About

Aaron Schwartz, MD, PhD, is an Assistant Professor in the Department of Medical Ethics and Health Policy and the Department of Medicine at the Perelman School of Medicine. His research focuses on understanding and improving health care efficiency and quality. Recent projects have examined the use of low-value care in Medicare, the incentive properties of hospital performance measures, and the role of private and public insurer coverage rules. Dr. Schwartz pursued combined MD-PhD training at Harvard, where he obtained a PhD in Health Policy (Economics) in 2015 and MD in 2017. He completed his residency in Internal Medicine (Primary Care Track) at Brigham and Women’s Hospital in 2020.

Research topics

• Internal medicine
• Medicine
• Anesthesia
• Surgery
• Computer Science
• Audiology
• Statistics
• Physical therapy

Selected publications

• Effects of controlled carbon dioxide delivery by a novel device on sleep in healthy subjects

  Journal of Applied Physiology · 2026-02-27

  articleOpen access

  A novel low-resistance open-mask system enables safe delivery of low-dose CO 2 during sleep. Inhaling 2.5% CO 2 improves sleep efficiency and increases N3 sleep in healthy adults, without CO 2 retention or physiological stress. CO 2 up to 3.5% was safe and well tolerated and showed no cumulative effects. These findings demonstrate the physiological safety of low-dose CO 2 during sleep and support its potential for treating hypocapnia-related central sleep apnea.

  PublisherOA PDFDOI

• Proximal Hypoglossal Nerve Stimulation for Obstructive Sleep Apnea in the OSPREY Study

  Annals of Internal Medicine · 2026-04-20

  article

  BACKGROUND: Hypoglossal nerve stimulation (HGNS) is used to treat obstructive sleep apnea (OSA); however, evidence is limited for patients who cannot tolerate positive airway pressure therapy. Proximal HGNS (pHGNS) provides multicontact stimulation of proximal nerve portions, with easier electrode implantation than distal nerve stimulation. OBJECTIVE: To evaluate the efficacy and safety of pHGNS in patients with moderate-to-severe OSA. DESIGN: 7-month randomized controlled trial followed by a 6-month open-label extension. (ClinicalTrials.gov: NCT04950894). SETTING: 23 U.S. health centers. PATIENTS: Adults aged 22 years or older with moderate-to-severe OSA. INTERVENTION: All patients underwent implantation with pHGNS at baseline and were randomly assigned in a 2:1 ratio to treatment or control. Therapy began at month 1 (treatment) or month 7 (control). MEASUREMENTS: Assessments included the proportion of patients achieving greater than 50% improvement from baseline in the apnea-hypopnea index (AHI) and AHI below 20 events/h at month 7 (primary end point), improvements in oxygen desaturation index (ODI) and patient-reported outcomes (for example, Epworth Sleepiness Scale [ESS]), and safety. RESULTS: = 37) achieved the primary end point, and ODI was reduced by at least 25% in 68.7% (CI, 56.2% to 79.4%) versus 37.8% (CI, 22.5% to 55.2%) of patients, respectively. Median ESS score improved from baseline to month 7 in the treatment group (10.0 [IQR, 7.0 to 14.0] to 6.0 [IQR, 5.0 to 9.0]) but not in the control group (9.0 [IQR, 7.0 to 11.0] to 9.0 [IQR, 6.0 to 11.0]). No serious procedure-related adverse events were reported. LIMITATIONS: Lack of blinding, small sample size, and short follow-up. CONCLUSION: pHGNS for OSA yielded clinically significant responses versus control at month 7, supporting pHGNS as a therapeutic option for OSA. PRIMARY FUNDING SOURCE: LivaNova PLC.

  PublisherDOI

• Intraoperative Identification and Stimulation of the Ansa Cervicalis Nerve Plexus

