Controlling confounding of treatment effects in administrative data in the presence of time-varying baseline confounders

The impact of lookback windows on the prevalence and incidence of chronic diseases among people living with HIV: an exploration in administrative health data in Canada

BACKGROUND

We described the impact of different lengths of lookback window (LW), a retrospective time period to observe diagnoses in administrative data, on the prevalence and incidence of eight chronic diseases.

METHODS

Our study populations included people living with HIV (N = 5151) and 1:5 age-sex-matched HIV-negative individuals (N = 25,755) in British Columbia, Canada, with complete follow-up between 1996 and 2012. We measured period prevalence and incidence of diseases in 2012 using LWs ranging from 1 to 16 years. Cases were deemed prevalent if identified in 2012 or within a defined LW, and incident if newly identified in 2012 with no previous cases detected within a defined LW. Chronic disease cases were ascertained using published case-finding algorithms applied to population-based provincial administrative health datasets.

RESULTS

Overall, using cases identified by the full 16-year LW as the reference, LWs ≥8 years and ≥ 4 years reduced the proportion of misclassified prevalent and incidence cases of most diseases to < 20%, respectively. The impact of LWs varied across diseases and populations.

CONCLUSIONS

This study underscored the importance of carefully choosing LWs and demonstrated data-driven approaches that may inform these choices. To improve comparability of prevalence and incidence estimates across different settings, we recommend transparent reporting of the rationale and limitations of chosen LWs.

Maternal Gestational Weight Gain in Relation to Antidepressant Continuation in Pregnancy

OBJECTIVE

Both excessive and inadequate gestational weight gain (GWG) are associated with adverse health outcomes for the woman and her child. Antidepressant use in pregnancy could affect GWG, based on evidence in nonpregnant women that some antidepressants may cause weight gain and others weight loss. Previous studies of antidepressant use and GWG were small with limited ability to account for confounding, including by maternal mental health status and severity. We assessed the association of antidepressant continuation in pregnancy with GWG among women using antidepressants before pregnancy.

STUDY DESIGN

Our retrospective cohort study included singleton livebirths from 2001 to 2014 within Kaiser Permanente Washington, an integrated health care system. Data were obtained from electronic health records and linked Washington State birth records. Among women with ≥1 antidepressant fill within 6 months before pregnancy, women who filled an antidepressant during pregnancy were considered "continuers;" women without a fill were "discontinuers." We calculated mean differences in GWG and relative risks (RR) of inadequate and excessive weight gain based on Institute of Medicine guidelines. Using inverse probability of treatment weighting with generalized estimating equations, we addressed differences in maternal characteristics, including mental health conditions.

RESULTS

Among the 2,887 births, 1,689 (59%) were to women who continued antidepressants in pregnancy and 1,198 (42%) were to discontinuers. After accounting for confounding, continuers had similar weight gain to those who discontinued (mean difference: 1.3 lbs, 95% confidence interval [CI]: -0.1 to 2.8 lbs) and similar risks of inadequate and excessive GWG (RR: 0.95, 95% CI: 0.80-1.14 and RR: 1.06, 95% CI: 0.98-1.14, respectively). Findings were comparable for specific antidepressants and trimesters of exposure.

CONCLUSION

We did not find evidence that continuation of antidepressants in pregnancy led to differences in GWG.

KEY POINTS

· Antidepressant use is associated with weight change in nonpregnant populations.. · Prior evidence on whether antidepressant use in pregnancy affects gestational weight gain is sparse.. · We accounted for confounding by characteristics such as mental health conditions and their severity.. · We found no association between pregnancy antidepressant continuation and gestational weight gain..

Contemporary Analysis of Electronic Frailty Measurement in Older Adults with Multiple Myeloma Treated in the National US Veterans Affairs Healthcare System

Simple Summary

Geriatric and frailty assessment are recommended for all older adults with cancer undergoing systemic therapy, but assessments remain difficult to scale. The aim of this study was to use an electronic frailty index based on data from administrative claims and electronic health records—the Veterans Affairs Frailty Index (VA-FI-10)—to estimate frailty and its impact on older United States (US) military veterans treated for multiple myeloma (MM) throughout the national VA Healthcare System. We found frailty to be prevalent and strongly associated with mortality and hospitalizations—independently of age, race, and MM stage. We also showed that changing the way in which the VA-FI-10 is measured affects its classification of frailty for individual veterans but not its association with mortality. These findings support the VA-FI-10’s use in research investigating outcomes in frail veterans treated with contemporary MM therapies. We provide further insights into the VA-FI-10’s potential use in clinical practice.

