Use of noisy labels as weak learners to identify incompletely ascertainable outcomes: A Feasibility study with opioid-induced respiratory depression

Objective

Assigning outcome labels to large observational data sets in a timely and accurate manner, particularly when outcomes are rare or not directly ascertainable, remains a significant challenge within biomedical informatics. We examined whether noisy labels generated from subject matter experts' heuristics using heterogenous data types within a data programming paradigm could provide outcomes labels to a large, observational data set. We chose the clinical condition of opioid-induced respiratory depression for our use case because it is rare, has no administrative codes to easily identify the condition, and typically requires at least some unstructured text to ascertain its presence.

Materials and methods

Using de-identified electronic health records of 52,861 post-operative encounters, we applied a data programming paradigm (implemented in the Snorkel software) for the development of a machine learning classifier for opioid-induced respiratory depression. Our approach included subject matter experts creating 14 labeling functions that served as noisy labels for developing a probabilistic Generative model. We used probabilistic labels from the Generative model as outcome labels for training a Discriminative model on the source data. We evaluated performance of the Discriminative model with a hold-out test set of 599 independently-reviewed patient records.

Results

The final Discriminative classification model achieved an accuracy of 0.977, an F1 score of 0.417, a sensitivity of 1.0, and an AUC of 0.988 in the hold-out test set with a prevalence of 0.83% (5/599).

Discussion

All of the confirmed Cases were identified by the classifier. For rare outcomes, this finding is encouraging because it reduces the number of manual reviews needed by excluding visits/patients with low probabilities.

Conclusion

Application of a data programming paradigm with expert-informed labeling functions might have utility for phenotyping clinical phenomena that are not easily ascertainable from highly-structured data.

Development of a 3-Step theory of suicide ontology to facilitate 3ST factor extraction from clinical progress notes

OBJECTIVE

Suicide risk prediction algorithms at the Veterans Health Administration (VHA) do not include predictors based on the 3-Step Theory of suicide (3ST), which builds on hopelessness, psychological pain, connectedness, and capacity for suicide. These four factors are not available from structured fields in VHA electronic health records, but they are found in unstructured clinical text. An ontology and controlled vocabulary that maps psychosocial and behavioral terms to these factors does not exist. The objectives of this study were 1) to develop an ontology with a controlled vocabulary of terms that map onto classes that represent the 3ST factors as identified within electronic clinical progress notes, and 2) to determine the accuracy of automated extractions based on terms in the controlled vocabulary.

METHODS

A team of four annotators did linguistic annotation of 30,000 clinical progress notes from 231 Veterans in VHA electronic health records who attempted suicide or who died by suicide for terms relating to the 3ST factors. Annotation involved manually assigning a label to words or phrases that indicated presence or absence of the factor (polarity). These words and phrases were entered into a controlled vocabulary that was then used by our computational system to tag 14 million clinical progress notes from Veterans who attempted or died by suicide after 2013. Tagged text was extracted and machine-labelled for presence or absence of the 3ST factors. Accuracy of these machine-labels was determined for 1000 randomly selected extractions for each factor against a ground truth created by our annotators.

RESULTS

Linguistic annotation identified 8486 terms that related to 33 subclasses across the four factors and polarities. Precision of machine-labeled extractions ranged from 0.73 to 1.00 for most factor-polarity combinations, whereas recall was somewhat lower 0.65-0.91.

CONCLUSION

The ontology that was developed consists of classes that represent each of the four 3ST factors, subclasses, relationships, and terms that map onto those classes which are stored in a controlled vocabulary (https://bioportal.bioontology.org/ontologies/THREE-ST). The use case that we present shows how scores based on clinical notes tagged for terms in the controlled vocabulary capture meaningful change in the 3ST factors during weeks preceding a suicidal event.

Use of Noisy Labels as Weak Learners to Identify Incompletely Ascertainable Outcomes: A Feasibility Study with Opioid-Induced Respiratory Depression

Objective

Assigning outcome labels to large observational data sets in a timely and accurate manner, particularly when outcomes are rare or not directly ascertainable, remains a significant challenge within biomedical informatics. We examined whether noisy labels generated from subject matter experts' heuristics using heterogenous data types within a data programming paradigm could provide outcomes labels to a large, observational data set. We chose the clinical condition of opioid-induced respiratory depression for our use case because it is rare, has no administrative codes to easily identify the condition, and typically requires at least some unstructured text to ascertain its presence.

Materials and Methods

Using de-identified electronic health records of 52,861 post-operative encounters, we applied a data programming paradigm (implemented in the Snorkel software) for the development of a machine learning classifier for opioid-induced respiratory depression. Our approach included subject matter experts creating 14 labeling functions that served as noisy labels for developing a probabilistic Generative model. We used probabilistic labels from the Generative model as outcome labels for training a Discriminative model on the source data. We evaluated performance of the Discriminative model with a hold-out test set of 599 independently-reviewed patient records.

