Reattendance and complications in a randomised trial of prescribing strategies for sore throat: the medicalising effect of prescribing antibioticsBMJ 1997; 315 doi: http://dx.doi.org/10.1136/bmj.315.7104.350 (Published 09 August 1997) Cite this as: BMJ 1997;315:350
- P Little, GP Wellcome training fellowa,
- C Gould, research assistanta,
- I Williamson, senior lecturer in primary carea,
- G Warner, general practitionerb,
- M Gantley, anthropologista,
- A L Kinmonth, professor of primary medical carea
- a Primary Medical Care Faculty of Health, Medicine, and Biological Sciences Aldermoor Health Centre Southampton University Southampton SO16 5ST
- b Nightingale Surgery Romsey Hants SO16 5ST
- Correspondence to: Dr Little
Objective: To assess the medicalising effect of prescribing antibiotics for sore throat.
Setting: 11 general practices in England.
Design: Randomised trial of three approaches to sore throat: a 10 day prescription of antibiotics, no antibiotics, or a delayed prescription if the sore throat had not started to settle after three days.
Patients: 716 patients aged 4 and over with sore throat and an abnormal physical sign: 84% had tonsillitis or pharyngitis.
Outcome measures: Number and rate of patients making a first return with sore throat, pharyngitis, or tonsillitis. Early returns (within two weeks) and complications (otitis media, sinusitis, quinsy). Outcomes were documented in 675 subjects (94%).
Results: Mean follow up time was similar (antibiotic group 1.07 years, other two groups 1.03 years). More of those initially prescribed antibiotics initially returned to the surgery with sore throat (38% v 27%, adjusted hazard ratio for return 1.39, 95% confidence interval 1.03 to 1.89). Antibiotics prescribed for sore throat during the previous year had an additional effect (hazard ratio 1.69, 1.20 to 2.37). Longer duration of illness (> 5 days) was associated with increased return within six weeks (hazard ratio 2.90, 1.70 to 4.92). Prior attendance with upper respiratory conditions was also associated with increased reattendance. There was no difference between groups in early return (13/238 (5.5%) v 27/437 (6%)), or complications (2/236 (0.8%) v3/434 (0.7%)).
Conclusions: Complications and early return resulting from no or delayed prescribing of antibiotics for sore throat are rare. Both current and previous prescribing for sore throat increase reattendance. To avoid medicalising a self limiting illness doctors should avoid antibiotics or offer a delayed prescription for most patients with sore throat.
Sore throat is one of the commonest presentations of upper respiratory illness in primary care and attendence is increasing
Complications are rare with no, or delayed, antibiotic prescription
Prescribing antibiotics increases reattendance for future episodes
Unless patients are very ill general practitioners should consider exploring concerns, explain the natural history, and avoid or delay prescribing antibiotics
A systematic review showed only marginal benefit from antibiotics for sore throat,1 which must be balanced against side effects, including “medicalisation”2—making people more likely to seek medical care for future illness. Assessing medicalisation requires open randomised trials.3 We previously reported that antibiotic prescribing increases patients' belief in antibiotics and intention to reconsult, compared with either not prescribing or offering a delayed prescription.3 Immediate prescribing should, therefore, increase reattendance. We report here a comparison of the complication and reattendance rates of patients given immediate prescriptions and those managed by other strategies and report other predictors of reattendance.
The methods have been fully reported:3 716 patients with sore throat and an abnormal physical sign were randomised to three prescribing approaches supported by advice sheets. These approaches were: (a) a 10 day prescription of phenyoxymethylpenicillin, (b) no antibiotics, and (c) a 10 day prescription of antibiotics to collect if the sore throat had not started to settle after three days. Groups were well differentiated in reported antibiotic use, attitude, and intention.3
Notes review—All patients had their notes inspected in summer 1996 (follow up 2 months to 2.5 years after trial entry) for subsequent presentations. We counted all entries where sore throat, pharyngitis, tonsillitis, otitis media, or quinsy was noted either in the main complaint or in the clinical description. Attendances before randomisation were also noted, and whether antibiotics had been prescribed.
Observer bias—Details of the randomised episode were removed from photocopies of 75 randomly chosen sets of notes from the largest practice, the notes reassessed, and the data re-entered. The original assessment was compared with the blinded assessment.
Sample size calculation (for 80% power, 95% confidence using the epi info program)—For the principal comparison (the initial antibiotic group versus the other two groups) a 40% relative increase in reattendances2—or a 15% absolute increase (antibiotics 40%, others 55%; hazard ratio 1.38)—required 417 patients, or 596 allowing for 30% loss to follow up.
