Frequently Asked Questions
What is ∆QT Database?
∆QT Database is a publicly available resource for exploring the effects of one or more drugs on the QT interval. Drug-induced prolongation of the QT interval can increase the risk of a potentially fatal arrhythmia. The effects of individual drugs on the QT interval are well-studied, and in recent work our group has computationally predicted and experimentally validated the QT-prolonging effects of multiple drugs taken at the same time (drug-drug interactions). See CredibleMeds.org for an excellent introduction to long QT syndrome.
This website provides a front-facing interface for users to investigate one or more drugs and drug classes. You can save the resulting plot, the subset of data currently viewed, or the entire database. All source code for the site is available on GitHub.
Why is my drug not in the database?
∆QT Database only includes 259 commonly prescribed drugs at NewYork-Presbyterian Hospital to ensure patient privacy. If the drug(s) you are interested in studying are not included, email Nick Tatonetti to discuss a potential collaboration where we can provide evidence using our entire electronic health record; in these cases we will not be able to share the underlying data.
Where did the data come from?
The ∆QT Database is a deidentified subset of electronic health record data originally collected at NewYork-Presbyterian Hospital/ Columbia University Medical Center and mapped to the OHDSI Common Data Model. To prepare the data for public release we performed the following deidentification procedure (available as a Python script on GitHub):
- Removed 18 HIPAA Safe Harbor identifiers.
- Only included patients taking one or more commonly prescribed drugs (>3000 patients prescribed drug in EHR, N=259) and with at least 2 electrocardiograms (ECGs).
- Excluded patients younger than 18 or older than 89 at time of ECG; randomly adjusted age ±0-5 years.
- Defined global baseline for each patient: median QTc (heart rate-corrected QT) interval across all of that patient's ECGs.
- Defined one or more "ECG era"s for each patient: one or more ECGs combined into an era, such that each subsequent ECG occurs no more than 36 days after the previous one (otherwise create new ECG era). For each ECG era, defined maxECG as the ECG date with the most prolonged QTc interval.
- Collected all drugs a patient was taking 0-36 days inclusive before the maxECG date.
- Randomly swapped small subset of drug exposures from one patient to another to further deidentify the data. 6.5% of patients had at least one swapped drug. To perform swap:
Each entry in the database then contains:
- For each drug, calculated median ∆QTc (maxECG QTc – baseline QTc) across all ECG eras for all patients.
- Binned distribution of median ∆QTcs into 10 bins (0–80 ms).
- Assigned decreasing swap frequency (max: 1% for drugs with small effects on QTc interval; min: 1/1000% for drugs with large effect on QTc interval).
- Randomly swapped (bin_swap_frequency × num_drug_exposures) from one patient ECG era containing the drug to another patient ECG era not containing the drug.
- pt_id_era: concatenated pt_id and era
- pt_id: database-specific arbitrary patient ID
- era: "ECG era" for linking all drug exposures up to 36 days before post-exposure ECG
- age, sex: patient age at time of post-exposure ECG, sex
- race: self-reported race; White, Black, Other (including Hispanic)
- num_drugs: number of drugs taken by patient in the given "ECG era"
- drug_concept_id, drug_name: numerical drug identifier (OHDSI), drug name
- pre_qt_500: value of 1 indicates baseline QTc interval ≥ 500 ms
- post_qt_500: value of 1 indicates post-drug exposure QTc interval ≥ 500 ms
- delta_qt: change in QTc interval from baseline to post-drug exposure (ms)
- electrolyte_imbalance: value of 1 indicates patient experienced hypokalemia and/ or hypomagnesemia up to 36 days before/ after post-exposure ECG
- cardiac_comorbidity: value of 1 indicates patient experienced atrial fibrillation, heart failure, myocardial infarction, left ventricular hypertrophy, structural heart disease, bradycardia, paced rhythms, premature complexes, and/ or conduction disorders (other than long QT syndrome) any time before and up to 36 days after post-exposure ECG
How can I use these data?
We are providing ∆QT Database as a publicly available dataset to enable researchers, clinicians, and patients to have access to investigate the effects of drug(s) of interest or to conduct new data mining studies. Take a look at an example here.