Does diagnostic uncertainty increase antibiotic prescribing in primary care?

Abstract

This study aimed to determine the association between factors relevant to diagnostic uncertainty and physicians' antibiotic-prescribing behaviour in primary care. A questionnaire survey was conducted on 327 physicians that measured their diagnostic ability, perceived frequency of diagnostic uncertainty, tolerance, and perceived patient tolerance of uncertainty. Physician antibiotic-prescribing behaviours were assessed based on their prescriptions (n = 207,804) of three conditions: upper respiratory tract infections (URTIs, antibiotics not recommended), acute tonsillitis (cautious use of antibiotics), and pneumonia (antibiotics recommended). A two-level logistic regression model determined the association between diagnostic uncertainty factors and physician antibiotic prescribing. Physicians perceived a higher frequency of diagnostic uncertainty resulting in higher antibiotic use for URTIs and less antibiotic use for pneumonia. Higher antibiotic use for acute tonsillitis was related to a low tolerance of uncertainty of physicians and patients. This study suggests that reducing diagnostic uncertainty and improving physician and patient uncertainty management could reduce antibiotic use.

Fig. 1. Data collection and extraction procedure.

This figure shows the current study design. Physicians’ characteristics regarding diagnostic uncertainty were collected by a questionnaire survey, and their antibiotic-prescribing behaviours in clinical practice were extracted from the outpatient prescription dataset. The surveyed characteristics of physicians were mapped with their prescriptions for three acute conditions commonly treated in primary care facilities, including upper respiratory tract infections, acute tonsillitis, and pneumonia with an unspecified organism. According to the guidelines, antibiotics are not recommended for URTIs, while cautious use of antibiotics can be applied for tonsillitis, and antibiotic prescriptions for pneumonia with an unspecified organism are recommended. These conditions have a varied likelihood of bacterial infection and are deemed sensitive to the physicians and patient tolerance of uncertainty.

Fig. 2. The interaction effect between physician tolerance of uncertainty and patient tolerance of uncertainty on antibiotic prescribing for acute tonsillitis.

This figure shows the interaction between physician tolerance of uncertainty and perceived patient tolerance of uncertainty regarding antibiotic use with 95% confidential intervals for outpatients with acute tonsillitis. Physician tolerance of uncertainty was assessed using a tolerance for ambiguity scale, with higher scores indicating lower physician tolerance of uncertainty. Patient tolerance of uncertainty was assessed based on physician estimation of patient ambiguity aversion, with higher scores indicating lower patient tolerance of uncertainty. For different perceived patient tolerances of uncertainty, the one-point decrease in physician tolerance of uncertainty played different roles. When physicians perceived high patient tolerance of uncertainty (scores ≤6), the lower physician tolerance of uncertainty further increased antibiotic use, while for perceived low level of patient tolerance of uncertainty (scores ≥7), the lower physician tolerance of uncertainty would reduce antibiotic use. However, the interaction between the physician tolerance of uncertainty and perceived patient tolerance of uncertainty on antibiotic use was only significant when perceived patient tolerance of uncertainty was ≤4 (marginal effect >0, p < 0.05) or ≥14 (marginal effect <0, p < 0.05).