Machine learning can predict hyperglycemia risk in patients with psoriasis, AJMC reports — a finding that pushes comorbidity screening in dermatology toward data-driven risk stratification rather than uniform testing schedules.

Psoriasis carries well-documented cardiometabolic associations, and hyperglycemia sits among the most clinically consequential of them. Patients with moderate-to-severe disease face elevated rates of insulin resistance and type 2 diabetes, and guidelines already encourage clinicians to monitor metabolic parameters in this population. The new work suggests algorithms trained on routinely collected variables can identify which patients actually warrant closer glycemic surveillance.

The report, covered by AJMC (A Journal of Managed Care), centers on a predictive model designed to flag psoriasis patients at heightened risk of developing hyperglycemia. By applying machine learning to patient-level data, the approach aims to surface risk earlier than conventional screening intervals would — before abnormal glucose values appear on a lab panel.

For dermatologists, the clinical logic is straightforward. Psoriasis is a systemic inflammatory condition, not merely a cutaneous one, and chronic inflammation contributes to metabolic dysfunction. Yet glycemic screening in dermatology practices remains inconsistent. A validated risk-prediction tool embedded in the electronic record could prompt dermatology teams to coordinate earlier referral, fasting glucose or HbA1c testing, and lifestyle intervention for the patients most likely to benefit.

The machine learning approach also matters for population health. If a model can stratify psoriasis patients by hyperglycemia risk using data already available — demographic and clinical variables captured during routine care — health systems could target diabetes-prevention resources more efficiently. That aligns with the managed-care audience AJMC serves, where identifying high-risk cohorts before disease onset reduces downstream spending on diabetes complications.

Interest in predicting comorbidities in psoriasis has grown as systemic therapies become more widespread. Biologic and small-molecule treatments have transformed disease control measured by instruments such as PASI, but they do not address the metabolic risk that accumulates over years of chronic inflammation. Predictive analytics fills that gap by giving clinicians a forward-looking view of comorbidity trajectory alongside the backward-looking measures of skin clearance.

Questions remain that the report does not fully resolve: the specific algorithm architecture, the training dataset and its size, the model's sensitivity and specificity, and whether performance holds across diverse populations and care settings. External validation in prospective dermatology cohorts will determine whether the tool moves from research demonstration to clinical decision support.

If validated, hyperglycemia prediction could become part of a broader pattern: machine learning models flagging cardiovascular, renal, and metabolic risk in inflammatory skin disease, with dermatologists acting as early touchpoints in prevention. The AJMC report signals that psoriasis care is moving in that direction, and further studies defining the model's accuracy and implementation pathway will shape whether risk-stratified metabolic screening becomes standard practice.