SIT741 – Statistical Data Analysis for Business
Assignment 2: Modelling Emergency Department Demand and Weather Impacts

Unit Information

Unit Code: SIT741

Unit Title: Statistical Data Analysis for Business

Institution: Deakin University

Assessment Title: Assignment 2 – Modelling Emergency Department Demand

Assessment Type: Individual Data Analysis Report and R Notebook

Weighting: 40% of final unit mark

Submission Method: CloudDeakin

Submission Format: Written report (Word or PDF) and executable R Notebook (.Rmd)

Recommended Length: 1,800–2,200 words excluding code and references

Assessment Context

Emergency departments frequently experience unpredictable surges in patient demand. Health service planners therefore rely on statistical modelling and predictive analytics to anticipate high-demand periods and allocate resources accordingly. Environmental conditions such as extreme heat have been repeatedly linked to increases in hospital admissions, ambulance callouts, and emergency department presentations.

This assessment extends the analytical work undertaken in Assignment 1. Students will use publicly available hospital attendance data together with historical weather data to investigate patterns in emergency department demand across Perth hospitals. The assignment introduces regression modelling, generalized additive models, and exploratory residual analysis in order to quantify relationships between environmental conditions and health service demand.

Students will construct models using the R programming environment and evaluate how weather indicators such as temperature, precipitation, and heatwave intensity relate to emergency department activity. The task emphasises statistical reasoning, data acquisition, reproducible analysis, and careful interpretation of modelling outputs.

Application Scenario

You are part of a data science team working with a regional health authority in Western Australia. The organisation wants to improve forecasting of emergency department demand across the Perth metropolitan area. Reliable forecasts help hospitals prepare staffing schedules, allocate beds, and manage crowding during peak periods.

Your responsibility is to analyse historical emergency department activity and determine whether weather patterns can help explain fluctuations in demand. The findings will support operational planning and may inform the development of future predictive models used by hospital administrators and public health analysts.

Data Sources

You are provided with the following primary datasets. You may incorporate additional credible sources where appropriate.

• Emergency department admissions and attendance dataset published by the Western Australian Department of Health.
• Daily temperature and precipitation records obtained from the NOAA Climate Data API.
• Global Historical Climatology Network (GHCN-Daily) weather dataset.

Students should document all data sources used and clearly describe any preprocessing steps applied during analysis.

Task 1: Weather Data Acquisition (5 marks)

Emergency department demand data were prepared during Assignment 1. The next stage requires sourcing relevant weather data covering the same time period.

1. Data source selection

   Identify a reliable source for historical weather observations relevant to Perth. Acceptable sources include NOAA climate datasets or the Australian Bureau of Meteorology historical observation records. Explain the rationale for your choice in terms of data availability, accessibility, completeness, and suitability for modelling.

2. Data retrieval

   Download daily weather observations for the study period used in Assignment 1. At minimum, retrieve daily temperature and precipitation variables. If using the NOAA API, request a web service access token and document the query process used.

3. Data description

   Provide a short summary of the weather dataset that answers the following questions:

   • How many rows or observations are contained in the dataset?
   • What date range does the dataset cover?
   • Which variables are included in the final dataset?

Task 2: Model Planning (5 marks)

Effective statistical modelling requires clear conceptual planning before model construction.

1. Model purpose

   Explain how the final predictive model could be used by healthcare administrators. Discuss how forecasting emergency department demand might assist hospitals in addressing overcrowding, staffing allocation, or patient flow management.

2. Variables and relationships

   Define the relationship you intend to model. Clearly identify:

   • The response variable representing emergency department demand.
   • The predictor variables included in the model.
   • Whether these predictors are routinely collected and available early enough to support real-time prediction.
3. Future data assumptions

   Discuss whether future data are likely to share similar characteristics with historical observations. Consider seasonal patterns, climate variability, and changes in healthcare utilisation.

4. Statistical approach

   Identify the statistical methods you plan to use for modelling and explain why they are appropriate. Methods may include linear regression, generalized additive models (GAM), or other regression-based approaches.

Task 3: Modelling Emergency Department Demand (10 marks)

Let the emergency department demand variable defined in Assignment 1 be represented as Y. You will construct and refine models that explain changes in this variable.

Task 3.1: Model Development for a Single Hospital

1. Select one hospital randomly from the Perth emergency department dataset and identify it in your report.
2. Fit a linear regression model using date as the predictor variable for Y. Produce plots of fitted values and residuals. Evaluate whether a linear trend adequately captures patterns in demand.
3. Relax the assumption of linearity by fitting a generalized additive model (GAM). Compare the results with the previous model and assess residual patterns for evidence of model inadequacy.
4. Extend the model to include weekly seasonality. Compare competing models using the Akaike Information Criterion (AIC). Present the best-fitting model using coefficient estimates and graphical outputs.
5. Analyse the residuals for remaining temporal correlation or systematic structure.
6. Discuss the treatment of the day-of-week variable. Explain whether it should be represented as numeric, ordinal, or categorical and describe how that decision influences model fit.

