Search

Abstract LLMs can accomplish specialized medical knowledge tasks, however, equitable access is hindered by the extensive fine-tuning, specialized medical data requirement, and limited access to proprietary models. Open-source (OS) medical LLMs show performance improvements and provide the transparency and compliance required in healthcare. We present OpenMedLM, a prompting platform delivering state-of-the-art (SOTA) performance for OS LLMs on medical benchmarks. We evaluated OS foundation LLMs (7B-70B) on medical benchmarks (MedQA, MedMCQA, PubMedQA, MMLU medical-subset) and selected Yi34B for developing OpenMedLM. Prompting strategies included zero-shot, few-shot, chain-of-thought, and ensemble/self-consistency voting. OpenMedLM delivered OS SOTA results on three medical LLM benchmarks, surpassing previous best-performing OS models that leveraged costly and extensive fine-tuning. OpenMedLM displays the first results to date demonstrating the ability of OS foundation models to optimize performance, absent specialized fine-tuning. The model achieved 72.6% accuracy on MedQA, outperforming the previous SOTA by 2.4%, and 81.7% accuracy on MMLU medical-subset, establishing itself as the first OS LLM to surpass 80% accuracy on this benchmark. Our results highlight medical-specific emergent properties in OS LLMs not documented elsewhere to date and validate the ability of OS models to accomplish healthcare tasks, highlighting the benefits of prompt engineering to improve performance of accessible LLMs for medical applications.

Type 2 diabetes (T2D) is a global health concern with increasing prevalence. Comorbid hypothyroidism (HT) exacerbates kidney, cardiac, neurological and other complications of T2D; these risks can be mitigated pharmacologically upon detecting HT. The current HT standard of care (SOC) screening in T2D is infrequent, delaying HT diagnosis and treatment. We present a first-to-date machine learning algorithm (MLA) clinical decision tool to classify patients as low vs. high risk for developing HT comorbid with T2D; the MLA was developed using readily available patient data from harmonized multinational datasets. The MLA was trained on data from NIH All of US (AoU) and UK Biobank (UKBB) (Combined dataset) and achieved a high negative predictive value (NPV) of 0.989 and an AUROC of 0.762 in the Combined dataset, exceeding AUROCs for the models trained on AoU or UKBB alone (0.666 and 0.622, respectively), indicating that increasing dataset diversity for MLA training improves performance. This high-NPV automated tool can supplement SOC screening and rule out T2D patients with low HT risk, allowing for the prioritization of lab-based testing for at-risk patients. Conversely, an MLA output that designates a patient to be at risk of developing HT allows for tailored clinical management and thereby promotes improved patient outcomes.

Background/Objective: Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by lifelong impacts on functional social and daily living skills, and restricted, repetitive behaviors (RRBs). Applied behavior analysis (ABA), the gold-standard treatment for ASD, has been extensively validated. ABA access is hindered by limited availability of qualified professionals and logistical and financial barriers. Scientifically validated, parent-led ABA can fill the accessibility gap by overcoming treatment barriers. This retrospective cohort study examines how our ABA treatment model, utilizing parent behavior technicians (pBTs) to deliver ABA, impacts adaptive behaviors and interfering behaviors (IBs) in a cohort of children on the autism spectrum with varying ASD severity levels, and with or without clinically significant IBs. Methods: Clinical outcomes of 36 patients ages 3–15 years were assessed using longitudinal changes in Vineland-3 after 3+ months of pBT-delivered ABA treatment. Results: Within the pBT model, our patients demonstrated clinically significant improvements in Vineland-3 Composite, domain, and subdomain scores, and utilization was higher in severe ASD. pBTs utilized more prescribed ABA when children initiated treatment with clinically significant IBs, and these children also showed greater gains in their Composite scores. Study limitations include sample size, inter-rater reliability, potential assessment metric bias and schedule variability, and confounding intrinsic or extrinsic factors. Conclusion: Overall, our pBT model facilitated high treatment utilization and showed robust effectiveness, achieving improved adaptive behaviors and reduced IBs when compared to conventional ABA delivery. The pBT model is a strong contender to fill the widening treatment accessibility gap and represents a powerful tool for addressing systemic problems in ABA treatment delivery.

