Antibiotic treatment of gastroenteritis in primary care

Abstract

Background

Gastroenteritis (GE) is a frequent reason for consultating a general practitioner. Yet little is known about antibiotic prescribing in primary care patients with GE. In this study, we quantified empirical and targeted antibiotic treatment of GE, compliance with recommendations from primary care clinical practice guidelines (CPGs) and the degree of antimicrobial resistance in patients receiving diagnostic faeces testing (DFT).

Methods

We performed a cohort study using routine care data of 160 general practitioners, including electronic patient records from 2013 to 2014. GE episodes were extracted and linked to microbiological laboratory records to retrieve results of DFT. For each episode, data on patient characteristics, DFT results including antimicrobial resistance testing, and antibiotic prescriptions were collected.

Results

We identified 13217 GE episodes. Antibiotic treatment was prescribed in 1163 (8.8%) episodes, most frequently with metronidazole (n = 646, 4.9%), azithromycin (n = 254, 1.9%) or ciprofloxacin (n = 184, 1.4%). Treatment was empirical for 641 (5%) GE episodes, of which 30% (n = 191) followed the CPG-recommended antibiotic choice. Targeted treatment following DFT results was prescribed for 537 GE episodes (4%), of which 99% (n = 529) followed CPG recommendations. Non-susceptibility to first- or second-choice antibiotics was demonstrated in three Salmonella isolates (9%–13% of all isolates) and one Campylobacter isolate (1%).

Conclusions

Antibiotic treatment of GE in primary care is relatively infrequent, with 1 in 11 episodes treated. Empirical treatment was more frequent compared with targeted treatment and mostly with non-CPG-recommended antibiotics. However, treatment based upon DFT results followed CPG recommendations.

Introduction

Every year, around one in three persons experiences an episode of gastroenteritis (GE).^1,2 In the Netherlands, 5%–12% of patients with GE consult their general practitioner (GP). With 240000–600000 thousand consultations annually, GE is imposing a substantial burden on primary healthcare.^1,2 Because GE is predominantly of viral aetiology and usually self-limiting,^2,3 routine prescription of antibiotics is not recommended in clinical practice guidelines (CPGs) for primary care.^4,5 Yet little is known about the frequency and pattern of antibiotic prescribing for GE in primary care. Additionally, antimicrobial resistance (AMR) of Campylobacter spp. and Salmonella spp. to ciprofloxacin was found to be considerable in the Netherlands,⁶ but primary-care-specific AMR data are lacking. These data are, however, important to inform guideline development on optimal antibiotic treatment. Internationally, AMR of Salmonella spp., Shigella spp. and Campylobacter spp. to first- and second-choice antibiotics is widespread and considered to be a substantial public health problem.^7–11

Here we report the frequency and determinants of antibiotic treatment for GE in primary care, the compliance with CPG recommendations on antibiotic treatment and the results of routine antimicrobial susceptibility testing (AST) in patients who underwent molecular diagnostic faeces testing (DFT).

Methods

Study design

We performed a retrospective cohort study using the routine primary care data from the Julius General Practitioner Network (JGPN), containing pseudonymized routine primary care data from 45 general practices (with 160 GPs and ∼290000 patients) in the academic primary care network of Utrecht, the Netherlands. Participating centres included both small- to medium-size practices and large primary care centres, all located in a typical Western European (sub)urban environment. The database contains detailed information on all patient contacts (telephone and practice consultations and home visits) of the participating GPs during office hours. Diagnoses during these contacts are coded according to the International Classification of Primary Care (ICPC)¹² and antibiotic prescriptions are registered according to the WHO guidelines for Anatomical Therapeutic Chemical (ATC) classification.¹³ GPs of participating practices are trained in correct use of ICPC coding and have on average 10 years experience in systematic coding of disease episodes.¹⁴

We selected patients with at least one GE episode that presented to GPs participating in the JGPN between 2013 and 2014. A GE episode was defined as a GP contact with assigned ICPC code D11 (diarrhoea), D70 (infectious diarrhoea) or D73 (suspected infectious diarrhoea). Multiple GP contacts for GE within a 60 day period were counted as one episode. More elaborate information on the study design and population can be found in the study design statement.¹⁵

