Please login or create an account. If you do not have access to this content, you will be shown a 30 second preview and licensing options.

Presentation

AI-Based Digital Phenotyping to Identify Undiagnosed Hidradenitis Suppurativa Earlier

Description

The speaker described findings from an AI-based digital phenotyping model aimed at identifying undiagnosed hidradenitis suppurativa (HS) earlier. The model performed well in distinguishing HS from non-HS patients and was especially informative when analyzing patients who had previously seen dermatologists versus those who had not. Among patients with dermatologic histories, the strongest HS signals included classic skin-related clues such as cutaneous abscesses, carbuncles, furuncles, sweat gland disorders, prior abscess treatment, recurrent cellulitis, acne, dermatitis, and use of topical treatments. The talk emphasized that dermatologists should ask about these common symptoms when abscesses or treatment-resistant skin problems appear in the chart, and should also educate primary care and emergency clinicians to recognize early HS. Other important predictors included female sex, younger age, obesity, chronic pain, recurrent pain treatment, gynecologic visits, repeated STI testing, anxiety, depression, and surprisingly, upper respiratory infections. The speaker noted that HS was highly concentrated in women, but that normal weight does not exclude HS. For patients who had not seen a dermatologist, the model relied more on general comorbidities such as obesity, tobacco use, hypertension, diabetes-related medications, and infection or pain treatments, rather than skin-specific findings. Overall, the presentation concluded that AI-based digital phenotyping can help identify high-risk patients and support earlier diagnosis and treatment of HS.

View more

Conclusions

AI-based digital phenotyping can identify patients at high risk for undiagnosed hidradenitis suppurativa (HS).
The predictive signals differ depending on whether patients have already seen a dermatologist, with skin-specific clues dominating in the dermatology group and more general comorbidities dominating in the non-dermatology group.
Classic HS clues such as cutaneous abscesses, carbuncles, furuncles, sweat gland disorders, and prior abscess treatment are the strongest indicators of missed disease.
Non-classic findings like severe acne, cellulitis, dermatitis, and topical treatment for dermatitis can also serve as meaningful early warning signs.
HS remains heavily concentrated in women, so undiagnosed disease should be strongly considered in female patients.
A substantial fraction of HS patients are not obese, meaning normal weight does not exclude the diagnosis.
Recurrent upper respiratory infections may be associated with HS, but this relationship needs further study before it can be used confidently.
Chronic pain and repeated pain treatment are important contextual signals that may point toward unrecognized HS.
Gynecologic complaints, repeated STI testing, and genital symptoms may reflect HS involvement in the groin and should not be overlooked.
Anxiety and depression may function as additional suspicion-raising clues when combined with other HS indicators.
The model performed better than chance and had useful discrimination, supporting its potential for clinical screening.
These findings suggest that earlier recognition and referral could improve HS diagnosis and treatment if clinicians actively look for these patterns.