# Exploiting Population Activity Dynamics to Predict Urban Epidemiological Incidence

## TL;DR

This paper analyzes ambulance call patterns and urban activity data to predict regional health incident risks, revealing stable temporal patterns and linking mobile activity to emergency demand for improved healthcare policy and operational planning.

## Contribution

It introduces a novel approach combining ambulance call data with urban activity patterns and semantic place information to predict regional health incident risks.

## Key findings

- Identified stable diurnal and weekly ambulance call patterns.
- Linked urban activity dynamics to emergency incident rates.
- Developed a new metric for assessing regional health risks.

## Abstract

Ambulance services worldwide are of vital importance to population health. Timely responding to incidents by dispatching an ambulance vehicle to the location a call came from can offer significant benefits to patient care across a number of medical conditions. Moreover, identifying the reasons that drive ambulance activity at an area not only can improve the operational capacity of emergency services, but can lead to better policy design in healthcare. In this work, we analyse the temporal dynamics of 5.6 million ambulance calls across a region of 7 million residents in the UK. We identify characteristic temporal patterns featuring diurnal and weekly cycles in ambulance call activity. These patterns are stable over time and across geographies. Using a dataset sourced from location intelligence platform Foursquare, we establish a link between the spatio-temporal dynamics of mobile users engaging with urban activities locally and emergency incidents. We use this information to build a novel metric that assesses the health risk of a geographic area in terms of its propensity to yield ambulance calls. Formulating then an online classification task where the goal becomes to identify which regions will need an ambulance at a given time, we demonstrate how semantic information about real world places crowdsourced through online platforms, can become a useful source of information in understanding and predicting regional epidemiological trends.