An Open-Source Tool for Mortality Forecasting for National and Local Health Planning in Portugal

Issue/problem

Health planning requires robust indicators to set evidence-based targets. Projections are increasingly used to anticipate trends, define measurable goals, and monitor progress. Yet their application in routine health planning remains limited. In Portugal, the National Health Plan (NHP) relies on forecasting to set national goals, while Local Health Plans (LHPs) often face methodological heterogeneity and limited technical resources to produce comparable forecasts.

Description of the problem

We developed an interactive R Shiny tool for age-standardised mortality rates by cause, using data from 1991 to 2023, that enables: (1) automated data retrieval from Statistics Portugal; (2) fitting multiple forecasting models (ARIMA, ETS, Holt, Holt-damped, Theta, TBATS, naive, random walk with drift); (3) model evaluation and selection based on residual diagnostics (including autocorrelation testing via Ljung–Box), forecast accuracy metrics (MASE, ME), and bias assessment; (4) national and municipal-level analyses.

Results

The tool generates projections for specific causes of death, stratified by sex and age group, at national and municipal levels. It supports updating the NHP and LHPs, enabling local health authorities to produce area-specific projections, identify divergent trends from national benchmarks, and prioritise local needs, ensuring alignment between national strategy and local implementation. The interface offers a standard mode for non-expert users and a customisable mode for advanced users.

Lessons

Embedding forecasting into an accessible, reproducible tool can bridge the gap between statistical methods and real-world planning. The approach is scalable and adaptable to other health indicators beyond mortality and to settings with open statistical data systems, providing a blueprint for data-driven health planning in decentralized settings and supporting more consistent, evidence-based target setting across governance levels.

Key messages

• An open-source R-based tool produces user-guided mortality projections from Statistics Portugal data, supporting evidence-based target setting for national and local health plans.

• Embedding automated data retrieval, multi-model methods, and user-friendly design enables scalable, transferable support for health planning.