Severe (level 3) hypoglycaemia occurrence in a real-world cohort of adults with type 1 or 2 diabetes mellitus (iNPHORM, United States).

AIMS: Among adults with insulin- and/or secretagogue-treated diabetes in the United States, very little is known about the real-world descriptive epidemiology of iatrogenic severe (level 3) hypoglycaemia. Addressing this gap, we collected primary, longitudinal data to quantify the absolute frequency of events as well as incidence rates and proportions. MATERIALS AND METHODS: iNPHORM is a US-wide, 12-month ambidirectional panel survey (2020-2021). Adults with type 1 diabetes mellitus (T1DM) or insulin- and/or secretagogue-treated type 2 diabetes mellitus (T2DM) were recruited from a probability-based internet panel. Participants completing ≥1 follow-up questionnaire(s) were analysed. RESULTS: Among 978 respondents [T1DM 17%; mean age 51 (SD 14.3) years; male: 49.6%], 63% of level 3 events were treated outside the health care system (e.g. by family/friend/colleague), and <5% required hospitalization. Following the 12-month prospective period, one-third of individuals reported ≥1 event(s) [T1DM 44.2% (95% CI 36.8%-51.8%); T2DM 30.8% (95% CI 28.7%-35.1%), p = .0404, α = 0.0007]; and the incidence rate was 5.01 (95% CI 4.15-6.05) events per person-year (EPPY) [T1DM 3.57 (95% CI 2.49-5.11) EPPY; T2DM 5.29 (95% CI 4.26-6.57) EPPY, p = .1352, α = 0.0007]. Level 3 hypoglycaemia requiring non-transport emergency medical services was more common in T2DM than T1DM (p < .0001, α = 0.0016). In total, >90% of events were experienced by <15% of participants. CONCLUSIONS: iNPHORM is one of the first long-term, prospective US-based investigations on level 3 hypoglycaemia epidemiology. Our results underscore the importance of participant-reported data to ascertain its burden. Events were alarmingly frequent, irrespective of diabetes type, and concentrated in a small subsample.

Development and validation of a real-world model to predict 1-year Level 3 (severe) hypoglycaemia risk in adults with diabetes (the iNPHORM study, United States).

AIMS: We aimed to develop and internally validate a real-world prognostic model for Level 3 hypoglycaemia risk compatible with outpatient care in the United States. MATERIALS AND METHODS: iNPHORM is a 12-month, US-based panel survey. Adults (18-90 years old) with type 1 diabetes mellitus or insulin- and/or secretagogue-treated type 2 diabetes mellitus were recruited from a nationwide, probability-based internet panel. Among participants completing ≥ 1 follow-up questionnaire(s), we modelled 1-year Level 3 hypoglycaemia risk using Andersen and Gill's Cox survival and penalized regression with multiple imputation. Candidate variables were selected for their clinical relevance and ease of capture at point-of-care. RESULTS: In total, 986 participants [type 1 diabetes mellitus: 17%; men: 49.6%; mean age: 51 (SD: 14.3) years] were analysed. Across follow-up, 035.1 (95% CI: 32.2-38.1)% reported ≥1 Level 3 event(s), and the rate was 5.0 (95% CI: 4.1-6.0) events per person-year. Our final model showed strong discriminative validity and parsimony (optimism corrected c-statistic: 0.77). Numerous variables were selected: age; sex; body mass index; marital status; level of education; insurance coverage; race; ethnicity; food insecurity; diabetes type; glycated haemoglobin value; glycated haemoglobin variability; number, type and dose of various medications; number of SH events requiring hospital care (past year and over follow-up); type and number of comorbidities and complications; number of diabetes-related health care visits (past year); use of continuous/flash glucose monitoring; and general health status. CONCLUSIONS: iNPHORM is the first US-based primary prognostic study on Level 3 hypoglycaemia. Future model implementation could potentiate risk-tailored strategies that reduce real-world event occurrence and overall diabetes burden.

Real-World Effects of Second-Generation Versus Earlier Intermediate/Basal Insulin Analogues on Rates of Hypoglycemia in Adults with Type 1 and 2 Diabetes (iNPHORM, US).

