On March 18, 2020, the doors to our sleep center were physically closed. Two potential exposures to COVID-19 within a few hours, the palpable anxiety of our team, and a poor grasp of the virus and the growing pandemic moved us to make this decision. Up to that point, we could not help but feel we were playing “catch up” with our evolving set of safety measures to the escalating risk. Like so many other sleep centers around the country, a complete transition to virtual care was needed to ensure the safety of our patients and our team. It was perhaps that moment that we felt the emotional impact that our world had changed, altering both our personal lives and sleep medicine practice as we knew it. This event, while unfortunate, also provided a transformative opportunity to reimagine our identity, accelerating the efforts to bring the future of sleep medicine into the present.
Our team’s clinical evolution and innovation efforts have been guided by efforts to reconsider sleep medicine paradigms. Innovation progress was deliberate with incremental implementations that typically required repeat business cases with multiple approving parties and budgetary access. Those barriers largely dissolved once COVID-19 intensified, and a large portion of the strategies on our roadmap were put into production. In a matter of a couple weeks, our services completely transitioned to remote and virtual care, while most of the team of 55 persons were moved to “work-from-home.” A suite of technologies (automated questionnaires, automated and two-way text messaging templates, consumer wearable technologies, and population management dashboards) were put on the table (Somnoware, Inc.), and each of our longitudinal care teams (eg, adult obstructive sleep apnea, pediatrics, chronic respiratory failure, commercial driver, insomnia programs, etc) worked to embed them into new care pathways. This effort further consolidated technology as the backbone of our work and the enabler of remote virtual collaboration between sleep center personnel (respiratory case managers, medical assistants and nursing team, and physician and leadership personnel) to enhance our team-based approach. Moreover, we felt this point in time was ripe to swallow the proverbial “red pill” and approach patient care with shifted paradigms. We discuss three areas of active effort to leverage technology in this COVID-19 environment to accelerate a transition toward how we envision the future of sleep medicine.
Reimagined sleep diagnostics
Our virtual obstructive sleep apnea (OSA) diagnostic process includes utilizing a disposable home sleep apnea test (HSAT) device with wireless data transfer (WatchPAT ONE, Itamar Medical) while HSAT and PAP (positive airway pressure) setups are supported by information sheets, online videos (YouTube), automated interactive platforms (Emmi Solutions; Hwang D. Am), and synchronous provider video visits. Our more radical shift, however, is in approaching OSA diagnosis based principally on symptoms and secondarily supported by physiologic measurements and response to therapy. This “clinical diagnosis” approach reduces our reliance on traditional sleep testing and allows patient wearables to provide supportive physiologic data (eg, oximetry) to help determine OSA severity and phenotype. Its immediate impact is in limiting the need to send and retrieve potentially contaminated equipment. Broader clinical advantages include overcoming the imprecise nature of the apnea-hypopnea index (which often has dramatic night-to-night variability) through data collection over extended durations, improving disease assessment due to availability of complementary sleep/activity data in the person’s usual setting, and tracking changes after therapy initiation.
Our post-COVID-19 re-opening of polysomnography (PSG) services, after a temporary shutdown, introduces home PSG (Type II) for approximately half our patients without suspected complex breathing conditions while reserving attended PSG (Type I) for those who may require noninvasive ventilation. The immediate incentive is in reducing viral exposure by limiting patient traffic and risk of PAP trial aerosolization while also improving access to accommodate the backlog of patients requiring PSG. This approach furthers the paradigm shift to emphasizing care in the home setting. Testing in the patient’s usual environment and enabling multiple night/day testing may be clinically advantageous.
