Conflicts of interest in electrophysiology and devices presentations.

AIMS: Industry collaboration with arrhythmia and devices research is common. However, this results in conflicts of interest (CoI) for researchers that should be disclosed. This study aimed to examine the quality of CoI disclosures in arrhythmia and devices presentations. METHODS: Recorded presentations from the Arrhythmia & Devices section of the ESC Annual Congress 2016-2020 were assessed. The number of words, conflicts, and time displayed was documented for CoI declarations. Meta-data including sponsorship by an industry partner, presenter sex, and institution were obtained. RESULTS: Of 1153 presentations assessed, 999 were suitable for inclusion. CoI statements were missing from 7.2% of presentations, and 58% reported ≥1 conflict. Those with conflicts spent less time-per-word on their disclosures (median 150 ms, interquartile range [IQR] 83-273 ms) compared with those without conflicts (median 250 ms, IQR 125-375 ms). One-in-eight presentations were sponsored (12.8%, n = 128). CoI statements were more likely to be missing in sponsored presentations (14.8%, n = 19) compared with non-sponsored presentations (6.1%, n = 53), P = 0.0003. Sponsored presentations contained a greater median number of CoIs (10, IQR 6-18) compared with non-sponsored sessions (1, IQR 0-5), P < 0.0001. Time-per-word spent on COI disclosures was 50% lower in sponsored sessions (125 ms, IQR 75-231 ms) compared with non-sponsored sessions (250 ms, IQR 125-375 ms), P < 0.0001. CONCLUSION: The majority of those presenting arrhythmia and devices research have CoIs to declare. Declarations were often missing or displayed for short periods of time. Presenters in sponsored sessions, while being more conflicted, had a lower standard of declaration suggesting a higher risk of potential bias which viewers had insufficient opportunity to assess.

How can Never Event data be used to reflect or improve hospital safety performance?

The absolute number of Never Events is used by UK regulators to help assess hospital safety performance, without account of hospital workload. We applied funnel plots, as an established means of taking workload into account, to published Never Event data for 151 acute Trusts in NHS England, matched to finished consultant episodes for 3 years, 2017-2020. Trusts with excess event rates should have the most Never Events if absolute number is a valid way to judge performance. The absolute number of Never Events was correlated with workload (r2 = 0.51, p < 0.001), but the five Trusts above the upper 95% confidence limit did not have the highest number of Never Events. However, a limitation to interpretation was that the data were skewed; 12 out of 151 Trusts lay below the lower 95% limit. This skew probably arises because funnel plots pool all Never Events and workload data; whereas, ideally, different Never Events should use as denominator only the relevant workload actions that could cause them. We conclude that the manner in which Never Event data are currently used by regulators, in part to judge or rate hospitals, is mathematically invalid. The focus should shift from identifying 'outlier' hospitals to reducing the overall national mean Never Event rate through shared learning and an integrated system-wide approach.