Optional Task 3.2: Models for All Hospitals (Bonus)

1. Use functional programming techniques such as the map() function to fit models for all hospitals in the dataset.
2. Plot model trends and residual structures across hospitals. Compare patterns and interpret differences in demand behaviour.

Task 4: Heatwaves and Emergency Department Demand (15 marks)

Extreme heat events have been associated with increased emergency department visits. The next stage investigates this relationship quantitatively.

Task 4.1: Measuring Heatwaves (7 marks)

1. Review the heatwave index proposed by John Nairn and Robert Fawcett known as the Excess Heat Factor (EHF).
2. Using NOAA temperature data, calculate daily EHF values for the Perth region during the study period.
3. Create a time-series plot showing daily EHF values and identify periods classified as heatwaves.

Task 4.2: Model Extension with Heatwave Indicator (8 marks)

1. Extend the previously developed model by including EHF as an additional predictor variable.
2. Estimate the effect of heatwave intensity on emergency department demand.
3. Evaluate whether the additional variable improves model performance using appropriate model comparison statistics.
4. Interpret the findings in the context of healthcare planning and emergency department capacity management.

Optional Task 4.3: Additional Weather Predictors

Propose additional meteorological indicators that could influence emergency department demand. Examples may include humidity, wind conditions, or heat index. Integrate one additional feature into the model and evaluate whether predictive performance improves.

Task 5: Reflection (5 marks)

Provide a short analytical reflection addressing the following questions.

1. What limitations arise when regression models are built solely from historical data?
2. Regression analysis can support process understanding or prediction. Which objective is more appropriate in this assignment and how would that choice influence modelling decisions?
3. Assess whether the analyses conducted answered the research questions originally proposed.

Submission Requirements

• One written report submitted as a Word document or PDF containing responses to all assignment tasks.
• An executable R Notebook file titled Assignment2_submission.Rmd containing all code used in the analysis.
• The notebook must run successfully and include explanatory comments describing each stage of the workflow.
• All required R packages must be clearly listed.

Marking Criteria

Sample Analytical Insight (Illustrative Example)

Emergency department demand often shows clear seasonal and short-term temporal patterns that simple linear models cannot adequately capture. A generalized additive modelling approach allows non-linear trends to emerge while preserving interpretability for health service planners. Evidence from Australian heatwave studies indicates that sustained high temperatures are associated with increased hospital presentations, particularly among older adults and patients with cardiovascular or respiratory conditions (Nairn and Fawcett, 2015). Incorporating a heatwave intensity indicator such as the Excess Heat Factor therefore provides a plausible explanatory variable within predictive models of ED activity. Statistical models that include temperature extremes, weekly seasonal effects, and long-term trends often produce more accurate demand forecasts than models based solely on historical admission counts. Such analyses support proactive hospital planning during periods of environmental stress.

Health system research consistently shows that climate conditions influence healthcare utilisation patterns. Australian studies have documented measurable increases in emergency presentations during prolonged heatwaves, with significant strain placed on urban hospitals during extreme summer events. Climate-sensitive modelling approaches therefore offer practical decision-support tools for hospital administrators responsible for staffing, bed capacity planning, and ambulance diversion strategies. Integrating meteorological predictors into demand forecasting models contributes to more resilient health systems under changing climate conditions.

Complete a 1,800–2,200 word SIT741 assignment analysing emergency department demand in Perth using regression models, weather data, and heatwave indicators. Analyse hospital ED demand using statistical models and weather data, then evaluate how heatwaves influence healthcare utilisation.

References

Nairn, J. and Fawcett, R. (2015) ‘The excess heat factor: a metric for heatwave intensity and its use in classifying heatwave severity’, International Journal of Environmental Research and Public Health, 12(1), pp. 227–253. https://doi.org/10.3390/ijerph120100227

Gasparrini, A. et al. (2018) ‘Heat-related mortality in 43 countries’, Environmental Health Perspectives, 126(8). https://doi.org/10.1289/EHP2236

Vicedo-Cabrera, A. et al. (2021) ‘The burden of heat-related mortality attributable to recent human-induced climate change’, Nature Climate Change, 11, pp. 492–500. https://doi.org/10.1038/s41558-021-01058-x

Zhao, Q. et al. (2019) ‘Global, regional, and national burden of mortality associated with non-optimal ambient temperatures’, The Lancet Planetary Health, 3(7), pp. e273–e282. https://doi.org/10.1016/S2542-5196(19)30135-7

Xu, Z., FitzGerald, G., Guo, Y., Jalaludin, B. and Tong, S. (2020) ‘Impact of heatwave on mortality under different heatwave definitions’, Environmental International, 134. https://doi.org/10.1016/j.envint.2019.105241