A novel electronic health record-based, machine-learning model to predict severe hypoglycemia leading to hospitalizations in older adults with diabetes: A territory-wide cohort and modeling study.

BACKGROUND: Older adults with diabetes are at high risk of severe hypoglycemia (SH). Many machine-learning (ML) models predict short-term hypoglycemia are not specific for older adults and show poor precision-recall. We aimed to develop a multidimensional, electronic health record (EHR)-based ML model to predict one-year risk of SH requiring hospitalization in older adults with diabetes. METHODS AND FINDINGS: We adopted a case-control design for a retrospective territory-wide cohort of 1,456,618 records from 364,863 unique older adults (age ≥65 years) with diabetes and at least 1 Hong Kong Hospital Authority attendance from 2013 to 2018. We used 258 predictors including demographics, admissions, diagnoses, medications, and routine laboratory tests in a one-year period to predict SH events requiring hospitalization in the following 12 months. The cohort was randomly split into training, testing, and internal validation sets in a 7:2:1 ratio. Six ML algorithms were evaluated including logistic-regression, random forest, gradient boost machine, deep neural network (DNN), XGBoost, and Rulefit. We tested our model in a temporal validation cohort in the Hong Kong Diabetes Register with predictors defined in 2018 and outcome events defined in 2019. Predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC) statistics, and positive predictive value (PPV). We identified 11,128 SH events requiring hospitalization during the observation periods. The XGBoost model yielded the best performance (AUROC = 0.978 [95% CI 0.972 to 0.984]; AUPRC = 0.670 [95% CI 0.652 to 0.688]; PPV = 0.721 [95% CI 0.703 to 0.739]). This was superior to an 11-variable conventional logistic-regression model comprised of age, sex, history of SH, hypertension, blood glucose, kidney function measurements, and use of oral glucose-lowering drugs (GLDs) (AUROC = 0.906; AUPRC = 0.085; PPV = 0.468). Top impactful predictors included non-use of lipid-regulating drugs, in-patient admission, urgent emergency triage, insulin use, and history of SH. External validation in the HKDR cohort yielded AUROC of 0.856 [95% CI 0.838 to 0.873]. Main limitations of this study included limited transportability of the model and lack of geographically independent validation. CONCLUSIONS: Our novel-ML model demonstrated good discrimination and high precision in predicting one-year risk of SH requiring hospitalization. This may be integrated into EHR decision support systems for preemptive intervention in older adults at highest risk.

Accuracy of continuous glucose monitoring during exercise-related hypoglycemia in individuals with type 1 diabetes.

Background: Hypoglycemia is common in individuals with type 1 diabetes, especially during exercise. We investigated the accuracy of two different continuous glucose monitoring systems during exercise-related hypoglycemia in an experimental setting. Materials and methods: Fifteen individuals with type 1 diabetes participated in two separate euglycemic-hypoglycemic clamp days (Clamp-exercise and Clamp-rest) including five phases: 1) baseline euglycemia, 2) plasma glucose (PG) decline ± exercise, 3) 15-minute hypoglycemia ± exercise, 4) 45-minute hypoglycemia, and 5) recovery euglycemia. Interstitial PG levels were measured every five minutes, using Dexcom G6 (DG6) and FreeStyle Libre 1 (FSL1). Yellow Springs Instruments 2900 was used as PG reference method, enabling mean absolute relative difference (MARD) assessment for each phase and Clarke error grid analysis for each day. Results: Exercise had a negative effect on FSL1 accuracy in phase 2 and 3 compared to rest (ΔMARD = +5.3 percentage points [(95% CI): 1.6, 9.1] and +13.5 percentage points [6.4, 20.5], respectively). In contrast, exercise had a positive effect on DG6 accuracy during phase 2 and 4 compared to rest (ΔMARD = -6.2 percentage points [-11.2, -1.2] and -8.4 percentage points [-12.4, -4.3], respectively). Clarke error grid analysis showed a decrease in clinically acceptable treatment decisions during Clamp-exercise for FSL1 while a contrary increase was observed for DG6. Conclusion: Physical exercise had clinically relevant impact on the accuracy of the investigated continuous glucose monitoring systems and their ability to accurately detect hypoglycemia.

Transitional Neonatal Hypoglycemia and Adverse Neurodevelopment in Midchildhood.