Results

The final Discriminative classification model achieved an accuracy of 0.977, an F1 score of 0.417, a sensitivity of 1.0, and an AUC of 0.988 in the hold-out test set with a prevalence of 0.83% (5/599).

Discussion

All of the confirmed Cases were identified by the classifier. For rare outcomes, this finding is encouraging because it reduces the number of manual reviews needed by excluding visits/patients with low probabilities.

Conclusion

Application of a data programming paradigm with expert-informed labeling functions might have utility for phenotyping clinical phenomena that are not easily ascertainable from highly-structured data.

Representing and utilizing clinical textual data for real world studies: An OHDSI approach

Clinical documentation in electronic health records contains crucial narratives and details about patients and their care. Natural language processing (NLP) can unlock the information conveyed in clinical notes and reports, and thus plays a critical role in real-world studies. The NLP Working Group at the Observational Health Data Sciences and Informatics (OHDSI) consortium was established to develop methods and tools to promote the use of textual data and NLP in real-world observational studies. In this paper, we describe a framework for representing and utilizing textual data in real-world evidence generation, including representations of information from clinical text in the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM), the workflow and tools that were developed to extract, transform and load (ETL) data from clinical notes into tables in OMOP CDM, as well as current applications and specific use cases of the proposed OHDSI NLP solution at large consortia and individual institutions with English textual data. Challenges faced and lessons learned during the process are also discussed to provide valuable insights for researchers who are planning to implement NLP solutions in real-world studies.

Suicide theory-guided natural language processing of clinical progress notes to improve prediction of veteran suicide risk: protocol for a mixed-method study

INTRODUCTION

The state-of-the-art 3-step Theory of Suicide (3ST) describes why people consider suicide and who will act on their suicidal thoughts and attempt suicide. The central concepts of 3ST-psychological pain, hopelessness, connectedness, and capacity for suicide-are among the most important drivers of suicidal behaviour but they are missing from clinical suicide risk prediction models in use at the US Veterans Health Administration (VHA). These four concepts are not systematically recorded in structured fields of VHA's electronic healthcare records. Therefore, this study will develop a domain-specific ontology that will enable automated extraction of these concepts from clinical progress notes using natural language processing (NLP), and test whether NLP-based predictors for these concepts improve accuracy of existing VHA suicide risk prediction models.

METHODS AND ANALYSIS

Our mixed-method study has an exploratory sequential design where a qualitative component (aim 1) will inform quantitative analyses (aims 2 and 3). For aim 1, subject matter experts will manually annotate progress notes of clinical encounters with veterans who attempted or died by suicide to develop a domain-specific ontology for the 3ST concepts. During aim 2, we will use NLP to machine-annotate clinical progress notes and derive longitudinal representations for each patient with respect to the presence and intensity of hopelessness, psychological pain, connectedness and capacity for suicide in temporal proximity of suicide attempts and deaths by suicide. These longitudinal representations will be evaluated during aim 3 for their ability to improve existing VHA prediction models of suicide and suicide attempts, STORM (Stratification Tool for Opioid Risk Mitigation) and REACHVET (Recovery Engagement and Coordination for Health - Veterans Enhanced Treatment).

ETHICS AND DISSEMINATION

Ethics approval for this study was granted by the Stanford University Institutional Review Board and the Research and Development Committee of the VA Palo Alto Health Care System. Results of the study will be disseminated through several outlets, including peer-reviewed publications and presentations at national conferences.

Information Extraction From Electronic Health Records to Predict Readmission Following Acute Myocardial Infarction: Does Natural Language Processing Using Clinical Notes Improve Prediction of Readmission?

Background

Social risk factors influence rehospitalization rates yet are challenging to incorporate into prediction models. Integration of social risk factors using natural language processing (NLP) and machine learning could improve risk prediction of 30‐day readmission following an acute myocardial infarction.

Methods and Results

Patients were enrolled into derivation and validation cohorts. The derivation cohort included inpatient discharges from Vanderbilt University Medical Center between January 1, 2007, and December 31, 2016, with a primary diagnosis of acute myocardial infarction, who were discharged alive, and not transferred from another facility. The validation cohort included patients from Dartmouth‐Hitchcock Health Center between April 2, 2011, and December 31, 2016, meeting the same eligibility criteria described above. Data from both sites were linked to Centers for Medicare & Medicaid Services administrative data to supplement 30‐day hospital readmissions. Clinical notes from each cohort were extracted, and an NLP model was deployed, counting mentions of 7 social risk factors. Five machine learning models were run using clinical and NLP‐derived variables. Model discrimination and calibration were assessed, and receiver operating characteristic comparison analyses were performed. The 30‐day rehospitalization rates among the derivation (n=6165) and validation (n=4024) cohorts were 15.1% (n=934) and 10.2% (n=412), respectively. The derivation models demonstrated no statistical improvement in model performance with the addition of the selected NLP‐derived social risk factors.

