The sample consisted of French-speaking patients who underwent various procedures, including orthopaedic, hand, plastic, and abdominal surgery, in three university hospitals in southeastern France. Several types of anaesthesia, ranging from regional to general, were represented. Ambulatory procedures were also included. Sociodemographic and other clinical data, such as the American Society of Anaesthesiologists (ASA) physical status measure, which assesses the fitness of patients before surgery, were collected.

The APAIS is a self-report questionnaire comprising six items (see Tables  1 and 2). Two items are dedicated to the assessment of anaesthesia-related anxiety, two items assess surgery-related anxiety, and two items evaluate the desire for information. Thus, the APAIS assesses anxiety about anaesthesia, anxiety about surgery (with the sum of both serving as the global anxiety index), and the desire for information. Patients older than 18, undergoing elective surgery (except obstetric), able to understand and read French, and able to complete a self-report questionnaire were eligible for inclusion in the study.

The items were answered during a consultation with the anaesthesiologist using a five-point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”). We evaluated the redundancy (inter-item correlation), the response rate (missing data), and the skewness in the distribution of the answers (floor and ceiling effect). A rate of missing data higher than 20% and floor or ceiling effects higher than 15% were considered high.

Within 4 to 48 h after surgery, the patients were asked to complete the “Evaluation du Vécu de l’ANesthésie” questionnaire (EVAN) 9 13 , which is a validated, multi-dimensional questionnaire that defines a patient’s reported outcome by assessing the patient’s experience during the perioperative period. This scale captures six dimensions (attention, privacy, information, pain, discomfort, and waiting) and a global satisfaction index. The score for each dimension was obtained by computing the mean of the scores for the items related to that dimension. All of the dimension scores were linearly transformed onto a 0-to-100 scale, where 100 indicates the highest possible level of satisfaction and 0 indicates the lowest. The global satisfaction score was computed as the mean of the dimension scores. The purpose of using the EVAN scores was to use a validated scale from the field of anaesthesia to assess the external validity of the APAIS in the perioperative period.

The validation process included two steps. The first step involved the production of a French version of the APAIS that is semantically equivalent to the original version. In the second step, we evaluated the psychometric properties of the French version, including its internal consistency and reliability, differential item functioning, and external validity.

This study meets the requirements of the Declaration of Tokyo 14 , and there was no interference in the physician-patient relationship.

A forward-backward translation was performed. A native English bilingual translator produced the first draft from the original version. A French bilingual expert then back-translated the items to cross-validate them.

The psychometric evaluation covered four domains: internal validity, differential item functioning, external validity, and acceptability.

The confirmatory factor analysis used the original APAIS questionnaire 10 as a reference. We sought to determine whether the model generated from the results of the original APAIS fit the data collected in France. For this purpose, a maximum likelihood confirmatory factor analysis was conducted using polychoric covariance 15 . The adequacy of the model was explored using a global index that is responsive to sample size and complexity: the root mean square error of approximation (RMSEA) 16 . We also used an incremental index that is less dependent on the sample size: the comparative fit index (CFI). An RMSEA lower than 0.08 indicates a fair fit, and a CFI higher than 0.9 indicates that the model satisfactorily fits the data.

The unidimensionality was assessed through a Rasch analysis. The Partial Credit Model (PCM), which is an extension of the Rasch model for Likert-type responses, was used 17 .

The scalability of each dimension was assessed by examining the pattern of item goodness-of-fit statistics (INFIT), and INFIT values between 0.7 and 1.2 indicate that all of the items on the scale tended to measure the same concept.

The dimensional structure of the questions on the APAIS questionnaire was also explored using a multi-trait, multi-item analysis. Each item was matched with its own dimension, and the item-internal consistency (IIC) was retained if the correlation exceeded the standard of 0.4 after the overlap correction was performed. If an item was more strongly correlated with its dimension than with the other dimensions, we considered this result evidence of the item’s discriminant validity (IDV) 18 19 .

The reliability of the internal consistency of each dimension was assessed using the Cronbach’s alpha coefficient. A Cronbach’s alpha coefficient higher than 0.7 was expected for each scale 20 .

