An observational cohort analysis was conducted within the Surveillance, Prevention, and Management of Diabetes Mellitus (SUPREME-DM) DataLink, a consortium of 11 integrated health-care delivery systems with electronic health records in 10 US states. Among nearly 7 million adults aged 20 years or older, we estimated annual diabetes incidence per 1,000 persons overall and by age, sex, race/ethnicity, and body mass index. We identified 289,050 incident cases of diabetes. Age- and sex-adjusted population incidence was stable between 2006 and 2010, ranging from 10.3 per 1,000 adults (95% confidence interval (CI): 9.8, 10.7) to 11.3 per 1,000 adults (95% CI: 11.0, 11.7). Adjusted incidence was significantly higher in 2011 (11.5, 95% CI: 10.9, 12.0) than in the 2 years with the lowest incidence. A similar pattern was observed in most prespecified subgroups, but only the differences for persons who were not white were significant. In 2006, 56% of incident cases had a glycated hemoglobin (hemoglobin A1c) test as one of the pair of events identifying diabetes. By 2011, that number was 74%. In conclusion, overall diabetes incidence in this population did not significantly increase between 2006 and 2010, but increases in hemoglobin A1c testing may have contributed to rising diabetes incidence among nonwhites in 2011.
diabetes mellitus; glycated hemoglobin; hemoglobin A1c; incidence; trends
Medication nonadherence is a major obstacle to better control of glucose, blood pressure (BP), and LDL cholesterol in adults with diabetes. Inexpensive effective strategies to increase medication adherence are needed.
RESEARCH DESIGN AND METHODS
In a pragmatic randomized trial, we randomly assigned 2,378 adults with diabetes mellitus who had recently been prescribed a new class of medication for treating elevated levels of glycated hemoglobin (A1C) ≥8% (64 mmol/mol), BP ≥140/90 mmHg, or LDL cholesterol ≥100 mg/dL, to receive 1) one scripted telephone call from a diabetes educator or clinical pharmacist to identify and address nonadherence to the new medication or 2) usual care. Hierarchical linear and logistic regression models were used to assess the impact on 1) the first medication fill within 60 days of the prescription; 2) two or more medication fills within 180 days of the prescription; and 3) clinically significant improvement in levels of A1C, BP, or LDL cholesterol.
Of the 2,378 subjects, 89.3% in the intervention group and 87.4% in the usual-care group had sufficient data to analyze study outcomes. In intent-to-treat analyses, intervention was not associated with significant improvement in primary adherence, medication persistence, or intermediate outcomes of care. Results were similar across subgroups of patients defined by age, sex, race/ethnicity, and study site, and when limiting the analysis to those who completed the intended intervention.
This low-intensity intervention did not significantly improve medication adherence or control of glucose, BP, or LDL cholesterol. Wide use of this strategy does not appear to be warranted; alternative approaches to identify and improve medication adherence and persistence are needed.
The Centers for Medicare and Medicaid Services (CMS) recently added medication adherence to antihypertensives, antihyperlipidemics, and oral antihyperglycemics to its Medicare STAR quality measures. These CMS metrics exclude patients with <2 medication fills (i.e. “early non-adherence”) and patients concurrently taking insulin. This study examined the proportion of diabetes patients prescribed cardiovascular disease (CVD) medications excluded from STAR adherence metrics, and assessed the relationship of both STAR-defined adherence and exclusion from STAR metrics with CVD risk factor control.
Cross-sectional, population-based analysis of 129,040 diabetics ≥65 in 2010 from three Kaiser Permanente regions.
We estimated adjusted risk ratios to assess the relationship between achieving STAR adherence, and exclusion from STAR adherence metrics, with CVD risk factor control(A1c<8.0%, LDL-C<100mg/dL, systolic blood pressure (SBP)<130mmHg) in diabetics.
STAR metrics excluded 27% of diabetes patients prescribed oral medications. STAR-defined non-adherence was negatively associated with CVD risk factor control (RR=0.95, 0.84, 0.96 for A1c, LDL-C, and SBP control; p<0.001). Exclusion from STAR metrics due to early non-adherence was also strongly associated with poor control (RR=0.83, 0.56, 0.87 for A1c, LDL-C, and SBP control; p<0.001). Exclusion for insulin use was negatively associated with A1c control (RR=0.78; p<.0001).
Medicare STAR adherence measures underestimate the prevalence of medication non-adherence in diabetes, and exclude patients at high risk for poor CVD outcomes. Up to 3 million elderly diabetes patients may be excluded from these measures nationally. Quality measures designed to encourage effective medication use should focus on all patients treated for CVD risk.