Abstract

Electronic frailty indices based on data from administrative claims and electronic health records can be used to estimate frailty in large populations of older adults with cancer where direct frailty measures are lacking. The objective of this study was to use the electronic Veterans Affairs Frailty Index (VA-FI-10)—developed and validated to measure frailty in the national United States (US) VA Healthcare System—to estimate the prevalence and impact of frailty in older US veterans newly treated for multiple myeloma (MM) with contemporary therapies. We designed a retrospective cohort study of 4924 transplant-ineligible veterans aged ≥ 65 years initiating MM therapy within VA from 2004 to 2017. Initial MM therapy was measured using inpatient and outpatient treatment codes from pharmacy data in the VA Corporate Data Warehouse. In total, 3477 veterans (70.6%) were classified as frail (VA-FI-10 > 0.2), with 1510 (30.7%) mildly frail (VA-FI-10 > 0.2–0.3), 1105 (22.4%) moderately frail (VA-FI-10 > 0.3–0.4), and 862 (17.5%) severely frail (VA-FI-10 > 0.4). Survival and time to hospitalization decreased with increasing VA-FI-10 severity (log-rank p-value < 0.001); the VA-FI-10 predicted mortality and hospitalizations independently of age, sociodemographic variables, and measures of disease risk. Varying data sources and assessment periods reclassified frailty severity for a substantial portion of veterans but did not substantially affect VA-FI-10’s association with mortality. Our study supports use of the VA-FI-10 in future research involving older veterans with MM and provides insights into its potential use in identifying frailty in clinical practice.

Updating and Validating the U.S. Veterans Affairs Frailty Index: Transitioning From ICD-9 to ICD-10

BACKGROUND

The Veterans Affairs Frailty Index (VA-FI) is an electronic frailty index developed to measure frailty using administrative claims and electronic health records data in Veterans. An update to ICD-10 coding is needed to enable contemporary measurement of frailty.

METHOD

International Classification of Diseases, ninth revision (ICD-9) codes from the original VA-FI were mapped to ICD-10 first using the Centers for Medicaid and Medicare Services (CMS) General Equivalence Mappings. The resulting ICD-10 codes were reviewed by 2 geriatricians. Using a national cohort of Veterans aged 65 years and older, the prevalence of deficits contributing to the VA-FI and associations between the VA-FI and mortality over years 2012-2018 were examined.

RESULTS

The updated VA-FI-10 includes 6422 codes representing 31 health deficits. Annual cohorts defined on October 1 of each year included 2 266 191 to 2 428 115 Veterans, for which the mean age was 76 years, 97%-98% were male, 78%-79% were White, and the mean VA-FI was 0.20-0.22. The VA-FI-10 deficits showed stability before and after the transition to ICD-10 in 2015, and maintained strong associations with mortality. Patients classified as frail (VA-FI > 0.2) consistently had a hazard of death more than 2 times higher than nonfrail patients (VA-FI ≤ 0.1). Distributions of frailty and associations with mortality varied with and without linkage to CMS data and with different assessment periods for capturing deficits.

CONCLUSIONS

The updated VA-FI-10 maintains content validity, stability, and predictive validity for mortality in a contemporary cohort of Veterans aged 65 years and older, and may be applied to ICD-9 and ICD-10 claims data to measure frailty.

Association of Antidepressant Continuation in Pregnancy and Infant Birth Weight

PURPOSE

The aim of the study was to evaluate the association of antidepressant continuation in pregnancy with infant birth weight among women using antidepressants before pregnancy.

METHODS

This retrospective cohort study used electronic health data linked with state birth records. We identified singleton live births (2001-2014) to enrolled women with 1 or more antidepressant prescriptions filled 6 months or less before pregnancy, including "continuers" (≥1 antidepressant fills during pregnancy, n = 1775) and "discontinuers" (no fill during pregnancy, n = 1249). We compared birth weight, small or large for gestational age (SGA or LGA), low birth weight (LBW; <2500 g), and macrosomia (>4500 g) between the 2 groups, using inverse probability of treatment weighting to account for pre-pregnancy characteristics, including mental health conditions.

