Longitudinal Association of a Medication Risk Score With Mortality Among Ambulatory Patients Acquired Through Electronic Health Record Data.

The use of electronic health records allows for the application of a novel medication risk score for the rapid identification of ambulatory patients at risk of adverse drug events. We sought to examine the longitudinal association of medication risk score with mortality. This retrospective cohort study included patients whose data were available through electronic health records from multiple health care organizations in the United States that provided data as part of a Patient Safety Organization. Patients were included if they had ≥1 visit and ≥1 medication in their record between January 1, 2011, to June 30, 2017. Cox proportional hazards regression was used to examine the association between continuous and categorized medication risk score with all-cause mortality. Among 427,103 patients, the median age was 50 years (interquartile range, 29-64 years); 61% were female; 50% were White, 11% were Black, and 38% were Hispanic; and 6873 had a death date recorded. Patients 30 to 49 years old had the highest hazard ratios (HRs), followed by the 50- to 64-year-olds and lastly those 65 years or older. Controlling for all covariates, 30- to 49-year-olds with a score of 20 to 30 (versus <10) had a 604% increase in the hazard of death (HR, 7.04; 95% confidence interval [CI], 3.86-12.85), 50- to 64-year-olds had a 254% increase (HR, 3.54; 95% CI, 2.71-4.63), and ≥65-year-olds had an 87% increase (HR, 1.87; 95% CI, 1.67-2.09). The medication risk score was independently associated with death, adjusting for multimorbidities and other conditions. Risk was found to vary by age group and score. Results suggest that pharmaceutical interventions among those with elevated scores could improve medication safety for patients taking multiple medications.

The Longitudinal Epidemiologic Assessment of Diabetes Risk (LEADR): Unique 1.4 M patient Electronic Health Record cohort.

BACKGROUND: The Longitudinal Epidemiologic Assessment of Diabetes Risk (LEADR) study uses a novel Electronic Health Record (EHR) data approach as a tool to assess the epidemiology of known and new risk factors for type 2 diabetes mellitus (T2DM) and study how prevention interventions affect progression to and onset of T2DM. We created an electronic cohort of 1.4 million patients having had at least 4 encounters with a healthcare organization for at least 24-months; were aged ≥18 years in 2010; and had no diabetes (i.e., T1DM or T2DM) at cohort entry or in the 12 months following entry. EHR data came from patients at nine healthcare organizations across the U.S. between January 1, 2010-December 31, 2016. RESULTS: Approximately 5.9% of the LEADR cohort (82,922 patients) developed T2DM, providing opportunities to explore longitudinal clinical care, medication use, risk factor trajectories, and diagnoses for these patients, compared with patients similarly matched prior to disease onset. CONCLUSIONS: LEADR represents one of the largest EHR databases to have repurposed EHR data to examine patients' T2DM risk. This paper is first in a series demonstrating this novel approach to studying T2DM. IMPLICATIONS: Chronic conditions that often take years to develop can be studied efficiently using EHR data in a retrospective design. LEVEL OF EVIDENCE: While much is already known about T2DM risk, this EHR's cohort's 160 M data points for 1.4 M people over six years, provides opportunities to investigate new unique risk factors and evaluate research hypotheses where results could modify public health practice for preventing T2DM.

Effect of 2 Clinical Decision Support Strategies on Chronic Kidney Disease Outcomes in Primary Care: A Cluster Randomized Trial.

Importance: Information is needed about optimal strategies to improve evidence-based treatment of chronic kidney disease (CKD) in primary care. Objective: To determine whether a multimodal intervention delays annualized loss of estimated glomerular filtration rate (eGFR) in stages 3 and 4 CKD. Design, Setting, and Participants: This pragmatic cluster randomized clinical trial enrolled 42 primary care practices located in nonhospital settings with electronic health record systems. Practices were recruited through the American Academy of Family Physicians National Research Network. The study was conducted January 2013 through January 2016. Interventions: Practices were randomized at the organization level to either the clinical decision support (CDS) plus practice facilitation (PF) group (n = 25) or CDS group (n = 17) using covariate constrained randomization. Both groups received point-of-care CDS to prompt screening, diagnosis, and treatment of CKD; the intervention group also received PF based on the 9-point TRANSLATE model (target, use point-of-care reminder systems, get administrative buy-in, network information systems using registries, site coordination, local physician champion, audit and feedback, team approach, and education). Main Outcomes and Measures: The primary outcome measure was eGFR over time. Secondary outcome measures were systolic blood pressure over time, change in hemoglobin A1c (HbA1c) over time, avoidance of nonsteroidal anti-inflammatory medications, use of angiotensin converting enzyme inhibitor or angiotensin-renin blocker medication, early recognition and diagnosis of CKD, blood pressure control, and smoking cessation. Results: In this cluster randomized trial of 30 primary care practices comprising 6699 patients, there were 1685 patients in the control group (10 practices) and 5014 patients in the intervention group (20 practices). The final sample of practices differed from the original set of randomized practices owing to dropout. Patients in the practices were similar at baseline for age (mean [SD], 71.3 [9.6] years), sex (2716 male [40.5%]), and eGFR. There was a significant difference in eGFR slopes for patients in the intervention vs control group practices. The mean (SE) annualized loss of eGFR was 0.95 (0.19) in the control group in propensity-adjusted longitudinal analyses and 0.01 (0.12) in the intervention group (mean [SE] difference in slopes, 0.93 [0.23]; P < .001). Among patients with HbA1c measures, slopes differed significantly for patients in intervention vs control practices, with a mean (SE) annualized increase of 0.14 (0.03) in HbA1c for patients in control practices and a mean (SE) decline of 0.009 (0.02) for patients in intervention practices. There was a significant difference in HbA1c slopes for patients in the intervention compared with control group practices (control vs intervention, -0.14; P < .001), but no difference in the other secondary outcomes. Conclusions and Relevance: A multimodal intervention in primary care, based on the TRANSLATE model, slowed annualized loss of eGFR. This study had several important strengths, weaknesses, and lessons learned regarding the implementation of pragmatic interventions in primary care to improve CKD outcomes. Trial Registration: ClinicalTrials.gov Identifier: NCT01767883.

Federated queries for comparative effectiveness research: performance analysis.

This paper presents a study of the performance of federated queries implemented in a system that simulates the architecture proposed for the Scalable Architecture for Federated Translational Inquiries Network (SAFTINet). Performance tests were conducted using both physical hardware and virtual machines within the test laboratory of the Center for High Performance Computing at the University of Utah. Tests were performed on SAFTINet networks ranging from 4 to 32 nodes with databases containing synthetic data for several million patients. The results show that the caGrid FQE (Federated Query Engine) is capable and suitable for comparative effectiveness research (CER) federated queries given its nearly linear scalability as partner nodes increase in number. The results presented here are also important for the specification of the hardware required to run a CER grid.