Agreement and validity of electronic health record prescribing data relative to pharmacy claims data: A validation study from a US electronic health record database

Diabetes status and other factors as correlates of risk for thrombotic and thromboembolic events during SARS-CoV-2 infection: A nationwide retrospective case-control study using Cerner Real-World Data™

OBJECTIVES

We sought to examine in individuals with SARS-CoV-2 infection whether risk for thrombotic and thromboembolic events (TTE) is modified by presence of a diabetes diagnosis. Furthermore, we analysed whether differential risk for TTEs exists in type 1 diabetes mellitus (T1DM) versus type 2 diabetes mellitus (T2DM).

DESIGN

Retrospective case-control study.

SETTING

The December 2020 version of the Cerner Real-World Data COVID-19 database is a deidentified, nationwide database containing electronic medical record (EMR) data from 87 US-based health systems.

PARTICIPANTS

We analysed EMR data for 322 482 patients >17 years old with suspected or confirmed SARS-CoV-2 infection who received care between December 2019 and mid-September 2020. Of these, 2750 had T1DM; 57 811 had T2DM; and 261 921 did not have diabetes.

OUTCOME

TTE, defined as presence of a diagnosis code for myocardial infarction, thrombotic stroke, pulmonary embolism, deep vein thrombosis or other TTE.

RESULTS

Odds of TTE were substantially higher in patients with T1DM (adjusted OR (AOR) 2.23 (1.93-2.59)) and T2DM (AOR 1.52 (1.46-1.58)) versus no diabetes. Among patients with diabetes, odds of TTE were lower in T2DM versus T1DM (AOR 0.84 (0.72-0.98)).

CONCLUSIONS

Risk of TTE during COVID-19 illness is substantially higher in patients with diabetes. Further, risk for TTEs is higher in those with T1DM versus T2DM. Confirmation of increased diabetes-associated clotting risk in future studies may warrant incorporation of diabetes status into SARS-CoV-2 infection treatment algorithms.

Concordance of opioid exposure in all-payer claims databases with prescription drug monitoring program database using Arkansas as a case example

OBJECTIVE

To assess the concordance between and benefit of adding prescription drug monitoring program (PDMP) data to all-payer claims database (APCD) data for identifying and classifying opioid exposure among insured individuals.

DATA SOURCES AND STUDY SETTING

Arkansas APCD and PDMP.

STUDY DESIGN

Enrollees in APCD were classified as (1) true positives: if they received opioids in both databases, (2) false positives: if they only received opioids in APCD, (3) true negatives: if they had no opioid exposure in both databases, (4) false negatives: if they only received opioids in the PDMP database. Specificity, sensitivity, negative, and positive predictive values were calculated using PDMP as the "gold standard" database source. Subjects were also categorized as those who received any opioid, chronic opioid, high-dose opioid, or high-risk opioid therapies.

DATA COLLECTION/EXTRACTION METHODS

Arkansas residents continuously enrolled with pharmacy coverage in 2016 were included. APCD and PDMP were linked using an encrypted enrollee identifier, gender, and year of birth.

PRINCIPAL FINDINGS

The degree of concordance in opioid exposure between the two databases among 1,411,565 enrollees was high (sensitivity = 92.67%, specificity = 96.13%, positive predictive value = 91.60%, negative predictive value = 96.65%). Enrollees classified as having any opioid (APCD: 31.64% vs. PDMP: 31.26% vs. APCD+PDMP: 33.93%), chronic opioid (APCD: 7.81% vs. PDMP: 7.54% vs. APCD+PDMP: 8.24%), high-dose opioid (APCD: 10.60% vs. PDMP: 9.62% vs. APCD+PDMP: 11.33%), or high-risk opioid (APCD: 5.28% vs. PDMP: 5.33% vs. APCD+PDMP: 6.20%) therapies, were similar using only APCD versus PDMP versus the combined APCD and PDMP data sources.

CONCLUSIONS

Claims data sources, such as APCDs, are fairly accurate in identifying opioid exposure and the level of opioid exposure among persons with continuous pharmacy coverage.

Temporal trends in the prevalence and incidence of depression and the interplay of comorbidities in patients with young- and usual-onset type 2 diabetes from the USA and the UK

AIMS/HYPOTHESIS

We aimed to investigate the prevalence and incidence of depression, and the interplay of cardiometabolic comorbidities, in the differentiation of depression risk between young-onset diabetes (diagnosis at age <40 years) and usual-onset diabetes (diagnosis at age ≥40 years).

METHODS

Using electronic medical records from the UK and USA, retrospective cohorts of adults with incident type 2 diabetes diagnosed between 2006 and 2017 were examined. Trends in the prevalence and incidence of depression, and risk of developing depression, in participants with young-onset type 2 diabetes compared with usual-onset type 2 diabetes were assessed separately by sex and comorbidity status.

RESULTS

In total 230,932/1,143,122 people with type 2 diabetes from the UK/USA (mean age 58/60 years, proportion of men 57%/46%) were examined. The prevalence of depression in the UK/USA increased from 29% (95% CI 28, 30)/22% (95% CI 21, 23) in 2006 to 43% (95% CI 42, 44)/29% (95% CI 28, 29) in 2017, with the prevalence being similar across all age groups. A similar increasing trend was observed for incidence rates. In the UK, compared with people aged ≥50 years with or without comorbidity, 18-39-year-old men and women had 23-57% and 20-55% significantly higher risks of depression, respectively. In the USA, compared with those aged ≥60 years with or without comorbidity, 18-39-year-old men and women had 5-17% and 8-37% significantly higher risks of depression, respectively.

