Since the late 1990s annual outpatient and emergency department visits for abscesses or cellulitis in the USA have almost doubled [Reference Hersh1]. In England there has been a fivefold increase in hospital admissions for staphylococcal septicaemia, pneumonia and impetigo and a threefold increase for staphylococcal abscesses and cellulitis [Reference Hayward2]. The burden of disease in the community is unknown because there have been no systematic surveys. We investigate whether consultations for boils and abscesses, and repeat episodes have increased in primary care using data from The Health Improvement Network (THIN) database. Further, we explore if there has been a sustained increase in hospital admissions for staphylococcal skin infections using data from Hospital Episode Statistics (HES).

We used individual-level data from THIN and aggregated data from HES. In the UK, 98% of the population is registered with a General Practitioner (GP) who acts as a gatekeeper to specialist services and provides advice, treatment and prescriptions. Practices contributing to THIN enter demographic and clinical data into the clinical records by using Vision software every time a consultation takes place, generating a longitudinal record. Since 1990, symptoms, diagnoses, treatments and referrals have been recorded using a hierarchical system of more than 100 000 Read codes. Prescriptions are recorded linking each formulation to the British National Formulary (www.bnf.org). Social deprivation is estimated by the Townsend score, a composite measure based on levels of unemployment, car ownership, household overcrowding and housing tenure, linked to the patient's postcode. THIN contains the medical records of more than 11 million patients [3], and is broadly representative of the UK population [Reference Blak4]. Prescription and consultation rates in the dataset are comparable to national statistics and external data sources [Reference Lewis5]. Adequacy of death recording is assessed by comparing mortality rates recorded by the practice to national age- and sex-standardized mortality rates [Reference Maguire, Blak and Thompson6]. We included data from when practices met the acceptable mortality recording criteria and were fully computerized.

HES is a national dataset containing information on every admission to a NHS hospital (www.hscic.gov.uk/hes). Admissions are coded by professional coders to identify the primary reason for admission, classified using the 10th edition of the International Classification of Diseases (ICD-10).

In our primary-care study, patients were eligible for inclusion if they were registered with a participating practice between 1 January 1995 and 31 December 2010. Read code lists were developed to identify patients seeking care for a boil or abscess. Boils and abscesses were selected because they are predominantly caused by Staphylococcus aureus and are easily identified in General Practice. To assess time trends, we included the first consultation for a boil or abscess for patients of all ages, so patients were counted once per practice. Patients left the study on the earliest of the following dates: date of consultation for a boil or abscess, date patient left the practice, date of death, date the study ended (31 December 2010). Recurrent infection was defined as between two and five consultations for a boil or abscess within a calendar year.

In our secondary-care study patient admissions were identified based on ICD-10 codes corresponding to abscesses, carbuncles, furuncles and cellulitis [Reference Hayward2]. These conditions were chosen because S. aureus is the main causative organism, or in the case of cellulitis, one of the most important organisms. We obtained an extract of aggregate data on all hospital admissions where the primary diagnosis was one of these codes, including repeat admissions. These data cover all NHS hospital admissions in England for each financial year (1 April to 31 March) from 1998/1999 to 2010/2011. We used annual population denominator information from the Office for National Statistics (www.statistics.gov.uk) to calculate age-specific hospital admission rates.

We used negative binomial regression to estimate changes in the incidence of primary-care consultation per year adjusted for age, sex and social deprivation (Townsend score). We used a random-effects model to account for clustering of observations by practice. The denominator was the total number of person-years contributed by patients in the sample population for each corresponding calendar year.

For the analysis of secondary-care admissions we used indirect age-standardization to allow meaningful comparison of annual rates, taking the population of England in 1998/1999 as the baseline.

The THIN programme of anonymized data provision for researchers was approved in 2002. This study was approved by the THIN scientific review committee, reference 11–504. Ethical approval is not required for analysis of anonymized HES data.

Overall, 164 461 people had at least one primary-care consultation for a boil or abscess between 1995 and 2010, corresponding to 211 243 consultations. There was a slight (5%) increase in the incidence of first consultation for a boil or abscess, from 411 [95% confidence interval (CI) 394–428] in 1995 to 433 (95% CI 426–440) per 100 000 person-years in 2010, although the overlapping confidence indicates this may be a statistical artefact [incidence rate ratio (IRR) 1·005, 95% CI 1·004–1·007]. The rate of repeat consultations increased from 66 (95% CI 59–73) in 1995 to peak at 97 (95% CI 94–101) per 100 000 person-years in 2006 (IRR 1·65, 95% CI 1·48–1·85) and remained stable thereafter (Fig. 1). Hospital admissions for abscesses, carbuncles, furuncles and cellulitis almost doubled, increasing in a linear fashion from 123 admissions per 100 000 in 1998/1999 to 236 admissions per 100 000 in 2010/2011.

Fig. 1. The rate of recurrent infection in patients consulting their General Practitioner for a boil or abscess, 1995–2010.

Although the number of patients consulting General Practice with a boil or abscess did not change substantially between 1995 and 2010, the frequency of repeat consultations in those diagnosed with a boil or abscess has increased. Furthermore, over the last 15 years there has been a sustained rise in hospital admissions in England for staphylococcal skin infections.

Our study suggests increasing hospitalizations cannot be explained by more cases in primary care. Instead staphylococcal strains may have become more severe or more difficult to treat, causing recurrent infection. This could be driven by a combination increased drug resistance, as in the USA where community-onset methicillin-resistant S. aureus (MRSA) is a major public health concern, or by increased virulence, conferred by a toxin such as Panton Valentine leukocidin (PVL), which has been epidemiologically linked to recurrent community-onset staphylococcal boils and abscesses [Reference Dufour7]. Without access to microbiological data we cannot confirm these hypotheses, although the prevalence of community-onset MRSA in the UK is thought to be low [Reference Holmes8].

The strengths of this study is its scale and that it uses data from two different, high-quality datasets. The primary-care dataset is large and nationally representative, with prescription and consultation rates that are comparable to those recorded by national statistics and external data sources. THIN contains the medical records of ~6% of the UK population registered with a GP, and is broadly representative of this population in terms of demographics, major condition prevalence and death rates adjusted for demographics and social deprivation [Reference Blak4]. The HES dataset covers all admissions. The limitations of this study relate to the fact that THIN data are recorded for patient management and not for the purpose of research. Thus, we used boils and abscess as a proxy for staphylococcal infection and recognize that up to a quarter of infections may have been caused by a different pathogen [Reference Issartel9]. Patients consulting for boils and abscesses were identified using diagnostic Read codes but some GPs might prescribe an antibiotic without recording a diagnostic code. Similarly patients with a boil or abscess may attend the Emergency department rather than visit their GP, potentially underestimating the incidence of infection in the community. We used repeat consultation as a crude measure of disease severity and acknowledge this does not differentiate between consultations for a new infection and a prolonged infection.

Our findings suggest the number of patients experiencing severe and recurrent staphylococcal skin disease in the community has increased, with a rise in the rate of repeat consultations in primary care and increased hospital admissions. By contrast, there is no evidence for an increase in primary-care consultations for uncomplicated boils and abscesses. The mainstay of treatment for recurrent infections is decolonization therapy using topical antibiotics such as mupirocin and chlorhexidine antiseptic. Given the paucity of data to support decolonization therapy in the community [Reference Fritz10], more work is needed to identify better treatments for recurrent staphylococcal skin infections in primary care.