Substance abuse is a rising epidemic in the United States and has resulted in nearly 92,800 overdose deaths in 2020, of which more than 74% involved the use of opioids [1], [2]. Opioids, such as oxycodone, heroin, and fentanyl, make up the majority of overdose deaths due to their highly addictive nature and rapid onset of symptoms [3]. Opioid overdose deaths are most often due to opioid-induced respiratory depression (OIRD) [4], [5]. OIRD is the physiological result of opioid agonists binding to μ-opioid receptors within the brain that result in suppression of unconscious breathing. Not only does this reaction result in a dangerously slow respiratory rate (RR), but it also reduces tidal volume and impedes neural feedback loops that would otherwise restore normal breathing [6]. Widely available medications, such as naloxone, can quickly reverse the dangerous effects of OIRD by displacing opioids at their binding sites and restoring rhythmic breathing [7]. The challenge in OIRD treatment, however, lies in timing. Opioids act quickly on the body, and without proper oxygen supply, the hypoxic brain can become permanently damaged in as little as 4 minutes, leading to fatality soon after [8]. This is particularly concerning considering the typical response time for emergency medical services (EMS) in the United States averages 7 minutes [9]. As a result, there is a need for a monitor that would recognize OIRD events and immediately deliver naloxone, while also alerting EMS in real time.

Previous studies have shown that patients are receptive to using continuous monitoring devices given their potential life-saving capabilities [10], [11], [12], [13], and a number of researchers have proposed to monitor OIRD [10], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24]. In 2014, a study performed by Carreiro et al. found that a portable biosensor quantifying electrodermal activity (EDA) could be used to identify intravenous use of both cocaine and opioids [22]. Despite this, the study concluded that the robustness of such devices was questionable outside of controlled settings due to their lack of specificity to stress responses. Another, more recent study published by Nandakumar et al. used a smartphone-based sonar to monitor RR and apnea during opioid use in harm reduction settings [23]. The device was perched within 1 meter of the user to monitor breathing, a use case not amenable to continuous monitoring of OIRD events in outpatient settings and non-stationary settings.

Beyond academic research studies, several wrist worn, consumer-grade smartwatches from Garmin [25], Apple [26], and Fitbit [27] now use optical monitoring via photoplethysmography (PPG) at peripheral sites to report RR in addition to other metrics such as heart rate. However, these RR measurements are taken only intermittently during sleep or once every few hours, which is not suitable for detecting OIRD in free-living settings. Furthermore, commercially-available smartwatches rely on respiratory sinus arrhythmia (RSA) [28], [29], [30] to derive RR from heart rate variability. However, not everyone exhibits RSA with studies seeing RSA exhibited in only 43% of their subject pool [31] or not at all in people with underlying cardiac abnormalities [32], [33].

Considering these limitations, it is necessary to develop a device to monitor OIRD events in real-time and in naturalistic settings by capturing the underlying baseline changes in the PPG signal due to respiration and not relying on RSA alone. This is the motivation behind our device. By continuously monitoring the wearer's RR we aim to detect OIRD and alert bystanders, and it can be paired with wireless communications to alert EMS in the event of a drug overdose or incorporate automated naloxone delivery systems [16], [24], [34], [35], [36]. Our device is specifically designed to monitor respiratory events with a “wear-and-work” modality, requiring minimal user intervention outside of charging and wearing the device, while still offering continuous monitoring.

To determine the functionality of our design, the study herein focused on the ability of our PPG device to detect OIRD in vivo. Mice were chosen as subjects because they are sensitive to general anesthesia (isoflurane), opioids (oxycodone), and naloxone, thereby enabling us to monitor respiration under anesthesia, during OIRD, and following the administration of naloxone for OIRD recovery [37], [38], [39]. Our work improves upon previous studies that relied on voluntary breath holds in humans, but were not able to directly monitor the real-time effects of the opioid and antidote on RRs due to practical and ethical concerns [14], [34], [40]. By using PPG, our study improves upon previous work that used whole-body plethysmography to detect OIRD in mice [41], [42]. PPG is much more amenable to real-time monitoring of human RR in real-life settings as it would be impractical to measure whole-body plethysmography in humans in everyday settings [41], [42], [43]. Additionally, given the agreement between our device and the reference pneumatic pillow system, we propose that our device can be used to monitor RR in rodents during surgical procedures where the reference system is not convenient, like during abdominal surgeries where the chest cavity would be compromised. Our device's performance was determined by two factors: First, the ability to accurately report RR in real-time when compared to a traditional pressure pillow based small animal instruments system, and second, the ability to detect true OIRD events.