CONCLUSIONS/INTERPRETATION

Depression risk has been increasing in people with incident type 2 diabetes in the UK and USA, particularly among those with young-onset type 2 diabetes, irrespective of other comorbidities. This suggests that proactive mental health assessment from the time of type 2 diabetes diagnosis in primary care is essential for effective clinical management of people with type 2 diabetes.

Emulating the GRADE trial using real world data: retrospective comparative effectiveness study

OBJECTIVE

To emulate the GRADE (Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study) trial using real world data before its publication. GRADE directly compared second line glucose lowering drugs for their ability to lower glycated hemoglobin A_1c (HbA_1c).

DESIGN

Observational study.

SETTING

OptumLabs® Data Warehouse (OLDW), a nationwide claims database in the US, 25 January 2010 to 30 June 2019.

PARTICIPANTS

Adults with type 2 diabetes and HbA_1c 6.8-8.5% while using metformin monotherapy, identified according to the GRADE trial specifications, who also used glimepiride, liraglutide, sitagliptin, or insulin glargine.

MAIN OUTCOME MEASURES

The primary outcome was time to HbA_1c ≥7.0%. Secondary outcomes were time to HbA_1c >7.5%, incident microvascular complications, incident macrovascular complications, adverse events, all cause hospital admissions, and all cause mortality. Propensity scores were estimated using the gradient boosting machine method, and inverse propensity score weighting was used to emulate randomization of the treatment groups, which were then compared using Cox proportional hazards regression.

RESULTS

8252 people were identified (19.7% of adults starting the study drugs in OLDW) who met eligibility criteria for the GRADE trial (glimepiride arm=4318, liraglutide arm=690, sitagliptin arm=2993, glargine arm=251). The glargine arm was excluded from analyses owing to small sample size. Median times to HbA_1c ≥7.0% were 442 days (95% confidence interval 394 to 480 days) for glimepiride, 764 (741 to not calculable) days for liraglutide, and 427 (380 to 483) days for sitagliptin. Liraglutide was associated with lower risk of reaching HbA_1c ≥7.0% compared with glimepiride (hazard ratio 0.57, 95% confidence interval 0.43 to 0.75) and sitagliptin (0.55, 0.41 to 0.73). Results were consistent for the secondary outcome of time to HbA_1c >7.5%. No significant differences were observed among treatment groups for the remaining secondary outcomes.

CONCLUSIONS

In this emulation of the GRADE trial, liraglutide was statistically significantly more effective at maintaining glycemic control than glimepiride or sitagliptin when added to metformin monotherapy. Generating timely evidence on medical treatments using real world data as a complement to prospective trials is of value.

Evolution of Clinical Complexity, Treatment Burden, Health Care Use, and Diabetes-Related Outcomes Among Commercial and Medicare Advantage Plan Beneficiaries With Diabetes in the U.S., 2006-2018

OBJECTIVE

To characterize trends in clinical complexity, treatment burden, health care use, and diabetes-related outcomes among adults with diabetes.

RESEARCH DESIGN AND METHODS

We used a nationwide claims database to identify enrollees in commercial and Medicare Advantage plans who met claims criteria for diabetes between 1 January 2006 and 31 March 2019 and to quantify annual trends in clinical complexity (e.g., active health conditions), treatment burden (e.g., medications), health care use (e.g., ambulatory, emergency department [ED], and hospital visits), and diabetes-related outcomes (e.g., hemoglobin A1c [HbA1c] levels) between 2006 and 2018.

RESULTS

Among 1,470,799 commercially insured patients, the proportion with ≥10 active health conditions increased from 33.3% (95% CI 33.1-33.4) in 2006 to 38.9% (38.8-39.1) in 2018 (P = 0.001) and the proportion taking three or more glucose-lowering medications increased from 11.6% (11.5-11.7) to 23.1% (22.9-23.2) (P = 0.007). The proportion with HbA1c ≥8.0% (≥64 mmol/mol) increased from 28.0% (27.7-28.3) in 2006 to 30.5% (30.2-30.7) in 2015, decreasing to 27.8% (27.5-28.0) in 2018 (overall trend P = 0.04). Number of ambulatory visits per patient per year decreased from 6.86 (6.84-6.88) to 6.19 (6.17-6.21), (P = 0.001) while ED visits increased from 0.26 (0.257-0.263) to 0.29 (0.287-0.293) (P = 0.001). Among 1,311,903 Medicare Advantage enrollees, the proportion with ≥10 active conditions increased from 51.6% (51.2-52.0) to 65.1% (65.0-65.2) (P < 0.001); the proportion taking three or more glucose-lowering medications was stable at 16.6% (16.3-16.9) and 18.1% (18.0-18.2) (P = 0.98), and the proportion with HbA1c ≥8.0% increased from 17.4% (16.7-18.1) to 18.6% (18.4-18.7) (P = 0.008). Ambulatory visits per patient per year remained stable at 8.01 (7.96-8.06) and 8.17 (8.16-8.19) (P = 0.23), but ED visits increased from 0.41 (0.40-0.42) to 0.66 (0.66-0.66) (P < 0.001).

CONCLUSIONS

Among patients with diabetes, clinical complexity and treatment burden have increased over time. ED utilization has also increased, and patients may be using ED services for low-acuity conditions.

