Disease risk score as a confounder summary method: systematic review and recommendations

Confounder Adjustment Using the Disease Risk Score: A Proposal for Weighting Methods

Propensity score analysis is a common approach to addressing confounding in nonrandomized studies. Its implementation, however, requires important assumptions (e.g., positivity). The disease risk score (DRS) is an alternative confounding score that can relax some of these assumptions. Like the propensity score, the DRS summarizes multiple confounders into a single score, on which conditioning by matching allows the estimation of causal effects. However, matching relies on arbitrary choices for pruning out data (e.g., matching ratio, algorithm, and caliper width) and may be computationally demanding. Alternatively, weighting methods, common in propensity score analysis, are easy to implement and may entail fewer choices, yet none have been developed for the DRS. Here we present 2 weighting approaches: One derives directly from inverse probability weighting; the other, named target distribution weighting, relates to importance sampling. We empirically show that inverse probability weighting and target distribution weighting display performance comparable to matching techniques in terms of bias but outperform them in terms of efficiency (mean squared error) and computational speed (up to >870 times faster in an illustrative study). We illustrate implementation of the methods in 2 case studies where we investigate placebo treatments for multiple sclerosis and administration of aspirin in stroke patients.

Renin-angiotensin-aldosterone system inhibitors as a risk factor for chronic subdural hematoma recurrence: A matter of debate

OBJECTIVES

Chronic subdural hematoma (cSDH) is a common central nervous system condition. Recent reports indicate that cSDH affects long-term prognosis; however, its definitive risk factors remain unknown. An antihypertensive drug, renin-angiotensin-aldosterone system inhibitors (RAASi), can affect vascular permeability and cell proliferation processes, which may suppress the recurrence of cSDH. However, several studies have reported negative results to this effect. Therefore, we aimed to evaluate antihypertensive drugs, including RAASi, as risk factors for recurrent cSDH.

MATERIALS AND METHODS

A total of 203 consecutive cases of surgically treated cSDH were retrospectively reviewed. Clinical and radiological parameters were compared between the groups with and without cSDH recurrence to identify risk factors.

RESULTS

Of the included cases, 68 (33.5%) used RAASi and 37 (18.2%) developed recurrence within 60 days of surgery. In the multiple logistic regression analysis adjusted by composite risk score, the odds ratios (95% confidence interval) of RAASi, calcium channel blockers, diuretics, β and α blockers, for the recurrent risk of cSDH after surgery were 2.49 (1.16, 5.42), 1.79 (0.84, 3.82), 1.83 (0.62, 4.87), 0.90 (0.28, 2.44), and 0.96 (0.21, 3.20), respectively. The Cox proportional hazard model also demonstrated that RAASi-use was an independent risk factor for cSDH recurrence.

CONCLUSIONS

Present series suggests RAASi-use as a risk factor for cSDH recurrence, although the role of RAASi-use in cSDH remains debatable. Further studies for deeper understanding of the microenvironment of hematoma and the surroundings are preferable. (235 words).

Real-World Trends in the Evaluation of Medical Products

There is a compelling need to evaluate the real-world health effects of medical products outside of tightly controlled preapproval clinical trials. This is done through pharmacoepidemiology, which is the study of the health effects of medical products (including drugs, biologicals, and medical devices and diagnostics) in populations, often using nonrandomized designs. Recent developments in pharmacoepidemiology span changes in the focus of research questions, research designs, data used, and statistical analysis methods. Developments in these areas are thought to improve the value of the evidence produced by such studies, and are prompting greater use of real-world evidence to inform clinical, regulatory, and reimbursement decisions.

Use of Instrumental Variable Analyses for Evaluating Comparative Effectiveness in Empirical Applications of Oncology: A Systematic Review

PURPOSE

This systematic review aims to characterize the use and trends of instrumental variables (IVs) in oncology research, assess the quality and completeness of IV reporting, and evaluate the agreement and interpretation of IV results in comparison with other techniques used for determining comparative effectiveness in observational research.

METHODS

We performed a systematic search of observational empirical oncology papers evaluating the comparative effectiveness of cancer treatments using IV methods. EMBASE and MEDLINE (through June 2021) were used for a keyword search; Scopus and Web of Science were used for a citation search. Publication details and characteristics of IV analysis and reporting were extracted from each study to examine the uptake and quality of IV applications.

RESULTS

Sixty-five empirical papers were identified from February 2001 through June 2021. Geographic variation (50.8%) was the most common type of IV used, and the majority of IV applications constructed binary instruments (53.8%). Concurrent analyses using another non-IV method to adjust for confounding were conducted in 56 (86.2%) studies, 17 (30.4%) of which produced results divergent from IV approaches. We observed a modest uptake of IV methods between 2011 and 2021 together with its dissemination, which remained fairly limited to the United States (76.9%). The quality and completeness of IV reporting varied greatly. The underlying assumptions required for a valid IV analysis were only accounted for in full by 20 (30.8%) studies.

CONCLUSION

There are limited use and variable quality of IV analyses in oncology. Future research should look to establish standards to better facilitate the quality, transparency, and completeness of IV reporting in this setting.

Factors Associated With Severity of COVID-19 Disease in a Multicenter Cohort of People With HIV in the United States, March-December 2020

BACKGROUND

Understanding the spectrum of COVID-19 in people with HIV (PWH) is critical to provide clinical guidance and risk reduction strategies.

SETTING

Centers for AIDS Research Network of Integrated Clinic System, a US multisite clinical cohort of PWH in care.

METHODS

We identified COVID-19 cases and severity (hospitalization, intensive care, and death) in a large, diverse HIV cohort during March 1, 2020-December 31, 2020. We determined predictors and relative risks of hospitalization among PWH with COVID-19, adjusted for disease risk scores.

RESULTS

Of 16,056 PWH in care, 649 were diagnosed with COVID-19 between March and December 2020. Case fatality was 2%; 106 (16.3%) were hospitalized, and 12 died. PWH with current CD4 count <350 cells/mm 3 [aRR 2.68; 95% confidence interval (CI): 1.93 to 3.71; P < 0.001] or lowest recorded CD4 count <200 cells/mm 3 (aRR 1.67; 95% CI: 1.18 to 2.36; P < 0.005) had greater risks of hospitalization. HIV viral load and antiretroviral therapy status were not associated with hospitalization, although most of the PWH were suppressed (86%). Black PWH were 51% more likely to be hospitalized with COVID-19 compared with other racial/ethnic groups (aRR 1.51; 95% CI: 1.04 to 2.19; P = 0.03). Chronic kidney disease, chronic obstructive pulmonary disease, diabetes, hypertension, obesity, and increased cardiovascular and hepatic fibrosis risk scores were associated with higher hospitalization risk. PWH who were older, not on antiretroviral therapy, and with current CD4 count <350 cells/mm 3 , diabetes, and chronic kidney disease were overrepresented among PWH who required intubation or died.

