Denmark covers approximately 42.952 km2 divided into 5 regions (i.e., the North Denmark Region, Central Denmark Region, Region of Southern Denmark, Region Zealand, and Capital Region of Denmark) and 98 municipalities [7, 8]. The annual Danish residential population was approximately 5.9 million individuals in 2023 [9]. The Danish healthcare system is free of charge as it is funded by the Danish tax system and managed by the central government, regions, and municipalities.

Denmark has a long history of recording health and social data on Danish residents, gathering this into various nationwide administrative registers. The Danish Civil Registration System, established in 1968, introduced the unique civil personal registration (CPR) number to all Danish residents, which has since been given to Danish residents either at birth or upon immigration [10]. The CPR number is included in all Danish nationwide administrative registers, making it possible to link information across different registers to the individual as well as other data sources if the CPR number is available. Most of the common Danish nationwide administrative registers have been previously described in detail elsewhere [10,11,12,13,14,15,16,17,18].

In accordance with the Danish legislation, it is possible to use the registers if the purpose is for scientific and statistical research of significant societal importance [19]. The variety of health, social, and economic data with long-term follow-up available from the registers provide a unique possibility to perform nationwide register-based studies on the Danish population.

The Danish Nationwide ECG Cohort is an open real-world cohort containing information on high-quality standard 12-lead digital ECG recordings from pre- and in-hospital settings in Denmark, in which prehospital refers to ambulance ECGs and in-hospital to in- and outpatient ECGs. The cohort includes both pediatric and adult patients with at least one ECG examination performed between January 01, 2000, and December 31, 2021, and is planned to undergo continuous updates with new ECGs from all Danish regions that is scheduled to occur annually. Of note, ECG data from primary care and private hospitals or clinics are not available in this cohort as these services are privatized and not operated by the Danish Regions.

All ECG data are stored securely on servers at Statistics Denmark, and data access is only accessible through affiliation to a Danish research institution, ensuring compliance with data protection regulations and preventing public access [20]. Thus, datasets cannot be made publicly available. Researchers that wish to access the data and programming codes can contact the authors of this study for collaboration on further studies.

The Danish Nationwide ECG Cohort contains information about the CPR number of patients, the date and time of ECG acquisition, the setting where the ECG was acquired, and in which region the ECG examination was performed. Furthermore, standardized ECG diagnostic statements and ECG measurements encompassing both global parameters and lead-specific measures of waveform amplitudes, durations, and intervals are available.

In the present paper, the Danish Civil Registration System has been used to report demographic data for the cohort description [10], and patients were not included in case of missing data on age or sex. Furthermore, data on vital status were obtained from the Danish Cause of Death Register to ensure that the included patients were not erroneously recorded with a death date prior to the ECG recording date [12].

As stated previously, it is possible to link the ECG data to all of the other Danish nationwide administrative registers through the CPR number to obtain health and social data such as blood test results, comorbidities, medication use, procedures, income, education, employment, and long-term clinical outcomes including mortality data.

Data analysis was performed using R version 4.2.1 (R Core Team, Vienna, Austria).

In both the pre- and in-hospital setting, trained healthcare personnel followed a standardized protocol to digitally record standard 12-lead ECGs, preferably while the patient was at rest and in supine position. All ECG data were stored in the MUSE Cardiology Information System (GE Healthcare, Wauwatosa, WI, USA). To ensure consistency and avoid discrepancies caused by different algorithms and vendors during the ECG data sampling period, the Marquette 12SL algorithm version 23 was utilized to reanalyze all ECG data [21]. This process ensured standardized and uniform ECG diagnostic statements as well as global and lead-specific ECG measurements.

All ECGs were filtered to within the band between 0.16 and 150 Hz. The Marquette 12SL Hookup Advisor was utilized to evaluate the quality of ECG leads and assign them to three levels (i.e., green, yellow, or red) based on factors such as muscle tremor, baseline sway, AC interference, electrode noise, and lead saturation [21]. We excluded ECGs flagged as red from further analysis, as were ECGs that were incompatible with the Marquette 12SL algorithm reanalysis, displayed flatline recordings in any lead, had a heart rate of 0 beats per minute, deviated from the standard 12-lead configuration, were shorter than 10 s, had a sampling rate below 500 Hz, were duplicates, or had invalid CPR numbers preventing linkage to Danish nationwide administrative registers. For the remaining ECGs, the MUSE Cardiology Information System, along with the Marquette 12SL algorithm, generated standardized ECG diagnostic statements enabling the identification of commonly encountered ECG abnormalities. These statements align with the recommendations of the AHA/ACC/HRS (i.e., the American Heart Association, the American College of Cardiology, and the Heart Rhythm Society) [22].

