Situation

A small biotechnology company had launched a therapy for a rare hematologic disease characterized by highly heterogeneous, non-specific symptoms, low physician awareness, and prolonged delays to diagnosis. Many patients experienced years of fragmented care before receiving the appropriate workup, limiting timely access to treatment and contributing to substantial clinical and healthcare burden. The client sought an innovative way to identify patients earlier and shorten an average diagnostic journey that can extend to nearly a decade.

Approach

HealthQuest designed and developed a symptom-driven predictive model to flag patients at elevated risk who may warrant further evaluation and testing. Using Komodo Health™ claims data spanning more than 4 million patients, the team applied a supervised learning approach to identify combinations of diagnoses and procedures most predictive of eventual disease, then translated those patterns into an interpretable rules-based algorithm suitable for practical use. SHAP analysis was used to quantify the influence of individual symptoms and procedures on model predictions and provide transparency into key drivers of risk.

Impact

The engagement delivered a trained and tested predictive model that demonstrated meaningful ability to distinguish at-risk patients from those unlikely to have disease, establishing a strong analytic foundation for future deployment in real-world care settings. The work provided the client with a scalable, data-driven approach to support earlier testing and referral, with the potential to reduce diagnostic delay, uncover previously unrecognized patients, and lower avoidable downstream utilization associated with years of missed or incorrect diagnosis.