Asthma Prediction & Diagnosis

Asthma is one of the most prevalent and costly chronic conditions in New Zealand and globally. New Zealand has a high prevalence of asthma, with one in seven children (13 percent) aged 2–14 years (110,000 children) and one in eight adults (12 percent, 452,000 adults) taking current asthma medication. Respiratory disease is New Zealand's third most common cause of death. OECD statistics indicate New Zealand has one of the highest hospital admission rates for asthma of OECD countries. According to  Asthma Infographic Impact Report 2019, on average 77 Kiwis die from asthma each year, and the death rate has not improved over the last few years. In terms of financial costs to the New Zealand economy, the cost burden of asthma to the country has increased over time, with recent reports suggesting costs over $1 billion. In 2015, asthma costed $1.018b, with $198.3m in direct costs, and $828.2 in indirect costs from workdays lost, disability affected life years, and mortality.

A key method to minimise the impact of asthma on outcomes and economic costs is to reduce exacerbations and hospitalisations, both of which are key contributors to morbidity, mortality, and costs. Early detection of worsening asthma can help providers to act proactively and minimise adverse outcomes. Our research aims to generate a prediction model leveraging diverse data sources to predict daily asthma related ED visits and admissions at population-level. As such we will be using climatic parameters and air quality predictors as the input data streams to predict the daily count of ED visits and admissions related to asthma.

The data sources for the input parameters will be Auckland Council and NIWA while the statistics for the de-identified hospital data would be obtained from ADHB. Thereupon, we will triangulate the effects of input attributes on the outcome by analysing all the de-identified data and generating correlation coefficients (using Pearson’s Coefficient). The strongly related attributes would then be used as input variables into a prediction model that would be executing a Machine Learning Algorithm and as an outcome, the findings of this research will help in forecasting the daily number of asthma related ED visits and admissions at the community level. It is a data driven approach that uses historic data, learns from it, find patterns, and generates accurate predictions from it. Knowledge gained from implementing such a model could potentially provide additional risk-stratification for more efficiency and timeliness in healthcare resource deployment.

For instance, it can help prioritise resources to patients who are at highest risk. The model can help predict the adverse impact of social factors such as overcrowding on asthma outcomes and inform proactive measures by health providers and organizations. It can also provide early warning signals to people at risk of asthma adverse events, and enable timely, proactive, and targeted preventive and therapeutic interventions, thereby reducing hospitalisations and hence cutting down the costs from ED visits related to asthma.