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  article

  Abstract Rationale: Ansa cervicalis (AC) stimulation of the infrahyoid strap muscles has been described as a promising neuromodulation strategy to treat patients with obstructive sleep apnea (OSA). The common sternothyroid (ST) trunk in particular is an attractive target given its reported anatomic consistency and complete ST muscle innervation. Here, we examine the natural anatomic variation of AC nerve branches and the effects of neurostimulation on hyolaryngeal excursion. Methods: This study was approved by the VUMC Institutional Review Board (#221998). Eligible participants were adult patients undergoing level IV neck dissection for standard head and neck cancer nodal management. Branches of the AC were isolated during routine surgical dissection. Photos and anatomic measurements were recorded. In a subset of patients, a cuff electrode was placed on the common ST trunk, ST branch, and sternohyoid (SH) branch. Videos and measurements of muscle response to neurostimulation were recorded. Results: Measurements were collected from 36 of 47 patients recruited for this study. Reasons for exclusion included significant radiation fibrosis (n=5), nodal disease burden (n=2), and other (n=4). Amongst included patients, 28 underwent unilateral and 8 underwent bilateral neck dissection. The AC was reliably identified in all patients. The average length of the common ST trunk was 38.0 (±14.7, 12-80) mm and the ST branch 20.6 (±8.3, 7-44) mm, and average diameter 2.1 (±0.6) mm and 1.5 (±0.5) mm, respectively. The distance from the clavicle to ST branch insertion into the ST muscle varied, with an average length of 11.7 (±14.3, 0-60) mm. Eight patients underwent neurostimulation. The cuff electrode was successfully placed on the common ST trunk in all 8 patients. The average amplitude of first observed muscle contraction was 0.36 (0.14-0.7) mA and maximal contraction was 0.63 (0.2-1.5) mA. Average descent of the hyolaryngeal complex was 14.7 (±3.7, 10-20) mm. Three patients had anatomy also amenable for cuff electrode placement on the ST and SH branches. In these patients, stimulation resulted in less robust muscle contraction when compared with the common ST trunk. Conclusions: The AC is anatomically variable, but the common ST trunk was reliably identified in all included patients. The minor variability observed in nerve diameter across patients and target branches suggests that a single, standardized electrode design would be feasible for an implantable neurostimulation therapy. Stimulation of the common ST trunk resulted in the greatest degree of hyolaryngeal descent amongst targeted nerve branches, further supporting its potential value as a neurostimulation target.

  PublisherDOI

• Combined Hypoglossal Nerve and Ansa Cervicalis Stimulation During Drug-induced Sleep Endoscopy in Patients With Obstructive Sleep Apnea

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01 · 2 citations

  article

  Abstract Rationale: Ansa cervicalis stimulation (ACS) has been proposed as a novel neurostimulation treatment for obstructive sleep apnea (OSA). ACS enhances pharyngeal patency by mimicking the caudal tracheal traction effect. Previous studies have demonstrated that ACS decreases airway collapsibility more effectively at the palatal, oropharyngeal lateral wall, and epiglottic levels than at the tongue base, whereas hypoglossal nerve stimulation (HNS) is expected to alleviate tongue-driven sources of airway obstruction at the tongue base and velopharynx. We hypothesized that ACS combined with hypoglossal nerve stimulation (HNS) would relieve airway obstruction to a greater level than the isolated HNS alone. This study aimed to quantify the effect of combined HNS and ACS on metrics of airway collapsibility in patients with OSA. Methods: Twenty-nine participants (Age: 54.2±11.8 years; Male: 18; BMI: 31.9±3.1kg/m2; AHI: 42.0±17.2 events/hour) with moderate-to-severe OSA underwent unilateral HNS with and without bilateral ACS via hook-wire percutaneous electrodes during drug-induced sleep endoscopy. Maximum inspiratory airflow was assessed with a pneumotachometer during flow-limited breathing across varying levels of positive airway pressure support. Pressure-flow relationships were constructed to quantify changes in pharyngeal critical closing (PCRIT) and opening pressures (POPEN) for each stimulation regimen. A linear mixed-effect model analysis was performed on a per-breath basis to characterize the effect of stimulation regimen on airflow. Results (see Figure): Combined HNS+ACS (n=22) significantly decreased PCRIT (median [Q1, Q3]: 3.1 [2.1, 5.8] cmH2O) and POPEN (5.8 [3.3, 9.8] cmH2O) from baseline (p<0.001). The decrease was greater than that of isolated HNS (n=26) in POPEN (-3.0 [-5.6, -0.9] cmH2O, p =0.02) and PCRIT (-1.7 [-4.3, -1.3] cmH2O, p=0.22) (Figure 1A). In 7 patients, ACS+HNS completely abolished flow limitation, dropping PCRIT and POPEN well below atmospheric pressure. The mixed-effect model demonstrated combined stimulation had a significantly greater effect on increasing airflow compared to HNS per unit increment of airway pressure (β = 25 mL/s/cmH2O, p=0.02), resulting in a greater reduction in POPEN (ΔΔPOPEN = -2.8 [95% CI: -4.0, -1.6] cmH2O, p < 0.001, Figure 1B). Conclusions: Combined HNS+ACS reduced pharyngeal collapsibility to a greater degree than isolated HNS, highlighting its potential as a neurostimulation strategy for OSA.

