Purpose: In studies of how practitioners weight clinical findings in making judgments about management or diagnosis, interactions among the findings are usually ignored or not analyzed. To determine if interactions influenced the decision to give antibiotics to patients presenting with symptoms of acute respiratory infections (ARI), we analyzed the first order interactions in community practitioners’ decisions to prescribe antibiotics.
Method: 101 community practitioners in Denver, CO estimated how likely they were to prescribe antibiotics in response to each of 20 paper cases of patients presenting with ARI. They also judged the likelihood that the patient had each of 4 clinical diagnoses (viral ARI, pneumonia, bronchitis, strep throat). Values of clinical findings were expressed at 2 levels (high/low or present/absent), following a fractional factorial design. We used judgment analysis (lens model) to determine the weight (size of the effect) of each clinical finding on the decision to prescribe antibiotics. Weights were calculated for each individual practitioner.
Result: In the decision to prescribe antibiotics, cue interactions outweighed all main effects for 41% of the 101 practitioners. The two most highly weighted interactions were temperature x duration (positive weighting) and temperature and runny nose (negative weighting). In diagnosing the cause of the illness, interactions outweighed the main effects for 50-60% of the participants, depending on the diagnosis.
Conclusion: We found that interactions between clinical findings were often more important than main effects in predicting practitioners' decisions about whether to give antibiotics in ARI. One important interaction was runny nose x temperature, which reduced the likelihood of prescribing antibiotics - a logical effect since runny nose is a feature of viral ARI (though not in the case of influenza). Temperature x long duration had a positive effect on prescribing antibiotics, consistent with bacterial pneumonia and sinusitis, but not influenza. Thus, interactions may amplify or attenuate the effect of the clinical finding on clinical judgment, as duration of illness here amplifies the effect of temperature. Since we do not know how practitioners process clinical findings (e.g. pattern recognition, analytical, narrative, cue weighting), understanding the role of interactions will need further research. Interactions may be important in teaching how clinical findings relate to diagnosis and management.