INTRODUCTION: Second-generation basal insulin analogues have been shown to reduce hypoglycemia in several trials and observational studies of select populations; however, it remains unclear whether these results persist in real-world settings. Using self-reported hypoglycemia events, we assessed whether second-generation basal insulin analogues reduce rates of hypoglycemia events (non-severe/severe; overall/daytime/nocturnal) compared to earlier intermediate/basal insulin analogues among people with insulin-treated type 1 or 2 diabetes. METHODS: We used prospectively collected data from the Investigating Novel Predictions of Hypoglycemia Occurrence Using Real-World Models (iNPHORM) panel survey. This US-wide, 1-year internet-based survey assessed hypoglycemia experiences and related sociodemographic and clinical characteristics of people with diabetes (February 2020-March 2021). We estimated population-average rate ratios for hypoglycemia comparing second-generation to earlier intermediate/basal insulin analogues using negative binomial regression, adjusting for confounders. Within-person variability of repeated observations was addressed with generalized estimating equations. RESULTS: Among iNPHORM participants with complete data, N = 413 used an intermediate/basal insulin analogue for ≥ 1 month during follow-up. After adjusting for baseline and time-updated confounders, average second-generation basal insulin analogue users experienced a 19% (95% CI 3-32%, p = 0.02) lower rate of overall non-severe hypoglycemia and 43% (95% CI 26-56%, p < 0.001) a lower rate of nocturnal non-severe hypoglycemia compared to earlier intermediate/basal insulin users. Overall severe hypoglycemia rates were similar among second-generation and earlier intermediate/basal insulin users (p = 0.35); however, the rate of severe nocturnal hypoglycemia was reduced by 44% (95% CI 10-65%, p = 0.02) among second-generation insulin users compared to earlier intermediate/basal insulin users. CONCLUSION: Our real-world results suggest second-generation basal insulin analogues reduce rates of hypoglycemia, especially nocturnal non-severe and severe events. Whenever possible and feasible, clinicians should prioritize prescribing these agents over first-generation basal or intermediate insulin in people with type 1 and 2 diabetes.

An introduction to machine learning for classification and prediction.

Classification and prediction tasks are common in health research. With the increasing availability of vast health data repositories (e.g. electronic medical record databases) and advances in computing power, traditional statistical approaches are being augmented or replaced with machine learning (ML) approaches to classify and predict health outcomes. ML describes the automated process of identifying ("learning") patterns in data to perform tasks. Developing an ML model includes selecting between many ML models (e.g. decision trees, support vector machines, neural networks); model specifications such as hyperparameter tuning; and evaluation of model performance. This process is conducted repeatedly to find the model and corresponding specifications that optimize some measure of model performance. ML models can make more accurate classifications and predictions than their statistical counterparts and confer greater flexibility when modelling unstructured data or interactions between covariates; however, many ML models require larger sample sizes to achieve good classification or predictive performance and have been criticized as "black box" for their poor transparency and interpretability. ML holds potential in family medicine for risk profiling of patients' disease risk and clinical decision support to present additional information at times of uncertainty or high demand. In the future, ML approaches are positioned to become commonplace in family medicine. As such, it is important to understand the objectives that can be addressed using ML approaches and the associated techniques and limitations. This article provides a brief introduction into the use of ML approaches for classification and prediction tasks in family medicine.

Real-world risk factors of confirmed or probable COVID-19 in Americans with diabetes: A prospective, community-based study (iNPHORM).

INTRODUCTION: Americans with diabetes are clinically vulnerable to worse COVID-19 outcomes; thus, insight into how to prevent infection is imperative. Using longitudinal, prospective data from the real-world iNPHORM study, we identify the intrinsic and extrinsic risk factors of confirmed or probable COVID-19 in people with type 1 or 2 diabetes. METHODS: The iNPHORM study recruited 1206 Americans (18-90 years) with insulin- and/or secretagogue-treated type 1 or 2 diabetes from a probability-based internet panel. Online questionnaires (screener, baseline and 12 monthly follow-ups) assessed COVID-19 incidence and various plausible intrinsic and extrinsic factors. Multivariable Cox regression was used to model the rate of COVID-19 (confirmed or probable). Risk factors were selected using a repeated backwards-selection 'voting' procedure. RESULTS: A sub-sample of 817 iNPHORM participants (type 1 diabetes: 16.9%; age: 52.1 [SD: 14.2] years; female: 50.2%) was analysed between May 2020 and March 2021. During this period, 13.7% reported confirmed or probable COVID-19. Age, body mass index, number of chronic comorbidities, most recent A1C, past severe hypoglycaemia, and employment status were selected in our final model. Body mass index ≥30 kg/m2 versus <30 kg/m2 (HR 1.63 [1.05; 2.52]95% CI ), and increased number of comorbidities (HR 1.16 [1.05; 1.27]95% CI ) independently predicted COVID-19 incidence. Marginally significant effects were observed for overall A1C (p = .06) and employment status (p = .07). CONCLUSIONS: This is the first US-based epidemiologic investigation to characterize community-based COVID-19 susceptibility in diabetes. Our results reveal specific and promising avenues to prevent COVID-19 in this at-risk population. CLINICALTRIALS: gov Identifier: NCT04219514.

