Wearable Biosensors to Evaluate Recurrent Opioid Toxicity After Naloxone Administration: A Hilbert Transform Approach

Abstract

Opioid abuse is a rapidly escalating problem in the United States. Effective opioid reversal is achieved with the antidote naloxone, but often does not last as long as the offending opioid, necessitating in-hospital observation. Continuous physiologic monitoring using wearable biosensors represents a potential option to extend monitoring capability outside the clinical setting across the spectrum of opioid abuse including post- naloxone administration. The present study aims to identify the physiologic change that marks the cessation of naloxone's effect. Eleven participants were recruited in the Emergency Department after naloxone administration for an opioid overdose and continuously monitored using a wearable biosensor measuring heart rate, temperature, electrodermal activity and accelerometry. Hilbert transform was used to evaluate a 90- minute post naloxone time point. Physiologic changes were consistent with the onset of opioid drug effect across parameters, but only changes in heart rate and skin temperature research statistical significance.

Figure 1

Empatica E4 wrist mounted biosensor

Figure 2

Accelerometer data from three axes XYZ from an individual (participant 10). The green line marked at time =0 represents N90. The yellow and green shading mark representative 5 minutes windows before and after N90 respectively.

Figure 3

Amplitude estimated using Hilbert transform for the 5 minutes window (shaded regions in Figure 1) for the same participant. Left panels with blue lines represents before and right panel with red lines represents after

Figure 4

Distribution of amplitudes along with the best fit for the 5-minute window (shaded regions of Figure 2). Gamma function with shape and scale characterizing parameters are found to be the best fit

Figure 5

D_k, Heart Rate(HR), EDA and Temperature estimated for 5-minute window with 4 minute overlapping before (blue curve) and after (red curve) for participant 10.

Figure 6

Mean and standard error of D_k, HR, EDA and Temperature values obtained from all subjects (N=11) before and after N90