Children with respiratory tract infections in Swedish primary care; prevalence of antibiotic resistance in common respiratory tract pathogens and relation to antibiotic consumption

Children, aged 0-10 years, consulting their PHCC with symptoms of RTI, between 1st Nov 2013 until 30th April 2014 and 1st Nov 2014 until 30th April 2015 were included. Information about the study and a written invitation was given to the parents at arrival at the clinic and informed parental consent was obtained for each case. Lack of knowledge of the Swedish language was the only exclusion criterion.

The parents were asked to fill out a questionnaire regarding factors believed to affect bacterial resistance, such as the child’s age, antibiotic consumption over the last year and specifically during the last 4 weeks, type of day care, number of siblings, travel abroad, hospital admissions etc. (Additional files 1 and 2). The same patient was only included once. The individual antibiotic consumption was evaluated via the questionnaire filled out by the parents.

NP specimens were obtained in a standardised manner by trained laboratory personnel at the PHCCs. A swab (ESwab ™ Liquid Amies Collection and Transport System, COPAN) [26] was taken from the rear wall of the nasopharynx, where the stick was kept still for 10 s before being withdrawn. The test swab was kept in a refrigerator until being sent to the Department of Clinical Microbiology in Malmö/Lund where it was analysed for S.pneumoniae, H.influenzae, Moraxella catarrhalis and beta-haemolytic streptococci group A, according to national recommendations [27]. Antibiotic susceptibility testing was performed according to recommendations of the EUCAST [28]. Isolated S.pneumoniae were screened for penicillin resistance by disc diffusion test using 1-microgram oxacillin discs. Minimum inhibitory concentration (MIC) for benzylpenicillin was determined by E-test in isolates with inhibition zones <20 mm. S.pneumoniae with MIC of ≥0.125 mg/L was considered penicillin non-susceptible (PNSP). Screening for beta-lactam resistance in H.influenzae was performed by disk diffusion test using benzylpenicillin discs. If found to be benzylpenicillin resistant, a beta-lactamase test was performed (Nitrocefin). For detection of chromosomal penicillin resistance (BLNAR), a disk diffusion test was used with a beta-lactamase stable cephalosporin (cefaclor) as an indicator.

In this paper, colonization with resistant bacteria was defined as the presence of PNSP and/or H.influenzae strains with beta-lactamase production and/or H.influenzae with chromosomal resistance, (BLNAR).

The main outcome was the presence of PNSP and resistant H.influenzae in the NP culture. Questionnaire and laboratory data were collected in Microsoft Excel and statistical analysis was performed using SPSS 22.0 software (IBM, Armonk, NY, USA). Descriptive statistics are presented as numbers and proportions. Comparisons between proportions for categorical variables in two independent groups were performed using the two-sided χ²-test. Multiple logistic regressions were used to model the relationship between the outcome variable (carriage of resistant bacteria) and several independent variables (risk factors for carriage of resistant bacteria retrieved from the parental questionnaire).

Of the 12 PHCCs, eight provided 90% of the NP cultures. We aimed at collecting 400 cultures but managed 340 due to declining recruitment. Altogether, 422 children were recruited, of which 340 participated with NP cultures and the parental questionnaire. A total of 82 children did not accept having nasopharyngeal culture taken (Fig. 1). The baseline characteristics of the children with and without a nasopharyngeal culture taken were similar (Table 1). The median age in both groups was 2.0 years (IQR = 1-5).

Of the 340 children, 86 (25%) carried S.pneumoniae, 129 (38%) H.influenzae, 185 (54%) M.catarrhalis, 24 (7%) beta haemolytic streptococci group A and 69 (20%) were culture negative or carried non-target organisms (Table 2). Of the 86 isolates with S.pneumoniae, five (6%) were PNSP all with MIC = 0.125 mg/L. Three of these PNSP isolates also showed resistance to trimethoprim, one showed additional resistance to tetracycline and one additional resistance to tetracycline, erythromycin and clindamycin. Resistance data are presented in Table 2.

