Air quality and cancer risk in the All of Us Research Program.

INTRODUCTION: The NIH All of Us Research Program has enrolled over 544,000 participants across the US with unprecedented racial/ethnic diversity, offering opportunities to investigate myriad exposures and diseases. This paper aims to investigate the association between PM2.5 exposure and cancer risks. MATERIALS AND METHODS: This work was performed on data from 409,876 All of Us Research Program participants using the All of Us Researcher Workbench. Cancer case ascertainment was performed using data from electronic health records and the self-reported Personal Medical History questionnaire. PM2.5 exposure was retrieved from NASA's Earth Observing System Data and Information Center and assigned using participants' 3-digit zip code prefixes. Multivariate logistic regression was used to estimate the odds ratio (OR) and 95% confidence interval (CI). Generalized additive models (GAMs) were used to investigate non-linear relationships. RESULTS: A total of 33,387 participants and 46,176 prevalent cancer cases were ascertained from participant EHR data, while 20,297 cases were ascertained from self-reported survey data from 18,133 participants; 9,502 cancer cases were captured in both the EHR and survey data. Average PM2.5 level from 2007 to 2016 was 8.90 µg/m3 (min 2.56, max 15.05). In analysis of cancer cases from EHR, an increased odds for breast cancer (OR 1.17, 95% CI 1.09-1.25), endometrial cancer (OR 1.33, 95% CI 1.09-1.62) and ovarian cancer (OR 1.20, 95% CI 1.01-1.42) in the 4th quartile of exposure compared to the 1st. In GAM, higher PM2.5 concentration was associated with increased odds for blood cancer, bone cancer, brain cancer, breast cancer, colon and rectum cancer, endocrine system cancer, lung cancer, pancreatic cancer, prostate cancer, and thyroid cancer. CONCLUSIONS: We found evidence of an association of PM2.5 with breast, ovarian, and endometrial cancers. There is little to no prior evidence in the literature on the impact of PM2.5 on risk of these cancers, warranting further investigation.

Evaluating Natural Language Processing Packages for Predicting Hospital-Acquired Pressure Injuries From Clinical Notes.

Incidence of hospital-acquired pressure injury, a key indicator of nursing quality, is directly proportional to adverse outcomes, increased hospital stays, and economic burdens on patients, caregivers, and society. Thus, predicting hospital-acquired pressure injury is important. Prediction models use structured data more often than unstructured notes, although the latter often contain useful patient information. We hypothesize that unstructured notes, such as nursing notes, can predict hospital-acquired pressure injury. We evaluate the impact of using various natural language processing packages to identify salient patient information from unstructured text. We use named entity recognition to identify keywords, which comprise the feature space of our classifier for hospital-acquired pressure injury prediction. We compare scispaCy and Stanza, two different named entity recognition models, using unstructured notes in Medical Information Mart for Intensive Care III, a publicly available ICU data set. To assess the impact of vocabulary size reduction, we compare the use of all clinical notes with only nursing notes. Our results suggest that named entity recognition extraction using nursing notes can yield accurate models. Moreover, the extracted keywords play a significant role in the prediction of hospital-acquired pressure injury.

Extracting medication information from unstructured public health data: a demonstration on data from population-based and tertiary-based samples.

BACKGROUND: Unstructured data from clinical epidemiological studies can be valuable and easy to obtain. However, it requires further extraction and processing for data analysis. Doing this manually is labor-intensive, slow and subject to error. In this study, we propose an automation framework for extracting and processing unstructured data. METHODS: The proposed automation framework consisted of two natural language processing (NLP) based tools for unstructured text data for medications and reasons for medication use. We first checked spelling using a spell-check program trained on publicly available knowledge sources and then applied NLP techniques. We mapped medication names into generic names using vocabulary from publicly available knowledge sources. We used WHO's Anatomical Therapeutic Chemical (ATC) classification system to map generic medication names to medication classes. We processed the reasons for medication with the Lancaster stemmer method and then grouped and mapped to disease classes based on organ systems. Finally, we demonstrated this automation framework on two data sources for Mylagic Encephalomyelitis/ Chronic Fatigue Syndrome (ME/CFS): tertiary-based (n = 378) and population-based (n = 664) samples. RESULTS: A total of 8681 raw medication records were used for this demonstration. The 1266 distinct medication names (omitting supplements) were condensed to 89 ATC classification system categories. The 1432 distinct raw reasons for medication use were condensed to 65 categories via NLP. Compared to completion of the entire process manually, our automation process reduced the number of the terms requiring manual labor for mapping by 84.4% for medications and 59.4% for reasons for medication use. Additionally, this process improved the precision of the mapped results. CONCLUSIONS: Our automation framework demonstrates the usefulness of NLP strategies even when there is no established mapping database. For a less established database (e.g., reasons for medication use), the method is easily modifiable as new knowledge sources for mapping are introduced. The capability to condense large features into interpretable ones will be valuable for subsequent analytical studies involving techniques such as machine learning and data mining.

