Quantifying statistical uncertainty in metrics of sleep disordered breathing

Area of Science:

• Sleep Medicine
• Respiratory Physiology
• Biostatistics

Background:

• The apnea-hypopnea index (AHI) is the standard metric for diagnosing sleep disordered breathing (SDB).
• Current AHI calculations treat the index as a precise point estimate, neglecting statistical uncertainty.
• This lack of uncertainty consideration can lead to variability in diagnoses and treatment eligibility.

Purpose of the Study:

• To quantify the statistical uncertainty associated with respiratory event indices in SDB.
• To evaluate the impact of this uncertainty on patient diagnosis and treatment eligibility.

Main Methods:

• Developed empirical (non-parametric bootstrap) and theoretical (Poisson) estimates of AHI uncertainty.
• Applied these methods to data from 2049 subjects in the Multi-Ethnic Study of Atherosclerosis (MESA).

Main Results:

• Mean 95% empirical confidence interval width for AHI was 11.5 events/hour, and theoretical Poisson was 6.0 events/hour.
• Significant percentages of subjects (27% symptomatic, 43% full population) had uncertain diagnoses due to AHI variability.
• Including uncertain cases increased eligible patients by up to 84.8%, while excluding them decreased eligibility by up to 34.8%.

Conclusions:

• AHI uncertainty is a major confounding factor in current SDB diagnostic frameworks.
• Incorporating AHI uncertainty is vital for improving clinical practice and diagnostic accuracy.
• Considering uncertainty in epidemiological studies can enhance the robust linking of AHI with comorbidities and outcomes.