Secondary Logo

Journal Logo

Technology, Computing, and Simulation: Research Reports

Clinical Testing of the Apnea Prevention Device

Proof of Concept Data

Zornow, Mark H., MD

Author Information
doi: 10.1213/ANE.0b013e318204e3cb

Attempts to treat postoperative pain have become increasingly aggressive over the past decade. This trend began in the year 2000 when the Joint Commission on Accreditation of Hospital Organizations designated pain as the “fifth” vital sign (Table 1, ref. #1). Unfortunately, as an unintended consequence, more and more patients have received overdoses of narcotics, either from patient-controlled analgesia (PCA) devices or from neuraxial narcotics. This has resulted in an increased incidence of profound respiratory depression, leading to apnea, hypoxic brain injury, and even death (Table 1, ref. #2 [see section on Closed Claims data presented by Dr. Caplan] and ref. #3). The incidence of respiratory depression in patients using PCA may be much greater than has been previously estimated by intermittent assessments. In a study using continuous oximetry and capnography, 12% of postsurgical patients had episodes of desaturation and 41% had episodes of bradypnea lasting >3 minutes.1 Even more worrisome are data that suggest that 0.1%–1.3% of patients receiving postoperative analgesia will have 1 or more episodes of respiratory depression so profound that they require treatment with naloxone.2 Narcotic overdoses have been reported as the most common cause of cardiorespiratory arrest in a hospital,3 and the Anesthesia Patient Safety Foundation has identified narcotic-induced postoperative respiratory depression as a major cause of perioperative morbidity (Table 1, ref. #2). Current monitoring modalities (e.g., intermittent nursing assessments and unmonitored pulse oximetry) are inadequate to detect and treat respiratory depression in extubated, postoperative patients (Table 1, ref. #2). Intermittent nursing assessments, even if frequently performed, may not capture the rapid onset of airway obstruction, apnea, and hypoxia that can occur in many of these patients. Continuous nursing observation in an intensive care unit setting is cost-prohibitive and impractical given the large number of patients at risk. Therefore, a device is urgently needed that will not only continuously monitor a patient's oxygenation and respiration but will also stimulate the patient's respiratory drive and summon medical assistance should respiratory depression occur.

Table 1
Table 1:
Website Citations

The current version of the prototype Apnea Prevention Device (APD) is designed to collect and analyze data from an oximeter and when indicated, to deliver a series of stimuli of increasing intensity to arouse patients from narcosis. The initial stimulus is a patient-specific verbal prompt telling the patient to take a deep breath (e.g., “BOB, TAKE A DEEP BREATH”). If the saturations do not increase after a short, user-defined period of time, a cutaneous stimulus (0.5- to 1.5-second duration, 50-Hz pulse) is delivered to the skin over the dorsum of the hand followed by the verbal prompt to breathe.

This manuscript describes the ability of the APD to treat episodes of respiratory depression occurring in postoperative patients in the postanesthesia care unit (PACU). Although the APD is intended for use on the surgical wards, testing the device in the PACU environment allowed us to safely collect efficacy data on patients who were sedated with narcotics.


This protocol was reviewed and approved by the Oregon Health and Sciences IRB. Written informed consent was obtained from adult patients scheduled for surgery under general anesthesia who were likely to be extubated before their arrival in the PACU. We enrolled patients who were at increased risk for respiratory events in the PACU: patients older than 65 years, morbidly obese patients (body mass index >40), patients with obstructive sleep apnea, current cigarette smokers, those with reactive airway disease or multiple pulmonary emboli, and patients who had an anticipated need for large doses of narcotics in the PACU. We analyzed only those data collected from patients who exhibited light to deep levels of sedation (Richmond Agitation Sedation Scale (RASS)4 of −2 or less), had not received ketamine as part of their anesthetic care, and demonstrated repeated desaturations while in the PACU.

Before surgery, patients were provided with a detailed written description of the protocol and were given a brief demonstration of the functioning of the APD, including the sensation produced by the cutaneous stimulator. Surface electrode pads were applied to the dorsum of the patient's hand near the thumb, and the 1-Hz pulses (square wave stimuli of 200 μs in duration) were delivered as the current was gradually increased. The patients were asked to indicate when they first felt the stimulus and then again when it became “annoying.” The milliamp currents associated with the threshold and the “annoying” levels of stimulation were recorded. The milliamp current that each patient identified before surgery as being “annoying” was used in the PACU during functioning of the APD.

The computer used for this study consisted of a laptop running a program written in C# (Oxymon v. 1.03). A Masimo SET oximeter (model RDS-1 running version software) was connected to this computer by way of an RS-232 port. Averaging time for saturations on the oximeter was set to the minimum of 2 seconds. The resolution for this oximeter is 1%. A DigiStimII (NeuroTechnologies, Kerrville, TX) peripheral nerve stimulator with variable current output was also connected to the laptop by means of a USB port.

