Predictive models of long COVID.

BACKGROUND: The cause and symptoms of long COVID are poorly understood. It is challenging to predict whether a given COVID-19 patient will develop long COVID in the future. METHODS: We used electronic health record (EHR) data from the National COVID Cohort Collaborative to predict the incidence of long COVID. We trained two machine learning (ML) models - logistic regression (LR) and random forest (RF). Features used to train predictors included symptoms and drugs ordered during acute infection, measures of COVID-19 treatment, pre-COVID comorbidities, and demographic information. We assigned the 'long COVID' label to patients diagnosed with the U09.9 ICD10-CM code. The cohorts included patients with (a) EHRs reported from data partners using U09.9 ICD10-CM code and (b) at least one EHR in each feature category. We analysed three cohorts: all patients (n = 2,190,579; diagnosed with long COVID = 17,036), inpatients (149,319; 3,295), and outpatients (2,041,260; 13,741). FINDINGS: LR and RF models yielded median AUROC of 0.76 and 0.75, respectively. Ablation study revealed that drugs had the highest influence on the prediction task. The SHAP method identified age, gender, cough, fatigue, albuterol, obesity, diabetes, and chronic lung disease as explanatory features. Models trained on data from one N3C partner and tested on data from the other partners had average AUROC of 0.75. INTERPRETATION: ML-based classification using EHR information from the acute infection period is effective in predicting long COVID. SHAP methods identified important features for prediction. Cross-site analysis demonstrated the generalizability of the proposed methodology. FUNDING: NCATS U24 TR002306, NCATS UL1 TR003015, Axle Informatics Subcontract: NCATS-P00438-B, NIH/NIDDK/OD, PSR2015-1720GVALE_01, G43C22001320007, and Director, Office of Science, Office of Basic Energy Sciences of the U.S. Department of Energy Contract No. DE-AC02-05CH11231.

Clinical decision support to reduce unnecessary diagnostic testing for heparin-induced thrombocytopenia.

Appropriate evaluation of heparin-induced thrombocytopenia (HIT) is imperative because of the potentially life-threatening complications. However, overtesting and overdiagnosis of HIT are common. Our goal was to evaluate the impact of clinical decision support (CDS) based on the HIT computerized-risk (HIT-CR) score, designed to reduce unnecessary diagnostic testing. This retrospective observational study evaluated CDS that presented a platelet count versus time graph and 4Ts score calculator to clinicians who initiated a HIT immunoassay order in patients with predicted low risk (HIT-CR score 0-2). The primary outcome was the proportion of immunoassay orders initiated but cancelled after firing of the CDS advisory. Chart reviews were conducted to assess anticoagulation usage, 4Ts scores and the proportion of patients who had HIT. In a 20-week period, 319 CDS advisories were presented to users who initiated potentially unnecessary HIT diagnostic testing. The diagnostic test order was discontinued in 80 (25%) patients. Heparin products were continued in 139 (44%) patients, and alternative anticoagulation was not given to 264 (83%). The negative predictive value of the advisory was 98.8% (95% CI: 97.2-99.5). HIT-CR score-based CDS can reduce unnecessary diagnostic testing for HIT in patients with a low pretest probability of HIT.

Relationship between a risk score for QT interval prolongation and mortality across rural and urban inpatient facilities.

OBJECTIVES: To evaluate the relationship between a modified Tisdale QTc-risk score (QTc-RS) and inpatient mortality and length of stay in a broad inpatient population with an order for a medication with a known risk of torsades de pointes (TdP). BACKGROUND: Managing the risk of TdP is challenging due to the number of medications with known risk of TdP and the complexity of precipitating factors. A model to predict risk of mortality may be useful to guide treatment decisions. METHODS: This was a retrospective observational study using inpatient data from 28 healthcare facilities in the western United States. This risk score ranges from zero to 23 with weights applied to each risk factor based on a previous validation study. Logistic regression and a generalized linear model were performed to assess the relationship between QTc-RS and mortality and length of stay. RESULTS: Between April and December 2020, a QTc-RS was calculated for 92,383 hospitalized patients. Common risk factors were female (55.0%); age > 67 years (32.1%); and receiving a medication with known risk of TdP (24.5%). A total of 2770 (3%) patients died during their hospitalization. Relative to patients with QTc-RS < 7, the odds ratio for mortality was 4.80 (95%CI:4.42-5.21) for patients with QTc-RS = 7-10 and 11.51 (95%CI:10.23-12.94) for those with QTc-RS ≥ 11. Length of hospital stay increased by 0.7 day for every unit increase in the risk score (p < 0.0001). CONCLUSION: There is a strong relationship between increased mortality as well as longer duration of hospitalization with an increasing QTc-RS.

