This prospective, observational study evaluates the clinical utility of an artificial intelligence (AI)-based computational software device designed to support primary care practitioners (PCPs) and dermatologists in managing skin pathologies. The research explores whether the device can enhance diagnostic accuracy and optimize the referral process from primary care to specialized dermatology services.
Study Methodology and Design
The investigation is designed as an analytical study of a clinical case series. Key technical aspects include:
* Investigational Tool: A software-only medical device using computer vision algorithms to analyze images of the epidermis and dermis to provide clinical data for assessment.
* Participant Roles: 15 HCPs (including PCPs and dermatologists) evaluated with a cohort of over 100 patients.
* Procedural Workflow: PCPs captured skin images using smartphones or mobile dermatoscopes, uploaded them to the platform, and provided a diagnosis guided by the AI results.
* Evaluation Baseline: HCPs acted as their own controls, allowing for a comparison of diagnostic performance with and without the AI tool.
Quality Assurance and Registry Procedures
To ensure the integrity of the data collected within this organized system, several quality control measures were implemented:
* Data Validation and Checks: The database utilized consistency rules and logical ranges to control errors during tabulation. Computerized validation filters automatically identified missing values or inconsistencies based on predefined rules.
* Source Data Verification (SDV): A designated independent clinical monitor performed verification of anonymized source documents (e.g., image files and clinical records) against Case Report Forms (CRFs) to ensure accuracy and completeness.
* Monitoring Plan: The research team held quarterly meetings to address data collection issues, while the monitor conducted remote and, if necessary, on-site visits to ensure compliance with the Clinical Investigation Plan (CIP) and ISO 14155 standards.
* Missing Data Management: Manual editing and exploratory statistical techniques were used to detect and resolve logical errors or inconsistent values before the database was considered closed.
Sample Size and Statistical Principles The study was powered to detect a 10% improvement in diagnostic accuracy.
* Assessment Power: A sample size of 100 patients was determined to provide a 95% confidence level with an 80% power and a margin of error between 9% and 10%.
* Analytical Techniques: Central tendency and variability statistics (mean, SD) were used for quantitative variables, while qualitative variables were analyzed through frequency distributions.
* In addition to parametric tests, the McNemar test was used to analyze the specific impact of AI on HCP diagnostic choices. Statistical significance was set at alpha = 0.05.
For qualitative data, Fisher's exact or Chi-square tests were employed. Statistical significance was set at alpha = 0.05.
Safety and Ethical Standards The study complied with Regulation (EU) 2017/745 (MDR) and ISO 14155:2021. Data protection followed GDPR and Spanish Organic Law 3/2018, utilizing encrypted patient information and alphanumeric identification codes to maintain participant anonymity. All clinical data stored on the device is permanently deleted upon study conclusion.
