To identify patients with URIs and acute bronchitis, we used International Classification of Diseases-ninth revision-Clinical Modification (ICD-9-CM) codes for acute bronchitis and URI that were entered with the diagnosis recorded in the electronic medical record. A search for all patients with a problem code 466.0 for acute bronchitis between June 1, 1996, and December 31, 1998, found 165 patient records. Using a similar approach and the ICD-9 code 465.9 for upper respiratory infection, we identified 526 patients with a diagnosis of URI in the same time period. We selected a random sample of 495 records to create a 3:1 match with the bronchitis cases. Patient visits that contained a secondary diagnosis of otitis media, sinusitis, asthma, chronic obstructive pulmonary disease, congestive heart failure, or pneumonia were excluded. The remaining number of total records was 544 (409 URI and 135 acute bronchitis).
Two medical students conducted a detailed chart review of these records. Information on demographics, symptoms, and physical findings were recorded, as well as the provider’s level of training for each record reviewed. To assess the interrater reliability of the reviewers, a random sample of 54 charts (10% of the total sample) was abstracted by both reviewers. Interrater agreement was excellent, ranging from 93% to 100%, depending on the variable.
Information was entered into Epi Info, version 6 (Centers for Disease Control, Atlanta, Ga), a standard epidemiologic database. Data were analyzed with bivariate comparisons of possible predictors. From the initial analysis, the statistically significant variables most often associated with either URI or acute bronchitis were used to perform a multiple regression to determine the independent predictive value of these variables. Regression analysis was performed using True Epistat software (Epistat Services, Richardson, Tex). Because of the large number of variables used in the study (4 demographic, 11 symptom, and 9 physical finding variables), we used the Bonferroni adjustment for multiple comparisons and adjusted the P value for 24 multiple comparisons to set a value of 0.002 (0.05 of 24 possible comparisons) as the indicator of statistical significance.14
Results
Table 1 shows the demographics and recent visit history for patients based on their diagnosis. Acute bronchitis patients were slightly older than those with URIs and were more likely to be smokers. Acute bronchitis patients also were more likely to have made previous visits for either URI or acute bronchitis within 3 months of the diagnosis. When we examined the training level of physicians who saw these patients, we found that attending physicians were more likely to diagnose acute bronchitis than were residents. Of the patients seen by attending physicians, 33% were were considered to have acute bronchitis, compared with 19% of the patients seen by residents (P <.001). Examining the reported symptoms of those patients with acute bronchitis and URI revealed that several symptoms were more often associated with each diagnosis Table 2. Cough was present in the majority of both conditions but occurred more often when acute bronchitis was the diagnosis. Chest pain, shortness of breath, and a history of wheezing also were associated with acute bronchitis, although each was present in only 8% to 12% of the cases. In contrast, symptoms associated with URI included runny nose and sore throat, but neither alone was seen in a majority of patients considered to have URIs.
Comparisons of physical findings provided similar results Table 3. A red throat, red nose, or cervical lymph nodes were all associated with the diagnosis of URI. The only physical finding associated with bronchitis was the presence of wheezing, found in approximately 30% of patients with acute bronchitis.
Since very few of the symptoms and physical findings associated with the diagnosis of acute bronchitis or URI appeared in the majority of patients, it was thought that clinicians probably rely on a constellation of these signs and symptoms to make the diagnosis of bronchitis. Also, because each of the symptoms and signs might not be independent predictors of URI as opposed to bronchitis, we entered symptoms and physical findings associated with each diagnosis into a multivariate logistic regression model along with possible demographic predictors. Table 4 shows the results of this logistic regression. All other things being equal, cough and wheezing were the strongest independent predictors of acute bronchitis, while nausea or a red or runny nose were the strongest predictors of URI. Also of note is that adults and patients who had been seen in the previous 3 months with bronchitis were more likely to receive a diagnosis of acute bronchitis independent of their other signs and symptoms.