Why we built SkinTracker Link to heading
Clinical trials for atopic dermatitis (AD) still rely on in-person visits, which drive up costs, limit geographic diversity, and create barriers for patients who can’t easily travel. Our goal was simple: test whether a smartphone-based app could capture disease severity as reliably as a clinician-examined photograph or a DSLR-grade image, while keeping participants engaged over an eight-week hybrid study.
Study design in a nutshell Link to heading
| Item | Details |
|---|---|
| Population | 28 volunteers (18 with physician-diagnosed AD, 10 healthy controls) |
| Duration | 8 weeks (baseline + week 8 in-clinic visits; bi-weekly remote tasks) |
| Assessments | Eczema Area Severity Index (EASI), Investigator Global Assessment (IGA), Patient-reported outcomes (POEM, NRS itch) |
| Data capture | Mobile-app photos (34 standardized poses per session), DSLR photos (baseline & week 8), Apple Watch activity & sleep data |
| Compliance checks | Bi-weekly phone check-ins, push notifications, auto-quality detection on the app |
Key quantitative findings Link to heading
- Agreement between modalities: Intraclass correlation coefficients (ICCs) for the primary scores were excellent:
- EASI ICCs ranged from 0.90–0.97 (app vs. in-person, app vs. DSLR).
- IGA ICCs ranged from 0.88–0.94.
- Component-level performance: The four EASI sub-domains (erythema, edema/papulation, excoriation, lichenification) showed moderate-to-good agreement (ICCs ≈ 0.50 – 0.83).
- Participant compliance: 97.6% of scheduled photo tasks were completed; >95% of POEM and itch surveys were submitted.
- User satisfaction: On a 1-7 Likert scale, the convenience of the app versus an all-in-person protocol scored a mean of 6.7 (SD = 0.5).
What worked (and what didn’t) Link to heading
What worked Link to heading
- Automatic blur/dark detection forced participants to retake unusable images, keeping data quality high.
- Bi-weekly phone check-ins were repeatedly cited as the biggest motivator for adherence. Human contact beats push notifications alone.
What fell short Link to heading
- The voice-diary feature lacked prompts, so participants rarely used it.
- Apple Watch bands caused skin irritation for a subset of AD patients, highlighting the need for hypoallergenic accessories in future protocols.
Limitations Link to heading
- Sample size: Only 18 AD patients; statistical power is modest and the confidence intervals around ICCs are relatively wide.
- Demographic skew: Median age ~34 years; younger, tech-savvy participants may over-represent compliance compared with older or less digitally literate populations.
- Device lock-in: All participants needed an iPhone (the app currently runs only on iOS). This excludes Android-only users and amplifies the “digital divide” we already warned about.
Implications for remote dermatology research Link to heading
The data show that a well-engineered mobile-photography workflow can deliver clinically acceptable severity scores. For sponsors, this means:
- Cost reduction: Fewer in-clinic visits translate directly into lower travel reimbursements and staff time.
- Broader reach: With proper device provisioning (e.g., loaner phones, cross-platform apps), trials could tap under-represented communities that are traditionally excluded by geography.
- Hybrid designs: Even with perfect remote data, occasional in-person checkpoints remain valuable for safety monitoring and calibration.
Next steps Link to heading
- Expand the app to Android and web interfaces.
- Integrate a structured voice-diary prompt library.
- Conduct a larger, multisite validation that deliberately oversamples older adults and non-English speakers.