA-screening results using modified Poisson regression adjusted for funding status (publicly funded vs self-paid), gestational age at screening, method of conception, and maternal age for autosomal aneuploidies. RESULTS: Our cohort included 35,146 pregnancies that had cell-free DNA screening during the study period. The overall cell-free DNA screening failure rate was 4.8% on first attempt and 2.2% after multiple attempts. An abnormal cytogenetic result for trisomies 21, 18, and 13, or sex chromosome aneuploidies was identified in 19.4% of pregnancies with a failed cell-free DNA screening for which cytogenetic testing was performed. Pregnancies with a failed cell-free DNA screen on first attempt had a relative risk of 130.3 (95% confidence interval, 64.7-262.6) for trisomy 21, trisomy 18, or trisomy 13, and a risk difference of 5.4% (95% confidence interval, 2.6-8.3), compared with pregnancies with a low-risk result. The risk of sex chromosome aneuploidies was not significantly greater in pregnancies with a failed result compared with pregnancies with a low-risk result (relative risk, 2.7; 95% confidence interval, 0.9-7.9; relative difference, 1.2%; 95% confidence interval, -0.9 to 3.2). CONCLUSION: Cell-free DNA screening test failures are relatively common. Although repeated testing improves the likelihood of an informative result, pregnancies with a failed cell-free DNA screen upon first attempt remain at increased risk for common autosomal aneuploidies, but not sex chromosome aneuploidies.