Strategies for diagnosing and treating suspected acute bacterial sinusitis: a cost-effectiveness analysis

Abstract

Objective: Symptoms suggestive of acute bacterial sinusitis are common. Available diagnostic and treatment options generate substantial costs with uncertain benefits. We assessed the cost-effectiveness of alternative management strategies to identify the optimal approach.

Design: For such patients, we created a Markov model to examine four strategies: 1) no antibiotic treatment; 2) empirical antibiotic treatment; 3) clinical criteria-guided treatment; and 4) radiography-guided treatment. The model simulated a 14-day course of illness, included sinusitis prevalence, antibiotic side effects, sinusitis complications, direct and indirect costs, and symptom severity. Strategies costing less than 50,000 dollars per quality-adjusted life year gained were considered "cost-effective."

Measurements and main results: For mild or moderate disease, basing antibiotic treatment on clinical criteria was cost-effective in clinical settings where sinusitis prevalence is within the range of 15% to 93% or 3% to 63%, respectively. For severe disease, or to prevent sinusitis or antibiotic side effect symptoms, use of clinical criteria was cost-effective in settings with lower prevalence (below 51% or 44%, respectively); empirical antibiotics was cost-effective with higher prevalence. Sinus radiography-guided treatment was never cost-effective for initial treatment.

Conclusions: Use of a simple set of clinical criteria to guide treatment is a cost-effective strategy in most clinical settings. Empirical antibiotics are cost-effective in certain settings; however, their use results in many unnecessary prescriptions. If this resulted in increased antibiotic resistance, costs would substantially rise and efficacy would fall. Newer, expensive antibiotics are of limited value. Additional testing is not cost-effective. Further studies are needed to find an accurate,low-cost diagnostic test for acute bacterial sinusitis.

FIGURE 1

Markov decision model, showing each of the four treatment and diagnostic options on the first four branches. The square at the far left of the diagram represents a decision node, with each branch representing the clinical management choices. The brackets indicate that patients who received each of the strategies proceeded to the subtree to the right. The circles at the start of each subsequent branching indicate chance nodes representing the uncertainty surrounding possible subsequent outcomes indicated in the branches to the right. This diagram shows the portion of the decision tree modeling the management strategies, prevalence of sinusitis, and diagnostic test performance (sensitivity and specificity). The Markov process simulates 14 days during which patients may make various possible transitions: if they are sick, they may remain sick, become well, become sick with an antibiotic side effect, or develop a serious disease complication; if they are sick with a side effect, the side effect symptoms will last for 2 days, after which they may remain sick or become well without a side effect; if they are well, they may remain well or develop a side effect which will last for 2 days, after which they will return to the well state.

FIGURE 2

Kaplan-Meier curves and fitted Weibull curves for the placebo and amoxicillin arms of trial evaluating amoxicillin for the treatment of acute bacterial sinusitis. Point estimates and 95% confidence intervals are shown at day 12 for treatment meta-analysis estimates of cure at days 10 to 14.

FIGURE 3

Optimal management strategies depending on symptom severity and prevalence of acute bacterial sinusitis in four groups of patients: 1) considering any symptom (sinusitis or antibiotic-related), 2) quality of life adjusted for severe sinusitis symptoms, 3) quality of life adjusted for moderate sinusitis symptoms, and 4) quality of life adjusted for mild sinusitis symptoms. Optimal strategies are effective and cost-saving, or cost-effective, at less than $137 per day [equivalent to $50,000 per quality-adjusted life year]. White bars at left indicate prevalence levels for which no antibiotics is the favored strategy. Gray bars in the middle indicate prevalence levels for which use of clinical criteria is favored. Black bars at right indicate prevalence levels for which empirical antibiotics is favored. The numbers within the bars indicate the exact threshold at which the optimal strategy changes.