Background: To describe trends in labor induction, including elective induction, from 2001 to 2007 for six U.S. health plans and to examine the validity of induction measures derived from birth certificate and health plan data.
Methods: This retrospective cohort study included 339,123 deliveries at 35 weeks' gestation or greater. Linked health plan and birth certificate data provided information about induction, maternal medical conditions, and pregnancy complications. Induction was defined from diagnosis and procedure codes and birth certificate data and considered elective if no accepted indication was coded. We calculated induction prevalence across health plans and years. At four health plans, we reviewed medical records to validate induction measures.
Results: Based on electronic data, induction prevalence rose from 28% in 2001 to 32% in 2005, then declined to 29% in 2007. The trend was driven by changes in the prevalence of apparent elective induction, which rose from 11% in 2001 to 14% in 2005 and then declined to 11% in 2007. The trend was similar for subgroups by parity and gestational age. Elective induction prevalence varied considerably across plans. On review of 86 records, 36% of apparent elective inductions identified from electronic data were confirmed as valid.
Conclusions: Elective induction appeared to peak in 2005 and then decline. The decrease may reflect quality improvement initiatives or changes in policies, patient or provider attitudes, or coding practices. The low validation rate for measures of elective induction defined from electronic data has important implications for existing quality measures and for research studies examining induction's outcomes.
Assessing the safety and effectiveness of medical products with observational electronic medical record data is challenging when the treatment is time-varying. The objective of this paper is to develop a Cox model stratified by event times with stabilized weights adjustment to examine the effect of time-varying treatment in observational studies.
Time-varying stabilized weights are calculated at unique event times and are used in a Cox model stratified by event times to estimate the effect of time-varying treatment. We applied this method in examining the effect of an anti-platelet agent, clopidogrel, on events, including bleeding, myocardial infarction (MI), and death after a Drug-Eluting Stent was implanted in coronary artery. Clopidogrel use may change over time based on patients' behavior (e.g., non-adherence) and physicians' recommendations (e.g., end of duration of therapy). We also compared the results to those from a Cox model for counting processes adjusting for all covariates used in creating stabilized weights.
We demonstrate that 1) results from the stratified Cox model without stabilized weights adjustment and the Cox model for counting processes without covariate adjustment are identical in analyzing the clopidogrel data; and 2) effects of clopidogrel on bleeding, MI and death are larger in the stratified Cox model with stabilized weights adjustment compared to those from the Cox model for counting processes with covariate adjustment.
The Cox model stratified by event times with time-varying stabilized weights adjustment is useful in estimating the effect of time-varying treatments in observational studies while balancing for known confounders.
Drug-Eluting Stent; Clopidogrel; Time-varying exposure; Cox model; Stabilized weights
To review the published, peer-reviewed literature on clinical research data warehouse governance in distributed research networks (DRNs).
Materials and methods
Medline, PubMed, EMBASE, CINAHL, and INSPEC were searched for relevant documents published through July 31, 2013 using a systematic approach. Only documents relating to DRNs in the USA were included. Documents were analyzed using a classification framework consisting of 10 facets to identify themes.
6641 documents were retrieved. After screening for duplicates and relevance, 38 were included in the final review. A peer-reviewed literature on data warehouse governance is emerging, but is still sparse. Peer-reviewed publications on UK research network governance were more prevalent, although not reviewed for this analysis. All 10 classification facets were used, with some documents falling into two or more classifications. No document addressed costs associated with governance.
Even though DRNs are emerging as vehicles for research and public health surveillance, understanding of DRN data governance policies and procedures is limited. This is expected to change as more DRN projects disseminate their governance approaches as publicly available toolkits and peer-reviewed publications.
While peer-reviewed, US-based DRN data warehouse governance publications have increased, DRN developers and administrators are encouraged to publish information about these programs.
Data governance; Data warehouse; Research networks; Clinical research; Data privacy
The Centers for Medicare and Medicaid Services provide significant incentives to health plans that score well on Medicare STAR metrics for cardiovascular disease risk factor medication adherence. Information on modifiable health system-level predictors of adherence can help clinicians and health plans develop strategies for improving Medicare STAR scores, and potentially improve cardiovascular disease outcomes.
To examine the association of Medicare STAR adherence metrics with system-level factors.
A cross-sectional study.
A total of 129,040 diabetes patients aged 65 years and above in 2010 from 3 Kaiser Permanente regions.
Adherence to antihypertensive, antihyperlipidemic, and oral antihyperglycemic medications in 2010, defined by Medicare STAR as the proportion of days covered ≥80%.