PMCVec: Distributed phrase representation for biomedical text processing.

Distributed semantic representation of biomedical text can be beneficial for text classification, named entity recognition, query expansion, human comprehension, and information retrieval. Despite the success of high-quality vector space models such as Word2Vec and GloVe, they only provide unigram word representations and the semantics for multi-word phrases can only be approximated by composition. This is problematic in biomedical text processing where technical phrases for diseases, symptoms, and drugs should be represented as single entities to capture the correct meaning. In this paper, we introduce PMCVec, an unsupervised technique that generates important phrases from PubMed abstracts and learns embeddings for single words and multi-word phrases simultaneously. Evaluations performed on benchmark datasets produce significant performance gains both qualitatively and quantitatively.

Phenotype Instance Verification and Evaluation Tool (PIVET): A Scaled Phenotype Evidence Generation Framework Using Web-Based Medical Literature.

BACKGROUND: Researchers are developing methods to automatically extract clinically relevant and useful patient characteristics from raw healthcare datasets. These characteristics, often capturing essential properties of patients with common medical conditions, are called computational phenotypes. Being generated by automated or semiautomated, data-driven methods, such potential phenotypes need to be validated as clinically meaningful (or not) before they are acceptable for use in decision making. OBJECTIVE: The objective of this study was to present Phenotype Instance Verification and Evaluation Tool (PIVET), a framework that uses co-occurrence analysis on an online corpus of publically available medical journal articles to build clinical relevance evidence sets for user-supplied phenotypes. PIVET adopts a conceptual framework similar to the pioneering prototype tool PheKnow-Cloud that was developed for the phenotype validation task. PIVET completely refactors each part of the PheKnow-Cloud pipeline to deliver vast improvements in speed without sacrificing the quality of the insights PheKnow-Cloud achieved. METHODS: PIVET leverages indexing in NoSQL databases to efficiently generate evidence sets. Specifically, PIVET uses a succinct representation of the phenotypes that corresponds to the index on the corpus database and an optimized co-occurrence algorithm inspired by the Aho-Corasick algorithm. We compare PIVET's phenotype representation with PheKnow-Cloud's by using PheKnow-Cloud's experimental setup. In PIVET's framework, we also introduce a statistical model trained on domain expert-verified phenotypes to automatically classify phenotypes as clinically relevant or not. Additionally, we show how the classification model can be used to examine user-supplied phenotypes in an online, rather than batch, manner. RESULTS: PIVET maintains the discriminative power of PheKnow-Cloud in terms of identifying clinically relevant phenotypes for the same corpus with which PheKnow-Cloud was originally developed, but PIVET's analysis is an order of magnitude faster than that of PheKnow-Cloud. Not only is PIVET much faster, it can be scaled to a larger corpus and still retain speed. We evaluated multiple classification models on top of the PIVET framework and found ridge regression to perform best, realizing an average F1 score of 0.91 when predicting clinically relevant phenotypes. CONCLUSIONS: Our study shows that PIVET improves on the most notable existing computational tool for phenotype validation in terms of speed and automation and is comparable in terms of accuracy.

Harnessing Peer Support in an Online Intervention for Older Adults with Depression.

