We followed the STROBE guidelines for observational studies.
The original questionnaire included 6 domains with 40 questions, each question was measured with a five-point Likert-scale ranging from very high  to very low  (9). Yang et al. (2018) proposed a model for the adoption of machine translations (MT) (14). Consequently, the survey was translated to spanish using MT. The translation was reviewed independently by every research team member, then a consensus was made regarding minor changes in the syntax of the survey to facilitate understanding.
The translated instrument was converted into a google form survey and was implemented as pilot test among 65 physicians. The survey objective was to ensure that the redaction, vocabulary, design, and time needed to complete the survey was appropriate enough to prove its stability and comprehension (15).
After the expert feedback and questionnaire indicators, the research team members proposed to reduce the instrument to 13-items in 4 domains, due similarity of questions. In the same way were excluded questions that could not be used due to differences in administrative characteristics (budget management, conferences, trainings or guidelines).
Sample size was determined by multiplying the number of questions (13 items) by 5 (lower limit) and 20 (upper limit), as proposed by Suhr (2005) and other studies of validation in Ecuadorian population (16–18). The total sample was 65 to 250 doctors.
Eligibility criteria included physicians from private and public services with an active medical/clinical practice. The questionnaire was answered by 404 participants and 22 of them were excluded because they did not belong to the Ecuadorian population. The sample was divided into 2 groups, randomly selected. The first group consisted of 198 doctors for sample 1 (S1) to perform exploratory factor analysis (EFA). The second group included 184 doctors for sample 2 (S2) for confirmatory factor analysis (CFA).
A descriptive statistical analysis was implemented for sociodemographic data. Descriptive variables were presented as frequencies and percentages, and quantitative variables as means and standard deviations. The perception of each aspect of telemedicine was represented by mean response value in the corresponding group of questions. Age, gender, educational level, and area information were used as independent variables on each analysis.
The Kaiser-Meyer-Olkin (KMO) measure was calculated to evaluate the sampling adequacy of each item on the anti-image of the correlation matrix (19). All sampling measure cutoff points were set at 0.5 to be considered acceptable (19). Bartlett’s test of sphericity indicated suitability for analysis being relevant to the scale (20). Commonality analysis was performed and after this procedure the questionnaire was reduced to ensure construct parsimony and usability.
The scale structure was determined by several indicators. Initially, eigenvalues were extracted. A factor structure with eigenvalues above 1 were selected as candidate for EFA. Maximum likelihood estimation (MLE) was performed. Therefore, MLE supported the reduction of the instrument in 4 domains, according to the model fit test. The chi square showed a non-significant result, supporting the MLE.
The internal consistency was calculated by Cronbach's alpha for both, the full scale and for each of the subscales found in EFA. Cronbach's alpha measures the interrelation of the items, with values ranging between 0 and 1 (21). The results were considered acceptable above the threshold of > 0.7.
Other tests were used to indicate goodness of fit such as Chi square, Roots mean square error of approximation-RMSEA (range 0–1 with a recommended result of ≤ 0.06), Normed fit index-NFI (range 0–1 recommending > 0.90), Tucker Lewis index-TLI (0–1 recommending > 0.90), Comparative fit index-CFI (range 0–1 recommending > 0.90), Parsimony ratio (range 0–1 recommending about 1), Parsimony normed fit index-PNFI (range 0–1 recommending > 0.50), and Parsimony comparative fit index-PCFI (range 0–1 recommending > 0.50) (22–23).
Fernandez-Huerta index and Crawford were used to assess readability and estimate grade level, respectively. Scores were reported (range 0-100), higher results indicated greater readability, and a result of 60–70 represented an easy understanding for the population of ~ 15 years old (24).