The user can program the APD with (a) up to 4 threshold values for oxygen saturation, (b) the number of consecutive saturations (associated with consecutive systoles) that must occur below a critical (threshold) level before the APD triggers an intervention (minimizes false triggering), and (c) the lockout period between interventions (allows patients to respond to an intervention and for the saturations to increase before any additional interventions are delivered). Whenever any of the 4 user-entered threshold saturations are exceeded, the laptop computer will trigger either a patient-specific verbal stimulus using the patient's preferred form of address (e.g., “BOB! Take a deep breath right now! BREATHE!”) or a graded (0.5 to 1.5 second) electrical stimulus to the skin followed by the voice prompt. Voice prompts were delivered via foam headphones positioned over the patient's ears. The volume was adjusted to approximately 86 dB (6 dB over ambient). For this particular study, the saturation threshold values were set relatively high to ensure sufficient triggering of the device to verify proper function. The variables used in the study were as follows: threshold saturations: 97%, 95%, 93%, and 91%; data verification: 3 consecutive saturations below a critical value; lockout interval: 60 seconds. The interventions that were delivered were as follows:

Intervention #1.

If the 1-minute lockout period is not in effect, for 3 or more consecutive heart beats with saturations below 97%, administer patient-specific verbal prompt (e.g., “BOB! Take a deep breathe right now. BREATHE!”).

Intervention #2.

If the 1-minute lockout period is not in effect, for 3 or more consecutive saturations below 95%, deliver 0.5-second, 50-Hz stimulus to dorsum of hand, followed by a more emphatic patient-specific verbal stimulus. The wording used for this verbal stimulus was similar to that of the first verbal prompt, but the tone of voice was more insistent. In no instance was the delivered current more than what the patient had identified as “annoying” before surgery.

Intervention #3.

If the 1-minute lockout period is not in effect, for 3 or more consecutive saturations below 93%, deliver 1.0-second, 50-Hz stimulus to dorsum of hand followed by the verbal prompt.

Intervention #4.

If the 1-minute lockout period is not in effect, for 3 or more consecutive saturations below 91%, deliver 1.5-sec, 50-Hz stimulus to dorsum of hand followed by verbal prompt.

All saturations and interventions were time-stamped and logged into a data file maintained by the APD. The investigator was continuously present at the patient's bedside during clinical testing of the APD and recorded the patient's responses to APD and the nature of respiratory-related nursing interventions.

Once a study patient arrived in the PACU, they received routine 1:1 nursing care, including standard physiologic monitoring and the administration of analgesics as specified by the anesthesia care team on the PACU order sheet. The PACU pulse oximeter was routinely set to alarm when the patient's saturation decreased below 90%. Supplemental oxygen was administered at the discretion of the nurse. Participation in the study did not alter in any way the types or amounts of analgesics administered to subjects. The PACU nurses caring for study patients were made familiar with the device, the protocol, and whether or not the APD was enabled to deliver interventions. In addition to the routine monitoring, for the purposes of this study, a second pulse oximeter was attached to the patient and electrode pads were placed on the dorsum of the hand and connected to the nerve stimulator. Nonocclusive headphones were placed over the patient's ears for delivery of verbal prompts. After verifying that the correct variables had been programmed into the APD, data collection was begun. The investigator observed and recorded the patient's responses to the various interventions delivered by the APD. The investigator also noted times and doses of narcotics given by the PACU nurse and any interventions made by the nurse to treat episodes of desaturation. Respiratory-related nursing interventions included verbally instructing the patient to breathe or physically stimulating the patient (shaking the patient's arm or delivering a sternal rub) and then instructing the patient to breathe.

Shaking the patient's arm was done by the nurse, using either his or her right or left hand, to grasp the patient's right or left shoulder at approximately the level of the proximal humerus and moving the arm approximately 3 cm in either an anterior/posterior direction or medially/ laterally. Usually, 4 or 5 such oscillations were delivered within 2 seconds. A sternal rub was delivered by the nurse by applying the knuckles of the proximal interphalangeal joints to the skin over the patient's sternum with several pounds of pressure and moving the knuckes approximately 1 or 2 cm, 4 or 5 times in a cephalad/caudad direction within 2 seconds.

Each intervention delivered by the APD or nurse was scored by an observer as a success if the patient took a large tidal volume breath as evidenced by an observable chest rise with a subsequent increase in oxygen saturation of 1% or more over the 60-second period that followed the stimulus. Depth of sedation using the RASS system was recorded.

