Setting
Eighteen general practitioners (GPs), nine from the Arkhangelsk region and nine from the Murmansk region, from both urban and rural offices, were randomized into intervention and control groups. The trial was conducted over 12 weeks (from 30 January to 30 April 2010). All registrations were made by GPs in their offices.
Study population
Patients with acute cough/LRTI (including acute bronchitis, pneumonia, and infectious exacerbations of COPD or asthma) were included. Other inclusion criteria were age 18 years or older, an illness of less than 28 days duration, first consultation for the illness episode, being seen in a physician’s office, and written consent to participate. Exclusion criteria were an inability to fill out study documentation, being previously included in the study, immunocompromised status (HIV patients, immunosuppressive treatment), and ongoing treatment with oral corticosteroids.
Design
Cluster randomization was performed with GPs as units with SPSS 18.0 (IBM, Armonk, NY, USA). The rationale for using a cluster design was to ensure consecutive recruitment of patients and to avoid the situation where experience gained from the use of the test contaminated the care of control patients.
The second author generated the allocation sequence using SPSS, and the first author enrolled the clusters and made a list of clusters. Based on this list and using the allocation sequence, the first author assigned clusters to interventions.
All the GPs worked in separate outpatient departments (polyclinics), some in single GP offices and others within a GP partnership (with doctors who did not participate in this study). The sample sizes were based on a hypothesis of 20% reduction in antibiotic prescribing in the intervention group compared with the control group. Based on the chi-square statistic, the required sample size in each group was 72 participants (with a power of 90% and a risk of false positive difference less than 5%).
Two months before the trial, a baseline study without CRP testing was conducted that included 13 of the 18 participating GPs, using the same case report form (CRF) and examination. This allowed observation of prescription rates before and after the clinical trial, serving as a sensitivity analysis.
The CONSORT checklist and flow diagram are included as additional files.
The case report form (CRF)
The CRF was similar to that used in the GRACE study (Genomics to combat Resistance against Antibiotics in Community-acquired lower respiratory tract infections in Europe [24]), describing symptoms, findings, and treatment in LRTI [4]. The GPs reported the following 15 symptoms in a questionnaire: cough, sputum production, shortness of breath, wheeze, coryza (blocked/runny nose), fever during this illness, chest pain, muscle aches, headache, disturbed sleep, feeling generally unwell, interference with normal activities, confusion/disorientation, and diarrhoea. If increased sputum production was reported, the colour of the sputum was recorded. Symptom severity scores were calculated using the scores for 13 symptoms (similar to that computed in the GRACE study) [4]. The categories for clinicians to rate the severity of each symptom as “no problem”, “mild problem”, “moderate problem”, or “severe problem” were scored 1, 2, 3, and 4, respectively. Scores were calculated for patients with a minimum of 85% (that is, 12 of 14 symptoms) of their symptoms recorded. This score was scaled to range between 0 and 100 so that it could be interpreted as a percentage of maximum symptom severity.
Other variables registered included sex, age and smoking status (never smoker, previous smoker and current smoker). For previous and current smokers, the average number of cigarettes per day and the number of smoking years were recorded. The pack-year criterion was calculated: one pack-year of smoking would mean that someone had smoked one pack of cigarettes (20 cigarettes) daily for one year.
The clinical examination included a chest examination and axillary temperature. The following chest findings were recorded: diminished vesicular breathing, wheezes, crackles and rhonchi.
After the clinical examination, the GPs recorded their provisional diagnosis, choosing from the following: URTI, acute bronchitis, pneumonia, COPD, asthma, non-infectious cough, and other diagnosis.
The comorbidities registered were: 1) pulmonary diseases, including COPD, asthma, tuberculosis, bronchiectasis, lung cancer, and other lung disease; 2) heart diseases, including heart failure, ischaemic heart disease, and other heart diseases (e.g., valvular lesions, cardiomyopathy); 3) diabetes; and 4) other chronic diseases.
Optional examinations
Chest radiography was accessible for all patients, and other investigations (e.g., culture of sputum, spirometry, electrocardiogram) could be ordered when necessary. All suspected pneumonia cases were confirmed by chest radiography, and films were routinely reviewed by specialists at radiology departments.
CRP testing
Before the clinical trial began, all GPs participated in two vocational training sessions concerning the CRP test, including theoretical and practical information. They were given guidelines about the interpretation of CRP results. This information included a summary of the literature on RTI and the role of CRP; paper cases of patients with different RTIs and different CRP values were discussed. They were told that antibiotics were usually not needed when the CRP value was below 20 mg/L and that a prescription could be indicated for CRP values above 50 mg/L, taking into account the duration of illness [15]. However, the management, including antibiotic treatment, should be decided for each patient on an individual basis.
The CRP test was performed in the intervention group at both the first and second consultations. The Afinion test system (Axis Shield) was used, which provides results within 5 minutes and before treatment is determined. This test is based on solid-phase sandwich immunometric analysis. The measurement range in whole blood samples is 8–200 mg/L. Test kits were supplied by Axis Shield.
Treatment
GPs could prescribe any treatment, including antibiotics and other drugs for the cough (e.g., cough mixture) and additional medication if deemed necessary. They were told that medication should be prescribed after the clinical examination (and after the CRP test in the intervention group), without waiting for chest radiography results.
Outcome measures
The primary outcome was the antibiotic prescribing rate. Secondary outcomes were referral to radiography and rate of recovery at the follow-up consultation after 2 weeks with the following five alternatives: “fully recovered”, “almost recovered”, “slightly improved”, “unchanged”, and “worse”. Reconsultations (another consultation with the GP within 2 weeks) and complications (in need of hospitalization) were recorded.
Statistical analysis
CRP values were divided into three groups: CRP < 20 mg/L, 20–50 mg/L and ≥ 50 mg/L. Differences in the rate of prescribing antibiotics and referral to chest radiography were calculated, as was the percentage of participants who stated that they had recovered or almost recovered after 2 weeks. Differences in patient characteristics at inclusion between the intervention and the control group were analysed to check for recruitment bias. Changes in prescription rate between baseline and the clinical trial for each GP were analysed as part of the sensitivity analyses. The chi-square test was used to assess differences, and a P-value < 0.05 was considered to indicate significance. As a second sensitivity analysis, the predictive value of CRP testing for antibiotic prescribing was evaluated by multivariable logistic regression, in a model including other relevant explanatory variables.
Statistical analysis was performed using SPSS 18.0.
Ethics
All patients gave informed consent. The study was approved by the local Ethics Committee of the Northern State Medical University (Arkhangelsk, Russia).