The resulting models can efficiently detect disease onset based on patient’s clinical features and their medical history and can be deployed at the point of care. Models’ clinical relevance is ensured thanks to an interpretability layer.
Use case examples include:
- Improving/accelerating early diagnosis in rare disease patients
- Estimating the true incidence of an under-diagnosed disease
- Assisting patients in self-diagnosis and self-administration of over-the-counter drugs
Diagnosis algorithm in a rare disease
Context: challenging diagnosis of rare conditions
Our algorithms and methodologies are typically impactful in rare diseases, and when leveraging real-world data. Rare diseases have an incidence lower than 5 cases per 10,000. Hence, rare diseases are often late or mis-diagnosed. This might be due to patients heterogeneity, low prevalence, or a very limited knowledge on the clinical course of disease. While late diagnosis translates into futile or ineffective treatment in the real world, testing unnecessarily “suspicious patients” can lead to useless healthcare resource use and costs.
Quinten Health’s solution: the combination of clustering and predictive machine learning
As an example, to address such issues observed in a rare metabolic disorder, Quinten Health combined clustering and predictive machine learning algorithms to develop an early diagnosis proxy. The proposed algorithm was trained and validated on linked electronic medical records and claims data of almost 1,000 patients.
Our ML algorithm was then compared with the current medical practice to test and diagnose potential patients. Using the patient’s characteristics available in routine practice, our ML algorithm proved to detect more true cases while performing twice less laboratory testing.
Impact: better diagnosis and lower public health cost
As a result, such approaches can largely reduce the number of misdiagnosed patients with substantial savings on the screening cost. In several indications, early diagnosis also translates into critical improvement in disease-modifying treatments effectiveness for which late treatments are often ineffective. Anticipating such issues in drug development and trials recruitment maximizes chances for treatment success in real world.
 Committee for Orphan Medicinal Products (COMP) – https://www.ema.europa.eu/en/committees/committee-orphan-medicinal-products-comp
 Blöß, S., 2017. Diagnostic needs for rare diseases and shared prediagnostic phenomena: Results of a German-wide expert Delphi survey. PLoS One, 12(2), pp. 1-12.