OBJECTIVE: This pilot study evaluated the feasibility and efficacy of two methods of delivering a cognitive behaviorally informed Internet intervention for depression for adults 65 years and older. METHODS: Forty-seven participants were enrolled and assigned to receive one of two versions of the Internet intervention, either delivered individually (III) or with peer support (II+PS), or to a wait list control group (WLC). Primary outcomes included change in depressive symptoms from baseline to post-intervention (week 8), site use, self-reported usability, and coach time. Secondary outcomes included measures of social support and isolation and anxiety. RESULTS: Follow-up data were provided by 85.1% (40 of 47) of enrolled participants. There were significant differences in depression change across groups (F(2,37) = 3.81, p = 0.03). Greater reductions in depressive symptoms were found for the III (p = 0.02) and II+PS (p = 0.03) compared with WLC, and significantly less coach time was required in the II+PS (p = 0.003). CONCLUSIONS: These results highlight the potential of cognitive-behaviorally informed Internet interventions for older adults with depression, and indicate that peer-supported programs are both acceptable and equivalent to individually delivered Internet interventions. Including peer support may be a viable and potentially more cost-effective option for disseminating online treatments for depression for older adults.

A Dysregulated Balance of Proinflammatory and Anti-Inflammatory Host Cytokine Response Early During Therapy Predicts Persistence and Mortality in Staphylococcus aureus Bacteremia.

OBJECTIVES: The contribution of individual immune response to Staphylococcus aureus bacteremia on outcome has not been well studied. The objective was to relate the host cytokine response to outcome of Staphylococcus aureus bacteremia. DESIGN: Prospective observational study. SETTING: Three U.S. university-affiliated medical centers. PATIENTS: Adult patients infected with Staphylococcus aureus bacteremia hospitalized between July 2012 and August 2014. INTERVENTIONS: Blood specimens were obtained at Staphylococcus aureus bacteremia onset and 72 hours after therapy initiation. Levels of tissue necrosis factor, interleukin-6, interleukin-8, interleukin-17A, and interleukin-10 were measured by enzyme-linked immunosorbent assay at each time point and compared between those with persistent bacteremia (≥ 4 d) and resolving bacteremia. Primary outcome was persistent bacteremia after 4 days of effective therapy. Secondary outcomes were 30-day mortality and 30-day recurrence. MEASUREMENTS AND MAIN RESULTS: A total of 196 patients were included (mean age, 59 yr); of them, 33% had methicillin-resistant Staphylococcus aureus bacteremia. Forty-seven percent of the methicillin-resistant Staphylococcus aureus strains were staphylococcal cassette chromosome mec IV. Persistent bacteremia occurred in 24% of patients (47/196); they were more likely to die than resolving bacteremia group (28% vs 5%; p < 0.001). Compared with resolving bacteremia group, persistent bacteremia patients had higher initial median levels of tissue necrosis factor (44.73 vs 21.68 pg/mL; p < 0.001), interleukin-8 (124.76 vs 47.48 pg/mL; p = 0.028), and interleukin-10 (104.31 vs 29.72 pg/mL; p < 0.001). Despite 72 hours of treatment, levels remained higher for the persistent bacteremia group than for the resolving bacteremia group (tissue necrosis factor: 26.95 vs 18.38 pg/mL, p = 0.02; interleukin-8: 70.75 vs 27.86 pg/mL, p = 0.002; interleukin-6: 67.50 vs 21.81 pg/mL, p = 0.005; and interleukin-10: 30.98 vs 12.60 pg/mL, p < 0.001). Interleukin-17A levels were similar between groups at both time points. After controlling for confounding variables by multivariate analysis, interleukin-10/tissue necrosis factor ratio at 72 hours most significantly predicted persistence (odds ratio, 2.98; 95% CI, 1.39-6.39; p = 0.005) and mortality (odds ratio, 9.87; 95% CI, 2.64-36.91; p < 0.001) at values more than 1.00 and more than 2.56, respectively. CONCLUSIONS: Sustained elevation of interleukin-10/tissue necrosis factor ratio at 72 hours suggests a dysregulated immune response and may be used to guide management to improve outcomes.

Limestone: high-throughput candidate phenotype generation via tensor factorization.

