CAT: computer aided triage improving upon the Bayes risk through ε-refusal triage rules.

BACKGROUND: Manual 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. RESULTS: This 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. CONCLUSIONS: From 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.

Environmental determinants of health: Measuring multiple physical environmental exposures at the United States census tract level.

Physical environment plays a key role in determining human health risks. Exposure to toxins, weather extremes, degraded air and water quality, high levels of noise and limited accessibility to green areas can negatively affect health. Furthermore, adverse environmental exposures are often correlated with each other and with socioeconomic status, thereby compounding disadvantages in marginalized populations. Moreover, despite their importance in determining human health risks, the role of multiple environmental exposures is not well studied, and only a few resources contain aggregate environmental exposure data and only for selected areas of the contiguous US. To fill these gaps, we took a cumulative approach to measuring the environment by generating a composite Multi-Exposure Environmental Index (MEEI) as a US Census Tract-level summary of key environmental factors with known health effects. This measure quantifies multiple environmental exposures in the same area that can result in additive and synergistic effects on health outcomes. This information is crucial to better understand and possibly leverage environmental determinants of health for informed policy-making and intervention.

Exploring the Spatial Patterning of Sociodemographic Disparities in Extreme Heat Exposure at Multiple Scales Across the Conterminous United States.

Climate change has led to an increase in heat-related morbidity and mortality. The impact of heat on health is unequally distributed amongst different socioeconomic and demographic groups. We use high-resolution daily air temperature-based heat wave intensity (HWI) and neighborhood-scale sociodemographic information from the conterminous United States to evaluate the spatial patterning of extreme heat exposure disparities. Assuming differences in spatial patterns at national, regional, and local scales; we assess disparities in heat exposure across race, housing characteristics, and poverty level. Our findings indicate small differences in HWI based on these factors at the national level, with the magnitude and direction of the differences varying by region. The starkest differences are present over the Northeast and Midwest, where primarily Black neighborhoods are exposed to higher HWI than predominantly White areas. At the local level, we find the largest difference by socioeconomic status. We also find that residents of nontraditional housing are more vulnerable to heat exposure. Previous studies have either evaluated such disparities for specific cities and/or used a satellite-based land surface temperature, which, although correlated with air temperature, does not provide the true measure of heat exposure. This study is the first of its kind to incorporate high-resolution gridded air temperature-based heat exposure in the evaluation of sociodemographic disparities at a national scale. The analysis suggests the unequal distribution of heat wave intensities across communities-with higher heat exposures characterizing areas with high proportions of minorities, low socioeconomic status, and homes in need of retrofitting to combat climate change.

Classifying cancer pathology reports with hierarchical self-attention networks.

We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks - site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data - Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

Return to work outcomes following accident compensation corporation work capacity assessment.

AIMS: To determine the proportion of Accident Compensation Corporation (ACC) claimants who have returned to fulltime work after ceasing to receive ACC weekly compensation following Work Capacity Assessment (WCAP). To assess what factors impact on return to work. To assess whether ACC's research findings into return to work outcomes WCAP are valid. METHODS: A structured questionnaire telephone follow-up survey was conducted with ACC claimants seen for WCAP. RESULTS: 43% of those exited from ACC weekly compensation after WCAP were currently working fulltime. Claimants who had exited ACC after WCAP were significantly more likely to be working than those remaining on ACC. Claimants over 40 years of age were significantly less likely to be working. Gender, race, length of time since injury, and retraining made no difference to return to work. 80% of claimants felt that the WCAP process was unfair. CONCLUSION: Nearly half of those claimants certified as being unfit for work but now exited from ACC via WCAP were working fulltime. This may indicate that ACC's rehabilitation is successful, or that claimants tend to remain on ACC for economic rather than injury reasons, or that WCAP results in claimants returning to physically unsuitable work putting them at risk of further injury. ACC's research finding, that 79% of claimants were working after WCAP does not appear to be valid.