Methodological considerations when translating "burnout"

No study has systematically examined how researchers address cross-cultural adaptation of burnout. We conducted an integrative review to examine how researchers had adapted the instruments to the different contexts. We reviewed the Content Validity Indexing scores for the Maslach Burnout Inventory-Human Services Survey from the 12-country comparative nursing workforce study, RN4CAST. In the integrative review, multiple issues related to translation were found in existing studies. In the cross-cultural instrument analysis, 7 out of 22 items on the instrument received an extremely low kappa score. Investigators may need to employ more rigorous cross-cultural adaptation methods when attempting to measure burnout.

1. Introduction

Burnout is a global work-related phenomenon that, as multiple studies in mostly English-speaking developed countries have demonstrated, is associated with the quality of working conditions, interpersonal relationships, role conflict, and workload (Maslach, Schaufeli, & Leiter, 2001). Burnout is specifically a work-related syndrome that is commonly found in people who work in human services that require significant human contact (Maslach et al., 2001). For example, high levels of burnout have been reported among nurses in multiple countries (Aiken et al., 2012; Hatcher & Laschinger, 1996; McHugh, Kutney-Lee, Cimiotti, Sloane, & Aiken, 2011; Poghosyan, Clarke, Finlayson, & Aiken, 2010). Worker burnout, as a concept for research in the health professions, remains a cross-culturally relevant subject.

The Maslach Burnout Inventory is composed Likert-type items that assess three distinct components of burnout: emotional exhaustion, depersonalization, and personal accomplishment (Maslach et al., 2001). Three versions of the MBI exist, but the Maslach Burnout Inventory-Human Services Survey (MBI-HSS) has been designed exclusively for professionals whose work involves intensive human contact and interaction, such as nurses. Recent work continues to demonstrate that the MBI-HSS (Maslach et al., 2001) is the standard for gauging burnout among English-speaking healthcare professionals, especially nurses (Aiken, Clarke, Sloane, Sochalski, & Silber, 2002; Aiken et al., 2010; Cimiotti, Aiken, Sloane, & Wu, 2012; Hatcher & Laschinger, 1996; Laschinger, Grau, Finegan, & Wilk, 2010; Losa Iglesias, Becerro de Bengoa Vallejo, & Salvadores Fuentes, 2010; Patrician, Shang, & Lake, 2010; Poghosyan et al., 2010; Santen, Holt, Kemp, & Hemphill, 2010; Stimpfel, Sloane, & Aiken, 2012). Originally developed in English, the MBI-HSS has been used to assess burnout in nurses in several English-speaking countries, including the US (Aiken, Clarke, Sloane, & Sochalski, 2001; Aiken et al., 2002,2010), Canada (Aiken et al., 2001; Estabrooks et al., 2002; Hatcher & Laschinger, 1996; Laschinger et al., 2010; Leiter & Spence Laschinger, 2006; Spence Laschinger, & Leiter, 2006; Tourangeau et al., 2007), England (Sheward, Hunt, Hagen, Macleod, & Ball, 2005; Rafferty et al., 2007), Scotland (Aiken et al., 2001), and New Zealand (Poghosyan et al., 2010). The MBI-HSS has been translated into German (Aiken et al., 2001), Hebrew (Chayu & Kreitler, 2011), Japanese (Poghosyan et al., 2010), Turkish (Akkuş, Karacan, Göker, & Aksu, 2010; Günüşen &Ustün, 2010) and Chinese (Yao, Yao, Wang, Li, & Lan, 2013).

