Validation of a gyroscope-based wearable device for real-time position monitoring of patients in a hospital

Abstract

BACKGROUND:

Monitoring patients’ position is important, but there have been few studies related to validation.

OBJECTIVE:

The objective of this study was to assess the validity of position monitoring measured using a wearable device by comparing the device’s measurements to a patient’s actual position.

METHODS:

We constructed a wearable device with a three-axis gyroscope and applied it to 10 patients who were unable to change their position independently. We compared the actual angle of the position and the angle transmitted from the wearable device using a Bland-Altman plot and a receiver operating characteristic curve.

RESULTS:

We compared the actual angle of the position and the angle transmitted from the wearable device using a Bland-Altman plot, but it was difficult to observe statistical similarity. The angles transmitted from the wearable device in the lateral and supine positions showed significant differences. The cutoff value separating the lateral and supine positions was found to be 27.1 ${}^{\circ}$ (sensitivity $=$ 100%, specificity $=$ 99.9%).

CONCLUSIONS:

Through our method, the measured values from the gyroscope-based wearable device did not accurately reflect the patient’s actual position. However, the wearable device was able to distinguish the lateral position from the supine position.

Keywords

Wearable device validation monitoring position change patients

1. Introduction

Changing patients’ position from supine to lateral is an important intervention to improve oxygenation, prevent pressure ulcer, and enhance recovery after surgery [1, 2, 3]. Some protocols associated with changing patients’ position have been proposed [4, 5]. However, compliance with these protocols has been low because tools to monitor patient positions are not widely available [6].

Recently, there has been increasing interest in all areas of the Fourth Industrial Revolution, which comprises artificial intelligence, big data, Internet of Things (IoT), and robotics [7, 8, 9]. Wearable devices, a type of IoT technology, have been applied in the healthcare field [8, 10]. Wearable devices with a small sensor applied to a micro-electro-mechanical system (MEMS) have been developed to monitor patients’ position [6, 11]. This type of wearable device is convenient because it can be simply attached to the patient’s body [4, 6, 11]. However, it is not well known how accurately the measured values from wearable devices reflect patients’ position. Thus, we collected the values measured using the wearable device with a three-axis MEMS gyroscope that corresponded with patient position and validated them against the actual values. In addition, when using a wearable device included a gyroscope, we attempted finding how much a value capable of recognizing position change.

2. Materials and methods

2.1 Hardware design and system architecture for position monitoring

We constructed a small wireless wearable device with a three-axis MEMS gyroscope (LSM6DSL, STMicroelectronics, Geneva, Switzerland) and a rechargeable battery. The device was placed in a 4 cm $\times$ 3 cm $\times$ 1 cm cuboid plastic case. The data generated by the gyroscope were the roll, pitch, and yaw values. Among them, the value that could measure the patient’s rotation angle was roll because the axis of the roll was parallel to the patient’s long axis. The serial number of each device and the rotation angle measured by the gyroscope were transmitted to the gateway through LoRa, a long range wireless telecommunications method every 10 seconds. We used the ARTIK module (Samsung Electronics Co., Ltd., Suwon, South Korea) as the gateway. The gateway was programmed to record the date and time when the rotation angle was sent. Then, the serial number of the device, the transmitted angle, and the corresponding date and time were bundled and stored.

2.2 Patients

The study was approved by the institutional review board of Pusan National University Hospital (Approval Number: 1903-015-077). This study was performed on 10 patients who were unable to change their position independently. Informed consents were obtained from all patients. Each patient wore a patient gown or suit that had the wearable device attached at the manubrium of the sternum (Fig. 1). Then, the wearable device was turned on and the caregiver changed the patient’s position at a set time for 24 hours. While the angle measured by the device was continuously transmitted to the gateway, a video was recorded for 24 hours above the patient’s head so that the real position could be checked at the same time.

Figure 1.

(a) Schematic diagram of the location of the wearable device. (b) The captured photo.

2.3 Actual data collection for validation and classification of actual positions

To check the status of the actual position, an image corresponding to the time at which the data were transmitted from the device was captured by the recorded video. In the captured picture, an imaginary line connecting the center of both shoulders was drawn. Then the angle between this line and the bed was measured to obtain the angle of the actual position (Fig. 2). Among the captured images, images captured when the caregiver changed the position were collected and classified into the lateral position group, and the supine position group. Then, the angles transmitted from the devices in each group were compared.

