This study provides an in-depth description of Lyme disease patients seen in English and Welsh hospitals, and addresses some of the NICE guidelines calls for new epidemiological data . Incidence rose over the study period, showing a similar trend, but at lower levels, compared to officially published figures based on laboratory confirmed cases [2,3,4]. This discrepancy is to be expected, as national laboratories will receive samples from both hospital and primary care patients, and will therefore have a higher incidence. Not all cases would need to be referred to a hospital clinician from primary care, as the majority of cases are likely to present with an uncomplicated erythema migrans rash . The cause for the increase in incidence is unknown, but may be the result of, among other causes; increased awareness by the public and/or hospital clinicians, increase in referrals by clinicians in primary care, or a true increase in incidence within England and Wales. Further research is needed to understand the drivers for this increase in incidence. Compared to other European countries the incidence we describe is lower. In France the annual hospitalisation rate due to Lyme diseases is 1.55 cases per 100,000 , with an estimated average national incidence of 42 cases per 100,000 population. Whereas in Germany the inpatient incidence was 9 cases per 100,000 population, but with large regional variation . The reasons for this are mixed and are likely due to; differences in Ixodes spp prevalence and Borrelia spp carriage rates, different levels of exposure to ticks by the general population, and differences in how patients access healthcare.
The seasonality observed here supports the known risk factors and epidemiology of Lyme disease. Tick populations in the UK have been shown to peak in June or July each year [25,26,27]. One would therefore expect to see tick bite incidence and exposure to Lyme disease to peak similarly. Clinical signs will appear anywhere from several days to a few weeks after a tick bite . Previous work in England and Wales showed a peak of serologically confirmed cases between July and September, with an assumed peak of symptoms earlier in the summer [4, 28]. This work would support this conclusion. This mirrors other Northern European countries, such as Finland and Germany, where clinically diagnosed cases peak throughout July and August [29, 30].
The age structure of this population compares closely with two recent studies performed in England and Wales [4, 15]. It shows the classic bimodal age distribution seen with Lyme disease, with an initial peak incidence in pre and peri-pubescent children, followed by a second larger peak from late middle age. The reasons for this age structure haven’t been formally assessed, however there is agreement that it likely reflects an increased exposure to tick habitats due to leisure behaviour rather than occupational exposure . These data display a predominance of female cases, unlike both studies referenced above. The reasons for this are hard to explain, but could be related to differences in health seeking behaviour .
Ninety-six percent of patients identified as being white, compared to 86% in the 2011 national census . There is no clear reason why ethnicity has any impact on a person’s susceptibility to Lyme disease. Instead, this apparent association is most likely due to sociocultural and behavioural reasons. Patients were found, in increasing numbers, living in less deprived areas. It must be noted that all ethnic minority groups were more likely to live in areas of higher deprivation than the white population , and this could explain the higher proportion of white patients within this population. Lyme disease patients were more likely than the national population to live in rural areas. The characterisation of Lyme disease patients as white and from suburban or rural areas with low deprivation may be explained by a complex combination of risk factors related to access to habitats which support ticks (either through work or recreation), and access to health care . As deprivation and rural-urban data are derived from aggregated data, the exact location of a case’s Lyme disease acquisition and socioeconomic status is unknown. These data are therefore acting as proxies and it is unknown how representative they are of the individual case. Despite the identification of clear trends and associations, these factors cannot be unravelled using these datasets, and so the degree of inherent bias remains unknown. Further research, utilising multivariable models, is required to understand the link, and any interaction and confounding, between ethnicity, deprivation, area of residence and presentation to hospitals with Lyme disease.
There is clear geographical variation in incidence between local authorities. The highest incidence is in southern-central and western England, which has traditionally been seen as a Lyme disease hotspot . Areas with no cases are unlikely to be due to an absence of disease but may reflect differences in case management or hospital coding practices. The remainder of England and Wales is a patchwork of low incidence with no obvious hotspots of disease. Interestingly, there are no clear foci of infection observed in either the Thetford Forest, the Lake District or the North Yorkshire Moors as identified previously by Public Health England (PHE) . In these areas the awareness, diagnosis and management of Lyme disease may differ from other areas, perhaps with primary care clinicians treating cases in the community and with fewer subsequent cases referred to hospitals. This shows a similar geographic distribution to laboratory-confirmed cases of Lyme disease . However, the areas of higher incidence in the hospital data expand further in to the south-west of England and into central England compared to laboratory cases. This is likely due to differences in case management. The high level of visual concordance between this research and the laboratory data suggest that both are accurately capturing the locations of Lyme disease patients. The geographical data collected by HES and PEDW is based upon the patient’s home address and no information is recorded on recent travel history or where a tick bite may have occurred, and so there may be an element of bias in the results. The data presented in this paper is at too low a geographical resolution and does not provide information on the patient’s tick bite location, to be able to hypothesise about any ecological associations with Lyme disease incidence.
