Several meta-analyses found an association between low maternal serum 25-hydroxyvitamin D (25(OH)D) level and gestational diabetes mellitus (GDM). However, some of them reported significant heterogeneity. We examined the association of serum 25(OH)D concentration measured in the first and in the second halves of pregnancy with the development of GDM in Russian women surveyed in the periods of 2012–2014 and 2018–2021. We conducted a case–control study (including 318 pregnant women) nested on two previous studies. In 2012–2014, a total of 214 women (83 GDM and 131 controls) were enrolled before 15 weeks of gestation and maternal serum 25(OH)D concentrations were measured twice: at 8th–14th week of gestation and simultaneously with two-hour 75 g oral glucose tolerance test (OGTT) at 24th–32nd week of gestation. In the period of 2018–2021, 104 women (56 GDM and 48 controls) were included after OGTT and 25(OH)D concentrations were measured at 24th–32nd week of gestation. Median 25(OH)D levels were 20.0 [15.1–25.7] vs. 20.5 [14.5–27.5] ng/mL (p = 0.565) in GDM and control group in the first half of pregnancy and 25.3 [19.8–33.0] vs. 26.7 [20.8–36.8] ng/mL (p = 0.471) in the second half of pregnancy, respectively. The prevalence rates for vitamin D deficiency (25(OH)D levels < 20 ng/mL) were 49.4% and 45.8% (p = 0.608) in the first half of pregnancy and 26.2% vs. 22.1% (p = 0.516) in the second half of pregnancy in women who developed GDM and in women without GDM, respectively. The frequency of vitamin D supplements intake during pregnancy increased in 2018–2021 compared to 2012–2014 (p = 0.001). However, the third trimester 25(OH)D levels and prevalence of vitamin D deficiency (25.5 vs. 23.1, p = 0.744) did not differ in women examined in the periods of 2012–2014 and 2018–2021. To conclude, there was no association between gestational diabetes risk and maternal 25(OH)D measured both in the first and in the second halves of pregnancy. The increased prevalence of vitamin D supplements intake during pregnancy by 2018–2021 did not lead to higher levels of 25(OH)D.

Background: Gestational diabetes mellitus (GDM) is a common complication of pregnancy associated with serious adverse outcomes for mothers and their offspring. Achieving glycaemic targets is the mainstream in the treatment of GDM in order to improve pregnancy outcomes. As GDM is usually diagnosed in the third trimester of pregnancy, the time frame for the intervention is very narrow. Women need to get new knowledge and change their diet very quickly. Usually, these patients require additional frequent visits to healthcare professionals. Recommender systems based on artificial intelligence could partially substitute healthcare professionals in the process of educating and controlling women with GDM, thus reducing the burden on the women and healthcare systems. We have developed a mobile-based personalized recommendation system DiaCompanion I with data-driven real time personal recommendations focused primarily on postprandial glycaemic response prediction. The study aims to clarify the effect of using DiaCompanion I on glycaemic levels and pregnancy outcomes in women with GDM.Methods: Women with GDM are randomized to 2 treatment groups: utilizing and not utilizing DiaCompanion I. The app provides women in the intervention group the resulting data-driven prognosis of 1-hour postprandial glucose level every time they input their meal data. Based on the predicted glucose level, they can adjust their current meal so that the predicted glucose level falls within the recommended range below 7 mmol/L. The app also provides reminders and recommendations on diet and lifestyle to the participants of the intervention group. All the participants are required to perform 6 blood glucose measurements a day. Capillary glucose values are retrieved from the glucose meter and if not available, from the woman’s diary. Additionally, data on glycaemic levels during the study and consumption of major macro- and micronutrients will be collected using the mobile app with electronic report forms in the intervention group. Women from the control group receive standard care without the mobile app. All participants are prescribed with insulin therapy if needed and modifications in their lifestyle. A total of 216 women will be recruited. The primary outcome is the percentage of postprandial capillary glucose values above target (>7.0 mmol/L). Secondary outcomes include the percentage of patients requiring insulin therapy during pregnancy, maternal and neonatal outcomes, glycaemic control using glycated hemoglobin (HbA1c), continuous glucose monitoring data and other blood glucose metrics, the number of patient visits to endocrinologists and acceptance/satisfaction of the two strategies assessed using a questionnaire.Discussion: We believe that the approach including DiaCompanion I will be more effective in patients with GDM for improving glycaemic levels and pregnancy outcomes. We also expect that the use of the app will help reduce the number of clinic visits.Trial registration number: ClinicalTrials.gov, Identifier NCT05179798.

This study introduces a multi-agent reinforcement learning approach to address the challenges of real-time scheduling in dynamic environments, with a specific focus on healthcare operations. The proposed system integrates the Human-in-the-Loop (HITL) paradigm, providing continuous feedback from human evaluators, and it employs a sophisticated reward function to attenuate the effects of human-driven events. Novel mapping between reinforcement learning (RL) concepts and the Belief–Desire–Intention (BDI) framework is developed to enhance the explainability of the agent’s decision-making. A system is designed to adapt to changes in patient conditions and preferences while minimizing disruptions to existing schedules. Experimental results show a notable decrease in patient waiting times compared to conventional methods while adhering to operator-induced constraints. This approach offers a robust, explainable, and adaptable solution for the challenging tasks of scheduling in the environments that require human-centered decision-making.

This paper presents a neuro-symbolic approach for constructing distributed knowledge graphs to facilitate cooperation through communication among spatially proximate agents. We develop a graph autoencoder (GAE) that learns rich representations from heterogeneous modalities. The method employs density-adaptive k-nearest neighbor (k-NN) construction with Gabriel pruning to build the proximity graphs that balance local density awareness with geometric consistency. When the agents enter the bridging zone, their individual knowledge graphs are aggregated into hypergraphs using a construction algorithm, for which we derive the theoretical bounds on the minimum number of hyperedges required for connectivity under arity and locality constraints. We evaluate the approach in PettingZoo’s communication-oriented environment, observing improvements of approximately 10% in episode rewards and up to 40% in individual agent rewards compared to Deep Q-Network (DQN) baselines, while maintaining comparable policy loss values. The explicit graph structures may offer interpretability benefits for applications requiring auditability. This work explores how structured knowledge representations can support cooperation in distributed multi-agent systems with heterogeneous observations.

The article considers the structures of representation of the graph of inter-agent connections for increasing the efficiency of agent interaction in cooperative competitive games using graph neural networks. A comparative assessment of metrics and adjacency matrices for graphs of connections defined using geometric and semantic metrics of proximity is performed. It is shown that semantic proximity is more effective in constructing a graph of inter-agent connections, and the use of oriented graphs ensures flexible management of information flows. The proposed patterns are important to consider when organizing multi-agent reinforcement learning in a wide range of application areas.