Shift in emphasis to care management
The emphasis of sleep medicine has traditionally focused on diagnostics through performing PSG and HSAT. Our field has invested tremendous effort in developing guidelines for processing sleep studies, but the scoring and interpretation of those studies is extremely labor intensive. Reimagining the diagnostic approach reduces the need to manually process studies—wearable data are produced automatically, HSAT can be auto-scored, and artificial intelligence platforms can score PSGs (Goldstein CA.), which allows a shift in resources and emphasis to follow-up care. A comprehensive discussion of technology-based tools to enhance care management is beyond the purview of this editorial. However, an overview of our current efforts includes: (1) utilizing population management dashboards to automatically risk stratify different cohorts of patients (eg, adult OSA, pediatrics, commercial drivers, chronic respiratory failure, etc) to identify patients “at-risk” (eg, based on OSA severity, symptoms, co-morbidities, and PAP adherence); (2) applying enhanced patient-provider interchange tools that include automated and “intelligent” electronic questionnaires, automated personalized text messaging/emails, and two-way messaging to deliver care; (3) utilizing remote patient monitoring to enhance holistic, personalized management, such as with remote activity/sleep trackers, blood pressure monitors, glucometers, and weight scales. We are engaged with efforts to validate the impact of these data to provide more personalized feedback, directly impact clinical outcomes, facilitate interdisciplinary collaboration, and identify acutely ill patients. Furthermore, a holistic approach beyond a narrow focus on PAP may create a positive collateral effect on adherence by targeting engagement with broader areas of health; and (4) implementing machine learning tools to directly support providers and patients (examples discussed in the next section.) Each of our teams has created workflows embedding these strategies throughout new care pathways.
Generally, our emphasis during the first 3 months after PAP initiation focuses on achieving therapy adherence, and the post-3-month period broadens the efforts to target clinical outcomes. Recent trials with low PAP usage that failed to confirm the benefit of PAP on cardiovascular outcomes (McEvoy DR, et al.) strongly suggest greater investment in cost-effective long-term strategies is imperative to increase our field’s relevance.
Application of artificial intelligence
We describe current efforts to apply artificial intelligence (AI) into clinical care: (1) We are implementing machine learning (ML) PSG scoring, which can potentially improve both the consistency and efficiency of scoring, further enabling greater investment in follow-up care. The future of sleep study processing, however, will likely depend on computer vision to “view” details inaccessible to the human eye and produce novel metrics that better inform clinical phenotypes (eg, cardiovascular risk, response to alternative therapies, etc). For example, “brain age” has been derived from EEG tracings that could reflect the degree of impact of sleep disorders on neurocognitive function (Fernandez C, unpublished data); (2) Machine learning clinical decision tools are in development to predict PAP adherence and timing of discontinuation, predict timing of cardiovascular disease onset and hospitalization, personalizing adherence targets, automating triaging of patients to home or PSG testing, and innumerable other predictions at clinical decision inflection points. Prediction outputs may be presented as risk profiles embedded in each patient’s “chart,” as personalized alerts, and in gamification strategies. For example, machine learning personalized cardiovascular risk scores can be regularly updated based on degree of PAP use to incentivize adherence; (3) Artificial providers may provide consistent, personalized, and holistic supplementary care. Many people rely on AI-bots for social support and cognitive-behavioral therapy (CBT) for depression. A sleep wellness bot, currently in planning stages, is intended to be the primary interface for many of the strategies described above that enhance engagement with PAP and therapies for comorbid conditions, provide CBT and lifestyle accountability, and collect patient reported data. This artificial provider would be a constant companion providing interactive, personalized, and continuous management to complement traditional intermittent live-person care.
The current health-care environment embodies the principle to “never let a serious crisis go to waste.” COVID-19 has accelerated the progression into the future by fostering an opening to embrace novel application of technologies to support changes in paradigms. Furthermore, health-care infrastructures that typically progress deliberately changed seemingly in a single moment. The Center for Medicare Services issued broad authorization to reimburse for telemedicine in response to COVID-19. Continued evolution in infrastructures will dictate progress with innovation, and a greater transition to outcomes-based incentives may be necessary to accommodate many of the strategies described above that rely on nonsynchronous care. But, we may be experiencing the moment when health care starts to catch up with the world in its embrace of technology. Sleep and pulmonary medicine can be a leader by providing a successful template for other specialties in optimizing chronic disease management.
Dr. Hwang is Medical Director, Kaiser Permanente SBC Sleep Center, and co-chair, Sleep Medicine, Kaiser Permanente Southern California.