  PublisherDOI

• Proximal Hypoglossal Nerve Stimulation Improves Sleep-disordered Breathing in Moderate to Severe Obstructive Sleep Apnea: The Osprey Randomized, Controlled Trial

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01 · 1 citations

  article

  Abstract RATIONALE: Treatment of obstructive sleep apnea(OSA) with distal hypoglossal nerve stimulation(HGNS) is becoming more prevalent despite having only uncontrolled single-arm clinical trial evidence. In contrast, proximal, targeted HGNS(pHGNS) is supported by the THN3 randomized, controlled trial(RCT) where an earlier generation system yielded superior responses in active Treatment compared to inactive Controls. The OSPREY RCT of next-generation pHGNS was undertaken to confirm efficacy and safety based on primary outcomes reported herein. METHODS: Subjects with moderate-to-severe OSA (apnea-hypopnea index, AHI,15-65/hr, body mass index, BMI≤35kg/m2, no sleep endoscopy required) were implanted with the pHGNS system and randomized 2:1 to receive 6 months of active Treatment or not (Control) beginning 1-month post-implantation. The difference in the proportion of Treatment and Control subjects at Month 7 achieving AHI <20/hr with ≥50% reduction versus Baseline (responder rates, RRs) served as the primary endpoint. A Bayesian “Goldilocks” design with pre-specified interim analyses at 20-subject intervals from 50 to 150 randomizations was used to predict early success or futility. Safety was assessed by adjudication of all adverse events by an independent Clinical Events Committee.RESULTS: On interim analysis of 90 subjects, predictive probability of success exceeded 97.5%. Recruitment was halted shortly thereafter, resulting in 104 subjects in the intention-to-treat cohort. Enrollees were middle-aged (55.6±9.0 years, Mean±SD), overweight to obese (BMI=30.6±3.0kg/m2), and population-representative in sex (28/104,26.9%, female) and ethnicity (18/104,17.3%, Hispanic/Latino). Baseline parameters were notable for an AHI=35.2±12.9/hr (63/104,60.6%, severe), 4% oxygen desaturation index of 36.2±13.6/hr, and percentage of sleep time with SpO2<90% of 11.9±14.7%. Median implant operative time was 72 minutes with all discharged home the same day. At Month 7, RR was 56.7% (38/67) in Treatment versus 13.5% (5/37) in Control, for a difference of 43.2% (95% Confidence Interval, 95%CI:[27.0%,59.4%], p<0.001). Likewise, Median AHI decreased in Treatment (34.3/hr→ 11.6/hr,ΔMedian=-17.7/hr,95%CI:[-23.7/hr,-10.1/hr]) but not in Control (33.7/hr→ 34.5/hr,ΔMedian=+2.2/hr,95%CI:[-10.6/hr,+8.9/hr]). No serious adverse events were adjudicated as related to the device or procedure.CONCLUSIONS: pHGNS in OSPREY was safe and effective at improving moderate-to-severe OSA. Responses compared favorably with other HGNS trials but in a more diverse subject population having greater OSA severity. OSPREY demonstrated high-quality evidence of definitive responses in treatment vs. controls, and rapid, simpler implantation of only 2 components, requiring minimal dissection. OSPREY confirmed AHI responses in the THN3 trial, further increasing confidence in therapeutic efficacy of pHGNS in clinical practice. Absolute AHI responses in the OSPREY RCT were comparable to those in non-randomized Phase III HGNS trials, providing compelling Level-1 evidence for the pHGNS therapeutic approach.

  PublisherDOI

• Intraoperative Identification and Stimulation of the Ansa Cervicalis Nerve Plexus

  Otolaryngology · 2025-12-07 · 2 citations

  articleOpen access

  OBJECTIVE: Quantify the anatomic variation of the ansa cervicalis and evaluate neurostimulation of the infrahyoid musculature. STUDY DESIGN: Intraoperative physiology study. SETTING: Tertiary referral center. METHODS: Adult patients undergoing lateral neck dissection including level IV for head and neck cancer were recruited. Ansa cervicalis plexus branches were documented during surgical dissection. In a subset of participants, an electrode was placed on branches of the common sternothyroid trunk innervating the sternothyroid and sternohyoid muscles. Hyolaryngeal excursion with neurostimulation was recorded. RESULTS: Measurements were collected from 39 of 50 participants. Reasons for intraoperative exclusion included significant radiation fibrosis (n = 4), nodal disease burden (n = 3), and surgeon preference (n = 4). The mean lengths of the common sternothyroid trunk and sternothyroid branch were 37.6 ± 15.0 mm and 20.4 ± 8.1 mm, respectively. Their respective mean diameters were 2.1 ± 0.7 mm and 1.5 ± 0.5 mm. The distance from sternothyroid branch muscle insertion to the sternum varied substantially (12.8 ± 14.8 mm). Nine patients underwent neurostimulation of the common sternothyroid trunk. The amplitude of first observed muscle contraction was 0.35 ± 0.18 mA and maximal was 0.57 ± 0.40 mA, during which the hyolaryngeal complex descended by 13.6 ± 4.6 mm. In patients with anatomy amenable to neurostimulation of other nerve branches, stimulation of the common sternothyroid trunk produced the greatest hyolaryngeal descent (P < .05). CONCLUSION: The minor variability observed in ansa cervicalis nerve diameter across patients and branches supports the feasibility of a standardized electrode design for an implantable neurostimulation device. Stimulation of the common sternothyroid trunk resulted in the greatest hyolaryngeal descent, highlighting its value as a potential neuromodulation target.

