Emma M. Eggleston

Automated Detection and Classification of Type 1 Versus Type 2 Diabetes Using Electronic Health Record Data

OBJECTIVE To create surveillance algorithms to detect diabetes and classify type 1 versus type 2 diabetes using structured electronic health record (EHR) data. RESEARCH DESIGN AND METHODS We extracted 4 years of data from the EHR of a large, multisite, multispecialty ambulatory practice serving ∼700,000 patients. We flagged possible cases of diabetes using laboratory test results, diagnosis codes, and prescriptions. We assessed the sensitivity and positive predictive value of novel combinations of these data to classify type 1 versus type 2 diabetes among 210 individuals. We applied an optimized algorithm to a live, prospective, EHR-based surveillance system and reviewed 100 additional cases for validation. RESULTS The diabetes algorithm flagged 43,177 patients. All criteria contributed unique cases: 78% had diabetes diagnosis codes, 66% fulfilled laboratory criteria, and 46% had suggestive prescriptions. The sensitivity and positive predictive value of ICD-9 codes for type 1 diabetes were 26% (95% CI 12–49) and 94% (83–100) for type 1 codes alone; 90% (81–95) and 57% (33–86) for two or more type 1 codes plus any number of type 2 codes. An optimized algorithm incorporating the ratio of type 1 versus type 2 codes, plasma C-peptide and autoantibody levels, and suggestive prescriptions flagged 66 of 66 (100% [96–100]) patients with type 1 diabetes. On validation, the optimized algorithm correctly classified 35 of 36 patients with type 1 diabetes (raw sensitivity, 97% [87–100], population-weighted sensitivity, 65% [36–100], and positive predictive value, 88% [78–98]). CONCLUSIONS Algorithms applied to EHR data detect more cases of diabetes than claims codes and reasonably discriminate between type 1 and type 2 diabetes.

Sex-Specific Associations of Gestational Glucose Tolerance With Childhood Body Composition

OBJECTIVE To examine the associations of maternal gestational glucose tolerance with offspring body composition in late childhood. RESEARCH DESIGN AND METHODS Among 958 women in the prebirth cohort Project Viva, glucose tolerance was assessed in the second trimester by nonfasting 50-g 1-h glucose challenge test (GCT), followed if abnormal by fasting 100-g 3-h oral glucose tolerance test (OGTT). We categorized women as normoglycemic (83.3%) if GCT was ≤140 mg/dL, isolated hyperglycemia (9.1%) if GCT was abnormal but OGTT normal, intermediate glucose intolerance (IGI) (3.3%) if there was one abnormal value on OGTT, or gestational diabetes mellitus (GDM) (4.5%) if there were two or more abnormal OGTT values. Using multivariable linear regression, we examined adjusted associations of glucose tolerance with offspring overall (N = 958) and central (N = 760) adiposity and body composition using dual X-ray absorptiometry (DXA) measured at the school-age visit (95 ± 10 months). RESULTS Compared with that in the male offspring of normoglycemic mothers, DXA fat mass was higher in male offspring of GDM mothers (1.89 kg [95% CI 0.33–3.45]) but not in male offspring of mothers with IGI (0.06 kg [−1.45 to 1.57]). DXA trunk-to-peripheral fat mass, a measure of central adiposity, was also somewhat higher in male offspring of GDM mothers (0.04 [−0.01 to 0.09]). In girls, DXA fat mass was higher in offspring of mothers with IGI (2.23 kg [0.12–4.34]) but not GDM (−1.25 kg [−3.13 to 0.63]). We showed no association of gestational glucose tolerance with DXA lean mass. CONCLUSIONS In this study, only male offspring of GDM mothers manifested increased adiposity, whereas only female offspring of mothers with IGI did so. Sex differences in glycemic sensitivity may explain these findings.