In a subset of patients who demonstrated frequent desaturations and were not ready for prompt discharge from the PACU, we compared the frequency and depth of desaturations that occurred when the patients were monitored and prompted by the APD (APD ON) with routine nursing care (APD OFF). Routine PACU nursing care in these patients involved individual care (1:1 staffing). In these patients, we collected saturation data with all APD interventions enabled and then informed the nurse that we would continue to record data, but that the APD-generated prompts would be deactivated. Data collection continued and the frequency and nature of nursing interventions were recorded by the investigator. The depth of desaturations with the APD ON versus OFF was compared with a paired t test. The frequency of nursing interventions (shaking the patient's arm, sternal rub, verbally encouraging the patient to breathe) with the APD ON versus OFF was compared using the Mann–Whitney test. All oxygen saturation values with the APD ON (5047 saturation values) versus OFF (4402 saturation values) were compared using an unpaired t test. Saturation data for each patient were plotted versus time as well as the time and saturation at which each intervention was delivered (Fig. 1). Summary statistics are reported as means ± SD.

Figure 1
Figure 1:
Patient (Pt.) 23 YE: Oxygen saturations versus time. Patient was breathing room air throughout data collection. Note maintenance of acceptable oxygen saturations (Ox Sat) when apnea prevention device (APD) interventions wereon in contrast to the profound desaturations despite frequent nursing interventions (nadir of 81% at 12:17), when APD interventions were disabled (APD OFF). No nursing prompts for the patient to breathe occurred during those periods during which the APD interventions were enabled. APD1 = verbal prompt by APD; APD2 = 0.5-second cutaneous stimulus followed by verbal prompt; APD3 = 1.0-second cutaneous stimulus followed by verbal prompt; Nurse1 = verbal prompt by nurse; Nurse2 = tactile stimulus by nurse followed by verbal prompt.


Informed consent was obtained from 17 patients. No data were collected on 3 of these patients because they were taken intubated and ventilated directly from the operating room to the intensive care unit. Two additional patients received ketamine as part of their anesthetic and were excluded from the study. Two more patients were too alert in the PACU and did not meet the sedation criteria. This left 10 patients who met all of the inclusion criteria and are the subject of this manuscript. In 5 of these 10 patients, it was possible to observe them for a sufficient length of time that comparisons could be made with and without the APD interventions enabled.

Demographics of the 10 patients in our study are shown in Table 2. They received general anesthetics consisting of varying concentrations of volatile anesthetics and narcotics at the discretion of the anesthesia providers for multilevel thoraco-lumbar spine surgery and major, open, intra-abdominal procedures. Muscle relaxants were reversed at the end of the procedures, and patients met clinical criteria for tracheal extubation before their arrival in the PACU. None of the patients in the PACU was hemodynamically unstable or required pressor support. Data collection was conducted with the patients breathing room air unless otherwise noted. All patients exhibited light to deep levels of sedation (RASS4 of −2 or less). Anesthesia times ranged from 3.5 to 7.2 hours with a mean of 5.2 hours.

Table 2
Table 2:
Characteristics of Patients Included in the Study

The mean current stimulus used in the PACU was 8.2 mA (range: 3.7 to 18.8 mA). The current used for each patient was the same as that which they had identified as “annoying” in the preoperative area. A total of 125 interventions were delivered by the APD with an overall success rate of 97% (Table 3). Although on occasion a verbal prompt alone failed, electrical stimulation followed by a verbal prompt was always successful in producing an observable chest rise and a 1% or more increase in oxygen saturation within 60 seconds of the stimulus.

Table 3
Table 3:
Success Rates for the Various Stimuli Delivered by the Apnea Prevention Device

During those periods when the APD was OFF (total of 87 minutes), 33 respiratory-related nursing interventions were made. When the APD was ON (total of 97 minutes), only 2 respiratory-related nursing interventions were made. This represents 1 respiratory-related nursing intervention every 2.3 minutes with the APD OFF versus 1 every 47 minutes when the APD was functioning.

Figure 1 shows a plot of oxygen saturation versus time in 1 of the 5 patients in whom it was possible to compare the functioning of the APD versus routine PACU nursing care. In the 5 patients in whom such comparisons were possible, with the APD ON, hypoxic episodes were less profound (91.02 ± 3.35% APD ON vs. 88.54 ± 3.27% APD OFF; P = 0.001) and the frequency of nursing interventions was much less (0.25 interventions per patient-hour with the APD ON vs. 4.55 interventions per patient-hour with the APD OFF; P < 0.008). A comparison of all oxygen saturation data (APD ON vs. OFF) revealed a small, but statistically significant improvement in saturations with the APD ON (95.0 ± 2.9% vs. 94.5 ± 3.3%; P < 0.001, t = 7.5, df = 9447).