Abstract Background Applied behavioral analysis (ABA) is regarded as the gold standard treatment for autism spectrum disorder (ASD) and has the potential to improve outcomes for patients with ASD. It can be delivered at different intensities, which are classified as comprehensive or focused treatment approaches. Comprehensive ABA targets multiple developmental domains and involves 20–40 h/week of treatment. Focused ABA targets individual behaviors and typically involves 10–20 h/week of treatment. Determining the appropriate treatment intensity involves patient assessment by trained therapists, however, the final determination is highly subjective and lacks a standardized approach. In our study, we examined the ability of a machine learning (ML) prediction model to classify which treatment intensity would be most suited individually for patients with ASD who are undergoing ABA treatment. Methods Retrospective data from 359 patients diagnosed with ASD were analyzed and included in the training and testing of an ML model for predicting comprehensive or focused treatment for individuals undergoing ABA treatment. Data inputs included demographics, schooling, behavior, skills, and patient goals. A gradient-boosted tree ensemble method, XGBoost, was used to develop the prediction model, which was then compared against a standard of care comparator encompassing features specified by the Behavior Analyst Certification Board treatment guidelines. Prediction model performance was assessed via area under the receiver-operating characteristic curve (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Results The prediction model achieved excellent performance for classifying patients in the comprehensive versus focused treatment groups (AUROC: 0.895; 95% CI 0.811–0.962) and outperformed the standard of care comparator (AUROC 0.767; 95% CI 0.629–0.891). The prediction model also achieved sensitivity of 0.789, specificity of 0.808, PPV of 0.6, and NPV of 0.913. Out of 71 patients whose data were employed to test the prediction model, only 14 misclassifications occurred. A majority of misclassifications (n = 10) indicated comprehensive ABA treatment for patients that had focused ABA treatment as the ground truth, therefore still providing a therapeutic benefit. The three most important features contributing to the model’s predictions were bathing ability, age, and hours per week of past ABA treatment. Conclusion This research demonstrates that the ML prediction model performs well to classify appropriate ABA treatment plan intensity using readily available patient data. This may aid with standardizing the process for determining appropriate ABA treatments, which can facilitate initiation of the most appropriate treatment intensity for patients with ASD and improve resource allocation.

Mild cognitive impairment (MCI) is cognitive decline that can indicate future risk of Alzheimer’s disease (AD). We developed and validated a machine learning algorithm (MLA), based on a gradient-boosted tree ensemble method, to analyze phenotypic data for individuals 55–88 years old (n = 493) diagnosed with MCI. Data were analyzed within multiple prediction windows and averaged to predict progression to AD within 24–48 months. The MLA outperformed the mini-mental state examination (MMSE) and three comparison models at all prediction windows on most metrics. Exceptions include sensitivity at 18 months (MLA and MMSE each achieved 0.600); and sensitivity at 30 and 42 months (MMSE marginally better). For all prediction windows, the MLA achieved AUROC ≥ 0.857 and NPV ≥ 0.800. With averaged data for the 24–48-month lookahead timeframe, the MLA outperformed MMSE on all metrics. This study demonstrates that machine learning may provide a more accurate risk assessment than the standard of care. This may facilitate care coordination, decrease healthcare expenditures, and maintain quality of life for patients at risk of progressing from MCI to AD.

Abstract Background Non‐alcoholic fatty liver (NAFL) can progress to the severe subtype non‐alcoholic steatohepatitis (NASH) and/or fibrosis, which are associated with increased morbidity, mortality, and healthcare costs. Current machine learning studies detect NASH; however, this study is unique in predicting the progression of NAFL patients to NASH or fibrosis. Aim To utilize clinical information from NAFL‐diagnosed patients to predict the likelihood of progression to NASH or fibrosis. Methods Data were collected from electronic health records of patients receiving a first‐time NAFL diagnosis. A gradient boosted machine learning algorithm (XGBoost) as well as logistic regression (LR) and multi‐layer perceptron (MLP) models were developed. A five‐fold cross‐validation grid search was utilized for hyperparameter optimization of variables, including maximum tree depth, learning rate, and number of estimators. Predictions of patients likely to progress to NASH or fibrosis within 4 years of initial NAFL diagnosis were made using demographic features, vital signs, and laboratory measurements. Results The XGBoost algorithm achieved area under the receiver operating characteristic (AUROC) values of 0.79 for prediction of progression to NASH and 0.87 for fibrosis on both hold‐out and external validation test sets. The XGBoost algorithm outperformed the LR and MLP models for both NASH and fibrosis prediction on all metrics. Conclusion It is possible to accurately identify newly diagnosed NAFL patients at high risk of progression to NASH or fibrosis. Early identification of these patients may allow for increased clinical monitoring, more aggressive preventative measures to slow the progression of NAFL and fibrosis, and efficient clinical trial enrollment.