After controlling for individual-level factors, the strongest predictor of achieving STAR-defined medication adherence was a mean prescribed medication days’ supply of >90 days (RR=1.61 for antihypertensives, oral antihyperglycemics, and statins; all P<0.001). Using mail order pharmacy to fill medications >50% of the time was independently associated with better adherence with these medications (RR=1.07, 1.06, 1.07; P<0.001); mail order use had an increased positive association among black and Hispanic patients. Medication copayments ≤$10 for 30 days’ supply (RR=1.02, 1.02, 1.02; P<0.01) and annual individual out-of-pocket maximums ≤$2000 (RR=1.02, 1.01, 1.02; P<0.01) were also significantly associated with higher adherence for all 3 therapeutic groupings.
Greater medication days’ supply and mail order pharmacy use, and lower copayments and out-of-pocket maximums, are associated with better Medicare STAR adherence. Initiatives to improve adherence should focus on modifiable health system-level barriers to obtaining evidence-based medications.
adherence; quality measurement; quality improvement
Inverse probability of treatment weighting (IPTW) has been used in observational studies to reduce selection bias. To obtain estimates of the main effects, a pseudo data set is created by weighting each subject by IPTW and analyzed with conventional regression models. Currently variance estimation requires additional work depending on type of outcomes. Our goal is to demonstrate a statistical approach to directly obtain appropriate estimates of variance of the main effects in regression models.
We carried out theoretical and simulation studies to show that the variance of the main effects estimated directly from regressions using IPTW is underestimated, and that the type I error rate is higher due to the inflated sample size in the pseudo data. The robust variance estimator using IPTW often slightly overestimates the variance of the main effects. We propose to use the stabilized weights to directly estimate both the main effect and its variance from conventional regression models.
We applied the approach to a study examining the effectiveness of serum potassium monitoring in reducing hyperkalemia-associated adverse events among 27,355 diabetic patients newly-prescribed a renin-angiotensin-aldosterone system (RAAS) inhibitor. The incidence rate ratio (with monitoring versus without monitoring) and confidence intervals were 0.46 (0.34, 0.61) using the stabilized weights compared to 0.46 (0.38, 0.55) using typical inverse probability of treatment weighting.
Our theoretical, simulation results and real data example demonstrate that the use of the stabilized weights in the pseudo data preserves the sample size of the original data, produces appropriate estimation of the variance of main effect, and maintains an appropriate type I error rate.
Inverse probability of treatment weighting; stabilized weights; type I error rates; incidence rate ratio; confidence intervals
The goal of this study was to determine changes in antibiotic-dispensing rates among children in 3 health plans located in New England [A], the Mountain West [B], and the Midwest [C] regions of the United States.
Pharmacy and outpatient claims from September 2000 to August 2010 were used to calculate rates of antibiotic dispensing per person-year for children aged 3 months to 18 years. Differences in rates by year, diagnosis, and health plan were tested by using Poisson regression. The data were analyzed to determine whether there was a change in the rate of decline over time.
Antibiotic use in the 3- to <24-month age group varied at baseline according to health plan (A: 2.27, B: 1.40, C: 2.23 antibiotics per person-year; P < .001). The downward trend in antibiotic dispensing slowed, stabilized, or reversed during this 10-year period. In the 3- to <24-month age group, we observed 5.0%, 9.3%, and 7.2% annual declines early in the decade in the 3 plans, respectively. These dropped to 2.4%, 2.1%, and 0.5% annual declines by the end of the decade. Third-generation cephalosporin use for otitis media increased 1.6-, 15-, and 5.5-fold in plans A, B, and C in young children. Similar attenuation of decline in antibiotic use and increases in use of broad-spectrum agents were seen in other age groups.
Antibiotic dispensing for children may have reached a new plateau. Along with identifying best practices in low-prescribing areas, decreasing broad-spectrum use for particular conditions should be a continuing focus of intervention efforts.
antibiotics; otitis media; respiratory tract infections
To evaluate the prevalence, trends, timing and duration of exposure to antiviral medications during pregnancy within a US cohort of pregnant women and to evaluate the proportion of deliveries with a viral infection diagnosis among women given antiviral medication during pregnancy.
Live-born deliveries between 2001 and 2007, to women aged 15 to 45 years, were included from the Medication Exposure in Pregnancy Risk Evaluation Program (MEPREP), a collaborative research program between the U.S. Food and Drug Administration and eleven health plans. They were evaluated for prevalence, timing, duration, and temporal trends of exposure to antiviral medications during pregnancy. We also calculated the proportion of deliveries with a viral infection diagnosis among those exposed to antiviral medications.