Medicare Claim-Based National Institutes of Health Stroke Scale to Predict 30-Day Mortality and Hospital Readmission

BACKGROUND

The Centers for Medicare and Medicaid Services (CMS) penalizes hospitals for higher than expected 30-day mortality rates using methods without accounting for condition severity risk adjustment. For patients with stroke, CMS claims did not quantify stroke severity until recently, when the National Institutes of Health Stroke Scale (NIHSS) reporting began.

OBJECTIVE

Examine the predictive ability of claim-based NIHSS to predict 30-day mortality and 30-day hospital readmission in patients with ischemic stroke.

DESIGN

Retrospective cohort study of Medicare claims data.

PATIENTS

Medicare beneficiaries with ischemic stroke (N=43,241) acute hospitalization between October 2016 and November 2017.

MEASUREMENTS

All-cause 30-day mortality and 30-day hospital readmission. NIHSS score was derived from ICD-10 codes and stratified into the following: minor to moderate, moderate, moderate to severe, and severe categories.

RESULTS

Among 43,241 patients with ischemic stroke with NIHSS from 2,659 US hospitals, 64.6% had minor to moderate stroke, 14.3% had moderate, 12.7% had moderate to severe, and 8.5% had a severe stroke,10.1% died within 30 days, 12.1% were readmitted within 30 days. The NIHSS exhibited stronger discriminant property (C-statistic 0.83, 95% CI: 0.82-0.84) for 30-day mortality compared to Elixhauser (0.74, 95% CI: 0.73-0.75). A monotonic increase in the adjusted 30-day mortality risk occurred relative to minor to moderate stroke category: hazard ratio [HR]=2.92 (95% CI=2.59-3.29) for moderate stroke, HR=5.49 (95% CI=4.90-6.15) for moderate to severe stroke, and HR=7.82 (95% CI=6.95-8.80) for severe stroke. After accounting for competing risk of mortality, there was a significantly higher readmission risk in the moderate stroke (HR=1.11, 95% CI=1.03-1.20), but significantly lower readmission risk in the severe stroke (HR=0.84, 95% CI=0.74-0.95) categories.

LIMITATION

Timing of NIHSS reporting during hospitalization is unknown.

CONCLUSIONS

Medicare claim-based NIHSS is significantly associated with 30-day mortality in Medicare patients with ischemic stroke and significantly improves discriminant property relative to the Elixhauser comorbidity index.

Trend in Blood Pressure Control Post Antihypertensive Drug Initiation in the U.S

INTRODUCTION

The aim of this study is to evaluate the temporal trends in systolic blood pressure control over 18 months after blood pressure‒lowering drug initiation in the U.S.

METHODS

From U.S. nationally representative electronic health records, 1,036,775 adults initiating and continuing blood pressure‒lowering drugs for ≥18 months during 2006-2018 were identified (January 2021). Prevalence trends of cardiovascular disease, diabetes, and depression at blood pressure‒lowering drug initiation, blood pressure‒lowering drug therapy intensification over 18 months, and the adjusted probability of achieving systolic blood pressure control 6 months after baseline and sustaining the control for over 18 months were evaluated.

RESULTS

At blood pressure‒lowering drug initiation, the prevalence of diabetes and depression consistently increased during the study period across all age groups, particularly in those aged 18-49 years, whereas the prevalence of cardiovascular disease was stable. Adjusted probabilities of achieving sustainable systolic blood pressure control by age group were 0.62 (95% CI=0.61, 0.63) for ages 18-39 years, 0.55 (95% CI=0.55, 0.56) for ages 40-49 years, 0.50 (95% CI=0.49, 0.50) for ages 50-59 years, 0.43 (95% CI=0.42, 0.43) for ages 60-69 years, and 0.37 (95% CI=0.37, 0.38) for ages 70-80 years. Those with cardiovascular disease or cardiovascular disease and diabetes had approximately 20% lower adjusted probability of achieving systolic blood pressure control (31%/29%) than those without these conditions (52%, p<0.01). Those with depression had a 4% higher probability of systolic blood pressure control than those without the condition (49% vs 45%, p<0.01).

CONCLUSIONS

In the U.S., only 30%-50% of the population are achieving sustainable blood pressure control over 18 months after blood pressure‒lowering drug initiation, with no indication of improvement in control over the last decade.

Comparing and validating medication complexity from insurance claims against electronic health records

BACKGROUND: Patient effort to comply with complex medication instructions is known to be related to nonadherence and subsequent medical complications or health care costs. A widely used Medication Regimen Complexity Index (MRCI) has been used with electronic health records (EHRs) to identify patients who could benefit from pharmacist intervention. A similar claims-derived measure may be better suited for clinical decision support, since claims offer a more complete view of patient care and health utilization. OBJECTIVE: To define and validate a novel insurance claims-based medication complexity score (MCS) patterned after the widely used MRCI, derived from EHRs. METHODS: Insurance claims and EHR data were provided by HealthPartners (N = 54,988) (Bloomington, Minnesota) and The Johns Hopkins Health System (N = 28,589) (Baltimore, Maryland) for years 2013 and 2017, respectively. Yearly measures of medication complexity were developed for each patient and evaluated with one another using rank correlation within different clinical subgroupings. Indicators for the presence of individually complex prescriptions were also developed and assessed using exact agreement. Complexity measures were then correlated with select covariates to further validate the concordance between MCS and MRCI with respect to clinical metrics. These included demographic, comorbidity, and health care utilization markers. Prescribed medications in each system's EHR were coded using the previously validated MRCI weighting rules. Insurance claims for retail pharmacy medications were coded using our novel MCS, which closely followed MRCI scoring rules. RESULTS: EHR-based MRCI and claims-based MCS were significantly correlated with one another for most clinical subgroupings. Likewise, both measures were correlated with several covariates, including count of active medications and chronic conditions. The MCS was, in most cases, more associated with key health covariates than was MRCI, although both were consistently significant. We found that the highest correlation between MCS and MRCI is obtained with patients who have similar counts of pharmacy records between EHRs and claims (HealthPartners: P = 0.796; Johns Hopkins Health System: P = 0.779). CONCLUSIONS: The findings suggest good correspondence between MCS and MRCI and that claims data represent a useful resource for assessing medication complexity. Claims data also have major practical advantages, such as interoperability across health care systems, although they lack the detailed clinical context of EHRs. DISCLOSURES: The Johns Hopkins University holds the copyright to the Adjusted Clinical Groups (ACG) system and receives royalties from the global distribution of the ACG system. This revenue supports a portion of the authors' salary. No additional or external funding supported this work. The authors have no conflict of interest to disclose.

