The cohort investigated in the current study was a control group of a 12-month cluster-randomized controlled trial (C-RCT) assessing the (cost-)effectiveness of a novel care approach (motivational interviewing) compared to usual care in improving adherence to wearing orthopaedic footwear [18]. The trial was registered in the Netherlands Trial Register, NL7710 [18] (Available on the International Clinical Trials Registry Platform). The trial was assessed as exempt from medical ethical approval by the ethical committee region Arnhem–Nijmegen, the Netherlands (NL68567.091.19) according to Dutch law, and its protocol has been published in detail elsewhere [18]. The study protocol was approved by the Ethical Committee of Behavioural, Management and Social Sciences faculty of the University of Twente (file number 190141) [18].

All participants had a temperature sensor built in their orthopaedic footwear to monitor daily wearing time (hours/day) during 12-month follow-up. The primary study outcome was mean overall daily wearing time. The secondary outcomes were wearing time patterns, assessed by calculating participants’ (in)consistency of wearing orthopaedic footwear, comparing differences between weekdays (Monday through Friday) and weekend days (Saturday and Sunday), and investigating seasonal differences. Factors potentially associated with orthopaedic footwear (i.e. participants’ demographic, disease-related characteristics, and footwear usability) were collected by questionnaires and from participants’ medical files.

Participants were recruited at locations of Voetencentrum Wender and Voetmax Orthopedie, located in the east of The Netherlands. Eligible participants were informed about the study by the podiatrist and received an information brochure and informed consent form. After participant’s permission, the coordinating investigator contacted the participant in order to further explain the study. Thereafter, the participant had minimal one week to decide to participate. Recruitment started in July 2019 and was completed in January 2021. Participants were followed for 12 months. The orthopaedic footwear were prescribed by a medical specialist who was experienced in treating people with diabetic foot disease. Participants received usual care, as provided in standard clinical practice in the Netherlands in accordance with evidence-based guidelines [19].

Inclusion criteria were: diagnosis of diabetes mellitus type 1 and 2 patients; age ≥ 18 years; loss of protective sensation (LOPS) and/or peripheral artery disease (PAD), and prescribed with orthopaedic footwear for foot deformities (International Working Group on the Diabetic Foot (IWGDF) risk 2–3) [11]. All participants were screened for eligibility by trained podiatrists. LOPS was measured using the 10 g Semmes–Weinstein monofilament [20] and PAD using an audible handheld Doppler (Huntley Digital Doppler®; Huntleigh Healthcare Ltd, Cardiff, Wales), with the diagnosis based on presence or absence of triphasic pedal Doppler waveforms [21]. Exclusion criteria were: inability to follow study instructions; active Charcot’s neuro-arthropathy; foot infection; or being unable to walk. Written informed consent was obtained from each participant prior to inclusion in the trial.

On the informed consent form, participants agreed to the sensor placement and data storage. In both the information brochure and informed consent form participants were not notified that the sensor was used to monitor daily wearing time; it was only described as temperature monitoring sensor. Logged temperature data were collected from the microsensors every three months. These moments were mostly combined with regular appointments with a pedorthist or podiatrist. Otherwise data were read out during an additional appointment or at the participant’s home. Participants who withdrew or were deceased before the first sensor reading were excluded from further analysis. Drop-outs after the three-month mark were included in the analysis, including reason registration for withdrawn.

Measuring days from periods in which participants (re-)experienced complications (e.g. diabetic foot ulcer, lower-extremity amputation, or hospitalization) that could have affected wearing time were excluded from analysis. These complication periods were selected by retrospectively screening participants’ medical files after study completion. Whenever either the start or end date of a complication period was unknown, an exclusion period of 165 days was used based on diabetic foot ulcer (DFU) healing time showed in a recent study conducted in the same geographical region [22].

Every pair of orthopaedic footwear that participants possessed and used at study entry (i.e. earlier prescriptions) or that was prescribed and provided during follow-up was included in the study and equipped with a microsensor (Orthotimer®; Rollerwerk medical engineering & consulting, Balingen, Germany). The sensor was placed in the medial arch of the shoe insole because of sufficient place in the insole, relatively low pressure from the foot, and its previous validation at this location [23]. The sensor stored temperature with a date- and timestamp every 20 min and had a storage capacity of 133 days before overwriting the oldest data. At 12 months, participants were asked to fill in the Monitor Orthopaedic Shoes (MOS) questionnaire to measure their perception regarding their orthopaedic footwear use and usability, and their subjective assessment of their wearing behaviour [24].

