Validation of the Spanish version of the Amsterdam Preoperative Anxiety and Information Scale (APAIS)

The aim of this study was to carry out a Spanish cultural adaptation of the and empirical psychometric validation of the APAIS for the Spanish population.

A psychometric validation study was conducted for adaptation of the APAIS scale.

The types of anesthesia administered included general, regional and local anesthesia in a hospitalization regimen of inpatient surgery and major and minor ambulatory surgery.

Some patients could have been administered an anxiolytic drug immediately before surgery, and were asked to complete the APAIS and GADS questionnaires prior to entering the operating room.

Sample size was calculated following the method of MacCallum et al. [31], namely: assuming a null hypothesis of a root mean square error of approximation between 0.04 and 0.08 with an alpha value of 0.05 and a statistical power of 0.8 and a maximum of 18 degrees of freedom, a sample size of 500 patients was obtained. This size was increased by 5% in anticipation of possible loss to follow-ups.

APAIS is a self-reported six-item questionnaire that has been validated for assessing preoperative anxiety. The scale is divided into two subscales exploring three aspects of preoperative anxiety: anesthesia, surgery (items 1, 2, 4 and 5) and need for information (items 3 and 6). Each question is rated on a five-point Likert’s scale, where a value of 1 indicates “not anxious at” all and 5 means “extremely anxious”. The cut-off points for the overall score established by the authors of the original version are 11 and up to 13, when used for research [20]. APAIS has also been reported to be useful as a predictor of early postoperative pain [32].

Patient sociodemographic variables, comorbidities, anesthetic risk (ASA) and history of previous surgery, amongst other data, were recorded. Anxiety was measured using the Spanish version of the Goldberg Depression and Anxiety scales (GADS) [33]. The scale had good psychometric properties (it has a one-dimensional structure that explained 72% of variance, a Cronbach’s alpha of 0.936, and a cut-off point of 10 obtained a sensitivity of 86.8%, and specificity of 93.4%), and confirmed that it could be reliably used by health professionals not specialized in mental health, such as anesthetists and surgical nurses.

Surgical nurses previously trained in the implementation of the scales informed the patients about the procedure and obtained informed consent from patients who met inclusion criteria. Data were collected through face-to-face interviews in the surgical area immediately prior to surgery.

This study was approved by the Local Research Ethics Committee Costa del Sol (Spain), CEI (002-ma-PR-APAIS) on 26 March 2015, and was carried out in accordance with the ethical principles as set out in the Declaration of Helsinki. Written consent was obtained from all the participants in the study.

It was performed according to the methodology described by Guillemin et al. [34] and ISPOR guidelines [35]. Such method includes the following stages: i) translation of the source version into the target language; ii) use of qualitative methods to check cultural adaptation to the local population; and iii) back translation from the target language to the source language to verify that the underlying meaning of questions was appropriately transferred (Fig.1).

First of all, authorization was obtained from the authors of the original scale to carry out the Spanish cultural adaptation of their scale. Next, two native Spanish language translators prepared draft translations of the original scale separately. The translators reviewed their two translations with the research group and produced an adapted version of the scale. Following this, a blinded back translation was performed on the adapted version and compared with the original scale to resolve discordances. A second adapted version of the questionnaire was then produced. The understandability, clarity and familiarity of the Spanish version were evaluated through cognitive debriefing using inquiry and reformulation techniques (Table 1).

Descriptive analysis was performed of the sociodemographic and clinical variables. The normality of the distribution of all variables was evaluated by the Kolmogorov-Smirnov test, and the skewness, kurtosis and histograms of the distributions were all ascertained. Bivariate analyses were carried out using the Student’s t and the chi-square tests, in accordance with the characteristics of the variables analysed, when they were normally distributed. Otherwise, non-parametric tests such as the Wilcoxon test and the Mann-Whitney U test were used. One-way ANOVA was used to determine quantitative and qualitative relationships where appropriate, with measures of central robustness in cases of non-homoscedasticity (determined by Levene’s test), using the Welch and Brown-Forsythe tests. If any of the assumptions necessary for the ANOVA test were not met, the Kruskall-Wallis test was performed.

