Describing the indirect impact of COVID-19 on healthcare utilisation using syndromic surveillance systems

The coronavirus pandemic (COVID-19) is currently causing an unprecedented global health crisis with significant impact on health and social care services and society worldwide. As of 21st June 2021, there have been over 178 million cases and 3.8 million deaths from SARS-CoV-2 globally [1]. The first cases in England were identified at the end of January 2020 [2, 3]. Widespread transmission in the UK led to a rapid increase in incidence during March reaching an initial peak of new cases in England in April [4]. During 2020, the daily number of deaths in England was at its highest at just over 1000 deaths in week 15 (6-12th April 2020).

The UK government’s COVID-19 public health information campaign was launched in early February encouraging individuals to increase handwashing and improve respiratory hygiene to ‘catch’ coughs and sneezes [5]. On the 12th March, the government announced a move from ‘contain’ to the ‘delay’ phase [6]. The announcement advised the public to use a new online version of the National Health Service (NHS) remote telephone health advice line; NHS 111 [7]. Primary care (general practitioner; GP) services in England started to conduct remote consultations in the first instance [8]. Stepwise restrictions on social mixing and daily life were introduced over the subsequent two weeks until the 23rd March when the first national ‘lockdown’ tightened restrictions further (Fig. 1). Ambulance services and emergency departments (EDs) continued to operate as usual throughout; however, significant changes in healthcare seeking behaviour were observed, potentially due to changes in need for urgent healthcare, possible fears of contracting the infection in healthcare settings, and a desire to ‘protect the NHS’ [10].

Public Health England (PHE; please note - on 1 October 2021 PHE was replaced by UK Health Security Agency and the Office of Health Improvement and Disparities) publishes weekly, publicly available, surveillance reports for the monitoring of the COVID-19 pandemic in England [11]. The COVID-19 surveillance activities reported are carried out across multiple data sources including laboratory confirmed cases; community outbreaks; internet surveys and search activity; hospitalisations; and mortality. These surveillance activities are further supplemented by syndromic surveillance of healthcare utilisation across five different areas of the NHS: NHS 111 (calls and online assessments); GP consultations, both In Hours (scheduled) and Out-of-Hours (unscheduled); ambulance calls; and ED attendances.

‘Syndromic’ surveillance is “the process of collecting, analysing and interpreting health-related data to provide an early warning of public health threats which require public health action” [12]. It is powerful at detecting changes in relatively non-specific symptoms or preliminary diagnosis information collected during routine healthcare provision. This can offer the near real-time ability to provide early warning, improve situational awareness, and monitor the emergence and spread of common infectious diseases and the public health impact of non-infectious diseases through the population. The important role of syndromic surveillance has been demonstrated during both major emergencies and mass gathering events, e.g. the 2009 global influenza pandemic [13, 14], during periods of extreme hot [15, 16] and cold [17] weather, and events such as the London 2012 Olympic and Paralympic Games [18]. COVID-19 has generated several new challenges for the syndromic surveillance systems, including: the impact of media reporting and social distancing measures on healthcare-seeking behaviour, changes in healthcare system delivery, and an increased demand for rapid analysis and dissemination of findings [19].

This study used PHE syndromic surveillance data from Wednesday 1st January 2020 to Sunday 24th May 2020, and Tuesday 1st January 2019 to Friday 24th May 2019 (one day fewer due to 2020 being a leap year). These figures represent the data routinely available through the syndromic surveillance systems, but for some systems this is not necessarily representative of the whole of England (Table 1). The total number of NHS 111 calls was 5,630,117 for this time period in 2019, and 6,175,249 in 2020, an increase of 9.68%. The total number of GP Out-of-Hours contacts was 3,803,553 in 2019, and 3,665,455 in 2020, a decrease of 3.63%. The number of GP Out-of-Hours contacts assigned a Read code was 1,634,938 in 2019 (43% of contacts), and 1,331,478 in 2020 (36% of contacts). The daily mean GPIH registered practice population was 23,998,209 in 2019, and 39,455,662 in 2020, an increase of 64.41%. The number of ambulance calls was 2,100,527 in 2019, and 2,105,082 in 2020, a slight increase of 0.22%. The total number of attendances at the 100 emergency departments included in the study was 3,909,796 in 2019, and 2,997,082 in 2020, a decrease of 23.34%. The number of ED attendances with a diagnosis was 3,138,470 in 2019 (80%), and 2,491,255 in 2020 (83%).