Validity of Methods to Identify Individuals With Lower Extremity Amputation Using Department of Veterans Affairs Electronic Medical Records

Objectives

To determine the positive predictive value (PPV) of algorithms to identify patients with major (at the ankle or more proximal) lower extremity amputation (LEA) using Department of Veterans Affairs electronic medical records (EMR) and to evaluate whether PPV varies by sex, age, and race.

Design

We conducted a validation study comparing EMR determined LEA status to self-reported LEA (criterion standard).

Setting

Veterans who receive care at the Department of Veterans Affairs.

Participants

We invited a national sample of patients (N=699) with at least 1 procedure or diagnosis code for major LEA to participate. We oversampled women, Black men, and men ≤40 years of age.

Interventions

Not applicable.

Main Outcome Measure

We calculated PPV estimates and false negative percentages for 7 algorithms using EMR LEA procedure and diagnosis codes relative to self-reported major LEA.

Results

A total of 466 veterans self-reported their LEA status (68%). PPVs for the 7 algorithms ranged from 89% to 100%. The algorithm that required a single diagnosis or procedure code had the lowest PPV (89%). The algorithm that required at least 1 procedure code had the highest PPV (100%) but also had the highest proportion of false negatives (66%). Algorithms that required at least 1 procedure code or 2 or more diagnosis codes 1 month to 1 year apart had high PPVs (98%-99%) but varied in terms of false negative percentages. PPV estimates were higher among men than women but did not differ meaningfully by age or race, after accounting for sex.

Conclusion

PPVs were higher if 1 procedure or at least 2 diagnosis codes were required; the difference between algorithms was marked by sex. Investigators should consider trade-offs between PPV and false negatives to identify patients with LEA using EMRs.

Risk of physical injuries in children and teens with ophthalmic diagnoses in the OptumLabs Data Warehouse

BACKGROUND

To evaluate how eye diseases affect the risk of injuries (fractures, musculoskeletal, head and eye injuries) in children and teens.

METHODS

Claims from the OptumLabs Data Warehouse, a longitudinal deidentified commercial insurance claims database, were analyzed. Eligible subjects were aged <19 years at the time of their eye disease diagnosis, enrolled in the health plan between 2007 and 2018, and had >1 visit for >1 significant eye disease (strabismus, amblyopia, nystagmus, structural diseases), based on ICD9/10 codes. Controls were children in the same database who had no eye disease codes reported. Demographics and injury claims (fractures, musculoskeletal injuries, head injuries and eye injuries) were compared.

RESULTS

The overall incidence of any subsequent physical injury (even into adulthood) was 29% in eye disease patients and 23% in controls (P < 0.001). After accounting for covariates, the hazard ratio for injury with any type of eye disease was 1.14 (95% CI, 1.13-1.15), 1.17 (95% CI, 1.16-1.18), 0.97 (95% CI, 0.96-0.98), and 1.63 (95% CI, 1.60-1.66) for musculoskeletal injuries, fractures, head injuries, and eye injuries, respectively. The eye disease with the highest adjusted injury risk was nystagmus (HR = 1.26; 95% CI, 1.23-1.28), followed by optic neuritis and pseudotumor cerebri (HR = 1.25). The eye diseases with the lowest risk included amblyopia, esotropia, and glaucoma (HR ≤ 1.06).

CONCLUSIONS

There was an increased long-term risk of physical injury among children and teens with eye disease. The clinical significance of these small differences is unclear.

Congruence of opioid prescriptions and dispensing using electronic records and claims data

OBJECTIVE

To quantify discrepancies between opioid prescribing and dispensing via the percentage of patients with Electronic Medical Record (EMR) prescriptions who subsequently filled the prescription within 90 days, defined as congruence, and compared opioid congruence with related medications.

DATA SOURCES

Deidentified data from the IBM MarketScan Explorys Claims-EMR Dataset.

STUDY DESIGN

In this retrospective, observational study, we examined congruence for commonly prescribed controlled substances-opioids, stimulants, and benzodiazepines. Congruence was stratified by age group and sex.

DATA COLLECTION/EXTRACTION METHODS

Continuously enrolled adults aged 18-64 years with an EMR encounter (excluding inpatient settings) and ≥ 1 prescription for selected classes between 1/1/2016 and 10/2/2017.

PRINCIPAL FINDINGS

During the study period, 1,353,478 adults had ≥1 EMR encounter. Patients with stimulants prescriptions had the highest congruence (83%) corresponding to 7151 claims for 8,635 EMR prescriptions, followed by opioids (66%; 62,766/95,690) and benzodiazepines (64%; 30,181/47,408). Chi-square testing showed congruence differed by age group within opioids (P < .0001) and benzodiazepines (P < .0001) and was higher among females within benzodiazepines (P < .0001).

