CD-10 (International Classification of Diseases, 10th revision) diagnoses and OPCS- 4 (OPCS Classification of Interventions and Procedures version 4) codes recorded during any admissions to an NHS hospital from the Hospital Episodes Statistics (HES) dataset and ICD-10 codes recorded on the death record were used to define atrial fibrillation (AF, which included atrial flutter) and non-fatal ventricular arrhythmia (VA) which included resuscitated cardiac arrest but excluded events occurring on the day of death (Table 1). VAs were restricted to non-fatal events because we could not reliably ascertain all such arrhythmias contributing to death, and it was not possible to determine which cardiac arrests without resuscitation had involved an arrhythmia; all AFs recorded were non-fatal. (Information was not available from primary care records, so any diagnoses of an arrhythmia by a participant’s general practitioner without admission to hospital would only be captured if the participant reported it).

The presence of any relevant diagnoses or procedures in HES records prior to the date of randomization was used to define prior history of AF (which was not recorded at baseline). The HES records can provide evidence of incident acute arrhythmias (ventricular arrhythmias) but only evidence of the presence of chronic arrhythmias (AFs) because admission records typically include several diagnoses and it is not possible to determine reliably which diagnosis was the cause of the admission. The arrhythmic outcomes defined were therefore: AF onset among participants without known AF at randomization; and the first recorded incidence of non-fatal VA. Onset of AF was defined as the first occurrence of one of the relevant codes (Table 1) in post-randomization HES or death records (in any diagnostic position) or a reported hospitalisation or serious event of AF in the trial among participants without prior history of AF. A relevant code (Table 1) in any diagnostic position within a post- randomization HES episode or a hospitalisation or serious event reported in the trial with survival to the next day was regarded as evidence of non-fatal ventricular arrhythmia.

Numbers of arrhythmias in electronic records versus reported in trial

Prior AF was recorded in 106 participants. Among the remaining 15 374 participants new onset of AF was identified in 1121 participants in HES, compared to 287 by self- report, including 231 reported in both, giving a total of 1177 from either source (Table 2, Figure 1). A VA was identified in 132 participants in HES, in only 12 in the trial data including 9 with a report in both sources yielding 135 in total.

Frequency of hospitalisation

The frequency of hospitalisation for a diagnosis other than an arrhythmia (providing the opportunity for incidental diagnosis of AF) was similar in the two treatment groups: 71% of participants in each treatment arm had a hospital episode prior to randomization for a non-arrhythmia diagnosis and the median (inter-quartile range) of the number of episodes was 1 (0-3) in each arm; three-quarters of participants (75.9% allocated omega-3 FA vs. 75.5% allocated placebo), had at least one such hospitalisation post-randomization and the number of episodes of hospitalisation (median 2, inter-quartile range 1-5) and time to first hospitalisation were similar in the two treatment groups (rate ratio 1.005, 95% CI, 0.97,1.04).

Table 1

International Classification of Disease (ICD-10) and OPCS Classification of Interventions and Procedures (OPCS-4) codes used for arrhythmia definitions

Table 2

Numbers of participants with post-randomization arrhythmia events from self- report and Hospital Episode Statistics data