Bed day analysis showed three distinct populations; those with one episode who weren’t admitted (35.5% of patients) or stayed for one night (12.5%), those with multiple episodes and a low number of bed days and those with one or many episodes that had a large number of bed days (Table 3). The first group is likely to represent patients with uncomplicated cases of Lyme disease. The second group often had consecutive daily episodes totalling 14 to 21 days, which could be consistent with daily intravenous doses of antibiotics as recommend by the British Infection Association and National Institute for Health and Care Excellence (NICE) guidelines [5, 8]. The final group appear to represent complicated cases of Lyme disease that require prolonged stays in hospital. It was not in the scope of this project to see whether any clinical presentations predisposed patients to these three groups, but further investigations are recommended.
Analysing the patient flow through the datasets has enabled better understanding of the care pathway for Lyme disease infected patients. Thirty percent of Lyme disease admissions in England, and 67.6% in Wales, originate from the A&E department. To place this into context, in 2011 69% of all NHS England admissions originate from A&E . The same report saw a decline in admissions through primary care referral and an increase through A&E between 2001 and 2011. It would be unlikely that the numbers of patients admitted in our study have more acute/severe presentations of disease that require immediate hospital attendance, however this cannot be ruled out. A combination of two factors possibly result in this finding; the lack of knowledge of the recommended care pathways for symptoms associated with Lyme disease (such as flu-like illness and rashes), and the difficulty in getting a prompt appointment in primary care [12, 34,35,36,37]. Peak non-urgent attendance at NHS emergency departments has been recorded at weekends , which may be due to the lack of access to primary care at the weekend [12, 35,36,37, 39]. However, our data show that the number of cases appearing in A&E is relatively evenly distributed throughout week, suggesting that the lack of knowledge of where to seek help with Lyme disease symptoms may be the predominant cause of the above findings. Further work is needed to explore why so many patients would seek treatment at a hospital when, for the majority of cases, management could occur at primary care level. By linking with primary care electronic health records, one may be able to see whether they had sought help first in primary care before arriving at A&E.
The major limitations of this study revolve around the use and validity of ICD-10 codes. A case of Lyme disease can be defined without laboratory confirmation, so there is no way to independently validate the accuracy of diagnostic coding in this context [5, 8]. Previous work has shown that coding practices in hospitals are not infallible, but are steadily improving; quality issues were primarily focused on patient management variables, rather than demographics and geography . Without such an audit, any potential inconsistencies in coding behaviour cannot be fully understood or quantified. Subjectively, admissions data in HES and PEDW were the most robust. As such, further work on the Lyme disease patient hospital population should primarily focus on admissions data.
Sixty-three treatment departments were recorded, some of which have no discernible link to Lyme disease. This may represent simple coding errors or that the code has been added for completeness when the primary reason for admission was unrelated to Lyme disease. The outpatient dataset was significantly overrepresented by two hospitals; both had the main treating department as dermatology and resulted in a high number of ACA codes. This is further seen, by the large number of outpatients seen on a Monday. These cases were all from one hospital, and likely represent one dermatology clinician’s outpatient clinic. This suggests that outpatient departments across England and Wales were not coding consistently and episodes may be being lost. The A&E dataset contained very low numbers of patients, in stark contrast to the large number being admitted through A&E as recorded in the APC dataset. The main reasons for these low numbers is not through lack of attendance but how coding is encouraged. Within A&E, coding is not required to be as specific as the admissions data, and is just needed to code a generalised condition, sub-analysis of more serious conditions and anatomical area involved . This results in Lyme disease potentially falling into multiple categories depending on symptoms, such as “Infectious disease”, “Local infection”, “Dermatological conditions” and “Facio-maxillary conditions”. This has been seen in previous work on arthropod bites, where all cases were recorded as “Bites/Stings” and routinely didn’t specify the causal arthropod .
PEDW only collects admission data and so some of the issues discussed above for the English dataset were negated. Unfortunately linkage between the PEDW and HES datasets was not possible; though, for reasons described above, these patients were likely to be unique. Without linkage there still is the potential of duplication of patients within the records and therefore there is a small degree of uncertainty attached to these results.