Importance: The circumstances under which neonatal hypoglycemia leads to brain damage remain unclear due to a lack of long-term data on the neurodevelopment of affected children. As a result, diagnostic strategies and treatment recommendations are inconsistent. Objective: To evaluate whether the occurrence of severe transitional neonatal hypoglycemia (defined as having at least 1 blood glucose measurement of 30 mg/dL or below) is associated with adverse neurodevelopment in midchildhood. Design, Setting, and Participants: This cohort study using neurodevelopmental testing of a retrospectively recruited cohort was conducted at a single-center tertiary hospital in Germany between March 2022 and February 2023. Children with neonatal blood glucose screening data were randomly selected from all births between 2010 and 2015. Frequency matching for sex, birth weight, gestational age, socioeconomic status, and primary risk factors for neonatal hypoglycemia was performed. Children with persistent hypoglycemia diseases or any risk factor for adverse neurodevelopment except hypoglycemia were excluded. Data were analyzed between February 2023 and March 2023. Exposure: At least 1 neonatal hypoglycemia measurement with blood glucose measuring 30 mg/dL or below vs all measured blood glucose levels above 30 mg/dL during postnatal blood glucose screening starting on the first day of life. Main Outcomes and Measures: Cognitive function measured by full-scale IQ test. Secondary outcomes included standardized scales of motor, visual, and executive functions, and child behavior, each measured at ages 7 to 11 years. Results: A total of 140 children (mean [SD] age 9.1 [1.3] years; 77 male [55.0%]) participated in the study. Children with severe neonatal hypoglycemia had a 4.8 points lower mean full-scale IQ than controls (107.0 [95% CI, 104.0-109.9] vs 111.8 [95% CI, 108.8-114.8]). They showed a 4.9-fold (95% CI, 1.5-15.5) increased odds of abnormal fine motor function and a 5.3-fold (95% CI, 2.1-13.3) increased odds of abnormal visual-motor integration. Significantly higher T scores for attention problems (58.2 [95% CI, 56.1-60.2] vs 54.6 [95% CI, 52.6-56.6]) and attention-deficit/hyperactivity disorder symptoms (58.2 [95% CI, 56.2-60.2] vs 54.7 [95% CI, 52.8-56.7]) were reported by parents. Conclusions and Relevance: Neonatal hypoglycemia with blood glucose levels of 30 mg/dL or below was associated with an increased risk for suboptimal neurodevelopmental outcomes in midchildhood. These findings imply that treatment strategies should aim to prevent episodes of hypoglycemia at these severely low levels.

[Continuous glucose monitoring data: how can they be collected and used in practice?] / Les données de la mesure continue du glucose : comment les récupérer et pour quel usage dans la pratique ?

CONTINUOUS GLUCOSE MONITORING DATA: HOW CAN THEY BE COLLECTED AND USED IN PRACTICE? Continuous glucose monitoring (CGM) is becoming an essential part of diabetes management. The AGP report is obtained over a 14-day period, with at least 70% of captured data. The time spent in the 70-180 mg/dl targel range, withe a target of over 70% or 50% in frail patients, is a new parameter that is essential for assessing glycemic control via CGM. Complemented by estimated HBA1c, now called GMI (Glucose Management Indicator), the time spent in hypoglycemia (target inférieur 5% or even inférieur 1% for frail patients) and the coefficient of variation (target inférieur 36%), the CGM offers a very comprehensive analysis of blood glucose levels, with individualized treatment adjustments based on ambulatory blood glucose profiles.

Diagnostic Yield of Critical Sample and Elective Fast-Test in Children After a Hypoglycemic Event: Experience From a Single Center in Israel.

OBJECTIVE: To determine the diagnostic yield of the critical sample and fast-tests as dynamic function tests for the work-up of hypoglycemia in children. METHODS: A retrospective record review of children (0-18 years) with a diagnosis of hypoglycemia (glucose ≤ 50 mg/dL) was performed. A comparison of results of critical sample (obtained during an episode of hypoglycemia) and fast-test (performed to induce hypoglycemia in fasting state) was done. RESULTS: In 317 patients with hypoglycemia, data of 89 critical samples and 52 fast-tests were taken. Only 7 (7.8%) patients who underwent critical sample testing received an endocrine or metabolic diagnosis. No confirmatory diagnoses were made using the fast-tests. Idiopathic ketotic hypoglycemia was detected in 33/89 (37.1%) of critical samples and 21/52 (40.4%) of fast-tests. The completeness of workup including the hormonal and metabolic profile was <80% in both tests. CONCLUSION: The confirmatory yield of critical sample was better than fast-test. The processing of metabolic analytes was incomplete in a few, suggesting the need to rationalize the dynamic function testing.

