Advancing PICO element detection in biomedical text via deep neural networks.

MOTIVATION: In evidence-based medicine, defining a clinical question in terms of the specific patient problem aids the physicians to efficiently identify appropriate resources and search for the best available evidence for medical treatment. In order to formulate a well-defined, focused clinical question, a framework called PICO is widely used, which identifies the sentences in a given medical text that belong to the four components typically reported in clinical trials: Participants/Problem (P), Intervention (I), Comparison (C) and Outcome (O). In this work, we propose a novel deep learning model for recognizing PICO elements in biomedical abstracts. Based on the previous state-of-the-art bidirectional long-short-term memory (bi-LSTM) plus conditional random field architecture, we add another layer of bi-LSTM upon the sentence representation vectors so that the contextual information from surrounding sentences can be gathered to help infer the interpretation of the current one. In addition, we propose two methods to further generalize and improve the model: adversarial training and unsupervised pre-training over large corpora. RESULTS: We tested our proposed approach over two benchmark datasets. One is the PubMed-PICO dataset, where our best results outperform the previous best by 5.5%, 7.9% and 5.8% for P, I and O elements in terms of F1 score, respectively. And for the other dataset named NICTA-PIBOSO, the improvements for P/I/O elements are 3.9%, 15.6% and 1.3% in F1 score, respectively. Overall, our proposed deep learning model can obtain unprecedented PICO element detection accuracy while avoiding the need for any manual feature selection. AVAILABILITY AND IMPLEMENTATION: Code is available at https://github.com/jind11/Deep-PICO-Detection.

Genetic Misdiagnoses and the Potential for Health Disparities.

BACKGROUND: For more than a decade, risk stratification for hypertrophic cardiomyopathy has been enhanced by targeted genetic testing. Using sequencing results, clinicians routinely assess the risk of hypertrophic cardiomyopathy in a patient's relatives and diagnose the condition in patients who have ambiguous clinical presentations. However, the benefits of genetic testing come with the risk that variants may be misclassified. METHODS: Using publicly accessible exome data, we identified variants that have previously been considered causal in hypertrophic cardiomyopathy and that are overrepresented in the general population. We studied these variants in diverse populations and reevaluated their initial ascertainments in the medical literature. We reviewed patient records at a leading genetic-testing laboratory for occurrences of these variants during the near-decade-long history of the laboratory. RESULTS: Multiple patients, all of whom were of African or unspecified ancestry, received positive reports, with variants misclassified as pathogenic on the basis of the understanding at the time of testing. Subsequently, all reported variants were recategorized as benign. The mutations that were most common in the general population were significantly more common among black Americans than among white Americans (P<0.001). Simulations showed that the inclusion of even small numbers of black Americans in control cohorts probably would have prevented these misclassifications. We identified methodologic shortcomings that contributed to these errors in the medical literature. CONCLUSIONS: The misclassification of benign variants as pathogenic that we found in our study shows the need for sequencing the genomes of diverse populations, both in asymptomatic controls and the tested patient population. These results expand on current guidelines, which recommend the use of ancestry-matched controls to interpret variants. As additional populations of different ancestry backgrounds are sequenced, we expect variant reclassifications to increase, particularly for ancestry groups that have historically been less well studied. (Funded by the National Institutes of Health.).

Bridging semantics and syntax with graph algorithms-state-of-the-art of extracting biomedical relations.

Research on extracting biomedical relations has received growing attention recently, with numerous biological and clinical applications including those in pharmacogenomics, clinical trial screening and adverse drug reaction detection. The ability to accurately capture both semantic and syntactic structures in text expressing these relations becomes increasingly critical to enable deep understanding of scientific papers and clinical narratives. Shared task challenges have been organized by both bioinformatics and clinical informatics communities to assess and advance the state-of-the-art research. Significant progress has been made in algorithm development and resource construction. In particular, graph-based approaches bridge semantics and syntax, often achieving the best performance in shared tasks. However, a number of problems at the frontiers of biomedical relation extraction continue to pose interesting challenges and present opportunities for great improvement and fruitful research. In this article, we place biomedical relation extraction against the backdrop of its versatile applications, present a gentle introduction to its general pipeline and shared resources, review the current state-of-the-art in methodology advancement, discuss limitations and point out several promising future directions.

