Changes in the utilisation of acute hospital care in Ireland during the first wave of the COVID-19 pandemic in 2020

Study design and setting

A retrospective, population-based, observational study was conducted to quantify and characterise acute hospital service utilisation events in Ireland and to compare these events across different time periods with reference to the epidemiology and public health management of COVID-19. Emergency Department (ED) presentations, admissions from ED and non-COVID-19 in-patient admissions to Health Service Executive (HSE) acute hospitals over a 26-week period from 6^th January 2020 to 5^th July 2020 were identified, analysed and compared with those observed over defined reference periods.

Data sources

Patient Experience Time (PET). National data ED attendances were obtained from the Patient Experience Time (PET) dataset which is an administrative dataset that contains observations of individual-level ED utilisation across 30 HSE-operated or funded hospitals¹². PET contains information on age, sex, discharge destination, mode of arrival and referral and triage status. Clinical information is not reported and therefore patients with and without COVID-19 were included in the data used for this study. PET data does not include Minor Injury Units (MIU), private EDs, specialist EDs or direct attendance at acute assessment units.

Hospital In-Patient Enquiry (HIPE). National acute hospitals discharge data were accessed from the Hospital In-Patient Enquiry (HIPE) data via the Health Intelligence Unit (HIU) Health Atlas Ireland Analyser. HIPE is managed by the Healthcare Pricing Office (HPO) and is a well-established, quality-assured health information system that is the primary source of episode-based, aggregate clinical, demographic and administrative data on discharges from acute public hospitals in Ireland¹³. It contains information on age, sex, area of residence, admission type, date of admission and discharge along with principal diagnosis coded using the International Classification of Diseases Tenth Revision (ICD-10)^14,15. It is used nationally to inform healthcare planning, management and activity-based funding¹⁶.

Variables

Exposure. The exposure was to COVID-19 and the associated public health restrictions and wider socioeconomic changes within 2020. The study period was divided into four sub-periods (Table 1). These time periods reflect levels of exposure based on the a priori knowledge of the epidemiology of COVID-19 during the first wave, and of the public health measures implemented. Period 1 was defined as prior to the beginning of the first wave, Periods 2 and 3 were periods where progressive public health restrictions were implemented and Period 4 commenced with the easing of public health restrictions. Period 1 was defined as starting on the first Monday of January for the study period and reference periods, which were divided into the same sub-periods. The historic reference period for the PET data was a 26-week time period beginning on the first Monday of January 2019. This dataset has increased in completeness year-on-year so restricting the reference period to 2019 allowed meaningful comparison. The historic reference period for the HIPE data was a 26-week time period beginning on the first Monday in January for 2017–2019. It was assumed that while there might be a slight variation year on year, the three-year average of hospital admissions would provide meaningful comparison¹³. For analysis within 2020, the reference period was Period 1 which was prior to the beginning of the first COVID-19 wave. To compare population rates of healthcare utilisation between pre- and post-COVID-19 time periods, two reference periods were used; the historic reference periods and Period 1 2020. The results reported in this paper primarily focus on the comparison within 2020 using Period 1 2020 as a reference period.

Outcomes. The outcomes were presentation to and admission from ED as recorded on PET and an acute hospital admission of any type for a non-COVID-19 illness. A non-COVID-19 hospital admission was defined as a hospital discharge (including death) recorded on HIPE where the diagnosis was a non-COVID-19 illness. Patients recorded with an ICD-10 diagnostic code for COVID-19 (U071 OR U072 OR B342 OR B972) were excluded for this purpose. The occurrence and characteristics of the outcomes were compared between exposure and reference periods. In order to describe stratified rates of each outcome and the characteristics of the population who experienced outcomes for the exposure and reference periods, relevant variables were included from the PET and HIPE datasets (Table 2). To explore trends further for selected clinical conditions to inform and aid recovery planning, ‘tracer diagnoses’ were chosen from within HIPE using defined ICD-10 codes (Table 3). These conditions were chosen following a review of the literature and from discussions with the HSE Lead for Integrated Care, for the Acute Hospitals and for Mental Health all of whom were providing frontline clinical care. The purpose of selecting the ‘tracer diagnoses’ was to explore healthcare utilisation trends in key clinical areas where changes in healthcare utilisation had been observed. The rationale for the selection of the ‘tracer diagnoses’ is outlined in Table 3.

