Methods utilizing instrumental variables have been a fundamental statistical approach to estimation in the presence of unmeasured confounding, usually occurring in non-randomized observational data common to fields such as economics and public health. However, such methods usually make constricting linearity and additivity assumptions that are inapplicable to the complex modeling challenges of today. The growing body of observational data being collected will necessitate flexible regression modeling while also being able to control for confounding using instrumental variables. Therefore, this article presents a nonlinear instrumental variable regression model based on Bayesian regression tree ensembles to estimate such relationships, including interactions, in the presence of confounding. One exciting application of this method is to use genetic variants as instruments, known as Mendelian randomization. Body mass index is one factor that is hypothesized to have a nonlinear relationship with cardiovascular risk factors such as blood pressure while interacting with age. Heterogeneity in patient characteristics such as age could be clinically interesting from a precision medicine perspective where individualized treatment is emphasized. We present our flexible Bayesian instrumental variable regression tree method with an example from the UK Biobank where body mass index is related to blood pressure using genetic variants as the instruments.

Using high-dimensional genetic variants such as single nucleotide polymorphisms (SNP) to predict complex diseases and traits has important applications in basic research and other clinical settings. For example, predicting gene expression is a necessary first step to identify (putative) causal genes in transcriptome-wide association studies. Due to weak signals, high-dimensionality, and linkage disequilibrium (correlation) among SNPs, building such a prediction model is challenging. However, functional annotations at the SNP level (e.g., as epigenomic data across multiple cell- or tissue-types) are available and could be used to inform predictor importance and aid in outcome prediction. Existing approaches to incorporate annotations have been based mainly on (generalized) linear models. Bayesian additive regression trees (BART), in contrast, is a reliable method to obtain high-quality nonlinear out of sample predictions without overfitting. Unfortunately, the default prior from BART may be too inflexible to handle sparse situations where the number of predictors approaches or surpasses the number of observations. Motivated by our real data application, this article proposes an alternative prior based on the logit normal distribution because it provides a framework that is adaptive to sparsity and can model informative functional annotations. It also provides a framework to incorporate prior information about the between SNP correlations. Computational details for carrying out inference are presented along with the results from a simulation study and a genome-wide prediction analysis of the Alzheimer's Disease Neuroimaging Initiative data.

Objective: Use the RE-AIM framework to examine the implementation of a patient contextual data (PCD) Tool designed to share patients' needs, values, and preferences with care teams ahead of clinical encounters.Materials & Methods: Observational study that follows initial PCD Tool scaling across primary care at a Midwestern academic health network. Program invitations, enrollment, patient submissions, and clinician views were tracked over a 1-year study period. Logistic regression modeled the likelihood of using the PCD Tool, accounting for patient covariates.Results: Of 58,874 patients who could be contacted by email, 9,183 (15.6%) became PCD Tool users. Overall, 76% of primary care providers had patients who used the PCD Tool. Older age, female gender, non-minority race, patient portal activation, and Medicare coverage were significantly associated with increased likelihood of use. Number of office visits, medical issues, and behavioral health conditions also associated with use. Primary care staff viewed 18.7% of available PCD Tool summaries, 1.1% to 57.6% per clinic. Discussion: The intervention mainly reached non-minority patients and patients who used more health services. Given the requirement for an email address on file, some patients may have been underrepresented. Overall, patient reach and adoption and clinician adoption, implementation, and maintenance of this Tool were modest but stable, consistent with a non-directive approach to fostering adoption by introducing the Tool in the absence of clear expectations for use.Conclusion: Healthcare organizations must implement effective methods to increase the reach, adoption, implementation, and maintenance of PCD tools across all patient populations. Assisting people, particularly racial minorities, with PCD Tool registration and actively supporting clinician use are critical steps in implementing technology that facilitates care.