Despite these translated uses of the MBI-HSS, reports of systematic translations of this instrument are few. The cross-cultural adaptation of any instrument designed in one country for use in another requires a rigorous and systematic process (Squires et al., 2013). For researchers in countries seeking to study burnout for the first time, both language and translatability may present problems to utilizing the MBI-HSS instrument itself. While studies in the literature have translated the MBI-HSS, authors have done so using a variety of techniques and analytic approaches. The variation in validation methods for translated versions of the MBI-HSS presents several methodological issues for researchers. One methodological challenge for studying burnout arises with the concept of “burnout” including the language used to express or define burnout and its dimensions, such as emotional exhaustion. These descriptions might not exist culturally or linguistically, or the general concept of burnout may be culturally taboo. For example, if the cultural norm for dealing with or expressing symptoms related to burnout is to endure them silently, the concept of burnout might not exist in the language of the culture or may remain unidentified or have a different descriptor. Yet we do know that it is a psychological syndrome that is commonly identified in people and thus, an initial study may be needed to identify the presence of the phenomenon in a new context or culture through a standardized measure. Comparatively, in cultures where emotional expressions of burnout are expected, burnout, or an equivalent concept may already exist in the language.

A good example of the cross-cultural and language translation challenges of burnout can present is how “burnout” as a concept has translated into Mexican Spanish. Balseiro Almario translated burnout as “fatiga emocional” and “desgaste emocional” when introducing this concept into the Mexican health services research literature (Balseiro-Almario, 2004, 2005). When translated back into English, the phrases respectively translate as “emotional fatigue” and “emotional exhaustion, which only describes one dimension of burnout. Despite coining these translated phrases, Balseiro Almario used the English word “burnout” in her articles, either for its ease of pronunciation in Spanish or because an adequate phrase in Spanish could not be found to encompass all the dimensions of burnout. The following questions then arise: (1) Why only focus the translation on one dimension of burnout? (2) What aspect of that phrasing makes the concept culturally presentable? These kinds of considerations and challenges are common when translating and applying complex concepts, like burnout, in new cultural and language contexts.

In the case of the RN4CAST (www.rn4cast.eu) study (Sermeus et al., 2011), in order to evaluate the effects of work environments on nurses’ perceptions about their job related burdens, the study integrated the Maslach Burnout Inventory into a 12-country comparative study of nursing professionals in Europe. The initial review of the instrument by the RN4CAST research team raised concerns about the language used in the MBI-HSS and its translatability to other languages, cultures, countries, and contexts. It inspired the team to investigate the issue further.

The purpose this study was first to evaluate translated versions of the MBI-HSS that exist in the current literature to explore how other studies have conducted the cross-cultural adaptation of the MBI-HSS. The team then evaluated the MBI-HSS specific results of the translation process developed for the RN4CAST study instrument, which was grounded in Flaherty et al.’s (1988) guidelines for cross-cultural evaluation of survey instruments. The systematic approach to translating the survey instrument included traditional forward and back translation techniques (Brislin, 1970), expert panel reviews, and a quantifiable technique to evaluate the relevance of questions and the quality of the translation, as described in detail by Squires et al. (2013). Overall, we seek to illustrate the challenges that can arise when translating complex concepts and to identify threats to the reliability and validity of study results when the translation of survey instruments is not conducted systematically and without consideration for the cross-cultural relevance of the topic.

1.1. Background

The World Health Report 2006 – Working Together for Health (World Health Organization, 2006) is an assessment of the global healthcare workforce that estimates a shortage of 4.3 healthcare personnel per 100,000 people worldwide. The report further cites job-related burnout and its associated outcomes as factors that contribute to healthcare workers intentions to leave their jobs. Healthcare worker burnout appears to be an economic burden to national health systems and its care organizations that are struggling to maintain adequate staff numbers to provide health services to those in need (Alameddine, Baumann, Laporte, & Deber, 2012; El-Jardali et al., 2011; Pomaki, Franche, Murray, Khushrushahi, & Lampinen, 2012; Stansfeld, Shipley, Head, & Fuhrer, 2012). It is beneficial for healthcare systems to evaluate levels of burnout among their staff to determine its effects on retention rates, attrition rates from the profession, or its effects on international migration (Alameddine et al., 2012).

