Table 2.

Details of the studies using temperature sensors alone to monitor compliance, ordered by year published.

Author and Year	Orthosis Prescribed	Study Aim(s)	Population Demographics	Description of Technology	Instructions for Use	Wear Time Estimation	Pros of Method	Cons of Method
Sangiorgio et al., 2016 [50]	Mitchell Ponseti brace sandals for clubfoot	Assess difference between prescribed and measured brace use, difference between parent-reported use and measured brace use, extent to which brace compliance affects risk of relapse.	48 patients (37 male and 11 female) aged 6 months to 4 years	Two wireless temperature loggers (SmartButton, ACR Systems, Surrey, BC, Canada) consisting of programmable data acquisition, a 3 V battery, and data storage capacity of 2048 readings for data collection every 90 min by each sensor, offset by 45 min, for 4 months. Sensors attached to outside of braces and above the heel.	Wear brace during night and naps.	Data imported to MATLAB and data from both sensors were synchronised to make a single dataset for each patient. A baseline temperature was established using the mean temperature of the dataset, and wear time was determined by finding data points above the baseline.	Objectively measured compliance. Beneficial to clinicians when interpreting parental reports of brace use.	Sampling rate was limited.
Ehrmann et al., 2018 [48]	Insoles for diabetes	Determine when patients become nonadherent to diabetic footwear. Observe possible effects of gender on adherence.	26 patients (18 male and 8 female) aged 59–76 years	Temperature sensor (Orthotimer, Rollerwerk Medical Engineering & Consulting, Balingen, Germany) was embedded into the longitudinal arch of one insole. Temperature was measured within footwear every 15 min. Sensors stored data for 100 days before overwriting the oldest data.	-	An optimal cutoff temperature of 25 °C was found by testing the sensor in healthy participants. Temperatures higher than cutoff temperature were classified as worn and lower temperatures as not worn.	Monitored compliance objectively. Improved long-term adherence and early detection of non-compliance. Patients could not feel the sensor.	No information regarding patient activity and mobility. Temperatures in footwear could have exceeded 25 °C due to environmental factors and not just wear.
Grubhofer et al., 2019 [21]	Abduction shoulder brace for post-arthroscopic rotator cuff repair	Analyse abduction brace wearing behaviour in patients who underwent arthroscopic rotator cuff surgery.	50 patients (23 male and 27 female) aged 28–79 years	Temperature sensor (Orthotimer, Rollerwerk Medical Engineering, Balingen, Germany) was invisibly placed in abdominal belt of brace. Sensor recorded surrounding temperature every 15 min.	23 h/day wear time for 6 weeks postoperatively.	If the measured temperature was above 35 °C, it was recorded as wear time. Data from the sensors were read out using computer software and displayed wear time for each day since sensor activation.	Clear to see overestimation in self-reported wearing time, so useful to have invisible sensor.	Self-reported wear time was done by patient estimation in outpatient visit; questionnaires would have been better to compare with temperature sensor. Software only shows hours worn per day and not when during the day.
Richards et al., 2020 [49]	Abduction brace for clubfoot	Observe daily orthosis wear time in patients successfully treated with the Ponseti method. Determine compliance of patient caretaker with prescribed brace treatment.	124 patients (83 male and 41 female) aged less than 3 months	Temperature sensor (iButton, Maxim Integrated Products, San Jose, CA, USA) embedded in shoe recorded temperature every 15 min. Held up to 3 months of data.	Prescribed 22 h/day for the first 90 days, then 12 h/night until 2 years old.	-	Highlighted true usage of braces.	Awareness of foot temperature being measured could have influenced brace wear. Few sensors failed to record data during a time interval.