The rapidly increasing availability of electronic health records (EHRs) from multiple heterogeneous sources has spearheaded the adoption of data-driven approaches for improved clinical research, decision making, prognosis, and patient management. Unfortunately, EHR data do not always directly and reliably map to medical concepts that clinical researchers need or use. Some recent studies have focused on EHR-derived phenotyping, which aims at mapping the EHR data to specific medical concepts; however, most of these approaches require labor intensive supervision from experienced clinical professionals. Furthermore, existing approaches are often disease-centric and specialized to the idiosyncrasies of the information technology and/or business practices of a single healthcare organization. In this paper, we propose Limestone, a nonnegative tensor factorization method to derive phenotype candidates with virtually no human supervision. Limestone represents the data source interactions naturally using tensors (a generalization of matrices). In particular, we investigate the interaction of diagnoses and medications among patients. The resulting tensor factors are reported as phenotype candidates that automatically reveal patient clusters on specific diagnoses and medications. Using the proposed method, multiple phenotypes can be identified simultaneously from data. We demonstrate the capability of Limestone on a cohort of 31,815 patient records from the Geisinger Health System. The dataset spans 7years of longitudinal patient records and was initially constructed for a heart failure onset prediction study. Our experiments demonstrate the robustness, stability, and the conciseness of Limestone-derived phenotypes. Our results show that using only 40 phenotypes, we can outperform the original 640 features (169 diagnosis categories and 471 medication types) to achieve an area under the receiver operator characteristic curve (AUC) of 0.720 (95% CI 0.715 to 0.725). Moreover, in consultation with a medical expert, we confirmed 82% of the top 50 candidates automatically extracted by Limestone are clinically meaningful.

Cognitive-behavioral therapy in depressed primary care patients with co-occurring problematic alcohol use: effect of telephone-administered vs. face-to-face treatment-a secondary analysis.

This secondary analysis of a larger study compared adherence to telephone-administered cognitive-behavioral therapy (T-CBT) vs. face-to-face CBT and depression outcomes in depressed primary care patients with co-occurring problematic alcohol use. To our knowledge, T-CBT has never been directly compared to face-to-face CBT in such a sample of primary care patients. Participants were randomized in a 1:1 ratio to face-to-face CBT or T-CBT for depression. Participants receiving T-CBT (n = 50) and face-to-face CBT (n = 53) were compared at baseline, end of treatment (week 18), and three-month and six-month follow-ups. Face-to-face CBT and T-CBT groups did not significantly differ in age, sex, ethnicity, marital status, educational level, severity of depression, antidepressant use, and total score on the Alcohol Use Disorders Identification Test. Face-to-face CBT and T-CBT groups were similar on all treatment adherence outcomes and depression outcomes at all time points. T-CBT and face-to-face CBT had similar treatment adherence and efficacy for the treatment of depression in depressed primary care patients with co-occurring problematic alcohol use. When targeting patients who might have difficulties in accessing care, primary care clinicians may consider both types of CBT delivery when treating depression in patients with co-occurring problematic alcohol use.

An AdaBoost-based algorithm to detect hospital-acquired pressure injury in the presence of conflicting annotations.

Hospital-acquired pressure injury is one of the most harmful events in clinical settings. Patients who do not receive early prevention and treatment can experience a significant financial burden and physical trauma. Several hospital-acquired pressure injury prediction algorithms have been developed to tackle this problem, but these models assume a consensus, gold-standard label (i.e., presence of pressure injury or not) is present for all training data. Existing definitions for identifying hospital-acquired pressure injuries are inconsistent due to the lack of high-quality documentation surrounding pressure injuries. To address this issue, we propose in this paper an ensemble-based algorithm that leverages truth inference methods to resolve label inconsistencies between various case definitions and the level of disagreements in annotations. Application of our method to MIMIC-III, a publicly available intensive care unit dataset, gives empirical results that illustrate the promise of learning a prediction model using truth inference-based labels and observed conflict among annotators.

Gene polymorphisms predict toxicity to neoadjuvant therapy in patients with rectal cancer.