CONCLUSIONS

PWH with CD4 count <350 cells/mm 3 , and a history of CD4 count <200 cells/mm 3 , have a clear excess risk of severe COVID-19, accounting for comorbidities associated with severe outcomes. PWH with these risk factors should be prioritized for COVID-19 vaccination and early treatment and monitored closely for worsening illness.

Racial and ethnic disparities in coronavirus disease 2019 disease incidence independent of comorbidities, among people with HIV in the United States

OBJECTIVES

To define the incidence of clinically detected coronavirus disease 2019 (COVID-19) in people with HIV (PWH) in the United States and evaluate how racial and ethnic disparities, comorbidities, and HIV-related factors contribute to risk of COVID-19.

DESIGN

Observational study within the CFAR Network of Integrated Clinical Systems cohort in seven cities during 2020.

METHODS

We calculated cumulative incidence rates of COVID-19 diagnosis among PWH in routine care by key characteristics including race/ethnicity, current and lowest CD4+ cell count, and geographic area. We evaluated risk factors for COVID-19 among PWH using relative risk regression models adjusted with disease risk scores.

RESULTS

Among 16 056 PWH in care, of whom 44.5% were black, 12.5% were Hispanic, with a median age of 52 years (IQR 40-59), 18% had a current CD4+ cell count less than 350 cells/μl, including 7% less than 200; 95.5% were on antiretroviral therapy (ART), and 85.6% were virologically suppressed. Overall in 2020, 649 PWH were diagnosed with COVID-19 for a rate of 4.94 cases per 100 person-years. The cumulative incidence of COVID-19 was 2.4-fold and 1.7-fold higher in Hispanic and black PWH respectively, than non-Hispanic white PWH. In adjusted analyses, factors associated with COVID-19 included female sex, Hispanic or black identity, lowest historical CD4+ cell count less than 350 cells/μl (proxy for CD4+ nadir), current low CD4+ : CD8+ ratio, diabetes, and obesity.

CONCLUSION

Our results suggest that the presence of structural racial inequities above and beyond medical comorbidities increased the risk of COVID-19 among PWH. PWH with immune exhaustion as evidenced by lowest historical CD4+ cell count or current low CD4+ : CD8+ ratio had greater risk of COVID-19.

Core Concepts in Pharmacoepidemiology: Fundamentals of the cohort and case-control study designs

In this review paper, I outline the principles of the cohort as a sampling frame and provide a basic introduction to the cohort study design and the case–control study design, two of the most important designs in the pharmacoepidemiologist's toolbox. Further, I discuss when to prefer one design over the other. The paper is intended as a primer for people new to the field of pharmacoepidemiology and contains a range of suggestions for additional reading regarding the study designs and related epidemiological topics.

Prevalence and Associated Factors of Depression in Medical Students in a Northern Thailand University: A Cross-Sectional Study

This study was conducted to investigate the prevalence and associated factors of depression in medical students. This cross-sectional study investigated the prevalence and associated factors of depression in medical students from May 2018 to April 2019. Depression was diagnosed using the nine-item Patient Health Questionnaire. We evaluated the following potential predictors: demographic data, stressors, psychiatric comorbidities, emotional intelligence (EI), and perceived social support. The association between potential factors and depression was analyzed using multiple logistic regression analysis. The prevalence of depression was 149 of 706 students with 12.5% suicidality. Second- and fourth-year medical students were high-risk groups. Risk factors identified were insufficient income, physical illness, and previous psychiatric illness. Depression in medical students likely coincides with anxiety, internet addiction, sleep problems, and loneliness. Highly associated stressors were personal relationships, physical health, mental health, difficulties in social relationships, satisfaction with grades, and boredom with medical education. Protective EI factors included emotional self-control, problem-solving abilities, inner peace, and life satisfaction. Up to 21.1% of medical students had depression. In this study, among multiple known risk factors of depression, we found that EI is the novel protective factor against depression among medical students. EI training might be protective intervention for medical students in the future.

Racial and ethnic disparities in COVID-19 disease incidence independent of comorbidities, among people with HIV in the US

OBJECTIVES

To define the incidence of clinically-detected COVID-19 in people with HIV (PWH) in the US and evaluate how racial and ethnic disparities, comorbidities, and HIV-related factors contribute to risk of COVID-19.

DESIGN

Observational study within the CFAR Network of Integrated Clinical Systems cohort in 7 cities during 2020.

METHODS

We calculated cumulative incidence rates of COVID-19 diagnosis among PWH in routine care by key characteristics including race/ethnicity, current and lowest CD4 count, and geographic area. We evaluated risk factors for COVID-19 among PWH using relative risk regression models adjusted with disease risk scores.

RESULTS

Among 16,056 PWH in care, of whom 44.5% were Black, 12.5% were Hispanic, with a median age of 52 years (IQR 40-59), 18% had a current CD4 count < 350, including 7% < 200; 95.5% were on antiretroviral therapy, and 85.6% were virologically suppressed. Overall in 2020, 649 PWH were diagnosed with COVID-19 for a rate of 4.94 cases per 100 person-years. The cumulative incidence of COVID-19 was 2.4-fold and 1.7-fold higher in Hispanic and Black PWH respectively, than non-Hispanic White PWH. In adjusted analyses, factors associated with COVID-19 included female sex, Hispanic or Black identity, lowest historical CD4 count <350 (proxy for CD4 nadir), current low CD4/CD8 ratio, diabetes, and obesity.

CONCLUSIONS

Our results suggest that the presence of structural racial inequities above and beyond medical comorbidities increased the risk of COVID-19 among PWHPWH with immune exhaustion as evidenced by lowest historical CD4 or current low CD4:CD8 ratio had greater risk of COVID-19.

Disparities by sex in P2Y12 inhibitor therapy duration, or differences in the balance of ischaemic-benefit and bleeding-risk clinical outcomes in older women versus comparable men following acute myocardial infarction? A P2Y12 inhibitor new user retrospective cohort analysis of US Medicare claims data

OBJECTIVES

To determine if comparable older women and men received different durations of P2Y₁₂ inhibitor therapy following acute myocardial infarction (AMI) and if therapy duration differences were justified by differences in ischaemic benefits and/or bleeding risks.