CONCLUSIONS

These findings demonstrate that relying on claims data alone for opioid prescribing measures might underestimate actual prescribing magnitude by as much as one-third in these data. Combined EMR and claims data can help future research better understand characteristics associated with congruence or incongruence between prescribing and dispensing.

Defining the duration of the dispensation of oral anticoagulants in administrative healthcare databases

PURPOSE

In clinical practice, warfarin therapy requires frequent dose adjustments. In pharmacy claims, the days supplied value may not reflect the true duration of warfarin dispensation. This may affect the measures of association comparing the safety of direct oral anticoagulants (DOACs) versus warfarin.

METHODS

Using Quebec healthcare administrative databases, we formed a cohort of 55 230 patients newly treated with oral anticoagulants between 2010 and 2016. The duration of dispensations was defined using two approaches: the recorded days supplied value, and the longitudinal coverage approximation (data-driven) that may account for individual variation in drug usage patterns. Propensity scores adjusted Cox proportional hazards regression models were used to estimate the hazard ratio (HR) of major bleeding with dabigatran or rivaroxaban versus warfarin.

RESULTS

Using the days supplied, the mean (and standard deviation) dispensation durations for dabigatran, rivaroxaban, and warfarin were 19 (15), 19 (14), and 13 (12) days, respectively. Using the data-driven approach, the durations were 20 (16), 19 (15), and 15 (16) days, respectively. The choice of the approach had no impact on the HR estimates.

CONCLUSIONS

In our settings, the data-driven approach closely approximated the recorded days supplied value for the standard dose therapies such as dabigatran and rivaroxaban. For warfarin, the data-driven approach captured more variability in the duration of dispensations compared to the days supplied value, which may better reflect the true drug-taking behavior of warfarin. Both approaches may provide valid estimates when comparing the safety of DOACs versus warfarin.

The need to shift pharmacogenetic research from candidate gene to genome-wide association studies

The primary research approach in pharmacogenetics has been candidate gene association studies (CGAS), but pharmacogenomic genome-wide association studies (GWAS) are becoming more common. We are now at a critical juncture when the results of those two research approaches, CGAS and GWAS, can be compared in pharmacogenetics. We analyzed publicly available databases of pharmacogenetic CGAS and GWAS (i.e., the Pharmacogenomics Knowledgebase [PharmGKB^®] and the NHGRI-EBI GWAS catalog) and the vast majority of variants (98%) and genes (94%) discovered in pharmacogenomic GWAS were novel (i.e., not previously studied CGAS). Therefore, pharmacogenetic researchers are not selecting the right candidate genes in the vast majority of CGAS, highlighting a need to shift pharmacogenetic research efforts from CGAS to GWAS.

Comparison of measures of medication adherence from pharmacy dispensing and insurer claims data

OBJECTIVE

Medication nonadherence is linked to worsened clinical outcomes and increased costs. Existing system-level adherence interventions rely on insurer claims for patient identification and outcome measurement, yet suffer from incomplete capture and lags in data acquisition. Data from pharmacies regarding prescription filling, captured in retail dispensing, may be more efficient.

DATA SOURCES

Pharmacy fill and insurer claims data.

STUDY DESIGN

We compared adherence measured using pharmacy fill data to adherence using insurer claims data, expressed as proportion of days covered (PDC) over 12 months. Agreement was evaluated using correlation/validation metrics. We also explored the relationship between adherence in both sources and disease control using prediction modeling.

DATA EXTRACTION METHODS

Large pragmatic trial of cardiometabolic disease in an integrated delivery network.

PRINCIPAL FINDINGS

Among 1113 patients, adherence was higher in pharmacy fill (mean = 50.0%) versus claims data (mean = 47.4%), although they had moderately high correlation (R = 0.57, 95% CI: 0.53-0.61) with most patients (86.9%) being similarly classified as adherent or nonadherent. Sensitivity and specificity of pharmacy fill versus claims data were high (0.89, 95% CI: 0.86-0.91 and 0.80, 95% CI: 0.75-0.85). Pharmacy fill-based PDC predicted better disease control slightly more than claims-based PDC, although the difference was nonsignificant.

CONCLUSIONS

Pharmacy fill data may be an alternative to insurer claims for adherence measurement.

Therapeutic inertia in the management of dyslipidaemia and hypertension in incident type 2 diabetes and the resulting risk factor burden: Real-world evidence from primary care

OBJECTIVE

To investigate trends in the prevalence of hypertension and dyslipidaemia in incident type 2 diabetes (T2DM), time to antihypertensive (AHT) and lipid-lowering therapy (LLT), and the association with systolic blood pressure (SBP) and lipid control.

RESEARCH DESIGN AND METHODS

Using The Health Improvement Network UK primary care database, 254 925 people with incident T2DM and existing dyslipidaemia or hypertension were identified. Among those without atherosclerotic cardiovascular disease (ASCVD) history and not on AHT or LLT at diagnosis, the adjusted median months to initiating an AHT or an LLT, and the probabilities of high SBP or lipid levels over 2 years in people initiating therapy within or after 1 year were evaluated according to high and low ASCVD risk status.

