Semantic and cultural equivalence of the working alliance inventory short-revised scale for therapeutic alliance in family medicine: Lessons learned in Slovenia

The long-term doctor-patient relationship can also be described by the concept of therapeutic alliance. The most commonly-cited definition of this was first articulated by Bordin (4) in 1979; he argued that the construct consists of three components: the bond between the therapist and the patient; therapist-patient agreement on the goals of treatment; and therapist-patient agreement on the tasks of treatment. Therapeutic alliance also involves an assessment of doctor-patient trust, communication, and patient cooperation (5). It thus exceeds the paternalistic doctor-patient relationship and implements a model of shared decision-making, where the relationship is based on mutual trust, understanding, and the doctor’s empathy. Scales for assessing therapeutic alliance were first developed and validated in psychotherapy (6), and proved to be a useful tool in assessing psychotherapeutic alliance when dealing with both in- and outpatients (7). The scales have not yet been specifically altered for use outside psychotherapy, but can nonetheless be helpful in the evaluation and improvement of the therapeutic relationship elsewhere (8). Several research studies show that therapeutic alliance is also associated with better treatment results in clinical medicine (9). A study evaluating patient outcomes in cardiac rehabilitation programmes showed that a strong therapeutic alliance could play an important role in achieving favourable results (10). Its use, therefore, also seems to be applicable to family medicine.

The importance of using validated assessment tools has been receiving growing attention, with researchers becoming more aware that tools and techniques with established validity and reliability produce more consistent and accurate results.

Internationally, several tools have been used to assess therapeutic alliance in previous research (8, 11, 12, 13, 14, 15). In Slovenia, different studies have been conducted assessing various doctor-patient relationship attributes, such as empathy (16) or patient satisfaction (2), but no study has validated or even used the therapeutic alliance scales. This survey is the Slovenian part of an international research study aiming to validate the WAI-SR scale for therapeutic alliance for patients and physicians throughout Europe (8).

A WAI-SR instrument that has been previously validated in one language is not automatically equivalent to the same instrument in another language and/or culture. The equivalence between translated versions of the questionnaire is important for its international comparison.

Slovene is a language spoken by only about two million people. However, scales from other languages still need to be translated and made equivalent to the original language in terms of concepts (the concept must exist in different cultures (17) and semantics, i.e. “equivalence in meaning between the source and the question wording” (18). International research also requires scales to be culturally equivalent, to enable understanding, interpretation and assessment of the subject, that is equal or similar across different cultures. However, cultural factors cannot be seen at the level of the form or meaning of language, and exist only in the background. Since cultural factors are those relating to value systems, geographical situation, traditions, religion, etc., it is important to consider any impact that a culture or way of life can have on wording. The procedure for testing cultural linguistic equivalence consists of the evaluation of the back-translated version, test-retest by bilingual respondents, adaptation of the translated version, and a final cultural check by a principal researcher in the target country (19).

The aim of this study was to obtain a culturally consistent translation of the two WAI-SR scales (one concerning physicians and one concerning patients) and their scoring key. The semantic and cultural equivalence process is also presented.

An international group of researchers was formed under the umbrella of the European General Practice Research Network (EGPRN), led by the University of Brest, consisting of ten national research teams simultaneously working on a translation procedure following the same protocol, aiming to develop a tool available and equivalent in different languages and cultures. In this paper, we present the Slovenian part of the study.

With the aim of obtaining semantic, idiomatic, experiential and conceptual equivalence in translation, both scales and the scoring key were translated by an e-mail with a forward and back translation using a Delphi procedure. Afterward, a cultural equivalence was performed to adapt the translations within the national context, in order to ensure the homogeneity of the scales throughout Europe.

A small group of four experts was formed for the forward translation from English to Slovene.

A convenient sample of 30 practising and academic Family Medicine Doctors (FMDs) were invited to participate in the Delphi method to achieve consensus (20). All participants were provided with a written explanation of the aims and procedure of the study, and signed a statement on voluntary participation. Among those thirty invited experts, four did not accept the invitation, with one saying he was too busy, and three not replying.

Two independent English language translators undertook back translation.

The WAI-SR scales are a 12-item questionnaire for the patient and a 10-item questionnaire for the physician, assessing the three main features of the therapeutic alliance between them: goal, tasks, and bond (8). In this scale, the patient or physician rates each item on a 5-point Likert scale from “1 – rarely or never” and “5 – always”. The higher the score, the better the therapeutic alliance. The scoring key provides instructions for the evaluation of the scales.

The group of four translators who did the forward translation consisted of two FMDs, one linguist, and one psychologist; three women and one man, with a mean age of 53 years (range 43–60). All were fluent in English.

26 experts (FMDs) participated in the Delphi part of the study. Among them, 3/26 were male and 23/26 female; their mean age was 40.7 years (range 27–60) and the mean number of years of working in the practice was 12.3 years (range 1–34). 18/26 experts were involved in teaching, and 19/26 were researchers. Of the whole group, 7/26 participants worked in a solo practice and 17/26 in a group practice, while 2/26 were still trainees; 13/26 worked in a rural or semirural environment.

For validation, each item had to be rated at 7 or more by at least 17 participants. The first Delphi round for the WAI-SR patient scale showed acceptable agreement in all but one statement (Q8), where only 12/26 (46%) of the participants rated the translation with 7 or above (Table 1).

In the second Delphi round concerning Q8, the participants proposed 18 alternative translations. We present some of the suggestions in Supplementary material – Table 1. Agreement was reached on a revised translation.

