Little is known about actual clinical practice regarding management of smokers compared with ex-smokers and nonsmokers presenting with acute cough in primary care, and whether a lower threshold for prescribing antibiotics benefits smokers.
This was a multicentre 13-country European prospective observational study of primary care clinician management of acute cough in consecutive immunocompetent adults presenting with an acute cough of ≤28 days duration.
There was complete smoking status data for 2,549 out of 3,402 participants. 28% were smokers, 24% ex-smokers and 48% nonsmokers. Smokers and ex-smokers had more chronic respiratory conditions (18.5% and 20.5% versus 12.5%). Median symptom severity scores were similar. Smokers were prescribed antibiotics more frequently (60%) than ex-smokers (51%) and nonsmokers (53%). After adjusting for clinical presentation and patient characteristics, the odds ratio of being prescribed antibiotics for smokers compared with nonsmokers was 1.44 (95% CI 1.12–1.86; p = 0.005). Patient recovery was not significantly different for smokers and nonsmokers, after adjusting for clinical presentation and patient characteristics.
Smoking status was used as an independent factor to determine whether or not to prescribe an antibiotic. Being prescribed an antibiotic was not associated with recovery in smokers.
Smoking increases risk for many diseases including cardiovascular and respiratory diseases and many types of cancer. People who smoke are more likely to die at an earlier age from most illnesses compared with nonsmokers 1.
Since clinicians are aware of the association between smoking and increased risk for disease and complications, they may treat smokers more aggressively when they develop a respiratory infection compared with otherwise comparable nonsmokers 2–4.
Smokers have more frequent respiratory tract infections than nonsmokers 5, 6. Smokers are also more likely to develop pneumonia than nonsmokers 5, 7, 8. There is no consensus about whether the course of an acute cough is prolonged in smokers; one observational study suggested there is no difference compared to nonsmokers 9, while another suggested that coughs last longer 10.
Antibiotic resistance is increasing, driven by antibiotic prescribing 11. Clinicians are therefore encouraged to better target antibiotic prescriptions to only those who will benefit, but there is insufficient evidence about which subgroups of patients with respiratory infections these are 12, 13.
This evidence gap is one reason why there is wide variation in antibiotic prescribing for both upper and lower respiratory tract infections 2, 3, 14–17. Our searches revealed no randomised placebo-controlled trials of antibiotic treatment for acute cough in smokers. A review using data on smokers included in larger trials concluded that there was insufficient data for meta-analysis. Subset analysis concluded that “any benefit of antibiotics for smokers is the same or less for smokers than for nonsmokers” 18.
We therefore used data from a 13-country, primary care observational study of acute cough presentation, management and outcomes to investigate whether clinicians were more likely to prescribe antibiotics for smokers compared with ex-smokers and nonsmokers presenting with acute cough, comparing symptom severity scores at presentation and recovery course for each smoking status category.
Genomics to combat Resistance against Antibiotics in Community-acquired lower respiratory tract infection (LRTI) in Europe (GRACE) is a network of excellence, with the aim of combating antimicrobial resistance through improved management of community-acquired LRTI (www.grace-lrti.org/portal/en-GB/). The first GRACE clinical study described differences in clinical presentation, management (more specifically, antibiotic prescribing) and recovery of patients consulting with an acute or worsened cough as the main or dominant symptom (or signs and symptoms suggestive of LRTI) in 14 primary care networks in 13 countries across Europe 19. We conducted further analyses on this data set to explore our study question.
Consecutive immunocompetent patients, aged 18 yrs or older, consulting for the first time in the illness episode with an acute cough of <28 days duration were eligible for inclusion. Following informed consent, the clinician completed a case report form (CRF) and the patient completed a daily symptom diary for 28 days. CRF data included presenting symptoms, as well as their severity, perceived diagnosis and management/treatment given. Patient diary data included symptom severity scores in the 28 days following the consultation, further demographic details and past medical history, along with information on smoking status. Diaries with complete smoking status data were matched to their respective CRF.