Solitary pleural fibroma causing IGF-2-mediated hypoglycaemia in a non-diabetic patient.

A patient without a diagnosis of diabetes mellitus presented to the hospital due to a fall and hypoglycaemia on admission. The patient was found to have recurrent nocturnal fasting hypoglycaemia. CT revealed a large lung mass consistent with a solitary pleural fibroma, a rare tumour associated with insulin-like growth factor 2 (IGF-2) production. This case is an important reminder that potential causes of hypoglycaemia should be considered in non-diabetic patients.

Use of machine learning to identify characteristics associated with severe hypoglycemia in older adults with type 1 diabetes: a post-hoc analysis of a case-control study.

INTRODUCTION: Severe hypoglycemia (SH) in older adults (OAs) with type 1 diabetes is associated with profound morbidity and mortality, yet its etiology can be complex and multifactorial. Enhanced tools to identify OAs who are at high risk for SH are needed. This study used machine learning to identify characteristics that distinguish those with and without recent SH, selecting from a range of demographic and clinical, behavioral and lifestyle, and neurocognitive characteristics, along with continuous glucose monitoring (CGM) measures. RESEARCH DESIGN AND METHODS: Data from a case-control study involving OAs recruited from the T1D Exchange Clinical Network were analyzed. The random forest machine learning algorithm was used to elucidate the characteristics associated with case versus control status and their relative importance. Models with successively rich characteristic sets were examined to systematically incorporate each domain of possible risk characteristics. RESULTS: Data from 191 OAs with type 1 diabetes (47.1% female, 92.1% non-Hispanic white) were analyzed. Across models, hypoglycemia unawareness was the top characteristic associated with SH history. For the model with the richest input data, the most important characteristics, in descending order, were hypoglycemia unawareness, hypoglycemia fear, coefficient of variation from CGM, % time blood glucose below 70 mg/dL, and trail making test B score. CONCLUSIONS: Machine learning may augment risk stratification for OAs by identifying key characteristics associated with SH. Prospective studies are needed to identify the predictive performance of these risk characteristics.

Therapeutic inertia in treatment of older adults with type II diabetes at high risk for hypoglycemia.

Patients 80 years or older with HbA1c <7.0% (53 mmol/mol) treated with multiple daily insulin injections had low rates of rapid-acting insulin deprescription and initiation of diabetes medications with lower risk of hypoglycemia. Further investigation is needed to elucidate factors contributing to potentially inappropriately aggressive treatment of these patients.

The impact of prior exposure to hypoglycaemia on the inflammatory response to a subsequent hypoglycaemic episode.

BACKGROUND: Hypoglycaemia has been shown to induce a systemic pro-inflammatory response, which may be driven, in part, by the adrenaline response. Prior exposure to hypoglycaemia attenuates counterregulatory hormone responses to subsequent hypoglycaemia, but whether this effect can be extrapolated to the pro-inflammatory response is unclear. Therefore, we investigated the effect of antecedent hypoglycaemia on inflammatory responses to subsequent hypoglycaemia in humans. METHODS: Healthy participants (n = 32) were recruited and randomised to two 2-h episodes of either hypoglycaemia or normoglycaemia on day 1, followed by a hyperinsulinaemic hypoglycaemic (2.8 ± 0.1 mmol/L) glucose clamp on day 2. During normoglycaemia and hypoglycaemia, and after 24 h, 72 h and 1 week, blood was drawn to determine circulating immune cell composition, phenotype and function, and 93 circulating inflammatory proteins including hs-CRP. RESULTS: In the group undergoing antecedent hypoglycaemia, the adrenaline response to next-day hypoglycaemia was lower compared to the control group (1.45 ± 1.24 vs 2.68 ± 1.41 nmol/l). In both groups, day 2 hypoglycaemia increased absolute numbers of circulating immune cells, of which lymphocytes and monocytes remained elevated for the whole week. Also, the proportion of pro-inflammatory CD16+-monocytes increased during hypoglycaemia. After ex vivo stimulation, monocytes released more TNF-α and IL-1ß, and less IL-10 in response to hypoglycaemia, whereas levels of 19 circulating inflammatory proteins, including hs-CRP, increased for up to 1 week after the hypoglycaemic event. Most of the inflammatory responses were similar in the two groups, except the persistent pro-inflammatory protein changes were partly blunted in the group exposed to antecedent hypoglycaemia. We did not find a correlation between the adrenaline response and the inflammatory responses during hypoglycaemia. CONCLUSION: Hypoglycaemia induces an acute and persistent pro-inflammatory response at multiple levels that occurs largely, but not completely, independent of prior exposure to hypoglycaemia. Clinical Trial information Clinicaltrials.gov no. NCT03976271 (registered 5 June 2019).

