Dataset Information

Pediatric sepsis phenotypes for enhanced therapeutics: An application of clustering to electronic health records

ABSTRACT: Abstract

Objective

The heterogeneity of pediatric sepsis patients suggests the potential benefits of clustering analytics to derive phenotypes with distinct host response patterns that may help guide personalized therapeutics. We evaluate the relative performance of latent class analysis (LCA) and K‐means, 2 commonly used clustering methods toward the derivation of clinically useful pediatric sepsis phenotypes.

Methods

Data were extracted from anonymized medical records of 6446 pediatric patients that presented to 1 of 6 emergency departments (EDs) between 2013 and 2018 and were thereafter admitted. Using International Classification of Diseases (ICD)‐9 and ICD‐10 discharge codes, 151 patients were identified with a sepsis continuum diagnosis that included septicemia, sepsis, severe sepsis, and septic shock. Using feature sets used in related clustering studies, LCA and K‐means algorithms were used to derive 4 distinct phenotypic pediatric sepsis segmentations. Each segmentation was evaluated for phenotypic homogeneity, separation, and clinical use.

Results

Using the 2 feature sets, LCA clustering resulted in 2 similar segmentations of 4 clinically distinct phenotypes, while K‐means clustering resulted in segmentations of 3 and 4 phenotypes. All 4 segmentations identified at least 1 high severity phenotype, but LCA‐identified phenotypes reflected superior stratification, high entropy approaching 1 (eg, 0.994) indicating excellent separation between estimated phenotypes, and differential treatment/treatment response, and outcomes that were non‐randomly distributed across phenotypes (P < 0.001).

Conclusion

Compared to K‐means, which is commonly used in clustering studies, LCA appears to be a more robust, clinically useful statistical tool in analyzing a heterogeneous pediatric sepsis cohort toward informing targeted therapies. Additional prospective studies are needed to validate clinical utility of predictive models that target derived pediatric sepsis phenotypes in emergency department settings.

SUBMITTER: Koutroulis I

PROVIDER: S-EPMC8790108 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundIdentifying implausible clinical observations (e.g., laboratory test and vital sign values) in Electronic Health Record (EHR) data using rule-based procedures is challenging. Anomaly/outlier detection methods can be applied as an alternative algorithmic approach to flagging such implausible values in EHRs.MethodsThe primary objectives of this research were to develop and test an unsupervised clustering-based anomaly/outlier detection approach for detecting implausible observations in EHR data as an alternative algorithmic solution to the existing procedures. Our approach is built upon two underlying hypotheses that, (i) when there are large number of observations, implausible records should be sparse, and therefore (ii) if these data are clustered properly, clusters with sparse populations should represent implausible observations. To test these hypotheses, we applied an unsupervised clustering algorithm to EHR observation data on 50 laboratory tests from Partners HealthCare. We tested different specifications of the clustering approach and computed confusion matrix indices against a set of silver-standard plausibility thresholds. We compared the results from the proposed approach with conventional anomaly detection (CAD) approaches, including standard deviation and Mahalanobis distance.ResultsWe found that the clustering approach produced results with exceptional specificity and high sensitivity. Compared with the conventional anomaly detection approaches, our proposed clustering approach resulted in significantly smaller number of false positive cases.ConclusionOur contributions include (i) a clustering approach for identifying implausible EHR observations, (ii) evidence that implausible observations are sparse in EHR laboratory test results, (iii) a parallel implementation of the clustering approach on i2b2 star schema, and (3) a set of silver-standard plausibility thresholds for 50 laboratory tests that can be used in other studies for validation. The proposed algorithmic solution can augment human decisions to improve data quality. Therefore, a workflow is needed to complement the algorithm's job and initiate necessary actions that need to be taken in order to improve the quality of data.

Project description:Background: Drug-induced coagulopathy (DIC) is a severe adverse reaction and has become a significantly increased clinical problem in children. It is crucial to the detection of the DIC safety signal for drug post-marketing scientific supervision purposes. Therefore, this study aimed to detect potential signals for DIC in children using the routine electronic medical record (EMR) data. Methods: This study extracted EMR data from Beijing Children's Hospital between 2009 and 2020. A two-stage modeling method was developed to detect the signal of DIC. We calculated the crude incidence by mining cases of coagulopathy to select the potential suspected drugs; then, propensity score-matched retrospective cohorts of specific screened drugs from the first stage were constructed and estimated the odds ratio (OR) and 95% confidence interval (CI) using conditional logistic regression models. The current literature evidence was used to assess the novelty of the signal. Results:In the study, from a total of 340 drugs, 22 drugs were initially screened as potentially inducing coagulopathy. In total, we identified 19 positive DIC associations. Of these, potential DIC risk of omeprazole (OR: 2.23, 95% CI: 1.88-2.65), chlorpheniramine (OR:3.04, 95% CI:2.56-3.60), and salbutamol sulfate (OR:1.36, 95% CI:1.07-1.73) were three new DIC signals in both children and adults. Twelve associations between coagulopathy and drugs, meropenem (OR: 3.38, 95% CI: 2.72-4.20), cefoperazone sulbactam (OR: 2.80, 95% CI: 2.30-3.41), fluconazole (OR: 2.11, 95% CI: 1.71-2.59), voriconazole (OR: 2.82, 95% CI: 2.20-3.61), ambroxol hydrochloride (OR: 2.12, 95% CI: 1.74-2.58), furosemide (OR: 2.36, 95% CI: 2.08-2.67), iodixanol (OR: 2.21, 95% CI: 1.72-2.85), cefamandole (OR: 1.82, 95% CI: 1.56-2.13), ceftizoxime (OR: 1.95, 95% CI: 1.44-2.63), ceftriaxone (OR: 1.95, 95% CI: 1.44-2.63), latamoxef sodium (OR: 1.76, 95% CI: 1.49-2.07), and sulfamethoxazole (OR: 1.29, 95% CI: 1.01-1.64), were considered as new signals in children. Conclusion: The two-stage algorithm developed in our study to detect safety signals of DIC found nineteen signals of DIC, including twelve new signals in a pediatric population. However, these safety signals of DIC need to be confirmed by further studies based on population study and mechanism research.