Generalisable long COVID subtypes: findings from the NIH N3C and RECOVER programmes.

BACKGROUND: Stratification of patients with post-acute sequelae of SARS-CoV-2 infection (PASC, or long COVID) would allow precision clinical management strategies. However, long COVID is incompletely understood and characterised by a wide range of manifestations that are difficult to analyse computationally. Additionally, the generalisability of machine learning classification of COVID-19 clinical outcomes has rarely been tested. METHODS: We present a method for computationally modelling PASC phenotype data based on electronic healthcare records (EHRs) and for assessing pairwise phenotypic similarity between patients using semantic similarity. Our approach defines a nonlinear similarity function that maps from a feature space of phenotypic abnormalities to a matrix of pairwise patient similarity that can be clustered using unsupervised machine learning. FINDINGS: We found six clusters of PASC patients, each with distinct profiles of phenotypic abnormalities, including clusters with distinct pulmonary, neuropsychiatric, and cardiovascular abnormalities, and a cluster associated with broad, severe manifestations and increased mortality. There was significant association of cluster membership with a range of pre-existing conditions and measures of severity during acute COVID-19. We assigned new patients from other healthcare centres to clusters by maximum semantic similarity to the original patients, and showed that the clusters were generalisable across different hospital systems. The increased mortality rate originally identified in one cluster was consistently observed in patients assigned to that cluster in other hospital systems. INTERPRETATION: Semantic phenotypic clustering provides a foundation for assigning patients to stratified subgroups for natural history or therapy studies on PASC. FUNDING: NIH (TR002306/OT2HL161847-01/OD011883/HG010860), U.S.D.O.E. (DE-AC02-05CH11231), Donald A. Roux Family Fund at Jackson Laboratory, Marsico Family at CU Anschutz.

Early Prediction of Cardiogenic Shock Using Machine Learning.

Cardiogenic shock (CS) is a severe condition with in-hospital mortality of up to 50%. Patients who develop CS may have previous cardiac history, but that may not always be the case, adding to the challenges in optimally identifying and managing these patients. Patients may present to a medical facility with CS or develop CS while in the emergency department (ED), in a general inpatient ward (WARD) or in the critical care unit (CC). While different clinical pathways for management exist once CS is recognized, there are challenges in identifying the patients in a timely manner, in all settings, in a timeframe that will allow proper management. We therefore developed and evaluated retrospectively a machine learning model based on the XGBoost (XGB) algorithm which runs automatically on patient data from the electronic health record (EHR). The algorithm was trained on 8 years of de-identified data (from 2010 to 2017) collected from a large regional healthcare system. The input variables include demographics, vital signs, laboratory values, some orders, and specific pre-existing diagnoses. The model was designed to make predictions 2 h prior to the need of first CS intervention (inotrope, vasopressor, or mechanical circulatory support). The algorithm achieves an overall area under curve (AUC) of 0.87 (0.81 in CC, 0.84 in ED, 0.97 in WARD), which is considered useful for clinical use. The algorithm can be refined based on specific elements defining patient subpopulations, for example presence of acute myocardial infarction (AMI) or congestive heart failure (CHF), further increasing its precision when a patient has these conditions. The top-contributing risk factors learned by the model are consistent with existing clinical findings. Our conclusion is that a useful machine learning model can be used to predict the development of CS. This manuscript describes the main steps of the development process and our results.

Clinician Responses to a Clinical Decision Support Advisory for High Risk of Torsades de Pointes.

Background Torsade de pointes (TdP) is a potentially fatal cardiac arrhythmia that is often drug induced. Clinical decision support (CDS) may help minimize TdP risk by guiding decision making in patients at risk. CDS has been shown to decrease prescribing of high-risk medications in patients at risk of TdP, but alerts are often ignored. Other risk-management options can potentially be incorporated in TdP risk CDS. Our goal was to evaluate actions clinicians take in response to a CDS advisory that uses a modified Tisdale QT risk score and presents management options that are easily selected (eg, single click). Methods and Results We implemented an inpatient TdP risk advisory systemwide across a large health care system comprising 30 hospitals. This CDS was programmed to appear when prescribers attempted ordering medications with a known risk of TdP in a patient with a QT risk score ≥12. The CDS displayed patient-specific information and offered relevant management options including canceling offending medications and ordering electrolyte replacement protocols or ECGs. We retrospectively studied the actions clinicians took within the advisory and separated by drug class. During an 8-month period, 7794 TdP risk advisories were issued. Antibiotics were the most frequent trigger of the advisory (n=2578, 33.1%). At least 1 action was taken within the advisory window for 2700 (34.6%) of the advisories. The most frequent action taken was ordering an ECG (n=1584, 20.3%). Incoming medication orders were canceled in 793 (10.2%) of the advisories. The frequency of each action taken varied by drug class (P<0.05 for all actions). Conclusions A modified Tisdale QT risk score-based CDS that offered relevant single-click management options yielded a high action/response rate. Actions taken by clinicians varied depending on the class of the medication that evoked the TdP risk advisory, but the most frequent was ordering an ECG.

