Prospective case-control study of role of infection in patients who reconsult after initial antibiotic treatment for lower respiratory tract infection in primary careBMJ 1997; 315 doi: https://doi.org/10.1136/bmj.315.7117.1206 (Published 08 November 1997) Cite this as: BMJ 1997;315:1206
- John Macfarlane, consultant physiciana,
- Janet Prewett, research administratora,
- Donald Rose, consultant radiologista,
- Philip Gard, principal in general practiceb,
- Richard Cunningham, senior registrar in microbiologyc,
- Pekka Saikku, research professord,
- Stephanie Euden, research techniciane,
- Steven Myint, professor of clinical microbiologye
- a Respiratory Infection Unit, Nottingham City Hospital, Nottingham NG5 1PB
- b Arnold Health Centre, Nottingham NG5 7BQ
- c Department of Microbiology and Public Health Laboratory Service, University Hospital, Nottingham NG7 2UH
- d Chlamydia Laboratory, National Public Health Institute Department in Oulu, 90101 Oulu, Finland
- e Department of Microbiology and Immunology, University of Leicester, Leicester LE1 9HN
- Correspondence to: Dr Macfarlane
- Accepted 1 August 1997
Objective: To assess direct and indirect evidence of active infection which may benefit from further antibiotics in adults who reconsult within 4 weeks of initial antibiotic management of acute lower respiratory tract infection in primary care.
Design: Observational study with a nested case-control group.
Setting: Two suburban general practices in Arnold, Nottingham, over 7 winter months.
Subjects: 367 adults aged 16 years and over fulfilling a definition of lower respiratory tract infection and treated with antibiotics. 74 (20%) patients who reconsulted within 4 weeks for the same symptoms and 82 “control” patients who did not were investigated in detail at follow up.
Main outcome measures: Direct and indirect evidence of active infection at the time of the reconsultation or the follow up visit with the research nurse for the controls. Investigations performed included sputum culture, pneumococcal antigen detection, serial serology for viral and atypical pathogens and C reactive protein, throat swabs for detecting viral and atypical pathogens by culture and polymerase chain reaction, and chest radiographs.
Results: Demographic and clinical features of the groups were similar. Two thirds of the 74 patients who reconsulted received another antibiotic because the general practitioner suspected continuing infection. Any evidence of infection warranting antibiotic treatment was uncommon at reconsultation. The findings for the two groups were similar for the occurrence of identified pathogens; chest x ray changes of infection (present in 13%); and C reactive protein concentrations, which had nearly all fallen towards normal. Only three patients in the reconsultation group had concentrations ≥40 mg/l. Pathogens identified at follow up in the 156 patients in both groups included ampicillin sensitive bacteria in six. Atypical infections diagnosed in 27 (Chlamydia pneumoniae in 22) and viral infections in 54 had probably been present at the initial presentation.
Conclusion: Our study suggests that active infection, which may benefit from further antibiotics, is uncommon in patients who reconsult after a lower respiratory tract infection, and a repeat antibiotic prescription should be the exception rather than the rule. Other factors, such as patients' perception of their illness, may be more important than disease and infection in their decision to reconsult.
Lower respiratory tract infections are very common, but even if they have been given antibiotics, a fifth to a quarter of patients reconsult and many receive further antibiotics
No demographic or clinical features at presentation identify those who may reconsult
Direct and indirect evidence of infection warranting antibiotics is uncommon in those who reconsult and no different to those who do not
Chlamydia pneumoniae is the commonest infection identified in this study population
Antibiotics should be the exception rather than the rule for patients who reconsult
Acute respiratory infections are the commonest reason for consulting a general practitioner, and over 30 million courses of antibiotics are prescribed annually for their treatment.1 Despite this management, up to a quarter of adults treated for lower respiratory tract infections return to see their doctor, and most receive a further course of antibiotic, suggesting concerns about continuing infection.2 3 4 Reconsultation causes considerable inconvenience for the patient and increased workload for the doctor and contributes to the documented excess use and costs of antibiotics in the community.5
We suggest that active infection is not the reason why patients with lower respiratory tract infection reconsult, and in an observational study including a nested case-control group, we investigated whether there is direct or indirect evidence of persisting or new infection which may justify the prescription of a further antibiotic in adults reconsulting after initial antibiotic management of lower respiratory tract infection in general practice.