The MBI-HSS has been the chosen instrument of many researchers looking to gain better insight into the problem of burnout among healthcare professionals. While many studies have used the MBI-HSS to measure burnout across different cultures and national contexts, our review of the literature for this study showed that translation methods and study designs appear inconsistent. Researchers with extensive cross-cultural research experience generally agree that a systematic approach is necessary when translating a concept like burnout. A solid translation process will manage the emic and etic aspects of translation, include forward and back translation completed through the use of qualified translators, and an expert panel review (Cha, Kim, & Erlen, 2007; Hilton & Skrutkowski, 2002; Hyrkäs, Appelqvist-Schmidlechner, & Oksa, 2003; Im, Page, Lin, Tsai, & Cheng, 2004; Jones, Lee, Phillips, Zhang, & Jaceldo, 2001; Sidani, Guruge, Miranda, Ford-Gilboe, & Varcoe, 2010; Temple, 2005; Wang, Lee, & Fetzer, 2006; Weeks, Swerissen, & Belfrage, 2007). In addition, a pilot study of the translated version of the instrument is also recommended.

More concerning, however, is that health researchers tend to only use forward and back translation techniques. Maneesriwongul and Dixon (2004) showed, through a rigorous integrative review of studies that required translating survey instruments, that simple forward and backward translation is insufficient to produce a valid translation. Therefore, the results from studies that only employ forward and backward translation of survey instruments for their translation approach might not be very reliable or valid because of the quality of the translation or produce artificially high or low results.

In light of the concerns about translation quality, when examining the MBI-HSS prior to study implementation, the RN4CAST research team found that the survey items include questions that use slang words or phrases in American English. Unless researchers conducting burnout studies were familiar with the appropriate meaning and interpretation of the slang words, or they used an American English speaker to conduct the translation, the validity of the translation and the respective survey findings could be called into question. Who conducted the translation can be an important factor in evaluating the quality of the translation and its threat to reliability and validity (Squires, 2008,2009; Temple & Young, 2004; Temple, 2002).

An evaluation of the cross-cultural relevance of an established instrument, which includes translation quality, is an important step prior to its use outside of its original context of development. While it may not be possible to alter the instrument when consistent results across countries are part of a study’s aims, a pre-data collection evaluation step may help researchers to identify potentially problematic items in the instrument that are specific to the setting, language, or culture (Squires et al., 2013). At the same time, if a valid translation exists, duplicating work wastes time.

2. Methods and results

We explored the issue of cross-cultural adaptation of the MBI-HSS in two ways. First, we conducted an integrative literature review focused on translation methods used in research studies that had translated the MBI-HSS, and second, a Content Validity Indexing (CVI) process to evaluate the cross-cultural relevance of the translations conducted by the study’s country teams. We present the approaches and results for both parts of the study in two sections and then synthesize the findings in the discussion.

2.1. Approach to the integrative review of the literature

For the integrative review, the team used MEDLINE, CINAHL and Google Scholar, and searched for articles about the MBI that involved language translation. Search terms included “Maslach”, “burnout”, “international,” “language,” “interpretation,” and “translation”. Over 400 articles were identified in the search process. Selection criteria for articles included their availability in English, Spanish, or Portuguese, publication date after 2000.

Once the team finalized article selection, directed content analysis techniques, defined as “a research method for the subjective interpretation of the content of text data through the systematic classification process of coding and identifying themes or patterns” (Hsieh & Shannon, 2005), provided structure to the review process. To code the articles, the team focused on identifying the presence of cross-cultural research characteristics, as defined by Flaherty et al. (1988) and illustrated in Table 1, the translation method used, and when applicable, the type of statistical analysis used to determine the reliability and validity of the translated instrument. Detailed notes about specific methodological issues related to translation or how researchers interpreted the results were also made during the coding process.

Table 1

Definitions of Flaherty’s criteria for evaluating cross-cultural equivalence of survey instrument items.