Real World Interstitial Glucose Profiles of a Large Cohort of Physically Active Men and Women.

The use of continuous glucose monitors (CGMs) in individuals living without diabetes is increasing. The purpose of this study was to profile various CGM metrics around nutritional intake, sleep and exercise in a large cohort of physically active men and women living without any known metabolic disease diagnosis to better understand the normative glycemic response to these common stimuli. A total of 12,504 physically active adults (age 40 ± 11 years, BMI 23.8 ± 3.6 kg/m2; 23% self-identified as women) wore a real-time CGM (Abbott Libre Sense Sport Glucose Biosensor, Abbott, USA) and used a smartphone application (Supersapiens Inc., Atlanta, GA, USA) to log meals, sleep and exercise activities. A total of >1 M exercise events and 274,344 meal events were analyzed. A majority of participants (85%) presented an overall (24 h) average glucose profile between 90 and 110 mg/dL, with the highest glucose levels associated with meals and exercise and the lowest glucose levels associated with sleep. Men had higher mean 24 h glucose levels than women (24 h-men: 100 ± 11 mg/dL, women: 96 ± 10 mg/dL). During exercise, the % time above >140 mg/dL was 10.3 ± 16.7%, while the % time <70 mg/dL was 11.9 ± 11.6%, with the remaining % within the so-called glycemic tight target range (70-140 mg/dL). Average glycemia was also lower for females during exercise and sleep events (p < 0.001). Overall, we see small differences in glucose trends during activity and sleep in females as compared to males and higher levels of both TAR and TBR when these active individuals are undertaking or competing in endurance exercise training and/or competitive events.

Burden of impaired awareness of hypoglycemia in people with diabetes undergoing hemodialysis.

INTRODUCTION: Impaired awareness of hypoglycemia (IAH) refers to a diminished capacity to detect hypoglycemia. IAH can result in severe and even life-threatening outcomes for individuals with diabetes, especially those in advanced stages of the disease. This study aimed to assess the prevalence of IAH in people with diabetes on hemodialysis. RESEARCH DESIGN AND METHODS: We conducted a single-center audit to assess the prevalence of IAH using the Clarke questionnaire. Simultaneously, we measured fear of hypoglycemia with an adapted version of the Hypoglycemia Survey and recorded the incidence of severe hypoglycemia. Data were presented as mean±SD or counts/percentages. Logistic regression was then employed to analyze the association between IAH and various sociodemographic and clinical factors. RESULTS: We included 56 participants with diabetes on hemodialysis, with a mean age of 67.2 years (±12.9), of whom 51.8% were male. The ethnic distribution was 23.2% white, 23.2% black, 19.6% Asian, and 33.9% unspecified. The mean HbA1c was 52 mmol/mol (±18.6). The majority (91.1%) had a diagnosis of type 2 diabetes, and 55.4% of those were treated with insulin. The use of diabetes technology was low, with 2.8% of the participants using a continuous glucose monitor. IAH prevalence was 23.2%, and among the 57 participants, 23.6% had a history of severe hypoglycemia, and 60.6% reported fear of hypoglycemia. There were no significant differences in sociodemographic and clinical characteristics between those with IAH and normal hypoglycemia awareness. CONCLUSIONS: We observed that 23.2% of individuals with diabetes undergoing hemodialysis had IAH. IAH was more prevalent in people who reported a fear of hypoglycemia and had a history of severe hypoglycemia episode. The study highlights the unmet needs and disparities in access to diabetes technology within this population.

Intraoperative segmental pancreatic occlusion and insulin assay combination optimizes the localization of lesions and confirmation of complete resection in pancreatic hypoglycemia patients.