Three smoking status categories were defined (table 1⇓). Since there are no standard definitions of “ex-smoker”, we combined a definition used by the American Morbidity and Mortality Monthly cross-sectional survey to define an “ever-smoker” and combined this with a cut-off of ≥1 yr of stopping smoking because of the high relapse rates in smokers who re-start smoking soon after quitting 20, 21. Throughout all analyses, nonsmokers are used as the reference category.
Smoking status was also examined using pack-years 22. These were calculated for patients who smoked cigarettes only, since we did not have information on the weight of rolled tobacco smoked, nor is there consensus on the amount of tobacco in cigars or cigarillos. Patients who had never smoked have a pack-year score of zero.
Clinicians classified the severity of 14 symptoms for each patient on a four-point scale. These symptoms were: cough, phlegm, shortness of breath, wheeze, coryza, fever, chest pain, muscle ache, headache, disturbed sleep, feeling generally unwell, interference with normal activities, confusion/disorientation and diarrhoea. Patients were asked to rate the same symptoms except for confusion/disorientation and diarrhoea daily for up to 28 days. In addition, patients were asked about interference with social activities. Clinicians therefore rated 14 symptoms at presentation and patients rated 13 each day. All patient-reported symptoms were on a seven-point scale.
Both clinician-rated (from the CRF) and patient-rated (from the diary) total symptom severity scores were calculated by summing the individual symptom scores for each individual and scaling them to range between 0 and 100.
Descriptive statistics are presented for each smoking status category by network and overall. They were calculated using mean±sd, median (interquartile range (IQR)) and proportions as appropriate. Presented standard deviations have been inflated for clustering at the clinician level.
Two-level logistic regression (patients nested within clinicians) was used to assess the association between smoking status and antibiotic prescribing, controlling for each individual clinical symptom and sign (excluding cough, as this was present in 99.8% of patients), comorbidities, age and duration of illness prior to consulting. Results are presented as odds ratios and 95% confidence intervals. A sub-analysis was performed to assess the association between pack-years and antibiotic prescribing for those individuals for whom a valid pack-year estimate could be estimated.
A three-level hierarchical ARMA (1,1) model (daily symptom scores nested within individuals nested within clinicians) modelled individual total symptom severity scores (logged) over time. This model tested the association between smoking status and symptoms scores over time and interactions with antibiotic treatment. Again, a pack-year sub-analysis was performed to test the association between pack-years and outcome.
2,549 (75%) of the 3,402 recruited patients had complete clinical information and patient diary including information on their smoking status (fig. 1⇓).
702 (28%) patients were smokers. This ranged from 20 (14%) in the Southampton, UK network to 91 (42%) in the Łódź, Poland network. 616 patients (24%) were classed as ex-smokers (24 (8%) in the Balatonfüred, Hungary network to 61 (38%) in the Utrecht, the Netherlands network). 48% of patients were classed as nonsmokers (from 77 (36%) in the Łódź network to 196 (66%) in the Bratislava, Slovakia network) (table 2⇓).
There was a wide range in the number of cigarettes smoked per day among current smokers from <1 to 45 per day (mean±sd 11.4±7.55). The number of smoking years ranged from <1 to 62 yrs.
The median (IQR) age for smokers was 44 (33–56) yrs with nonsmokers being older (47 (34–60) yrs) and ex-smokers older still (56 (44–67) yrs). 273 (39%) patients of smokers were male, compared with 359 (29%) of nonsmokers and 280 (46%) of ex-smokers. Ex-smokers were more likely to suffer from coexistent respiratory conditions. Smokers had more respiratory illnesses compared with nonsmokers. Smokers and ex-smokers were unwell for a median of 5 days, with nonsmokers unwell for a median of 4 days before consulting their clinician (table 3⇓).