These results suggest that the APD can reliably reverse mild to moderate degrees of narcotic-induced respiratory depression in postoperative surgical patients. The prototype APD used in this study continuously monitors oxygen saturations and autonomously intervenes to stimulate the patient to initiate one or more breaths when oxygen saturations decrease below threshold values. The device nearly eliminated the need for PACU nurses to stimulate the patients to breathe. The current practice of using intermittent nursing assessments on hospital wards to monitor for respiratory depression is inadequate because it cannot detect the often rapid onset of airway obstruction and profound hypoxia3 (Table 1, ref. #2). Although supplemental oxygen may reduce the incidence of hypoxic episodes in postoperative patients, it is not without risk because patients may develop more profound levels of respiratory depression before medical personnel are aware of a potentially life-threatening situation3 (Table 1, ref. #2). Future enhancements will incorporate respiratory rate monitoring and mechanisms by which health care providers can be notified of respiratory depression in a timely fashion. Such notification may occur by way of hospital paging systems or WiFi-enabled devices (e.g., Vocera™). Respiratory data can be obtained either from commercially available capnometers or a novel bioacoustic respiratory sensor, which is now on the market (Table 1, refs. #5 and #6).

This study was conducted in the PACU on patients at increased risk for respiratory depression and desaturation. For the purposes of this study, the critical limits for saturations on the APD were intentionally set higher than what might be used on patients who had been transferred to the wards to generate a sufficient number of interventions to verify the ability of the device to sense and respond to signs of respiratory depression. To minimize the incidence of false triggering of the device, the APD was programmed to require 3 consecutive heart beats with associated oxygen saturations below a threshold value before the delivery of a stimulus. The primary clinical application of the APD will be to immediately intervene to prevent patient harm while alerting health care providers to the occurrence of repeated or significant episodes of respiratory depression. Once notified of the potential for patient harm, providers can act by decreasing PCA dosing or transferring the patient to a higher-acuity setting for closer supervision.

To make the APD clinically useful, a number of enhancements will need to be implemented. Besides the addition of a respiratory sensor as previously mentioned, the stimulating electrodes should be incorporated either into the oximetry sensor or, if a bioacoustic respiratory sensor is used, into that device (Table 1, refs. #5 and #6). Headphones are likely to be impractical for use on a ward setting; hence, placing the speakers used to deliver the verbal stimulus into the pillows or side-rails of the bed are possible options. Both of these modifications will reduce the number of wires connected to the patient and thereby decrease the likelihood of device failure. The likelihood of interventions being successful may be further enhanced if both the amplitude (decibels) of the voice prompt and the intensity (milliamps) of the cutaneous stimulus are incrementally increased. Additional refinements to the algorithm controlling the delivery of interventions may include an analysis of the rate of decrease (or increase) in saturations and trending of respiratory rate over time. Finally, an interface could be designed to automatically terminate the administration of narcotics from a PCA pump in the face of evidence of significant respiratory depression.

Although the results of this study are very encouraging, it is limited by the number of patients who were studied and the levels of hypoxia that were successfully treated. Because we felt it would have been unethical and potentially dangerous not to inform the PACU nurses when we disabled the APD interventions, they were always made aware of the state of the device and this may have influenced their behavior. Future studies will be designed to assess the APD's utility on surgical wards. Outcome variables for such studies might include the number of false alarms, oxygen saturations in patients monitored by the APD versus routine care, and the incidence of respiratory events requiring emergency interventions. Whether or not this device can be shown to reduce the morbidity and mortality associated with narcotic-induced respiratory depression in surgical patients will require large, prospective, randomized studies.


Name: Mark H. Zornow, MD.

Contribution: This author designed the study, conducted the study, analyzed the data, and wrote the manuscript.

Attestation: This author has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.

Conflict of Interest: This author notes that he holds a provisional patent on this device, but all potential rights, royalties, title, and interest in the device have been transferred to Oregon Health and Science University as mandated by his employment by the university.


The author would like to thank Mr. John Hunt and Mr. J. D. Harris for their assistance in the creation of the software and hardware interfaces used in the prototype apnea prevention device.


1. Overdyk FJ, Carter R, Maddox RR, Callura J, Herrin AE, Henriquez C. Continuous oxymetry/capnometry monitoring reveals frequent desaturation and bradypnea during patient-controlled analgesia. Anesth Analg 2007;105:412–8
2. Cashmand JN, Dolin SJ. Respiratory and haemodynamic effects of postoperative pain management: evidence from published data. BJA 2004;93:212–23
3. Fecho K, Joyner L, Pfeiffer DL. Opioids and code blue emergencies. Anesthesiology 2008;109:A34
4. Sessler CN, Gosnell M, Grap MH, Brophy GT, O'Neal PV, Keane KA, Tesoro EP, Elswick RK. The Richmond Agitation–Sedation Scale: validity and reliability in adult intensive care patients. Am J Respir Crit Care Med 2002;166:1338–44
© 2011 International Anesthesia Research Society