BACKGROUND: Toxicity from neoadjuvant chemoradiation therapy (NT) increases morbidity and limits therapeutic efficacy in patients with rectal cancer. The objective of this study was to determine whether specific polymorphisms in genes associated with rectal cancer response to NT were correlated with NT-related toxicity. METHODS: One hundred thirty-two patients with locally advanced rectal cancer received NT followed by surgery. All patients received 5-fluorouracil (5-FU) and radiation (RT), and 80 patients also received modified infusional 5-FU, folinic acid, and oxaliplatin chemotherapy (mFOLFOX-6). Grade ≥3 adverse events (AEs) that occurred during 5-FU/RT and during combined 5-FU/RT + mFOLFOX-6 were recorded. Pretreatment biopsy specimens and normal rectal tissues were collected from all patients. DNA was extracted and screened for 22 polymorphisms in 17 genes that have been associated with response to NT. Polymorphisms were correlated with treatment-related grade ≥3 AEs. RESULTS: Overall, 27 of 132 patients (20%) had grade ≥3 AEs; 18 patients had a complication associated only with 5-FU/RT, 3 patients experienced toxicity only during mFOLFOX-6, and 6 patients had grade ≥3 AEs associated with both treatments before surgery. Polymorphisms in the genes x-ray repair complementing defective repair in Chinese hamster cells 1 (XRCC1), xeroderma pigmentosum group D (XPD), and tumor protein 53 (TP53) were associated with grade ≥3 AEs during NT (P < .05). Specifically, 2 polymorphisms-an arginine-to-glutamine substitution at codon 399 (Q399R) in XRCC1 and a lysine-to-glutamine substitution at codon 751 (K751Q) in XPD-were associated with increased toxicity to 5-FU/RT (P < .05), and an arginine-to-proline substitution at codon 72 (R72P) in TP53 was associated with increased toxicity to mFOLFOX-6 (P = .008). CONCLUSIONS: Specific polymorphisms in XRCC1, XPD, and TP53 were associated with increased toxicity to NT in patients with rectal cancer. The current results indicated that polymorphism screening may help tailor treatment for patients by selecting therapies with the lowest risk of toxicity, thus increasing patient compliance.

Short- and long-term outcomes of intracorporeal versus extracorporeal ileocolic anastomosis in laparoscopic right hemicolectomy for colon cancer.

BACKGROUND: We assessed the short- and long-term outcomes of intracorporeal ileocolic anastomosis (IA) in laparoscopic right hemicolectomy for colon cancer compared with extracorporeal anastomosis (EA). METHODS: A retrospective chart review of 86 consecutive patients who underwent laparoscopic right hemicolectomy for colon cancer from March 2005 to June 2010 was performed. RESULTS: There were 51 and 35 patients who underwent intracorporeal and extracorporeal anastomosis, respectively. The two groups were demographically comparable. The conversion rate to open surgery was 8.6 % in the EA group, but none in the IA group (p = 0.064). There was no significant difference in operative time, estimated blood loss, complications (intra-abdominal abscess, anastomotic leak, ileus, and wound infection), and length of hospital stay between the groups. There was no perioperative mortality in both groups. There was no significant difference in median number of retrieved lymph node. The overall survival and the disease-free survival at 3 years were not different between the groups. CONCLUSIONS: Compared with the extracorporeal anastomosis technique, intracorporeal ileocolic anastomosis produces comparable short- and long-term outcomes in laparoscopic right hemicolectomy for colon cancer.

Interest and use of mental health and specialty behavioral medicine counseling in US primary care patients.

BACKGROUND: Counseling interventions have the potential to improve health and quality of life for primary care patients, but there are few studies describing the interest in and utilization of counseling among this patient population in the USA. PURPOSE: The purpose of the study was to evaluate interest in mental health and specialty behavioral medicine counseling and predictors of utilization over 1 year among US primary care patients. METHOD: Participants in this two-survey longitudinal study included 658 primary care patients in an urban US academic medical center (461 females, age M = 51.05, SD = 15.46 years). Retention rate was 61.2% at survey 2. Patient demographics, depression, anxiety, and interest in counseling services were assessed through a survey mailed 1 week following an outpatient appointment. Respondents to survey 1 were re-contacted 1 year later to assess. Interest and use of the following counseling services were evaluated in the relevant subgroups: mental health (the entire sample and patients with elevated anxiety and/or depression), health/lifestyle (overweight and obese participants), smoking cessation (current and occasional smokers), and pain management (participants with elevated daily pain ratings). RESULTS: At survey 1, 45.7% of the sample reported interest in mental health counseling, and 58.9% of the sample reported interest in behavioral medicine counseling. Among overweight or obese participants, 59.9% were interested in health/lifestyle counseling. Among smokers, 55.3% were interested in smoking cessation, and among participants with chronic pain, 33.8% were interest in pain management. Rates of utilization of services at survey 2 were 21.3% for mental health, 7.7% for health/lifestyle, 6.7% for smoking cessation, and 6.6% for pain management. Interest in receiving services at survey 1 was the strongest predictor of utilization. CONCLUSION: Results demonstrate high interest but low utilization over 1 year among US primary care patients. Identifying patients interested in counseling services and reducing barriers may help facilitate receipt of services for those with interest and need for behavioral treatments.