Among 664,297 live births, the overall prevalence of antiviral exposure during pregnancy was 4% (n=25,155). Between 2001 and 2007, antiviral medication exposure during pregnancy doubled from 2.5% to 5%. The most commonly used antiviral medication was acyclovir, with 3% of the deliveries being exposed and most of the exposure occurring after the 1st trimester. Most deliveries exposed to antiviral medications were exposed for less than 30 days (2% of all live births). Forty percent of the women delivering an infant exposed to antiviral medications had a herpes diagnosis.
Our findings highlight the increased prevalence of women delivering an infant exposed to antiviral medications over time. These findings support the need for large, well-designed studies to assess the safety and effectiveness of these medications during pregnancy.
To understand the burden of medication use for newly-diagnosed diabetes patients both before and after diabetes diagnosis, and to identify subpopulations of newly-diagnosed diabetes patients who face a relatively high drug burden.
Eleven U.S. integrated health systems.
196,654 insured adults aged ≥20 diagnosed with newly-diagnosed diabetes from 1/1/2005 – 12/31/2009.
Main Outcome Measure
Number of unique therapeutic classes of drugs dispensed in the 12 months prior to, and 12 months post, the diagnosis of diabetes in 5 categories: overall, antihypertensive, antihyperlipidemic, mental health, and antihyperglycemic (post-period only).
The mean number of drug classes used by newly-diagnosed diabetes patients is high before diagnosis (5.0), and increases significantly afterwards (6.6, p<.001). Eighty-one percent of this increase is due to antihyperglycemic initiation and increased use of medications to control hypertension and lipid levels. Multivariate analyses showed that overall drug burden after diabetes diagnosis was higher in female, older, white, and obese patients, as well as among those with higher A1cs and comorbidity levels (p<.001 for all comparisons). The overall number of drug classes used by newly-diagnosed diabetes patients after diagnosis decreased slightly between 2005 and 2009 (p<.001).
Diabetes patients face significant drug burden to control diabetes and other comorbidities, and our data indicate an increased focus on cardiovascular disease risk factor control after diabetes diagnosis. However, total drug burden may be slightly decreasing over time. This information can be valuable to pharmacists working with newly-diagnosed diabetes patients to address their increasing drug regimen complexity.
diabetes; medication burden; surveillance
Clinical trials are unlikely to ever be launched for many Comparative Effectiveness Research (CER) questions. Inferences from hypothetical randomized trials may however be emulated with marginal structural modeling (MSM) using observational data but success in adjusting for time-dependent confounding and selection bias typically relies on parametric modeling assumptions. If these assumptions are violated, inferences from MSM may be inaccurate. In this article, we motivate the application of a data-adaptive estimation approach called Super Learning to avoid reliance on arbitrary parametric assumptions in CER.
Study Design and Setting
Using the electronic health records data from adults with new onset type 2 diabetes, we implemented MSM with inverse probability weighting estimation to evaluate the effect of three oral anti-diabetic therapies on the worsening of glomerular filtration rate.
Inferences from IPW estimation were noticeably sensitive to the parametric assumptions about the associations between both the exposure and censoring processes and the main suspected source of confounding, i.e., time-dependent measurements of hemoglobin A1c. Super Learning was successfully implemented to harness flexible confounding and selection bias adjustment from existing machine learning algorithms.
Erroneous IPW inference about clinical effectiveness due to arbitrary and incorrect modeling decisions may be avoided with Super Learning.
super learning; marginal structural model; inverse probability weighting; comparative effectiveness research; time-dependent confounding; selection bias
To propose a unifying set of definitions for prescription adherence research utilizing electronic health record prescribing databases, prescription dispensing databases, and pharmacy claims databases and to provide a conceptual framework to operationalize these definitions consistently across studies.
We reviewed recent literature to identify definitions in electronic database studies of prescription-filling patterns for chronic oral medications. We then develop a conceptual model and propose standardized terminology and definitions to describe prescription-filling behavior from electronic databases.
The conceptual model we propose defines two separate constructs: medication adherence and persistence. We define primary and secondary adherence as distinct sub-types of adherence. Metrics for estimating secondary adherence are discussed and critiqued, including a newer metric (New Prescription Medication Gap measure) that enables estimation of both primary and secondary adherence.