The total daily wearing time of all pairs of orthopaedic footwear during the 12-month follow-up was based on logged temperature data with date- and timestamps from the sensors, and calculated with the validated Groningen algorithm, version 2, using Matlab (R2017a, The MathWorks, Inc., Natick, Massachusetts, United States) [23, 25]. The primary outcome was the participants’ mean overall daily wearing time (hours/day) during the study, and was calculated as:

$$mean\,daily\,wearing\,time= \frac^}\sum_^}daily\,wearing\,time(\frac)}_}$$

Besides wearing time, adherence to wearing orthopaedic footwear was calculated as percentage of wearing time of a total assumed 16 h out-of-bed daytime, to compare outcomes with previous studies using the same adherence definition (adherent ≥ 80%, medium adherent ≥ 60% < 80%, non-adherent < 60%) [10, 14, 26]. Missing data (i.e. due to delayed sensor readings or drop-outs after three-months) or invalid data (i.e. summed daily wearing time ≥ 24 h or measuring days from periods in which participants (re-)experienced complications) were not imputed.

Secondary outcomes were the wearing time patterns and factors potentially associated with wearing time. Patterns based on (in)consistency of wearing orthopaedic footwear were assessed by calculating the coefficient of variation (CV) for each participant over the 12-month follow-up, defined as the ratio of the standard deviation to the mean wearing time [27]. The CV is a standardized measure of dispersion. Participants were split into tertiles from low to high CV. Participants in the low CV tertile had the most consistent wearing pattern and those in the high CV tertile had the most inconsistent wearing pattern. To assess seasonal differences in wearing time, astronomical seasonal periods were used; Spring (21st of March – 20th of June), Summer (21st of June – 20th September), Autumn (21st of September – 20th of December), and Winter (21st of December – 20th of March). Participants were included in the comparison of seasonal wearing times when at least 50% of seasonal days were assigned as valid during each season.

Demographic data (i.e. gender, age, body mass index (BMI), education level, working situation, living situation, self-reliance, dependence on an assistive device) and disease-related characteristics (i.e. diabetes type, diabetes duration, IWGDF risk profile) were collected using participants’ medical files and self-report at study entry. Footwear usability variables (i.e. walking ability, perceived walking change by orthopaedic footwear, shoe fit, shoe walking, shoe weight, donning and doffing, aesthetic, aesthetic perceived by others, number of orthopaedic footwear pairs, footwear possession, owns regular off-the-shelf shoes, satisfaction with my wear of orthopaedic footwear, orthopaedic footwear wearing goal reached) were collected using the MOS-questionnaire at 12 months.

Statistical analysis was performed using SPSS statistical software (V.28.0, SPSS, New York, USA), with significance level of p < 0.05. Wearing time was stated to fit a normal distribution (Anderson–Darling test; p = 0.368). Descriptive statistics for wearing time were calculated as the mean (SD) for all participants, wearing (in)consistency subgroups (low CV, medium CV, and high CV), adherent subgroups (non-adherent, medium adherent, adherent), weekdays, and weekend days.

One-way analyses of variance (ANOVA) tested for differences between (in)consistency subgroups, adherent subgroups, week and weekend days, and seasonal periods. Tukey–Kramer post-hoc analyses were applied for pairwise comparisons. Univariate linear regression tested the associations with the dependent variable daily wearing time for all dichotomous and continuous independent variables. Variables with p < 0.20 were entered into a forward multivariate linear regression analysis to identify unique determinants of wearing time. Collinearity between independent variables was tested by linear regression, where Pearson’s correlation coefficients ≥ 0.70 were defined as correlated. In the event of collinearity where both variables also had a near significant (p < 0.20) correlation with wearing time, only the variable with highest association with daily wearing time was entered in the multivariate linear regression model. Post-hoc power analyses based on a two-sided alpha of 0.05 and power of 0.80 were performed (version 3.1.9.7, G*Power, Germany) to test whether the sample size met for subgroups comparisons and multivariate linear regression analysis.