Sign and pseudo-sign of Leser-Trélat: case reports and a review of the literature.

BACKGROUND: Leser-Trélat is distinguished by a rare paraneoplastic sign that is characterized by the sudden eruption of multiple seborrheic keratoses (SKs), associated with underlying internal malignancies. Similar non-malignancy-associated SK eruptions are referred to as the "pseudo-sign of Leser-Trélat" (PLT). OBJECTIVE: Two cases of rapid SK eruptions, one the sign of Leser-Trélat (SLT) and one PLT, are presented, and the literature on SLT and PLT is reviewed. METHODS: A literature review of SLT/PLT was performed by searching the PubMed database for all related English published cases. RESULTS: We identified 109 cases of SLT and 12 cases of PLT, with a mean patient age of 61.8 years. SK eruptions were observed before (68.3%), after (22.1%), and at the time of (9.6%) malignancy diagnosis. The malignancy most frequently associated with SLT was gastric adenocarcinoma. The most common anatomical location of SK eruptions was the trunk (18.9%). Frequently reported associated signs and symptoms included pruritus (52%) and acanthosis nigricans (38.7%). The most common treatment included surgery (35.8%), chemotherapy (26.9%), and radiation therapy (26.9%). Treatment resulted in clinical improvement (45%), no change (30%), exacerbation (15%), or initial improvement followed by exacerbation of SKs. Patient outcomes included disease stability/ improvement (48.4%), recurrence (9.7%), exacerbation/metastasis/new malignancy (4.8%), and death (37.1%). LIMITATIONS: This was a retrospective study and excluded non-English published cases. CONCLUSION: This review updates the existing SLT literature and emphasizes the presence of PLT. Clinicians should be aware that SK eruptions may be early manifestations of an internal malignancy or other pathology. To our knowledge, this is the first review examining both SLT and PLT.

Weakly-Supervised Scientific Document Classification via Retrieval-Augmented Multi-Stage Training.

Scientific document classification is a critical task for a wide range of applications, but the cost of collecting human-labeled data can be prohibitive. We study scientific document classification using label names only. In scientific domains, label names often include domain-specific concepts that may not appear in the document corpus, making it difficult to match labels and documents precisely. To tackle this issue, we propose WanDeR, which leverages dense retrieval to perform matching in the embedding space to capture the semantics of label names. We further design the label name expansion module to enrich its representations. Lastly, a self-training step is used to refine the predictions. The experiments on three datasets show that WanDeR outperforms the best baseline by 11.9%. Our code will be published at https://github.com/ritaranx/wander.

Hypergraph Transformers for EHR-based Clinical Predictions.

Electronic health records (EHR) data contain rich information about patients' health conditions including diagnosis, procedures, medications and etc., which have been widely used to facilitate digital medicine. Despite its importance, it is often non-trivial to learn useful representations for patients' visits that support downstream clinical predictions, as each visit contains massive and diverse medical codes. As a result, the complex interactions among medical codes are often not captured, which leads to substandard predictions. To better model these complex relations, we leverage hypergraphs, which go beyond pairwise relations to jointly learn the representations for visits and medical codes. We also propose to use the self-attention mechanism to automatically identify the most relevant medical codes for each visit based on the downstream clinical predictions with better generalization power. Experiments on two EHR datasets show that our proposed method not only yields superior performance, but also provides reasonable insights towards the target tasks.

A Comprehensive and Improved Definition for Hospital-Acquired Pressure Injury Classification Based on Electronic Health Records: Comparative Study.