Predicting Real-world Hypoglycemia Risk in American Adults With Type 1 or 2 Diabetes Mellitus Prescribed Insulin and/or Secretagogues: Protocol for a Prospective, 12-Wave Internet-Based Panel Survey With Email Support (the iNPHORM [Investigating Novel Predictions of Hypoglycemia Occurrence Using Real-world Models] Study).

BACKGROUND: Hypoglycemia prognostic models contingent on prospective, self-reported survey data offer a powerful avenue for determining real-world event susceptibility and interventional targets. OBJECTIVE: This protocol describes the design and implementation of the 1-year iNPHORM (Investigating Novel Predictions of Hypoglycemia Occurrence Using Real-world Models) study, which aims to measure real-world self-reported severe and nonsevere hypoglycemia incidence (daytime and nocturnal) in American adults with type 1 or 2 diabetes mellitus prescribed insulin and/or secretagogues, and develop and internally validate prognostic models for severe, nonsevere daytime, and nonsevere nocturnal hypoglycemia. As a secondary objective, iNPHORM aims to quantify the effects of different antihyperglycemics on hypoglycemia rates. METHODS: iNPHORM is a prospective, 12-wave internet-based panel survey that was conducted across the United States. Americans (aged 18-90 years) with self-reported type 1 or 2 diabetes mellitus prescribed insulin and/or secretagogues were conveniently sampled via the web from a pre-existing, closed, probability-based internet panel (sample frame). A sample size of 521 baseline responders was calculated for this study. Prospective data on hypoglycemia and potential prognostic factors were self-assessed across 14 closed, fully automated questionnaires (screening, baseline, and 12 monthly follow-ups) that were piloted using semistructured interviews (n=3) before fielding; no face-to-face contact was required as part of the data collection. Participant responses will be analyzed using multivariable count regression and machine learning techniques to develop and internally validate prognostic models for 1-year severe and 30-day nonsevere daytime and nocturnal hypoglycemia. The causal effects of different antihyperglycemics on hypoglycemia rates will also be investigated. RESULTS: Recruitment and data collection occurred between February 2020 and March 2021 (ethics approval was obtained on December 17, 2019). A total of 1694 participants completed the baseline questionnaire, of whom 1206 (71.19%) were followed up for 12 months. Most follow-up waves (10,470/14,472, 72.35%) were completed, translating to a participation rate of 179% relative to our target sample size. Over 70.98% (856/1206) completed wave 12. Analyses of sample characteristics, quality metrics, and hypoglycemia incidence and prognostication are currently underway with published results anticipated by fall 2022. CONCLUSIONS: iNPHORM is the first hypoglycemia prognostic study in the United States to leverage prospective, longitudinal self-reports. The results will contribute to improved real-world hypoglycemia risk estimation and potentially safer, more effective clinical diabetes management. TRIAL REGISTRATION: ClinicalTrials.gov NCT04219514; https://clinicaltrials.gov/ct2/show/NCT04219514. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/33726.

COVID-19 hinterland: surveilling the self-reported impacts of the pandemic on diabetes management in the USA (cross-sectional results of the iNPHORM study).

MAIN OBJECTIVE: To determine how and to what extent COVID-19 has affected real-world, self-reported glycaemic management in Americans with type 1 or type 2 diabetes taking insulin and/or secretagogues, with or without infection. DESIGN: A cross-sectional substudy using data from the Investigating Novel Predictions of Hypoglycemia Occurrence using Real-world Models panel survey. SETTING: USA. PARTICIPANTS: Americans 18-90 years old with type 1 or 2 diabetes taking insulin and/or secretagogues were conveniently sampled from a probability-based internet panel. PRIMARY OUTCOME MEASURE: A structured, COVID-19-specific questionnaire was administered to assess the impact of the pandemic (irrespective of infection) on socioeconomic, behavioural/clinical and psychosocial aspects of glycaemic management. RESULTS: Data from 667 respondents (type 1 diabetes: 18%; type 2 diabetes: 82%) were analysed. Almost 25% reported A1c values ≥8.1%. Rates of severe and non-severe hypoglycaemia were 0.68 (95% CI 0.5 to 0.96) and 2.75 (95% CI 2.4 to 3.1) events per person-month, respectively. Ten respondents reported a confirmed or probable COVID-19 diagnosis. Because of the pandemic, 24% of respondents experienced difficulties affording housing; 28% struggled to maintain sufficient food to avoid hypoglycaemia; and 19% and 17% reported challenges accessing diabetes therapies and testing strips, respectively. Over one-quarter reported issues retrieving antihyperglycaemics from the pharmacy and over one-third reported challenges consulting with diabetes providers. The pandemic contributed to therapeutic non-adherence (14%), drug rationing (17%) and reduced monitoring (16%). Many struggled to keep track, and in control, of hypoglycaemia (12%-15%) and lacked social support to help manage their risk (19%). Nearly half reported decreased physical activity. Few statistically significant differences were observed by diabetes type. CONCLUSIONS: COVID-19 was found to cause substantial self-reported deficiencies in glycaemic management. Study results signal the need for decisive action to restabilise routine diabetes care in the USA. TRIAL REGISTRATION NUMBER: NCT04219514.