Project description:BackgroundThe main goal of this study is to explore the use of features representing patient-level electronic health record (EHR) data, generated by the unsupervised deep learning algorithm autoencoder, in predictive modeling. Since autoencoder features are unsupervised, this paper focuses on their general lower-dimensional representation of EHR information in a wide variety of predictive tasks.MethodsWe compare the model with autoencoder features to traditional models: logistic model with least absolute shrinkage and selection operator (LASSO) and Random Forest algorithm. In addition, we include a predictive model using a small subset of response-specific variables (Simple Reg) and a model combining these variables with features from autoencoder (Enhanced Reg). We performed the study first on simulated data that mimics real world EHR data and then on actual EHR data from eight Advocate hospitals.ResultsOn simulated data with incorrect categories and missing data, the precision for autoencoder is 24.16% when fixing recall at 0.7, which is higher than Random Forest (23.61%) and lower than LASSO (25.32%). The precision is 20.92% in Simple Reg and improves to 24.89% in Enhanced Reg. When using real EHR data to predict the 30-day readmission rate, the precision of autoencoder is 19.04%, which again is higher than Random Forest (18.48%) and lower than LASSO (19.70%). The precisions for Simple Reg and Enhanced Reg are 18.70 and 19.69% respectively. That is, Enhanced Reg can have competitive prediction performance compared to LASSO. In addition, results show that Enhanced Reg usually relies on fewer features under the setting of simulations of this paper.ConclusionsWe conclude that autoencoder can create useful features representing the entire space of EHR data and which are applicable to a wide array of predictive tasks. Together with important response-specific predictors, we can derive efficient and robust predictive models with less labor in data extraction and model training.

Project description:BackgroundThe exponential growth of digital healthcare data is fueling the development of Knowledge Discovery in Databases (KDD). Extracting temporal relationships between medical events is essential to reveal hidden patterns that can help physicians find optimal treatments, diagnose illnesses, detect drug adverse reactions, and more. This paper presents an approach for the extraction of patient evolution patterns from electronic health records written in Catalan and/or Spanish.MethodsWe propose a robust formulation for extracting Temporal Association Rules (TARs) that goes beyond simple rule extraction by considering the sequence of multiple visits. Our highly configurable algorithm leverages this formulation to extract Temporal Association Rules from sequences of medical instances. We can generate rules in the desired format, content, and temporal factors while accounting for different levels of abstraction of medical instances. To demonstrate the effectiveness of our methodology, we applied it to extract patient evolution patterns from clinical histories of multimorbid patients suffering from heart disease and stroke who visited Primary Care Centers (CAP) in Catalonia. Our main objective is to uncover complex rules with multiple temporal steps, that comprise a set of medical instances.ResultsAs we are working with real-world, error-prone data, we propose a process of validation of the results by expert practitioners in primary care. Despite our limited dataset, the high percentage of patterns deemed correct and relevant by the experts is promising. The insights gained from these patterns can inform preventive measures and help detect risk factors, ultimately leading to better treatments and outcomes for patients.ConclusionOur algorithm successfully extracted a set of meaningful and relevant temporal patterns, especially for the specific type of multimorbid patients considered. These patterns were evaluated by experts and demonstrated the ability to predict risk factors that are commonly associated with certain diseases. Moreover, the average time gap between the occurrence of medical events provided critical insight into the term of these risk factors. This information holds significant value in the context of primary healthcare and preventive medicine, highlighting the potential of our method to serve as a valuable medical tool.

Project description:BackgroundNational electronic health records (EHR) are increasingly used for research but identifying disease cases is challenging due to differences in information captured between sources (e.g. primary and secondary care). Our objective was to provide a transparent, reproducible model for integrating these data using atrial fibrillation (AF), a chronic condition diagnosed and managed in multiple ways in different healthcare settings, as a case study.MethodsPotentially relevant codes for AF screening, diagnosis, and management were identified in four coding systems: Read (primary care diagnoses and procedures), British National Formulary (BNF; primary care prescriptions), ICD-10 (secondary care diagnoses) and OPCS-4 (secondary care procedures). From these we developed a phenotype algorithm via expert review and analysis of linked EHR data from 1998 to 2010 for a cohort of 2.14 million UK patients aged ? 30 years. The cohort was also used to evaluate the phenotype by examining associations between incident AF and known risk factors.ResultsThe phenotype algorithm incorporated 286 codes: 201 Read, 63 BNF, 18 ICD-10, and four OPCS-4. Incident AF diagnoses were recorded for 72,793 patients, but only 39.6% (N = 28,795) were recorded in primary care and secondary care. An additional 7,468 potential cases were inferred from data on treatment and pre-existing conditions. The proportion of cases identified from each source differed by diagnosis age; inferred diagnoses contributed a greater proportion of younger cases (? 60 years), while older patients (? 80 years) were mainly diagnosed in SC. Associations of risk factors (hypertension, myocardial infarction, heart failure) with incident AF defined using different EHR sources were comparable in magnitude to those from traditional consented cohorts.ConclusionsA single EHR source is not sufficient to identify all patients, nor will it provide a representative sample. Combining multiple data sources and integrating information on treatment and comorbid conditions can substantially improve case identification.