Criteria	Definition
Content equivalence	The content of each item of the instrument is relevant to the phenomena of each culture being studied.
Semantic equivalence	The meaning of each item is the same in each culture after translation into the language and idiom (written or oral) of each culture.
Technical equivalence	The method of assessment is comparable in each culture with respect to the data that it yields.
Criterion equivalence	The interpretation of the measurement of the variable remains the same when compared with the norm for each culture studied.
Conceptual equivalence	The instrument is measuring the same theoretical construct in each culture.

Adapted from Flaherty et al. (1988), p. 258.

2.2. Results of the integrative review

In the review, 30 articles met the criteria for the analysis, representing 26 countries, 20 languages, and 8 of the RN4CAST countries. There were 7 languages with regional or geographical dialects represented, including Spanish (Spain, Argentina, Colombia), German (Germany and Switzerland), Italian (Italy and Switzerland), French (France and Switzerland), Dutch (Belgium and the Netherlands), Chinese (Cantonese, Mandarin, Macau dialect, and Taiwanese), and Arabic (Yemeni dialect).

Table 2 summarizes the findings, including the cross-cultural analysis against Flaherty et al.’s (1988) criteria. Within the 30 articles analyzed, there were vast differences in each approach to translate the MBI-HSS. Literature analyzed for this study offered no consistent findings about the cross-cultural relevance of burnout and showed that researchers used no consistent approach to translating the instrument, even when a translated version already existed in their own language.