Development of a prognostic prediction model to estimate the risk of multiple chronic diseases: constructing a copula-based model using Canadian primary care electronic medical record data.

INTRODUCTION: The ability to estimate risk of multimorbidity will provide valuable information to patients and primary care practitioners in their preventative efforts. Current methods for prognostic prediction modelling are insufficient for the estimation of risk for multiple outcomes, as they do not properly capture the dependence that exists between outcomes. OBJECTIVES: We developed a multivariate prognostic prediction model for the 5-year risk of diabetes, hypertension, and osteoarthritis that quantifies and accounts for the dependence between each disease using a copula-based model. METHODS: We used data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN) from 2009 onwards, a collection of electronic medical records submitted by participating primary care practitioners across Canada. We identified patients 18 years and older without all three outcome diseases and observed any incident diabetes, osteoarthritis, or hypertension within 5-years, resulting in a large retrospective cohort for model development and internal validation (n=425,228). First, we quantified the dependence between outcomes using unadjusted and adjusted Ø coefficients. We then estimated a copula-based model to quantify the non-linear dependence between outcomes that can be used to derive risk estimates for each outcome, accounting for the observed dependence. Copula-based models are defined by univariate models for each outcome and a dependence function, specified by the parameter θ. Logistic regression was used for the univariate models and the Frank copula was selected as the dependence function. RESULTS: All outcome pairs demonstrated statistically significant dependence that was reduced after adjusting for covariates. The copula-based model yielded statistically significant θ parameters in agreement with the adjusted and unadjusted Ø coefficients. Our copula-based model can effectively be used to estimate trivariate probabilities. DISCUSSION: Quantitative estimates of multimorbidity risk inform discussions between patients and their primary care practitioners around prevention in an effort to reduce the incidence of multimorbidity.

Development and evaluation of an osteoarthritis risk model for integration into primary care health information technology.

BACKGROUND: We developed and evaluated a prognostic prediction model that estimates osteoarthritis risk for use by patients and practitioners that is designed to be appropriate for integration into primary care health information technology systems. Osteoarthritis, a joint disorder characterized by pain and stiffness, causes significant morbidity among older Canadians. Because our prognostic prediction model for osteoarthritis risk uses data that are readily available in primary care settings, it supports targeting of interventions delivered as part of clinical practice that are aimed at risk reduction. METHODS: We used the CPCSSN (Canadian Primary Sentinel Surveillance Network) database, which contains aggregated electronic health information from a cohort of primary care practices, to develop and evaluate a prognostic prediction model to estimate 5-year osteoarthritis risk, addressing contextual challenges of data availability and missingness. We constructed a retrospective cohort of 383,117 eligible primary care patients who were included in the cohort if they had an encounter with their primary care practitioner between 1 January 2009 and 31 December 2010. Patients were excluded if they had a diagnosis of osteoarthritis prior to their first visit in this time period. Incident cases of osteoarthritis were observed. The model was constructed to predict incident osteoarthritis based on age, sex, BMI, previous leg injury, and osteoporosis. Evaluation of the model used internal 10-fold cross-validation; we argue that internal validation is particularly appropriate for a model that is to be integrated into the same context from which the data were derived. RESULTS: The resulting prediction model for 5-year risk of osteoarthritis diagnosis demonstrated state-of-the-art discrimination (estimated AUROC 0.84) and good calibration (assessed visually.) The model relies only on information that is readily available in Canadian primary care settings, and hence is appropriate for integration into Canadian primary care health information technology. CONCLUSIONS: If the contextual challenges arising when using primary care electronic medical record data are appropriately addressed, highly discriminative models for osteoarthritis risk may be constructed using only data commonly available in primary care. Because the models are constructed from data in the same setting where the model is to be applied, internal validation provides strong evidence that the resulting model will perform well in its intended application.