When analysing the 340 children from which NP cultures were taken, we found that having received antibiotic treatment (predominately Penicillin V) within the previous 4 weeks predisposed for resistant bacteria (OR = 3.08, 95% CI (1.13-8.42). Antibiotic consumption during the last year did not significantly affect the prevalence of resistant bacteria in our analysis. But, since it could interact with the variable antibiotic consumption the previous 4 weeks we decided to exclude it in the regression analysis. The risk of carrying resistant bacteria also decreased with age (OR = 0.75, 95% CI (0.57-0.99) (Table 3).

In 178 (52%) of the cultures, there was growth of either S.pneumoniae or H.influenzae or both, of which 40 (22%) isolates showed some sort of resistance to beta-lactam antibiotics, either PNSP, and/or beta-lactamase producing H.influenzae and/or BLNAR (Fig. 1).

When analysing only the 178 children with growth of either S. pneumoniae or H. influenzae, or both, we noted that boys were more commonly colonised with resistant bacteria (OR = 2.95, 95% CI (1.25-6.98) (Table 4). No significant association was found between resistance and hospital care, travelling abroad, parental smoking, day care attendance, respiratory tract disease (asthma/allergy) or pneumococcal vaccination (Tables 3 and 4).

The prevalence of S.pneumoniae and H.influenzae were similar to the data from the local clinical microbiology laboratory for the same period and age group (Table 2). We could not identify any statistically significant difference in the prevalence of PNSP (p = 0.31), beta-lactamase producing H.influenzae (p = 0.33) or BLNAR (p = 0.08) between our community study and the data from the local clinical microbiology laboratory (Table 2).

We found a low prevalence of PNSP and resistant H.influenzae among respiratory tract pathogens in children with symptoms of RTIs in primary care, indicating that phenoxymethylpenicillin is valid as empirical treatment for these patients, if antibiotics are required. We noted no significant differences in prevalence of PNSP and resistant H. influenzae in our results compared to the local microbiological laboratory. Studies based on primary care populations are important for validation of resistance measures from microbiological laboratory findings due to differences in case-mix. Treatment with antibiotics, within the previous 4 weeks predisposed for resistant bacteria supporting continued efforts to rationalise antibiotic use.

Previous attempts to investigate bacterial resistance in RTIs in primary care have studied populations containing a mix of children attending PHCCs as well as those seeking care at hospital emergency rooms [29, 30]. The strength of this study is that the population consists of primary care patients only. Bearing in mind that most antibiotics are prescribed in primary care, monitoring resistance levels in this population should be of importance to appreciate the true burden of resistance as well as for the development of guidelines and recommendation of empirical antibiotic choice. Based on the carriage rate of S.pneumoniae in Swedish children we estimated a need of 400 cultures to identify about 100 isolates of S.pneumoniae, but we settled with 340 cultures and 86 isolates due to a declining recruitment rate. We chose to recruit during the winter season because the consultations for RTIs are more common at that time of the year. We do not believe that the few cases missed during the rest of the year would affect our result. We deem the risk of patient selection bias to be low because although we received a variable number of cultures from the different PHCCs, indicating difficulties in recruiting patients, we noted that most parents were willing to participate in the study once they were asked. Furthermore, the only exclusion criterion was lack of knowledge of the Swedish language. Ongoing treatment with antibiotics was not asked about, and therefore not an exclusion criterion. We had a good response rate regarding antibiotic use the previous 4 weeks, which to most parents include whether the child are on any current antibiotic treatment. The use of routine sample collection and well-established methods for analysis of NP cultures should ensure the reliability of the results.

The number of antibiotic prescriptions during the last year was difficult for some parents to answer for in the questionnaire and therefore we saw around 20% missing data regarding this question. However, since the missing data were equally distributed in the groups, with and without resistance, it should not affect the regression analysis.

The association between antibiotic consumption during the previous 4 weeks and bacterial resistance showed an OR with a wide 95% confidence interval due to a small number of children that had taken antibiotics within the 4 weeks prior to the culture and a small number with beta-lactam resistant strains. Although there is some insecurity in the result due to the small sample size it should be possible to infer to similar populations.

