AI on the GPP case
Experts employ real-world data to understand GPP, diagnose it early, and target treatment.
One of the more frustrating aspects of having a dermatologic disorder, in addition to the pain and suffering, is a lack of diagnosis or misdiagnosis. According to the speakers at the Aug. 3 Sunrise Session, “Phenotyping Generalized Pustular Psoriasis (GPP): Findings of a Collaborative Artificial Intelligence (AI) Initiative,” this is a common occurrence for patients with GPP — a rare, hard-to-diagnose inflammatory disease.
GPP is a poorly understood disease in general. Therefore, its presentation can frequently be mistaken for a cutaneous or systemic infection, said session director Marta Van Beek, MD, MPH, FAAD. An improper or delayed diagnosis can have a ripple effect that impacts the patient — most of all — but also the clinician and health system.
“Earlier diagnosis allows earlier access to the right treatment and decreased suffering from this very symptomatic disease,” said Dr. Van Beek, who is executive medical director at the University of Iowa Ambulatory Clinics, and C. William Hanke professor of dermatologic surgery at University of Iowa Health Care in Iowa City.
The importance of early diagnosis
Patients living with GPP often have a “circuitous path” to learning about and confirming their condition, said Lawrence J. Green, MD, FAAD. A patient may not have access to a board-certified dermatologist, or they could be misdiagnosed by a primary care physician or other specialist without ever seeing a dermatologist.
An inadvertent misdiagnosis is extra cause for concern, said Dr. Green, who is clinical professor of dermatology at George Washington University School of Medicine in Rockville, Maryland. It can lead to inadequate, ineffective, and costly outpatient or inpatient treatments that prolong, exacerbate, or add to existing symptoms.
Tina Bhutani, MD, MAS, FAAD, further explained that not all types of psoriasis are created equal. People who are living with GPP can suffer great physical and psychological symptoms that affect their overall quality of life. Due to its difficult diagnosis, GPP also puts an increased burden on health care systems and resource utilization, said Dr. Bhutani, who is assistant professor of dermatology at the University of California San Francisco School of Medicine.
“A recent report showed that approximately 36% of patients required hospitalization following their initial encounter,” Dr. Bhutani said. “Additionally, patients may be re-hospitalized frequently, like every one to five years, for disease control and pain management, with average hospitalizations ranging from 10 to 14 days.”
The importance of data
To learn more about GPP, a group of dermatologists used AAD’s DataDerm™ registry to access data on the disease and create an AI model. The innovative technology allowed them to discover the GPP phenotype and detect patterns in patients’ health histories that signal an increased risk of GPP.
Having this “digital phenotypic pattern,” said Dr. Van Beek, can help dermatologists and other health care workers to more easily identify GPP patients and get them the right treatment years earlier than they might have otherwise. Joseph Zabinski, PhD, vice president of strategy & AI for OM1 in Boston, illustrated this concept with a graphic in his presentation.
Although AI can have promising care benefits and health outcomes, Dr. Van Beek cautioned that it can also have severe implications if physicians are not part of the process. In this particular effort, dermatologists collaborated with AI researchers employed by OM1 to thoroughly develop and test the technology prior to patient use.
“Dermatologists need to be involved and at the table to make sure that any type of augmented intelligence is used to mitigate bias, purport to do what it says it does, and minimize harm,” she said.
Prioritizing quality of life
The panel of experts collectively spoke in more detail about GPP and how the clinical insight that this AI model provides can significantly impact patient quality of life via advanced intervention. Similar technology has been used to diagnose non-dermatologic diseases, said Dr. Van Beek, and could be modified for multiple applications in the future.
“We hope to use this technology to identify other dermatologic conditions earlier in the patient journey to connect them with the most efficacious treatment,” said Dr. Van Beek.
The research completed with DataDerm also helped to corroborate claims from previous studies, said Dr. Bhutani. For example, the large dataset was able to validate a recent study’s report of GPP patients having a greater prevalence of cardiometabolic comorbid disease than patients with psoriasis vulgaris.
The topic of GPP was new for Innovation Academy 2024, and the session highlighted how real-world information can transform practices and lives. The information exists thanks to the participation of AAD members in the DataDerm registry, and speaker Elizabeth A. Kiracofe, MD, FAAD, founder of Airia Comprehensive Dermatology in Chicago, urged those who are not already contributing to the registry to sign up.
“I think members are always impressed and shocked by just how much data the registry has…and how valuable this information is for our patients,” Dr. Kiracofe said.