RESULTS

At diabetes diagnosis, 66% and 66% had dyslipidaemia and hypertension, respectively. During 2005 to 2016, dyslipidaemia prevalence increased by 10% in people aged <60 years, while hypertension prevalence remained stable in all age groups. Among those with high ASCVD risk status in the age groups 18 to 39, 40 to 49, and 50 to 59 years, the median number of months to initiation of therapy were 20.4 (95% confidence interval [CI] 20.3-20.5), 10.9 (95% CI 10.8-11.0), and 9.5 (95% CI 9.4-9.6) in the dyslipidaemia subcohort, and 28.1 (95% CI 28.0-28.2), 19.2 (95% CI 19.1-19.3), and 19.9 (95% CI 19.8-20.0) in the hypertension subcohort. Among people with high and low ASCVD risk status, respectively, compared to early LLT initiators, those who initiated LLT after 1 year had a 65.3% to 85.3% and a 65.0% to 85.3% significantly higher probability of failing lipid control at 2 years of follow-up, while late AHT initiators had a 46.5% to 57.9% and a 40.0% to 58.7% significantly higher probability of failing SBP control.

CONCLUSIONS

Significant delay in initiating cardioprotective therapies was observed, and time to first prescription was similar in the primary prevention setting, irrespective of ASCVD risk status across all T2DM diagnosis age groups, resulting in poor risk factor control at 2 years of follow-up.

Association of biological antirheumatic therapy with risk for type 2 diabetes: a retrospective cohort study in incident rheumatoid arthritis

OBJECTIVE

To explore possible associations of treatment with biological disease-modifying antirheumatic drugs (bDMARDs), including T-cell-based and interleukin-6 inhibition (IL-6i)-based therapies, and the risk for type 2 diabetes mellitus (T2DM) in patients with rheumatoid arthritis (RA).

STUDY DESIGN, SETTING AND PARTICIPANTS

Five treatment groups were selected from a United States Electronic Medical Records database of 283 756 patients with RA (mean follow-up, 5 years): never received bDMARD (No bDMARD, n=125 337), tumour necrosis factor inhibitors (TNFi, n=34 873), IL-6i (n=1884), T-cell inhibitors (n=5935) and IL-6i+T cell inhibitor abatacept (n=1213). Probability and risk for T2DM were estimated with adjustment for relevant confounders.

RESULTS

In the cohort of 169 242 patients with a mean 4.5 years of follow-up and a mean 641 200 person years of follow-up, the adjusted probability of developing T2DM was significantly lower in the IL-6i (probability, 1%; 95% CI 0.6 to 2.0), T-cell inhibitor (probability, 3%; 95% CI 2.3 to 3.3) and IL-6i+T cell inhibitor (probability, 2%; 95% CI 0.1 to 2.9) groups than in the No bDMARD (probability, 5%; 95% CI 4.6 to 4.9) and TNFi (probability, 4%; 95% CI 3.7 to 4.7) groups. Compared with No bDMARD, the IL-6i and IL-6i+T cell inhibitor groups had 37% (95% CI of HR 0.42 to 0.96) and 34% (95% CI of HR 0.46 to 0.93) significantly lower risk for T2DM, respectively; there was no significant difference in risk in the TNFi (HR 0.99; 95% CI 0.93 to 1.06) and T-cell inhibitor (HR 0.96; 95% CI 0.82 to 1.12) groups.

CONCLUSIONS

Treatment with IL-6i, with or without T-cell inhibitors, was associated with reduced risk for T2DM compared with TNFi or No bDMARDs; a less pronounced association was observed for the T-cell inhibitor abatacept.

Agreement in Medications Reported in Medical and Dental Electronic Health Records

INTRODUCTION

The historical separation between medicine and dentistry has resulted in the creation of separate health records, which have the potential to negatively impact patient care and safety. Of particular importance, errors or omissions in medication lists in separate electronic health records (EHRs) may lead to medical errors and serious adverse outcomes.

OBJECTIVE

This study aimed to compare medication lists reported in the EHRs of active patients treated by both the University of Michigan School of Dentistry and Michigan Medicine to determine if differences exist.

METHODS

In this cohort study, EHRs of a population of 159,733 patients that the University of Michigan medical and dental clinics share in common were investigated for agreement in the reporting of 16 medications. After exclusion of minors and patients not seen in the last 5 y, records of 27,277 patients were examined.

RESULTS

The maximum percentage of agreement in medications reported in both records was 52% for levothyroxine, and the minimum was 7% for sildenafil. The medical record had a significantly higher number of unique medications than the dental record, suggesting higher underreporting in the dental setting.

CONCLUSION

The lack of agreement in the report of medications with serious dental and medical implications argues in favor of unification of records and use of available technology to increase accurate medication reporting.

KNOWLEDGE TRANSFER STATEMENT

The results demonstrate a lack of agreement between medications reported in medical and dental records, which can have serious implications to patients' health. A unified health record, employing available technology to increase accurate medication reporting, would mitigate this problem.

Selective serotonin reuptake inhibitor use patterns among commercially insured US pregnancies (2005-2014)

The goal of this study was to describe patterns of selective serotonin reuptake inhibitor (SSRI) use during pregnancy in a US cohort (2005-2014) of > 1 million commercially insured women using administrative claims. We used international classification of disease (ICD-9) diagnosis and procedure and current procedural terminology codes in the OptumLabs® Data Warehouse to identify deliveries (including losses) among US women aged 15-45 (n = 1,061,023). SSRI dispensings that overlapped with the timing of pregnancy were identified using national drug codes in linked pharmacy claims. Demographic characteristics were imputed based on residential location, census data, and consumer information. We investigated patterns by trimester, agent, and demographic subgroups. A total of 46,087 of women (4.34%) were dispensed SSRIs during the estimated pregnancy period. Sertraline was the most common overall and had the highest initial use after trimester 1, including women who switched from another SSRI, although dispensing for > 1 SSRI during pregnancy was uncommon. Use of vilazodone was rare and had the highest discontinuation after trimester 1, followed by paroxetine. SSRI use was more common among women who were older, White, college-educated, higher income (≥ $100,000), or resided in the Midwest. Paroxetine and dispensings for > 1 SSRI were more common in lower education subgroups. White women had the highest proportion of use in all trimesters of pregnancy, whereas Hispanic women had the lowest. Among commercially insured US women, SSRI use during pregnancy differed by agent and demographics. More research is needed to understand whether these differences are due to symptom reporting, cultural beliefs, and/or physician preferences.