The first Delphi round for the WAI-SR Family physician scale showed agreement in all but two statements (Q2, Q10). Q2 was rated as adequate by 15/26 (58%) of participants, but Q10 by only 10/26 (38%) participants (Table 2).

Numerous alternative translations were again proposed in the second Delphi round – 6 for Q2 and 5 for Q10. Some of these are presented in Supplementary material – Tables 2 and 3. Again, agreement was reached on a revised translation.

The Slovene version of the translation for both scales was issued after the second Delphi round (Table 3, Supplementary material – Table 4).

Consensus was achieved in two rounds of the agreement process between the two professional translators for both WAI SR scales. Consensus on the back-translation of the scoring key was achieved in four rounds.

The work on cultural equivalence highlighted three potential problems with translation: In Q10 of the patient scale, “treatment” was translated as “consultation”. We concluded that, considering its original use in psychotherapy, the word “treatment” did not relate solely to medical treatment but to the entire process of doctor-patient consultation. For this reason, the translation relates to the entire process of consultation and not only to treatment actions.

Also, in the patient scale, Q12 was originally in the active voice, but was translated to the passive voice in the validated Slovene translation. After the cultural check was carried out, the national team agreed that the use of the active voice was more suitable, since it emphasized the patient’s active role in the consultation and corresponded to the “shared decision-making” model (Supplementary material – Table 5).

In Q8 of the physician scale, the discussion on cultural equivalence revealed that there was a difference between ‘the common perception of a goal’ (as in the original version) and ‘common agreement on the goal’ (as translated). We consulted the linguist and appropriately altered the validated translation so that the original meaning of the statement was retained (Supplementary material – Table 6).

The final version of the WAI-SR Slovene translation was accepted after the second Delphi round, including these cultural adaptations.

The scoring key contains instructions concerning the evaluation of the scale. The same procedure was used for the translation of the scoring key as for the WAI-SR items. It was validated in the first Delphi round with Q1, Q2, and Q3 each having one evaluation of <7, and all the others (except the last item) >7. Item 11 was adapted after consulting the author of the scale, AO Horvath, who gave additional instructions.

Only two rounds of the Delphi method were needed to achieve a consensus on the translations of all the items. Cultural equivalence of the back-translation was obtained after some minor adaptations were made. The process showed that a simple literal translation was inappropriate, and rigorous efforts must be made to ensure the meaning and intent of the original items are maintained so the scale remains relevant.

The equivalence procedure in the translation of the two scales assessing therapeutic alliance was complex and time-consuming, but it served well for the purpose of semantic validation. The same procedure was used to validate and achieve equivalence in the translation of the definition of multi-morbidity (21, 22) and to validate the WAI-SR questionnaire in other countries (23). The translation into Polish showed the feasibility of the procedure, taking only one Delphi round to achieve consensus (23). The advantage of this procedure also lies in the fact that it was simultaneously taking place in several European countries with different linguistic bases, which provided the opportunity to discuss the difficulties national and local research groups met with while translating the original WAI-SR scales.

The Delphi method was used to validate the agreed forward translation and has been shown to be suitable for exploring areas where controversy, debate or a lack of clarity exist. Within this process, translations of WAI-SR scales were actively tested in representatives of the target population or language group to determine whether the respondents understood the questionnaire in the same way as the original. We feel that the use of this method for translation was legitimate, since it provided an accurate consensus technique (24).

Ideally, every questionnaire translation should undergo a cultural equivalence to identify and resolve any inadequate expressions in the translation, as well as to sort out any other discrepancies between the original items and the back-translated ones. The first steps in the process were inspired by the work of Streiner et al. (25). The standardized approach for the cultural adaptation of patient-measured outcomes was confirmed in recent guidelines (26, 27). In this study, we followed the recommendations at all stages: in the first part of cultural adaptation by using the Delphi method, because we recognized this as the best option given the specifics of our language, social and cultural context, and then by the supervision of the researchers led by the University of Brest, who oversaw the adaptation of the questionnaire and the cultural adaptation based on the back-translation. This was to ensure that the items were translated considering their structure as well as the suitability of their content.

Given that translation is the most common method for preparing instruments for cross-cultural research, we must be alert to the pitfalls that threaten validity.

Firstly, when translating scales such as the WAI-SR, it would be best if the forward translation was carried out by professionals who fit these criteria: familiar with the terminology used in the questionnaire; knowledgeable about the subjects covered; experienced in translating scales from (as here) English; and have Slovene as their native language. The content of the WAI-SR covers the fields of psychology and medicine, and its translation must be understandable by both physicians and patients. In Slovenia, we were unlikely to find a professional translator who would meet all these criteria. Creating a group of two family medicine doctors, a psychologist and a linguist to carry out the forward translation solved this problem.

Secondly, we stated that the experts carrying out the consensus procedure consisted of individuals who were fluent in English. However, the method of evaluating fluency in the language is debatable. Proficiency in English was assessed in two ways: one was self-evaluation, and the other was the number of English publications of each of the participants. The latter, in particular, may not be a powerful tool for showing language fluency; however, it was a pragmatic and feasible solution.

Thirdly, the Delphi group was not representative of the community of Slovenian family medicine doctors (FMDs) – men were underrepresented and the percentage of the academic FMDs involved was higher than the Slovenian average. But considering that the Delphi method is a qualitative one, population representativeness is not necessary. It is more important that all the characteristics of the participants that can influence decisions regarding validation are represented, such as different ages, location of practice, years of experience and involvement in the academic side of family medicine.

Finally, it would have been preferable if the back-translation had been made by an independent translator fluent in Slovene but whose native language was English. Since no such translators were available, we settled for two independent licensed Slovenian translators who had no previous knowledge of the WAI-SR scale.