In this article, we introduce the BART R package which is an acronym for Bayesian additive regression trees. BART is a Bayesian nonparametric, machine learning, ensemble predictive modeling method for continuous, binary, categorical and time-to-event outcomes. Furthermore, BART is a tree-based, black-box method which fits the outcome to an arbitrary random function, f , of the covariates. The BART technique is relatively computationally efficient as compared to its competitors, but large sample sizes can be demanding. Therefore, the BART package includes efficient state-of-the-art implementations for continuous, binary, categorical and time-to-event outcomes that can take advantage of modern off-the-shelf hardware and software multi-threading technology. The BART package is written in C++ for both programmer and execution efficiency. The BART package takes advantage of multi-threading via forking as provided by the parallel package and OpenMP when available and supported by the platform. The ensemble of binary trees produced by a BART fit can be stored and re-used later via the R predict function. In addition to being an R package, the installed BART routines can be called directly from C++. The BART package provides the tools for your BART toolbox.

Precision medicine is an active area of research that could offer an analytic paradigm shift for clinical trials and the subsequent treatment decisions based on them. Clinical trials are typically analyzed with the intent of discovering beneficial treatments if the same treatment is applied to the entire population under study. But, such a treatment strategy could be suboptimal if subsets of the population exhibit varying treatment effects. Identifying subsets of the population experiencing differential treatment effect and forming individualized treatment rules is a task well‐suited to modern machine learning methods such as tree‐based ensemble predictive models. Specifically, Bayesian additive regression trees (BART) has shown promise in this regard because of its exceptional performance in out‐of‐sample prediction. Due to the unique inferential needs of precision medicine for clinical trials, we have proposed the BART extensions explicated here. We incorporate random effects for longitudinal repeated measures and subject clustering within medical centers. The addition of a novel interaction detection prior to identify treatment heterogeneity among clinical trial patients and its association with patient characteristics. These extensions are unified under a framework that we call mixedBART. Simulation studies and applications of precision medicine based on real randomized clinical trials data examples are presented.

Background Patient contextual data (PCD) are often missing from electronic health records, limiting the opportunity to incorporate preferences and life circumstances into care. Engaging patients through tools that collect and summarize such data may improve communication and patient activation. However, differential tool adoption by race might widen health care disparities. Objective Determine if a digital tool designed to collect and present PCD improves communication and patient activation; secondarily, evaluate if use impacts outcomes by race. Design, Setting, and Participants A pragmatic, two-armed, non-blinded, randomized controlled trial conducted during 2019 in a primary care setting. Intervention The PCD tool (PatientWisdom) invited patients to identify preferences, values, goals, and barriers to care. Patients were randomized to a standard pre-visit email or facilitated enrollment with dedicated outreach to encourage use of the tool. Main Outcomes and Measures Outcomes of interest were post-visit patient communication and patient activation measured by the Communication Assessment Tool (CAT) and Patient Activation Measure (PAM), respectively. Outcomes were evaluated using treatment-on-the-treated (TOT) and intention-to-treat (ITT) principles. Key Results A total of 301 patients were enrolled. Facilitated enrollment resulted in a five-fold increase in uptake of the PCD tool. TOT analysis indicated that the PCD tool was associated with notable increases in specific CAT items rated as excellent: “treated me with respect” (+ 13 percentage points; p = 0.04), “showed interest in my ideas” (+ 14 percentage points; p = 0.03), “showed care and concern” (+ 16 percentage points; p = 0.02), and “spent about the right amount of time with me” (+ 11 percentage points; p = 0.05). There were no significant pre/post-visit differences in PAM scores between arms (− 4.41 percentage points; p = 0.58). ITT results were similar. We saw no evidence of the treatment effect varying by race in ITT or TOT analyses. Conclusions and Relevance The inclusion of PCD enhanced essential aspects of patient-provider communication but did not affect patient activation. Outcomes did not differ by race. Trial Registration Clincaltrials.gov identifier: NCT03766841

