Tree-informed Bayesian multi-source domain adaptation: cross-population probabilistic cause-of-death assignment using verbal autopsy.

Determining causes of deaths (CODs) occurred outside of civil registration and vital statistics systems is challenging. A technique called verbal autopsy (VA) is widely adopted to gather information on deaths in practice. A VA consists of interviewing relatives of a deceased person about symptoms of the deceased in the period leading to the death, often resulting in multivariate binary responses. While statistical methods have been devised for estimating the cause-specific mortality fractions (CSMFs) for a study population, continued expansion of VA to new populations (or "domains") necessitates approaches that recognize between-domain differences while capitalizing on potential similarities. In this article, we propose such a domain-adaptive method that integrates external between-domain similarity information encoded by a prespecified rooted weighted tree. Given a cause, we use latent class models to characterize the conditional distributions of the responses that may vary by domain. We specify a logistic stick-breaking Gaussian diffusion process prior along the tree for class mixing weights with node-specific spike-and-slab priors to pool information between the domains in a data-driven way. The posterior inference is conducted via a scalable variational Bayes algorithm. Simulation studies show that the domain adaptation enabled by the proposed method improves CSMF estimation and individual COD assignment. We also illustrate and evaluate the method using a validation dataset. The article concludes with a discussion of limitations and future directions.

Using source-associated mobile genetic elements to identify zoonotic extraintestinal E. coli infections.

A one-health perspective may provide new and actionable information about Escherichia coli transmission. E. coli colonizes a broad range of vertebrates, including humans and food-production animals, and is a leading cause of bladder, kidney, and bloodstream infections in humans. Substantial evidence supports foodborne transmission of pathogenic E. coli strains from food animals to humans. However, the relative contribution of foodborne zoonotic E. coli (FZEC) to the human extraintestinal disease burden and the distinguishing characteristics of such strains remain undefined. Using a comparative genomic analysis of a large collection of contemporaneous, geographically-matched clinical and meat-source E. coli isolates (n = 3111), we identified 17 source-associated mobile genetic elements - predominantly plasmids and bacteriophages - and integrated them into a novel Bayesian latent class model to predict the origins of clinical E. coli isolates. We estimated that approximately 8 % of human extraintestinal E. coli infections (mostly urinary tract infections) in our study population were caused by FZEC. FZEC strains were equally likely to cause symptomatic disease as non-FZEC strains. Two FZEC lineages, ST131-H22 and ST58, appeared to have particularly high virulence potential. Our findings imply that FZEC strains collectively cause more urinary tract infections than does any single non-E. coli uropathogenic species (e.g., Klebsiella pneumoniae). Our novel approach can be applied in other settings to identify the highest-risk FZEC strains, determine their sources, and inform new one-health strategies to decrease the heavy public health burden imposed by extraintestinal E. coli infections.

Integrating sample similarities into latent class analysis: a tree-structured shrinkage approach.

This paper is concerned with using multivariate binary observations to estimate the probabilities of unobserved classes with scientific meanings. We focus on the setting where additional information about sample similarities is available and represented by a rooted weighted tree. Every leaf in the given tree contains multiple samples. Shorter distances over the tree between the leaves indicate a priori higher similarity in class probability vectors. We propose a novel data integrative extension to classical latent class models with tree-structured shrinkage. The proposed approach enables (1) borrowing of information across leaves, (2) estimating data-driven leaf groups with distinct vectors of class probabilities, and (3) individual-level probabilistic class assignment given the observed multivariate binary measurements. We derive and implement a scalable posterior inference algorithm in a variational Bayes framework. Extensive simulations show more accurate estimation of class probabilities than alternatives that suboptimally use the additional sample similarity information. A zoonotic infectious disease application is used to illustrate the proposed approach. The paper concludes by a brief discussion on model limitations and extensions.

Factors Associated With Inferior Vena Cava Filter Placement and Retrieval for Patients With Cancer-Associated Thrombosis.

BACKGROUND: Venous thromboembolism is a leading cause of death in patients with cancer. Inferior vena cava filters are utilized to mitigate the risk of pulmonary embolism for patients who have contraindication to, or failure of, anticoagulation. METHODS: We reviewed an insurance claims database to identify adults receiving cancer-directed therapy and had a new diagnosis of venous thromboembolism. We then evaluated clinical and sociodemographic characteristics in patients with and without filter placement and retrieval. RESULTS: There were 25,788 patients (mean [SD] age: 68.3 [12.7] years) who met the study inclusion criteria, with 2111 individuals (8.2%) undergoing filter placement. Filter placement was associated with the type of thrombosis, malignancy, recent surgery, comorbidities, and income. A total of 137 patients (6.5%) newly started anticoagulation within 3 days of filter placement, and 612 (29%) patients received anticoagulation within 30 days after filter placement. Despite this, only 159 (7.5%) patients had their filters retrieved during the study period. Patients with income of $75-99K (odds ratio 2.13, P = .012) or above $100K (odds ratio 1.8, P = .038) were more likely to have filter retrieval compared with those with income <$50K. Filter retrieval was also more likely in younger patients and those with fewer comorbidities or without central nervous system or lung malignancies. CONCLUSIONS: Inferior vena cava filter placement and retrieval are associated with several sociodemographic factors. Filter retrieval rates are low despite re-initiation of anticoagulation in many patients. Efforts are needed to address disparities in filter use and improve retrieval rates.