Tensor factorization toward precision medicine.

Precision medicine initiatives come amid the rapid growth in quantity and variety of biomedical data, which exceeds the capacity of matrix-oriented data representations and many current analysis algorithms. Tensor factorizations extend the matrix view to multiple modalities and support dimensionality reduction methods that identify latent groups of data for meaningful summarization of both features and instances. In this opinion article, we analyze the modest literature on applying tensor factorization to various biomedical fields including genotyping and phenotyping. Based on the cited work including work of our own, we suggest that tensor applications could serve as an effective tool to enable frequent updating of medical knowledge based on the continually growing scientific and clinical evidence. We encourage extensive experimental studies to tackle challenges including design choice of factorizations, integrating temporality and algorithm scalability.

Learning a Comorbidity-Driven Taxonomy of Pediatric Pulmonary Hypertension.

RATIONALE: Pediatric pulmonary hypertension (PH) is a heterogeneous condition with varying natural history and therapeutic response. Precise classification of PH subtypes is, therefore, crucial for individualizing care. However, gaps remain in our understanding of the spectrum of PH in children. OBJECTIVE: We seek to study the manifestations of PH in children and to assess the feasibility of applying a network-based approach to discern disease subtypes from comorbidity data recorded in longitudinal data sets. METHODS AND RESULTS: A retrospective cohort study comprising 6 943 263 children (<18 years of age) enrolled in a commercial health insurance plan in the United States, between January 2010 and May 2013. A total of 1583 (0.02%) children met the criteria for PH. We identified comorbidities significantly associated with PH compared with the general population of children without PH. A Bayesian comorbidity network was constructed to model the interdependencies of these comorbidities, and network-clustering analysis was applied to derive disease subtypes comprising subgraphs of highly connected comorbid conditions. A total of 186 comorbidities were found to be significantly associated with PH. Network analysis of comorbidity patterns captured most of the major PH subtypes with known pathological basis defined by the World Health Organization and Panama classifications. The analysis further identified many subtypes documented in only a few case studies, including rare subtypes associated with several well-described genetic syndromes. CONCLUSIONS: Application of network science to model comorbidity patterns recorded in longitudinal data sets can facilitate the discovery of disease subtypes. Our analysis relearned established subtypes, thus validating the approach, and identified rare subtypes that are difficult to discern through clinical observations, providing impetus for deeper investigation of the disease subtypes that will enrich current disease classifications.

Medical subdomain classification of clinical notes using a machine learning-based natural language processing approach.

BACKGROUND: The medical subdomain of a clinical note, such as cardiology or neurology, is useful content-derived metadata for developing machine learning downstream applications. To classify the medical subdomain of a note accurately, we have constructed a machine learning-based natural language processing (NLP) pipeline and developed medical subdomain classifiers based on the content of the note. METHODS: We constructed the pipeline using the clinical NLP system, clinical Text Analysis and Knowledge Extraction System (cTAKES), the Unified Medical Language System (UMLS) Metathesaurus, Semantic Network, and learning algorithms to extract features from two datasets - clinical notes from Integrating Data for Analysis, Anonymization, and Sharing (iDASH) data repository (n = 431) and Massachusetts General Hospital (MGH) (n = 91,237), and built medical subdomain classifiers with different combinations of data representation methods and supervised learning algorithms. We evaluated the performance of classifiers and their portability across the two datasets. RESULTS: The convolutional recurrent neural network with neural word embeddings trained-medical subdomain classifier yielded the best performance measurement on iDASH and MGH datasets with area under receiver operating characteristic curve (AUC) of 0.975 and 0.991, and F1 scores of 0.845 and 0.870, respectively. Considering better clinical interpretability, linear support vector machine-trained medical subdomain classifier using hybrid bag-of-words and clinically relevant UMLS concepts as the feature representation, with term frequency-inverse document frequency (tf-idf)-weighting, outperformed other shallow learning classifiers on iDASH and MGH datasets with AUC of 0.957 and 0.964, and F1 scores of 0.932 and 0.934 respectively. We trained classifiers on one dataset, applied to the other dataset and yielded the threshold of F1 score of 0.7 in classifiers for half of the medical subdomains we studied. CONCLUSION: Our study shows that a supervised learning-based NLP approach is useful to develop medical subdomain classifiers. The deep learning algorithm with distributed word representation yields better performance yet shallow learning algorithms with the word and concept representation achieves comparable performance with better clinical interpretability. Portable classifiers may also be used across datasets from different institutions.