Data entry and analysis—Data were entered and analysed on an intention to treat basis using SPSS and STATA for windows. Cox proportional hazard regression was used to estimate hazard ratio of first return to the surgery for sore throat, pharyngitis, or tonsillitis (a “failure”), data being censored then or at the end of follow up. We tested predictive features in the model using the likelihood ratio χ2 test by forward selection of significant terms (at the 5% level), terms being retained if there was no evidence of significant multi-collinearity. The proportional hazards assumption—that the effect of regressors does not vary with time—was assessed using interaction terms according to three follow up periods (0-45, 46-179, >179 days) defined by tertiles of the distribution of returns to the surgery.
Notes were reviewed for 675 (94%) subjects. Results are presented for the intial antibiotic group versus the other two groups. Group characteristics (table 1) and mean follow up time were similar (1.07, 1.03 years respectively, P=0.2). Prescribing antibiotics increased return to the surgery (38% versus 27%), with an additional effect from previous prescribing (see table 2). A longer duration of illness increased the return rate, confined to the first follow up period (hazard ratios for 0-45, 46-179, and >179 days respectively 2.86, 0.83, 1.13, likelihood ratio χ2 (2 df) 11.05, P=0.0004); 23 of these 34 returns (68%) occurred within two weeks and 17/34 (50%) within eight days. Increasing prior attendance with upper respiratory illness was also associated with future reattendance for sore throat (z test for trend 2.03, P=0.04); this effect was not confounded by the frequency of sore throats, as reported by patients in the end of study questionnaire. There was no difference between the antibiotic and other groups in the proportion of early returns (respectively 13/238 (5.5%) v 27/437 (6%)) or complications (otitis media, sinusitis, quinsy: 2/236 (0.8%) v3/434 (0.7%)). Sociodemographic or psychosocial factors measured at the index consultation did not predict future reattendance (table 3).
Although the principal analysis compared the initial antibiotics groups with the other two groups, the “delayed” group had the lowest rates of reattendence (hazard ratio of reattendence: delayed 1.00, no antibiotic 1.3 (95% confidence interval 0.86 to 1.97), antibiotic 1.61 (1.09 to 2.38)).
General practitioners' attitude to prescribing antibiotics at the end of the study (very, moderately, slightly, not at all comfortable prescribing antibiotics immediately) did not significantly predict reattendence (hazard ratio 1, 1.24, 0.77, 0.90; z trend -0.88, P=0.39).
Outcome documentation bias—We minimised possible bias from general practitioners' relabelling episodes of sore throat by assessing the main text of the consultation and not just the problem summary. If relabelling bias was significant the difference between antibiotic and other groups should have disappeared when all other upper respiratory illnesses were included (sore throat, pharyngitis, tonsillitis, quinsy, and other upper respiratory labels—URTI, cold, otitis media, sinusitis—where no sore throat was documented), but it did not (respectively 106/238 (45%) and 146/437 (33%)).
Observer bias—Only 5/73 (7%) disagreements occurred between the unblinded and blinded assessment of notes (agreement 68/73 (93%), κ 0.83), all in the no antibiotics or delayed group, and there was no evidence of systematic bias (three attendances coded unblinded as no sore throat became sore throat, and two were recoded in the opposite direction).
This trial confirms that complications of sore throat are rare, and that prescribing antibiotics increases reattendance.2 8 We have shown that trial groups were similar, well differentiated,3 and that selection,3 non-response, relabelling, general practitioners' attitude, and outcome assessment variation are not likely to bias the results.
Consultations for respiratory conditions in British general practice have increased by 14% in 10 years.9 Since the pathogenic basis is not likely to be changing,1 10 11 psychosocial factors may explain changes in attendance—for example, patient expectations, altered social support networks, employers' attitudes, or changes in doctors' behaviour.12 This study suggests that an effective way of counteracting increased consulting is for general practitioners not to prescribe antibiotics—or delay prescribing them—for self limiting illness in individuals who are not very ill and where complications are rare.
Prolonged duration of sore throat also increased the rate of return within six weeks of the original illness, and half the subjects with a longer duration of illness who returned did so within eight days. This suggests that explanation of the natural history—that the average duration is five days after consultation and that almost 40% of people have a sore throat for longer than five days—may reduce expectations and possibly alter subsequent attendance behaviour. General practitioners should consider careful counselling in subjects who have attended with upper respiratory illness twice or more in the past year, as this is a marker for reattendance.
We thank these general practitioners for their help in recruitment, advice, and enthusiasm: Nigel Dickson, Graham Newman, Peter Willicombe, Peter White, Sue Tippett, Richard Peace, Julie Chinn, Katie Warner, Neil Ball, Tim Taylor, Deidre Durrant, Mark Rickenbach, Sally Bacon, Tim Whelan, Peter Markby, Simon Goodison, D Traynor, R Briggs, Evelyn Beale, Fiona Bradley, Simon Smith, and Stephen Morgan.
Funding: Wessex NHS regional research and development funds. PL is supported by the Wellcome Trust.
Conflict of interest: None.