BACKGROUND: Patients develop pressure injuries (PIs) in the hospital owing to low mobility, exposure to localized pressure, circulatory conditions, and other predisposing factors. Over 2.5 million Americans develop PIs annually. The Center for Medicare and Medicaid considers hospital-acquired PIs (HAPIs) as the most frequent preventable event, and they are the second most common claim in lawsuits. With the growing use of electronic health records (EHRs) in hospitals, an opportunity exists to build machine learning models to identify and predict HAPI rather than relying on occasional manual assessments by human experts. However, accurate computational models rely on high-quality HAPI data labels. Unfortunately, the different data sources within EHRs can provide conflicting information on HAPI occurrence in the same patient. Furthermore, the existing definitions of HAPI disagree with each other, even within the same patient population. The inconsistent criteria make it impossible to benchmark machine learning methods to predict HAPI. OBJECTIVE: The objective of this project was threefold. We aimed to identify discrepancies in HAPI sources within EHRs, to develop a comprehensive definition for HAPI classification using data from all EHR sources, and to illustrate the importance of an improved HAPI definition. METHODS: We assessed the congruence among HAPI occurrences documented in clinical notes, diagnosis codes, procedure codes, and chart events from the Medical Information Mart for Intensive Care III database. We analyzed the criteria used for the 3 existing HAPI definitions and their adherence to the regulatory guidelines. We proposed the Emory HAPI (EHAPI), which is an improved and more comprehensive HAPI definition. We then evaluated the importance of the labels in training a HAPI classification model using tree-based and sequential neural network classifiers. RESULTS: We illustrate the complexity of defining HAPI, with <13% of hospital stays having at least 3 PI indications documented across 4 data sources. Although chart events were the most common indicator, it was the only PI documentation for >49% of the stays. We demonstrate a lack of congruence across existing HAPI definitions and EHAPI, with only 219 stays having a consensus positive label. Our analysis highlights the importance of our improved HAPI definition, with classifiers trained using our labels outperforming others on a small manually labeled set from nurse annotators and a consensus set in which all definitions agreed on the label. CONCLUSIONS: Standardized HAPI definitions are important for accurately assessing HAPI nursing quality metric and determining HAPI incidence for preventive measures. We demonstrate the complexity of defining an occurrence of HAPI, given the conflicting and incomplete EHR data. Our EHAPI definition has favorable properties, making it a suitable candidate for HAPI classification tasks.

Hepatitis C care cascade in a large academic healthcare system, 2012 to 2018.

To determine the hepatitis C virus (HCV) care cascade among persons who were born during 1945 to 1965 and received outpatient care on or after January 2014 at a large academic healthcare system. Deidentified electronic health record data in an existing research database were analyzed for this study. Laboratory test results for HCV antibody and HCV ribonucleic acid (RNA) indicated seropositivity and confirmatory testing. HCV genotyping was used as a proxy for linkage to care. A direct-acting antiviral (DAA) prescription indicated treatment initiation, an undetectable HCV RNA at least 20 weeks after initiation of antiviral treatment indicated a sustained virologic response. Of the 121,807 patients in the 1945 to 1965 birth cohort who received outpatient care between January 1, 2014 and June 30, 2017, 3399 (3%) patients were screened for HCV; 540 (16%) were seropositive. Among the seropositive, 442 (82%) had detectable HCV RNA, 68 (13%) had undetectable HCV RNA, and 30 (6%) lacked HCV RNA testing. Of the 442 viremic patients, 237 (54%) were linked to care, 65 (15%) initiated DAA treatment, and 32 (7%) achieved sustained virologic response. While only 3% were screened for HCV, the seroprevalence was high in the screened sample. Despite the established safety and efficacy of DAAs, only 15% initiated treatment during the study period. To achieve HCV elimination, improved HCV screening and linkage to HCV care and DAA treatment are needed.

Neighborhood-Regularized Self-Training for Learning with Few Labels.

Training deep neural networks (DNNs) with limited supervision has been a popular research topic as it can significantly alleviate the annotation burden. Self-training has been successfully applied in semi-supervised learning tasks, but one drawback of self-training is that it is vulnerable to the label noise from incorrect pseudo labels. Inspired by the fact that samples with similar labels tend to share similar representations, we develop a neighborhood-based sample selection approach to tackle the issue of noisy pseudo labels. We further stabilize self-training via aggregating the predictions from different rounds during sample selection. Experiments on eight tasks show that our proposed method outperforms the strongest self-training baseline with 1.83% and 2.51% performance gain for text and graph datasets on average. Our further analysis demonstrates that our proposed data selection strategy reduces the noise of pseudo labels by 36.8% and saves 57.3% of the time when compared with the best baseline. Our code and appendices will be uploaded to https://github.com/ritaranx/NeST.