Finding the Role of Health Care in Population Health

Enhanced focus on population health is increasingly invoked as a potential solution to the persistent problems facing the US health care system including failures to achieve targets for health outcomes, eliminate disparities in health and health care, and function within a sustainable budget Shortell1 recently discussed how the Affordable Care Act (ACA) and related developments could change incentives to align health care, public health, and social services. Sox2 further specified the inherent tensions in allocation of resources to balance the needs of individuals and those of the population overall and highlighted new aspects of medical professionalism that will be needed to improve population health outcomes However, additional clarity is needed regarding the specific health system–based activities that may contribute most to improvements in population health and well-being and the barriers that must be overcome for them to succeed. These include stakeholder interests that may not be aligned with investments in population health, barriers to information transfer and service integration between health care and other sectors, and persistent difficulties in addressing health care disparities. Without solutions to these challenges, integration of population health–focused activities into the routine work of health care systems will be neither robust nor sustainable.

Rational use of electronic health records for diabetes population management.

Population management is increasingly invoked as an approach to improve the quality and value of diabetes care. Recent emphasis is driven by increased focus on both costs and measures of care as the US moves from fee for service to payment models in which providers are responsible for costs incurred, and outcomes achieved, for their entire patient population. The capacity of electronic health records (EHRs) to create patient registries, apply analytic tools, and facilitate provider- and patient-level interventions has allowed rapid evolution in the scope of population management initiatives. However, findings on the efficacy of these efforts for diabetes are mixed, and work remains to achieve the full potential of an-EHR based population approach. Here we seek to clarify definitions and key domains, provide an overview of evidence for EHR-based diabetes population management, and recommend future directions for applying the considerable power of EHRs to diabetes care and prevention.

Variation in Postpartum Glycemic Screening in Women With a History of Gestational Diabetes Mellitus.

OBJECTIVE: To assess patterns and predictors of postpartum diabetes screening in a commercially insured, geographically and sociodemographically diverse sample of women with gestational diabetes mellitus. METHODS: Using commercial insurance claims (2000–2012) from all 50 states, we conducted a retrospective cohort study in 447,556 women with at least one delivery and continuous enrollment 1 year before and after delivery. We identified women with a gestational diabetes mellitus pregnancy and examined postpartum diabetes screening type and timing and performed logistic regression to identify screening predictors. RESULTS: Gestational diabetes mellitus was diagnosed in 32,253 (7.2%) women during the study timeframe. Three fourths received no screening within 1 year postpartum. Rates of recommended 75-g oral glucose tolerance testing within 6–12 weeks were low but increased over time (27 [2%] in 2001 compared with 249 [7%] in 2011, adjusted odds ratio [OR] 3.1, 95% confidence interval [CI] 2.0–47). Among women screened, those in the Northeast (19%) and South (18%) were least likely to receive a 75-g oral glucose tolerance test within 0–12 weeks (adjusted OR 0.4 for each, CI 0.4–0.5) compared with the West (36%). Asian women were most likely to receive any screening (18%; adjusted OR 1.5, CI 1.3–1.6) compared with white women (12%). Black women were most likely to receive hemoglobin A1c (21%; adjusted OR 2.0, CI 1.3–3.2) compared with white women (11%). Antepartum antiglycemic medication (21%; adjusted OR 2.1, CI 2.0–2.3) or visit to a nutritionist–diabetes educator (19%; adjusted OR 1.6, CI 1.4–1.7) or endocrinologist (23%; adjusted OR 1.7, CI 1.6–1.9) predicted screening within 12 weeks postpartum. CONCLUSION: Postpartum diabetes screening remains widely underused among commercially insured women with gestational diabetes mellitus. Differences in screening by geography, race, and antepartum care can inform health system and public health interventions to increase diabetes detection in this high-risk population.

Harnessing Electronic Health Records for Public Health Surveillance

Electronic medical record (EMR) systems are a rich potential source for detailed, timely, and efficient surveillance of large populations. We created the Electronic medical record Support for Public Health (ESP) system to facilitate and demonstrate the potential advantages of harnessing EMRs for public health surveillance. ESP organizes and analyzes EMR data for events of public health interest and transmits electronic case reports or aggregate population summaries to public health agencies as appropriate. It is designed to be compatible with any EMR system and can be customized to different states’ messaging requirements. All ESP code is open source and freely available. ESP currently has modules for notifiable disease, influenza-like illness syndrome, and diabetes surveillance. An intelligent presentation system for ESP called the RiskScape is under development. The RiskScape displays surveillance data in an accessible and intelligible format by automatically mapping results by zip code, stratifying outcomes by demographic and clinical parameters, and enabling users to specify custom queries and stratifications. The goal of RiskScape is to provide public health practitioners with rich, up-to-date views of health measures that facilitate timely identification of health disparities and opportunities for targeted interventions. ESP installations are currently operational in Massachusetts and Ohio, providing live, automated surveillance on over 1 million patients. Additional installations are underway at two more large practices in Massachusetts.

