Epidemiology Forecasting
Forecasting the available patient population of a disease area using a natural history model
Problem Statement
A US-based Big Pharma company wanted to assess the market opportunity 5 years down the line when their in-development drug would launch. It was necessary to estimate the patient population in multiple geographies and by specific high-risk sub-categories. The company had a previously developed natural history model that they wanted to use as a starting point.
Challenges
- Lack of good documentation in the previously developed forecasting model
- No sources available for the numbers given out by competitors and thus hard to evaluate their accuracy
- Niche disease area which lacked longitudinal studies
- Large number of geographies to be evaluated, each with its own epidemiology and healthcare system
Solution
Summary
I enhanced the existing natural history model by incorporating new features and adapting it for diverse geographic regions. Through an extensive literature review, I assessed the robustness of each study, identified key data points, and used them to calibrate the models for improved accuracy.
Key Features
- Used exponential growth curves to model incidence and logistic growth curves to model changes in diagnosis and treatment rates
- Triangulation of calibration points based on multiple research sources including academic research and commercial research
- Detailed estimation of incident, prevalent, diagnosed and treated populations by year from 1980 to 2030
- Estimation of patient population by disease severity and by patient sub-population
Key Outcomes
- Unique modeling approach lead to interesting insights around gaps in knowledge about disease progression among researchers
- Generation of insights around the evolution of diagnosis and treatment rates as curing therapies are discovered
- Recommendations on Go/No-Go decisions for senior management of the client organization based on estimated market landscape around the time of launch
Example Incidence Curve
Analyzed Data Source
Literature Review that yielded Critical Data Points for Model Calibration
Nanda has demonstrated an ability to translate between business question and analysis design, and back to the business answer from the results. I consider him a partner in helping me think through some complicated questions.
Josh H.
Business Analytics Manager, ZS
Have a similar challenge?
Let's discuss how we can develop solutions for your specific use case.