Over 99% of patients had a cough. The next most common presenting symptoms were phlegm production (from 73% in nonsmokers to 82% in smokers), feeling generally unwell (from 78% in ex-smokers to 82% nonsmokers) and interference with normal activities (from 68% in nonsmokers and 71% in ex-smokers). A higher percentage of smokers than nonsmokers presented with shortness of breath (55% versus 46%) and wheeze (40% versus 34%).
Overall, clinicians recorded an average of eight symptoms for patients from the three smoking categories. The number of symptoms patients presented with recorded ranged from one to 14 in each category. The median severity scores of the symptoms that smokers, ex-smokers and nonsmokers presented with was similar (31, 30 and 29 out of 100, respectively). The interquartile ranges were also very similar for all groups (6–10) (table 3⇑).
The 14 individual symptom severity scores rated by clinicians and ranging from scores of 0–4 were similar across all smoking status categories (fig. 2⇓).
Clinicians assigned each person a working diagnosis based on the history and examination findings. The proportions of patients with a working diagnosis of LRTI were not statistically different across the three smoking status categories (p = 0.12) (table 3⇑).
Clinicians were asked to respond to two statements: 1) “this patient wanted me to prescribe antibiotics for them”, and 2) “antibiotics will help this patient get better quicker” using a five-point Likert-scale ranging from “strongly agree” to “strongly disagree”. 48.7% of clinicians were more likely to agree or strongly agree to both these statements for smokers than nonsmokers (32.7%).
Patients were asked to respond to several statements regarding their understanding of antibiotics on the same five-point Likert-scale. Regarding the statement “I believe that most coughs lasting more than a few days should be treated with antibiotics”, smokers were more likely to agree or strongly agree than subjects in either of the other smoking status categories.
Overall, antibiotics were prescribed for 421 (60%) smokers, 655 (53%) nonsmokers and 314 (51%) ex-smokers, all for a median of 7 days (table 3⇑).
Apart from Bratislava (the overall highest antibiotic prescribing network) and Barcelona, Spain (the overall lowest antibiotic prescribing network), smokers in all other networks were prescribed antibiotics more frequently than nonsmokers (fig. 3⇓).
Antibiotic prescribing between networks and smoker categories adjusted for clinical presentation
After adjusting for variables associated with antibiotic prescribing, the odds ratio of a smoker being prescribed antibiotics was 1.44 (95% CI 1.12–1.86; p≤0.01). Ex-smokers had an odds ratio of 1.18 (95% CI 0.90–1.53; p = 0.23); there is insufficient evidence to show a significant difference in antibiotic prescribing between ex-smokers and nonsmokers (table 4⇓).
With the smoker status category included in the model, significant variation between networks remained, with the overall odds of antibiotic prescribing being significantly different from nine other networks (four networks prescribing more than average, five prescribing less) (table 4⇑).
Fitting the same model as above but including pack-year values to the subgroup with valid pack-years (2,346 individuals comprising 1,234 individuals who never smoked and 1,112 current or ex-cigarette smokers) indicated that pack-years was not a significant predictor of antibiotic prescribing (OR 1.00, 95% CI 0.99–1.01).
Table 5⇓ summarises the results of the patient recovery model. Of particular interest are the parameters related to smoking status.
The three-way interaction between smoking status, being prescribed an antibiotic, and day of illness post-consulting was not significant. Antibiotic prescribing was not associated with a difference in patient recovery for either smoking category (smokers or ex-smokers) compared with nonsmokers.
Fitting the same model as above but including pack-years to the subset of data with valid pack-year values (2,172 individuals comprising 1,154 never-smokers and 1,018 current or ex-cigarette smokers) indicated that patient outcome was not associated with patient pack-years (coefficient 0.001, p = 0.47).