Abstract Diagnosis and appropriate intervention for myocardial infarction (MI) are time‐sensitive but rely on clinical measures that can be progressive and initially inconclusive, underscoring the need for an accurate and early predictor of MI to support diagnostic and clinical management decisions. The objective of this study was to develop a machine learning algorithm (MLA) to predict MI diagnosis based on electronic health record data (EHR) readily available during Emergency Department assessment. An MLA was developed using retrospective patient data. The MLA used patient data as they became available in the first 3 h of care to predict MI diagnosis (defined by International Classification of Diseases, 10th revision code) at any time during the encounter. The MLA obtained an area under the receiver operating characteristic curve of 0.87, sensitivity of 87% and specificity of 70%, outperforming the comparator scoring systems TIMI and GRACE on all metrics. An MLA can synthesize complex EHR data to serve as a clinically relevant risk stratification tool for MI.

BackgroundShort-term fall prediction models that use electronic health records (EHRs) may enable the implementation of dynamic care practices that specifically address changes in individualized fall risk within senior care facilities. ObjectiveThe aim of this study is to implement machine learning (ML) algorithms that use EHR data to predict a 3-month fall risk in residents from a variety of senior care facilities providing different levels of care. MethodsThis retrospective study obtained EHR data (2007-2021) from Juniper Communities’ proprietary database of 2785 individuals primarily residing in skilled nursing facilities, independent living facilities, and assisted living facilities across the United States. We assessed the performance of 3 ML-based fall prediction models and the Juniper Communities’ fall risk assessment. Additional analyses were conducted to examine how changes in the input features, training data sets, and prediction windows affected the performance of these models. ResultsThe Extreme Gradient Boosting model exhibited the highest performance, with an area under the receiver operating characteristic curve of 0.846 (95% CI 0.794-0.894), specificity of 0.848, diagnostic odds ratio of 13.40, and sensitivity of 0.706, while achieving the best trade-off in balancing true positive and negative rates. The number of active medications was the most significant feature associated with fall risk, followed by a resident’s number of active diseases and several variables associated with vital signs, including diastolic blood pressure and changes in weight and respiratory rates. The combination of vital signs with traditional risk factors as input features achieved higher prediction accuracy than using either group of features alone. ConclusionsThis study shows that the Extreme Gradient Boosting technique can use a large number of features from EHR data to make short-term fall predictions with a better performance than that of conventional fall risk assessments and other ML models. The integration of routinely collected EHR data, particularly vital signs, into fall prediction models may generate more accurate fall risk surveillance than models without vital signs. Our data support the use of ML models for dynamic, cost-effective, and automated fall predictions in different types of senior care facilities.

Abstract Background Acute heart failure (AHF) is associated with significant morbidity and mortality. Effective patient risk stratification is essential to guiding hospitalization decisions and the clinical management of AHF. Clinical decision support systems can be used to improve predictions of mortality made in emergency care settings for the purpose of AHF risk stratification. In this study, several models for the prediction of seven-day mortality among AHF patients were developed by applying machine learning techniques to retrospective patient data from 236,275 total emergency department (ED) encounters, 1881 of which were considered positive for AHF and were used for model training and testing. The models used varying subsets of age, sex, vital signs, and laboratory values. Model performance was compared to the Emergency Heart Failure Mortality Risk Grade (EHMRG) model, a commonly used system for prediction of seven-day mortality in the ED with similar (or, in some cases, more extensive) inputs. Model performance was assessed in terms of area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity. Results When trained and tested on a large academic dataset, the best-performing model and EHMRG demonstrated test set AUROCs of 0.84 and 0.78, respectively, for prediction of seven-day mortality. Given only measurements of respiratory rate, temperature, mean arterial pressure, and FiO2, one model produced a test set AUROC of 0.83. Neither a logistic regression comparator nor a simple decision tree outperformed EHMRG. Conclusions A model using only the measurements of four clinical variables outperforms EHMRG in the prediction of seven-day mortality in AHF. With these inputs, the model could not be replaced by logistic regression or reduced to a simple decision tree without significant performance loss. In ED settings, this minimal-input risk stratification tool may assist clinicians in making critical decisions about patient disposition by providing early and accurate insights into individual patient’s risk profiles.