We evaluated the differential item functioning (DIF) to assess the APAIS’ cross-population properties. DIF analyses were performed to explore the performance of the items and dimensions across several groups of patients (e.g., based on age, gender, type of anaesthesia and surgery, premedication, ASA physical status, and ambulatory course). If an item functioned differently in a subgroup of patients, the DIF would be increased. We calculated the uniform DIF to determine the probability of giving a particular answer at a given level of anxiety or desire for information across the subgroups. The Crane and Larson model 21 was used to detect the DIF, which allowed us to quantify the magnitude of the DIF that signifies an increase in the explained variance of an item by including the variable for each subgroup.

We explored the external validity with t-tests by gathering various sets of data, such as age, gender, ASA status score, and EVAN score, to assess the perioperative patient experience.

The percentage of missing answers was used to explore the global acceptability of the French APAIS among the patients. To ensure data quality, the validation analysis was not performed on records with more than 25% of the responses missing.

A database with 175 patients was created. Women represented 57% of the population, and the mean age was 51 years. Of the patients, 56% had an ASA status score of less than 2 (Table  3). The mean APAIS scores were as follows: anxiety for anaesthesia (3.3 ± 1.8); anxiety for surgery (3.9 ± 2.3); global anxiety (7.2 ± 3.7), and desire for information (5.7 ± 2.3) (Table  4).

IIC the correlation between the item scores and their dimension score (corrected for overlap). The numbers shown are the lowest and highest Pearson’s correlation coefficients. IDV the correlation between the item scores for a given dimension and the other dimension scores. The numbers shown are the lowest and highest Pearson’s correlation coefficients. NA not applicable.

There were no mismatches in the cross-validation of the items after they were back-translated; see Table  1.

The French model had the same structure as the original APAIS and explained 85% of the total variance.

The principal component factor analysis revealed the same three-dimensional structure as the original scale: anxiety about anaesthesia (two items), anxiety about surgery (two items), and the desire for information (two items). The characteristics of the items and the scales for each dimension are reported in Table  2.

The confirmatory factor analysis indicated a good fit (RMSEA = 0.069 and CFI = 1.00).

Of the items fitted to the Rash model, none produced an INFIT statistic outside the acceptable range, which indicates that the French version of the APAIS is scalable.

The item internal consistency (IIC) ranged from 0.61 to 0.72. The correlations between the items and the other dimensions (i.e., the item discriminant validity (IDV)) ranged from 0.16 to 0.61.

The internal consistency reliability and construct validity were high, i.e., the Cronbach's alpha values ranged from 0.76 to 0.84.

The level of DIF was low between the subgroups of patients based on age, gender, type of anaesthesia or surgery, premedication, ASA physical status, and ambulatory course (Table  5).

IIC the correlation between the item scores and their dimension score (corrected for overlap). The numbers shown are the lowest and highest Pearson’s correlation coefficients. IDV the correlation between the item scores for a given dimension and the other dimension scores. The numbers shown are the lowest and highest Pearson’s correlation coefficients. NA not applicable.

The French version of the APAIS was compared with other concurrent data to assess its convergent validity (Table  6). Patients older than 55 tended to feel more anxious about anaesthesia. Female patients experienced significantly greater anxiety about anaesthesia and a greater desire for information. There was no difference between the patients who received general anaesthesia compared with the patients who received regional anaesthesia. The APAIS was strongly correlated with the dimensions of the EVAN (Table  7). Anxiety about anaesthesia, global anxiety, and the desire for information were correlated with the dimensions of discomfort, waiting, and pain and with the global index of the EVAN. As expected, because the EVAN aims to evaluate a patient’s experiences with anaesthesia, anxiety about surgery was not correlated with any of the EVAN dimensions except the discomfort dimension.

*p < 0.05, **p < 0.01. The numbers shown are the Pearson’s correlation coefficients.

For each dimension, the proportion of missing values was low (ranged from 0.6 to 2.9%). These results indicate that the questionnaire was well accepted.