Terminology currently used in prescription adherence research employing electronic databases lacks consistency. We propose a clear, consistent, broadly applicable conceptual model and terminology for such studies. The model and definitions facilitate research utilizing electronic medication prescribing, dispensing, and/or claims databases and encompasses the entire continuum of prescription-filling behavior.
Employing conceptually clear and consistent terminology to define medication adherence and persistence will facilitate future comparative effectiveness research and meta-analytic studies that utilize electronic prescription and dispensing records.
medication adherence; medication persistence; medication discontinuation; refill compliance; refill persistence; administrative; database; electronic health record; computerized medical record systems
Research on medication safety in pregnancy often utilizes health plan and birth certificate records. This study discusses methods used to link mothers with infants, a crucial step in such research.
We describe how 8 sites participating in the Medication Exposure in Pregnancy Risk Evaluation Program created linkages between deliveries, infants and birth certificates for the 2001–2007 birth cohorts. We describe linkage rates across sites and, for two sites, we compare the characteristics of populations linked using different methods.
Of 299,260 deliveries, 256,563 (86%; range by site, 74–99%) could be linked to infants using a deterministic algorithm. At two sites, using birth certificate data to augment mother-infant linkage increased the representation of mothers who were Hispanic or non-white, younger, Medicaid recipients, or had low educational level. A total of 236,460 (92%; range by site, 82–100%) deliveries could be linked to a birth certificate.
Tailored approaches enabled linking most deliveries to infants and to birth certificates, even when data systems differed. The methods used may affect the composition of the population identified. Linkages established with such methods can support sound pharmacoepidemiology studies of maternal drug exposure outside the context of a formal registry.
Birth Certificates; Medicaid; Pregnancy Outcome/epidemiology; Medical Record Linkage
To validate an algorithm that uses delivery date and diagnosis codes to define gestational age at birth in electronic health plan databases.
Using data from 225,384 live born deliveries among women aged 15–45 years in 2001–2007 within 8 of the 11 health plans participating in the Medication Exposure in Pregnancy Risk Evaluation Program, we compared 1) the algorithm-derived gestational age versus the “gold-standard” gestational age obtained from the infant birth certificate files; and 2) the prenatal exposure status of two antidepressants (fluoxetine and sertraline) and two antibiotics (amoxicillin and azithromycin) as determined by the algorithm-derived versus the gold-standard gestational age.
The mean algorithm-derived gestational age at birth was lower than the mean obtained from the birth certificate files among singleton deliveries (267.9 versus 273.5 days) but not among multiple-gestation deliveries (253.9 versus 252.6 days). The algorithm-derived prenatal exposure to the antidepressants had a sensitivity and a positive predictive value (PPV) of ≥95%, and a specificity and a negative predictive value (NPV) of almost 100%. Sensitivity and PPV were both ≥90%, and specificity and NPV were both >99% for the antibiotics.
A gestational age algorithm based upon electronic health plan data correctly classified medication exposure status in most live born deliveries, but misclassification may be higher for drugs typically used for short durations.
algorithm; database; gestational age; maternal exposure; pregnancy; validation studies
Anti-TNF-α agents have been hypothesized to increase the risk of interstitial lung disease (ILD), including its most severe manifestation, pulmonary fibrosis.
We conducted a cohort study among autoimmune disease patients who were members of Kaiser Permanente Northern California, 1998–2007. We obtained therapies from pharmacy data and diagnoses of ILD from review of X-ray and computed tomography reports. We compared new users of anti-TNF-α agents to new users of non-biologic therapies using Cox proportional hazards analysis to adjust for baseline propensity scores and time-varying use of glucocorticoids. We also made head-to-head comparisons between anti-TNF-α agents.
Among the 8,417 persons included in the analysis, 38 (0.4%) received a diagnostic code for ILD by the end of follow-up, including 23 of 4,200 (0.5%) who used anti-TNF-α during study follow-up, and 15 of 5,423 (0.3%) who used only non-biologic therapies. The age- and gender-standardized incidence rate of ILD, per 100 person-years, was 0.21 (95% CI 0–0.43) for rheumatoid arthritis and appreciably lower for other autoimmune diseases. Compared to use of non-biologic therapies, use of anti-TNF-α therapy was not associated with a diagnosis of ILD among RA patients (adjusted hazard ratio, 1.03; 95% CI 0.51–2.07). Nor did head-to-head comparisons across anti-TNF-α agents suggest important differences in risk, although the number of cases available for analysis was limited.
The study provides evidence that compared to non-biologic therapies anti-TNF-α therapy does not increase the occurrence of ILD among patients with autoimmune diseases, and informs research design of future safety studies of ILD.