Abstract Background Severe sepsis and septic shock are among the leading causes of death in the United States and sepsis remains one of the most expensive conditions to diagnose and treat. Accurate early diagnosis and treatment can reduce the risk of adverse patient outcomes, but the efficacy of traditional rule-based screening methods is limited. The purpose of this study was to develop and validate a machine learning algorithm (MLA) for severe sepsis prediction up to 48 h before onset using a diverse patient dataset. Methods Retrospective analysis was performed on datasets composed of de-identified electronic health records collected between 2001 and 2017, including 510,497 inpatient and emergency encounters from 461 health centers collected between 2001 and 2015, and 20,647 inpatient and emergency encounters collected in 2017 from a community hospital. MLA performance was compared to commonly used disease severity scoring systems and was evaluated at 0, 4, 6, 12, 24, and 48 h prior to severe sepsis onset. Results 270,438 patients were included in analysis. At time of onset, the MLA demonstrated an AUROC of 0.931 (95% CI 0.914, 0.948) and a diagnostic odds ratio (DOR) of 53.105 on a testing dataset, exceeding MEWS (0.725, P

Abstract Background Pulmonary embolisms (PE) are life‐threatening medical events, and early identification of patients experiencing a PE is essential to optimizing patient outcomes. Current tools for risk stratification of PE patients are limited and unable to predict PE events before their occurrence. Objective We developed a machine learning algorithm (MLA) designed to identify patients at risk of PE before the clinical detection of onset in an inpatient population. Materials and Methods Three machine learning (ML) models were developed on electronic health record data from 63,798 medical and surgical inpatients in a large US medical center. These models included logistic regression, neural network, and gradient boosted tree (XGBoost) models. All models used only routinely collected demographic, clinical, and laboratory information as inputs. All were evaluated for their ability to predict PE at the first time patient vital signs and lab measures required for the MLA to run were available. Performance was assessed with regard to the area under the receiver operating characteristic (AUROC), sensitivity, and specificity. Results The model trained using XGBoost demonstrated the strongest performance for predicting PEs. The XGBoost model obtained an AUROC of 0.85, a sensitivity of 81%, and a specificity of 70%. The neural network and logistic regression models obtained AUROCs of 0.74 and 0.67, sensitivity of 81% and 81%, and specificity of 44% and 35%, respectively. Conclusions This algorithm may improve patient outcomes through earlier recognition and prediction of PE, enabling earlier diagnosis and treatment of PE.

BackgroundStrokes represent a leading cause of mortality globally. The evolution of developing new therapies is subject to safety and efficacy testing in clinical trials, which operate in a limited timeframe. To maximize the impact of these trials, patient cohorts for whom ischemic stroke is likely during that designated timeframe should be identified. Machine learning may improve upon existing candidate identification methods in order to maximize the impact of clinical trials for stroke prevention and treatment and improve patient safety.MethodsA retrospective study was performed using 41,970 qualifying patient encounters with ischemic stroke from inpatient visits recorded from over 700 inpatient and ambulatory care sites. Patient data were extracted from electronic health records and used to train and test a gradient boosted machine learning algorithm (MLA) to predict the patients' risk of experiencing ischemic stroke from the period of 1 day up to 1 year following the patient encounter. The primary outcome of interest was the occurrence of ischemic stroke.ResultsAfter training for optimization, XGBoost obtained a specificity of 0.793, a positive predictive value (PPV) of 0.194, and a negative predictive value (NPV) of 0.985. The MLA further obtained an area under the receiver operating characteristic (AUROC) of 0.88. The Logistic Regression and multilayer perceptron models both achieved AUROCs of 0.862. Among features that significantly impacted the prediction of ischemic stroke were previous stroke history, age, and mean systolic blood pressure.ConclusionMLAs have the potential to more accurately predict the near risk of ischemic stroke within a 1-year prediction window for individuals who have been hospitalized. This risk stratification tool can be used to design clinical trials to test stroke prevention treatments in high-risk populations by identifying subjects who would be more likely to benefit from treatment.