Although gastrostomy placement is one of the most common procedures performed in children, the optimal technique remains unclear. The purpose of this study was to evaluate variability in the method of gastrostomy tube placement in children in the United States. Patients <18 years old undergoing percutaneous endoscopic gastrostomy (PEG) or surgical gastrostomy (SG) (including open or laparoscopic) from 1997 to 2012 were identified using the Kids’ Inpatient Database. Method of gastrostomy placement was evaluated using a multivariable mixed-effects logistic regression model with a random intercept term and a patient-age random-effect term. A total of 67,811 gastrostomy placements were performed during the study period. PEG was used in 36.6% of entries overall and was generally consistent over time. PEG placement was less commonly performed in infants (adjusted odds ratio [aOR] 0.30, 95%CI 0.26–0.33), children at urban hospitals (aOR: 0.38, 95%CI 0.18–0.82), and children cared for at children’s hospitals (aOR 0.57, 95%CI 0.48–0.69) and was more commonly performed in children with private insurance (aOR 1.17, 95%CI 1.09–1.25). Dramatic variability in PEG use was identified between centers, ranging from 0% to 100%. The random intercept and slope terms significantly improved the model, confirming significant center-level variability and increased variability among patients <1 year old. These findings emphasize the need to further evaluate the safest method of gastrostomy placement in children, in particular among the youngest patients in whom practice varies the most.

BackgroundPatient-centered care is respectful of and responsive to individual patient preferences, needs, and values. To provide patient-centered care, clinicians need to know and incorporate patients’ context into their communication and care with patients. Patient contextual data (PCD) encompass social determinants of health and patients’ needs, values, goals, and preferences relevant to their care. PCD can be challenging to collect as a routine component of the time-limited primary care visit. ObjectiveThis study aims to determine if patient-provider communication and patient activation are different for patient users and patient nonusers of an electronic health record (EHR)–integrated PCD tool and assess if the impact of using PCD on patient-provider communication and patient activation differs for Black and White patients. MethodsWe describe a randomized controlled trial of a prospective cohort of non-Hispanic White and Black patients who receive primary care services at a midwestern academic health care system in the United States. We will evaluate whether providing PCD through a consumer informatics tool enhances patient-provider communication, as measured by the Communication Assessment Tool, and we will evaluate patient activation, as measured by the Patient Activation Measure for PCD tool users and nonusers. Furthermore, owing to racial disparities in care and communication, we seek to determine if the adoption and use of the tool might narrow the differences between patient groups. ResultsThe trial was funded in November 2017 and received local ethics review approval in February 2019. The study began recruitment in April 2019 and enrollment concluded in October 2019 with 301 participants. The analysis was completed in May 2020, and trial results are expected to be published in winter 2020. ConclusionsRecently, there has been increased attention to the role of health information technology tools to enable patients to collaborate with providers through the sharing of PCD. The adoption of such tools may overcome the barriers of current EHRs by directly engaging patients to submit their contextual data. Effectively, these tools would support the EHR in providing a more holistic understanding of the patient. Research further supports that individuals who have robust digital engagement using consumer informatics tools have higher participation in treatment follow-up and self-care across populations. Therefore, it is critical to investigate interventions that elicit and share patients’ social risks and care preferences with the health care team as a mechanism to improve individualized care and reduce the gap in health outcomes. Trial RegistrationClinicalTrials.gov NCT03766841; https://clinicaltrials.gov/ct2/show/NCT03766841 International Registered Report Identifier (IRRID)RR1-10.2196/20309

Objective. To examine morphine milligram equivalent (MME) trends, use of concurrent opioids and benzodiazepines, and opioid-related emergency department (ED) visits or hospitalizations in a national cohort of patients on chronic opioid therapy. Design. Retrospective cohort analysis of prospectively collected data from the Truven Health MarketScan Commercial Claims and Encounters database from 2009 to 2015. This includes individuals in both the Commercial Claims and Medicare Supplemental databases of MarketScan. Methods. MME comparisons of 1) patients on chronic opioids with and without opioid-related ED visits or hospitalizations, 2) patients on concurrent opioids and benzodiazepines with and without opioid-related ED visits or hospitalizations, and 3) patients on chronic opioids compared with those on concurrent opioid and benzodiazepine using the Student t test. Results. MME decreased from 194 mg in 2009 to 119 mg in 2015 among patients on chronic opioids. Patients on opioids and benzodiazepines had higher doses than those on opioids alone for all years (P< 0.001). Those with an opioid-related ED visit or hospitalization had a higher average MME than those without, for all years except 2009 (P< 0.05). Patients on chronic opioids or on concurrent benzodiazepine with an MME >50 had a twofold increased risk of having an opioid-related ED visit or hospitalization compared with those with an MME <50, for all years. Conclusions. Although the average MME decreased over time, patients on combination opioid and benzodiazepine and those with opioid-related ED visits and hospitalizations had significantly higher doses.