Improving the power of genetic association tests with imperfect phenotype derived from electronic medical records.

To reduce costs and improve clinical relevance of genetic studies, there has been increasing interest in performing such studies in hospital-based cohorts by linking phenotypes extracted from electronic medical records (EMRs) to genotypes assessed in routinely collected medical samples. A fundamental difficulty in implementing such studies is extracting accurate information about disease outcomes and important clinical covariates from large numbers of EMRs. Recently, numerous algorithms have been developed to infer phenotypes by combining information from multiple structured and unstructured variables extracted from EMRs. Although these algorithms are quite accurate, they typically do not provide perfect classification due to the difficulty in inferring meaning from the text. Some algorithms can produce for each patient a probability that the patient is a disease case. This probability can be thresholded to define case-control status, and this estimated case-control status has been used to replicate known genetic associations in EMR-based studies. However, using the estimated disease status in place of true disease status results in outcome misclassification, which can diminish test power and bias odds ratio estimates. We propose to instead directly model the algorithm-derived probability of being a case. We demonstrate how our approach improves test power and effect estimation in simulation studies, and we describe its performance in a study of rheumatoid arthritis. Our work provides an easily implemented solution to a major practical challenge that arises in the use of EMR data, which can facilitate the use of EMR infrastructure for more powerful, cost-effective, and diverse genetic studies.

Genetic basis of autoantibody positive and negative rheumatoid arthritis risk in a multi-ethnic cohort derived from electronic health records.

Discovering and following up on genetic associations with complex phenotypes require large patient cohorts. This is particularly true for patient cohorts of diverse ancestry and clinically relevant subsets of disease. The ability to mine the electronic health records (EHRs) of patients followed as part of routine clinical care provides a potential opportunity to efficiently identify affected cases and unaffected controls for appropriate-sized genetic studies. Here, we demonstrate proof-of-concept that it is possible to use EHR data linked with biospecimens to establish a multi-ethnic case-control cohort for genetic research of a complex disease, rheumatoid arthritis (RA). In 1,515 EHR-derived RA cases and 1,480 controls matched for both genetic ancestry and disease-specific autoantibodies (anti-citrullinated protein antibodies [ACPA]), we demonstrate that the odds ratios and aggregate genetic risk score (GRS) of known RA risk alleles measured in individuals of European ancestry within our EHR cohort are nearly identical to those derived from a genome-wide association study (GWAS) of 5,539 autoantibody-positive RA cases and 20,169 controls. We extend this approach to other ethnic groups and identify a large overlap in the GRS among individuals of European, African, East Asian, and Hispanic ancestry. We also demonstrate that the distribution of a GRS based on 28 non-HLA risk alleles in ACPA+ cases partially overlaps with ACPA- subgroup of RA cases. Our study demonstrates that the genetic basis of rheumatoid arthritis risk is similar among cases of diverse ancestry divided into subsets based on ACPA status and emphasizes the utility of linking EHR clinical data with biospecimens for genetic studies.

Thromboprophylaxis is associated with reduced post-hospitalization venous thromboembolic events in patients with inflammatory bowel diseases.

