Antibiotic prescribing and antibiotic resistance in community practice: retrospective study, 1996-8

We describe a retrospective survey of antibiotic prescribing in general practitioners' surgeries and resistance to antibiotics in Wales from March 1996 to April 1998.

Methods and results

Data on the susceptibility to antibiotics of coliform organisms in routine urine samples taken by general practitioners for diagnosis of urinary tract infections were collected from the Bangor, Cardiff, and Rhyl Public Health Laboratories and the East Glamorgan, Prince Charles, and Wrexham Maelor Hospitals. Data on the prescribing practices of surgeries were obtained from the Welsh Prescription Pricing Service. Rates of prescribing (the number of prescriptions/1000 patients per year) and resistance rates (which excluded multiple isolates of organisms with the same susceptibility from the same patient) were calculated for each surgery. The use of broad spectrum penicillin formulations without a β lactamase inhibitor (such as ampicillin and amoxicillin) was estimated by subtracting the number of prescriptions for co-amoxiclav from the total number of prescriptions for all other broad spectrum penicillins. We use the term amoxicillin below to refer to these broad spectrum penicillins without a β lactamase inhibitor.

Resistance rates for surgeries which were based on fewer than 50 isolates were excluded, leaving data on about 30 000 isolates from 190 general practitioner surgeries serving about 1 200 000 patients. We sought to identify the effects of bias caused by the selective submission of urine samples by examining the relation between resistance rates and sampling (number of urine specimens/1000 registered patients) and the relation between positivity (number of coliform isolates/100 samples or 1000 registered patients) and prescribing or sampling.

The use of antibiotics and rates of resistance to antibiotics varied between surgeries; the correlation between the prescribing of an antibiotic and resistance to the same antibiotic was often significant (table). The correlation was also significant between the use of amoxicillinand resistance to trimethoprim and vice versa. Combined resistance to ampicillin and trimethoprim occurred in 21% (6782/32 532) of isolates and was significantly associated with the use of both trimethoprim and amoxicillin (P<0.001). The correlation between the use of amoxicillin and resistance to trimethoprim and vice versa was lost when strains exhibiting combined resistance to both agents were removed from the analysis.

There was no significant correlation between antibiotic use and the number of urine specimens submitted for testing per 1000 registered patients or the number of coliform isolates in urine samples per 1000 registered patients. The number of isolates per 1000 registered patients correlated linearly with the number of urine specimens submitted per 1000 registered patients (P=0.001, r_s=0.9585).

Comment

The results show that there is a correlation between antibiotic resistance in coliform organisms in urine samples and the use of antibiotics by a general practice. This is the first survey to suggest that geographically localised effects from antibiotic use occur in communities.

The dynamics of the emergence, spread, and maintenance of antibiotic resistance in populations are still unclear.¹ Much of the prescribing described here is likely to have been related to treatment of respiratory infections, and this may have been an important factor in determining the observed resistance. Resistance could be occurring through the prior selection of antibiotic resistant coliform organisms in the faecal flora² of patients presenting with urinary infections or by transmission of such organisms by others in the community.

Coselection of resistance to trimethoprim and ampicillin is explainable. Transmissible plasmids that code for combined resistance to ampicillin and trimethoprim are common in Escherichia coli³; therefore, selection pressure for resistance to one of these antibiotics is likely to select for resistance to the other.

The association between prescribing and resistance could have been caused by sampling bias if practices that had high rates of prescribing antibiotics were more selective in submitting samples for analysis, reserving testing for cases of treatment failure or complicated urological problems. In such cases it is likely that increased resistance would have been associated with lower rates of sampling. This was not observed. Few of the species of the coliform isolates were identified. Most were probably Escherichia coli, and it is unlikely that the small proportion of other coliform organisms (which often show broader resistance to antibiotics) would vary significantly between practices and be associated only with surgeries with high rates of prescribing. Other confounding variables may explain the observed correlations but it seems probable that the relations reflect a causal Darwinian association between prescribing and resistance.

These findings bring the debate on prescribing in the community from the national to the local level and provide preliminary evidence that practitioners may have to face the broader consequences of their antibiotic prescribing among their own patients.

Table.

P value and Spearman's correlation coefficient (95% confidence interval) for use of antibiotics and rate of antibiotic resistance (excluding repeat isolates) among general practitioner surgeries in Wales, 1996-8

Antibiotic used	Ampicillin			Co-amoxiclav			Cephalosporin*			Trimethoprim			Ciprofloxacin			Mean (range) No of prescriptions/ 1000 patients/year		Mean (range) percentage of organisms resistant to antibiotics†
P value	rs	P value	rs	P value	rs	P value	rs	P value	rs
Amoxicillin	<0.001	0.394 (0.269 to 0.506)		0.659			0.527			0.015	0.172 (0.033 to 0.304)		0.376			352 (81-912)		53.2 (34.2-67.6)
Co-amoxiclav	0.576			0.714			0.145			0.936			0.309			75 (7-239)		8.4 (0.0-26.1)
Cephalosporin	0.108			0.753			0.018	0.167 (0.028 to 0.299)		0.134			0.478			107 (16-532)		6.5 (0.0-19.0)
Trimethoprim	0.004	0.204 (0.066 to 0.334)		0.003	0.229 (0.081 to 0.367)		0.001	0.230 (0.094 to 0.358)		<0.001	0.331 (0.201 to 0.450)		0.056			63 (13-215)		26.3 (12.7-44.3)
Quinolones	0.069			0.873			0.103			0.306			0.015	0.225 (0.045 to 0.391)		34 (7-96)		2.2 (0.0-10.1)
All antibiotics	<0.001	0.319 (0.188 to 0.439)		0.581			0.411			0.031	0.153 (0.014 to 0.286)		0.141			971 (400- 1739)

^*Resistance to cefalexin was identified at six laboratories and resistance to cefradine at one. 

^†Overall mean of resistance rates for individual surgeries based on number of isolates varying from 50 to about 800.