BackgroundA high number of patients who are hospitalized with COVID-19 develop acute respiratory distress syndrome (ARDS). ObjectiveIn response to the need for clinical decision support tools to help manage the next pandemic during the early stages (ie, when limited labeled data are present), we developed machine learning algorithms that use semisupervised learning (SSL) techniques to predict ARDS development in general and COVID-19 populations based on limited labeled data. MethodsSSL techniques were applied to 29,127 encounters with patients who were admitted to 7 US hospitals from May 1, 2019, to May 1, 2021. A recurrent neural network that used a time series of electronic health record data was applied to data that were collected when a patient’s peripheral oxygen saturation level fell below the normal range (

Deep venous thrombosis (DVT) is associated with significant morbidity, mortality, and increased healthcare costs. Standard scoring systems for DVT risk stratification often provide insufficient stratification of hospitalized patients and are unable to accurately predict which inpatients are most likely to present with DVT. There is a continued need for tools which can predict DVT in hospitalized patients. We performed a retrospective study on a database collected from a large academic hospital, comprised of 99,237 total general ward or ICU patients, 2,378 of whom experienced a DVT during their hospital stay. Gradient boosted machine learning algorithms were developed to predict a patient’s risk of developing DVT at 12- and 24-hour windows prior to onset. The primary outcome of interest was diagnosis of in-hospital DVT. The machine learning predictors obtained AUROCs of 0.83 and 0.85 for DVT risk prediction on hospitalized patients at 12- and 24-hour windows, respectively. At both 12 and 24 hours before DVT onset, the most important features for prediction of DVT were cancer history, VTE history, and internal normalized ratio (INR). Improved risk stratification may prevent unnecessary invasive testing in patients for whom DVT cannot be ruled out using existing methods. Improved risk stratification may also allow for more targeted use of prophylactic anticoagulants, as well as earlier diagnosis and treatment, preventing the development of pulmonary emboli and other sequelae of DVT.

BackgroundThe COVID-19 pandemic remains a significant global threat. However, despite urgent need, there remains uncertainty surrounding best practices for pharmaceutical interventions to treat COVID-19. In particular, conflicting evidence has emerged surrounding the use of hydroxychloroquine and azithromycin, alone or in combination, for COVID-19. The COVID-19 Evidence Accelerator convened by the Reagan-Udall Foundation for the FDA, in collaboration with Friends of Cancer Research, assembled experts from the health systems research, regulatory science, data science, and epidemiology to participate in a large parallel analysis of different data sets to further explore the effectiveness of these treatments.MethodsElectronic health record (EHR) and claims data were extracted from seven separate databases. Parallel analyses were undertaken on data extracted from each source. Each analysis examined time to mortality in hospitalized patients treated with hydroxychloroquine, azithromycin, and the two in combination as compared to patients not treated with either drug. Cox proportional hazards models were used, and propensity score methods were undertaken to adjust for confounding. Frequencies of adverse events in each treatment group were also examined.ResultsNeither hydroxychloroquine nor azithromycin, alone or in combination, were significantly associated with time to mortality among hospitalized COVID-19 patients. No treatment groups appeared to have an elevated risk of adverse events.ConclusionAdministration of hydroxychloroquine, azithromycin, and their combination appeared to have no effect on time to mortality in hospitalized COVID-19 patients. Continued research is needed to clarify best practices surrounding treatment of COVID-19.