Figure 2.

Schematic diagram of the measurement of the angle of the patient’s actual position. An imaginary line was drawn to connect the center of both shoulders (dotted line). The angle between the dotted line and the bed was measured.

2.4 Statistical analysis

Spearman correlation coefficient and Bland-Altman plot analysis were performed to compare the similarity between the angle measured by the wearable device and the angle of the actual position. Student’s $t$ -test was used to compare the angles transmitted from the wearable device between the supine position group and the lateral position group. Then, a receiver operating characteristic (ROC) curve was used to determine the angle that would be considered the lateral position among the angles transmitted from the wearable device. MedCalc Statistical Software version 19.3.1 for Biomedical Research (MedCalc Software Ltd., Ostend, Belgium) was used for all statistical analyses.

3. Results

A total of 86,400 angles were collected from the device worn by 10 patients over 24 hours. Images of the patients’ positions were captured by a video recorded over the 24 hours in which angle data being transmitted. A total of 86,400 actual angles of the patients’ positions were collected. When the device was turned counterclockwise, the transmitted angles were negative (range, $-$ 180 ${}^{\circ}$ $\sim$ 0 ${}^{\circ}$ ). Thus, in the captured image, the angle was measured as a negative number when the patient was turned counterclockwise. The angle in the remaining patients turned clockwise was measured as positive numbers (range, 0 ${}^{\circ}$ $\sim$ 180 ${}^{\circ}$ ). Four thousand captured images were classified as the lateral position, and the rest were classified as the supine position.

For statistical analysis, all values were converted to absolute values. A positive correlation was identified between the actual angles measured from the captured images and the angle transmitted by the wearable device ( $r=$ 0.36, $p<$ 0.001). However, when the Bland-Altman plots of actual angles and transmitted angles were analyzed, the result did not indicate high agreement between the two types of angles (Fig. 3).

Figure 3.

Bland-Altman plot of a relatively low agreement, given the 95% confidence interval.

The mean transmitted angle from the wearable device in the supine position group was 9.1 $\pm$ 4.4 ${}^{\circ}$ , while the mean angle in the lateral position group was 61.9 $\pm$ 12.9 ${}^{\circ}$ . The two groups showed significant differences ( $p<$ 0.0001). The predictive value of the transmitted angles for determining whether the patient’s position was lateral was evaluated by ROC analysis. The area under the curve value for the transmitted angles was 1.000, and the cutoff value, which was calculated by the Youden index, was 27.1 ${}^{\circ}$ , with 100% sensitivity and 99.9% specificity ( $p<$ 0.0001).

4. Discussion

Wearable devices are increasingly applied in the healthcare field [8, 12]. However, validating the accuracy and sensitivity of such device is a necessary first toward obtaining accurate and objective information [12, 13]. Some studies associated with position monitoring with wearable devices have already been reported [4, 6]. Notwithstanding, there is a paucity of validation studies on the accuracy of wearable devices in the realm of position monitoring.

The gold standard method of position monitoring is to directly measure the degree of rotation. To measure this rotation, we had to look at the patient’s body above the head or under the soles. However, the method of watching from under the soles was difficult to measure because the body might be invisible depending on the position of the legs. Thus, we looked at the patient’s body from above the head and measured the actual angle of rotation. Then, the wearable device was placed on the sternum, the centerline of both shoulders. Thus, the actual measurement site and the wearable device were kept as close as possible (Fig. 1).

In this study, the wearable device could not accurately monitor patient position. We suspect that this may be due to the attachment location of the wearable device. We conducted this study by fixing the wearable device on the patient’s clothing. In this situation, it is possible that the wearable device was not parallel to the patient’s body due to wrinkling of the patient’s clothing such that the angle could not be accurately reflected. Although the device could be attached to the patient’s skin, this was not followed in this study. In general, monitoring of a patient’s position could last for several days, there is a risk that a skin rash or lesion may occur if the device is attached directly the skin. Even if the wearable device was attached to the skin, however, the body was not completely flat; therefore there may still be some errors. Furthermore, the curvature of the chest wall depending on the development of the pectoralis major muscle would also need to be taken into consideration (Fig. 1).

