apacitive sensor constructed of thin polyurethane foam between conductive silver fabric. A 2-inch by 2-inch, 0.4-cm thick sensor was developed to measure pressure between the lymphedema garment and the skin (see Figure 2). The polyurethane foam has negligible “memory”: it has been found in mechanical engineering testing to retain 98% of its original thickness after 22 hours of 30% compression at room temperature.19,20 Hence, it retains its initial capacity to store charge after prolonged compression. The sensor is not yet commercially available.19
The sensor was incorporated within a Bluetooth peripheral device (LightBlue Bean; Punchthrough Technologies, San Francisco, California) that communicates with a small microcomputer/Bluetooth low energy (BLE) transmitter/receiver. The peripheral device uses the ATmega328 microprocessor (Atmel Corp, San Jose, California), with software based on Arduino architecture.21 An Arduino C++ sketch reads the capacitance, finds a running average, and transforms the capacitance into pressure in mm Hg.
The sensor, microprocessor, and Bluetooth transmitter/receiver were wrapped in polyethylene under vacuum packaging.22 The advantages of using this technique include its strength and compactness, ability to measure pressure (rather than force), and inertness (polyethylene wrap is used for food storage) (see Figure 2).
A smartphone (with either an iOS Phone or Android OS operating system) served as the central Bluetooth device. The Bluefruit LE Connect Version 3.3.121 (Adafruit Industries; New York, NY)23 is available for both devices. The Bean Console, Version 1.0.0 (Punch Through Design, San Francisco, California), which is not longer sold or supported, is iOS only.24 Pressure readings on these apps scroll up from the bottom of an emulated serial terminal screen at a rate of 1 per second (see Figure 3).
Calibration. Before the pilot study, the author calibrated each device. The device was placed in a small pneumatic pressure chamber, the air pressure was increased from 0 mm Hg to 100 mm Hg in 20 mm Hg increments over a period of less than 3 minutes, and pressure measurements were recorded at approximately 10-second intervals. The sensor output was measured against the output of a commercially available pressure monitor (Thermo Fisher Scientific, Hampton, New Hampshire. A traceability certificate for the Thermo Fisher Scientific monitor has been filed in compliance with the National Institute of Standards and Technology).25 The published accuracy of the Thermo Fisher Scientific monitor is ± 2.3 mm Hg. The study device calibration data were analyzed as follows: a typical calibration curve is a second-order polynomial. This polynomial takes the form Ax2+Bx+C. For the calibration curve illustrated (see Figure 4) A = -0.00004, B = 0.2075 and C = 0.8399. For a representative sample of calibration curves (n=10), the coefficient of variation R2 ranges from 0.9992 to 0.9999.
The global error of calculation was defined as the average |(P-Pref)/Pref|. At a given capacitance, Pref is the reference pressure on the calibration curve, and P is the measured pressure. For 50 determinations over 10 sensors using the same reference pressure monitor, the global error was 2.2%.15,26
Calibration constants A and B of this equation were entered into a program in the computer language C++ to yield a digital output pressure. The constant C (typically less than 1 mm Hg) was neglected. The pressure was verified within 24 hours of device fabrication. When verified, the unit typically displays pressures ±2 mm Hg. At the present level of development of this device, the author (not a manufacturer or testing laboratory) calibrated the sensor.
Procedure. Figure 5, Figure 6, and Figure 7 illustrate device performance in practice once the calibration constants are entered. The investigator inserted the sensor/peripheral device into the lowest pocket on the posterior side of the hook-and-loop closure with the sensor visible (see Figure 5). As the practitioner affixed straps distally to proximally, the display on the mobile device (on the right of the figures) scrolled up from the bottom of the screen at 1 reading per second. Midway through the process, the display showed that the pressure on the sensor was 17 mm Hg (see Figure 6). The pressure reached a maximum of 44 mm Hg as the practitioner secured all the straps (see Figure 7). It should be noted that 44 mm Hg is higher that the study target of 35 mm Hg. These illustrations are from a video created before the start of the current study.
Study protocol. Study participants comprised staff and personnel of the Hyperbaric Wound and Edema Center, Fort HealthCare and Fort HealthCare Therapy and Fitness Physical Therapy Center (Fort HealthCare, Fort Atkinson, Wisconsin). The author verbally invited staff to participate; additionally, he invited employees with rehabilitation experience to participate. Participation was voluntary, and staff incurred no repercussions if they chose not to participate.
The author collected data in real time with pen and notebook. The participants’ professional and/or work title was recorded as a measure of experience. This was a low-risk, noninvasive, supervised trial on healthy employees at a community hospital without an Institutional Review Board. Therefore, IRB approval was deferred, and no formal informed consent documents were provided. Participants were divided into 2 groups: the instruction+feedback group and instruction only group. The selection process utilized the Excel random number generator function (Excel version 15.19.1; Microsoft Corporation, Redmond, Washington). Once randomized, participants were informed of their assigned group.
A hook-and-loop garment (medium ReadyWrap; see Figure 1) was applied on the right leg of the author, referred to herein as “the model.” All practitioners that applied a hook-and-loop pressure garment on the model received instructions from a trained and certified lymphedema specialist. The author (as the model) provided some information during training, which was limited to subjective levels of tightness (for the instruction group) or the pressure from the sensor (for the instruction+feedback group). The model was seated during training and measurement. Table 1 summarizes the teaching sequence.
At several points in the process, the practitioner applied the hook-and-loop closure to the model. For both groups, the first step in measuring the pressure on the model was to determine the zero pressure. First, the author inserted the sensor into the posterior pocket of the first strap of the garment (see Figure 5). Thee practitioner then affixed the most distal strap to the corresponding section of the loop. At this point, the author’s hand positioned the garment in place lightly against the skin on the surface of the posterior leg. In this position, the sensor was contiguous with the skin and garment. The investigator “zeroed” the sensor with a command on the mobile device. Next, the practitioner applied the remaining 4 straps distally to proximally, attempting to reach the correct tension.
For the pretraining session, the practitioner was asked to affix the hook-and-loop closure garment without specific direction. The only words stated were “try to reach 35 mm Hg.” The pressure was recorded but not shared with the practitioner.
The next step was a 10-minute hands-on training session during which the practitioner applied the garment on the model’s leg and tightened the straps. During these training sessions, the practitioners received verbal instructions from the lymphedema therapist. Practitioners in the instruction+feedback group received feedback on the actual pressure of the garment they were applying to the model. Feedback was subjective or subjective/objective, depending on the treatment arm. The after-training session was identical to the before-training session. The practitioners decided themselves when they had enough training.
Data collection and analysis. Each practitioner’s before-training and after-training performance was logged in a laboratory notebook and entered into Excel. The statistical paradigm stipulated practitioners with hands-on training with a pressure biofeedback device would be closer to the target than those who received hands-on training alone (between-group significance) and that compression approached the target in each group (within-group significance). The specific endpoint was the response to the question of how close each practitioner came to achieving the target pressure of 35 mm Hg before and after training. The “difference from target” was the absolute value of the following quantity: measured pressure minus 35 mm Hg. For each group, the pretraining pressure (Ppre) and post-training pressure (Ppost) were measured. The hypothesis was that the instruction+feedback group would be closer to the target post training than the instruction only group (ie, the mean |Ppost – 35| would approach zero for the instruction+feedback group but not for the instruction group). A paired t test was used to analyze within-group performance improvements; an unpaired t test was used assess between-group performance.