DESIGN

Retrospective cohort.

SETTING

20% sample of 2007-2015 US Medicare fee-for-service administrative claims data.

PARTICIPANTS

≥66-year-old P2Y₁₂ inhibitor new users following 2008-2013 AMI hospitalisation (N=30 613). Older women compared to older men with similar predicted risks of study outcomes.

PRIMARY AND SECONDARY OUTCOME MEASURES

Primary outcome: P2Y₁₂ inhibitor duration (modelled as risk of therapy discontinuation).

SECONDARY OUTCOMES

clinical events while on P2Y₁₂ inhibitor therapy, including (1) death/hospice admission, (2) composite of ischaemic events (AMI/stroke/revascularisation) and (3) hospitalised bleeds. Cause-specific risks and relative risks (RRs) estimated using Aalen-Johansen cumulative incidence curves and bootstrapped 95% CIs.

RESULTS

10 486 women matched to 10 486 men with comparable predicted risks of all 4 study outcomes. No difference in treatment discontinuation was observed at 12 months (women 31.2% risk; men 30.9% risk; RR 1.01; 95% CI 0.97 to 1.05), but women were more likely than men to discontinue therapy at 24 months (54.4% and 52.9% risk, respectively; RR 1.03; 95% CI 1.00 to 1.05). Among patients who did not discontinue P2Y₁₂ inhibitor therapy, women had lower 24-month risks of ischaemic outcomes than men (13.1% and 14.7%, respectively; RR 0.90; 95% CI 0.84 to 0.96), potentially lower 24-month risks of death/hospice admission (5.0% and 5.5%, respectively; RR 0.91; 95% CI 0.82 to 1.02), but women and men both had 2.5% 24-month bleeding risks (RR 0.98; 95% CI 0.82 to 1.14).

CONCLUSIONS

Risks for death/hospice and ischaemic events were lower among women still taking a P2Y₁₂ inhibitor than comparable men, with no difference in bleeding risks. Shorter P2Y₁₂ inhibitor durations in older women than comparable men observed between 12 and 24 months post-AMI may reflect a disparity that is not justified by differences in clinical need.

Factors associated with severity of COVID-19 disease in a multicenter cohort of people with HIV in the United States, March-December 2020

BACKGROUND

Understanding the spectrum of SARS-CoV-2 infection and COVID-19 disease in people with HIV (PWH) is critical to provide clinical guidance and implement risk-reduction strategies.

OBJECTIVE

To characterize COVID-19 in PWH in the United States and identify predictors of disease severity.

DESIGN

Observational cohort study.

SETTING

Geographically diverse clinical sites in the CFAR Network of Integrated Clinical Systems (CNICS).

PARTICIPANTS

Adults receiving HIV care through December 31, 2020.

MEASUREMENTS

COVID-19 cases and severity (hospitalization, intensive care, death).

RESULTS

Of 16,056 PWH in care, 649 were diagnosed with COVID-19 between March-December 2020. Case fatality was 2%; 106 (16.3%) were hospitalized and 12 died. PWH with current CD4 count <350 cells/mm ³ (aRR 2.68; 95%CI 1.93-3.71; P<.001) or lowest recorded CD4 count <200 (aRR 1.67; 95%CI 1.18-2.36; P<.005) had greater risk of hospitalization. HIV viral load suppression and antiretroviral therapy (ART) status were not associated with hospitalization, although the majority of PWH were suppressed (86%). Black PWH were 51% more likely to be hospitalized with COVID-19 compared to other racial/ethnic groups (aRR 1.51; 95%CI 1.04-2.19, P=.03). Chronic kidney disease (CKD), chronic obstructive pulmonary disease, diabetes, hypertension, obesity, and increased cardiovascular and hepatic fibrosis risk scores were associated with higher risk of hospitalization. PWH who were older, not on ART, with current CD4 <350, diabetes, and CKD were overrepresented amongst PWH who required intubation or died.

LIMITATIONS

Unable to compare directly to persons without HIV; underestimate of total COVID-19 cases.

CONCLUSIONS

PWH with CD4 <350 cells/mm ³ , low CD4/CD8 ratio, and history of CD4 <200, have a clear excess risk of severe COVID-19, after accounting for comorbidities also associated with severe outcomes. PWH with these risk factors should be prioritized for COVID-19 vaccination, early treatment, and monitored closely for worsening illness.

Reducing Bias Due to Exposure Measurement Error Using Disease Risk Scores

Suppose that an investigator wants to estimate an association between a continuous exposure variable and an outcome, adjusting for a set of confounders. If the exposure variable suffers classical measurement error, in which the measured exposures are distributed with independent error around the true exposure, then an estimate of the covariate-adjusted exposure-outcome association may be biased. We propose an approach to estimate a marginal exposure-outcome association in the setting of classical exposure measurement error using a disease score-based approach to standardization to the exposed sample. First, we show that the proposed marginal estimate of the exposure-outcome association will suffer less bias due to classical measurement error than the covariate-conditional estimate of association when the covariates are predictors of exposure. Second, we show that if an exposure validation study is available with which to assess exposure measurement error, then the proposed marginal estimate of the exposure-outcome association can be corrected for measurement error more efficiently than the covariate-conditional estimate of association. We illustrate both of these points using simulations and an empirical example using data from the Orinda Longitudinal Study of Myopia (California, 1989-2001).

Standardizing Discrete-Time Hazard Ratios With a Disease Risk Score

The disease risk score (DRS) is a summary score that is a function of a potentially large set of covariates. The DRS can be used to control for confounding by the covariates that went into estimation of the DRS and obtain a standardized estimate of an exposure's effect on disease. However, to date, literature on the DRS has not addressed analyses that focus on estimation of survival or hazard functions, which are common in epidemiologic analyses of cohort data. Here, we propose a method for standardization of hazard ratios using the DRS in longitudinal analyses of the association between a binary exposure and an outcome. This approach to handling a potentially large set of covariates through a model-based approach to standardization may provide a useful tool for cohort analyses of hazard ratios and may be particularly well-suited to settings where an exposure propensity score is difficult to model. Simulations are used in this paper to illustrate the approach, and an empirical example is provided.

Association between anxiolytic/hypnotic drugs and suicidal thoughts or behaviors in a population-based cohort of students

AIMS

To investigate the association between the use of anxiolytic/hypnotic drugs and suicidal thoughts and/or behavior (STB) in students.