Although the measured values of the wearable device did not accurately reflect the actual angle, it had no problem distinguishing the supine position from the lateral position. In this study, when the angle transmitted from the wearable device was 27.1 ${}^{\circ}$ or more, the device’s sensitivity judging the lateral position was 100% and the specificity was very high at 99.9%. In previous studies related to position monitoring using wearable devices, the angle that could be distinguished by the lateral position was defined as 20 ${}^{\circ}$ , a value similar to our cutoff value [4, 6, 11].

5. Conclusion

In our method, the wearable device could not reflect the patient’s actual position exactly. However, the wearable device could distinguish the lateral position from the supine position sufficiently and help monitoring of patients’ position change in hospitals. For clinical application, sufficient verification of these devices should be made, which is a priority of future research.

Footnotes

Acknowledgments

Support was received from the Technology Innovation Program of the Ministry of Trade, Industry & Energy (MOTIE), funded by the South Korean government (Grant Number 20000515).

Conflict of interest

None to report.

References

Burgess

Wainwright

. What is the evidence for early mobilisation in elective spine surgery? A narrative review. Healthcare. 2019; 7(3): 92. doi: 10.3390/healthcare7030092.

Marklew

. Body positioning and its effect on oxygenation – a literature review. Nurs Crit Care. 2006; 11(1): 16–22. doi: 10.1111/j.1362-1017.2006.00141.x.

Sen

McNeill

Mendelson

Dunn

Hickle

. A new vision for preventing pressure ulcers: wearable wireless devices could help solve a common-and serious-problem. IEEE Pulse. 2018; 9(6): 28–31. doi: 10.1109/mpul.2018.2869339.

Pickham

Berte

Pihulic

Valdez

Mayer

Desai

. Effect of a wearable patient sensor on care delivery for preventing pressure injuries in acutely ill adults: a pragmatic randomized clinical trial (LS-HAPI study). Int J Nurs Stud. 2018; 80: 12–19. doi: 10.1016/j.ijnurstu.2017.12.012.

Tayyib

Coyer

. Effectiveness of pressure ulcer prevention strategies for adult patients in intensive care units: a systematic review. Worldviews Evid Based Nurs. 2016; 13(6): 432–444. doi: 10.1111/wvn.12177.

Renganathan

Nagaiyan

Preejith

Gopal

Mitra

Sivaprakasam

. Effectiveness of a continuous patient position monitoring system in improving hospital turn protocol compliance in an ICU: a multiphase multisite study in India. J Intensive Care Soc. 2019; 20(4): 309–315. doi: 10.1177/1751143718804682.

Lee

Han

Kim

Lee

Song

, et al. Spine computed tomography to magnetic resonance image synthesis using generative adversarial networks: a preliminary study. J Korean Neurosurg Soc. 2020; 63(3): 386–396. doi: 10.3340/jkns.2019.0084.

Nam

Kim

Choi

Han

. Internet of things, digital biomarker, and artificial intelligence in spine: current and future perspectives. Neurospine. 2019; 16(4): 705–711. doi: 10.14245/ns.1938388.194.

Nam

Seo

Kim

Lee

Choi

Han

. Machine learning model to predict osteoporotic spine with hounsfield units on lumbar computed tomography. J Korean Neurosurg Soc. 2019; 62(4): 442–449. doi: 10.3340/jkns.2018.0178.

10.

Kim

Nam

Choi

Han

Jeon

Park

. The usefulness of a wearable device in daily physical activity monitoring for the hospitalized patients undergoing lumbar surgery. J Korean Neurosurg Soc. 2019; 62(5): 561–566. doi: 10.3340/jkns.2018.0131.

11.

Schutt

Tarver

Pezzani

. Pilot study: assessing the effect of continual position monitoring technology on compliance with patient turning protocols. Nurs Open. 2018; 5(1): 21–28. doi: 10.1002/nop2.105.

12.

Lewis

Directo

Kim

Dolezal

. Validation of biofeedback wearables for photoplethysmographic heart rate tracking. J Sports Sci Med. 2016; 15(3): 540–547.

13.

Bassett

Rowlands

Trost

. Calibration and validation of wearable monitors. Med Sci Sports Exerc. 2012; 44: S32–S38. doi: 10.1249/mss.0b013e3182399cf7.