Beyond the AHI for New Sleep Diagnostic Paradigms


Nox Medical’s satellite symposium at the 2020 SLEEP conference concluded with a session by Dr. Dennis Hwang, MD, Medical Director of Sleep Medicine at Kaiser Permanente.

Dr. Hwang specializes in the implementation of technology and big data tools to innovate sleep medicine care delivery, with a particular interest in technology integration and development of artificial intelligence tools. His session discussed new diagnostic paradigms for sleep physicians to consider and introduced an alternative approach to diagnosing sleep disorders.

Small Changes

Dr. Hwang began his talk by discussing what he called “smaller changes” for sleep professionals to consider based on the notion that the HSAT (home sleep apnea testing) is superior to the PSG (polysomnography) for diagnosing obstructive sleep apnea.

Dr. Hwang referenced a study of 141 patients where a negative PSG was followed by a positive HSAT in 84% of patients, and 20% of the time, those patients ended up having moderate to severe OSA.

“It begs the question that in a real-world scenario, when you have a negative HSAT and the PSG ends up being positive, it very possibly has nothing to do with the superiority of the PSG over HSAT, but rather you’re doing a second test and giving the patient another chance of being positive,” Dr. Hwang said.

According to Dr. Hwang, one concern related to the use of HSAT devices is that the AHI is thought to be potentially inaccurate for patients who are poor sleepers. This is because HSAT devices historically calculated AHI based on the number of events per hour recording rather than per hour of sleep. This older tech would lead to an underestimate of the true AHI and an underdiagnosis of OSA. Modern HSAT devices, however, use estimated sleep time as the denominator, helping to mitigate any concerns around accuracy.

In another study comparing PSA data with that of three different HSAT devices, results showed that all three HSAT devices had higher OSA diagnostic rates compared to the PSG. Dr. Hwang believes the HSAT to be superior in this regard, due to its ability to capture night-to-night variability.

In response to the question, “Can a ‘negative’ HSAT be considered diagnostic for ‘no OSA?,’ Dr. Hwang answered, “We believe emphatically that yes, it can,” citing the following reasons and recommendations:

  • Devices with sleep estimation are accurate and do not underdiagnose OSA, even in poor sleepers
  • Negative HSAT studies in those with a high pretest probability should have a repeat 2-night HSAT rather than PSG
  • Multi-night HSAT should be considered standard

Bigger Changes

Moving into what he called “bigger changes,” Dr. Hwang discussed the inherent unreliability and variability in using the AHI as a singular index to measure sleep apnea. Dr. Hwang is in favor of making OSA into clinical diagnosis, rather than being completely dependent on a sleep test. In shifting toward clinical diagnosis, he also supports the use of wearables (specifically oximeters) to enable multi-night testing and follow-up assessment.

“This is not to say there is no role for PSG,” Dr. Hwang explained. “We believe that PSG is very important. The key for us is to identify the proper patients and allocate them to the proper testing modality.”

COVID-19 has catalyzed what Dr. Hwang called a downshift in testing, meaning instead of doing PSG’s for all patients, his team will be implementing home PSG’s (Type II studies) using the Nox A1 for non-respiratory failure patients while reserving attended PSG’s for those with respiratory failure. This “downshifting” concept extends to using HSAT as the primary and best modality for OSA instead of PSG and even using wearables for many patients to support a clinical diagnosis for OSA.

“Much of the inspiration for us to move toward a clinical diagnosis approach and to downshift the intensity of our testing approaches for the diagnosis OSA is based on the unreliability of the apnea hypopnea index,” Dr. Hwang shared. “It’s been well shown that the AHI does a poor job of phenotyping patients…one of the directions we’d like to move in is to develop new metrics that are more informative of patient phenotypes.”

Really Big Changes

Looking ahead, Dr. Hwang is excited for the potential of using artificial intelligence and machine learning to develop metrics that better inform physicians of patients’ cardiovascular risk, neurocognitive impairment, and the likelihood of responding to non-CPAP treatments. Using this ML approach would involve allowing a computer to create metrics to inform a given patient’s outcomes.

“The data set that we could be looking to integrate into this would be PSG and HSAT metrics,” Dr. Hwang shared. “It would also include the raw data from sleep studies, as well as complimentary non-sleep study data, and then allowing the computer to develop metrics that can be more informative for us and make better decisions or help patients make better decisions in regards to their care.”

Ultimately, Dr. Hwang believes the time saved from not scoring studies can be converted to emphasizing follow-up care and care management, which he sees as an investment in improving clinical outcomes.

Click below to experience Dr. Hwang’s full presentation on how new diagnostic paradigms and alternative approaches to diagnosing sleep disorders are shaping the future of the industry.

Topic: Research & Publications