Data underlying the publication: Effect of grace period on false alarm rates of smartwatch-based OHCA detection systems: a pilot study

The data were collected during a human study with 26 participants: 15 in a ‘young’ (21-27 years, 12 male) and 11 in a ‘old’ (56-65 year, 5 male). All participants were instructed to wear a LilyGo T-Watch 2020 V3 programmable smartwatch between 12 PM and 10 PM and to cancel any alarm produced by the watch as quickly as possible by tapping its screen. The smartwatch recorded the response time from alarm onset to cancellation.<br>Prior to the experiment, the participants were informed that in real situations, the smartwatch would alert emergency medical services if an alarm was not cancelled in time. Further incentives included extra monetary compensation (€5) for the two fastest participants in each age group on top of base compensation (€12-€15). Nevertheless, we insisted on prioritizing safety over speed.<br>During each trial, the watch produced 9-18 alarms on a predefined schedule. Before triggering, the watch checked for 10 seconds of motionlessness within 10 minutes using its accelerometer, as motionlessness was expected to be an indicator considered by OHCA detection algorithms. If no motionlessness could be detected, the scheduled alarm was excluded from analysis. Furthermore, if an alarm was not cancelled within 60 seconds, the alarm would cease and be marked as censored.<br>Three alarm types were randomly scheduled:Auditory: 4 beeps (at 2050 and 4100 Hz) of 60 ms with pauses of 60 ms, followed by 580 ms of silence.Tactile: 3 vibration pulses (at 60 Hz) of 500 ms followed by an 800 ms pulse, with pauses of 400 ms.Audiotactile: Combination of auditory and tactile alarms.During all alarms, the display blinked (500 ms on, 500 ms off). The watches produced 139 auditory, 144 tactile, and 142 audiotactile alarms, yielding a total of 425 alarms.<br>Furthermore, participants were requested to keep a diary of their activities at the time of an alarm. With limited compliance to this instruction, not every activity could be matched to a smartwatch alarm or vice versa.<br>This study was performed in compliance with relevant laws and institutional guidelines, including the Declaration of Helsinki. Approval of the study was obtained from the Ethical Review Board of the Human Technology Interaction group at Eindhoven University of Technology (ref. 1906). The privacy rights of the participants have been observed, and written informed consent was obtained from each participant.

本数据集的数据来自一项包含26名受试者的人体研究：其中15人为青年组（21~27岁，男性12名），11人为老年组（56~65岁，男性5名）。所有受试者均被要求在每日12时至22时期间佩戴LilyGo T-Watch 2020 V3可编程智能手表，并通过轻触屏幕以最快速度取消手表触发的所有警报。该智能手表会记录从警报触发至警报取消的响应时长。
实验开始前，受试者被告知，在真实场景中，若警报未被及时取消，智能手表将自动呼叫紧急医疗服务。此外，本研究设置了额外激励：每个年龄组内响应最快的两名受试者可获得12至15欧元基础报酬之外，额外获得5欧元奖金。尽管如此，本研究始终坚持安全优先于速度的原则。
每次试验期间，手表将按照预设计划触发9~18次警报。在警报触发前，手表会通过加速度传感器检测受试者在10分钟内是否存在10秒的静止状态——因为院外心脏骤停（Out-of-Hospital Cardiac Arrest, OHCA）检测算法通常将静止状态作为一项核心判断指标。若未检测到静止状态，则该预设警报将被排除在分析之外。此外，若警报在60秒内未被取消，该警报将自动停止，并标记为删失数据。
本次研究共随机设置三类警报：1. 听觉警报：以2050Hz与4100Hz频率发出4段60ms蜂鸣声，每段蜂鸣间隔60ms，随后静默580ms；2. 触觉警报：3段500ms的60Hz振动脉冲，随后跟随一段800ms振动脉冲，脉冲间间隔400ms；3. 听触混合警报：结合听觉与触觉警报的复合警报。所有警报触发期间，手表屏幕会以500ms亮、500ms灭的频率闪烁。本次实验中共生成139次听觉警报、144次触觉警报与142次听触混合警报，总计425次警报。
此外，受试者被要求记录警报触发时的活动日志。但由于受试者依从性有限，并非所有活动日志均可匹配到智能手表记录的警报，反之亦然。
本研究严格遵循相关法律法规与机构指南开展，包括《赫尔辛基宣言》。本研究已获得埃因霍温理工大学人机交互研究团队伦理审查委员会的批准（编号：1906）。所有受试者的隐私权益均得到保障，且每名受试者均签署了书面知情同意书。