Incidence, healthcare-seeking behaviours, antibiotic use and natural history of common infection syndromes in England: results from the Bug Watch community cohort study

Antimicrobial resistance (AMR) is a serious global problem that has resulted in increasing numbers of infections becoming untreatable [1]. There were over 60,000 severe antibiotic resistant infections in England in 2018; approximately 165 new antibiotic resistant infections per day [2]. As AMR is largely driven by selective pressure exerted through antibiotic use [1], ensuring that antibiotics are used responsibly (antibiotic stewardship) is a priority [3]. Antibiotic prescribing in England reduced by 9% from 2014 to 2018, but the number of drug-resistant infections continues to rise [2]. The UK’s national action plan on tackling AMR aims to further reduce prescribing by 15% by 2024 whilst reducing the number of drug-resistant infections [4].

Achieving this aim requires identification of opportunities to safely reduce prescribing. The majority of antibiotic prescribing in England is in general practice, followed by hospitals, other community settings, and dental practice [2]. Most prescriptions made in general practice are for infections of the respiratory tract, urogenital tract, and skin/wounds [5]. It has been estimated that up to 23% of prescriptions in general practice, and over half of those in dental practice, are inappropriate [6, 7]. This includes prescriptions for mild illnesses that would get better on their own and has led to initiatives to reduce inappropriate prescribing in primary care, and efforts to understand levels of knowledge of antibiotic stewardship amongst professionals and the general public [8, 9]. Encouraging patients to manage their symptoms without medical consultation could be an effective complementary approach to reducing antibiotic prescribing.

There is evidence from large-scale community cohort studies in England that most people can safely self-manage respiratory and gastrointestinal infections without visiting a GP: The Flu Watch study estimated that 21% of influenza-like illnesses led to a GP consultation [10], and the IID2 study found approximately one GP consultation for every 15 cases of infectious intestinal disease in the community [11]. This abundance of undetected community cases is referred to as a clinical “iceberg” of infection and has been used as a evidence for promoting safe self-management of these symptoms [12]. However, there is limited information on how people manage symptoms of other common infections as similar large-scale studies have not been conducted.

Information on the usual duration of symptoms has also been used to develop educational resources for antibiotic stewardship campaigns. For example, Public Health England’s Keep Antibiotics Working campaign lists how long most people take to recover from five common illnesses (cough, cold, ear ache, sore throat, and sinusitis) with guidance on what to do if symptoms persist [13]. Establishing these estimates for a wider range of infections requires information on the typical duration of symptoms in the community (i.e., including episodes of symptoms that do not lead to healthcare consultation). Measuring the impacts of infections on quality of life can also inform health economic assessments of potential interventions. Data from the Flu Watch study has been used to generate these estimates for influenza-like illnesses [14] but there is a lack of similar information for other infections.

Bug Watch was a prospective community cohort study in England in 2018–2019 that collected information on symptoms of a range of acute common infections (respiratory, gastrointestinal, mouth/dental, skin/soft tissue, urinary tract, and eye) [15]. We quantified the incidence of these common infection syndromes in the community, determined the proportions of syndromes that led to various healthcare-seeking behaviours and antibiotic use, and estimated the average duration of symptoms and their impacts on health-related quality of life.

We have measured incidence rates, frequencies of healthcare seeking behaviours and antibiotic use, the average duration of symptoms, and the impacts on health-related quality of life for a range of common acute infection syndromes in the community. Using a novel and efficient method of data collection, we have captured reports of over 40,000 symptoms of respiratory, gastrointestinal, mouth/dental, skin/soft tissue, urinary tract, and eye infections across more than 440 person-years of follow up. Most people did not seek medical advice about their symptoms and one in ten infection syndromes led to antibiotic use. A considerable proportion of people went to hospital for their symptoms (particularly for urinary tract infection syndromes) but the NHS111 service was not used often. Symptoms usually resolved within a week with the worst day of symptoms on average after 3 days.

To our knowledge, this is the first study to measure community-level incidence rates across a range of infection syndromes, and there are no comparable estimates from equivalent large-scales studies for most of the types of syndromes covered. The incidence of respiratory tract infection has been estimated in various settings [10, 23‐25] including two large UK cohorts (Flu Survey [23] and Flu Watch [10]). Although these two studies used slightly different case definitions to Bug Watch, and covered winter months rather than the whole year, the estimates were comparable: 3·3 episodes of influenza-like illness per person year (age-standardised) in Flu Survey [23]; 1.7 episodes of any respiratory illness (crude) in Flu Watch [10], and 2·03 syndromes per person year (age and sex-adjusted) in Bug Watch. The IID2 study of infectious intestinal disease in the UK estimated a rate of 0·52 episodes per person year (definite and probable cases) [11], similar to our estimate of 0·68 gastrointestinal syndromes per person year.

We found substantial clinical “icebergs” for all infection types in Bug Watch, with most syndromes not leading to medical consultation. This pattern has been observed previously for respiratory tract [10, 24, 26‐28] and gastrointestinal [11, 29] infections, but contrasts with evidence from a cross-sectional study of urinary tract infections which found that 65% adult females in England contacted a GP about their most recent episode [30]. It is likely that this difference is due to omission of less severe cases from the previous study, which involved a one-off interview and relied on recall rather than prospective symptom reporting. Pharmacies and the internet were the other most frequently used sources of information, but the NHS 111 service was not used often. These findings suggest that interventions encouraging safe self-management, for example provision of better information online and in pharmacies, may be effective across all infection categories. Further work could also explore how to promote NHS 111, or to develop online services with access to advice from nurses or pharmacists, which could reduce use of GPs or Accident and Emergency services for these syndromes.