Rheumatoid arthritis; psoriatic arthritis; psoriasis; Crohn’s Disease; ulcerative colitis; inflammatory bowel disease; pharmacoepidemiology; drug safety; drug toxicity; adverse events; cohort studies; propensity scores; automated healthcare data; interstitial lung disease; pulmonary fibrosis
To estimate the prevalence of and temporal trends in prenatal antipsychotic medication use within a cohort of pregnant women in the U.S.
We identified live born deliveries to women aged 15–45 years in 2001–2007 from 11 U.S. health plans participating in the Medication Exposure in Pregnancy Risk Evaluation Program (MEPREP). We ascertained prenatal exposure to antipsychotics from health plan pharmacy dispensing files, gestational age from linked infant birth certificate files, and ICD-9-CM diagnosis codes from health plan claims files. We calculated the prevalence of prenatal use of atypical and typical antipsychotics according to year of delivery, trimester of pregnancy, and mental health diagnosis.
Among 585,615 qualifying deliveries, 4,223 (0.72%) were to women who received an atypical antipsychotic and 548 (0.09%) were to women receiving a typical antipsychotic any time from 60 days before pregnancy through delivery. There was a 2.5-fold increase in atypical antipsychotic use during the study period, from 0.33% (95% confidence interval: 0.29%, 0.37%) in 2001 to 0.82% (0.76%, 0.88%) in 2007, while the use of typical antipsychotics remained stable. Depression was the most common mental health diagnosis among deliveries to women with atypical antipsychotic use (63%), followed by bipolar disorder (43%) and schizophrenia (13%).
The number and proportion of pregnancies exposed to atypical antipsychotics has increased dramatically in recent years. Studies are needed to examine the comparative safety and effectiveness of these medications relative to other therapeutic options in pregnancy.
Antipsychotics; database; pregnancy; prevalence
To evaluate the validity of health plan and birth certificate data for pregnancy research.
A retrospective study was conducted using administrative and claims data from 11 U.S. health plans, and corresponding birth certificate data from state health departments. Diagnoses, drug dispensings, and procedure codes were used to identify infant outcomes (cardiac defects, anencephaly, preterm birth, and neonatal intensive care unit [NICU] admission) and maternal diagnoses (asthma and systemic lupus erythematosus [SLE]) recorded in the health plan data for live born deliveries between January 2001 and December 2007. A random sample of medical charts (n = 802) was abstracted for infants and mothers identified with the specified outcomes. Information on newborn, maternal, and paternal characteristics (gestational age at birth, birth weight, previous pregnancies and live births, race/ethnicity) was also abstracted and compared to birth certificate data. Positive predictive values (PPVs) were calculated with documentation in the medical chart serving as the gold standard.
PPVs were 71% for cardiac defects, 37% for anencephaly, 87% for preterm birth, and 92% for NICU admission. PPVs for algorithms to identify maternal diagnoses of asthma and SLE were ≥ 93%. Our findings indicated considerable agreement (PPVs > 90%) between birth certificate and medical record data for measures related to birth weight, gestational age, prior obstetrical history, and race/ethnicity.
Health plan and birth certificate data can be useful to accurately identify some infant outcomes, maternal diagnoses, and newborn, maternal, and paternal characteristics. Other outcomes and variables may require medical record review for validation.
administrative databases; birth certificate; positive predictive value; pregnancy; validation
To describe the prevalence, trends, and patterns in use of antidiabetic medications to treat hyperglycemia and insulin resistance prior to and during pregnancy in a large U.S. cohort of insured pregnant women.
Pregnancies resulting in livebirths were identified (N=437,950) from 2001–2007 among 372,543 women 12–50 years of age at delivery from 10 health maintenance organizations participating in the Medication Exposure in Pregnancy Risk Evaluation Program. Information for these descriptive analyses, including all antidiabetic medications dispensed during this period, was extracted from electronic health records and infant birth certificates.
Just over one percent (1.21%) of deliveries were to women dispensed antidiabetic medication(s) in the 120 days before pregnancy. Use of antidiabetic medications before pregnancy increased from 0.66% of deliveries in 2001 to 1.66% of deliveries in 2007 (p<0.001) due to a rise in metformin use. Most women using metformin before pregnancy had a diagnosis code for polycystic ovaries or female infertility (67.2%) while only 13.6% had a diagnosis code for diabetes. The use of antidiabetic medications during the second or third trimester of pregnancy increased from 2.8% of deliveries in 2001 to 3.6% in 2007 (p <0.001). Approximately two-thirds (68%) of women using metformin before pregnancy did not use any antidiabetic medications during pregnancy.