Innovative uses of electronic health records and social media for public health surveillance.

Electronic health records (EHRs) and social media have the potential to enrich public health surveillance of diabetes. Clinical and patient-facing data sources for diabetes surveillance are needed given its profound public health impact, opportunity for primary and secondary prevention, persistent disparities, and requirement for self-management. Initiatives to employ data from EHRs and social media for diabetes surveillance are in their infancy. With their transformative potential come practical limitations and ethical considerations. We explore applications of EHR and social media for diabetes surveillance, limitations to approaches, and steps for moving forward in this partnership between patients, health systems, and public health.

Diabetes Outpatient Care and Acute Complications Before and After High-Deductible Insurance Enrollment: A Natural Experiment for Translation in Diabetes (NEXT-D) Study.

Importance High-deductible health plans (HDHPs) have expanded under the Affordable Care Act and are expected to play a major role in the future of US health policy. The effects of modern HDHPs on chronically ill patients and adverse outcomes are unknown. Objective To determine the association of HDHP with high-priority diabetes outpatient care and preventable acute complications. Design, Setting, and Participants Controlled interrupted-time-series study using a large national health insurer database from January 1, 2003, to December 31, 2012. A total of 12 084 HDHP members with diabetes, aged 12 to 64 years, who were enrolled for 1 year in a low-deductible (⩽

Participatory surveillance of diabetes device safety: a social media-based complement to traditional FDA reporting

500) plan followed by 2 years in an HDHP (≥

‘Hard,’ ‘soft’ and ‘surrogate’ endpoints in diabetes

1000) after an employer-mandated switch were included. Patients transitioning to HDHPs were propensity-score matched with contemporaneous patients whose employers offered only low-deductible coverage. Low-income (n = 4121) and health savings account (HSA)–eligible (n = 1899) patients with diabetes were subgroups of interest. Data analysis was performed from February 23, 2015, to September 11, 2016. Exposures Employer-mandated HDHP transition. Main Outcomes and Measures High-priority outpatient visits, disease monitoring tests, and outpatient and emergency department visits for preventable acute diabetes complications. Results In the 12 084 HDHP members included after the propensity score match, the mean (SD) age was 50.4 (10.0) years; 5410 of the group (44.8%) were women. The overall, low-income, and HSA-eligible diabetes HDHP groups experienced increases in out-of-pocket medical expenditures of 49.4% (95% CI, 40.3% to 58.4%), 51.7% (95% CI, 38.6% to 64.7%), and 67.8% (95% CI, 47.9% to 87.8%), respectively, compared with controls in the year after transitioning to HDHPs. High-priority primary care visits and disease monitoring tests did not change significantly in the overall HDHP cohort; however, high-priority specialist visits declined by 5.5% (95% CI, −9.6% to −1.5%) in follow-up year 1 and 7.1% (95% CI, −11.5% to −2.7%) in follow-up year 2 vs baseline. Outpatient acute diabetes complication visits were delayed in the overall and low-income HDHP cohorts at follow-up (adjusted hazard ratios, 0.94 [95% CI, 0.88 to 0.99] for the overall cohort and 0.89 [95% CI, 0.81 to 0.98] for the low-income cohort). Annual emergency department acute complication visits among HDHP members increased by 8.0% (95% CI, 4.6% to 11.4%) in the overall group, 21.7% (95% CI, 14.5% to 28.9%) in the low-income group, and 15.5% (95% CI, 10.5% to 20.6%) in the HSA-eligible group. Conclusions and Relevance Patients with diabetes experienced minimal changes in outpatient visits and disease monitoring after an HDHP switch, but low-income and HSA-eligible HDHP members experienced major increases in emergency department visits for preventable acute diabetes complications.