We found that primary care clinicians participating in this 13 European country, observational study of the presentation, management and outcomes of acute cough prescribed antibiotics more frequently to smokers than nonsmokers. This suggests that, despite differences in training and practice setting, clinicians may have similar attitudes towards prescribing antibiotics for smokers.
One reason for the higher rate of prescribing for smokers may be because clinicians feel that smokers do worse if they do not receive antibiotics. However, symptom severity scores did not differ significantly at presentation or throughout recovery for smokers compared with nonsmokers. Antibiotic prescribing was not associated with recovery among smokers; recovery rate was no different between smokers and nonsmokers.
Strengths and weaknesses
A double-blind randomised controlled trial would answer the question of whether antibiotics benefit smokers with acute cough; however, this observational study design is best suited to describe actual clinician behaviour for a broader spectrum of patients. We were able to control for symptoms and signs, including for sicker patients receiving antibiotics more frequently.
Questions on smoking status were not included in the clinician-completed CRF, since we did not want to influence clinicians’ behaviour during the consultation. Clinicians may have had no prior knowledge of smoking status before choosing whether to prescribe antibiotics. However, since cough was an inclusion prerequisite, we believe that the majority of clinicians would have enquired about, or known about, the smoking status of the patient. This assumption is borne out by qualitative interviews suggesting that the majority of clinicians would enquire about smoking status during a consultation for cough (analysis in progress).
Determining criteria to define smoking status and allow the classification of participants was complex. There is no consensus on either a clinical or a research definition of “ever-smoker” or “ex-smoker”. Since relapse rates are high, the length of time since stopping smoking and defining a person as an ex-smoker varies by study; cut-off time periods produce very different results when analysing data. Similarly, there is no consensus regarding how many cigarettes need to have been smoked in a lifetime in order to be defined as an “ever-smoker”. Pack-years were not found to be associated with either antibiotic prescribing or patient recovery.
Comparison to current literature
This study found lower rates of antibiotic prescribing compared with some other observational studies on LRTIs 23. This is probably because of the inclusion of “acute cough” rather than diagnosis of LRTI. However, we did not find that smokers were more likely to be labelled as having a LRTI compared with nonsmokers.
The duration of illness of smokers is similar to other studies, and the finding that smokers do not recover any more slowly than nonsmokers is consistent 9. Antibiotic prescribing was not associated with any extra benefit to smokers with acute cough, which is in agreement with the conclusions of a review looking at antibiotic prescribing for acute bronchitis in smokers 18.
Our study confirmed that clinicians perceive antibiotics to be more beneficial to smokers 2, and that clinicians are more likely to prescribe antibiotics for smokers 4. Clinicians are more likely to prescribe antibiotics if a patient expects them 24, 25. In our study, smokers, more than nonsmokers, believed antibiotics would help them if cough lasted more than a few days.
Our study found that smokers believe that antibiotics are indicated for their cough if it lasts longer than a few days, confirming previous studies 26. There is a strong association between patients wanting antibiotics and antibiotics being prescribed to them 26, so this may account for some of the noticeable increase in prescriptions for antibiotics for smokers.
Smoking status is an independent factor for antibiotic prescribing in the European networks we investigated. However, antibiotic prescribing is not associated with recovery in smokers.
Implications for practice
Smoking status alone should not be a determinant for antibiotic prescribing.
This observational study has highlighted that antibiotic prescribing for smokers presenting with acute cough is higher than for nonsmokers in primary care in Europe. To our knowledge, no double-blind randomised controlled trial of antibiotic prescribing in smokers with acute cough has been conducted.
Funding: 6th Framework Programme of the European Commission under the reference LSHM-CT-2005-518226.
This study is registered at clinicaltrials.gov, identifier number NCT00353951.
Statement of interest
Earn CME accreditation by answering questions about this article. You will find these at the back of the printed copy of this issue or online at www.erj.ersjournals.com/misc/cmeinfo.dtl
- Received October 23, 2009.
- Accepted December 16, 2009.
- © ERS Journals Ltd