Data collection

For each GE episode, we collected information on age, gender, number of GP contacts per disease episode, comorbidities, immunocompromising disorders, immunosuppressive therapy (Table S1, available as Supplementary data at JAC Online), antimicrobial drug prescriptions and, if performed, results of DFT. According to the Dutch primary care guideline on GE, DFT and antibiotic treatment are only recommended in patients with severe illness, those with compromised immunity or those with increased risk of disease or pathogen transmission from a public health perspective (i.e. food handlers or healthcare professionals).^4,16 DFT results were retrieved from Saltro Diagnostic Center, the main regional primary care laboratory, and linked to the JGPN cohort by a pseudonymization procedure performed by a ‘trusted third party’ in accordance with privacy regulations. Approximately 96% of the GPs in JGPN use the laboratory facilities of Saltro Diagnostic Center for performing microbiological diagnostics.

Table 1.

Characteristics of patients with GE episodes with and without antibiotic treatment

Patient characteristic	Antibiotic treatment (n = 1163)	No antibiotic treatment (n = 12108)	P
Age (years), median (IQR)	35 (12.5–53)	28 (4–53)	<0.001
Female	651 (56.0)	6548 (54.1)	0.227
Number of consultations, median (IQR)a	3 (1–4)	1 (1–2)	<0.001
Comorbiditiesb
any	369 (31.7)	3616 (29.9)	0.197
intestinal	169 (14.5)	1470 (12.1)	0.02
malignancy	36 (3.1)	317 (2.6)	0.384
diabetes mellitus	73 (6.3)	923 (7.6)	0.108
COPD	54 (4.6)	413 (3.4)	0.036
asthma	136 (11.7)	1261 (10.4)	0.191
immunocompromising condition	52 (4.5)	453 (3.7)	0.245
Chronic medicationb
systemic corticosteroids	97 (8.3)	616 (5.1)	<0.001
chemotherapy	4 (0.3)	47 (0.4)	1
immunosuppressants	11 (0.9)	65 (0.5)	0.118
acid-suppressive drug	290 (24.9)	2726 (22.5)	0.065
DFT
any DFT	684 (58.8)	1745 (14.4)	<0.001
bacteria	493 (42.4)	1378 (11.4)	<0.001
parasites	632 (54.3)	1457 (12.0)	<0.001
viruses	78 (6.7)	228 (1.9)	<0.001
any positive	560 (81.9% of tests)	517 (29.6% of tests)	<0.001
bacterium positive	107 (21.7% of tests)	177 (12.8% of tests)	<0.001
parasite positive	481 (76.1% of tests)	350 (24.0% of tests)	<0.001
virus positive	3 (3.8% of tests)	17 (7.5% of tests)	0.396

Numbers are counts of patients with percentages (%) unless otherwise indicated.

a

In-office consultations, home visits and telephone consultations per GE episode.

b

See Table S1.

We extracted data on all ATC-coded antimicrobial drug prescriptions that are presently recommended for the treatment of GE by the Dutch primary and secondary care CPGs.^4,16 These included ‘intestinal anti-infectives’ (paromomycin and vancomycin), ‘antibacterials for systemic use’ (co-trimoxazole, erythromycin, azithromycin, ciprofloxacin, teicoplanin, fidaxomicin) and ‘antiprotozoal agents’ (metronidazole and clioquinol). Antibiotic treatment was classified as either empirical or targeted, the latter defined as antibiotics prescribed in the presence of positive DFT results and empirical defined as prescriptions that were issued for GE episodes without DFT being performed, initiated pending DFT results or with negative DFT results. CPGs in the Netherlands advocate restrictive use of empirically prescribed antibiotics for GE, but recommend considering empirical treatment with azithromycin in severely ill or immunocompromised patients with suspected bacterial infection. In these CPGs, empirical treatment of patients with suspected parasitic GE is not recommended.^4,16 Targeted treatment was defined as an antibiotic prescription for a GE episode where a presumed causal pathogen was identified through DFT. For both empirical and targeted treatment we assessed whether the prescribed antibiotic was in accordance with the recommendations described in the Dutch CPGs^4,16 (Table S2).