Data analysis

Using Census 2016 data as the denominator¹⁷, overall, age-specific and gender-specific population rates for each outcome were calculated with 95% Confidence Intervals (CI) for weekly counts across the 26-week study period and total and average weekly counts across the defined sub-periods. Rate differences with 95% CIs, and rate ratios with 95% CIs were used to compare the occurrence of the outcome across exposure and reference periods. A chi-squared test was used to test the hypothesis that there was no difference between the proportion of the population who experienced an outcome across exposure and reference periods. The characteristics of those who experienced an outcome were compared between exposure period and reference periods using a chi-squared test to investigate the null hypothesis that there was no difference in the characteristics across exposure and reference periods. Using PET data, the effect of patient-level characteristics, including time period, were compared for association with the likelihood of admission from ED using a binary logistic regression model. Adjusted Odds Ratios (AOR) were calculated to measure the independent likelihood of admission from ED for a specific level of a characteristic relative to the reference level within the model. The purpose of the regression analysis was to assess if presentation to ED within the specific study time periods was associated with an increased likelihood of admission from ED as admission following ED presentation is an indicator of acuity²⁷. The multiple logistic regression analysis could only be conducted using PET data as HIPE data are aggregate data and therefore regression analysis was not suitable. Within HIPE, initial data analysis was for all admission types, which was followed by further sub-group analysis of rates of elective and emergency non-COVID-19 hospital admissions by diagnostic group (CCS-IM)²⁸. All statistical analysis was carried out using Microsoft Excel, SPSS version 26.0, Stata 15 (Stata Corporation) and Open-Source Epidemiologic Statistics for Public Health version 3.01. Level of significance for all group differences in this study was set at 5% (p-value <0.05).

Reporting guideline

The Reporting of Studies Conducted using Observational Routinely-collected Data (RECORD) guideline extended from the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement on reporting guidelines was used for this study.

Ethical approval

Ethical review was not required for this study as the research is secondary analysis of anonymised data sets. The data used in the study are controlled by the HSE in Ireland. The study authors (LM and PK) conducted data processing for the study at the HSE National Health Intelligence Unit to inform the statutory function of the HSE in Ireland to improve, promote and protect the health and welfare of the public health²⁹. HIPE data are anonymised for users and usual practices regarding statistical disclosure control were applied.

Overall trends - total population, gender and age group

There was a substantial reduction in population rates of ED presentation and admission from ED in Periods 2–4 2020 compared to the historic reference period in 2019 (Figure 1 and Figure 2). Similarly, there were reductions in non-COVID-19 admissions of all types compared to historic reference periods from 2017–2019. As the reductions in non-COVID-19 hospital admissions were predominantly for elective and emergency admissions and as there were notable differences in the patterns and trends observed between elective and emergency admissions, these are presented separately in Figure 3–Figure 5 and Figure 7. Trends in elective and emergency non COVID-19 hospital admission are shown in Figure 3 and Figure 4. Figure 5 outlines the results of the internal comparison within 2020 and shows reductions in Periods 2–4 2020 compared to Period 1 2020. Figure 8 shows reductions in non-COVID-19 hospital admissions for all admission types.

Figure 1. Weekly rate of ED presentation per 100,000 population week 1–26 2019 vs. 2020.

Figure 2. Weekly rate of admission from ED per 100,000 population week 1–26 2019 vs. 2020.

Figure 3. Weekly rate of non-COVID-19 emergency admission per 100,000 population 2017–19 vs. 2020.

Figure 4. Weekly rate of non-COVID-19 elective admission per 100,000 population 2017–2019 vs 2020.

Figure 5. Rate ratios of average weekly ED presentation, admission from ED and non-COVID-19 acute hospital elective and emergency admission 2020.

The greatest absolute and relative rate reductions were seen in Period 3 2020 compared to Period 1 2020 for the following:

-   ED presentation (Rate Difference -142.1 per 100,000 population, 95% CI, -150.4, -133.6, p-value <0.0001 and Rate Ratio 0.72, 95% CI 0.71, 0.73, p-value <0.0001)

-   Admission from ED (Rate Difference -33.2 per 100,00 population, 95% CI, -37.5, -28.9, p-value <0.0001, Rate Ratio 0.75, 95% CI 0.72, 0.78, p-value <0.0001)

-   Overall non-COVID-19 acute hospital admission (Rate Difference -329.5 per 100,00 population, 95% CI, -338.8, -320.2, p-value <0.0001 and Rate Ratio 0.53, 95% CI 0.52, 0.54, p-value <0.0001)

-   Non-COVID-19 emergency hospital admission (Rate Difference -50.0 per 100,00 population, 95% CI, -54.9, -45.2, p-value <0.0001 and Rate Ratio 0.71, 95% CI 0.69, 0.73, p-value <0.0001)

-   Non-COVID-19 elective hospital admission (Rate Difference -270.2 per 100,00 population, 95% CI, -277.6, -262.9, p-value <0.0001 and Rate Ratio, 0.43 95% CI, 0.42, 0.44, p-value <0.0001)

Similar reductions were observed for both genders and across all age groups. The greatest relative rate reductions were in younger age groups (<45 years) while the greatest absolute rate reductions were seen in older age groups, particularly those aged over 80 years (Figure 6 and Figure 7).

