Interpretable Machine Learning Prediction of Drug-Induced QT Prolongation: Electronic Health Record Analysis.

BACKGROUND: Drug-induced long-QT syndrome (diLQTS) is a major concern among patients who are hospitalized, for whom prediction models capable of identifying individualized risk could be useful to guide monitoring. We have previously demonstrated the feasibility of machine learning to predict the risk of diLQTS, in which deep learning models provided superior accuracy for risk prediction, although these models were limited by a lack of interpretability. OBJECTIVE: In this investigation, we sought to examine the potential trade-off between interpretability and predictive accuracy with the use of more complex models to identify patients at risk for diLQTS. We planned to compare a deep learning algorithm to predict diLQTS with a more interpretable algorithm based on cluster analysis that would allow medication- and subpopulation-specific evaluation of risk. METHODS: We examined the risk of diLQTS among 35,639 inpatients treated between 2003 and 2018 with at least 1 of 39 medications associated with risk of diLQTS and who had an electrocardiogram in the system performed within 24 hours of medication administration. Predictors included over 22,000 diagnoses and medications at the time of medication administration, with cases of diLQTS defined as a corrected QT interval over 500 milliseconds after treatment with a culprit medication. The interpretable model was developed using cluster analysis (K=4 clusters), and risk was assessed for specific medications and classes of medications. The deep learning model was created using all predictors within a 6-layer neural network, based on previously identified hyperparameters. RESULTS: Among the medications, we found that class III antiarrhythmic medications were associated with increased risk across all clusters, and that in patients who are noncritically ill without cardiovascular disease, propofol was associated with increased risk, whereas ondansetron was associated with decreased risk. Compared with deep learning, the interpretable approach was less accurate (area under the receiver operating characteristic curve: 0.65 vs 0.78), with comparable calibration. CONCLUSIONS: In summary, we found that an interpretable modeling approach was less accurate, but more clinically applicable, than deep learning for the prediction of diLQTS. Future investigations should consider this trade-off in the development of methods for clinical prediction.

A polygenic risk score for the QT interval is an independent predictor of drug-induced QT prolongation.

Drug-induced QT prolongation (diLQTS), and subsequent risk of torsade de pointes, is a major concern with use of many medications, including for non-cardiac conditions. The possibility that genetic risk, in the form of polygenic risk scores (PGS), could be integrated into prediction of risk of diLQTS has great potential, although it is unknown how genetic risk is related to clinical risk factors as might be applied in clinical decision-making. In this study, we examined the PGS for QT interval in 2500 subjects exposed to a known QT-prolonging drug on prolongation of the QT interval over 500ms on subsequent ECG using electronic health record data. We found that the normalized QT PGS was higher in cases than controls (0.212±0.954 vs. -0.0270±1.003, P = 0.0002), with an unadjusted odds ratio of 1.34 (95%CI 1.17-1.53, P<0.001) for association with diLQTS. When included with age and clinical predictors of QT prolongation, we found that the PGS for QT interval provided independent risk prediction for diLQTS, in which the interaction for high-risk diagnosis or with certain high-risk medications (amiodarone, sotalol, and dofetilide) was not significant, indicating that genetic risk did not modify the effect of other risk factors on risk of diLQTS. We found that a high-risk cutoff (QT PGS ≥ 2 standard deviations above mean), but not a low-risk cutoff, was associated with risk of diLQTS after adjustment for clinical factors, and provided one method of integration based on the decision-tree framework. In conclusion, we found that PGS for QT interval is an independent predictor of diLQTS, but that in contrast to existing theories about repolarization reserve as a mechanism of increasing risk, the effect is independent of other clinical risk factors. More work is needed for external validation in clinical decision-making, as well as defining the mechanism through which genes that increase QT interval are associated with risk of diLQTS.

A Systems-Based Morbidity and Mortality Conference Was Associated With a Transient Reduction in ECG Completion Times.

BACKGROUND AND OBJECTIVES: During its monthly morbidity and mortality conference (MMC), the University of Colorado Division of Cardiology reviewed a "near-miss" patient safety event involving the delayed completion of a Stat-priority (ie, statim, meaning high priority) electrocardiogram (ECG). Because critical and interprofessional stakeholders participated in the conference, we hypothesized that the MMC would be associated with reduced ECG completion times. METHODS: Data were collected for in-hospital ECGs performed at the University of Colorado Hospital between January 1, 2017, and June 30, 2018. An interrupted time series analysis was used to estimate the immediate and ongoing impact of the MMC (held on February 28, 2018) on ECG completion times, stratified by order priority (Stat, Now, or Routine). The percentage of delayed Stat-priority ECGs was analyzed as a secondary outcome. RESULTS: Before the MMC, ECG completion times were stable for all order priorities ( P > .2), but the proportion of delayed Stat-priority ECGs increased from 5% in January 2017 to 20% in February 2018 ( P < .01). The MMC was associated with an immediate reduction in average daily ECG completion times for Routine (-18.4 minutes, P = .03) and Now (-8 minutes, P = .024) priority ECGs. No reduction was seen for Stat ECGs ( P = .97), though the percentage of delayed Stat ECGs stopped increasing ( P = .63). In the post-MMC period, completion times for Routine-priority ECGs increased and approached pre-MMC levels. CONCLUSIONS: The MMC was associated with an immediate, but temporary, improvement in ECG completion times. Although the observed clinical benefit of the MMC is novel, these data support the need for more durable reforms to sustain initial improvements.

