JMIR Publications

ABSTRACT

Background:

Digital health platforms that integrate patient-reported outcome measures (PROMs) with wound image submissions offer new opportunities for remote wound surveillance. However, the alignment between patient-reported symptoms and physician clinical judgment remains underexplored, particularly in real-world settings.

Objective:

This study aimed to evaluate the diagnostic performance of PROM-reported wound infection in predicting physician-initiated callbacks and to explore the symptom features associated with patients' perception of infection.

Methods:

We conducted a retrospective observational study at a tertiary medical center in Taipei, Taiwan. Patients with acute or chronic wounds were enrolled in a chatbot-assisted digital monitoring program between June 30, 2022, and March 1, 2023. Using their mobile devices, patients submitted wound photographs and completed a structured symptom checklist, including indicators such as redness, darkness, and infection. A senior plastic surgeon independently reviewed each image to determine the need for clinical follow-up (callback), which served as the reference outcome. The presence of “infection” in the PROM checklist served as the primary predictor. Sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were used to assess predictive accuracy. A secondary analysis examined associations between symptom features and infection reporting using logistic regression.

Results:

Among 2,297 wound image entries from 270 patients, PROM-reported infection showed high sensitivity (94.0%) and an AUC of 0.9575 (95% CI: 0.9502–0.9648) for predicting physician callbacks. In the acute wound subgroup, the AUC remained high (0.9335). Redness was the strongest correlate of infection reporting (OR = 31.6; 95% CI: 23.1–43.2), while skin darkness was negatively associated with perceived infection in acute wounds (OR = 0.415; 95% CI: 0.203–0.850), indicating potential misinterpretation.

Conclusions:

Patient-reported infection through a digital platform demonstrated high sensitivity in identifying wounds requiring medical attention. However, notable false-negative rates and symptom misinterpretation underscore the need for improved patient education and real-time decision support. These findings support the utility of PROM-based systems for remote triage and highlight the importance of integrating patient-clinician feedback loops to enhance wound care safety.