Table 2.

Empirical and targeted antibiotic treatment for GE episodes (n = 13271)

Antibiotic	GE episodes with treatmenta
	empiricalb (%)	targetedc (%)	empirical/targeted (% of GE)
Any prescription	641 (100)	537 (100)	1163 (8.8)
Metronidazole	205 (32)	442 (82)	646 (4.9)
Azithromycin	191 (30)	65 (12)	254 (1.9)
Ciprofloxacin	165 (26)	19 (4)	184 (1.4)
Co-trimoxazole	56 (9)	3 (1)	59 (0.4)
Clioquinol	31 (5)	27 (5)	57 (0.4)
Erythromycin	12 (2)	1 (0)	13 (0.1)
Paromomycin	4 (1)	3 (1)	7 (0.1)
Vancomycin	3 (0)	2 (0)	5 (<0.1)
CPG recommended	191 (30)	529 (99)	–

Numbers are counts of patients with percentages.

a

Results in each column are not mutually exclusive as multiple antibiotics could have been prescribed for each GE episode.

b

Prescriptions issued for a GE episode without DFT, preceding a DFT result or after a negative DFT result.

c

Prescriptions issued after reporting of a positive DFT result.

Microbiological diagnostic testing

The presence of viruses was tested by immunochromatographic strip tests for each individual pathogen (i.e. adenovirus 40/41, norovirus, rotavirus). Bacteria and parasites were tested for each pathogen group by real-time multiplex PCR methods, which are described elsewhere.¹⁵ Bacterial pathogen testing included Salmonella spp., Shigella spp., Yersinia spp., Campylobacter spp. and Plesiomonas spp., and parasite testing included Blastocystis hominis, Cryptosporidium spp., Dientamoeba fragilis, Entamoeba histolytica/dispar and Giardia spp. We performed AST on bacterial isolates using disc diffusion for Campylobacter and automated susceptibility testing systems (Vitek 2, bioMérieux) for other bacteria. Campylobacter isolates were tested for resistance to ciprofloxacin, erythromycin and tetracycline; other bacteria were tested for resistance to ciprofloxacin, co-trimoxazole, amoxicillin and amoxicillin/clavulanic acid.

Analysis

We calculated the incidence rate of GP consultation for GE in the population by dividing the total number of GE episodes by the person-years of observation. We assessed the proportion of episodes treated with any antibiotic and for each antimicrobial agent. Patient characteristics were compared between episodes with and without antibiotic treatment, using Pearson’s χ² test statistic for dichotomous variables and the Kruskal–Wallis test for continuous and categorical variables. Variation in antibiotic treatment proportions among months and age groups was assessed using Pearson’s χ² test statistic; a subsequent pair-wise comparison was performed and included Holm’s correction for multiple testing.¹⁷ We classified antibiotic prescriptions as either empirical or targeted and calculated proportions of GE episodes treated according to CPG-recommended antibiotic choices. Furthermore, for each gastrointestinal pathogen species detected through DFT, we assessed the proportion of infections treated with antibiotics.

Finally, antibiotic resistance levels were assessed for bacterial species detected with DFT by calculating the proportions and 95% CIs of resistant isolates for each antibiotic, but only for those bug–drug combinations with AST results available for at least 20 isolates per species. Statistical analysis was conducted in R 3.3.2, using the tableone, dplyr, xts and rms packages.

Ethics

The Institutional Review Board of the University Medical Center Utrecht considered this study not subject to the Medical Research Involving Human Subjects Act, therefore full ethics assessment was not deemed required (case number 13-480).

Results

Between January 2013 and December 2014, the JGPN database comprised a total of 515811 person-years from 277578 enlisted patients. In total, 13217 episodes of GE were recorded in the database, resulting in an overall incidence rate of primary care consultations for GE of 26 per 1000 person-years (Figure 1).