Subjects and methods
A stable suburban population of 20 848 adults aged 16 and over is served by two general practices with six partners in Arnold, Nottingham (Stenhouse Medical Centre and Arnold Health Centre). Both practices are experienced in conducting research into lower respiratory tract infection and identifying all suitable patients, having participated in several studies.2 3 6 We studied all adults who presented to the practices with acute community acquired lower respiratory tract infection between October 1994 and April 1995. The study was approved by the Nottingham City Hospital ethics committee. The case definition of lower respiratory tract infection was as previously reported2 3 and included a new or increasing, productive cough, associated with another symptom or sign of lower respiratory tract involvement including dyspnoea, wheeze, chest pain, or new signs on chest examination; and one or more constitutional symptom, including fever, sweating, headaches, aches and pains, sore throat, or coryza. Patients had not been given an antibiotic in the past 14 days but had been prescribed one for the illness at the consultation. The primary outcome measure was reconsultation with the doctor for the same illness within 28 days. Patients who returned for other reasons or only to obtain sickness certification were not considered to have reconsulted.
Each patient who entered the study had a standard history taken and examination performed by the doctor with previously designed data sheets,2 3 provided informed consent, and completed a questionnaire. Both the doctor and the patient recorded independently an estimate of the severity of the illness on the basis of a whole number scale from 0 (perfectly well) to 5 (very severe). An initial blood sample for serology was taken from each patient. The choice of antibiotic was left to the doctor.
For those who reconsulted, the doctor filled out a standard follow up data sheet recording details of symptoms and signs, management including further antibiotic treatment and investigations, and the doctor's view as to why the patient reconsulted.
Nested case-control study
A nested case-control study was included to investigate in detail clinical and microbiological factors associated with reconsultation after initial treatment. To provide a “control” group, every third patient entered into the study was asked to come back to see the research nurse about 10 days later. Our previous studies had shown that this was the median time between initial consultation for lower respiratory tract infection and reconsultation.2 3
Patients who reconsulted and control patients were investigated in a standard way when they returned. Sputum samples (assisted by nebulisation of isotonic saline when necessary), throat swabs, and a second blood sample for serology were taken. Chest radiographs obtained within 3 days of the return visit were reviewed “blind” by a radiologist (DHR) experienced in research into lower respiratory tract infection.2 3 6 Patients returned a month later for a third blood sample for serology.
Samples were held at 4°C and transported rapidly to the laboratory for standard investigations.2 Isolation of bacterial respiratory pathogens from diluted sputum or the detection of pneumococcal capsular antigen was taken as evidence of infection, as was a fourfold rise in antibody titre to viral and atypical pathogens. Antibodies to Chlamydia pneumoniae were measured by microimmunofluorescence in the National Public Health Institute Laboratory in Oulu, Finland, by using C pneumoniae AR 39 or Kajaani 6 epidemic strains, or both.7 A fourfold or greater rise in antibody titre, a high titre of 2048 (for IgG or IgA), or an IgM titre of 16 or more was considered evidence of infection (either reinfection or primary infection if IgM was present initially). C reactive protein concentrations were measured by the turbidometric method; values of 40 mg/l or over were taken as indirect evidence of active bacterial infection.8 9 10 11
Throat swabs were collected into ice cold sterile phosphate buffered saline and human placental ribonuclease inhibitor and subsequently stored at −70°C in 10% dimethyl sulphoxide. All samples were inoculated into tubes of MRC5, C16, and Ohio HeLa cells. Influenza viruses A and B, parainfluenza, adenovirus, and respiratory syncytial virus were identified by culture and confirmed by immunofluorescent staining. Aliquots were examined by gene amplification by using the nested polymerase chain reaction for coronavirus, human rhinovirus, Mycoplasma pneumoniae, and C pneumoniae based on previously published methods.12 13 14
Levels of significance were computed with the χ2 test for categorical variables and Students' t test for continuous variables.
