Brandon Koch’s research while affiliated with University of Nevada, Reno and other places

Treatment effect heterogeneity occurs when individual characteristics influence the effect of a treatment. We propose a novel approach that combines prognostic score matching and conditional inference trees to characterize effect heterogeneity of a randomized binary treatment. One key feature that distinguishes our method from alternative approaches is that it controls the Type I error rate, that is, the probability of identifying effect heterogeneity if none exists and retains the underlying subgroups. This feature makes our technique particularly appealing in the context of clinical trials, where there may be significant costs associated with erroneously declaring that effects differ across population subgroups. Treatment effect heterogeneity trees are able to identify heterogeneous subgroups, characterize the relevant subgroups and estimate the associated treatment effects. We demonstrate the efficacy of the proposed method using a comprehensive simulation study and illustrate our method using a nutrition trial dataset to evaluate effect heterogeneity within a patient population.

It is unclear how broadly aware parents are of the concept of herd immunity and whether parents consider community benefits of vaccination when making decisions about their child’s vaccinations. We aimed to determine whether educating parents about community-level benefits of measles, mumps, and rubella (MMR) vaccination and local vaccination rates would impact concern about their child’s risk of measles and risk of a measles outbreak. We conducted an electronic survey among Minnesota parents of children aged 6–18 years in August 2016. We assessed baseline knowledge of herd immunity, asked participants to estimate MMR vaccination coverage in their county, and asked participants to estimate the minimum coverage needed to prevent measles outbreaks. We then delivered a short, educational intervention via the survey to inform participants about the benefits of herd immunity, the actual MMR vaccination coverage in their county, and that at least 95% MMR vaccination coverage is needed to prevent measles outbreaks. Pre- and post-intervention, participants were asked to report how concerned they were that their child might get measles. We used logistic regression models to assess factors associated with awareness of herd immunity, change in concern about one’s child’s measles risk, and overall concern for a measles outbreak. Among 493 participants, 67.8% were aware of herd immunity at baseline. Post-intervention, 40.2% (n = 198) of parents learned that MMR vaccination rates in their county were higher than they expected. All participants found out that their county MMR rates were lower than the measles herd immunity threshold of 95%. Overall, 27.0% (n = 133) of participants reported an increase in concern that their child might get measles after learning about local vaccination coverage and the coverage needed to achieve herd immunity. We found that our short, educational intervention aimed to increase awareness about herd immunity and local vaccination led to an increase in concern about disease risk among less than a third of parents.

Unbiased estimation of causal effects with observational data requires adjustment for confounding variables that are related to both the outcome and treatment assignment. Standard variable selection techniques aim to maximize predictive ability of the outcome model, but they ignore covariate associations with treatment and may not adjust for important confounders weakly associated to outcome. We propose a novel method for estimating causal effects that simultaneously considers models for both outcome and treatment, which we call the bilevel spike and slab causal estimator (BSSCE). By using a Bayesian formulation, BSSCE estimates the posterior distribution of all model parameters and provides straightforward and reliable inference. Spike and slab priors are used on each covariate coefficient which aim to minimize the mean squared error of the treatment effect estimator. Theoretical properties of the treatment effect estimator are derived justifying the prior used in BSSCE. Simulations show that BSSCE can substantially reduce mean squared error over numerous methods and performs especially well with large numbers of covariates, including situations where the number of covariates is greater than the sample size. We illustrate BSSCE by estimating the causal effect of vasoactive therapy vs. fluid resuscitation on hypotensive episode length for patients in the Multiparameter Intelligent Monitoring in Intensive Care III critical care database.

Treatment effect heterogeneity occurs when individual characteristics influence the effect of a treatment. We propose a novel, intuitive approach that combines prognostic score matching and conditional inference trees to characterize effect heterogeneity of a randomized treatment. One key feature that distinguishes our method from alternative approaches is that it controls the Type I error rate, i.e., the probability of identifying effect heterogeneity if none exists. This feature makes our technique particularly appealing in the context of clinical trials, where there may be significant costs associated with erroneously declaring that effects differ across population subgroups.

Background Maintaining high coverage of measles, mumps, and rubella (MMR) vaccination is important for preventing outbreaks and maintaining herd immunity (HI), which benefits both individuals and communities. We aimed to determine whether information about the benefits of HI and local MMR vaccination rates could change a parent’s concern about their child’s risk of contracting measles. Methods We conducted a survey at the 2016 Minnesota State Fair among Minnesota residents ≥18 years who had at least one child aged 6–18 years. Participants were asked to choose the correct definition of HI, to estimate the MMR vaccination coverage in their county, and guess the minimum MMR vaccination coverage needed to prevent measles outbreaks. We delivered an educational intervention through the interactive survey informing participants about the benefits of herd immunity, the actual MMR coverage in their county, and that ≥95% coverage is needed to prevent outbreaks. Before and after the educational intervention, participants were asked to report their level of concern about their child contracting measles. We calculated adjusted predicted percentages from logistic regression models to evaluate changes in concern about risk pre- and post-intervention and to assess factors associated with concern about measles. Results Among the 493 participants, 92.7% reported vaccinating their child with MMR, though one third were not familiar with HI. Prior to receiving information, those knowledgeable about HI were significantly more likely to be concerned about their child getting measles (predicted percentage 80.2% [95% CI: 75.7–84.6]) than those who were unfamiliar with HI (predicted percentage 69.8% [95% CI: 62.1–77.5]), P-value for the difference = 0.027. Participants believed that MMR vaccination was, on average, 9.0% [95% CI: 6.9–11.0] lower than the actual coverage in their local area. Conclusion Information about HI and local vaccination coverage rates did not impact parental concern about their child being at risk for getting measles. Overall, parents learned that local MMR vaccination rates were higher than they had expected. Disclosures All authors: No reported disclosures.