Figure 6. Rate ratio of average weekly ED presentation and admission from ED and non-COVID-19 hospital admission by age 2020.

Figure 7. Rate ratio of average weekly non-COVID-19 elective and emergency admission by age 2020.

Figure 8. Rate ratios of average weekly non-COVID-19 admission by admission type 2020.

Within Periods 2–4 there were 81,712 fewer ED presentations (-18.8%), 19,692 fewer admissions from ED (-17.4%) and 210,357 fewer non-COVID-19 hospital admissions (-35.0%) than what would have been expected based on Period 1 2020. This included 173,688 fewer elective care admissions (-42.8%) and 30,899 fewer non-COVID-19 emergency care admissions (21.0%) (Supplementary Tables A1–A3).

Trends in emergency department presentations and admissions

Analysis of ED activity indicated that a greater proportion of those presenting to ED in Periods 2–4 2020 were admitted and a greater proportion of both presentations and admissions were from older age groups (p-value <0.0001), had arrived by ambulance (p-value <0.0001) and were admitted (p-value <0.0001) compared to Period 1 2020 (Table 4 and Table 5). Factors associated with conversion to hospital admission following presentation to ED were examined for their independent association through a logistic regression model (Table 6). Being in a higher triage category (AOR 8.88, 95% CI 8.64, 9.13, p-value <0.0001), followed by older age (AOR 5.00, 95% CI 4.84, 5.17, p-value <0.0001) were the greatest predictors of hospital admission from ED. Independent of other factors included in the model, there was an increased likelihood of being admitted to hospital following ED presentation in Periods 2–4 compared to Period 1, which was most marked in Period 3 (AOR, 1.10, 95% CI, 1.07, 1.12, p-value <0.0001). In the recovery period (Period 4), ED presentations and admissions from ED returned to pre-COVID-19 levels for those aged over 45 years but remained reduced for those aged below 45 years (Figure 6).

Trends in non-COVID-19 hospital admissions

Analysis of non-COVID-19 hospital admissions using HIPE data found reductions across all diagnostic groups and all admission types including elective, emergency, maternity and newborn admissions (Figure 8).

Trends in elective and emergency admissions for selected diagnostic groups are shown in Table 7. Comparing elective admissions in Periods 2–4 2020 to what would have been expected based on Period 1 2020, there were particularly large reductions in cancer (36,120 fewer episodes of admission, -33.8%), gastroenterology (26,895 fewer episodes of admission, -56.1%), dermatology (12,180 fewer episodes of admission, -66.8%), respiratory (8,021 fewer episodes of admission, -65.8%) and cardiovascular (6,637 fewer episodes of admission, -58.5%) admissions. Further analysis of emergency admissions with specific selected ‘tracer diagnoses’ showed reductions in admissions with stroke and transient ischaemic attack (TIA) (411 fewer episodes of admission, -12.1%) and acute myocardial infarction (AMI) (395 fewer episodes of admission, -14.7%) in Periods 2–4 2020 compared to expected based on Period 1 2020. There were also reductions in emergency admission with injury (2,059 fewer episodes of admission, -21.4%) and post-road traffic accident (RTA) (182 fewer episodes of admission, -24.4%) in Periods 2 and 3 2020 compared to expected based on Period 1 (Table 8).

For all non-COVID-19 hospital admission types there was a small overall increase in in-hospital mortality in Period 3 compared to Period 1 (0.9% vs. 0.6%, p-value 0.004) and a higher proportion of patients discharged in Periods 2–4 had a Charlson co-morbidity index (CCI)³⁰ score of over 10 compared to Period 1 (19.9% vs. 13.5%, p-value <0.0001). Patients experiencing an emergency admission are generally more acutely unwell compared to other admission types. These observed differences in outcomes were no longer statistically significant when analysis was limited to emergency admissions only. In-hospital mortality for emergency admissions was 2.5% in Period 3 vs. 2.4% in Period 1 (p-value 0.888), while the proportion of those with a CCI score of over 10 was 12.2% vs. 11.7% (p-value 0.627) (Table 9) Analysis of specific emergency tracer diagnoses also showed no difference in severity as measured with CCI and in-hospital mortality.