Background Severe sepsis and septic shock are among the leading causes of death in the USA. While early prediction of severe sepsis can reduce adverse patient outcomes, sepsis remains one of the most expensive conditions to diagnose and treat.Objective The purpose of this study was to evaluate the effect of a machine learning algorithm for severe sepsis prediction on in-hospital mortality, hospital length of stay and 30-day readmission.Design Prospective clinical outcomes evaluation.Setting Evaluation was performed on a multiyear, multicentre clinical data set of real-world data containing 75 147 patient encounters from nine hospitals across the continental USA, ranging from community hospitals to large academic medical centres.Participants Analyses were performed for 17 758 adult patients who met two or more systemic inflammatory response syndrome criteria at any point during their stay (‘sepsis-related’ patients).Interventions Machine learning algorithm for severe sepsis prediction.Outcome measures In-hospital mortality, length of stay and 30-day readmission rates.Results Hospitals saw an average 39.5% reduction of in-hospital mortality, a 32.3% reduction in hospital length of stay and a 22.7% reduction in 30-day readmission rate for sepsis-related patient stays when using the machine learning algorithm in clinical outcomes analysis.Conclusions Reductions of in-hospital mortality, hospital length of stay and 30-day readmissions were observed in real-world clinical use of the machine learning-based algorithm. The predictive algorithm may be successfully used to improve sepsis-related outcomes in live clinical settings.Trial registration number NCT03960203

Background: Early detection of pediatric severe sepsis is necessary in order to optimize effective treatment, and new methods are needed to facilitate this early detection.Objective: Can a machine-learning based prediction algorithm using electronic healthcare record (EHR) data predict severe sepsis onset in pediatric populations?Methods: EHR data were collected from a retrospective set of de-identified pediatric inpatient and emergency encounters for patients between 2–17 years of age, drawn from the University of California San Francisco (UCSF) Medical Center, with encounter dates between June 2011 and March 2016.Results: Pediatric patients (n = 9,486) were identified and 101 (1.06%) were labeled with severe sepsis following the pediatric severe sepsis definition of Goldstein et al. (1). In 4-fold cross-validation evaluations, the machine learning algorithm achieved an AUROC of 0.916 for discrimination between severe sepsis and control pediatric patients at the time of onset and AUROC of 0.718 at 4 h before onset. The prediction algorithm significantly outperformed the Pediatric Logistic Organ Dysfunction score (PELOD-2) (p < 0.05) and pediatric Systemic Inflammatory Response Syndrome (SIRS) (p < 0.05) in the prediction of severe sepsis 4 h before onset using cross-validation and pairwise t-tests.Conclusion: This machine learning algorithm has the potential to deliver high-performance severe sepsis detection and prediction through automated monitoring of EHR data for pediatric inpatients, which may enable earlier sepsis recognition and treatment initiation.

Background: Machine learning methods have been developed to predict the likelihood of a given event or classify patients into two or more diagnostic categories. Digital twin models, which forecast entire trajectories of patient health data, have potential applications in clinical trials and patient management. Methods: In this study, we apply a digital twin model based on a variational autoencoder to a population of patients who went on to experience an ischemic stroke. The digital twin’s ability to model patient clinical features was assessed with regard to its ability to forecast clinical measurement trajectories leading up to the onset of the acute medical event and beyond using International Classification of Diseases (ICD) codes for ischemic stroke and lab values as inputs. Results: The simulated patient trajectories were virtually indistinguishable from real patient data, with similar feature means, standard deviations, inter-feature correlations, and covariance structures on a withheld test set. A logistic regression adversary model was unable to distinguish between the real and simulated data area under the receiver operating characteristic (ROC) curve (AUCadversary = 0.51). Conclusion: Through accurate projection of patient trajectories, this model may help inform clinical decision making or provide virtual control arms for efficient clinical trials.

Therapeutic agents for the novel coronavirus disease 2019 (COVID-19) have been proposed, but evidence supporting their use is limited. A machine learning algorithm was developed in order to identify a subpopulation of COVID-19 patients for whom hydroxychloroquine was associated with improved survival; this population might be relevant for study in a clinical trial. A pragmatic trial was conducted at six United States hospitals. We enrolled COVID-19 patients that were admitted between 10 March and 4 June 2020. Treatment was not randomized. The study endpoint was mortality; discharge was a competing event. Hazard ratios were obtained on the entire population, and on the subpopulation indicated by the algorithm as suitable for treatment. A total of 290 patients were enrolled. In the subpopulation that was identified by the algorithm, hydroxychloroquine was associated with a statistically significant (p = 0.011) increase in survival (adjusted hazard ratio 0.29, 95% confidence interval (CI) 0.11–0.75). Adjusted survival among the algorithm indicated patients was 82.6% in the treated arm and 51.2% in the arm not treated. No association between treatment and mortality was observed in the general population. A 31% increase in survival at the end of the study was observed in a population of COVID-19 patients that were identified by a machine learning algorithm as having a better outcome with hydroxychloroquine treatment. Precision medicine approaches may be useful in identifying a subpopulation of COVID-19 patients more likely to be proven to benefit from hydroxychloroquine treatment in a clinical trial.

