Modeling Volatility Clustering in Daily Demand for Emergency Healthcare Services in The University of Cape Coast Hospital, Ghana: A Longitudinal Study

Abstract

Introduction: Since early 2020, demand for emergency healthcare services in the University of Cape Coast hospital has become increasingly volatile. This phenomenon is a cause for concern because sudden and unexpected shifts in demand for emergency healthcare services can and have created expensive disruptions in operational activities even for very well-run and sophisticated emergency healthcare systems elsewhere. These phenomena of shifts in demand often create shocks in the system called volatility clustering. This is used as a crude measure of the risk of operations in the emergency healthcare industry for strategic planning, readiness, and swift response to demand for emergency healthcare services. Modeling these volatilities is crucial for prudent risk management practices such as proactively curtailing losses or liabilities, negative spillover effects, medicinal supply chain disruptions, and overall costs among key stakeholders. Despite its relevance, the knowledge of volatility density clustering in the demand for emergency healthcare in the hospital is complex and imprecise and the model(s) explaining it remains unknown. Methods: a longitudinal study was conducted by using high-frequency data of daily returns of the demand for emergency healthcare services in the hospital from January 2020 through June 2022. The Generalized Autoregressive Conditionally Heteroscedastic (GARCH) models were deployed to analyze the data. Eviews 10 statistical software package was used to process the data with statistical significance set at 5%. Results: this unbiased research study revealed a current daily average admissions of 35 patients on an 11-bed capacity emergency ward and an overall emergency admission of 13,656 patients over the study period. The results showed that the daily returns of the demand for emergency healthcare in the hospital exhibited high and persistent volatility clustering with the observed R-squared LM-statistic = 20.15; p < .001, for the returns series and suggesting as the optimal model out of six explanatory candidate volatility models. By implication, the associated volatility with the current operational capacity is woefully inadequate. Conclusion: countermeasures and supporting tools need to be designed and developed by management to guarantee readiness for rapid response, monitor and minimize any undesirable outcomes, and ultimately maximize quality patient care and safety.