Shifting patterns in emergency department attendance: a time series analysis

Background Rising demand and limited capacity in primary care are often cited as reasons for the increasing pressure on emergency departments (EDs). The COVID-19 pandemic further strained but also reshaped healthcare services and their accessibility. However, an equally critical yet often overlooked factor is the increasing complexity of cases. This study assessed ED attendance trends for the Northwest London (NWL) population between February 2017 and September 2023, before, during and after the pandemic lockdown measures (March 2020–March 2021) in the UK as a whole and across sociodemographic and multimorbidity profiles. Method We used the Whole System Integrated Care data warehouse in NWL. We conducted a segmented time-series quasi-Poisson regression for weekly ED attendance for two periods, before and after the pandemic lockdown measures, adjusting for seasonality and autocorrelation. We stratified the model by age, sex and quintiles of the 2019 Index of Multiple Deprivation. We analysed ED attendance trends by multimorbidity groups. Results There were 3 365 279 ED attendances from February 2017 to September 2023. Before the pandemic, there was a statistically significant annual growth rate of 3.4% (rate ratio (RR) 1.034; CI 1.026 to 1.042), with a rising trend in attendance observed in all patient groups. After the pandemic, the overall trend stabilised (RR 1.002; CI 0.993 to 1.009). However, attendances have continued to rise for older age groups (61–75 years and 76+ years) and have increased for patients with multimorbidity and complex multimorbidity. Meanwhile, attendance has declined for the least deprived. For the other patient groups, attendance has plateaued. Conclusion Following the pandemic, total ED attendances stabilised, but have continued to rise for older people, particularly those requiring complex care. This has implications for hospital capacity and places an increased strain on urgent and emergency care. We used high-quality representative population-level linked patient records data. The study was observational with limited causality. Further research should explore specific reasons behind these changes.