Figure 1.

Flow chart of primary-care GE episodes and antibiotic treatment.

Antibiotic treatment of GE

Antibiotic treatment was prescribed in 1163 (8.8%) episodes (Table 1), with 1393 antibiotic prescriptions in total (105 per 1000 GE episodes). In most episodes (n = 982, 84.5%) a single antibiotic was prescribed. In 145 (12.5%) and 36 (3%) episodes two and three or more antibiotics were prescribed, respectively. Metronidazole was most frequently prescribed (n = 646, 4.9% of all GE episodes), accounting for more than half of the antibiotic treatment courses for GE, followed by azithromycin (n = 254, 1.9% of all GE episodes), ciprofloxacin (n = 184, 1.4%), co-trimoxazole (n = 59, 0.4%) and clioquinol (n = 57, 0.4%) (Table 2).

Determinants of antibiotic treatment

Compared with patients without antibiotic treatment, those treated were older (median 35 versus 28 years, P < 0.001), more often had intestinal comorbidity (14.5% versus 12.1%, P = 0.02) and chronic obstructive pulmonary disease (4.6% versus 3.4%, P = 0.036) and were more often prescribed systemic corticosteroids (8.3% versus 5.1%, P < 0.001). Patients treated with antibiotics had on average more GP consultations per episode (median 3 versus 1, P < 0.001) and higher DFT testing rates (59% versus 14%, P < 0.001). Patients who were 0 and 1–4 years old (1.7% and 5%, respectively) were less frequently treated with antibiotics than older (>4 years) patients (between 8.1% and 11.4%, P < 0.05) (Figure 2a). Prescription rates differed significantly between months of the year (P = 0.003) (Figure 2b).

Figure 2.

(a) Antibiotic treatment of patients with GE stratified by age category. (b) Monthly (n = 24) antibiotic prescription for patients with GE cases (n = 13217).

Empirical and targeted antibiotic treatment

Empirical treatment was prescribed in 641 (4.9%) of the patients presenting with GE episodes; 479 (74.7%) of these 641 patients did not have a DFT request, 41 (6.4%) received the prescription before DFT results became available and 121 (18.9%) received the prescription after a negative DFT result. Targeted treatment was prescribed in 537 (4.1%) of the patients presenting with a GE episode. Empirical prescription preceded subsequent targeted antibiotic prescription in 1.3% (n = 15) of the GE episodes with treatment (Table 2).

Azithromycin, which is the recommended choice for empirical treatment according to the CPGs, was prescribed in 30% (n = 191) of the empirically treated patients. Metronidazole (n = 205, 32%) and ciprofloxacin (n = 165, 26%) were also prescribed frequently, even though they are not recommended for empirical treatment (Table 2).

Metronidazole was most frequently prescribed as targeted antibiotic treatment (n = 442, 82%) (Table 2). The antibiotic choice for targeted treatment was in accordance with CPG recommendations in 99% (n = 529) of GE episodes with targeted antibiotics.

Pathogen-specific antibiotic treatment

Routine DFT testing identified a total of 1437 potential GE-causative enteropathogens in 2429 GE episodes. Overall, 751 (52%) of the identified enteropathogens were treated with a CPG-recommended antibiotic and 32 (2%) with a non-recommended antibiotic. The remaining 654 patients (46%) with identified enteropathogens did not get antibiotic treatment (Table 3). In the patients with a confirmed bacterial or protozoal enteropathogen, an antibiotic was prescribed in 32% (n = 94) and in 58% (n = 657), respectively. Blastocystis and Dientamoeba were relatively often treated with antibiotics, in 52% (n = 234) and 60% (n = 292) of the positive patients, respectively. Patients receiving targeted treatment for bacteria or parasites had more GP contacts per GE episode (median 4 versus 2 contacts, P < 0.001) and higher specialist referral rates (9% versus 5%, P = 0.05) compared with patients that did not receive treatment. Other patient characteristics (age, comorbidities and chronic medication use) did not significantly differ between those treated and those not treated.