RESULTS

After weighting, infants born to antidepressant continuers weighed 71.9 g less than discontinuers' infants (95% confidence interval [CI], -115.5 to -28.3 g), with a larger difference for female infants (-106.4 g; 95% CI, -164.6 to -48.1) than male infants (-48.5 g; 95% CI, -107.2 to 10.3). For female infants, SGA risk was greater in continuers than discontinuers (relative risk [RR],1.54; 95% CI, 1.02 to 2.32). Low birth weight risk was greater in continuers with 50% or more of days covered (RR, 1.69; 95% CI, 1.11 to 2.58) and exposure in the second trimester (RR, 1.53; 95% CI, 1.02 to 2.29), as compared with discontinuers.

CONCLUSIONS

Depending on infant sex, as well as duration and timing of use, continuation of antidepressant use during pregnancy may be associated with lower infant birth weight, with corresponding increases in LBW and SGA.

Effectiveness of two-drug therapy versus monotherapy as initial regimen in hypertension: A propensity score-matched cohort study in the UK Clinical Practice Research Datalink

Purpose

To compare the effectiveness on blood pressure (BP) of initial two‐drug therapy versus monotherapy in hypertensive patients.

Methods

Using the Clinical Practice Research Datalink, linked with Hospital Episode Statistics and Office for National Statistics, we identified a cohort of adults with uncontrolled hypertension, initiating one or two antihypertensive drug classes between 2006 and 2014. New users of two drugs and monotherapy were matched 1:2 by propensity score. Main exposure was “as‐treated,” ie, until first regimen change. Primary and secondary endpoints were systolic and diastolic BP control and major adverse cardiovascular event (MACE), respectively. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using Cox proportional hazard models.

Results

Of 54 523 eligible patients, 3256 (6.0%) were initiated to a two‐drug combination. Of these, 2807 were matched to 5614 monotherapy users. Mean exposure duration was 12.7 months, with 76.5% patients changing their initial regimen. Two‐drug therapy was associated with a clinically significant BP control increase in all hypertensive patients (HR = 1.17 [95%CI: 1.09‐1.26]), more so in patients with grade 2‐3 hypertension (HR = 1.28 [1.17‐1.41]). An increase of 27% in BP control (HR = 1.27 [1.08‐1.49]) was observed in patients initiating an ACEi+CCB combination compared with initiators of either single class. No significant association was found between two‐drug therapy and MACE. Several sensitivity analyses confirmed the main findings.

Conclusions

Few patients initiated therapy with two drugs, reflecting UK guidelines' recommendation to start with monotherapy. This study supports the greater effectiveness of two‐drug therapy as the initial regimen for BP control.

Antidepressant continuation in pregnancy and risk of gestational diabetes

PURPOSE

Previous studies observed modestly higher risk of gestational diabetes (GDM) associated with antidepressant use in pregnancy, potentially due to confounding by indication. We assessed the association of antidepressant continuation in pregnancy with GDM, as well as blood glucose levels, after accounting for confounding.

METHODS

We conducted a retrospective cohort study of singleton live births from 2001 to 2014 to women enrolled in Kaiser Permanente Washington, an integrated health care delivery system, utilizing electronic health data and linked Washington State birth records. We required that women have ≥1 antidepressant prescription fills ≤6 months before pregnancy. Women with an antidepressant fill during pregnancy were categorized as "continuers" (n = 1634); those without a fill were "discontinuers" (n = 1211). We calculated relative risks (RRs) for GDM and mean differences in screening blood glucose levels using generalized estimating equations with inverse probability of treatment weighting to account for baseline characteristics, including mental health conditions and indicators of mental health severity.

RESULTS

Compared with discontinuers, antidepressant continuers had comparable risk of GDM (RR: 1.10; 95% confidence interval [CI], 0.84-1.44) and blood glucose levels (mean difference: 2.3 mg/dL; 95% CI, -1.5 to 6.1 mg/dL). We observed generally similar results for specific antidepressants, with the potential exceptions of risk of GDM associated with sertraline (RR: 1.30; 95% CI, 0.90-1.88) and venlafaxine (RR: 1.52; 95% CI, 0.87-2.68), but neither association was statistically significant.