METHODS

12,112 participants who completed the baseline questionnaire in the i-Share cohort between April 2013 and March 2017 were included. STB were defined at inclusion as suicidal thoughts over the previous year and/or a lifetime suicide attempt. The use of prescribed anxiolytic/hypnotic drugs over the previous 3 months was measured at baseline and follow-up time points (in 2,919 students). Psychiatric disorders were assessed through validated scales. Multivariate logistic regression models were run using disease risk score.

RESULTS

At inclusion, 25.2% of students had STB and 10.3% used anxiolytics/hypnotics. STB at baseline were associated with a more frequent use of anxiolytics/hypnotics in the previous 3 months, after adjustment for covariates including anxiety, depression, sleep, impulsivity, and substance use. The use of anxiolytics/hypnotics at baseline was not associated with the occurrence, persistence or remission of STB one year later. STB at baseline were associated with a new anxiolytic/hypnotic treatment one year later.

CONCLUSIONS

Anxiolytic/hypnotic drug use was associated with STB in students independently of many risk factors of suicide and most psychiatric disorders that require such treatment, which raises drug safety concerns.

Factors Associated with Failure of Intermittent Nebulization with Short-Acting Beta-Agonists in Children with Severe Asthma Exacerbation

Purpose

Intermittent nebulization of short-acting beta-agonists (SABA) is the initial treatment of choice for children with asthma exacerbation. However, children with severe asthma exacerbation (SAE) may not show an adequate response and need aggressive stepwise therapy. We aimed to explore factors associated with a poor response to intermittent nebulized SABA in children with SAE.

Methods

A retrospective cohort study of children with SAE diagnosed according to the definition of the British Guidelines on the Management of Asthma, who were admitted at Hat Yai Hospital from January 1, 2015, to December 31, 2017. All children were treated with intermittent SABA nebulization. Treatment failure was defined as children needing escalated therapy. Logistic regression with confounding score adjustment was used to explore the predictors of treatment failure.

Results

One hundred thirty-three children were included in the analysis, 59 were in the failure group and 74 were in the success group. After adjusting for potential confounders, they were significantly associated with a previous history of intubation (adjusted OR 6.46, 95% CI 1.13 to 36.79, p=0.036), receiving <3 doses of nebulized salbutamol in the emergency room (ER, aOR 3.21, 95% CI 1.15 to 9.02, p=0.027), ER measured oxygen saturation (SpO₂) <92% (adjusted OR 3.02, 95% CI 1.18 to 7.75, p=0.022), and exacerbation triggered by pneumonia (adjusted OR 2.67, 95% CI 1.19 to 6.00, p=0.017).

Conclusion

We identified four prognostic factors of treatment failure in children with SAE: a previous history of intubation; receiving <3 doses of nebulized salbutamol in the ER, SpO₂ at ER <92%; and exacerbation triggered by pneumonia. Further prospective studies are required to confirm our findings before clinical implementation.

On the aggregation of published prognostic scores for causal inference in observational studies

As real world evidence on drug efficacy involves nonrandomized studies, statistical methods adjusting for confounding are needed. In this context, prognostic score (PGS) analysis has recently been proposed as a method for causal inference. It aims to restore balance across the different treatment groups by identifying subjects with a similar prognosis for a given reference exposure ("control"). This requires the development of a multivariable prognostic model in the control arm of the study sample, which is then extrapolated to the different treatment arms. Unfortunately, large cohorts for developing prognostic models are not always available. Prognostic models are therefore subject to a dilemma between overfitting and parsimony; the latter being prone to a violation of the assumption of no unmeasured confounders when important covariates are ignored. Although it is possible to limit overfitting by using penalization strategies, an alternative approach is to adopt evidence synthesis. Aggregating previously published prognostic models may improve the generalizability of PGS, while taking account of a large set of covariates-even when limited individual participant data are available. In this article, we extend a method for prediction model aggregation to PGS analysis in nonrandomized studies. We conduct extensive simulations to assess the validity of model aggregation, compared with other methods of PGS analysis for estimating marginal treatment effects. We show that aggregating existing PGS into a "meta-score" is robust to misspecification, even when elementary scores wrongfully omit confounders or focus on different outcomes. We illustrate our methods in a setting of treatments for asthma.

Controlling Confounding in a Study of Oral Anticoagulants: Comparing Disease Risk Scores Developed Using Different Follow-Up Approaches

Purpose:

Little is known about how disease risk score (DRS) development should proceed under different pharmacoepidemiologic follow-up strategies. In an analysis of dabigatran vs. warfarin and risk of major bleeding, we compared the results of DRS adjustment when models were developed under “intention-to-treat” (ITT) and “as-treated” (AT) approaches.

Methods:

We assessed DRS model discrimination, calibration, and ability to induce prognostic balance via the “dry run analysis”. AT treatment effects stratified on each DRS were compared with each other and with a propensity score (PS) stratified reference estimate. Bootstrap resampling of the historical cohort at 10 percent–90 percent sample size was performed to assess the impact of sample size on DRS estimation.

Results:

Historically-derived DRS models fit under AT showed greater decrements in discrimination and calibration than those fit under ITT when applied to the concurrent study population. Prognostic balance was approximately equal across DRS models (–6 percent to –7 percent “pseudo-bias” on the hazard ratio scale). Hazard ratios were between 0.76 and 0.78 with all methods of DRS adjustment, while the PS stratified hazard ratio was 0.83. In resampling, AT DRS models showed more overfitting and worse prognostic balance, and led to hazard ratios further from the reference estimate than did ITT DRSs, across sample sizes.

Conclusions:

In a study of anticoagulant safety, DRSs developed under an AT principle showed signs of overfitting and reduced confounding control. More research is needed to determine if development of DRSs under ITT is a viable solution to overfitting in other settings.

Marginal Structural Models for Risk or Prevalence Ratios for a Point Exposure Using a Disease Risk Score

The disease risk score is a summary score that can be used to control for confounding with a potentially large set of covariates. While less widely used than the exposure propensity score, the disease risk score approach might be useful for novel or unusual exposures, when treatment indications or exposure patterns are rapidly changing, or when more is known about the nature of how covariates cause disease than is known about factors influencing propensity for the exposure of interest. Focusing on the simple case of a binary point exposure, we describe a marginal structural model for estimation of risk (or prevalence) ratios. The proposed model incorporates the disease risk score as an offset in a regression model, and it yields an estimate of a standardized risk ratio where the target population is the exposed group. Simulations are used to illustrate the approach, and an empirical example is provided. Confounder control based on the proposed method might be a useful alternative to approaches based on the exposure propensity score, or as a complement to them.