Item-endorsement was determined by observing ceiling and floor effects. Internal consistency was assessed by Cronbach’s alpha; inter-item correlation and homogeneity index were determined. Construct validity was assessed by exploratory factor analysis using both principal axis factoring with oblique rotation (Oblimin), and principal component extraction with Varimax rotation on a randomized sub-sample of 155 subjects. Sample adequacy and level of inter-item correlation were previously evaluated by the Kaiser-Meyer-Olkin test and Bartlett’s sphericity test. Confirmatory factor analysis was subsequently performed both in the 374 resting subjects and in the whole sample, using the following indices of fit: CMIN/DF, root mean square error of approximation index (RMSEA) and its confidence interval (90% CI), normed fit index (NFI), comparative fit index (CFI) and goodness of fit index (GFI). Multivariate normality was determined by Mardia’s coefficient. Statistical analysis was performed using IBM SPSS version 22 and AMOS 21.

A panel of six experts (an anesthetist, two PhD nurses and three nurses with a master’s degree) reviewed the initial two translations performed by the two native speakers to produce a final version that was back translated into the source language. The panel verified the conceptual equivalence of this version and decided to replace “intervención” with “operación” (Items 4, 5 and 6), a much more common term in Spanish. Next, 10 patients aged 45 to 65 years equally distributed by age and sex underwent individual semi-structured cognitive debriefing separately. The purpose of cognitive debriefing was to assess the interpretation of questions by respondents, explore if the terms employed were appropriate, and verify that the items were culturally applicable. All respondents described the items of the questionnaire as clear and understandable. Consequently, it was deemed that no items required modification due to misinterpretation or lack of understanding.

Of the 549 subjects recruited, 20 were excluded for not meeting inclusion criteria, and 18 due to errors in questionnaire completion (Fig. 2). The characteristics of the 529 study subjects are detailed in Table 2, being general surgery the highest proportion of patients (n = 129; 24.4%), with a mean age of 50.6 years (SD: 14.54), 288 (54.4%) were men, of whom 54 (18.8%) used anxiolytics regularly, versus 5.8% in women (p < 0.0001).

According to the scores obtained from GAD7, anxiety levels in our study subjects were generally low (mean: 1.79; SD: 2.63), being higher in men with regards to women (2.35; SD; 2.93 vs 1.12; SD: 2.02; p < 0.001). No significant differences were observed in anxiety levels between the group that received anxiolytics prior to the procedure and the group who did not receive any drug (1.53; SD: 2.45 vs 1.58; SD: 2.70; p = 0.07). The highest anxiety levels were observed in patients undergoing gynecological surgery (2.72; SD: 3.18; p < 0.0001), receiving general anesthetics (2.17; SD: 2.78; p = 0.001), or using anxiolytics regularly (mostly men) (2.96 SD: 3.11 vs 1.62; SD: 2.51 p = 0.001) [36, 37]. Conversely, no significant differences in anxiety levels were observed regarding age, education level or history of previous surgery. No differences were found either between anxiety level and anesthetic risk (ASA I; 2.05; SD: 2.88; ASA II: 1.67; SD: 2.46; ASA III: 1.39; SD: 2.41; p = 0.121).

No ceiling or floor effects were detected in any item. No cases of score bunching above 85% were detected. The mean score on the scale was 12.87 (SD: 6.08) (range of possible values: 5–30). All items showed item-to-total correlations >0.20 (mean: 0.49) (Table 3). Cronbach’s alpha was 0.84 (Table 4).

Exploratory factor analysis (EFA) by oblique rotation (Oblimin) with 155 subjects, yielded a two-factor model that explained 75.78% of variance, whereas exploratory factor analysis of main axis with Varimax rotation produced a two-factor model that explained 74.8% (75%) of variance. Confirmatory factor analysis (CFA) showed that the one-factor model was better fitted than the two-factor model based on EFA, with good fitting patterns (Fig. 3). Multinormality analysis confirmed the validity of parameters (Table 5).

Criteria validity was tested by means of ICC and ROC area from APAIS scores and Goldberg Depression and Anxiety scale, obtaining an intra-class correlation coefficient (ICC) of 0.62 (95% CI: 0.55 to 0.68). The area under the curve (Fig. 4) for anxiety, as assessed by the APAIS scale, was 0.85 (95% CI: 0.81 to 0.88) with a cut-off point of 14.