CONCLUSIONS

Our study suggests that overall, women who continue antidepressants in pregnancy are not at increased risk for GDM or higher blood glucose, although further study may be warranted for sertraline and venlafaxine.

Challenges and Opportunities for Using Big Health Care Data to Advance Medical Science and Public Health

Methodological advancements in epidemiology, biostatistics, and data science have strengthened the research world's ability to use data captured from electronic health records (EHRs) to address pressing medical questions, but gaps remain. We describe methods investments that are needed to curate EHR data toward research quality and to integrate complementary data sources when EHR data alone are insufficient for research goals. We highlight new methods and directions for improving the integrity of medical evidence generated from pragmatic trials, observational studies, and predictive modeling. We also discuss needed methods contributions to further ease data sharing across multisite EHR data networks. Throughout, we identify opportunities for training and for bolstering collaboration among subject matter experts, methodologists, practicing clinicians, and health system leaders to help ensure that methods problems are identified and resulting advances are translated into mainstream research practice more quickly.

Comparative Cardiac Safety of Selective Serotonin Reuptake Inhibitors among Individuals Receiving Maintenance Hemodialysis

BACKGROUND

Individuals receiving maintenance hemodialysis may be particularly susceptible to the lethal cardiac consequences of drug-induced QT prolongation because they have a substantial cardiovascular disease burden and high level of polypharmacy, as well as recurrent exposure to electrolyte shifts during dialysis. Electrophysiologic data indicate that among the selective serotonin reuptake inhibitors (SSRIs), citalopram and escitalopram prolong the QT interval to the greatest extent. However, the relative cardiac safety of SSRIs in the hemodialysis population is unknown.

METHODS

In this retrospective cohort study, we used data from a cohort of Medicare beneficiaries receiving hemodialysis included in the US Renal Data System registry (2007-2014). We used a new-user design to compare the 1-year risk of sudden cardiac death among hemodialysis patients initiating SSRIs with a higher potential for prolonging the QT interval (citalopram, escitalopram) versus the risk among those initiating SSRIs with lower QT-prolonging potential (fluoxetine, fluvoxamine, paroxetine, sertraline). We estimated adjusted hazard ratios using inverse probability of treatment weighted survival models. Nonsudden cardiac death was treated as a competing event.

RESULTS

The study included 30,932 (47.1%) hemodialysis patients who initiated SSRIs with higher QT-prolonging potential and 34,722 (52.9%) who initiated SSRIs with lower QT-prolonging potential. Initiation of an SSRI with higher versus lower QT-prolonging potential was associated with higher risk of sudden cardiac death (adjusted hazard ratio, 1.18; 95% confidence interval, 1.05 to 1.31). This association was more pronounced among elderly individuals, females, patients with conduction disorders, and those treated with other non-SSRI QT-prolonging medications.

CONCLUSIONS

The heterogeneous QT-prolonging potential of SSRIs may differentially affect cardiac outcomes in the hemodialysis population.

Quantifying bias reduction with fixed-duration versus all-available covariate assessment periods

PURPOSE

Implementing a cohort study in longitudinal healthcare databases requires looking back over some covariate assessment period (CAP) preceding cohort entry to measure confounders. We used simulations to compare fixed-duration versus all-available CAPs for confounder adjustment in the presence of differences in available baseline time between exposure groups.

METHODS

We simulated cohorts of 10 000 patients with binary variables for a single confounder, exposure, and outcome. Baseline time was simulated based on the observed distribution in a claims-based comparison of statin users versus nonusers. We compared bias after measuring confounders using fixed-duration and all-available CAPs, both when exposure groups had similar and discrepant amounts of available baseline time.

RESULTS

When the comparison groups had similar amounts of baseline time, an all-available CAP was less biased than a fixed-duration CAP. When baseline time differed between comparison groups, the preferable CAP approach depended on the direction of confounding and which exposure group had higher covariate sensitivity. These findings were consistent in direction across sensitivity analyses.

CONCLUSION

In certain settings of differential available baseline time between exposure groups, the all-available CAP was more biased than the fixed-duration CAP. The relative directions and strengths of confounding and misclassification biases are an important consideration when choosing between a fixed-duration or all-available CAP, but they are often unknown. Therefore, we recommend comparing the amount of available baseline time between exposure groups. When there is a large discrepancy, despite appropriate design choices, we recommend a fixed-duration approach to avoid potential increases in bias because of differential data availability.