Use of propensity score and disease risk score for multiple treatments with time-to-event outcome: a simulation study

Propensity score (PS) and disease risk score (DRS) are often used in pharmacoepidemiologic safety studies. Methods of applying these two balancing scores are extensively studied in binary treatment settings. However, the use of PS and DRS is not well understood in the case of non-ordinal multiple treatments. Some PS methods of multiple treatments have been implemented since the theoretical establishment. Nevertheless, most of the work applies to continuous or binary outcomes. Little work has been done for time-to-event outcomes. In this study, we extend the application of the PS and DRS methods to time-to-event outcomes in multiple treatment settings. The analytical approaches include weighing, matching, stratification, and regression. Simulation studies with rare event rates are conducted to evaluate the performances of different methods. Different treatment-covariates and outcome-covariates strength of associations are considered. Additionally, the impacts of imbalanced designs and large or limited PS overlaps are investigated on various analytical approaches. We found that the inverse probability treatment weighting with bootstrap variance estimator, the generalized PS matching, and the Cox regression estimated DRS in full cohort generally performed well in multiple treatment settings. This study aims to provide additional guidance for researchers on PS and DRS analyses in pharmacoepidemiologic observational studies.

Outcomes in human immunodeficiency virus-infected recipients of heart transplants

BACKGROUND

With the advent of combined antiretroviral therapy (cART), growing evidence has shown human immunodeficiency virus (HIV) may no longer be an absolute contraindication for solid organ transplantation. This study compares outcomes of heart transplantations between HIV-positive and HIV-negative recipients using SRTR transplant registry data.

METHODS

Patient survival, overall graft survival and death-censored graft survival were compared between HIV-positive and HIV-negative recipients. Multivariate Cox regression and Cox regression with a disease risk score (DRS) methodology were used to estimate the adjusted hazard ratios among heart transplant recipients (HTRs).

RESULTS

In total, 35 HTRs with HIV+ status were identified. No significant differences were found in patient survival (88% vs 77%; P = 0.1493), overall graft survival (85% vs 76%; P = 0.2758), and death-censored graft survival (91% vs 91%; P = 0.9871) between HIV-positive and HIV-negative HTRs in 5-year follow-up. No significant differences were found after adjusting for confounders.

CONCLUSIONS

This study supports the use of heart transplant procedures in selected HIV-positive patients. This study suggests that HIV-positive status is not a contraindication for life-saving heart transplant as there were no differences in graft, patient survival.

Association of Oral Anticoagulants and Proton Pump Inhibitor Cotherapy With Hospitalization for Upper Gastrointestinal Tract Bleeding

IMPORTANCE

Anticoagulant choice and proton pump inhibitor (PPI) cotherapy could affect the risk of upper gastrointestinal tract bleeding, a frequent and potentially serious complication of oral anticoagulant treatment.

OBJECTIVES

To compare the incidence of hospitalization for upper gastrointestinal tract bleeding in patients using individual anticoagulants with and without PPI cotherapy, and to determine variation according to underlying gastrointestinal bleeding risk.

DESIGN, SETTING, AND PARTICIPANTS

Retrospective cohort study in Medicare beneficiaries between January 1, 2011, and September 30, 2015.

EXPOSURES

Apixaban, dabigatran, rivaroxaban, or warfarin with or without PPI cotherapy.

MAIN OUTCOMES AND MEASURES

Hospitalizations for upper gastrointestinal tract bleeding: adjusted incidence and risk difference (RD) per 10 000 person-years of anticoagulant treatment, incidence rate ratios (IRRs).

RESULTS

There were 1 643 123 patients with 1 713 183 new episodes of oral anticoagulant treatment included in the cohort (mean [SD] age, 76.4 [2.4] years, 651 427 person-years of follow-up [56.1%] were for women, and the indication was atrial fibrillation for 870 330 person-years [74.9%]). During 754 389 treatment person-years without PPI cotherapy, the adjusted incidence of hospitalization for upper gastrointestinal tract bleeding (n = 7119) was 115 per 10 000 person-years (95% CI, 112-118). The incidence for rivaroxaban (n = 1278) was 144 per 10 000 person-years (95% CI, 136-152), which was significantly greater than the incidence of hospitalizations for apixaban (n = 279; 73 per 10 000 person-years; IRR, 1.97 [95% CI, 1.73-2.25]; RD, 70.9 [95% CI, 59.1-82.7]), dabigatran (n = 629; 120 per 10 000 person-years; IRR, 1.19 [95% CI, 1.08-1.32]; RD, 23.4 [95% CI, 10.6-36.2]), and warfarin (n = 4933; 113 per 10 000 person-years; IRR, 1.27 [95% CI, 1.19-1.35]; RD, 30.4 [95% CI, 20.3-40.6]). The incidence for apixaban was significantly lower than that for dabigatran (IRR, 0.61 [95% CI, 0.52-0.70]; RD, -47.5 [95% CI,-60.6 to -34.3]) and warfarin (IRR, 0.64 [95% CI, 0.57-0.73]; RD, -40.5 [95% CI, -50.0 to -31.0]). When anticoagulant treatment with PPI cotherapy (264 447 person-years; 76 per 10 000 person-years) was compared with treatment without PPI cotherapy, risk of upper gastrointestinal tract bleeding hospitalizations (n = 2245) was lower overall (IRR, 0.66 [95% CI, 0.62-0.69]) and for apixaban (IRR, 0.66 [95% CI, 0.52-0.85]; RD, -24 [95% CI, -38 to -11]), dabigatran (IRR, 0.49 [95% CI, 0.41-0.59]; RD, -61.1 [95% CI, -74.8 to -47.4]), rivaroxaban (IRR, 0.75 [95% CI, 0.68-0.84]; RD, -35.5 [95% CI, -48.6 to -22.4]), and warfarin (IRR, 0.65 [95% CI, 0.62-0.69]; RD, -39.3 [95% CI, -44.5 to -34.2]).

CONCLUSIONS AND RELEVANCE

Among patients initiating oral anticoagulant treatment, incidence of hospitalization for upper gastrointestinal tract bleeding was the highest in patients prescribed rivaroxaban, and the lowest for patients prescribed apixaban. For each anticoagulant, the incidence of hospitalization for upper gastrointestinal tract bleeding was lower among patients who were receiving PPI cotherapy. These findings may inform assessment of risks and benefits when choosing anticoagulant agents.