Daily trends in total activity across all five syndromic surveillance systems showed significant variation during the COVID-19 pandemic in England compared to equivalent months in 2019 (Fig. 2). The daily mean percentage changes in activity during the peak pandemic period in 2020 (Weeks 13 to 22) compared to the same period in 2019 are shown in Table 1. These showed the least difference for NHS 111 calls (12% increase; 95%CI: 9 to 15%) and GPOOH consultations (11% decrease; 95%CI: − 8 to − 14%). The total number of ambulance calls varied markedly throughout the pandemic period, increasing during the introduction of social distancing, before reducing during lockdown, such that the mean daily number during the peak pandemic period hides the marked variation seen in Fig. 2. The total number of attendances at ED nearly halved during the peak pandemic (daily mean percentage change: -49% (95%CI: − 45% to − 52%). (Trends in healthcare usage during the peak pandemic period in 2020 compared to 2019, stratified by weekdays, and weekend and bank holidays, for each syndromic surveillance system and indicators are available in Additional Table 1.)

Eye conditions, monitored through GPIH, GPOOH, and NHS 111, showed a sharp decline in mid-February (earlier than other syndromic indicators). GPIH conjunctivitis remained low throughout the 2020 peak pandemic period (64% reduction; 95%CI: − 63% to − 65%, compared to 2019 during peak pandemic period) as did GPOOH eye irritation (71% reduction; 95%CI: − 68% to − 74%). In contrast, though the number of calls to NHS 111 for eye problems initially reduced during March, call numbers returned to baseline (2019) by May, resulting in a smaller (25%; 95%CI: − 20% to − 31%) reduction across the entire peak pandemic period (Fig. 3).

GP In Hours mumps consultations showed a dramatic decline after the introduction of NPIs, following high activity early in 2020. During the peak pandemic period, the number of GP In Hours consultations for mumps showed a 31% (95%CI: − 11% to − 54%) reduction compared to 2019 (Fig. 4).

Fractures presenting to EDs showed a sharp decline from the 12th March 2020 when COVID-19 restrictions were introduced, before starting to increase during May (Fig. 5a). During the peak pandemic period, there was a 49% (95%CI: − 46% to − 52%) reduction in attendance at ED for fractures compared to the same period in 2019. In 2019, weeks 13 to 22, the mean number of females attending ED with a fracture was highest on Monday mornings and for males there were peaks on both Sunday late morning and Monday morning. The trends of ED attendances by hour of day and day of week for the peak pandemic period in 2020 demonstrated that there was no longer any weekend effect, resulting in similar patterns of attendance across each day of the week for both males and females (Fig. 5b, c).

GP In Hours consultations for herpes zoster showed a decline following the interventions in March and an overall 27% reduction (95%CI: − 23% to − 31%) in consultations during the peak pandemic period compared to 2019 (Fig. 4).

ED attendances for myocardial ischaemia also showed a marked decline following the interventions in March, with a 35% reduction (95%CI: − 31% to − 39%) in attendance for the peak pandemic period in 2020 compared to 2019 (Fig. 6). Opposing trends were observed for chest pain in ambulance calls and chest pain/myocardial infarction in GPOOH.

There were significant changes to healthcare utilisation in the first few months of the pandemic response in England. Total numbers of contacts with healthcare services in England, as demonstrated through the NHS 111, Ambulance, and both GP syndromic surveillance systems, increased in early March. A marked decline in activity, compared to 2019, was then observed across all syndromic systems (including ED) around the time of the introduction of COVID-19 NPIs and restrictions to activities in mid to late March. Similar rapid reductions in ED visits have also been reported in the USA, where ED visits declined by 42% during the early COVID-19 pandemic period, particularly among children and females, and in Thailand where ED visits overall declined by 36% [31, 32]. A systematic review summarising the impact of COVID-19 pandemic on healthcare utilisation from 81 studies across 20 countries published before 10 August 2020 found a median 37% reduction in services overall compared to pre-pandemic trends [33].

Healthcare contacts for different conditions displayed different trends during the peak pandemic period. Eye conditions showed a decline in activity earlier than other indicators, which may be explained by healthcare utilisation changing most significantly for milder conditions, or this may reflect a true decrease in incidence due to changes in hygiene behaviour reducing the spread of infectious conjunctivitis [23]. The persistence of low activity of eye conditions may be due to the closure of schools and nurseries during the peak pandemic period. Improvements in hygiene behaviour, combined with social distancing and in particular the closure of universities in mid-March, may also explain the dramatic reduction in mumps consultations in GPIH, bringing a rapid end to the national mumps outbreak of 2019/2020 that was concentrated particularly in university student populations [24].