Prediction of Drug-Induced Long QT Syndrome Using Machine Learning Applied to Harmonized Electronic Health Record Data.

BACKGROUND: Drug-induced QT prolongation is a potentially preventable cause of morbidity and mortality, however there are no widespread clinical tools utilized to predict which individuals are at greatest risk. Machine learning (ML) algorithms may provide a method for identifying these individuals, and could be automated to directly alert providers in real time. OBJECTIVE: This study applies ML techniques to electronic health record (EHR) data to identify an integrated risk-prediction model that can be deployed to predict risk of drug-induced QT prolongation. METHODS: We examined harmonized data from the UCHealth EHR and identified inpatients who had received a medication known to prolong the QT interval. Using a binary outcome of the development of a QTc interval >500 ms within 24 hours of medication initiation or no ECG with a QTc interval >500 ms, we compared multiple machine learning methods by classification accuracy and performed calibration and rescaling of the final model. RESULTS: We identified 35,639 inpatients who received a known QT-prolonging medication and an ECG performed within 24 hours of administration. Of those, 4,558 patients developed a QTc > 500 ms and 31,081 patients did not. A deep neural network with random oversampling of controls was found to provide superior classification accuracy (F1 score 0.404; AUC 0.71) for the development of a long QT interval compared with other methods. The optimal cutpoint for prediction was determined and was reasonably accurate (sensitivity 71%; specificity 73%). CONCLUSIONS: We found that deep neural networks applied to EHR data provide reasonable prediction of which individuals are most susceptible to drug-induced QT prolongation. Future studies are needed to validate this model in novel EHRs and within the physician order entry system to assess the ability to improve patient safety.

Medication adherence in cardiovascular medicine.

Cardiovascular disease is the leading cause of death globally. While pharmacological advancements have improved the morbidity and mortality associated with cardiovascular disease, non-adherence to prescribed treatment remains a significant barrier to improved patient outcomes. A variety of strategies to improve medication adherence have been tested in clinical trials, and include the following categories: improving patient education, implementing medication reminders, testing cognitive behavioral interventions, reducing medication costs, utilizing healthcare team members, and streamlining medication dosing regimens. In this review, we describe specific trials within each of these categories and highlight the impact of each on medication adherence. We also examine ongoing trials and future lines of inquiry for improving medication adherence in patients with cardiovascular diseases.

MEG and EEG demonstrate similar test-retest reliability of the 40Hz auditory steady-state response.

The auditory steady-state response (ASSR) is increasingly being used as a biomarker in neuropsychiatric disorders, but research investigating the test-retest reliability of this measure is needed. We previously reported ASSR reliability, measured by electroencephalography (EEG), to 40Hz amplitude-modulated white noise and click train stimuli. The purpose of the current study was to (a) assess the reliability of the MEG-measured ASSR to 40Hz amplitude-modulated white noise and click train stimuli, and (b) compare test-retest reliability between MEG and EEG measures of ASSR, which has not previously been investigated. Additionally, impact of stimulus parameter choice on reliability was assessed, by comparing responses to white noise and click train stimuli. Test-retest reliability, across sessions approximately one week apart, was assessed in 17 healthy adults. On each study day, participants completed two passive listening tasks (white noise and click train stimuli) during separate MEG and EEG recordings. Between-session correlations for evoked power and inter-trial phase coherence (ITPC) were assessed following source-space projection. Overall, the MEG-measured ASSR was significantly correlated between sessions (p<0.05, FDR corrected), suggesting acceptable test-retest reliability. Results suggest greater response reproducibility for ITPC compared to evoked responses and for click train compared to white noise stimuli, although further study is warranted. No significant differences in reliability were observed between MEG and EEG measures, suggesting they are similarly reliable. This work supports use of the ASSR as a biomarker in clinical interventions with repeated measures.

Test-retest reliability of the 40 Hz EEG auditory steady-state response.

Auditory evoked steady-state responses are increasingly being used as a marker of brain function and dysfunction in various neuropsychiatric disorders, but research investigating the test-retest reliability of this response is lacking. The purpose of this study was to assess the consistency of the auditory steady-state response (ASSR) across sessions. Furthermore, the current study aimed to investigate how the reliability of the ASSR is impacted by stimulus parameters and analysis method employed. The consistency of this response across two sessions spaced approximately 1 week apart was measured in nineteen healthy adults using electroencephalography (EEG). The ASSR was entrained by both 40 Hz amplitude-modulated white noise and click train stimuli. Correlations between sessions were assessed with two separate analytical techniques: a) channel-level analysis across the whole-head array and b) signal-space projection from auditory dipoles. Overall, the ASSR was significantly correlated between sessions 1 and 2 (p<0.05, multiple comparison corrected), suggesting adequate test-retest reliability of this response. The current study also suggests that measures of inter-trial phase coherence may be more reliable between sessions than measures of evoked power. Results were similar between the two analysis methods, but reliability varied depending on the presented stimulus, with click train stimuli producing more consistent responses than white noise stimuli.