BACKGROUND & AIMS: Patients with inflammatory bowel diseases (IBDs) have increased risk for venous thromboembolism (VTE); those who require hospitalization have particularly high risk. Few hospitalized patients with IBD receive thromboprophylaxis. We analyzed the frequency of VTE after IBD-related hospitalization, risk factors for post-hospitalization VTE, and the efficacy of prophylaxis in preventing post-hospitalization VTE. METHODS: In a retrospective study, we analyzed data from a multi-institutional cohort of patients with Crohn's disease or ulcerative colitis and at least 1 IBD-related hospitalization. Our primary outcome was a VTE event. All patients contributed person-time from the date of the index hospitalization to development of VTE, subsequent hospitalization, or end of follow-up. Our main predictor variable was pharmacologic thromboprophylaxis. Cox proportional hazard models adjusting for potential confounders were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). RESULTS: From a cohort of 2788 patients with at least 1 IBD-related hospitalization, 62 patients developed VTE after discharge (2%). Incidences of VTE at 30, 60, 90, and 180 days after the index hospitalization were 3.7/1000, 4.1/1000, 5.4/1000, and 9.4/1000 person-days, respectively. Pharmacologic thromboprophylaxis during the index hospital stay was associated with a significantly lower risk of post-hospitalization VTE (HR, 0.46; 95% CI, 0.22-0.97). Increased numbers of comorbidities (HR, 1.30; 95% CI, 1.16-1.47) and need for corticosteroids before hospitalization (HR, 1.71; 95% CI, 1.02-2.87) were also independently associated with risk of VTE. Length of hospitalization or surgery during index hospitalization was not associated with post-hospitalization VTE. CONCLUSIONS: Pharmacologic thromboprophylaxis during IBD-related hospitalization is associated with reduced risk of post-hospitalization VTE.

Serum inflammatory markers and risk of colorectal cancer in patients with inflammatory bowel diseases.

BACKGROUND & AIMS: Patients with inflammatory bowel diseases (IBDs) (Crohn's disease, ulcerative colitis) are at increased risk of colorectal cancer (CRC). Persistent inflammation is hypothesized to increase risk of CRC in patients with IBD; however, the few studies in this area have been restricted to cross-sectional assessments of histologic severity. No prior studies have examined association between C-reactive protein (CRP) or erythrocyte sedimentation rate (ESR) elevation and risk of CRC in an IBD cohort. METHODS: From a multi-institutional validated IBD cohort, we identified all patients with at least one measured CRP or ESR value. Patients were stratified into quartiles of severity of inflammation on the basis of their median CRP or ESR value, and subsequent diagnosis of CRC was ascertained. Logistic regression adjusting for potential confounders was used to identify the independent association between CRP or ESR elevation and risk of CRC. RESULTS: Our study included 3145 patients with at least 1 CRP value (CRP cohort) and 4008 with at least 1 ESR value (ESR cohort). Thirty-three patients in the CRP cohort and 102 patients in the ESR cohort developed CRC during a median follow-up of 5 years at a median age of 55 years. On multivariate analysis, there was a significant increase in risk of CRC across quartiles of CRP elevation (P(trend) = .017; odds ratio for quartile 4 vs quartile 1, 2.72; 95% confidence interval, 0.95-7.76). Similarly higher median ESR was also independently associated with risk of CRC across the quartiles (odds ratio, 2.06; 95% confidence interval, 1.14-3.74) (P(trend) = .007). CONCLUSIONS: An elevated CRP or ESR is associated with increased risk of CRC in patients with IBD.

Association between reduced plasma 25-hydroxy vitamin D and increased risk of cancer in patients with inflammatory bowel diseases.

BACKGROUND & AIMS: Vitamin D deficiency is common among patients with inflammatory bowel diseases (IBD) (Crohn's disease or ulcerative colitis). The effects of low plasma 25-hydroxy vitamin D (25[OH]D) on outcomes other than bone health are understudied in patients with IBD. We examined the association between plasma level of 25(OH)D and risk of cancers in patients with IBD. METHODS: From a multi-institutional cohort of patients with IBD, we identified those with at least 1 measurement of plasma 25(OH)D. The primary outcome was development of any cancer. We examined the association between plasma 25(OH)D and risk of specific subtypes of cancer, adjusting for potential confounders in a multivariate regression model. RESULTS: We analyzed data from 2809 patients with IBD and a median plasma level of 25(OH)D of 26 ng/mL. Nearly one-third had deficient levels of vitamin D (<20 ng/mL). During a median follow-up period of 11 years, 196 patients (7%) developed cancer, excluding nonmelanoma skin cancer (41 cases of colorectal cancer). Patients with vitamin D deficiency had an increased risk of cancer (adjusted odds ratio, 1.82; 95% confidence interval, 1.25-2.65) compared with those with sufficient levels. Each 1-ng/mL increase in plasma 25(OH)D was associated with an 8% reduction in risk of colorectal cancer (odds ratio, 0.92; 95% confidence interval, 0.88-0.96). A weaker inverse association was also identified for lung cancer. CONCLUSIONS: In a large multi-institutional IBD cohort, a low plasma level of 25(OH)D was associated with an increased risk of cancer, especially colorectal cancer.