Background: Vaccination protects individuals directly and communities indirectly by reducing transmission. We aimed to determine whether information about herd immunity and local vaccination coverage could change an individual's vaccination plans and concern about influenza. Methods: We surveyed Minnesota residents ≥18 years during the 2016 Minnesota State Fair. Participants were asked to identify the definition of herd immunity, to report their history of and plans to receive influenza vaccine, to report their concern about influenza, and to estimate the reported influenza vaccination coverage in their county. After providing educational information about herd immunity and local vaccination rates, we reassessed vaccination plans and concerns. We used logistic regression to estimate predicted percentages for those willing to be vaccinated, for concern about influenza, and for changes in these outcomes after the intervention. We then compared those individuals with and without prior knowledge of herd immunity, accounting for other characteristics. Results: Among 554 participants, the median age was 57 years; most were female (65.9%), white (91.0%), and non-Hispanic/Latino (93.9%). Overall, 37.2% of participants did not know about herd immunity and 75.6% thought that the influenza vaccination coverage in their county was higher than it was reported. Those not knowledgeable about herd immunity were significantly less likely than those knowledgeable about the concept to report plans to be vaccinated at baseline (67.8% versus 78.9%; p = 0.004). After learning about herd immunity and influenza vaccination coverage, the proportion of those not knowledgeable about herd immunity who were willing to be vaccinated increased significantly by 7.3 percentage points (p = 0.001). Educating participants eliminated the significant difference in the proportion planning to be vaccinated between these two groups (80.1% of those knowledgeable and 75.1% of those who were not initially knowledgeable became willing; p = 0.148). Conclusions: Education about herd immunity and local vaccination coverage could be a useful tool for increasing willingness to vaccinate, generating benefits both to individuals and communities.

Background: HPV vaccination uptake remains lower than other recommended adolescent vaccines in the United States. Parental attitudes are important predictors of vaccine uptake, yet little is known about how they have changed over time. Methods: Participants included U.S. residents aged 13-17 years with documented vaccination status who were unvaccinated (had not initiated) or undervaccinated (initiated, but had not completed) with the HPV vaccination series whose parents responded to the National Immunization Surveys-Teen, 2010-2015. Results: Of the 76,971 participants, 63.0% were male, 58.8% were non-Hispanic white, and 14.4 years was the median age. The percentage of unvaccinated teens decreased 2010-2015, yet, annually, parents of unvaccinated teens of both sexes most often reported that they were "not likely at all" to vaccinate their teen. The percentage decreased significantly from 41.5% to 31.2% (P<.001) for parents of unvaccinated females from 2010-2015 but did not change among parents of males from 2012-2015. Conversely, parents of undervaccinated teens of both sexes reported higher and increasing vaccination intent over time. In 2015, nearly one third of parents of unvaccinated teens reported that their low intent stemmed from the belief that the vaccine was "not needed/necessary." Concerns about vaccine safety and side effects declined over time among parents of unvaccinated females but increased among parents of males (7.3% to 14.8%; P<.001). Conclusions: Although parental intent and knowledge increased over time, lack of perceived importance of HPV vaccination and parental concerns about vaccine safety remain barriers to both HPV vaccination series initiation and completion in the U.S.

The efficiency of doubly robust estimators of the average causal effect (ACE) of a treatment can be improved by including in the treatment and outcome models only those covariates which are related to both treatment and outcome (i.e., confounders) or related only to the outcome. However, it is often challenging to identify such covariates among the large number that may be measured in a given study. In this article, we propose GLiDeR (Group Lasso and Doubly Robust Estimation), a novel variable selection technique for identifying confounders and predictors of outcome using an adaptive group lasso approach that simultaneously performs coefficient selection, regularization, and estimation across the treatment and outcome models. The selected variables and corresponding coefficient estimates are used in a standard doubly robust ACE estimator. We provide asymptotic results showing that, for a broad class of data generating mechanisms, GLiDeR yields a consistent estimator of the ACE when either the outcome or treatment model is correctly specified. A comprehensive simulation study shows that GLiDeR is more efficient than doubly robust methods using standard variable selection techniques and has substantial computational advantages over a recently proposed doubly robust Bayesian model averaging method. We illustrate our method by estimating the causal treatment effect of bilateral versus single-lung transplant on forced expiratory volume in one year after transplant using an observational registry.