In order to understand how clinical factors influence practitioners’ decisions about prescribing antibiotics in ARIs, we studied the decisions of 101 community practitioners in response to 20 case vignettes of patients with respiratory tract infection. On average, they prescribed antibiotics in 44.5% of the cases. Although the weighting of specific findings varied from individual to individual, 72% of practitioners placed the greatest weight on the duration of illness in making the decision to start antibiotics. In addition, there were significant interactions between the most important clinical factors. Non-clinical patient factors such as the patient’s desire for antibiotics did not play a significant role in the decisions, although many physicians say they are influenced by patient factors in antibiotic treatment decisions for ARIs4,5,14–16
To provide a benchmark, eight internal medicine faculty members reviewed the published guidelines from a CDC expert panel before deciding whether to prescribe antibiotics in the same 20 cases. They prescribed antibiotics in 20% of the cases and thought that 19.4% of the cases were bacterial in nature. Four of the panelists placed the greatest weight on duration of illness, three on sinus pressure and pain, and one on colored nasal drainage.
These results suggest that duration of illness plays a major role in the decision to prescribe antibiotics. Duration of illness has not received much attention previously. Dosh and colleagues had found that duration greater than 14 days was associated with increased prescribing in their univariate analysis, but the effect disappeared when the model was adjusted for other variables9
. Does longer duration increase the likelihood that the illness is bacterial? In these 20 cases, the possible diagnoses included viral URI, influenza, bronchitis, rhinosinusitis, or pneumonia. Of these, only sinusitis and pneumonia would be treated with antibiotics according to the CDC guidelines20,21
. Pharyngitis was not a consideration in these cases because none had a sore throat. Because there is evidence that rhinosinusitis of longer duration is more likely to be a bacterial infection21
, it is reasonable to consider duration of illness in that setting. Routine antibiotic treatment of uncomplicated acute bronchitis, however, is not recommended in healthy individuals, regardless of the duration of the illness19
. The median duration of viral bronchitis is 2 to 3 weeks, and longer duration does not correlate with a bacterial cause, with the notable exception of pertussis in patients with cough lasting more than 2–3 weeks24
. With regard to pneumonia, a study of 1,436 patients with respiratory tract infection seen in the emergency department found that duration of illness was not an independent predictor of infiltrate on chest X-ray25
, and it was not an important factor in the diagnosis of pneumonia in the studies reviewed by Metlay26
Other findings given significant weight by the community practitioners were temperature, sinus tenderness, colored nasal drainage, and productive cough. In these case vignettes, the level of fever (101.5° F) does not reliably distinguish between viral and bacterial infections, but is a predictor of pulmonary infiltrate if pneumonia is under consideration25
. Colored nasal drainage, but not sinus pressure and pain, is helpful in distinguishing bacterial from viral rhinosinusitis7
. Productive cough usually suggests bronchitis unless a patient also presents with marked sinus congestion, but cough did not achieve significance as an independent predictor of pneumonia in two studies25,27
. Gonzales found that findings suggesting purulence are associated with increased rates of antibiotic use2
. Purulence, however, has not proved to be a reliable way to distinguish bacterial from viral bronchitis28,29
. None of the three patient factors appeared to influence the decisions. Several studies have suggested that patient factors are important4,5,13–15
, but an observational study of 482 patients found no independent effect on prescribing decisions9
. There are several possible explanations for these differences. Since most studies of patient factors rely on physicians’ self-report of what influenced their decision, it may be that practitioners overestimate the effect of patient factors or may use them to rationalize a decision to prescribe. Previous studies have shown that decision makers often have poor insight into their own decision policies and that the weights derived from judgment analysis are better at predicting future decisions30
. An alternative explanation is that patient factors are indeed important, but that the written descriptions of the patients’ wishes in this study did not have the impact of a face-to-face presentation and thus underestimated the effect.
Most of the practitioners were more comfortable with their decisions in the cases where they did not give antibiotics. The study cases were designed to have few clear-cut diagnoses, and it may be that the cases in which respondents prescribed antibiotics were those with more uncertainty about the diagnosis. Alternatively, practitioners may have been more concerned about error when giving antibiotics than when omitting them. A critical incident study of British practitioners found that the clinical situation that most frequently made them uncomfortable was deciding whether to prescribe antibiotics in respiratory tract conditions31
What accounted for the lower antibiotic prescribing rate among the faculty applying the CDC guidelines (20% versus 44.5%)? The lower rate in the faculty could have been due to differences in the overall proportion of cases they thought should get antibiotics (the base rate). The general tone of the published CDC guidelines is that most respiratory tract infections are viral and do not need antibiotics. Additionally, the decreased weight given productive cough and the lack of interaction between cough and duration by the faculty members could have accounted for some or all of the lower rate. The CDC guidelines for bronchitis stated, “Routine antibiotic treatment of uncomplicated acute bronchitis is not recommended, regardless of duration of cough”20
. The faculty group prescribing was consistent with this. They had a slightly negative weight for productive cough and no interaction between cough and duration.
There are several limitations to generalization of these findings. The decisions were made in response to paper case vignettes limited to nine features and not actual patients. Descriptions of clinical findings and patient factors may have lacked the force they would have in patient encounters. Practice patterns of these Colorado practitioners might not generalize to other geographic areas. Strongly diagnostic findings (e.g., tonsillar exudate, unilateral maxillary pain) were purposely omitted, and this could have increased the overall uncertainty and affected the rate of antibiotic prescribing. Finally, the vignettes did not allow ordering further examinations or tests as might have been appropriate if pneumonia had been suspected.
These results indicate that these practitioners are strongly influenced by the duration of illness in deciding to prescribe antibiotics. The influence of duration of illness is particularly strong in cases where the patient had a productive cough. Omission of this effect would bring the prescription rate close to that of faculty members following the published CDC guidelines. This demonstrates the importance of the weight the clinician gives to individual findings in deciding whether to prescribe antibiotics. It suggests, further, that clinical studies aimed specifically at whether antibiotics are effective in cough illnesses that last 2 weeks or more are particularly important. Targeted education of practitioners about the utility of antibiotics in this specific situation may be of great help in reducing antibiotic use.