Association of Cardiometabolic Multimorbidity and Depression With Cardiovascular Events in Early-Onset Adult Type 2 Diabetes: A Multiethnic Study in the U.S

OBJECTIVE

To evaluate the temporal patterns of cardiometabolic multimorbidity (CM) and depression in White Caucasians (WCs) and African Americans (AAs) with early-onset type 2 diabetes and their impact on long-term atherosclerotic cardiovascular disease (ASCVD).

RESEARCH DESIGN AND METHODS

From U.S. electronic medical records, 101,104 AA and 505,336 WC subjects with type 2 diabetes diagnosed between 2000 and 2017 were identified (mean follow-up 5.3 years). Among those without ASCVD at diagnosis, risk of ASCVD and three-point major adverse cardiovascular events (MACE-3) (heart failure, myocardial infarction, or stroke) was evaluated between ethnicities by age-groups.

RESULTS

The proportion of patients diagnosed at <50 years of age increased during 2012-2017 (AA 34-38%, WC 26-29%). Depression prevalence increased during 2000-2017 (AA 15-23%, WC 20-34%), with an increasing trend for CM at diagnosis in both groups. Compared with WC, the adjusted MACE-3 risk was significantly higher in AA across all age-groups, more pronounced in the 18-39-year age-group (hazard ratio 95% CI 1.42, 1.88), and in patients with and without depression. AAs had a 17% (1.05, 1.31) significantly higher adjusted ASCVD risk in the 18-39-year age-group only. Depression was independently associated with ASCVD and MACE-3 risk in both ethnic groups across all age-groups. Other comorbidities were independently associated with ASCVD and MACE-3 risk only among WCs.

CONCLUSIONS

AAs have higher cardiovascular risk compared with WCs, particularly in early-onset type 2 diabetes. CM and depression at diabetes diagnosis have been increasing over the past two decades in both ethnic groups. Strategies for screening and optimal management of CM and depression, particularly in early-onset type 2 diabetes, may result in a lower cardiovascular risk.

Cardiometabolic risk factor control in black and white people in the United States initiating sodium-glucose co-transporter-2 inhibitors: A real-world study

AIMS

To explore cardiometabolic risk profiles, the probability of sustainable control, and the effectiveness of treatment with sodium-glucose co-transporter-2 (SGLT2) inhibitors in black and white adults in the United States with type 2 diabetes.

MATERIALS AND METHODS

Using nationally representative US electronic medical records, 72 690 white and 10 004 black adults diagnosed with type 2 diabetes initiating SGLT2 inhibitors during the period 2013 to 2018, continuing it for ≥6 months, and with follow-up of ≥12 months, were identified. Glycated haemoglobin (HbA1c), body weight, systolic blood pressure (SBP) and lipid changes at 6 months, and sustainability of control over 18 months post SGLT2 inhibitor initiation were explored, separately in those with and without atherosclerotic cardiovascular disease (ASCVD).

RESULTS

The white group was older (58 years) with lower mean HbA1c (8.5%), compared to the black group (age 54 years, HbA1c 9.0%). Body mass index distribution was similar. The proportions of people with uncontrolled SBP, LDL cholesterol, non-HDL cholesterol and triglyceride levels were 24%, 42%, 51% and 62%, respectively, in white patients, and 31%, 51%, 49% and 32%, respectively, in black patients. At 6-month follow-up white and black patients had similar adjusted reductions in HbA1c (1.1%), SBP (8-10 mmHg), LDL cholesterol (0.26 - 0.34 mmol / L) and body weight (1.1-1.4 kg). However, over 18 months' follow-up, compared to white patients, black patients were significantly less likely to achieve sustainable control in HbA1c (odds ratio [OR] 0.67, 95% confidence interval [CI] 0.63-0.72), body weight (OR 0.81, 95% CI 0.72-0.91), SBP (OR 0.67, 95% CI 0.61-0.74) and LDL cholesterol (OR 0.77, 95% CI 0.67-0.89). Triglyceride control was significantly better among black patients. Black patients had a significantly higher risk factor burden, irrespective of ASCVD status.

CONCLUSIONS

While the effectiveness of SGLT2 inhibitors was similar among black and white patients, irrespective of ASCVD status, black patients continued to have worse cardiometabolic risk factor burden after SGLT2 inhibitor initiation.

Validation of drug prescription records for senior patients in Alberta's Tomorrow Project: Assessing agreement between two population-level administrative pharmaceutical databases in Alberta, Canada

PURPOSE

To assess agreement between the Pharmaceutical Information Network (PIN), a newly implemented medication data repository in Alberta, Canada, and the Alberta Blue Cross (ABC) database, a long established database with medication records of all senior patients in Alberta.

