# Real-World Application of an AI-Assisted Digital Workflow for Clinical Observational Data Collection in Dermatology Residency Programs: An Innovative Pilot Report From India

## TL;DR

A low-cost AI-assisted digital workflow using Google tools significantly improves data collection efficiency and accuracy in dermatology residency programs.

## Contribution

A novel, low-cost AI-assisted workflow using Google Workspace tools for efficient and accurate clinical data collection in resource-limited settings.

## Key findings

- AI-assisted workflow reduced data entry time by 16.9 minutes per record compared to paper-based methods.
- Error rates dropped from 8.54% to 2.39%, with no new errors introduced by the digital workflow.
- The workflow is scalable, rapidly deployable, and suitable for resource-constrained academic centers.

## Abstract

Introduction: Residency research is a cornerstone of academic dermatology training, yet paper-based data collection remains labor-intensive, error-prone, and a major source of stress for trainees. Electronic data capture (EDC) platforms can mitigate these issues but are often costly and difficult to implement in resource-limited settings. We evaluated a low-cost, artificial intelligence (AI)-assisted workflow leveraging widely available Google Workspace tools to streamline data collection and enhance data integrity during dermatology residency research.

Methods: In this prospective, single-center simulation study, 100 hypothetical patient records with 41 variables each (total 4,100 fields) were independently entered using (i) conventional paper-based data collection with subsequent manual digitization, and (ii) a digital workflow using Google Forms, automated coding in Google Sheets, PDF generation, and real-time AI-assisted validation via ChatGPT. Mean entry time per record, error rates per field, and absolute risk reduction were calculated. Paired t-tests, McNemar’s test with continuity correction, exact binomial p-values, and Newcombe 95% confidence intervals were used for analysis.

Results: The AI-assisted workflow reduced mean entry time from 24.4 ± 1.5 minutes to 7.5 ± 1.1 minutes per record (mean paired difference=16.9 minutes, 95% CI: 16.61-17.19, p<0.001, Cohen’s d=11.66). Error rates decreased from 8.54% (350/4,100) to 2.39% (98/4,100), yielding an absolute risk reduction of 6.15 percentage points (95% CI: 4.82-7.47) and a 72% relative reduction in errors (McNemar’s χ²(1)=6.13, p=0.013; exact binomial p=0.0078). No cases demonstrated new errors unique to the digital workflow.

Conclusion: An AI-assisted, Google Workspace-based workflow significantly reduced both time and error rates in simulated dermatology research data collection. This approach is low-cost, rapidly deployable, and scalable to resource-constrained academic centers. Adoption of such workflows has the potential to improve research efficiency, enhance data integrity, and reduce resident stress, ultimately fostering a stronger research culture in dermatology training programs.