The trends in attendances at ED for fractures, a high acuity condition that is more likely to represent true incidence, showed overall reduced numbers during the peak pandemic study period. Other data sources for incidence of trauma during COVID-19 show the same trend: referrals to specialist spinal surgery decreased significantly compared to the same period in 2019, most apparent for the number of high-energy traumatic presentations, which declined by 72% (p < 0.002) [34]. This is likely because of significant changes to daily activities, including the closing of schools and less participation in team sports, during the peak pandemic period. Data from the UK Government Department for Transport show mean car traffic use during the peak pandemic (weeks 13 to 22) was reduced by 61% compared to the baseline first week of February 2020 [35] likely resulting in fewer road traffic collisions. The trends in attendances for fractures show a loss of the weekend effect which may be partly due to a reduction in team sports and a reduction in alcohol-related attendances, as a consequence of the closure of the service industry, which has previously been shown to account for a peak of 72% of attendances in males in early hours at weekends [36]. The significant decline in ED attendance due to myocardial ischaemia during the COVID-19 pandemic was of concern to clinicians especially given the association between acute infection and myocardial infarction [37]. Further analysis of ED data showed the reduction in attendance was greatest at lower levels of acuity, and yet level 1 acuity ‘immediate’ attendances still reduced by 31% during COVID-19 [10]. This is in keeping with other reports on the reduction in the number of stroke patients attending, where the greatest decline was among mild severity stroke [38] and systematic review finding larger reductions in utilisation among people with milder spectrum of illness [33]. These dramatic declines in attendances at ED prompted public health messaging urging patients to continue to seek medical care as required [39].

A strength of this study is its use of the breadth of data across five national syndromic surveillance systems to demonstrate changes in healthcare utilisation across several healthcare systems and for a wide range of mild to severe conditions in England over time. This routine, real-time monitoring of syndromic indicators of disease provides a greater understanding of trends in healthcare utilisation, particularly for infectious diseases [13, 40]. The ability of ED data to add insight into changing trends in attendances by hour of day and day of week provides an added benefit of working with this syndromic data. This has previously been used to analyse the impact of sporting events on the timing of ED attendances [18, 41].

One limitation of the study is that some changes in indicator trends may result from modifications that have been made to the underlying health care systems to improve patient management rather than actual changes in activity. However, experience and detailed knowledge of the surveillance system and data by those undertaking the surveillance can help to minimise these limitations. For example, opposing trends in chest pain syndromes were observed throughout the first lockdown, however, the marked reduction in chest pain in the Ambulance system followed the introduction of a ‘pandemic’ chief complaint pathway (the basis of a newly defined COVID-19-like syndromic indicator) in early March. This suggested that the earlier peak in chest pain may have been detecting increased incidence of COVID-19 and/or anxiety, rather than cardiac-related chest pain. A COVID-19-like indicator was not introduced in the GPOOH system and therefore the chest pain/myocardial infarction indicator may have continued to capture some COVID-19 activity (albeit a small increase in percentage of Read code consultations). This demonstrates the importance of detailed interpretation from experts who understand the data’s strengths and limitations. Additionally, it is difficult to disentangle behavioural changes in healthcare usage and changes in disease incidence. This was in part minimised by exploring a range of different surveillance systems and indicators, for both mild and more severe health conditions.

This study improves our understanding of how healthcare usage changed during the first national lockdown, which began on the 23rd March 2020, both in total and for a range of syndromic indicators. This understanding is key when interpreting syndromic surveillance data and considering healthcare utilisation during the second and third national lockdowns (which were introduced in November 2020 and January 2021 respectively), when the NHS continued much of its routine healthcare provision and public health messaging about the importance of attending healthcare when needed was generally better emphasised [42]. Future work could describe how trends in syndromic indicators changed throughout summer 2020 when fewer restrictions were in place. Future sub-analyses could include exploring the trends in attendances for fractures by gender and age-groups and myocardial ischaemia by severity of illness markers. Importantly, our increased knowledge of how trends in syndromic surveillance are changing throughout the COVID-19 pandemic will be important when interpreting trends in other conditions, for example the monitoring of influenza throughout the seasonally expected peak influenza activity season, against a background of changing healthcare usage, behaviours, and incidence of other conditions.

This descriptive analysis of syndromic surveillance system data demonstrates the significant impact of COVID-19 on England’s health, National Health Service and society. The restrictions introduced to contain and delay the spread of COVID-19 and to ease pressure on the health service resulted in a combination of a likely true decrease in incidence and a change in healthcare seeking behaviour for certain conditions. Syndromic surveillance systems have a valued and important role in describing and interpreting these trends.