Screening for Obstructive Sleep Apnea in Stroke Patients: A Cost-Effectiveness Analysis

The primary outcome was the utility gained through OSA screening and treatment in relation to 2 common willingness-to-pay thresholds of $50 000 and $100 000 per quality-adjusted life year (QALY). Utilities summarize patient preferences for specific health states, ranging from 0 for death to 1 for perfect health. A decision tree (Figure) modeled 2 alternative strategies (screening

versus no screening) for a hypothetical adult with recent stroke resulting in a moderate deficit, with an associated health state utility of 0.6. In the screening pathway, identification of OSA led to CPAP titration, followed by 3 months of CPAP treatment. It was assumed that there were no deaths or recurrent strokes over this short time period because these are patients who survived their initial stroke hospitalization.³ The 3-month time horizon also allowed us to assume that no patient in the no-screening arm would come to medical attention due to symptomatic OSA and “crossover” to receive CPAP, thus allowing for a simple model. Reference case estimates of prevalence of OSA and probability of CPAP acceptance in addition to estimated costs and utilities are found in Table 1. The actual short-term acceptance of CPAP in stroke patients in unknown, but this value was allowed to vary in sensitivity analysis. The analysis was performed from a societal perspective using a 3-month time horizon. Costs were estimated based on Medicare reimbursement.^4,5 No discounting of costs or utilities was needed given the time horizon. One-way sensitivity analyses were conducted using the ranges found in Table 1. Two-way sensitivity analyses, in which 2 variables were allowed to vary over a plausible range, were conducted with respect to the utility estimates. The decision tree was analyzed by Data 4.0 (TreeAge Inc).

Model results are found in Table 2. In the base case, the incremental cost-effectiveness ratio for screening was $49 421 per QALY. Two-way sensitivity analysis performed on the utilities of stroke states showed that PSG has a cost-effectiveness ratio <$50 000 per QALY as long as the utility for those with OSA on CPAP is ≥0.2 higher than for those with OSA not on CPAP. Therefore, for a willingness-to-pay of $50 000 per QALY, the relative increment in utility would have to be ≥50% (from 0.4 to 0.6), meaning that screening is cost-effective as long as the treatment of stroke patients with OSA by CPAP improves quality of life (QOL) by ≥50%. For a willingness-to-pay of $100 000 per QALY, the relative increment in utility would have to be only ≥25% for screening to be cost-effective.

This analysis suggests that if CPAP were shown to improve stroke recovery, it would need to improve patients’ utilities by 0.2 in order to be cost-effective given a willingness-to-pay of $50 000 per QALY and by 0.1 for a willingness-to-pay of $100 000 per QALY. In general, interventions that cost <$50 000 to $100 000 per QALY gained are judged to be worthwhile healthcare expenditures.⁶ The aforementioned changes in utility represent a 25% to 50% relative improvement in QOL (given the assumption of 0.4 for the utility of a stroke in an OSA patient).

One method for lowering costs associated with PSG screening and CPAP titration would be to attempt to combine the 2 in a split-night study. If the apnea–hypopnea index is ≥20 in the first half of the night, the second half of the night may be used for CPAP titration.⁷ This would cut costs considerably to $26 918 per QALY. With the use of a split-night study in all patients, only 16% and 32% improvement in utility would be required to make screening cost-effective given a willingness-to-pay of $100 000 and $50 000 per QALY, respectively.

Another potential cost-saving mechanism would be to use clinical criteria to help identify patients with OSA. Unfortunately, clinical criteria currently identify less than two-thirds of stroke patients with OSA correctly.⁸ Therefore, presently, PSG is necessary to screen stroke patients for OSA.

The results of this analysis cannot be extended to those with very mild or very severe strokes. Additionally, this analysis is limited by the validity of the utility estimates used. The utilities used were gathered from patients at risk for stroke given the societal perspective taken. We also performed sensitivity analyses on these values in attempt to compensate for this limitation. Further studies obtaining more precise estimates of QOL and cost information in stroke patients with OSA are needed.