Extension of Disease Risk Score-Based Confounding Adjustments for Multiple Outcomes of Interest: An Empirical Evaluation

Use of disease risk score (DRS)-based confounding adjustment when estimating treatment effects on multiple outcomes is not well studied. We designed an empirical cohort study to compare dabigatran initiators and warfarin initiators with respect to risks of ischemic stroke and major bleeding in 12 sequential monitoring periods (90 days each), using data from the Truven Marketscan database (Truven Health Analytics, Ann Arbor, Michigan). We implemented 2 approaches to combine DRS for multiple outcomes: 1) 1:1 matching on prognostic propensity scores (PPS), created using DRS for bleeding and stroke as independent variables in a propensity score (PS) model; and 2) simultaneous 1:1 matching on DRS for bleeding and stroke using Mahalanobis distance (M-distance), and compared their performance with that of traditional PS matching. M-distance matching appeared to produce more stable results in the early marketing period than both PPS and traditional PS matching; hazard ratios from unadjusted analysis, traditional PS matching, PPS matching, and M-distance matching after 4 periods were 0.72 (95% confidence interval (CI): 0.51, 1.03), 0.61 (95% CI: 0.31, 1.09), 0.55 (95% CI: 0.33, 0.91), and 0.78 (95% CI: 0.45, 1.34), respectively, for stroke and 0.65 (95% CI: 0.53, 0.80), 0.78 (95% CI: 0.60, 1.01), 0.75 (95% CI: 0.59, 0.96), and 0.78 (95% CI: 0.64, 0.95), respectively, for bleeding. In later periods, estimates were similar for traditional PS matching and M-distance matching but suggested potential residual confounding with PPS matching. These results suggest that M-distance matching may be a valid approach for extension of DRS-based confounding adjustments for multiple outcomes of interest.

Optimal duration for continuation of statin therapy in bacteremic patients

Background

Evidence suggests statins may improve survival in patients with bloodstream infections. However, there is no consensus on optimal timing and duration of exposure.

Objectives

To quantify statin therapy duration associated with decreased mortality in bacteremic statin users.

Methods

We conducted a case-control study using OptumClinformatics™ with matched Premier hospital data (1 October 2009-31 March 2013). Cases who died during the hospitalization were matched 1:1 to survivors on disease risk scores (DRSs). Post-admission statin therapy duration was evaluated in patients with at least 90 days of pre-admission continuous statin use. Classification and regression tree (CART) analysis was conducted to identify the optimal duration of statin continuation which provided the lowest inpatient mortality. Logistic regression was used to calculate the odds of mortality.

Results

We included 58 DRS matched pairs of cases and controls: 47 patients (41%) continued statin therapy during the hospital admission, 15 (32%) cases and 32 (68%) controls. The CART analysis partitioned the continuation of statin therapy at ⩾2 days, representing lower mortality for patients who continued statins for 2 days or more and higher mortality for patients who did not continue or remained on statins for only 1 day. Inpatient mortality was 76% lower among those with at least 2 days of continued statin use (odds ratio 0.24, 95% confidence interval 0.11-0.55).

Conclusion

Among matched cases and controls with at least 90 days of baseline statin use prior to the admission, the continuation of statins for at least 2 days after admission demonstrated a survival benefit among bacteremic patients.

Second-generation anti-depressants and risk of new-onset seizures in the elderly

BACKGROUND

Anti-depressants are among the most widely-prescribed medications. It is unknown whether the risk of seizure during therapeutic use differs by drug. We ranked the seizure risk of popular anti-depressants.

METHODS

We conducted a population-based case-control study between April 2002 and March 2015 in Ontario, Canada. Cases were Ontario residents aged ≥65 years hospitalized for a first-ever seizure within 60 d of filling a prescription for one of nine second-generation anti-depressants, each dispensed more than 1 million times (range: 1,196,810 [fluvoxamine] to 19,849,930 [citalopram]) during the study period. For each case, we identified up to four seizure-free controls receiving a similar anti-depressant, and matched on age, sex, date and a pre-defined seizure-specific disease risk index.

RESULTS

We identified 5701 patients hospitalized with a first-ever seizure and matched them with 21,872 controls. Relative to bupropion, the risk of new-onset seizure during therapeutic use was highest for escitalopram (adjusted odds ratio [OR] 1.79; 95% confidence interval [CI] 1.42-2.25) and citalopram (OR 1.67; 95% CI 1.35-2.07), while no incremental risk was found for fluoxetine (OR 1.02; 95%CI 0.78-1.33) and duloxetine (OR 0.94; 95%CI 0.75-1.22). Other anti-depressants were associated with modest increase in seizure risk.

CONCLUSIONS

The risk of seizure during therapeutic use among elderly patients varies among second-generation anti-depressants. Escitalopram and citalopram are associated with the highest risk. Prescribers should consider the seizure risk of individual anti-depressants and use discretion when selecting an anti-depressant, especially for patients with other risk factors for seizure. Frontline clinicians should be cognizant of this differential risk.

Use of antipsychotics and risk of venous thromboembolism in postmenopausal women

Despite continued uncertainty of venous thromboembolism (VTE) caused from antipsychotic agents, this safety issue has not been examined in postmenopausal women, a population with high usages of antipsychotics and at high risk for VTE. We assessed whether antipsychotic use was associated with an increased VTE risk in women after menopause. We conducted a nested case-control study of all Taiwanese women aged ≥ 50 years (n = 316,132) using a nationwide healthcare claims database between 2000 and 2011. All newly diagnosed VTE patients treated with an anticoagulant or thrombectomy surgery were identified as cases (n = 2,520) and individually matched to select controls (n = 24,223) by cohort entry date, age, cancer diagnosis and major surgery procedure. The odds ratios (ORs) and 95 % confidence interval (CI) of VTE associated with antipsychotics were estimated by multivariate conditional logistic regressions. Current use of antipsychotics was associated with a 1.90-fold (95 % CI = 1.64–2.19) increased VTE risk compared with nonuse in postmenopausal women. The VTE risk existed in a dose-dependent fashion (test for trend, p<0.001), with a more than quadrupled risk for high-dose antipsychotics (adjusted OR = 4.60; 95 % CI = 2.88–7.33). Current parenteral administration of antipsychotics also led to a 3.46-fold increased risk (95 % CI = 2.39–5.00). Conversely, there was no increased VTE risk when antipsychotics were discontinued for > 30 days. In conclusion, current use of antipsychotics is significantly associated with a dose-dependent increased risk of VTE in postmenopausal women, especially for those currently taking high-dose or receiving parenteral antipsychotics.