To obtain ECG measurements encompassing both global parameters, such as heart rate, P-wave duration, PR interval, QRS duration, QT interval, corrected QT interval, frontal axis of P, QRS, and T waves, as well as lead-specific measures of waveform amplitudes, durations, and intervals, the Marquette 12SL algorithm generates a representative median beat in each lead, formed by aligning similar-shaped P-QRS-T complexes. Subsequently, fiducial points, including onset, offset, and peak points of ECG waveforms and segments, are derived from the temporally aligned complexes and utilized for the calculation of waveform amplitudes and durations. The Marquette 12SL algorithm adjusts the median complex in such a way that the voltage at the QRS onset is defined as 0. Consequently, all lead-specific amplitudes and ST levels are measured in µV relative to the voltage at the QRS onset (Fig. 1A). In addition, the global intervals measured by the 12SL algorithm represent the duration between the earliest and latest waveform deflection observed in any lead (Fig. 1B). Waveform areas computed by the Marquette 12SL algorithm require multiplication by 19.52 µV⋅ms to make the areas comparable to area measurements by non-GE Healthcare software [23, 24]. A comprehensive description of all derived ECG variables in the Danish Nationwide ECG Cohort is reported in Table 1.

The ECG reflects specific cardiac events and electrical activities. The P wave represents atrial depolarization, initiated by electrical impulses from the sinoatrial node. The PR interval represents conduction through the atrioventricular node. The QRS complex represents ventricular depolarization, marking the onset of systole and ventricular contraction. The T wave represents ventricular repolarization, with the ST segment representing an electrically neutral phase between ventricular depolarization (QRS complex) and repolarization (T wave). Finally, the QT interval represents the time taken for both ventricular depolarization and repolarization, effectively marking the time from ventricular isovolumetric contraction to relaxation. (A) The P and T waves can exhibit either unipolar or bipolar morphologies, resulting in positive, negative, or zero values for P, P′, T, and T′ waves depending on the waveform configuration. Due to the standard definition of the Q, S, and S′ waves as negative deflections, their amplitudes are represented as positive values, with the implicit understanding that they are negative deflections. STJ, commonly referred to as the J point, is defined as the ST level at QRS offset relative to QRS onset. The ST level at the QRS offset plus 1/16 of the average RR interval represents STM. Similarly, STE refers to the ST level at the QRS offset plus 1/8 of the average RR interval. (B) As opposed to human readers who may only inspect the QRS duration in any single lead of the ECG, the Marquette 12SL algorithm measures global intervals from the earliest to the latest waveform deflection across all 12 leads as represented by the red dots

The digital waveforms are also available as Extensible Markup Language (.XML) files containing the waveform data as well as global and lead-specific measurements. GE Healthcare supported the export of ECGs to .XML files during the data collection phase, ensuring that all ECGs could undergo reanalysis using the same version of the Marquette 12SL algorithm. GE Healthcare also assisted in exporting the reanalyzed ECGs to make the .XML files processed by the same version of 12SL accessible.

Several studies have leveraged the stability and accuracy of the Marquette 12SL algorithm for ECG measurements including amplitudes, durations, and intervals [25], and there is little evidence to suggest that manual methods are advantageous for large clinical trials or epidemiological studies compared with automated methods.

Details and criteria on the ECG diagnostic statements and ECG measurements, as generated by the Marquette 12SL algorithm, have been published in detail previously [21, 25].

Approval to collect the nationwide ECG data was granted by the Record Data Team, Center for Health in the Capital Region of Denmark (approval number: R-21032357). Additional approval to process the data sources for statistics and scientific research purposes was granted by the data responsible institute of the Capital Region of Denmark (approval number: P-2019-533) in compliance with both the Danish Data Protection Act and the General Data Protection Regulation [19, 26].