In this article, we introduce the BART R package which is an acronym for Bayesian Additive Regression Trees. BART is a Bayesian nonparametric, machine learning, ensemble predictive modeling method for continuous, binary, categorical and time-to-event outcomes. Furthermore, BART is a tree-based, black-box method which fits the outcome to an arbitrary random function, f , of the covariates. The BART technique is relatively computationally efficient as compared to its competitors, but large sample sizes can be demanding. Therefore, the BART package includes efficient state-of-the-art implementations for continuous, binary, categorical and time-to-event outcomes that can take advantage of modern off-the-shelf hardware and software multi-threading technology. The BART package is written in C++ for both programmer and execution efficiency. The BART package takes advantage of multi-threading via forking as provided by the parallel package and OpenMP when available and supported by the platform. The ensemble of binary trees produced by a BART fit can be stored and re-used later via the R predict function. In addition to being an R package, the installed BART routines can be called directly from C++. The BART package provides the tools for your BART toolbox.

There is limited information on the oral health of Albanian immigrant population residing in the U.S. This creates a hinderance to developing and implementing appropriate dental care programs for the population. This study investigated oral health practices, beliefs, dental visits and associated factors of Albanian adults living in Milwaukee, Wisconsin. Purposive and snowball sampling methodswere employed. Self-administered questionnaires were used to collect data on oral health practices, beliefs, dental visits and socio-demographic information. Descriptive and multivariable logistics regression were conducted. Overall, 266 adults were recruited, 54% male, 56% have lived 10 or more years in the U.S., 95% rated their oral health as excellent/good and 87% reported having a dental visit in the last year. Age, ability to speak English, having a usual source of dental care, and reporting excellent/good oral health were associated with having a dental visit in the last year. A substantial number of Albanians adult reported a dental visit in the last year and those that did not write or read in English had lower odds of reporting a dental visit.

Objective: This study examined the prevalence of ambulance use for nontraumatic tooth pain (NTP) visit to emergency departments (EDs) and the factors associated with ambulance use for NTP in the USA. Materials and methods: Data from the National Hospital Ambulatory Medical Care survey conducted in the USA from 2003 to 2012 were analyzed. Descriptive statistics were obtained, and multivariable logistic regression was used to determine associations with ambulance use for NTP. Results: The total proportion of ED visits due to NTP by ambulance was 1.1%, lowest in 2008 (0.43%) and highest in 2011 (2.28%). The proportion of ED visits due to NTP by ambulance was highest among public insurance enrollees (1.9%), Hispanics (2.3%) and those aged 45-64 years (2.7%). In the multivariable analysis, those aged 45-64 years had approximately four times higher odds of an ED visit for NTP by ambulance compared to those aged 25-44 years. Conclusion: This study demonstrates that transport to EDs by ambulance for NTP does occur at a measurable rate and adults aged 45-64 years had significantly higher odds of ED visits for NTP by ambulance.

Disparities exist between Latinos and non-Latino Whites in cognitive function. Dance is culturally appropriate and challenges individuals physically and cognitively, yet the impact of regular dancing on cognitive function in older Latinos has not been examined. A two-group pilot trial was employed among inactive, older Latinos. Participants (N = 57) participated in the BAILAMOS© dance program or a health education program. Cognitive test scores were converted to z-scores and measures of global cognition and specific domains (executive function, episodic memory, working memory) were derived. Results revealed a group × time interaction for episodic memory (p < .05), such that the dance group showed greater improvement in episodic memory than the health education group. A main effect for time for global cognition (p < .05) was also demonstrated, with participants in both groups improving. Structured Latin dance programs can positively influence episodic memory, and participation in structured programs may improve overall cognition among older Latinos.