To differentiate whether our results represent colonisation or infection is not possible. The fact that the children all presented with symptoms of RTIs, such as fever, cough, sore throat, earache or runny nose supports the theory of it representing an infection, but, the culture might not represent the etiological agent. We did not perform any respiratory virus testing, which could explain some of the symptoms. Treatment information and outcome data could also have helped in the differentiation between colonisation and infection, although most RTIs, whether caused by bacteria or virus are self-healing. Since serotyping of the pneumococci was not performed we don’t know if they were covered in the 13-valent vaccine used in Sweden.

We found that 25% of the children seeking care at their PHCC for RTI symptoms and who were recruited to our study carried S.pneumoniae and 38% carried H.influenzae. These figures were in line with other recent European and American studies [7, 29, 31, 32].

By contrast, the level of PNSP (MIC ≥ 0.125 mg/L) was lower in our study than in many other countries [24, 32, 33]. PNSP levels are known to be correlated to antibiotic consumption [6]. Over the last 20 years, great reductions in antibiotic prescribing for RTIs in Swedish PHC have been achieved [34], which is likely to have played an important role in the reduction of PNSP over time. Swedish guidelines for RTIs recommend phenoxymethylpenicillin as first line treatment to target the pneumococci, and the low level of resistance found in this study supports that recommendation. By contrast, the prevalence of beta-lactamase producing H.influenzae does not seem equally influenced by the reduced levels of antibiotics. The beta-lactamase producing H.influenzae has remained between 15 and 18% in Sweden since 2007 and we see a continuous increase in BLNAR [18]. Previously, we have seen a similar development for Moraxella catarrhalis, of which now 95-98% is beta-lactamase producing in every setting [24, 33].

Antibiotics have been shown to affect our bacterial flora so that penicillin sensitive pneumococci decreases 1 month after antibiotic therapy [7] as does penicillin sensitive H.influenzae [35]. Moreover, antibiotic use is associated with increased prevalence of PNSP [12, 36‐38] and BLNAR [39], but only some studies show antibiotic consumption to be a risk factor for carrying beta-lactamase producing H.influenzae [40‐42] while others do not [31, 43, 44].

When we considered only the children with growth of either S.pneumoniae or H.influenzae (n = 178) we found that boys were more often colonised with resistant bacteria than girls, which is supported by some previous studies [36] but not by others [12]. Boys have also been shown to catch more RTIs in younger ages than girls [45, 46] and develop complications to AOM more often [47]. The reasons for these differences between the sexes are not known.

Other risk factors for carrying PNSP (MIC ≥ 0.125 mg/L) have been reported in previous studies with various results. Some studies support our findings that younger age predisposes to PNSP carriage [12, 31, 36], and also longer duration of PNSP carriage [36] Other studies show that attending a day care centre increases the risk [31] while the number of siblings, presence of respiratory tract disease and parental smoking did not affect resistance levels in the studies mentioned above, nor in our results.

A lower PNSP prevalence in a PHC population compared to local microbiology laboratory data, as was seen in Iceland in 1993 [11, 12] may be due to patient selection. In the Swedish guidelines, it is only recommended to perform NP swabs on patients with RTIs when first-line treatment has failed or in recurrent infections. The local microbiology laboratory data also includes cultures from hospital patients, both admitted and attending Emergency services and in some cases also patients referred from primary care. The differences in case-mix provide support for regular measurements of resistance in primary care populations to validate laboratory resistance data. In many counties in Sweden there has also been a relative reduction in the number of performed NP swabs (Swedres/ SWARM 2001-2016, www.​folkhalsomyndigh​eten.​se) and the cultured isolates therefore may exaggerate the prevalence of resistance, which in turn could have implications for treatment guidelines and empirical choice of antibiotics.

The carriage rates of H.influenzae were similar in our findings and the local microbiological laboratory data, as were the levels of beta-lactamase producing H.influenzae and BLNAR.