Generalizable Long COVID Subtypes: Findings from the NIH N3C and RECOVER Programs.

Accurate stratification of patients with post-acute sequelae of SARS-CoV-2 infection (PASC, or long COVID) would allow precision clinical management strategies. However, the natural history of long COVID is incompletely understood and characterized by an extremely wide range of manifestations that are difficult to analyze computationally. In addition, the generalizability of machine learning classification of COVID-19 clinical outcomes has rarely been tested. We present a method for computationally modeling PASC phenotype data based on electronic healthcare records (EHRs) and for assessing pairwise phenotypic similarity between patients using semantic similarity. Our approach defines a nonlinear similarity function that maps from a feature space of phenotypic abnormalities to a matrix of pairwise patient similarity that can be clustered using unsupervised machine learning procedures. Using k-means clustering of this similarity matrix, we found six distinct clusters of PASC patients, each with distinct profiles of phenotypic abnormalities. There was a significant association of cluster membership with a range of pre-existing conditions and with measures of severity during acute COVID-19. Two of the clusters were associated with severe manifestations and displayed increased mortality. We assigned new patients from other healthcare centers to one of the six clusters on the basis of maximum semantic similarity to the original patients. We show that the identified clusters were generalizable across different hospital systems and that the increased mortality rate was consistently observed in two of the clusters. Semantic phenotypic clustering can provide a foundation for assigning patients to stratified subgroups for natural history or therapy studies on PASC.

Clinician Satisfaction With Advanced Clinical Decision Support to Reduce the Risk of Torsades de Pointes.

OBJECTIVES: Clinical decision support (CDS) can potentially help clinicians identify and manage patients who are at risk for torsades de pointes (TdP). However, computer alerts are often ignored and might contribute to alert fatigue. The goals of this project were to create an advanced TdP CDS advisory that presents patient-specific, relevant information, including 1-click management options, and to determine clinician satisfaction with the CDS. METHODS: The advanced TdP CDS was developed and implemented across a health system comprising 29 hospitals. The advisory presents patient-specific information including relevant risk factors, laboratory values, and 1-click options to help manage the condition in high-risk patients. A short electronic survey was created to gather clinician feedback on the advisory. RESULTS: After implementation, an email invitation to complete the anonymous advisory-related survey was sent to 442 clinicians who received the advisory. Among the 38 respondents, feedback was generally positive, with 79% of respondents reporting that the advisory helps them care for their patients and 87% responding that alternative actions for them to consider were clearly specified. However, 46% of respondents indicated the alert appeared too frequently. CONCLUSIONS: Advanced TdP risk CDS that provides relevant, patient-specific information and 1-click management options can be generally viewed favorably by clinicians who receive the advisory.

A Retrospective Cohort Study of Clinical Factors Associated with Transitions of Care among COVID-19 Patients.

Coronavirus Disease 2019 (COVID-19) is an international health crisis. In this article, we report on patient characteristics associated with care transitions of: 1) hospital admission from the emergency department (ED) and 2) escalation to the intensive care unit (ICU). Analysis of data from the electronic medical record (EMR) was performed for patients with COVID-19 seen in the ED of a large Western U.S. Health System from April to August of 2020, totaling 10,079 encounters. Of these, 5172 resulted in admission as an inpatient within 72 h. Inpatient encounters (n = 6079) were also considered for patients with positive COVID-19 test results, of which 970 resulted in a transfer to the ICU or in-hospital mortality. Laboratory results, vital signs, symptoms, and comorbidities were investigated for each of these care transitions. Different top risk factors were found, but two factors common to hospital admission and ICU transfer were respiratory rate and the need for oxygen support. Comorbidities common to both settings were cerebrovascular disease and congestive heart failure. Regarding laboratory results, the neutrophil-to-lymphocyte ratio was associated with transitions to higher levels of care, along with the ratio of aspartate aminotransferase (AST) to alanine aminotransferase (ALT).