  PublisherOA PDFDOI

• Characterizing IDC Compensation to Flow Limitation During Drug-Induced Sleep Endoscopy in OSA

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  article

  Abstract Introduction Obstructive Sleep Apnea (OSA) results from upper airway collapse during sleep, driven by anatomical, neuromuscular, and respiratory factors, causing apnea, hypoxia, and sleep disturbances. Respiratory compensatory mechanisms mitigate reduced ventilation from airway collapse by increasing inspiratory duty cycle (IDC), the ratio of inspiratory time to total breath time. This prospective study uses drug-induced sleep endoscopy (DISE) to examine IDC response to airway collapse in OSA patients. Methods This prospective study, approved by Emory University's Institutional Review Board (STUDY00003579), involved OSA patients undergoing DISE. During DISE, airflow was monitored using a pneumotachometer attached to a nasal mask, and CPAP titration was performed until non-flow-limited breathing (Popen) was achieved. Airflow (Vmax), tidal volume (TV), and IDC were calculated from the average of three breaths at each pressure. IDC was calculated as the percentage of inspiratory time over total breath duration. TV at each CPAP level was standardized as a percent of the TV during unobstructed breathing at Popen (%PopenTV). IDC compensation was defined as the slope of IDC versus %PopenTV, with steeper negative slopes indicating stronger compensation. IDC compensation was compared amongst patients with high versus low airway collapsibility (Popen &amp;gt;8cmH2O vs. Popen ≤8 cmH2O). Results Between July 2022 and July 2024, 72 patients enrolled with 58 patients completing the study. These patients had a mean (standard deviation) age of 62yrs (11), BMI 31.0kg/m2(3.4), AHI of 35.3events/hr (23.0), and Popen of 9.2cmH2O (3.8). 72% of patients were men. Both Vmax and TV increased linearly with CPAP titration while IDC decreased until Popen. High Popen (n=28 vs. n=30) had greater Vmax (29.37L/min ± 12.92 vs. 21.54L/min ± 6.68, p = 0.007), and lower IDC (0.38 ± 0.06 vs. 0.41 ± 0.05, p = 0.013) during non-flow limited breathing. The slope of IDC versus %PopenTV (IDC compensation metric) was more negative in the high Popen group (-0.18 ± 0.11 vs. -0.09 ± 0.11, p = 0.002). Conclusion This study demonstrates that IDC changes can be evaluated during DISE as a model for OSA airway collapse. Differences in non-flow limited IDC were observed between high versus low Popen subjects. High Popen patients also had greater IDC compensation suggesting their compensation allowed them to tolerate higher levels of airway collapse relative patients with low Popen. Understanding variations in IDC compensation can enhance our ability to characterize individual OSA presentations, leading to more personalized and effective management and treatment.

  PublisherDOI

• An informatics system for breath-by-breath analysis of large-scale multimodal time-series data in sleep research

  2025-04-10 · 1 citations

  article

  Sleep medicine involves handling large volumes of multi-modal time-series data that capture diverse biological signals. Analyzing these signals on a breath-by-breath basis is crucial for understanding intricate respiratory patterns, which yield valuable scientific insights and inform clinical decision-making. However, manual analysis of such data is labor-intensive and prone to error, and there's a shortage of easy-to-use analytical tool for processing data at scale. To address these challenges, we have developed a comprehensive informatics system that automates breathing cycle segmentation, feature engineering, recognition of specific respiratory patterns, and visualization of findings, using standard physiological signals from sleep studies. Our pipeline includes a deep learning model for identifying flow limitation — the definitive indicator of airway collapse with significant analytic and clinical implications. We evaluated this system using real-world patient data from 41 individuals undergoing drug-induced sleep endoscopy (DISE) procedures. The system has been deployed as a web-based platform with a graphical user interface (GUI). This intuitive application is anticipated to enhance the efficiency of breath-level sleep data analysis and expand its accessibility to a broader scientific community.