In the recovery period (Period 4), HIPE analysis found rates of non-COVID-19 hospital admission remained below expected levels for all age groups compared to Period 1 (Rate Ratio 0.72, 95% CI 0.71, 0.73, p-value <0.0001) (Figure 7). There was less recovery for elective admissions (Rate Ratio 0.63, 95% CI 0.62, 0.64, p-value <0.0001) compared to emergency admissions (Rate Ratio 0.90, 95% CI 0.87, 0.93, p-value <0.0001). During the recovery period there were increases in emergency mental health admissions (Rate Ratio 1.28, 95% CI 1.03, 1.60, p-value 0.028) (Table 7), emergency alcohol-related admissions (Rate Ratio 1.22, 95% CI 1.03, 1.43, p-value 0.016) and emergency admissions with self-harm (Rate Ratio 1.39, 95% CI 1.01, 1.91, p-value 0.043) (Table 8).

Summary of key findings

This study reports on the changes in healthcare utilisation in acute hospitals in Ireland during the first wave of the COVID-19 pandemic in 2020. There was reduced healthcare utilisation for elective and emergency acute public hospital care. This reduction began in early March 2020, following the beginning of the first wave (Period 2), and overall persisted for the duration of this study which included the recovery period up to 5^th July 2020 (Period 4). During the recovery period, population rates of elective non-COVID-19 care did not recover. In contrast, there was greater recovery of emergency healthcare utilisation rates, however, activity still remained below pre COIVD-19 levels, particularly among younger age groups. In particular, this study finds evidence of increased emergency alcohol and emergency mental health related admissions in the recovery period (Period 4) which began with phase 1 of reopening of society on 18^th May 2020 until 5^th July 2020.

Those who presented to ED during the first wave of the COVID-19 pandemic had an increased likelihood of admission which may suggest increased severity of illness²⁷. However, there is no evidence of an immediate increase in in-hospital mortality or an increase in co-morbidity on discharge. The full consequences of the impact of changes due to delayed or missed care on population health may only become apparent over time.

Comparison with other studies

The findings of this study are consistent with other published reports and literature describing disruption to healthcare services during the first wave of the COVID-19 pandemic in multiple countries^5,31–33. In particular, reductions in ED presentations and hospital admissions which persisted following the easing of restrictions are reported^32–34. While a proportion of these reductions were likely due to decreased incidence of certain conditions related to population-level restrictions, some necessary care was not accessed for acute medical emergencies (e.g., stroke and AMI)^3–5,32,35. The greatest reductions in presentations were reported among vulnerable groups such as lower socioeconomic groups, those at extremes of age and ethnic minorities^5,36.

This study found no evidence of immediate harm related to delayed or lost presentations. This is in contrast to other studies, which reported evidence of increased morbidity and mortality associated with changes in healthcare utilisation^4,5,37. This impact may only become fully apparent over time and through examination of wider health information datasets.

Reasons for changes in acute hospital utilisation

The reasons for the changes in healthcare utilisation during the study period are likely multifactorial. The COVID-19 pandemic highlighted well-established weaknesses in the Irish health system that pre-date the COVID-19 pandemic. These include the absence of universal healthcare, acute hospital capacity deficits and a service configuration with overreliance on the acute hospital system to provide scheduled as well as unscheduled care due to poor orientation to primary and community care^{7,10,11,38,39}. In order to create capacity to manage acute COVID-19 and non-COVID-19 illness, it was necessary to postpone elective care in the acute hospitals³⁸. While some time-critical elective care was diverted to private hospitals⁴⁰, analysis of clinical patterns of elective admissions in this study suggest that there are large backlogs in care. In the recovery period following the first wave of the COVID-19 pandemic in 2020, ongoing capacity restrictions in healthcare settings and the need to provide care for those with COVID-19 infection meant that it was not possible to resume elective activity at pre-COVID-19 levels or to provide the level of services required to fully address backlogs in elective care^41,42. For emergency care, the reduction may have been due to reduced incidence of some medical conditions, e.g., injuries and non-COVID-19 infections, due to population-level restrictions and/or due to a reduction in unnecessary emergency attendances^{26,35,43–48}. However, the scale of the reductions shown in this study and reduction in presentations for conditions such as stroke/TIA and AMI which are non-discretionary and time-sensitive suggest that necessary care was avoided or delayed. This may have been due to a fear of exposure to COVID-19 in hospital^49,50. Increased utilisation of acute health services for emergency alcohol and self-harm admissions in the recovery period in wave 1 suggests that the pandemic, and associated restrictions, are negatively impacting population health and wellbeing. This finding is consistent with published data reporting increased mental distress and increased utilisation of secondary mental health services due to the COVID-19 pandemic^51–54. This burden of unmet need is likely greatest among vulnerable groups most affected by COVID-19 such as those living in poverty, ethnic minority groups and older people^2,24,55–57.

