Leticia Cuellar-Hengartner

We present an ensemble transfer learning method to predict suicide from Veterans Affairs (VA) electronic medical records (EMR). A diverse set of base models was trained to predict a binary outcome constructed from reported suicide, suicide attempt, and overdose diagnoses with varying choices of study design and prediction methodology. Each model used twenty cross-sectional and 190 longitudinal variables observed in eight time intervals covering 7.5 years prior to the time of prediction. Ensembles of seven base models were created and fine-tuned with ten variables expected to change with study design and outcome definition in order to predict suicide and combined outcome in a prospective cohort. The ensemble models achieved c-statistics of 0.73 on 2-year suicide risk and 0.83 on the combined outcome when predicting on a prospective cohort of [Formula: see text] 4.2 M veterans. The ensembles rely on nonlinear base models trained using a matched retrospective nested case-control (Rcc) study cohort and show good calibration across a diversity of subgroups, including risk strata, age, sex, race, and level of healthcare utilization. In addition, a linear Rcc base model provided a rich set of biological predictors, including indicators of suicide, substance use disorder, mental health diagnoses and treatments, hypoxia and vascular damage, and demographics.

The Transport Airplane Risk Assessment Methodology (TARAM) is a process for calculating risk associated with continued operational safety issues in the U.S. transport airplane fleet. TARAM is important because its risk-analysis calculations are used when making determinations of unsafe conditions in transport airplanes and when selecting and implementing corrective actions. This report assesses the TARAM process used by the FAA in its efforts to improve the overall safety of the transport airplane fleet. A healthy safety culture requires commitment to continuous improvement. This report provides recommendations to the FAA to address the gaps and strengthen the TARAM.

BackgroundManual extraction of information from electronic pathology (epath) reports to populate the Surveillance, Epidemiology, and End Result (SEER) database is labor intensive. Systematizing the data extraction automatically using machine-learning (ML) and natural language processing (NLP) is desirable to reduce the human labor required to populate the SEER database and to improve the timeliness of the data. This enables scaling up registry efficiency and collection of new data elements. To ensure the integrity, quality, and continuity of the SEER data, the misclassification error of ML and NPL algorithms needs to be negligible. Current algorithms fail to achieve the precision of human experts who can bring additional information in their assessments. Differences in registry format and the desire to develop a common information extraction platform further complicate the ML/NLP tasks. The purpose of our study is to develop triage rules to partially automate registry workflow to improve the precision of the auto-extracted information.ResultsThis paper presents a mathematical framework to improve the precision of a classifier beyond that of the Bayes classifier by selectively classifying item that are most likely to be correct. This results in a triage rule that only classifies a subset of the item. We characterize the optimal triage rule and demonstrate its usefulness in the problem of classifying cancer site from electronic pathology reports to achieve a desired precision.ConclusionsFrom the mathematical formalism, we propose a heuristic estimate for triage rule based on post-processing the soft-max output from standard machine learning algorithms. We show, in test cases, that the triage rule significantly improve the classification accuracy.

The Probabilistic Effectiveness Methodology (PEM) is a simulation tool with a holistic approach to risk assessment of nuclear smuggling. PEM simulates valid representations of threat motivation, capabilities, and intent, threat transportation pathways (air, land, and sea), the performance of detector architectures, and individual detector performance associated with preventive radiological and nuclear detection. Further, it analyses from a Red/Adversary perspective, gaps, seams and vulnerabilities of the Global Nuclear Detection Architecture (GNDA). This paper presents the different PEM components and illustrates (through use of notional data) several examples of how PEM can support the decision making process for GNDA problems.

The shortest path problem on a network with fixed weights is a well studied problem with applications to many diverse areas such as transportation and telecommunications. We are particularly interested in the scenario where a nuclear material smuggler tries to succesfully reach herlhis target by identifying the most likely path to the target. The identification of the path relies on reliabilities (weights) associated with each link and node in a multi-modal transportation network. In order to account for the adversary's uncertainty and to perform sensitivity analysis we introduce random reliabilities. We perform some controlled experiments on the grid and present the distributional properties of the resulting stochastic shortest paths.

Nuclear weapons proliferation is an existing and growing worldwide problem. To help with devising strategies and supporting decisions to interdict the transport of nuclear material, we developed the Pathway Analysis, Threat Response and Interdiction Options Tool (PATRIOT) that provides an analytical approach for evaluating the probability that an adversary smuggling radioactive or special nuclear material will be detected during transit. We incorporate a global, multi-modal transportation network, explicit representation of designed and serendipitous detection opportunities, and multiple threat devices, material types, and shielding levels. This paper presents the general structure of PATRIOT, and focuses on the theoretical framework used to model the reliabilities of all network components that are used to predict the most likely pathways to the target.