Association between low density lipoprotein and rheumatoid arthritis genetic factors with low density lipoprotein levels in rheumatoid arthritis and non-rheumatoid arthritis controls.

OBJECTIVES: While genetic determinants of low density lipoprotein (LDL) cholesterol levels are well characterised in the general population, they are understudied in rheumatoid arthritis (RA). Our objective was to determine the association of established LDL and RA genetic alleles with LDL levels in RA cases compared with non-RA controls. METHODS: Using data from electronic medical records, we linked validated RA cases and non-RA controls to discarded blood samples. For each individual, we extracted data on: first LDL measurement, age, gender and year of LDL measurement. We genotyped subjects for 11 LDL and 44 non-HLA RA alleles, and calculated RA and LDL genetic risk scores (GRS). We tested the association between each GRS and LDL level using multivariate linear regression models adjusted for age, gender, year of LDL measurement and RA status. RESULTS: Among 567 RA cases and 979 controls, 80% were female and mean age at the first LDL measurement was 55 years. RA cases had significantly lower mean LDL levels than controls (117.2 vs 125.6 mg/dl, respectively, p<0.0001). Each unit increase in LDL GRS was associated with 0.8 mg/dl higher LDL levels in both RA cases and controls (p=3.0×10(-7)). Each unit increase in RA GRS was associated with 4.3 mg/dl lower LDL levels in both groups (p=0.01). CONCLUSIONS: LDL alleles were associated with higher LDL levels in RA. RA alleles were associated with lower LDL levels in both RA cases and controls. As RA cases carry more RA alleles, these findings suggest a genetic basis for epidemiological observations of lower LDL levels in RA.

Associations of autoantibodies, autoimmune risk alleles, and clinical diagnoses from the electronic medical records in rheumatoid arthritis cases and non-rheumatoid arthritis controls.

OBJECTIVE: The significance of non-rheumatoid arthritis (RA) autoantibodies in patients with RA is unclear. The aim of this study was to assess associations of autoantibodies with autoimmune risk alleles and with clinical diagnoses from the electronic medical records (EMRs) among RA cases and non-RA controls. METHODS: Data on 1,290 RA cases and 1,236 non-RA controls of European genetic ancestry were obtained from the EMRs of 2 large academic centers. The levels of anti-citrullinated protein antibodies (ACPAs), antinuclear antibodies (ANAs), anti-tissue transglutaminase antibodies (AGTAs), and anti-thyroid peroxidase (anti-TPO) antibodies were measured. All subjects were genotyped for autoimmune risk alleles, and the association between number of autoimmune risk alleles present and number of types of autoantibodies present was studied. A phenome-wide association study (PheWAS) was conducted to study potential associations between autoantibodies and clinical diagnoses among RA cases and non-RA controls. RESULTS: The mean ages were 60.7 years in RA cases and 64.6 years in non-RA controls. The proportion of female subjects was 79% in each group. The prevalence of ACPAs and ANAs was higher in RA cases compared to controls (each P < 0.0001); there were no differences in the prevalence of anti-TPO antibodies and AGTAs. Carriage of higher numbers of autoimmune risk alleles was associated with increasing numbers of autoantibody types in RA cases (P = 2.1 × 10(-5)) and non-RA controls (P = 5.0 × 10(-3)). From the PheWAS, the presence of ANAs was significantly associated with a diagnosis of Sjögren's/sicca syndrome in RA cases. CONCLUSION: The increased frequency of autoantibodies in RA cases and non-RA controls was associated with the number of autoimmune risk alleles carried by an individual. PheWAS of EMR data, with linkage to laboratory data obtained from blood samples, provide a novel method to test for the clinical significance of biomarkers in disease.

Decision support from local data: creating adaptive order menus from past clinician behavior.