Findings at initial presentation
The general practitioners treated 440 patients for lower respiratory tract infection during the 7 winter months of the study. Seventy three patients (mean age 53 years; 40% men) were not entered into the study because they were unwilling or unable to participate (49), no research nurse was available (19), and for other reasons (5).
Details of the cohort of 367 patients studied are shown in table 1. Nearly two thirds had been previously fit and well; the commonest pre-existing conditions present in 40% included chronic lung disease in 82 and cardiac disease in 45.
The doctors' subjective illness severity score for the patients was a score 1 for 49 (13%), 2 for 157 (43%), 3 for 139 (38%), 4 for 22 (6%). The patients' scores were higher (that is, recording more troublesome symptoms) than the corresponding doctors' scores on 54% of occasions, the same for 43%, and less for only 3%. Amoxycillin was prescribed for 333 patients (91%); 29 (8%) received erythromycin and 5 (1%) received other antibiotics.
Patients who reconsulted
A total of 20% of patients (74/367) reconsulted with their general practitioner after a median of 9 (interquartile range 7-15) days. Of these, nine reconsulted a further time, making a total of 23% (83/367) extra consultations for the same symptoms within 4 weeks of the index consultation. The reasons patients gave were: symptoms no better or worse (24; 33%), symptoms only marginally improved (48; 65%); and antibiotic side effects (2; 2%). Patients had one or more continuing problems: cough with discoloured sputum (61%), dyspnoea (55%), wheeze (46%), general malaise (59%), and signs on chest examination (29%).
The doctors assessed that 40 (54%) patients had continuing infection and 19 (26%) had “postinfective reactivity,”; in 11 (15%) the infection was gone but symptoms were slow to clear, and four had other problems. They prescribed antibiotics to 62% (46/74) of patients: macrolide for 34, cephalosporin for 8, amoxycillin for 1, coamoxiclav for 1, and others for 2. Only one patient was referred to the hospital outpatient clinic.
Control patients in nested case-control study
One hundred and twelve patients were identified as potential controls at their initial visit. Twenty (18%) of these arranged to see the doctor again and so joined the “reconsultation” group and 10 did not turn up for the follow up visit, leaving 82 in the “control” group. They returned to see the practice research nurse after a median of 10 (interquartile range 8-14) days. The patients in the control group were selected in a systematic rather than random way, but they did not differ in demographic and clinical characteristics from those who did not reconsult. They therefore seemed to be representative of the population of patients who did not return to see the doctor.
Comparative features of patients who reconsulted and those who did not return to see the GP
The demographic and clinical features of the 74 patients who reconsulted were similar to the 211 who did not return and the 82 control patients (see table 1).
More patients who reconsulted than control patients had spent time in bed after initial consultation (19/64 v 9/72; P<0.01), and 47 (73%) recorded restriction of normal activities due to the illness compared with 31 (43%) control patients (P<0.001).
Results of investigations at follow up
Direct and indirect evidence of infection likely to benefit from antibiotics was uncommon in both groups at follow up (table 2). Bacterial pathogens, all ampicillin sensitive, were cultured from only six samples. No coliforms were isolated. Twenty seven atypical infections were diagnosed serologically, most commonly Chlamydia pneumoniae (primary infection in six, reinfection in the remainder). Chlamydial infection was not detected in any throat swabs by polymerase chain reaction. Of the 54 viral infections detected in 46 patients, the diagnosis was made by sequential serology in 21, viral culture in 9, and polymerase chain reaction in 25. One patient had positive results on both culture and serology (to influenza B).
