AAD
Author: Willhemenia Loremipsum Special session: Lorem ipsum dolor sit amet Category: Lorem ipsum View all results

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

Using AI and Dataderm to Identify Undiagnosed Hidradenitis Suppurativa Earlier

Description

The presentation described a collaboration between HS experts, the AAD Dataderm registry, and OM1 to use artificial intelligence to identify undiagnosed hidradenitis suppurativa earlier. Dataderm’s large dataset includes millions of patient visits, making it useful for studying even difficult diseases like HS. The speaker emphasized the serious burden of HS, including pain, sleep and sexual health disruption, unemployment, depression, anxiety, and increased suicide risk, along with the long diagnostic delays and repeated misdiagnoses patients often experience. Because early treatment matters and delayed care reduces response to therapies, the team built an AI tool called Patient Finder to detect clues in patient journeys that suggest HS before diagnosis. The project had three phases: developing and testing the model, sharing educational findings, and planning a quality improvement effort to make the insights actionable. The model performed well, with an AUC of 0.83, and worked across age groups, demographics, and regions. It identified expected signals such as abscesses and also unexpected ones, like more upper respiratory infections among HS patients. The team plans to continue education through podcasts, questions of the week, and a manuscript, while beginning phase three to improve access and earlier care.

View more

Conclusions

  • The project suggests that combining a large dermatology registry with broader real-world health data can successfully identify previously undiagnosed hidradenitis suppurativa patients using AI.
  • The Patient Finder model showed strong performance, with an AUC of about 0.83 and consistent results across age, demographic, and geographic subgroups.
  • The algorithm’s highest-risk patients were much more likely to truly have HS, indicating that the approach could meaningfully shorten time to diagnosis.
  • In addition to classic clues like abscesses and cysts, the model surfaced broader signals such as chronic pain, upper respiratory infections, gynecologic care, STI testing, anxiety, depression, and metabolic or lifestyle factors.
  • The findings support the idea that HS is often missed because patients move through repeated visits and misdiagnoses before receiving appropriate care.
  • Earlier diagnosis matters because delayed treatment appears to reduce response to biologics and other therapies, reinforcing a real clinical window of opportunity.
  • The project’s next likely benefit is educational: the discovered patterns can be used to improve clinician awareness and recognition of HS.
  • A structured quality-improvement effort may turn the AI findings into better access to care and more timely management for HS patients.
  • JAAD. 2024 Dec;91(6S):S8-S11.
  • J Am Acad Dermatol. 2022 May;86(5):1092-1101.
  • J Eur Acad Dermatol Venereol. 2016 Nov;30(11):1965-1970.
  • Lancet 2025.
  • Br J Dermatol 2021.
  • JAAD 2021.
  • Dermatol Ther 2024.