AIM: To evaluate the efficacy of the application of intraoperative segmental pancreatic occlusion and insulin assay in surgical procedures for pancreatic hypoglycemia. METHODS: We retrospectively analyzed the clinical data of 11 pancreatic hypoglycemia cases treated in the China-Japan Friendship Hospital between September 2015 and August 2021. Intraoperative segmental pancreatic occlusion and insulin assay were used to enhance hypersecretory pancreatic tissues' localization and to achieve a complete resection. Intraoperative testing of insulin levels (peripheral venous blood) was carried out at several time points starting from before the resection of hypersecretory tissues (base value) and at 1 minute, 5 minutes, 15 minutes, 30 minutes, and 60 minutes after resection. Additional testing every 30 minutes until the end of the operation was carried out when necessary. RESULTS: A total of 11 pancreatic hypoglycemia cases were included; 9 cases were insulinomas (all with single pancreatic lesions, with 4 located in the head, 1 in the body, and 4 in the tail), 1 MEN-1, and 1 nesidioblastosis. The insulin assay (30 minutes after the resection of hypersecretory tissues) enhanced the ability to locate target tissues and the accuracy of complete resection to 100%. As for intraoperative blood glucose monitoring, the accuracy 30 minutes after resection was as low as 36.6%. Postoperative levels of insulin and glucose were normal in all patients, with no recurrence of hypoglycemic symptoms during postoperative follow-up visits (9 to 72 months). CONCLUSION: Intraoperative segmental pancreatic occlusion and insulin assay in pancreatic hypoglycemia is a simple, accurate, and fast approach that enhances the localization and complete resection of hypersecretory tissues. Such a combination is highly significant in challenging cases of hypoglycemia.

Neonatal hypoglycemia and neurodevelopmental outcomes: Yesterday, today, tomorrow.

Neonatal hypoglycemia is a major source of concern for pediatricians since it has commonly been related to poor neurodevelopmental outcomes. Diagnosis is challenging, considering the different operational thresholds provided by each guideline. Screening of infants at risk plays a crucial role, considering that most hypoglycemic infants show no clinical signs. New opportunities for prevention and treatment are provided by the use of oral dextrose gel. Continuous glucose monitoring systems could be a feasible tool in the next future. Furthermore, there is still limited evidence to underpin the current clinical practice of administering, in case of hypoglycemia, an intravenous "mini-bolus" of 10% dextrose before starting a continuous dextrose infusion. This brief review provides an overview of the latest advances in this field and neurodevelopmental outcomes according to different approaches.   Conclusion: To adequately define if a more permissive approach is risk-free for neurodevelopmental outcomes, more research on continuous glucose monitoring and long-term follow-up is still needed. What is Known: • Neonatal hypoglycemia (NH) is a well-known cause of brain injury that could be prevented to avoid neurodevelopmental impairment. • Diagnosis is challenging, considering the different suggested operational thresholds for NH (<36, <40, <45, <47 or <50 mg/dl). What is New: • A 36 mg/dl threshold seems to be not associated with a worse psychomotor development at 18 months of life when compared to the "traditional" threshold (47 mg/dl). • Further studies on long-term neurodevelopmental outcomes are required before suggesting a more permissive management of NH.

Validation of an international classification of disease, tenth revision, clinical modification (ICD-10-CM) algorithm in identifying severe hypoglycaemia events for real-world studies.

AIM: The transition to the ICD-10-CM coding system has reduced the utility of hypoglycaemia algorithms based on ICD-9-CM diagnosis codes in real-world studies of antidiabetic drugs. We mapped a validated ICD-9-CM hypoglycaemia algorithm to ICD-10-CM codes to create an ICD-10-CM hypoglycaemia algorithm and assessed its performance in identifying severe hypoglycaemia. MATERIALS AND METHODS: We assembled a cohort of Medicare patients with DM and linked electronic health record (EHR) data to the University of North Carolina Health System and identified candidate severe hypoglycaemia events from their Medicare claims using the ICD-10-CM hypoglycaemia algorithm. We confirmed severe hypoglycaemia by EHR review and computed a positive predictive value (PPV) of the algorithm to assess its performance. We refined the algorithm by removing poor performing codes (PPV ≤0.5) and computed a Cohen's κ statistic to evaluate the agreement of the EHR reviews. RESULTS: The algorithm identified 642 candidate severe hypoglycaemia events, and we confirmed 455 as true severe hypoglycaemia events, PPV of 0.709 (95% confidence interval: 0.672, 0.744). When we refined the algorithm, the PPV increased to 0.893 (0.862, 0.918) and missed <2.42% (<11) true severe hypoglycaemia events. Agreement between reviewers was high, κ = 0.93 (0.89, 0.97). CONCLUSIONS: We translated an ICD-9-CM hypoglycaemia algorithm to an ICD-10-CM version and found its performance was modest. The performance of the algorithm improved by removing poor performing codes at the trade-off of missing very few severe hypoglycaemia events. The algorithm has the potential to be used to identify severe hypoglycaemia in real-world studies of antidiabetic drugs.