Renal Dialysis Forecasting

The Challenge: Rising Demand for Dialysis Services

Renal dialysis places a growing burden on the state health system, with prevalence and incidence increasing steadily. Planning for future demand is complex, as projections must cover twenty years ahead based on only twenty years of historical data, cohorts vary widely, and Local Health Districts show sharply different patterns.

Our FDE Approach: Forecasting with Clinical Precision

Working with the State Ministry of Health, our Forward Deployed Engineers (FDEs) delivered a robust forecasting framework engineered for health planning, as detailed in the official government report:

• Parametric Poisson Models: Tailored to prevalence and incidence counts, capturing deterministic long-run trends without overfitting short-term fluctuations.
• Cohort-Level Forecasting: Separate models built for each age group and each Local Health District, then aggregated to a state-wide picture.
• Scenario Testing: Ability to adjust assumptions such as population growth, age distribution, and health service boundaries to observe the net impact.
• Governance and Transparency: Forecast intervals provided at 80 and 95 percent, allowing planners to weigh risk ranges rather than point estimates alone.

Empowering Planners with a Superior Model

Making Life Easier for Decision-Makers

We translate complex statistical outputs into a simple, interactive planning tool. Instead of dense spreadsheets, decision-makers get clear, actionable insights:

• Clear Visuals: At-a-glance charts show projected demand for the entire state, specific regions, and different age groups, making it easy to see where investment is needed most.
• Understand Uncertainty: By providing 80% and 95% confidence intervals, we show the range of possible futures, allowing for robust planning that accounts for best-case and worst-case scenarios.
• Test Your Assumptions: Planners can instantly see how changes in population growth or other factors would impact future demand, allowing for proactive strategy adjustments.

Why Our Approach is Better

Simple trend lines are not enough for long-term health planning with noisy data. Our Parametric Poisson Model is superior because it is purpose-built for this exact challenge:

• Focus on the Long-Term Signal: Unlike simpler models that can be confused by short-term spikes or dips in data, our model identifies the true underlying trend, providing a stable and reliable 20-year forecast.
• Designed for Health Data: It is a statistical method created specifically for 'count' data (like the number of patients), which is common in healthcare. This makes it more accurate and reliable than generic forecasting models.
• Handles Sparse Data: The model excels even with limited historical data and small case numbers in certain regions, ensuring that forecasts for all districts are statistically sound.