Conclusions

Social risk factors extracted using NLP did not significantly improve 30‐day readmission prediction among hospitalized patients with acute myocardial infarction. Alternative methods are needed to capture social risk factors.

Adaptation of an NLP system to a new healthcare environment to identify social determinants of health

Social determinants of health (SDoH) are increasingly important factors for population health, healthcare outcomes, and care delivery. However, many of these factors are not reliably captured within structured electronic health record (EHR) data. In this work, we evaluated and adapted a previously published NLP tool to include additional social risk factors for deployment at Vanderbilt University Medical Center in an Acute Myocardial Infarction cohort. We developed a transformation of the SDoH outputs of the tool into the OMOP common data model (CDM) for re-use across many potential use cases, yielding performance measures across 8 SDoH classes of precision 0.83 recall 0.74 and F-measure of 0.78.

Determining post-test risk in a national sample of stress nuclear myocardial perfusion imaging reports: Implications for natural language processing tools

BACKGROUND

Reporting standards promote clarity and consistency of stress myocardial perfusion imaging (MPI) reports, but do not require an assessment of post-test risk. Natural Language Processing (NLP) tools could potentially help estimate this risk, yet it is unknown whether reports contain adequate descriptive data to use NLP.

METHODS

Among VA patients who underwent stress MPI and coronary angiography between January 1, 2009 and December 31, 2011, 99 stress test reports were randomly selected for analysis. Two reviewers independently categorized each report for the presence of critical data elements essential to describing post-test ischemic risk.

RESULTS

Few stress MPI reports provided a formal assessment of post-test risk within the impression section (3%) or the entire document (4%). In most cases, risk was determinable by combining critical data elements (74% impression, 98% whole). If ischemic risk was not determinable (25% impression, 2% whole), inadequate description of systolic function (9% impression, 1% whole) and inadequate description of ischemia (5% impression, 1% whole) were most commonly implicated.

CONCLUSIONS

Post-test ischemic risk was determinable but rarely reported in this sample of stress MPI reports. This supports the potential use of NLP to help clarify risk. Further study of NLP in this context is needed.

Who said it? Establishing professional attribution among authors of Veterans' Electronic Health Records

BACKGROUND

A practical data point for assessing information quality and value in the Electronic Health Record (EHR) is the professional category of the EHR author. We evaluated and compared free form electronic signatures against LOINC note titles in categorizing the profession of EHR authors.

METHODS

A random 1000 clinical document sample was selected and divided into 500 document sets for training and testing. The gold standard for provider classification was generated by dual clinician manual review, disagreements resolved by a third reviewer. Text matching algorithms composed of document titles and author electronic signatures for provider classification were developed on the training set.

RESULTS

Overall, detection of professional classification by note titles alone resulted in 76.1% sensitivity and 69.4% specificity. The aggregate of note titles with electronic signatures resulted in 95.7% sensitivity and 98.5% specificity.

CONCLUSIONS

Note titles alone provided fair professional classification. Inclusion of author electronic signatures significantly boosted classification performance.

Detecting temporal expressions in medical narratives

BACKGROUND

Clinical practice and epidemiological information aggregation require knowing when, how long, and in what sequence medically relevant events occur. The Temporal Awareness and Reasoning Systems for Question Interpretation (TARSQI) Toolkit (TTK) is a complete, open source software package for the temporal ordering of events within narrative text documents. TTK was developed on newspaper articles. We extended TTK to support medical notes using veterans' affairs (VA) clinical notes and compared it to TTK.

METHODS

We used a development set consisting of 200 VA clinical notes to modify and append rules to TTK's time tagger, creating Med-TTK. We then evaluated the performances of TTK and Med-TTK on an independent random selection of 100 clinical notes. Evaluation tasks were to identify and classify time-referring expressions as one of four temporal classes (DATE, TIME, DURATION, and SET). The reference standard for this test set was generated by dual human manual review with disagreements resolved by a third reviewer. Outcome measures included recall and precision for each class, and inter-rater agreement scores.

RESULTS

There were 3146 temporal expressions in the reference standard. TTK identified 1595 temporal expressions. Recall was 0.15 (95% confidence interval [CI] 0.12-0.15) and precision was 0.27 (95% CI 0.25-0.29) for TTK. Med-TTK identified 3174 expressions. Recall was 0.86 (95% CI 0.84-0.87) and precision was 0.85 (95% CI 0.84-0.86) for Med-TTK.

CONCLUSION

The algorithms for identifying and classifying temporal expressions in medical narratives developed within Med-TTK significantly improved performance compared to TTK. Natural language processing applications such as Med-TTK provide a foundation for meaningful longitudinal mapping of patient history events among electronic health records. The tool can be accessed at the following site: http://code.google.com/p/med-ttk/.