Clinical and sociodemographic factors associated with anticoagulant use for cancer associated venous thromboembolism.

Cancer associated thrombosis (CAT) is a leading cause of death among patients with cancer. It is not clear if non-clinical factors are associated with anticoagulation receipt. We conducted a retrospective cohort study of Optum's de-identified Clinformatics® Database of adults with cancer diagnosed between 2009 and 2016 who developed CAT, treated with an outpatient anticoagulant (warfarin, low molecular weight heparin (LMWH), or a direct oral anticoagulant (DOAC)). Of 12,622 patients, three months after an episode of CAT, 1,485 (12%) were on LMWH, 1,546 (12%) on DOACs, and 9,591 (76%) were on warfarin. When controlling for other factors, anticoagulant use was significantly associated with socioeconomic factors, region, co-morbidities, type of thrombosis, and cancer subtype. Patients with a bachelor's degree or greater level of education were less likely to receive warfarin (OR: 0.77; 95% CI: [0.59, 0.99]; p < 0.046) or DOACs (OR: 0.67; 95% CI: [0.55, 0.82]; p < 0.001) compared to LMWH. Patients with higher income levels were more likely to receive LMWH or DOACs compared to warfarin, while patients across all income levels were equally likely to receive LMWH or DOACs. Non-clinical factors including income, education, and region, are associated with anticoagulation receipt three months after an episode of CAT. Sociodemographic factors may result in some patients receiving suboptimal care and contribute to non-guideline concordant care for CAT.

Anticoagulant medication adherence for cancer-associated thrombosis: A comparison of LMWH to DOACs.

Essentials It is not clear if patients are less adherent to low molecular weight heparin (LMWH) compared to direct oral anticoagulants (DOACs) for cancer-associated thrombosis (CAT). We evaluated medication adherence among two propensity-matched groups of patients with CAT by comparing the proportion of days covered (PDC). Median treatment persistence on DOACs was more than 80 days longer than LMWH. Medication adherence was high (~95%) and was similar with LMWH compared to DOACs. ABSTRACT: Background Low molecular weight heparin (LMWH) and direct oral anticoagulants (DOACs) are used to treat cancer-associated thrombosis (CAT). It is not clear if patients are less adherent to LMWH compared to DOACs. Objectives To compare medication persistence and adherence between LMWH and DOACs. Patients/Methods We analyzed Optum's de-identified Clinformatics® Data Mart Database of privately insured adults with cancer diagnosed between January 2009 and October 2015 who were undergoing chemotherapy, immunotherapy, targeted or hormonal therapies; developed CAT; and were treated with an outpatient anticoagulant. The proportion of days covered (PDC) was calculated from the date of anticoagulant prescription until the anticoagulant was switched, stopped, or the study end. Medication adherence was defined as PDC ≥ 80%, ≥95%, and by comparing the mean PDC. Results Two propensity-matched groups of 1128 patients were identified. Patient persistence was higher with DOACs compared to LMWH (median 116 days versus 34 days). With adherence defined as PDC ≥ 80%, we found no significant difference (95.6% versus 94.6% adherence with DOACs versus LMWH, P = .33). The mean difference of PDC between the two groups was also similar. With medication adherence defined as PDC ≥ 95%, adherence was evident in 73% of DOAC users and 81% of patients on LMWH (P < .001). Prescription copayments were higher on average for LMWH compared to DOACs (mean $153.61 versus 40.67; standard deviation $306.74 versus $33.11). Conclusion Patients remain on DOACs longer than LMWH, but medication adherence is similar with LMWH.

Robotic-Assisted Live Donor Ileal Segmentectomy for Intestinal Transplantation.

BACKGROUND: Every effort should be made to optimize surgical techniques and to minimize potential morbidity rates associated with live donor operations. Advances in a minimally invasive approach by robotic surgery to donor nephrectomy have raised the possibility of applying this technique to live donor bowel resections for intestinal transplantation. METHODS: We report the first 5 consecutive cases of a robotic-assisted live donor ileal segmentectomy. We describe the technical aspects of the procedure, discuss the rationale for considering this option, and evaluate potential advantages of this approach. RESULTS: We found that this new approach is associated with less postoperative discomfort, a shorter hospital length of stay, and a faster recovery of bowel function compared to our previous open surgery. CONCLUSIONS: Our initial experience suggests that robotic surgery is a safe and feasible procedure for live donor ileal resection for intestinal transplantation and is a useful alternative to conventional open surgery.