Fig 1.
The decision tree structure models the risk-stratification of an indeterminate pulmonary nodule utilizing artificial intelligence-assistance compared to the clinician alone. Repeated portions of the model have been collapsed into subtrees (A-G) for readability, each of which represents a diagnostic or management pathway that appears in various parts of the model (‘A’ = surveillance; ‘B’ = PET-CT evaluation; C = minimally invasive surgical (MIS) lobectomy; ‘D’ = low-risk surveillance; ‘E’ = intermediate-risk; ‘F’ = MIS wedge resection; ‘G’ = initial risk classification).
Table 1.
Model Parameters.
Table 2.
Chemotherapeutic Regimen by Stage.
Table 3.
Incremental Cost and Effectiveness of Clinician Compared to AI-Assisted Pulmonary Nodule Risk Stratification.
Fig 2.
Tornado Diagram of Incremental Cost-Effectiveness Ratio Range of Four Most Influential Model Parameters.
The tornado diagram demonstrates the incremental cost-effectiveness ratio (ICER) range for the 4 most influential parameters in one-way sensitivity analyses. Blue indicates the lower end of the varied parameter range and orange indicates the higher end of the range. The ICER exceeds the willingness-to-pay threshold only for malignancy prevalence when the prevalence is below.051. All other parameters remain cost-effective over the full tested range.
Fig 3.
One-Way Sensitivity Analysis of Malignancy Prevalence.
The x-axis shows the population malignancy prevalence, ranging from 0% on the left to 80% (0.8) on the right. The y-axis shows the incremental cost-effectiveness ratio (ICER) at the corresponding malignancy prevalence. The ICER exceeds the willingness-to-pay threshold when malignancy prevalence of the population is less than 5%.
Fig 4.
Sensitivity Analysis of Clinician Accuracy for Predicting Malignancy.
The x-axis shows the probability a clinician correctly classifies a malignant nodule as high-risk. The y-axis shows the incremental cost-effectiveness ratio (ICER) corresponding to that probability. The ICER exceeds the willingness-to-pay threshold when the clinician accuracy for correctly identifying malignant nodules as high-risk exceeds 65%.
Fig 5.
Probabilistic Sensitivity Analysis: Cost-Effectiveness Acceptability Curve.
The x-axis represents varying willingness-to-pay thresholds. The y-axis represents the percentage of iterations of the model that resulted in either strategy being cost-effective at the corresponding willingness-to-pay. The curves cross at $4,058, the willingness-to-pay threshold where 50% of the iterations were cost-effective for each strategy. At the base case willingness-to-pay of $100,000/life year gained, AI-assistance is cost-effective in 71% of iterations.