Although most antibiotics reported in Bug Watch were prescribed by GPs, a substantial proportion were from hospital doctors and dentists (for mouth/dental infections). This suggests that some patients may be bypassing primary care services and highlights the need for continued attention to antibiotic stewardship for acute infections across healthcare settings. Despite evidence that delayed prescriptions for respiratory tract infections can reduce use of antibiotics with no difference in symptoms or patient satisfaction [31], they were used infrequently. Half of the antibiotics reported were prescribed within 3 days of the onset of symptoms. There may therefore be scope for wider use of “safety netting” stewardship interventions that give GPs flexibility not to have to prescribe an antibiotic immediately without compromising patient safety.

We have estimated average symptom duration and impacts on health-related quality of life for a range of infection syndromes in the community. For respiratory tract infections, our estimates were comparable to those from Flu Watch (crude mean 9·8 days duration in Bug Watch; 9·0 days for influenza-like-illness and 6·9 days for other acute respiratory tract infections in Flu Watch; mean 0·7 QALD lost per syndrome in Bug Watch; 0·93 for influenza-like illness and 0·26 for other acute respiratory tract infections in Flu Watch) [14]. Other previous estimates of duration have tended to be based on patients who initially presented at primary care, and are therefore likely to have excluded less severe cases. This includes estimates of lower respiratory tract infection (average duration 3 weeks [32]) and urinary tract infection (mean 5 days of symptoms after GP consultation for infections not treated with antibiotics [33]). To our knowledge there are no other community-based estimates of QALD loss for other infection types. We have also provided data on the frequency and duration of individual symptoms of infection. These estimates will be useful in informing messages about symptom expectation for antibiotic stewardship campaigns and for modelling cost effectiveness of potential interventions. In the context of emerging infections such as COVID-19, knowledge of the background frequency of symptoms characteristic of the new infection (Table S10) could be used to inform public health advice regarding testing and self-isolation. For example, data from Bug Watch have been used to estimate the baseline demand for COVID-19 tests due to background cases of cough and fever [34]. All data from the Bug Watch study are available so that different combinations of symptoms can be assessed in future studies [35].

A limitation of this study was that a relatively low proportion of people from the HSE sample responded to the invitation to participate, and some did not consent to be re-contacted about future research. The results may therefore not be representative, with participants more likely to be older, female, and living in less deprived areas than the general population. To partially account for this, we adjusted incidence rate estimates to the age and sex structure of England and reported healthcare-seeking behaviours separately for adults and children and by sex. Whilst there are likely to be differences in healthcare-seeking behaviours for groups not well represented in this study, our data provides community-level information that is not easily obtained from other study designs. Further work is needed to understand how to engage a more diverse population in future similar studies. All surveys for children were completed by their parents/guardians, and healthcare-seeking behaviours are likely to have different drivers for this group, which we plan to explore through future qualitative interviews [36].

It is also possible that people who were motivated to sign up to participate in Bug Watch tended to have more infections or have a particular interest in antibiotic stewardship. Comparisons with the Wellcome Trust Monitor survey suggested that the Bug Watch cohort had a greater awareness of antibiotic resistance but that it was not reflected in large differences in behaviour in terms of antibiotic use. Once registered, the retention of the cohort was good, with 72% of people who signed up completing at least three quarters of their weekly surveys. This suggests that the online survey methodology was acceptable and not too burdensome, so should be considered for future similar studies. Our measures to maintain engagement in the study included provision of regular newsletters with study updates and small tokens of appreciation (£5 vouchers) for those who completed the majority of surveys.

We combined self-reported symptoms of infection on different days into infection syndromes. This relied on assumptions about the plausible length of time between symptoms that could have been caused by the same infection, and could have led to over- or under- ascertainment of incidence rates. Although we asked participants to only report symptoms that they suspected to be caused by an acute infection, there was no microbiological confirmation and it is likely that some symptoms (for example symptoms due to chronic conditions or allergies like hay fever) were mistakenly reported, leading to overestimation of rates. Similarly, the self-reports of healthcare-seeking behaviours were not corroborated with health records, we do not have information on overnight hospitalisation or death, and reports rely on participants correctly interpreting the survey questions. For example, if a participant initially phoned the NHS 111 service but were subsequently transferred to a telephone GP, we cannot be sure if they would have included both behaviours when completing the survey.

Most people are able to manage symptoms of common infection without medical consultation. The information on consultation and prescribing frequencies and symptom duration will be useful in the design of national and international antibiotic stewardship programs, public health campaigns, and to support GPs in communicating with patients about infections. Interventions encouraging safe self-management across a range of acute infection syndromes could decrease pressure on primary healthcare services and support targets for reducing antibiotic prescribing. Our health-related quality of life estimates could be used to in health economic models of the impacts of interventions. The novel data on symptom duration and frequency that we have captured is also valuable for assessing characteristics of emerging infections such as COVID-19 [34].