Geospatial Analysis to Optimise NHS Diagnostic Capacity

This repository contains code for the study “Geospatial Analysis to Optimise Diagnostic Capacity Across the NHS South West Region.” It uses population health data and geospatial methods to model accessibility and demand for CT, MRI, and Endoscopy services, informing strategic expansion of Community Diagnostic Centres (CDCs).

Table of Contents

1. Introduction
2. Methodology
3. Key Results
4. Relevance and Impact
5. Technical Stack

---

Introduction

In line with the NHS Long Term Plan and the Richards Review (2020), CDCs are being rolled out to expand diagnostic capacity and separate elective from acute patient flows. This project addresses a central planning question: where to place new facilities to maximise equitable access.

Using a competition-aware geospatial framework, it quantifies current accessibility, identifies underserved populations, and simulates the impact of targeted capacity uplift across the NHS South West region.

---

Methodology

It follows a three-stage pipeline: (1) build the geospatial and demographic foundation, (2) model modality-specific demand, and (3) run spatial access models and scenario tests.

Data Sources

• Geospatial spine: ONS 2021 Census for 3451 LSOAs in NHS SW.
• Diagnostic activity: NHS Diagnostic Imaging Dataset (DIDS) 2024 (3 989 188 anonymised tests across 142 providers aggregated in demand/1000 5 year age-band).
• Health infrastructure: Site locations and capabilities from SHAPE Place Atlas and NHS asset registers.

Demand & Access Modelling

1. Age-based demand (CT/MRI): For each LSOA, expected demand is calculated by applying 5-year age-band usage rates (derived from DIDS) to local population structure.
2. CT & MRI accessibility (E3SFCA): We compute an accessibility score (F_i) per LSOA using the Enhanced 3-Step Floating Catchment Area method:
   • 60-minute car catchments,
   • capacity competition,
   • stepwise travel-time decay weighting nearer facilities more highly.
3. Endoscopy accessibility: Access measured as competition-adjusted rooms per 100 000 population (aged 50–74), aligning with screening cohorts. LSOAs are classified as:
   • Adequate (≥ 4.0),
   • Marginal (3.5–< 4.0),
   • Low (< 3.5),
   • No Access (0).
4. Scenario testing: Candidate new sites are placed in clusters with bottom tertile access + above median demand. We re-run models with capacity uplifts of +5%, +10%, and +20% to quantify improvement.

---

Key Results

• CT: A +10% capacity uplift produced a strong system-wide response, improving access for 90.2% of LSOAs. The median accessibility score increased by +6.34%.
• MRI: A larger +20% uplift was needed for a systemic response; 61.8% of LSOAs improved, with a median gain of +10.11%.
• Endoscopy: Modest, targeted investments yielded significant gains. Adding 7 rooms (≈ +7.8%) increased Adequate LSOAs from 1,386 → 1,767 and halved No Access (32 → 16).

---

Relevance and Impact

This framework supports NHS planners to:

• Operationalise policy: Link capacity planning to population need and spatial equity as set out by the Richards Review.
• Tailor strategy by modality: Protect CT at acute sites for emergency resilience; prioritise broader community rollout for MRI and Endoscopy.
• Run “what-if” analyses: Test proposed CDC sites so investment reduces access deserts and benefits the most underserved communities.

---

Technical Stack

Developed in Python 3.11 with:

• pandas — data manipulation
• geopandas — spatial processing
• scikit-learn — KMeans clustering
• matplotlib — visualisation
• pyproj — CRS transformations