A Comparative Study of Carvedilol Versus Metoprolol Initiation and 1-Year Mortality Among Individuals Receiving Maintenance Hemodialysis

BACKGROUND

Carvedilol and metoprolol are the β-blockers most commonly prescribed to US hemodialysis patients, accounting for ∼80% of β-blocker prescriptions. Despite well-established pharmacologic and pharmacokinetic differences between the 2 medications, little is known about their relative safety and efficacy in the hemodialysis population.

STUDY DESIGN

A retrospective cohort study using a new-user design.

SETTING & PARTICIPANTS

Medicare-enrolled hemodialysis patients treated at a large US dialysis organization who initiated carvedilol or metoprolol therapy from January 1, 2007, through December 30, 2012.

PREDICTOR

Carvedilol versus metoprolol initiation.

OUTCOMES

All-cause mortality, cardiovascular mortality, and intradialytic hypotension (systolic blood pressure decrease ≥ 20mmHg during hemodialysis plus intradialytic saline solution administration) during a 1-year follow-up period.

MEASUREMENTS

Survival models were used to estimate HRs and 95% CIs in mortality analyses. Poisson regression was used to estimate incidence rate ratios (IRRs) and 95% CIs in intradialytic hypotension analyses. Inverse probability of treatment weighting was used to adjust for several demographic, clinical, laboratory, and dialysis treatment covariates in all analyses.

RESULTS

27,064 individuals receiving maintenance hemodialysis were included: 9,558 (35.3%) carvedilol initiators and 17,506 (64.7%) metoprolol initiators. Carvedilol (vs metoprolol) initiation was associated with greater all-cause (adjusted HR, 1.08; 95% CI, 1.02-1.16) and cardiovascular mortality (adjusted HR, 1.18; 95% CI, 1.08-1.29). In subgroup analyses, similar associations were observed among patients with hypertension, atrial fibrillation, heart failure, and a recent myocardial infarction, the main cardiovascular indications for β-blocker therapy. During follow-up, carvedilol (vs metoprolol) initiators had a higher rate of intradialytic hypotension (adjusted IRR, 1.10; 95% CI, 1.09-1.11).

LIMITATIONS

Residual confounding may exist.

CONCLUSIONS

Relative to metoprolol initiation, carvedilol initiation was associated with higher 1-year all-cause and cardiovascular mortality. One potential mechanism for these findings may be the increased occurrence of intradialytic hypotension after carvedilol (vs metoprolol) initiation.

Thirty-five-year Trends in First-time Hospitalization for Hip Fracture, 1-year Mortality, and the Prognostic Impact of Comorbidity: A Danish Nationwide Cohort Study, 1980-2014

BACKGROUND

We examined trends in hip fracture incidence in Denmark from 1980 to 2014, trends in subsequent 1-year mortality, and the prognostic impact of sex, age, and comorbidity.

METHODS

This nationwide cohort study prospectively collected data from population-based Danish registries. We included 262,437 patients with incident hip fracture and assessed comorbidity using the Charlson Comorbidity Index (CCI).

RESULTS

Despite slight increases in incidence rates (IRs) of hip fracture up to the mid-1990s, the annual IR decreased by 29% from 1980 to 2014 in women but remained stable in men. Decrease affected all age groups. IR decreased in patients without comorbidity but increased with increasing comorbidity (13% in patients with moderate and 510% in patients with very severe comorbidity). Adjusted mortality rate ratios (MRRs) following hip fracture in 2010-2014 compared with 1980-1984 were 0.68 (95% confidence interval [CI] = 0.65, 0.71) within 30 days and 0.63 (95% CI = 0.61, 0.66) within 31-365 days. The mortality decreased up to 40% irrespective of comorbidity. Compared with patients with no comorbidity, those with very severe comorbidity had adjusted MRRs of 2.48 (95% CI = 2.39, 2.56) and 2.81 (95% CI = 2.74, 2.88) within 30 days and 31-365 days post-hip fracture, respectively.