The "Dry-Run" Analysis: A Method for Evaluating Risk Scores for Confounding Control

A propensity score (PS) model's ability to control confounding can be assessed by evaluating covariate balance across exposure groups after PS adjustment. The optimal strategy for evaluating a disease risk score (DRS) model's ability to control confounding is less clear. DRS models cannot be evaluated through balance checks within the full population, and they are usually assessed through prediction diagnostics and goodness-of-fit tests. A proposed alternative is the "dry-run" analysis, which divides the unexposed population into "pseudo-exposed" and "pseudo-unexposed" groups so that differences on observed covariates resemble differences between the actual exposed and unexposed populations. With no exposure effect separating the pseudo-exposed and pseudo-unexposed groups, a DRS model is evaluated by its ability to retrieve an unconfounded null estimate after adjustment in this pseudo-population. We used simulations and an empirical example to compare traditional DRS performance metrics with the dry-run validation. In simulations, the dry run often improved assessment of confounding control, compared with the C statistic and goodness-of-fit tests. In the empirical example, PS and DRS matching gave similar results and showed good performance in terms of covariate balance (PS matching) and controlling confounding in the dry-run analysis (DRS matching). The dry-run analysis may prove useful in evaluating confounding control through DRS models.

Diffusion of Innovations model helps interpret the comparative uptake of two methodological innovations: co-authorship network analysis and recommendations for the integration of novel methods in practice

OBJECTIVE

The objective of this study was to characterize the diffusion of methodological innovation.

STUDY DESIGN AND SETTING

Comparative case study analysis of the diffusion of two methods that summarize confounder information into a single score: disease risk score (DRS) and high-dimensional propensity score (hdPS). We completed systematic searches to identify DRS and hdPS papers in the field of pharmacoepidemiology through to the end of 2013, plotted the number of papers and unique authors over time, and created sociograms and animations to visualize co-authorship networks. First and last author affiliations were used to ascribe institutional contributions to each paper and network.

RESULTS

We identified 43 DRS papers by 153 authors since 1981, reflecting slow uptake during initial periods of uncertainty and broader diffusion since 2001 linked to early adopters from Vanderbilt. We identified 44 hdPS papers by 147 authors since 2009, reflecting rapid and integrated diffusion, likely facilitated by opinion leaders, early presentation at conferences, easily accessible statistical code, and improvement in funding. Most contributions (87% DRS, 96% hdPS) were from North America.

CONCLUSION

When proposing new methods, authors are encouraged to consider innovation attributes and early evaluation to improve knowledge translation of their innovations for integration into practice, and we provide recommendations for consideration.

Risk Factors for Late-Onset Group B Streptococcal Disease Before and After Implementation of Universal Screening and Intrapartum Antibiotic Prophylaxis

BACKGROUND

It is unclear whether risk factors for late-onset Group B Streptococcus disease (LOD) have changed since the introduction of universal screening and treatment in 2002.

METHODS

We conducted a case-control study using linked birth certificates and hospital discharge records. All infants born in Washington State from 1992 to 2011 and hospitalized between 7 and 89 days of life with a Group B Streptococcus (GBS)-related International Classification of Diseases (ICD)-9 code were included. Controls were matched 4:1 by birth year. Multivariate logistic regression was used to evaluate the association between clinical characteristics and LOD. We compared differences in the effect of risk factors on LOD between infants born before and after 2002 using likelihood ratio tests.

RESULTS

We identified 138 cases of LOD. In multivariate analyses, prematurity and young maternal age were significantly associated with risk of LOD throughout the study period; positive GBS screen was associated with LOD from 2003 to 2011. Each week of decreasing gestation was associated with a 1.24 (95% confidence interval: 1.15-1.35) times greater likelihood of LOD. We did not detect differences in the association between prematurity or young maternal age and LOD comparing infants born before and after 2002. Compared with infants of non-Hispanic white mothers, risk of LOD among infants of non-Hispanic black mothers decreased after 2002 (adjusted odds ratio [aOR] = 2.74 vs 0.64; p_interaction = 0.02), whereas risk of LOD among infants of Hispanic mothers increased (aOR = 0.80 vs 2.23; p_interaction ≤ 0.001).

CONCLUSIONS

Our results confirm studies conducted before 2002, which found that prematurity and young maternal age were associated with increased risk of LOD. Ethnicity-associated LOD risk differed before and after 2002, which may be related to healthcare access.

"Threshold-crossing": A Useful Way to Establish the Counterfactual in Clinical Trials?

A central question in the assessment of benefit/harm of new treatments is: how does the average outcome on the new treatment (the factual) compare to the average outcome had patients received no treatment or a different treatment known to be effective (the counterfactual)? Randomized controlled trials (RCTs) are the standard for comparing the factual with the counterfactual. Recent developments necessitate and enable a new way of determining the counterfactual for some new medicines. For select situations, we propose a new framework for evidence generation, which we call "threshold-crossing." This framework leverages the wealth of information that is becoming available from completed RCTs and from real world data sources. Relying on formalized procedures, information gleaned from these data is used to estimate the counterfactual, enabling efficacy assessment of new drugs. We propose future (research) activities to enable "threshold-crossing" for carefully selected products and indications in which RCTs are not feasible.

Estimation of conditional and marginal odds ratios using the prognostic score

Introduced by Hansen in 2008, the prognostic score (PGS) has been presented as 'the prognostic analogue of the propensity score' (PPS). PPS-based methods are intended to estimate marginal effects. Most previous studies evaluated the performance of existing PGS-based methods (adjustment, stratification and matching using the PGS) in situations in which the theoretical conditional and marginal effects are equal (i.e., collapsible situations). To support the use of PGS framework as an alternative to the PPS framework, applied researchers must have reliable information about the type of treatment effect estimated by each method. We propose four new PGS-based methods, each developed to estimate a specific type of treatment effect. We evaluated the ability of existing and new PGS-based methods to estimate the conditional treatment effect (CTE), the (marginal) average treatment effect on the whole population (ATE), and the (marginal) average treatment effect on the treated population (ATT), when the odds ratio (a non-collapsible estimator) is the measure of interest. The performance of PGS-based methods was assessed by Monte Carlo simulations and compared with PPS-based methods and multivariate regression analysis. Existing PGS-based methods did not allow for estimating the ATE and showed unacceptable performance when the proportion of exposed subjects was large. When estimating marginal effects, PPS-based methods were too conservative, whereas the new PGS-based methods performed better with low prevalence of exposure, and had coverages closer to the nominal value. When estimating CTE, the new PGS-based methods performed as well as traditional multivariate regression. Copyright © 2016 John Wiley & Sons, Ltd.