Table 3.

Pathogen-specific antibiotic treatment for GE and accordance with first- and second-choice CPG-recommended antibiotics

Pathogen	Infection count	Antibiotic treatment, no. of patients (%)
		CPG recommendeda	not CPG recommended	no treatment
Bacteria	293	94 (32)	11 (4)	188 (64)
Clostridium difficile	12	10 (83)	10 (83)	2 (17)
Campylobacter	204	67 (33)	–	127 (62)
Shigella	36	10 (28)	1 (3)	25 (69)
Salmonella	30	6 (20)	–	24 (80)
Yersinia	7	1 (14)	–	6 (86)
Plesiomonas	4	–	–	4 (100)
Parasites	1124	657 (58)	20 (2)	447 (40)
Giardia lamblia	160	140 (88)	3 (2)	17 (11)
D. fragilis	477	283 (59)	9 (2)	185 (39)
B. hominis	451	226 (50)	8 (2)	217 (48)
E. histolytica/dispar	18	8 (44)	–	10 (56)
Cryptosporidium	18	–	–	18 (100)
Viruses	20	–	1 (5)	19 (95)
adenovirus	3	–	–	3 (100)
norovirus	10	–	–	10 (100)
rotavirus	7	–	1 (14)	6 (86)
Total	1437	751 (52)	32 (2)	654 (46)

a

See Table S2 for specification of CPG-recommended antibiotics.

Antimicrobial susceptibility

Microbiological culture and AST were performed in 144 (71%) of 204 episodes in which Campylobacter isolates were detected and in 23 of 30 (77%) episodes in which Salmonella was detected. One Campylobacter isolate (1%, 95% CI 0.1%–4%) was resistant to erythromycin (Table 4). Three Salmonella isolates (13%, 95% CI 3%–35%) were resistant to ciprofloxacin and two to co-trimoxazole (9%, 95% CI 2%–30%). None of the patients with resistant bacteria was treated with an antibiotic for which resistance was demonstrated.

Table 4.

Number of isolates (for n ≥ 20) tested for antimicrobial resistance and proportion of resistant isolates for each bacterial species

Antibiotica	Campylobacter spp.			Salmonella spp.
	isolates, n	resistant isolates		isolates, n	resistant isolates
		n	percentage (95% CI)		n	percentage (95% CI)
Ciprofloxacin	147	90	61 (53–69)	23	3	13 (3–35)
Erythromycinb	144	1	1 (<0.1–4)	–		–
Tetracycline	147	60	41 (34–50)	–		–
Co-trimoxazole	–		–	23	2	9 (2–30)
Amoxicillin	–		–	23	5	22 (8–44)
Amoxicillin/clavulanic acid	–		–	23	0	0 (0–18)

a

Indicated in bold are the first- and second-choice antibiotics recommended by the Dutch CPGs for treatment of infectious GE.^4,16

b

Resistance to erythromycin tends to correspond with cross-resistance to other macrolides, such as azithromycin.²⁰

Discussion

Main findings

We found that GPs prescribe antibiotic treatment in 1 in 11 patients who presented with a GE episode in primary care. Metronidazole is most frequently prescribed, followed by azithromycin and ciprofloxacin. About half of antibiotic treatment for GE is prescribed on an empirical basis, without microbiological test results available. Although azithromycin is the first-choice empirical treatment according to the CPGs, metronidazole was more frequently prescribed, closely followed by azithromycin and ciprofloxacin. Targeted treatment is, in nearly all cases, in agreement with CPG recommendations. After DFT confirmation, GPs refrain from antibiotic treatment in most patients with bacterial infections, but they do treat the majority of patients with parasitic infections. AST results suggest that almost all Campylobacter and Salmonella infections in patients with DFT are susceptible to the antibiotics currently recommended by CPGs.