CONCLUSIONS

Although the incidence rate of hip fracture increased substantially with increasing comorbidity, the following 1-year mortality decreased by 40% from 1980 through 2014 irrespective of sex, age, and comorbidity level, suggesting improvement in both treatment and rehabilitation of patients with hip fracture. Comorbidity burden was, however, a strong prognostic factor for 1-year mortality after hip fracture.

Comorbidity and the increased mortality after hospitalization for stroke: a population-based cohort study

Essentials Comorbidity is prevalent in the stroke population and affects post-stroke survival. A stroke patient cohort (n = 201 691) and a general population cohort were followed for survival. Cancer and advanced renal/liver disease substantially increased one-year stroke mortality. Tailoring stroke interventions according to comorbidity may reduce excess mortality.

SUMMARY

Background Comorbidity is prevalent among stroke patients, affecting post-stroke survival. It remains unknown whether comorbidity impacts post-stroke mortality beyond the combined individual effects of stroke and comorbidity. Methods Using nationwide Danish databases, we performed a cohort study of 201 691 patients ≥ 18 years old with incident ischemic stroke, intracerebral or subarachnoid hemorrhage, or unspecified stroke during 1995-2012, and 992 942 adults from the general population, matched to stroke patients by birth year, sex and individual comorbidities in the Charlson Comorbidity Index. During up to 5 years of follow-up, we computed standardized mortality rates (SMRs) to assess interaction contrasts as a measure of excess mortality not explained by the additive effects of stroke and comorbidity acting alone. Results Five-year post-stroke mortality was 48%, corresponding to an SMR of 187 deaths per 1000 person-years. During the 30-day peak post-stroke mortality (SMR, 180 per 1000 person-months), interaction with comorbidity represented 23%, 34% and 51% of post-stroke mortality rates among patients with low (score = 1), moderate (score = 2-3) and high (score = 4+) comorbidity based on Charlson Comorbidity Index scores. The interaction accounted for 5% to 32% of subsequent 31-365-day post-stroke mortality rates, depending on comorbidity level. The interaction contrasts were most notable among comorbid patients with cancer, particularly with hematological or metastatic disease, followed by patients with moderate-to-severe liver or renal disease. Conclusion Comorbidity, notably cancer and advanced renal or liver disease, increased 1-year mortality after stroke beyond the combined effects expected from either disease acting alone.

Propensity Scores in Pharmacoepidemiology: Beyond the Horizon

Purpose of review

Propensity score methods have become commonplace in pharmacoepidemiology over the past decade. Their adoption has confronted formidable obstacles that arise from pharmacoepidemiology's reliance on large healthcare databases of considerable heterogeneity and complexity. These include identifying clinically meaningful samples, defining treatment comparisons, and measuring covariates in ways that respect sound epidemiologic study design. Additional complexities involve correctly modeling treatment decisions in the face of variation in healthcare practice, and dealing with missing information and unmeasured confounding. In this review, we examine the application of propensity score methods in pharmacoepidemiology with particular attention to these and other issues, with an eye towards standards of practice, recent methodological advances, and opportunities for future progress.

Recent findings

Propensity score methods have matured in ways that can advance comparative effectiveness and safety research in pharmacoepidemiology. These include natural extensions for categorical treatments, matching algorithms that can optimize sample size given design constraints, weighting estimators that asymptotically target matched and overlap samples, and the incorporation of machine learning to aid in covariate selection and model building.

Summary

These recent and encouraging advances should be further evaluated through simulation and empirical studies, but nonetheless represent a bright path ahead for the observational study of treatment benefits and harms.

The impact of the lookback period and definition of confirmatory events on the identification of incident cancer cases in administrative data

Background

This cohort study examined the impact of the lengths of lookback and confirmation periods as well as the definition of confirmatory events on the number of incident cancer cases identified and age-standardized cumulative incidences (ACI) estimated in administrative data using German cancer registry data as a benchmark.

Methods

ACI per 100,000 insured persons for breast, prostate and colorectal cancer were estimated using BARMER Statutory Health Insurance claims data. Incident cancer cases were defined as having an in- or outpatient diagnosis in 2013, no diagnosis in a lookback period of 1 year and a second diagnosis (or death) in a confirmation period of 1 quarter. We varied lookback periods from 1 to 7 years, confirmation periods from 1 to 4 quarters as well as the definition of confirmatory events and compared ACI estimates to cancer registry data.