  PublisherDOI

• Ansa Cervicalis Stimulation During Non–Rapid Eye Movement Sleep in a Patient with Obstructive Sleep Apnea

  American Journal of Respiratory and Critical Care Medicine · 2025-06-13 · 1 citations

  articleOpen access
  PublisherDOI

• 0761 Detailed Three-Year Effects of Proximal Targeted Hypoglossal Nerve Stimulation on Apnea-Hypopnea Index in Moderate to Severe Obstructive Sleep Apnea

  SLEEP · 2025-05-01

  articleOpen access1st authorCorresponding

  Abstract Introduction Proximal hypoglossal nerve stimulation (pHGNS) has been shown in the THN3 randomized, controlled trial to produce durable three-year improvements in quality of life, oxygenation, and sleep-disordered breathing (Apnea-Hypopnea Index, AHI) in moderate to severe obstructive sleep apnea (OSA). Detailed AHI responses (apneas vs. hypopneas, supine vs. non-supine, REM vs. non-REM) have not previously been reported. Methods THN3 patients (no sleep endoscopy screening, AHI 20-65/hr, Body Mass Index ≤35 kg/m2) who completed 3 years annual follow up were studied (N=98), as prespecified by protocol. The non-parametric probability of superiority (A) with bootstrapped confidence intervals was used to compare median values and effect sizes at Follow-up to Baseline. Effect size was measured by A, the probability that a value in a sample (herein follow-up) was superior to a reference (herein at Baseline) with values bound by 0 and 1. A=1 would have indicated 100% likelihood of superiority while A=0.5 would have indicated equal likelihood (null result). Large effect sizes were characterized by A≥0.71 with a 95% confidence interval (95%CI) lower bound of &amp;gt;0.5. Results AHI fell from 36.2/hr at Baseline to 18.9/hr at 3 years [A=0.79 (95%CI:0.70-0.87)], due to reductions in apnea index from 8.3/hr to 2.4/hr [A=0.76 (95%CI:0.67-0.85)] and hypopnea index from 27.4/hr to 14.3/hr [A=0.76 (95%CI:0.68-0.85)]. Supine AHI decreased from 59.1/hr to 48.4/hr [A=0.71 (95%CI:0.62-0.81)], while Non-supine AHI improved from 26.0 /hr to 14.5/hr [A=0.71 (95%CI:0.63-0.81)]. Finally, REM AHI improved from 40.3/hr to 22.2/hr [A=0.77 (95%CI:0.68-0.85)] while non-REM AHI improved from 34.6/hr to 17.3/hr [A=0.77 (95%CI:0.68-0.85)]. These paired findings were comparable to unpaired responses for all outcomes from all available 3 year follow up data. Conclusion In moderate to severe OSA, pHGNS produces large, durable, clinically and statistically significant improvements in apneas, hypopneas and total AHI across all body positions and sleep stages. Sleep apnea severity was reduced in non-supine compared to supine sleep at Baseline and Follow up, suggesting that positional therapy can enhance pHGNS therapeutic responses, irrespective of sleep stage. Support (if any) LivaNova PLC

  PublisherOA PDFDOI

Recent grants

• Postural therapy reduces sleep disordered breathing and cardiometabolic stress at altitude

  NIH · $657k · 2017–2020

• NIH Grant R01HL050381

  NIH · $5.5M · 2014

• Treatment of Sleep Apnea by Targeting Leptin Signaling

  NIH · $5.4M · 2020–2025

• Lingual muscle control of upper airway patency in obstructive sleep apnea

  NIH · $1.6M · 2019–2024

• NIH Grant R01HL071506

  NIH · $1.4M · 2007

Frequent coauthors

• Philip L. Smith

  Armed Forces Research Institute of Medical Science

  263 shared

• Vsevolod Y. Polotsky

  Johns Hopkins University

  179 shared

• Hartmut Schneider

  153 shared

• Susheel P. Patil

  University Hospitals of Cleveland

  92 shared

• Christopher P. O’Donnell

  University of Pittsburgh

  82 shared

• Thomaz Fleury Curado

  Case Western Reserve University

  82 shared

• Huy Pho

  Johns Hopkins Medicine

  79 shared

• Jason P. Kirkness

  University of Miami

  78 shared

Education

• Ph.D., Health Policy (Economics)

  Harvard

  2015

• M.D.

  Harvard

  2017

• Other, Internal Medicine (Primary Care Track)

  Brigham and Women’s Hospital

  2020