Interpretation of results

Antibiotic treatment of GE is less frequent compared with other infectious diseases in primary care.^18,19 For example, antibiotic treatment is prescribed in 55% of the patients with acute otitis media antibiotic treatment is prescribed, and 14% of the episodes with acute upper respiratory tract infections.¹⁹

When receiving antibiotic treatment for GE, more than half (55%) of the prescriptions are made empirically, mainly in patients without testing for enteropathogens. This proportion seems relatively high when considering that CPGs discourage the prescription of empirical antibiotics.^4,16 Specifically, metronidazole and ciprofloxacin are frequently prescribed as empirical treatment, despite the recommendation of CPGs for azithromycin as empirical treatment for suspected bacterial GE.^4,16 Possibly, GPs prescribe metronidazole when suspecting a parasitic infection based on current clinical presentation or due to a previous parasitic infection of the patient or of family members. Ciprofloxacin is often prescribed as pre-emptive therapy for travellers’ diarrhoea,^4,16 which potentially explains the relatively high empirical prescription rate of this antibiotic. Although with GE the prescription of empirical antibiotics may be justified in some patients, it remains to be investigated whether this can be considered good clinical practice and whether CPGs should expand the recommendations for empirical treatment of GE in primary care, including alternative antibiotics such as metronidazole and ciprofloxacin.²⁰ Our study was, however, not designed to answer this question.

Microbiological testing may allow empirically started antibiotic treatment to be switched to targeted treatment. Our findings, though, demonstrate that GPs await microbiological test results to guide subsequent antibiotic treatment. In fact, only 1.3% of all empirical treatment episodes were followed by targeted antibiotic treatment.

We found a positive association between age and antibiotic prescribing, with antibiotics prescribed the least in children under 4 years old. This contrasts with other (mainly respiratory) infections, where antibiotic prescription rates are highest in younger children.¹⁹ Apparently, GPs consider GE in children as mild and self-limiting, and presumably of viral origin.^2,21

Only 32% of the patients with GE episodes in which a potential bacterial cause of infection was identified with DFT were actually treated with antibiotics. These patients had fewer GP contacts and lower specialist referral rates compared with patients with targeted treatment, suggesting a benign and self-limiting course of GE. As GPs did not regard post hoc antibiotic treatment necessary in these patients, it remains unclear what the actual indication for the diagnostic testing was. It can be argued that, if no clear indication for treatment exists, DFT has no direct clinical value and can be withheld in these patients. Furthermore, around 60% of the GE episodes in which DFT revealed Blastocystis and Dientamoeba were treated with antibiotics, although the clinical relevance of these parasites in GE remains unclear. As treatment of these parasites may not always be clinically beneficial^22–25 and in many cases even ineffective,^24,26,27 further research needs to evaluate the role of antibiotics in the treatment of these parasites and identify when antibiotic treatment of primary care patients with GE is appropriate.

Strengths and limitations

We used a large routine-care dataset, consisting of a representative sample of the general Dutch population,¹⁴ and included detailed laboratory results of molecular PCR faeces testing to determine infection status.

The main limitation of our study is that we could only perform a descriptive analysis of diagnostic practices and antibiotic use in a primary care population. We were not able to relate determinants of diagnostic behaviour of GPs and antibiotic prescribing to the clinical disease course, as the data did not provide such in-depth information. The same holds for the assessment of CPG adherence to prescribe no, empirical or targeted antibiotic treatment. Finally, we only obtained AMR data for a small number of isolates. The usefulness of these data in guiding antibiotic policy of CPGs is therefore limited.

Conclusions

Antibiotic treatment of GE in primary care is relatively infrequent, with 1 in 11 episodes treated. Empirical treatment was more frequent compared with targeted treatment and was mostly with non-CPG-recommended antibiotics. However, treatment based upon DFT results followed CPG recommendations. As bacterial infections are largely untreated, the clinical value of DFT in patients with suspected bacterial GE is unclear. Non-susceptibility to first- or second-choice antibiotics was infrequent.

Acknowledgements

We thank Marloes van Beurden and Marcel Moes for their valuable contributions regarding collection of the study data.

Funding

This study was supported by internal funding.

Transparency declarations

None to declare.