Implications for health policy and health system reform in Ireland

Harnessing the COVID-19 shock to manifest health system change. COVID-19 is a shock to the health system⁵⁸. However, despite the challenges, the system has responded and shown innovation and flexibility in work practices and delivery of services, which demonstrate capacity and readiness to reform^11,38. Lessons must be learned from COVID-19 to build health system resilience and increase preparedness for the future, including future pandemic preparedness^58,59. In the long-term, further strategic reform aligned with Sláintecare should be progressed building on this innovation and change capability shown during the COVID-19 pandemic. Internationally, there have been calls to ‘build back better’ and also to ‘build back fairer’ to achieve sustainable, resilient health systems and deliver universal healthcare. Such an endeavour will require political leadership, human and financial resources and investment in information technology (IT) infrastructure and public health expertise^11,60–64. The findings of this study were disseminated nationally to the director of Sláintecare and to the national leads for Integrated Care, the Acute Hospitals and Mental Health. This study quantified the changes in healthcare utilisation during the first wave of the COVID-19 pandemic and identified key clinical areas to focus on for population health recovery. The findings are important in the context of the ongoing reform of the Irish health system and the findings of this study informed the HSE National Service Plan for 2022 which has a focus on scheduled care recovery⁶⁵.

Public health should be core to health reform. Public health has been frontline in confronting initial waves of COVID-19 in Ireland. With the development and arrival of the COVID-19 vaccination in 2021, Ireland has entered a new phase of the COVID-19 pandemic. However, COVID-19 and its associated consequences will continue to impact population health and the health system for many years. Therefore, strong public health leadership and advocacy are required to seize the opportunity to control COVID-19 infection, to guide population health recovery from COVID-19 and to progress health system reform in Ireland

Limitations of this study

Due to the data available at the time of analysis, this study focuses only on the first wave of COVID-19. While the patterns observed in this study may predict healthcare utilisation in subsequent waves, there are likely differences as some lessons learned from the first wave may have been acted on. PET and HIPE datasets do not allow for identification of repeat episodes of care which may overestimate population rates of healthcare utilisation. However, such an overestimate is likely to be minimal due to the large size of the datasets. During the first wave of the COVID-19 pandemic, some time-sensitive elective care was provided in the private hospitals, these data were not available for this study. Therefore, the reduction in elective hospital activity may be overestimated. Data on GP utilisation were not analysed in this study, changes in provision of GP care may explain some of the changes reported. Hospital outpatient department (OPD) activity was also not examined, this may underestimate need for services as there are backlogs for OPD appointments. PET does not contain clinical information therefore the impact on non-COVID-19 care was not quantified. HIPE reports data on patients discharged from acute hospitals. Therefore, patients who remained in hospital at the end of the study period are not included in this study. As those who are more unwell may have longer admissions with poorer outcomes, co-morbidity and in-hospital mortality may have been underestimated.

Underlying data

Open Science Framework. The public health and health system implications of changes in the utilisation of acute hospital care in Ireland during the first wave of COVID-19: Lessons for recovery planning. DOI: https://doi.org/10.17605/OSF.IO/D56SZ⁶⁶

This project contains the following underlying data:

-   The public health and health system implications of changes in the utilisation of acute hospital care_Supplementary Tables.pdf

-   The public health and health system implications of changes in the utilisation of acute hospital care_RECORD Checklist.pdf

Data are available under the terms of the Creative Commons Attribution 4.0 International license

The datasets processed for this study were derived from special categories of personal data concerning health. The datasets are controlled by the HSE, not the authors, and so the authors cannot determine requests for data access. Further information on HSE data protection policy can be located at hse.ie/eng/gdpr/hse-data-protection-policy/. Reasonable requests to access the two datasets used in this study, HIPE and PET, can be directed to the data controller by contacting the HSE Healthcare Pricing Office (https://hpo.ie/) in the case of HIPE and to the HSE Special Delivery Unit https://www.hse.ie/eng/about/who/acute-hospitals-division/special-delivery-unit/ in the case of PET.

Reporting guideline

OSF registries. RECORD guideline checklist, extended from the STROBE statement. DOI: https://doi.org/10.17605/OSF.IO/D56SZ⁶⁶

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).