Performance of Disease Risk Score Matching in Nested Case-Control Studies: A Simulation Study

In a case-control study, matching on a disease risk score (DRS), which includes many confounders, should theoretically result in greater precision than matching on only a few confounders; however, this has not been investigated. We simulated 1,000 hypothetical cohorts with a binary exposure, a time-to-event outcome, and 13 covariates. Each cohort comprised 2 subcohorts of 10,000 patients each: a historical subcohort and a concurrent subcohort. DRS were estimated in the historical subcohorts and applied to the concurrent subcohorts. Nested case-control studies were conducted in the concurrent subcohorts using incidence density sampling with 2 strategies-matching on age and sex, with adjustment for additional confounders, and matching on DRS-followed by conditional logistic regression for 9 outcome-exposure incidence scenarios. In all scenarios, DRS matching yielded lower average standard errors and mean squared errors than did matching on age and sex. In 6 scenarios, DRS matching also resulted in greater empirical power. DRS matching resulted in less relative bias than did matching on age and sex at lower outcome incidences but more relative bias at higher incidences. Post-hoc analysis revealed that the effect of DRS model misspecification might be more pronounced at higher outcome incidences, resulting in higher relative bias. These results suggest that DRS matching might increase the statistical efficiency of case-control studies, particularly when the outcome is rare.

Maternal Tenofovir Disoproxil Fumarate Use in Pregnancy and Growth Outcomes among HIV-Exposed Uninfected Infants in Kenya

BACKGROUND

Tenofovir disoproxil fumarate (TDF) is commonly used in antiretroviral treatment (ART) and preexposure prophylaxis regimens. We evaluated the relationship of prenatal TDF use and growth outcomes among Kenyan HIV-exposed uninfected (HEU) infants.

MATERIALS AND METHODS

We included PCR-confirmed HEU infants enrolled in a cross-sectional survey of mother-infant pairs conducted between July and December 2013 in Kenya. Maternal ART regimen during pregnancy was determined by self-report and clinic records. Six-week and 9-month z-scores for weight-for-age (WAZ), weight-for-length (WLZ), length-for-age (LAZ), and head circumference-for-age (HCAZ) were compared among HEU infants with and without TDF exposure using t-tests and multivariate linear regression models.

RESULTS

Among 277 mothers who received ART during pregnancy, 63% initiated ART before pregnancy, of which 89 (32%) used TDF. No differences in birth weight (3.0 kg versus 3.1 kg, p = 0.21) or gestational age (38 weeks versus 38 weeks, p = 0.16) were detected between TDF-exposed and TDF-unexposed infants. At 6 weeks, unadjusted mean WAZ was lower among TDF-exposed infants (-0.8 versus -0.4, p = 0.03), with a trend towards association in adjusted analyses (p = 0.06). There were no associations between prenatal TDF use and WLZ, LAZ, and HCAZ in 6-week or 9-month infant cohorts.

CONCLUSION

Maternal TDF use did not adversely affect infant growth compared to other regimens.

Performance of the disease risk score in a cohort study with policy-induced selection bias

AIM

To examine the performance of the disease risk score (DRS) in a cohort study with evidence of policy-induced selection bias.

METHODS

We examined two cohorts of new users of bisphosphonates. Estimates for 1-year hip fracture rates between agents using DRS, exposure propensity scores and traditional multivariable analysis were compared.

RESULTS

The results for the cohort with no evidence of policy-induced selection bias showed little variation across analyses (-4.1-2.0%). Analysis of the cohort with evidence of policy-induced selection bias showed greater variation (-13.5-8.1%), with the greatest difference seen with DRS analyses.

CONCLUSION

Our findings suggest that caution may be warranted when using DRS methods in cohort studies with policy-induced selection bias, further research is needed.

Atypical antipsychotics olanzapine, quetiapine, and risperidone and risk of acute major cardiovascular events in young and middle-aged adults: a nationwide register-based cohort study in Denmark

BACKGROUND

A number of serious cardiovascular safety concerns related to the use of atypical antipsychotics, compared with no use, have emerged, but nearly all reports are from studies of older patients. We aimed to compare the risk of cardiovascular events between the three most commonly used atypical antipsychotics in young and middle-aged adults.

METHODS

We conducted a nationwide register-based cohort study in Denmark, 1997-2011, including adults aged 18-64 years, who started treatment with oral or intramuscular olanzapine (n = 15,774), oral quetiapine (n = 18,717), and oral or intramuscular risperidone (n = 14,134). The primary outcome was any major cardiovascular event (composite of cardiovascular mortality, acute coronary syndrome, or ischemic stroke) within 1 year following treatment initiation. Cox regression was used to estimate hazard ratios (HRs) while on current antipsychotic monotherapy in the outpatient setting, adjusting for an outcome-specific disease risk score.

RESULTS

The crude rate of any major cardiovascular event was 5.3 per 1,000 person-years among olanzapine users, 3.4 in quetiapine users, and 5.2 in risperidone users. Compared with risperidone, the risk of any major cardiovascular event was not significantly different in olanzapine users (HR 0.90, 95 % confidence interval [CI] 0.53-1.52) and quetiapine users (HR 0.79, 95 % CI 0.45-1.39). The absolute risk difference per 1,000 person-years on treatment was -0.5 (95 % CI -2.4 to 2.7) events for olanzapine and -1.1 (95 % CI -2.9 to 2.0) events for quetiapine.

CONCLUSIONS

Among young and middle-aged outpatients, the risk of acute major cardiovascular events was similar with use of olanzapine, quetiapine, and risperidone. Although moderate relative differences cannot be ruled out, any differences are small in absolute terms.

Reducing Bias Amplification in the Presence of Unmeasured Confounding Through Out-of-Sample Estimation Strategies for the Disease Risk Score

The prognostic score, or disease risk score (DRS), is a summary score that is used to control for confounding in non-experimental studies. While the DRS has been shown to effectively control for measured confounders, unmeasured confounding continues to be a fundamental obstacle in non-experimental research. Both theory and simulations have shown that in the presence of unmeasured confounding, controlling for variables that affect treatment (both instrumental variables and measured confounders) amplifies the bias caused by unmeasured confounders. In this paper, we use causal diagrams and path analysis to review and illustrate the process of bias amplification. We show that traditional estimation strategies for the DRS do not avoid bias amplification when controlling for predictors of treatment. We then discuss estimation strategies for the DRS that can potentially reduce bias amplification that is caused by controlling both instrumental variables and measured confounders. We show that under certain assumptions, estimating the DRS in populations outside the defined study cohort where treatment has not been introduced, or in outside populations with reduced treatment prevalence can control for the confounding effects of measured confounders while at the same time reduce bias amplification.