Background: A major problem in treating acute kidney injury (AKI) is that clinical criteria for recognition are markers of established kidney damage or impaired function; treatment before such damage manifests is desirable. Clinicians could intervene during what may be a crucial stage for preventing permanent kidney injury if patients with incipient AKI and those at high risk of developing AKI could be identified. Objective: In this study, we evaluate a machine learning algorithm for early detection and prediction of AKI. Design: We used a machine learning technique, boosted ensembles of decision trees, to train an AKI prediction tool on retrospective data taken from more than 300 000 inpatient encounters. Setting: Data were collected from inpatient wards at Stanford Medical Center and intensive care unit patients at Beth Israel Deaconess Medical Center. Patients: Patients older than the age of 18 whose hospital stays lasted between 5 and 1000 hours and who had at least one documented measurement of heart rate, respiratory rate, temperature, serum creatinine (SCr), and Glasgow Coma Scale (GCS). Measurements: We tested the algorithm’s ability to detect AKI at onset and to predict AKI 12, 24, 48, and 72 hours before onset. Methods: We tested AKI detection and prediction using the National Health Service (NHS) England AKI Algorithm as a gold standard. We additionally tested the algorithm’s ability to detect AKI as defined by the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines. We compared the algorithm’s 3-fold cross-validation performance to the Sequential Organ Failure Assessment (SOFA) score for AKI identification in terms of area under the receiver operating characteristic (AUROC). Results: The algorithm demonstrated high AUROC for detecting and predicting NHS-defined AKI at all tested time points. The algorithm achieves AUROC of 0.872 (95% confidence interval [CI], 0.867-0.878) for AKI detection at time of onset. For prediction 12 hours before onset, the algorithm achieves an AUROC of 0.800 (95% CI, 0.792-0.809). For 24-hour predictions, the algorithm achieves AUROC of 0.795 (95% CI, 0.785-0.804). For 48-hour and 72-hour predictions, the algorithm achieves AUROC values of 0.761 (95% CI, 0.753-0.768) and 0.728 (95% CI, 0.719-0.737), respectively. Limitations: Because of the retrospective nature of this study, we cannot draw any conclusions about the impact the algorithm’s predictions will have on patient outcomes in a clinical setting. Conclusions: The results of these experiments suggest that a machine learning–based AKI prediction tool may offer important prognostic capabilities for determining which patients are likely to suffer AKI, potentially allowing clinicians to intervene before kidney damage manifests.

The theory and computational tools developed to interpret and explore energy landscapes in molecular science are applied to the landscapes defined by local minima for neural networks. These machine learning landscapes correspond to fits of training data, where the inputs are vital signs and laboratory measurements for a database of patients, and the objective is to predict a clinical outcome. In this contribution, we test the predictions obtained by fitting to single measurements, and then to combinations of between 2 and 10 different patient medical data items. The effect of including measurements over different time intervals from the 48 h period in question is analysed, and the most recent values are found to be the most important. We also compare results obtained for neural networks as a function of the number of hidden nodes, and for different values of a regularization parameter. The predictions are compared with an alternative convex fitting function, and a strong correlation is observed. The dependence of these results on the patients randomly selected for training and testing decreases systematically with the size of the database available. The machine learning landscapes defined by neural network fits in this investigation have single-funnel character, which probably explains why it is relatively straightforward to obtain the global minimum solution, or a fit that behaves similarly to this optimal parameterization.

Sepsis, a dysregulated host response to infection, is a major health burden in terms of both mortality and cost. The difficulties clinicians face in diagnosing sepsis, alongside the insufficiencies of diagnostic biomarkers, motivate the present study. This work develops a machine-learning-based sepsis diagnostic for a high-risk patient group, using a geographically and institutionally diverse collection of nearly 500,000 patient health records. Using only a minimal set of clinical variables, our diagnostics outperform common severity scoring systems and sepsis biomarkers and benefit from being available immediately upon ordering.