Antidiabetic medication use prior to and during pregnancy rose from 2001–2007, possibly due to increasing prevalence of gestational diabetes mellitus, type 1 and type 2 diabetes, and other conditions associated with insulin resistance.
To compare mortality among patients with selected autoimmune diseases treated with anti-tumor necrosis factor alpha (TNF-α) agents with similar patients treated with non-biologic therapies.
Cohort study set within several large health care programs, 1998–2007. Autoimmune disease patients were identified using diagnoses from computerized healthcare data. Use of anti-TNF-α agents and comparison non-biologic therapies were identified from pharmacy data and mortality was identified from vital records and other sources. We compared new users of anti-TNF-α agents to new users of non-biologic therapies using propensity scores and Cox proportional hazards analysis to adjust for baseline differences. We also made head-to-head comparisons among anti-TNF-α agents.
Among the 46,424 persons included in the analysis, 2,924 (6.3%) had died by the end of follow-up, including 1,754 (6.1%) of the 28,941 with a dispensing of anti-TNF-α agent and 1,170 (6.7%) of the 17,483 who used non-biologic treatment alone. Compared to use of non-biologic therapies, use of anti-TNF-α therapy was not associated with an increased mortality in patients with rheumatoid arthritis (adjusted hazard ratio [aHR] 0.93 with 95% CI 0.85–1.03); psoriasis, psoriatic arthritis, or ankylosing spondylitis (combined aHR 0.81 with CI 0.61–1.06; or inflammatory bowel disease (aHR 1.12 with CI 0.85–1.46). Mortality rates did not differ to an important degree between patients treated with etanercept, adalimumab, or infliximab.
Anti-TNF-α therapy was not associated with increased mortality among patients with autoimmune diseases.
Rheumatoid arthritis; psoriatic arthritis; psoriasis; Crohn’s Disease; ulcerative colitis; inflammatory bowel disease; pharmacoepidemiology; drug safety; drug toxicity; adverse events; cohort studies; propensity scores; automated healthcare data; mortality
Answers to clinical and public health research questions increasingly require aggregated data from multiple sites. Data from electronic health records and other clinical sources are useful for such studies, but require stringent quality assessment. Data quality assessment is particularly important in multisite studies to distinguish true variations in care from data quality problems.
We propose a “fit-for-use” conceptual model for data quality assessment and a process model for planning and conducting single-site and multisite data quality assessments. These approaches are illustrated using examples from prior multisite studies.
Critical components of multisite data quality assessment include: thoughtful prioritization of variables and data quality dimensions for assessment; development and use of standardized approaches to data quality assessment that can improve data utility over time; iterative cycles of assessment within and between sites; targeting assessment toward data domains known to be vulnerable to quality problems; and detailed documentation of the rationale and outcomes of data quality assessments to inform data users. The assessment process requires constant communication between site-level data providers, data coordinating centers, and principal investigators.
A conceptually based and systematically executed approach to data quality assessment is essential to achieve the potential of the electronic revolution in health care. High-quality data allow “learning health care organizations” to analyze and act on their own information, to compare their outcomes to peers, and to address critical scientific questions from the population perspective.
data quality; data quality assessment; single-site studies; multisite studies
To describe a program to study medication safety in pregnancy, the Medication Exposure in Pregnancy Risk Evaluation Program (MEPREP). MEPREP is a multi-site collaborative research program developed to enable the conduct of studies of medication use and outcomes in pregnancy. Collaborators include the U.S. Food and Drug Administration and researchers at the HMO Research Network, Kaiser Permanente Northern and Southern California, and Vanderbilt University. Datasets have been created at each site linking healthcare data for women delivering an infant between January 1, 2001 and December 31, 2008 and infants born to these women. Standardized data files include maternal and infant characteristics, medication use, and medical care at 11 health plans within 9 states; birth certificate data were obtained from the state departments of public health. MEPREP currently involves more than 20 medication safety researchers and includes data for 1,221,156 children delivered to 933,917 mothers. Current studies include evaluations of the prevalence and patterns of use of specific medications and a validation study of data elements in the administrative and birth certificate data files. MEPREP can support multiple studies by providing information on a large, ethnically and geographically diverse population. This partnership combines clinical and research expertise and data resources to enable the evaluation of outcomes associated with medication use during pregnancy.
