Currently submitted to: JMIR Perioperative Medicine
Date Submitted: Nov 2, 2026
Open Peer Review Period: Nov 13, 2026 - Jan 8, 2027
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Leveraging Artificial Intelligence to Reduce Manual Burden in Perioperative Activity Data Collection: Descriptive Operational Evaluation
ABSTRACT
Background:
Commercial activity trackers support longitudinal perioperative monitoring, but manual extraction and database entry limit study scalability.
Objective:
To describe the development of an artificial intelligence (AI)–assisted pipeline for wearable device data retrieval, preparation, validation, and database upload, and to characterize its operational performance.
Methods:
This descriptive operational evaluation examined a single combined execution (September 28, 2026) of a pipeline used weekly in two ongoing prospective perioperative cohort studies at a single academic medical center, with database agreement checks. Study records were for adults aged 50 years or older undergoing elective surgery and assigned Garmin vívofit 4 or vívoactive 5 activity trackers. The Python pipeline was developed with investigator-directed large language model (LLM) assistance. Routine retrieval and database upload ran within the study environment without invoking an LLM. Outcomes were processing completion, elapsed time, agreement of database values with exported data and Garmin Connect, data availability, and estimated manual workload avoided, based on research staff–reported time per device per week.
Results:
157 work book rows passed input checks, 143 completed Garmin retrieval, and 142 completed verified REDCap import. Full execution required 79.4 minutes. Mean processing time among verified records was 33.1 seconds (SD, 18.5). Uploaded values agreed with entries reconstructed from exported data at all 152,054 positions, and a step value was available for 56.9% of 18,191 requested record-days. In a random manual audit of 100 record-days from a later run, all stored step values matched Garmin Connect (100%; 95% CI, 96.3%-100%). Five research staff members reported a mean manual workload of 1.50 hours per device per week (SD, 0.61), corresponding to an estimated 3,298 hours for the 15,389 record-days within the stored data windows.
Conclusions:
AI-assisted programming supported development of an automated wearable data workflow with verified transfer fidelity. The efficiency is attributable to automation. LLM assistance made that automation attainable for a small clinical research team, but stored data still require independent plausibility review. The community-developed interface served as a bridge to official application programming interface (API) access.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.