Sood et al., 2021 [45]	Shoulder sling for postoperative use	Investigate accuracy of temperature sensors positioned in shoulder slings. Assess whether sensor could discern difference between body temperature and hot environment.	4 healthy participants (3 male and 1 female) aged 25–32 years	Compact (3.35 × 5.64 × 1.8 cm; 12.75 g) data loggers (Onset HOBO MX2201, Onset Computer Corporation, Bourne, MA, USA) with internal microprocessor, data storage, and sensors to measure contact temperature. Data was sampled every 15 min and transferred via Bluetooth. Sensors were in three locations within the sling: inner region of the bolster touching the abdomen, medial elbow region, and palmar surface of the carpometacarpal joint.	Wear the sling as much as possible but free to remove the sling to perform daily activities.	An algorithm was used to estimate wear time. Start of a wear period was determined by a temperature increase. Temperature had to remain above a threshold value for at least 30 min and not exceed a maximum threshold value. End of a wear period was categorised by a fall in temperature and the temperature being lower than a threshold value for 30 min.	Accurately measured compliance (>99% accuracy). Algorithm could discern temperature difference when donned/doffed.	Data recorded at 15 min intervals, so could underestimate wear time. Sensors with less body contact overestimated compliance. User could “cheat” by leaving sling in a warm environment. Hawthorne effect could alter patient behaviour if they are aware they are being monitored.
Swarup et al., 2021 [47]	Abduction brace for residual acetabular dysplasia	Validate efficacy of part-time bracing. Determine relationship between brace wear time and correction of pathology.	26 patients around 6 months of age	Temperature sensor (iButton, Maxim Integrated Products, San Jose, USA), costing USD 75, placed in the posterior thigh region of the abduction brace.	Wear brace for nights/naps and return in 6 months for follow-up.	Temperature higher that 75 °F was defined as the orthosis being worn and less than or equal to 75 °F was defined was not worn.	Used differences in body and ambient temperatures, and wear patterns to determine temperature thresholds.	Inaccuracies could arise from temperature sensor depending on whether the brace was worn over or under clothing.
Grubhofer et al., 2022 [46]	Abduction shoulder brace for post-arthroscopic rotator cuff repair	Investigate whether compliance with immobilisation influences healing. Define compliance rate associated with tendon-repair post rotator cuff repair.	50 patients (23 male and 27 female) aged 28–79 years	Temperature sensor (Orthotimer, Rollerwerk Medical Engineering, Balingen, Germany) was invisibly placed in abdominal belt of brace. Sensor recorded surrounding temperature every 15 min.	23 h/day wear time for 6 weeks postoperatively.	If the measured temperature was above 35 °C, it was recorded as wear time. Data from the sensors was read out using computer software and displayed wear time for each day since sensor activation.	Monitored compliance objectively.	Patient not informed about sensor until after 6 weeks—could affect behaviour.
Haarman et al., 2022 [34]	Upper limb orthoses for impairment of the shoulder, arm, or hands	Validate method to estimate orthosis wear time using temperature sensors attached to the upper body. Assess if two temperature sensors are better than one to estimate wear time. Investigate the effect of sampling time on wear time estimation.	15 healthy participants (7 male and 8 female) aged 24–67 years	Miniature (diameter: 17 mm, height: 6 mm) data loggers (DS1922L Thermochron iButtons, Maxim Integrated Products, San Jose, USA) that measure and store temperature. Attached to the body using elastic straps positioned around the chest and forearm. Data were sampled at 1 min. Android smartphone application cued user to don or doff.	Remove and re-attach straps at specified time-points. Sixteen hours of non-use and eight hours of donning and doffing as instructed by the smartphone app (intervals between cues ranged from 15–60 min).	Data obtained from temperature sensors were used to train decision tree classification algorithm to estimate wear time.	Accurate wear time estimation without direct sensor–skin contact. Accurate estimation during donning and doffing. Algorithm was evaluated with unseen data to minimise bias.	As sampling time increased, the data stored increased, and wear time error and estimation error range increased. Potential discrepancy between actual and reported timestamp due to reaction time of user (on smartphone app). Data not collected on warm days, so not trained for high temperatures.