The majority of RTIs observed in our study were cases of rhinopharyngitis, and 32 cases were combined with an additional diagnosis of RTI (as detailed in Table 2). Notably, our findings indicated that antibiotic prescription were the least appropriate in bronchitis and otitis cases. Our analysis did not reveal any statistically significant influence stemming from factors such as the patient’s social status, age, gender, socioprofessional background, social category, or the presence of a long-term condition on the appropriateness of prescription for RTIs.

We observed an overall rate of inappropriate antibiotic-related management of 26.6% (95% CI 19.2–32.6%). Among consultations that resulted in an antibiotic prescription, 60.5% (95% CI 49.0–71.1%) were deemed inappropriate.

The proportion of inappropriate prescriptions per clinician ranged from 0 to 100%, representing extreme values mainly driven by clinicians with very few RTI consultations observed (e.g., one inappropriate prescription out of a single consultation). This variability was accounted for by including a random clinician effect in our statistical models.

Prescription patterns and decision-making factors

In our study, the most frequently prescribed antibiotics were amoxicillin, followed by amoxicillin-clavulanic acid, as summarized in Table 3. Bronchitis displayed the highest diversity in prescribed antibiotic molecules. Daily doses and treatment duration varied, without any overdoses, and a duration ranging from 5 to 10 days, with a majority between 5 and 7 days.

Factors influencing clinical decisions

Table 4 shows the clinical factors influencing GP antibiotic prescription.

No significant differences were found for most demographic variables. Patient’s General Condition: We found that the practitioner’s perception of the patient being in “good condition” was associated with more appropriate prescribing practices (odds ratio [OR] = 2.1, 95% [CI]: 1.1–4.3, p = 0.03). This indicates that when physicians perceived patients as generally healthy, they were more likely to make appropriate prescribing decisions.

Patient’s “At-Risk” Status: A patient was considered “at risk” if he had underlying comorbidities (e.g., diabetes, heart failure, kidney failure). Patients deemed “at risk” due to their medical conditions received more inappropriate antibiotic prescriptions (OR = 0.3, 95% CI 0.2–0.7, p = 0.005).

Difficult Social Context: Patients facing challenging social contexts, often indicated by their status as beneficiaries of universal health coverage, were associated with less appropriate prescriptions (OR = 0.14, 95% CI 0.03–0.73, p = 0.008). This suggests that social factors influenced prescribing decisions.

Frequent consultations for the same reason were linked to less suitable prescriptions. If a patient had a history of prior consultations for the same medical concern, the appropriateness of prescriptions decreased (OR = 0.2, 95% CI 0.1–0.5, p 

Notably, other clinical decision factors, such as the presence of fever, duration of symptom evolution, number of RTIs presented by the patient, and smoking status, did not demonstrate statistical significance in bivariate analysis.

When examining decision-making elements related to physicians’ profiles and clinical reasoning, our bivariate analyses did not uncover any significant differences associated with the age of the physicians, their practice mode in solo or group settings, or in rural or urban locations, the number of years of practice, or the workload.

In our study, the majority of doctors performed commented or ritualized clinical examinations, as presented in Table 5. These examinations involved systematic and thorough physical assessments following standard protocols, in contrast to “light examinations,” which consisted of brief or limited assessments that omitted certain standard steps. The distinction between these types of examinations was defined according to the data collection grid and reinforced during investigator training. Notably, when physicians performed light clinical examinations, this was associated with a 3.3 increase in inappropriate prescriptions (p = 0.002, OR: 0.3, 95% CI 0.1–0.6). This finding suggests that more comprehensive clinical examinations were associated with a lower rate of inappropriate prescriptions.

Explaining the diagnosis was relatively common and associated with more appropriate management (p = 0.03, OR: 2.2, 95% CI 1.1–4.6).

No treatment bargaining was observed in our study. Additionally, the explicit request for antibiotics by patients did not emerge as a significant decision factor influencing the appropriateness of care.

These findings suggest that thorough clinical examinations and effective communication of diagnoses played pivotal roles in determining the appropriateness of care decisions in our study.

Multivariate analysis

In the first multivariate analysis model (Online Appendix 5), we examined clinical factors associated with inappropriate antibiotic prescriptions. In cases where “fever” was present, prescription was significantly 3.65 times less appropriate (95% CI [1.54; 8.65], p p p 

In the second multivariate analysis model (Online Appendix 6), which focused on the specific context of the consultation, we found that explaining the diagnosis was associated with a 2.44-fold higher likelihood of appropriate prescription (OR: 0.41, 95% CI [0.17; 0.99], p = 0.047). Offering follow-up increased the risk of inappropriate prescription by a factor of 5.45 (95% CI [1.53; 19.40], p = 0.009). A recent consultation for the same reason increased the risk of inappropriate prescriptions by a factor of 4.13 (95% CI [1.51; 11.27], p = 0.006). Diagnostic doubt increased the risk of inappropriate prescription by a factor of 5 (95% CI [1.40; 17.71], p = 0.013).

Notably, variables related to social context, therapeutic alternatives, and nonmedicinal advice were not significantly associated with the prescription appropriateness.

In the third multivariate analysis (Online Appendix 7), the only variables significantly associated with inappropriate prescription were the number of drugs prescribed for conditions other than RTI treatments (OR: 1.36, 95% CI [1.07–1.73], p = 0.01) and receiving visits from medical representatives (OR: 4.59, 95% CI [1.51–13.95], p = 0.007). Variables such as consultation time, the number of drugs prescribed specifically for RTI, physician practice characteristics, and workload were not significantly associated with inappropriate prescriptions.