Diagnostic behaviour of general practitioners when suspecting Lyme disease: a database study from 2010-2015

Lyme Borreliosis (LB) is the most common tick-borne disease in the world [1‐4] caused by Borrelia burgdorferi sensu lato-spirochetes, of which Borrelia afzellii and Borrelia garinii are most prevalent in Europe [1, 5]. Over the past decades, the incidence of LB in Europe has increased to approximately 65,500 patients per year [6‐9]. Raised public awareness of LB, its various clinical manifestations, its treatable character, the fear of disease among patients and the increased availability of serological tests have driven a rising demand for diagnostic testing [8, 10, 11]. Current European guidelines, although not tailored to primary care, clearly state that serological tests must only be done when there is a high suspicion of LB. This high a priori probability is mainly based on LB-specific symptoms (like symptoms suggesting Lyme arthritis or neuroborreliosis) [12, 13]. Requesting non-indicated tests may lead to false-positive results, since - depending on the assay - 3-9% of healthy controls is seropositive for Borrelia antibodies [14]. This is because serological tests do not differentiate between an active LB and an (asymptomatic) infection from the past [11]. Other reasons for a false-positive test result are infections with other related pathogens (e.g. syphilis), autoimmune disorders, or cross-reactivity between spirochetes. [15] In general, according to Bayes’ theorem, the usefulness of a serological test for LB depends on the pre-test probability and the subsequent predictive values in the setting where the test is being used [2, 16, 17]. However, even in a tertiary, multidisciplinary setting it is challenging to either rule out or demonstrate an association with Borrelia burgdorferi sensu lato. [18] The seropositivity rates for LB vary throughout Europe, from 3.4% in the general population in Italy to 15.2% in France [14, 19, 20]. When testing high-risk populations, these seropositive rates will be even higher. In Poland, healthy forestry workers reach a seropositivity rate of 45% [19, 21]. Because positive test results are often interpreted as Borrelia being the causal agent of the symptoms, overtreatment is common in cases of false-positive test results. Previous studies in general practices have only focused on tick bites and erythema migrans (EM). In Belgium 50% of the patients presenting with a typical EM got serologically tested by the GP [9], contrary to national/international guidelines (i.e. patients presenting with EM should always receive antibiotic treatment, so serological testing has no additional value; moreover, a false negative test result may occur, which can be misleading [12, 13].

The aim of the present study was to describe GPs’ diagnostic behaviour with respect to LB. Our study is the first to describe seasonal trends in LB contacts, to link symptoms to serology requests, and to investigate if GPs’ diagnostic behaviour on LB has changed over the years 2010-2015.

An observational study was carried out with anonymized data extracted from the database of the Academic Network of General Practice of VU University Medical Center (ANH VUmc), Amsterdam. The database contains pseudonimized primary health care data. We used the anonymized data of 56,996 patients registered in 12 general practices (with more than 60 GPs) in Amsterdam, from January 2010 to June 2015. In the Netherlands, almost all non-institutionalised citizens are registered to a general practice and the GP acts as gatekeeper. We developed an extensive search and selection strategy, based on the International Classification of Primary Care (ICPC-1) codes, free text and diagnostic test codes, to identify contacts related to LB (see Fig. 1 for the flow chart, Appendix 1 for the search strategy, and Appendix 2 for the origin of the identified patients, divided by search strategy). Annotations of the GPs from all identified contacts (consultations, telephone, home visits, e-mails) were reviewed by one of the investigators (EB). Contacts not related to LB or tick bites were excluded (e.g. other insect bites). A random selection of 5% of the identified contacts was reviewed by a second, independent reviewer to check the reliability of the data extraction. Disagreements were resolved during a consensus meeting between investigator and independent reviewer.

The following data were extracted from the selected medical records (analytical sample): basic characteristics of the patients (i.e. sex, age at the time of the first LB contact), characteristics of the LB contact and episode (i.e. date of contact, type of contact, number of contacts), ICPC-1 code, symptoms registered, findings during physical examination, diagnosis and management (i.e. Lyme serology or referral). Contacts belonging to the same episode were merged for the analysis of presenting symptoms and diagnostic testing. In the analysis, a distinction was made between individual contacts and episodes that include all contacts concerning the same complaint. Also, a distinction was made between a definite tick bite and a possible tick bite; the latter was called ‘insect sting’ in the analysis.