Results

ACI were higher for breast (138.7) and prostate (103.6) but lower for colorectal cancer (42.1) when compared to cancer registries (119.3, 98.0 and 45.5, respectively). Extending the lookback period to 7 years reduced ACI to 129.0, 95.1 and 38.3. An extended confirmation period of 4 quarters increased ACI to 151.3, 114.9 and 46.8. Including breast and colorectal surgeries as a confirmatory event reduced ACI to 114.9 and 37.1, respectively.

Conclusions

The choice of lookback and confirmation periods and the definition of confirmatory events have considerable impact on the number of incident cancer cases identified and ACI estimated. Researchers need to be aware of potential misclassification when identifying incident cancer cases in administrative data. Further validation studies as well as studies using administrative data to estimate cancer incidences should consider several choices of the lookback and confirmation periods and the definition of confirmatory events to show how these parameters impact the validity and robustness of their results.

Electronic supplementary material

The online version of this article (doi:10.1186/s12874-017-0407-4) contains supplementary material, which is available to authorized users.

Effects of expanding the look-back period to all available data in the assessment of covariates

BACKGROUND

A fixed baseline period has been a common covariate assessment approach in pharmacoepidemiological studies from claims but may lead to high levels of covariate misclassification. Simulation studies have recommended expanding the look-back approach to all available data (AAD) for binary indicators of diagnoses, procedures, and medications, but there have been few real data analyses using this approach.

OBJECTIVE

The objective of the study is to explore the impact on treatment effect estimates and covariate prevalence of expanding the look-back period within five validated studies in the Aetion system, a rapid cycle analytics platform.

METHODS

We reran the five studies and assessed covariates using (i) a fixed window approach (usually 180 days before treatment initiation), (ii) AAD prior to treatment initiation, and (iii) AAD with a categorized by recency approach, where the most recent occurrence of a covariate was labeled as recent (occurring within the fixed window) or past (before the start of the fixed window). For each covariate assessment approach, we adjusted for covariates via propensity score matching.

RESULTS

All studies had at least one covariate that had an increase in prevalence of 15% or higher from the fixed window to the AAD approach. However, there was little change in treatment effect estimates resulting from differing covariate assessment approaches. For example, in a study of acute coronary syndrome in high-intensity versus low-intensity statin users, the estimated hazard ratio from the fixed window approach was 1.11 (95% confidence interval 0.98, 1.25) versus 1.21 (1.07, 1.37) when using AAD and 1.19 (1.05, 1.35) using categorized by recency.

CONCLUSION

Expanding the baseline period to AAD improved covariate sensitivity by capturing data that would otherwise be missed yet did not meaningfully change the overall treatment effect estimates compared with the fixed window approach. Copyright © 2017 John Wiley & Sons, Ltd.

Epoetin Alfa and Outcomes in Dialysis amid Regulatory and Payment Reform

Erythropoiesis-stimulating agents (ESAs) are commonly used to treat anemia in patients with CKD, including those receiving dialysis, although clinical trials have identified risks associated with ESA use. We evaluated the effects of changes in dialysis payment policies and product labeling instituted in 2011 on mortality and major cardiovascular events across the United States dialysis population in an open cohort study of patients on dialysis from January 1, 2005, through December 31, 2012, with Medicare as primary payer. We compared observed rates of death and major cardiovascular events in 2011 and 2012 with expected rates calculated on the basis of rates in 2005-2010, accounting for differences in patient characteristics and influenza virulence. An abrupt decline in erythropoietin dosing and hemoglobin concentration began in late 2010. Observed rates of all-cause mortality, cardiovascular mortality, and myocardial infarction in 2011 and 2012 were consistent with expected rates. During 2012, observed rates of stroke, venous thromboembolic disease (VTE), and heart failure were lower than expected (absolute deviation from trend per 100 patient-years [95% confidence interval]: -0.24 [-0.08 to -0.37] for stroke, -2.43 [-1.35 to -3.70] for VTE, and -0.77 [-0.28 to -1.27] for heart failure), although non-ESA-related changes in practice and Medicare payment penalties for rehospitalization may have confounded the results. This initial evidence suggests that action taken to mitigate risks associated with ESA use and changes in payment policy did not result in a relative increase in death or major cardiovascular events and may reflect improvements in stroke, VTE, and heart failure.