Pregnancy; Birth outcomes; Distributed data network
Background: Drug adverse event (AE) signal detection using the Gamma Poisson Shrinker (GPS) is commonly applied in spontaneous reporting. AE signal detection using large observational health plan databases can expand medication safety surveillance. Methods: Using data from nine health plans, we conducted a pilot study to evaluate the implementation and findings of the GPS approach for two antifungal drugs, terbinafine and itraconazole, and two diabetes drugs, pioglitazone and rosiglitazone. We evaluated 1676 diagnosis codes grouped into 183 different clinical concepts and four levels of granularity. Several signaling thresholds were assessed. GPS results were compared to findings from a companion study using the identical analytic dataset but an alternative statistical method—the tree-based scan statistic (TreeScan). Results: We identified 71 statistical signals across two signaling thresholds and two methods, including closely-related signals of overlapping diagnosis definitions. Initial review found that most signals represented known adverse drug reactions or confounding. About 31% of signals met the highest signaling threshold. Conclusions: The GPS method was successfully applied to observational health plan data in a distributed data environment as a drug safety data mining method. There was substantial concordance between the GPS and TreeScan approaches. Key method implementation decisions relate to defining exposures and outcomes and informed choice of signaling thresholds.
pharmacovigilance; drug safety surveillance; adverse events data mining; gamma Poisson shrinkage; tree-based scan statistic
Information comparing characteristics of patients who do and do not pick up their prescriptions is sparse, in part because adherence measured using pharmacy claims databases does not include information on patients who never pick up their first prescription, that is, patients with primary non-adherence. Electronic health record medication order entry enhances the potential to identify patients with primary non-adherence, and in organizations with medication order entry and pharmacy information systems, orders can be linked to dispensings to identify primarily non-adherent patients.
This study aims to use database information from an integrated system to compare patient, prescriber, and payment characteristics of patients with primary non-adherence and patients with ongoing dispensings of newly initiated medications for hypertension, diabetes, and/or hyperlipidemia.
This is a retrospective observational cohort study.
PARTICIPANTS (OR PATIENTS OR SUBJECTS)
Participants of this study include patients with a newly initiated order for an antihypertensive, antidiabetic, and/or antihyperlipidemic within an 18-month period.
Proportion of patients with primary non-adherence overall and by therapeutic class subgroup. Multivariable logistic regression modeling was used to investigate characteristics associated with primary non-adherence relative to ongoing dispensings.
The proportion of primarily non-adherent patients varied by therapeutic class, including 7% of patients ordered an antihypertensive, 11% ordered an antidiabetic, 13% ordered an antihyperlipidemic, and 5% ordered medications from more than one of these therapeutic classes within the study period. Characteristics of patients with primary non-adherence varied across therapeutic classes, but these characteristics had poor ability to explain or predict primary non-adherence (models c-statistics = 0.61–0.63).
Primary non-adherence varies by therapeutic class. Healthcare delivery systems should pursue linking medication orders with dispensings to identify primarily non-adherent patients. We encourage conduct of research to determine interventions successful at decreasing primary non-adherence, as characteristics available from databases provide little assistance in predicting primary non-adherence.
medication adherence; primary non-adherence; antihypertensive adherence; antidiabetic adherence; antihyperlipidemic adherence
Inverse probability of treatment weighted (IPTW) Kaplan-Meier estimates have been developed to compare two treatments in the presence of confounders in observational studies. Recently, stabilized weights were developed to reduce the influence of extreme IPTW weights in estimating treatment effects. The objective of this paper was to use adjusted Kaplan-Meier estimates and modified log-rank and Wilcoxon tests to examine the effect of a treatment which varies over time in an observational study.
In this paper, we propose stabilized weight (SW) adjusted Kaplan-Meier estimates and modified log-rank and Wilcoxon tests when the treatment is time-varying over the follow-up period. We applied these new methods in examining the effect of an anti-platelet agent, clopidogrel, on subsequent events, including bleeding, myocardial infarction, and death after a Drug-Eluting Stent was implanted into a coronary artery. In this population, clopidogrel use may change over time based on patients' behavior (e.g., non-adherence) and physicians' recommendations (e.g., end of duration of therapy). Consequently, clopidogrel use was treated as a time-varying variable.
We demonstrate that 1) the sample sizes at three chosen time points are almost identical in the original and weighted datasets, and 2) the covariates between patients on and off clopidogrel were well balanced after SWs were applied to the original samples.
The SW-adjusted Kaplan-Meier estimates and modified log-rank and Wilcoxon tests are useful in presenting and comparing survival functions for time-varying treatments in observational studies while adjusting for known confounders.
Observational study; Kaplan Meier estimates; Stabilized weights; Time-varying treatment; Stents