Mental Health Service Demand Forecasting
ML & AITime series forecasting system predicting mental health service demand across Scottish health boards using ARIMA modeling with interactive Streamlit dashboard.
Overview
💡 Challenge
Scottish health boards struggle to anticipate mental health service demand, leading to inadequate resource allocation, long wait times, and crisis-driven care. Traditional planning methods lack predictive capability for seasonal patterns and trend changes.
⚡ Solution
Developed ARIMA time series forecasting models analyzing 6 years of synthetic mental health presentation data (2019-2024) across 14 Scottish health boards. Built interactive Streamlit dashboard with multi-board comparison, seasonal decomposition, and demographic breakdowns enabling proactive capacity planning.
🎯 Impact
Enables evidence-based resource allocation with 3-6 month advance forecasting. System identifies seasonal peaks, board-specific trends, and demographic patterns, supporting strategic planning for mental health services across Scotland. Provides actionable insights for staffing, budgeting, and service expansion decisions.
Technical Details
🛠️ Tech Stack
✨ Key Features
- ✓ARIMA time series forecasting with automated parameter selection
- ✓Multi-health board comparison across 14 Scottish NHS boards
- ✓Seasonal decomposition showing trend, seasonality, and residuals
- ✓Interactive 3-6 month demand forecasts with confidence intervals
- ✓Demographic breakdown by age group and sex
- ✓COVID-19 impact analysis showing pandemic effects on presentations
- ✓Board-specific trend analysis identifying growth patterns
- ✓Historical data visualization with monthly aggregations
- ✓Year-over-year comparison charts
- ✓Interactive filtering by health board, time period, and demographics
- ✓Synthetic dataset generation with realistic healthcare patterns
- ✓Comprehensive data quality validation
- ✓Professional healthcare-themed UI with gradient designs
- ✓Export capabilities for forecast data and visualizations
Key Learnings
- →ARIMA models effectively capture seasonal patterns in healthcare demand data
- →Mental health service utilization shows strong seasonal variations requiring trend-seasonal decomposition
- →COVID-19 created structural breaks in historical patterns requiring careful model calibration
- →Synthetic data generation must preserve realistic correlations between demographics and service use
- →Health board-specific models perform better than single national model due to regional variation
- →Interactive dashboards enable stakeholders to explore forecasts across multiple dimensions
- →Time series forecasting provides actionable 3-6 month planning horizon for healthcare services
- →Statsmodels ARIMA implementation works well for univariate healthcare demand forecasting
- →Python 3.12/3.13 compatibility required careful dependency management for scipy and statsmodels
📊 Data Notes
This project uses synthetic/open data to demonstrate capabilities while maintaining privacy and confidentiality. All methods and approaches are applicable to real-world scenarios.