The number of chest radiographs showing changes suggestive of infection at follow up was similar both for those who reconsulted (11/70; 15%) and the control group who did not (9/80; 11%), as were the concentrations of C reactive protein at initial consultation and at follow up. Overall 19/156 (12%) of both groups had concentrations of 40 mg/l or more at initial consultation as did 3/72 (4%) of those who reconsulted, but none of the controls, at the time of follow up. Only one patient (in the reconsultation group) showed a rise in concentration (26 mg/l to 68 mg/l) between the time of the initial contact and reconsultation. All other values had fallen or returned to normal at follow up apart from one patient in the reconsultation group and one control patient in whom there was no significant change (17 mg/l to 18 mg/l and 23 mg/l to 25 mg/l, respectively).
Of the 39 patients thought by the doctor to have continuing infection at reconsultation and given antibiotics, only three had changes consistent with infection on chest radiography and only three had a C reactive protein concentration of 40 mg/l or more. Two had evidence of bacterial infection, 10 of viral infection, and five of atypical infection on investigation.
We confirmed that lower respiratory tract infection was common in the adult population presenting to this group of general practitioners, with an annual incidence of 36 per 1000. This is similar to our previous finding of 44 per 1000 population, an estimate which, unlike in this study, also included patients seen on home visits.2 Our study only ran over 7 winter months, but the results are probably applicable to the whole year as most respiratory pathogens, apart from legionellae, are commoner in the winter. Longitudinal studies of several years would be needed to pick up cyclical epidemics of pathogens such as Mycoplasma pneumoniae, which may influence the findings.15
Selection of controls
We chose our control group in a systematic way at the time of the initial consultation as we considered that it would be impractical either to ask all patients to return to see the research nurse for full investigations or to wait until a patient reconsulted before identifying a matched control and asking them to attend immediately for investigations. This method raises the possibility of bias, but we were satisfied that the control patients were similar to and seemed representative of those who did not return, and the interval between initial consultation and a second visit to the surgery was similar for those who reconsulted and the control group. In addition, the same proportion of patients identified as controls at the initial visit arranged to reconsult with their doctor as did those not asked to be controls.
One in five patients reconsulted at least once within the month after initial antibiotic treatment for lower respiratory tract infection because they were not satisfied with their progress. The demographic and clinical features at initial presentation were similar for those who did and did not need to reconsult. During this reconsultation, the doctor prescribed a further course of antibiotics to nearly two thirds, usually because continuing infection was considered likely. We found that the doctors' impression of continuing infection, however, was rarely matched by objective evidence of infection. Near complete investigations in both those who reconsulted and control patients showed similar patterns with little direct or indirect evidence of active infection warranting antibiotics in either group. The numbers in each group were, however, too small to allow more accurate analysis of this observational data.
Infections identified in cases and controls
In only three of the patients who reconsulted were bacterial pathogens identified and none were ampicillin resistant. We did not find colonisation and infection by coliforms or other antibiotic resistant respiratory pathogens after initial antibiotic treatment to be a problem in this community. The limitations of sputum culture are well known, but it can be a valuable investigation in the primary care setting, even if antibiotics have been given.16
Chlamydial infection, largely due to C pneumoniae, was diagnosed serologically in 15% of all patients studied. This confirms reports that this is a not uncommon cause of lower respiratory tract infection in adults in the community.7 17 18 It also provides some support to the doctors' choice of a macrolide for those patients who reconsult, although macrolides seem to offer no advantage over amoxycillin for initial treatment for lower respiratory tract infection.3 Our detection rate of viral infection in a third of cases was considerably higher than in our previous study of lower respiratory tract infection,2 largely because of better detection methods, including polymerase chain reaction, which doubled our diagnostic rate. We speculate that the viral and atypical infections were present at initial consultation rather than representing a secondary infection resulting in reconsultation. This is supported by the equal incidence of these infections in both reconsultation and control patients and the absence of chlamydial antigen and infrequency of viral isolation from throat swabs at relapse. It has been suggested that viral infections of the upper respiratory tract increase bacterial colonisation of the nasopharynx and the risks of secondary bacterial infection of the lower respiratory tract,19 but we found no evidence for this.