OBJECTIVE: Reducing care variability through guidelines has significantly benefited patients. Nonetheless, guideline-based Clinical Decision Support (CDS) systems are not widely implemented or used, are frequently out-of-date, and cannot address complex care for which guidelines do not exist. Here, we develop and evaluate a complementary approach - using Bayesian Network (BN) learning to generate adaptive, context-specific treatment menus based on local order-entry data. These menus can be used as a draft for expert review, in order to minimize development time for local decision support content. This is in keeping with the vision outlined in the US Health Information Technology Strategic Plan, which describes a healthcare system that learns from itself. MATERIALS AND METHODS: We used the Greedy Equivalence Search algorithm to learn four 50-node domain-specific BNs from 11,344 encounters: abdominal pain in the emergency department, inpatient pregnancy, hypertension in the Urgent Visit Clinic, and altered mental state in the intensive care unit. We developed a system to produce situation-specific, rank-ordered treatment menus from these networks. We evaluated this system with a hospital-simulation methodology and computed Area Under the Receiver-Operator Curve (AUC) and average menu position at time of selection. We also compared this system with a similar association-rule-mining approach. RESULTS: A short order menu on average contained the next order (weighted average length 3.91-5.83 items). Overall predictive ability was good: average AUC above 0.9 for 25% of order types and overall average AUC .714-.844 (depending on domain). However, AUC had high variance (.50-.99). Higher AUC correlated with tighter clusters and more connections in the graphs, indicating importance of appropriate contextual data. Comparison with an Association Rule Mining approach showed similar performance for only the most common orders with dramatic divergence as orders are less frequent. DISCUSSION AND CONCLUSION: This study demonstrates that local clinical knowledge can be extracted from treatment data for decision support. This approach is appealing because: it reflects local standards; it uses data already being captured; and it produces human-readable treatment-diagnosis networks that could be curated by a human expert to reduce workload in developing localized CDS content. The BN methodology captured transitive associations and co-varying relationships, which existing approaches do not. It also performs better as orders become less frequent and require more context. This system is a step forward in harnessing local, empirical data to enhance decision support.

Using machine learning to develop smart reflex testing protocols.

OBJECTIVE: Reflex testing protocols allow clinical laboratories to perform second line diagnostic tests on existing specimens based on the results of initially ordered tests. Reflex testing can support optimal clinical laboratory test ordering and diagnosis. In current clinical practice, reflex testing typically relies on simple "if-then" rules; however, this limits the opportunities for reflex testing since most test ordering decisions involve more complexity than traditional rule-based approaches would allow. Here, using the analyte ferritin as an example, we propose an alternative machine learning-based approach to "smart" reflex testing. METHODS: Using deidentified patient data, we developed a machine learning model to predict whether a patient getting CBC testing will also have ferritin testing ordered. We evaluate applications of this model to reflex testing by assessing its performance in comparison to possible rule-based approaches. RESULTS: Our underlying machine learning models performed moderately well in predicting ferritin test ordering (AUC=0.731 in reference to actual ordering) and demonstrated promising potential to underlie key clinical applications. In contrast, none of the many traditionally framed, rule-based, hypothetical reflex protocols we evaluated offered sufficient agreement with actual ordering to be clinically feasible. Using chart review, we further demonstrated that the strategic deployment of our model could avoid important ferritin test ordering errors. CONCLUSIONS: Machine learning may provide a foundation for new types of reflex testing with enhanced benefits for clinical diagnosis.

Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study.

BACKGROUND: Large language models (LLMs) such as GPT-4 hold great promise as transformative tools in health care, ranging from automating administrative tasks to augmenting clinical decision making. However, these models also pose a danger of perpetuating biases and delivering incorrect medical diagnoses, which can have a direct, harmful impact on medical care. We aimed to assess whether GPT-4 encodes racial and gender biases that impact its use in health care. METHODS: Using the Azure OpenAI application interface, this model evaluation study tested whether GPT-4 encodes racial and gender biases and examined the impact of such biases on four potential applications of LLMs in the clinical domain-namely, medical education, diagnostic reasoning, clinical plan generation, and subjective patient assessment. We conducted experiments with prompts designed to resemble typical use of GPT-4 within clinical and medical education applications. We used clinical vignettes from NEJM Healer and from published research on implicit bias in health care. GPT-4 estimates of the demographic distribution of medical conditions were compared with true US prevalence estimates. Differential diagnosis and treatment planning were evaluated across demographic groups using standard statistical tests for significance between groups. FINDINGS: We found that GPT-4 did not appropriately model the demographic diversity of medical conditions, consistently producing clinical vignettes that stereotype demographic presentations. The differential diagnoses created by GPT-4 for standardised clinical vignettes were more likely to include diagnoses that stereotype certain races, ethnicities, and genders. Assessment and plans created by the model showed significant association between demographic attributes and recommendations for more expensive procedures as well as differences in patient perception. INTERPRETATION: Our findings highlight the urgent need for comprehensive and transparent bias assessments of LLM tools such as GPT-4 for intended use cases before they are integrated into clinical care. We discuss the potential sources of these biases and potential mitigation strategies before clinical implementation. FUNDING: Priscilla Chan and Mark Zuckerberg.