METHODS

PIN data (2008-2015) were cross-validated with ABC medication records for senior participants (older than 65 years old) in Alberta's Tomorrow Project (ATP), a longitudinal cohort study in Alberta. The completeness and accuracy of PIN were respectively calculated as the percentage of ABC records coexisting (concordant) in PIN and the percentage of concordant records having mutually agreeable information on drug quantity. Generalized linear models were used to examine potential association of PIN completeness and accuracy with sociodemographic factors.

RESULTS

A total of 1 218 191 drug prescription records from 13 143 ATP participants were captured by PIN and ABC in 2008-2015, among which 91.6% were from PIN, 82.5% from ABC, and 74.2% coexisted in PIN and ABC. The overall completeness of PIN in capturing ABC medication records was 89.9%, with small variations (less than ±5%) across types of drugs. The completeness of PIN was improved on average by 1.3% annually over time (P < .001). PIN had 100% accuracy as defined by drug quantity data agreeable with ABC records. No significant associations were observed with age, sex, ethnicity, rural/urban areas, and socioeconomic status of the participants.

CONCLUSIONS

Cross-validated with the ABC dataset, our study showed that irrespective of drug type, PIN has a fairly good completeness (approximately 90%) and accuracy (100%) in capturing the ABC claimed medications for senior patients in Alberta.

The associations between self-reported depression, self-reported chronic inflammatory conditions and cognitive abilities in UK Biobank

Background

Depression and chronic inflammatory medical conditions have been linked to impaired cognitive ability. However despite frequent comorbidity, their combined association with cognitive ability has rarely been examined.

Methods

This study examined associations between self-reported depression and chronic inflammatory diseases and their interaction with cognitive performance in 456,748 participants of the UK Biobank, adjusting for sociodemographic and lifestyle factors. Numbers with available data ranged from 94,899 to 453,208 depending on the cognitive test.

Results

Self-reported depression was associated with poorer performance compared to controls in several cognitive tests (fully adjusted models, reaction time: B = 6.08, 95% CI = 5.09, 7.07; pairs matching: incidence rate ratio = 1.02, 95% CI = 1.02, 1.03; Trail Making Test B: B = 1.37, 95% CI = 0.88, 1.87; Digit Symbol Substitution Test (DSST): B = −0.35, 95% CI = −0.44, −0.27). Self-reported chronic inflammatory conditions were associated with slower reaction time (B = 3.79, 95% CI = 2.81, 4.78) and lower DSST scores (B = −0.21, 95% CI = −0.30, −0.13). No interaction effects were observed.

Discussion

In this large, population-based study we provide evidence of lower cognitive performance in both depression and a comprehensive category of chronic inflammatory conditions. Results are consistent with additive effects of both types of disorder on cognitive ability. Clinicians should be aware of such effects, particularly as cognitive impairment is linked to poorer disease outcomes and quality of life.

Measurement Error and Misclassification in Electronic Medical Records: Methods to Mitigate Bias

PURPOSE OF REVIEW

We sought to: 1) examine common sources of measurement error in research using data from electronic medical records (EMR), 2) discuss methods to assess the extent and type of measurement error, and 3) describe recent developments in methods to address this source of bias.

RECENT FINDINGS

We identified eight sources of measurement error frequently encountered in EMR studies, the most prominent being that EMR data usually reflect only the health services and medications delivered within the specific health facility/system contributing to the EMR data. Methods for assessing measurement error in EMR data usually require gold standard or validation data, which may be possible using data linkage. Recent methodological developments to address the impact of measurement error in EMR analyses were particularly rich in the multiple imputation literature.

SUMMARY

Presently, sources of measurement error impacting EMR studies are still being elucidated, as are methods for assessing and addressing them. Given the magnitude of measurement error that has been reported, investigators are urged to carefully evaluate and rigorously address this potential source of bias in studies based in EMR data.

Identifying Asthma Exacerbation-Related Emergency Department Visit Using Electronic Medical Record and Claims Data

Background  Asthma exacerbation leading to emergency department (ED) visit is prevalent, an indicator of poor control of asthma, and is a potentially preventable clinical outcome.

Objective  We propose to utilize multiple data elements available in electronic medical records (EMRs) and claims database to create separate algorithms with high validity for clinical and research purposes to identify asthma exacerbation-related ED visit among the general population.

Methods  We performed a retrospective study with inclusion criteria of patients aged 4 to 40 years, a visit to Geisinger ED from January 1, 2006, to October 28, 2013, with asthma on their problem list. Different electronic data elements including chief complaints, vitals, season, smoking, medication use, and discharge diagnoses were obtained to create the algorithm. A stratified random sample was generated to select the charts for review. Chart review was performed to classify patients with asthma-related ED visit, that is, the gold standard. Two reviewers performed the chart review and validation was done on a small subset.

Results  There were 966 eligible ED visits in the EMR sample and 731 in the claims sample. Agreement between reviewers was 95.45% and kappa statistic was 0.91. Mean age of the EMR sample was 22 years, and mostly white (93%). Multiple models conventionally used in studies were evaluated and the final model chosen included principal diagnosis, bronchodilator, and steroid use for both algorithms, chief complaints for EMR, and secondary diagnosis for claims. Area under the curve was 0.93 (95% confidence interval: 0.91–0.94) and 0.94 (0.93–0.96), respectively, for EMR and claims data, with positive predictive value of > 94%. The algorithms are visually presented using nomograms.

Conclusion  We were able to develop two separate algorithms for EMR and claims to identify asthma exacerbation-related ED visit with excellent diagnostic ability and varying discrimination threshold for clinical and research purposes.