Indirect evidence of active infection
C reactive protein concentrations were similar for reconsultation and control groups, and only 4% of patients had a concentration of 40 mg/l or over at reconsultation, suggesting that active bacterial infection was unusual in this group. These concentrations seem useful in differentiating viral and bacterial lower respiratory tract infection in adults20 21 22 and children,8 9 10 11 for guiding the decision about antibiotic treatment,21 and for assessing severity of infection.22 Concentrations at reconsultation had risen in only one patient, suggesting resolution of infection in the others. Initial C reactive protein concentrations or chest radiographic changes consistent with infection, when taken as markers of severity of infection, did not identify those more likely to reconsult. The proportion of all patients who had radiographic changes consistent with infection at follow up (13%) was the same as the proportion (12%) found at initial presentation in our previous study of lower respiratory tract infection,2 suggesting that the changes were probably a feature of the primary illness rather than a later complication causing reconsultation.
Other factors affecting reconsultation
Clearly, other factors cause some patients to reconsult and some doctors to prescribe further antibiotics.23 24 25 26 Patients' anxieties and previous consulting habits may be more important than factors related to infection.27 In a recent study of sore throat, “legitimation” of the illness—to explain to work or school in 60% of cases and to family and friends in 37%—was an important reason for consultation.28 The illness had considerably more impact on the quality of life of the patients who reconsulted as measured by days in bed and inability to perform normal activities. The patient's perception that an “infection” is the problem and antibiotics the answer29 may result in dissatisfaction at the speed of the natural recovery of the illness, particularly after treatment has been completed. The median reconsultation time was 2 days after antibiotics finished. Over half of the patients judged their illness to be more severe than their doctor did and very few less so, as has been noted before in general practice.24 The prescription of a further antibiotic may perpetuate the concept of infection to the patient and heighten the consulting habit for the next episode of lower respiratory tract symptoms, producing a cycle of consultations powered by antibiotic prescriptions. In the absence of an alternative strategy, and when prescribing decisions are invariably made without the aid of investigations in general or the identification of an infection in particular, doctors have an uphill, time consuming, and difficult task to educate their patients not to require antibiotics. The results of our study may aid them in this task.
Our study shows that infection likely to benefit from antibiotics is uncommon in patients who reconsult after lower respiratory tract infection, and it suggests that a repeat antibiotic prescription should be the exception rather than the rule. If further antibiotics are judged necessary, a macrolide seems most appropriate. We found nothing to support the use of other groups such as ß lactamase stable ß lactams or quinolones. Our results should increase the confidence with which doctors can discuss these issues with their patients and thus contribute to reducing unnecessary antibiotic prescribing in the community.
We are most grateful to the following general practitioners and nurses from the practices who participated, including G Bajek, M Broadley, S Bolsher, A Cockburn, F Coutts, M Elliott (Dr), M Elliott, B Hammersley, R Howard, C Leiper, S Marie-Jeanne, G Mills, S Newton, S Patel, P Pavier, I Schofield, M Smith, and D Thornhill. We also thank the departments of biochemistry, radiology, and microbiology for their cooperation; Dr A Guion, Dr W Holmes, and Mr T Brown for help with study design; and Mr A Barnett, medical statistician from the Medical Research Council's epidemiology and medical care unit, for help with data presentation.
Funding: This study was supported by a project grant from the British Lung Foundation. Abbott Laboratories provided a grant for the serological testing for C pneumoniae.
Conflict of interest: None.