Similar risk of depression and anxiety following surgery or hospitalization for Crohn's disease and ulcerative colitis.

OBJECTIVES: Psychiatric comorbidity is common in Crohn's disease (CD) and ulcerative colitis (UC). Inflammatory bowel disease (IBD)-related surgery or hospitalizations represent major events in the natural history of the disease. The objective of this study is to examine whether there is a difference in the risk of psychiatric comorbidity following surgery in CD and UC. METHODS: We used a multi-institution cohort of IBD patients without a diagnosis code for anxiety or depression preceding their IBD-related surgery or hospitalization. Demographic-, disease-, and treatment-related variables were retrieved. Multivariate logistic regression analysis was performed to individually identify risk factors for depression and anxiety. RESULTS: Our study included a total of 707 CD and 530 UC patients who underwent bowel resection surgery and did not have depression before surgery. The risk of depression 5 years after surgery was 16% and 11% in CD and UC patients, respectively. We found no difference in the risk of depression following surgery in the CD and UC patients (adjusted odds ratio, 1.11; 95% confidence interval, 0.84-1.47). Female gender, comorbidity, immunosuppressant use, perianal disease, stoma surgery, and early surgery within 3 years of care predicted depression after CD surgery; only the female gender and comorbidity predicted depression in UC patients. Only 12% of the CD cohort had ≥4 risk factors for depression, but among them nearly 44% subsequently received a diagnosis code for depression. CONCLUSIONS: IBD-related surgery or hospitalization is associated with a significant risk for depression and anxiety, with a similar magnitude of risk in both diseases.

Clinical Artificial Intelligence: Design Principles and Fallacies.

Clinical artificial intelligence (AI)/machine learning (ML) is anticipated to offer new abilities in clinical decision support, diagnostic reasoning, precision medicine, clinical operational support, and clinical research, but careful concern is needed to ensure these technologies work effectively in the clinic. Here, we detail the clinical ML/AI design process, identifying several key questions and detailing several common forms of issues that arise with ML tools, as motivated by real-world examples, such that clinicians and researchers can better anticipate and correct for such issues in their own use of ML/AI techniques.

Using Machine Learning to Develop Smart Reflex Testing Protocols.

Objective: Reflex testing protocols allow clinical laboratories to perform second line diagnostic tests on existing specimens based on the results of initially ordered tests. Reflex testing can support optimal clinical laboratory test ordering and diagnosis. In current clinical practice, reflex testing typically relies on simple "if-then" rules; however, this limits their scope since most test ordering decisions involve more complexity than a simple rule will allow. Here, using the analyte ferritin as an example, we propose an alternative machine learning-based approach to "smart" reflex testing with a wider scope and greater impact than traditional rule-based approaches. Methods: Using patient data, we developed a machine learning model to predict whether a patient getting CBC testing will also have ferritin testing ordered, consider applications of this model to "smart" reflex testing, and evaluate the model by comparing its performance to possible rule-based approaches. Results: Our underlying machine learning models performed moderately well in predicting ferritin test ordering and demonstrated greater suitability to reflex testing than rule-based approaches. Using chart review, we demonstrate that our model may improve ferritin test ordering. Finally, as a secondary goal, we demonstrate that ferritin test results are missing not at random (MNAR), a finding with implications for unbiased imputation of missing test results. Conclusions: Machine learning may provide a foundation for new types of reflex testing with enhanced benefits for clinical diagnosis and laboratory utilization management.