The data contained different Lyme serology tests from different laboratories and it was not possible to identify the exact assay that was used for each patient. Based on the quantitative value and the cut-off level we could infer that the majority of the samples were screened with the Immunetics C6 ELISA kit (Immunetics Inc., Boston, USA). We used the available laboratory interpretation of the screening test. This included the requirement that all equivocal or positive screening test results had to be confirmed by an immunoblot. Laboratories classified immunoblot outcomes as ‘negative’, ‘inconclusive’ or ‘positive’.

Descriptive analysis was performed using SPSS 23.0. Chi-square tests for trend (linear-by-linear) were used for comparing differences between years regarding contacts, episodes and serological tests.

The ANH database is run according to Dutch privacy legislation and contains pseudonymized general practice care data from all patients of the participating general practices, excluding those patients who object to this. Observational studies based on anonymized data from the ANH VUmc database are exempted from informed consent of patients.

Figure 1 provides a flowchart of the study. Twelve general practices were included (N = 56,996 patients), resulting in 2311 patients who visited the GP because of complaints or concerns related to LB in the period January 2010 to June 2015 (see Appendix 3 for practice characteristics). The reliability of the data extraction was acceptable: from a random selection of 5% of identified LB contacts, 6 out of 350 contacts (1.7%) had been classified incorrectly.

The basic characteristics of patients and their LB contacts are shown in Table 1.

The overall incidence of episodes related to LB was 8.8 per 1000 person years within the dataset of 56,996 patients. The incidence of LB episodes across the years 2010-2014 did not show a significant change (p = 0.932). Since data for 2015 were only partially available, 2015 was not included in this analysis. Each year, LB related contacts showed a peak during spring and summer (May until September; Fig. 2). The majority of patients (78.2%) contacted their GP only once within an LB episode. For LB episodes with 2 or more contacts, the mean interval between first and last contact was 45 days. Most patients (88%) experienced only one LB episode over the period January 2010 to June 2015.

Table 2 provides an overview of reasons for encounter during episodes related to LB (N = 2619 episodes). Also, the table illustrates per reason for encounter how often serological testing was ordered and how often this resulted in a positive immunoblot test result.

Symptoms most frequently registered by GPs in LB episodes were recent insect sting, defined as ‘recent’ when stung within three months prior to contact (N = 1162, 44.4% of episodes), skin abnormalities (N = 1080, 41.2%) and general symptoms like fatigue and malaise (N = 966, 36.9%; Table 2).

Lyme serology was requested during 953 out of 2619 episodes (36.4%). GPs mostly requested serology for episodes in which patients presented with general symptoms (N = 690, 71.4%). The percentage of positive test results was highest for the episodes in which patients presented with skin abnormalities (14.5%) and for episodes in which patients requested for a test themselves (13.4%; Table 2). GPs requested serology for 18.4% (N = 25 out of 136) of the episodes in which patients presented with typical EM, for 10.1% (N = 33 out of 327) of the episodes in which patients presented with a locally irritated tick bite, and for 13.2% (N = 121 out of 914) of the episodes in which patients presented with an asymptomatic tick bite.

From January 2010 to June 2015, the number of requested serological tests did not change significantly (p = 0.190). Of the 953 serological tests, 19.3% was screened positively. Only 5.9% of these were confirmed by the immunoblot. Referral to a specialist occurred in 7.6% of all LB episodes, mostly to neurologists (24.2%) or internists (21.7%; Table 3).

The general practices showed considerable differences with respect to serological testing during LB episodes, ranging from 19.2% to 75.8% of the LB episodes. There was a tendency for practices with lower serological testing rates to have higher positive tests rates (Fig. 3).

This descriptive GP database study, covering the period January 2010 to June 2015, was the first to investigate seasonal trends in LB contacts, to link symptoms to serology requests, and to investigate GPs’ diagnostic behaviour over time. The distribution of LB contacts showed annual peaks during spring and summer. Episodes in which patients presented with general symptoms, like fatigue, were most frequently followed by serological testing. Serological testing was performed in 36.4% of LB episodes; only 5.9% of these tests turned out to be positive. The high number of LB consultations with nonspecific complaints together with the low frequency of positive serological tests, and the considerable inter-practice differences indicate a high number of inappropriate tests and underscores the need for an easy-to-use diagnostic algorithm that is (also) applicable in daily general practice.