Table 2

Integrative review of studies that translated the MBI-HSS, MBI-GS, & MBI-ES.a

Authors	Year	Country	Language	PT	Translation Method	SAM	CE1	SE	TE	CE2	CE3
Ahola, et al.	2006	Finland	Finnish	3+4	MBI-GS – no mention	Gender/ageadjusted logisticalregression	Y	N	Y	N	?
Al-Dubai and  Rumpal	2010	Yemen	Arabic	2	– MBI-HSS “translated intoArabic by a professionaltranslator. The translatedversion was compared with theEnglish version by theprincipal author to ensure itreflected original method”– Exploratory andconfirmatory factor analysis toconfirm accuracy	Descriptivestatistics, multiplelogistic regression	Y	Y	Y	N	?
Asai et al.	2007	Japan	Japanese	2	No mention, however,“psychometric properties ofthe Japanese version of the MBIare controversial”	Logical regressionanalysis	Y	N	Y	N	?
Berg et al.	2006	Norway	Norwegian	4	No mention of translation	Logical regressionanalysis	Y	N	Y	N	?
Bressi et al.	2009	Italy	Italian	2	States use of “Italian Version”	Linear regressionanalysis	Y	N	Y	N	?
Chen et al.	2013	Malaysia	Malay	3	MBI-HSS – forward/backwardtranslation. Multi-disciplinaryteam to reconcile theinstruments.	Cronbach’s alpha	Y	N	Y	N	?
Córdoba et al.	2011	Columbia	Spanish	1,2and3	MBI-HSS – forward/backwardtranslation. Judges evaluatedthe level of pertinence for eachitem	Descriptivestatistics andCronbach’s alpha	Y	N	Y	N	?
Embriaco et al.	2006	France	French	2	No mention of translation	Ordinal logisticalregression	Y	N	Y	N	?
Embriaco et al.	2007	France	French	3	No mention of translation	No mention	Y	N	Y	N	?
Glasberg et al.	2007	Sweden	Swedish	3+4	Validated Swedish translation	Regression analysis	Y	?	Y	N	?
Goehring et al.	2005	Switzerland	French,German,Italian	2	MBI-HSS – validated Germanand French versions. Theycreated their own version ofthe Italian.	Logistic regression	Y	Y	Y	N	?
Hu and  Schaufeli	2009	China	Chinese	4	MBI-ES – “translated fromEnglish into Chinese by threenative Chinese speakingmaster’s students…semanticdifferences were agreedupon…”	Confirmatoryfactor analysis	Y	?	Y	N	?
Iglesias et al.	2009	Spain	Spanish	1	“Both questionnaires havebeen validated internationallyand have been adapted for theSpanish population”	Descriptivestatistics	Y	?	Y	N	?
Juthberg et al.	2010	Sweden	Swedish	1	A Swedish translation, but notstated if it is validated.	Partial leastsquares regression	Y	?	Y	N	?
Kanste et al.	2006	Finland	Finnish	1	MBI-HS – “translation-backtranslation procedure”	Exploratory andconfirmatory factoranalysis	Y	?	Y	N	?
Klersy et al.	2007	Italy	Italian	1+2	“Burnout was assessed withthe validated Italian-languageversion of the Maslach BurnoutInventory”	Population-averagedregression models	Y	?	Y	N	?
Lee et al.	2012	Taiwan	Chinese	1	MBI-HSS-forward/backwardtranslation. Pilot study of theChinese version.	Exploratory andconfirmatory factoranalysis	Y	Y	Y	N	?
Liakopoulou  et al.	2007	Greece	Greek	3	No mention of translation	Descriptivestatistics	Y	N	Y	N	?
Luk et al.	2009	Macau	Chinese	4	C-MBI – stated used a Chineseversion, but no mention ofvalidity	ANOVA	Y	?	Y	N	?
Mészáros et al.	2013	Hungary	Hungarian	1,2,and3	MBI-HSS – forward/backwardtranslation. Pilot test	Confirmatoryfactor analysis	Y	Y	Y	N	?
Ndetei et al.	2008	Kenya	Swahili	3	MBI HS and GS - no mention oftranslation	SPSS	Y	?	Y	N	?
Pisanti et al.	2012	Italy	Italian	1	MBI-HSS - forward/backwardtranslation. States it is“substantially equivalent” toanother Italian version	Confirmatoryfactor analysis	Y	N	Y	N	?
Schaufeli et al.	2009	TheNetherlands	Dutch	2	They use the Dutch version ofthe MBI-HSS	Structural equationmodeling	Y	Y	Y	N	?
Soler et al.	2006	ManyEuropeancountries	Many	2	“In those countries where theuse of an English-languageinstrument could potentiallypose language barriers, thequestionnaire was translatedinto the native language by aFD”	Descriptivestatistics	Y	N	Y	N	?
Tokuda et al.	2009	Japan	Japanese	2	Reliable and valid Japaneseversion	Path analysis	Y	Y	Y	N	?
Unterbrink  et al.	2007	Germany	German	4	Use of the MBI-D which is theGerman version the MBI scale?	Descriptivestatistics	Y	N	Y	N	?
Van Bogaert  et al.	2009	Belgium	Dutch	1	A previously used translatedversion of the MBI-HSS, notmention of validity	Structural equationmodeling	Y	?	Y	N	?
van der Ploeg  et al.	2003	TheNetherlands	Dutch	4	They use the Dutch version ofMBI	Multiplehierarchicalregression analysis	Y	Y	Y	N	?
Waldman et al.	2009	Argentina	Spanish	2	“Employed the Spanishversion, which has shown tohave adequate reliability andvalidity in previous studies”	Multivariatelogistic regression	Y	?	Y	N	?
Wu et al.	2007	China	MandarinChinese	1	MBI-GS-translated fromEnglish to Chinese – then backfrom Chinese to English. TheChinese version was reviewedand was deemed to have highvalidity	Parametricstatistics	Y	?	Y	N	?

Note: PT (provider type): RN = 1; MD = 2; Multiple = 3; Other=4; SAM (statistical analysis method); CE1 (content equivalence); SE (semantic equivalence); TE (technical equivalence); CE2 (criterion equivalence); CE3 (conceptual equivalence); Y= Present; N = Not present; ? = Unable to determine.

Among the 30 reviewed articles, 7 (22%) articles did not mention any method of translation and 9 (28%) articles cited other validated versions of MBI-HSS in their respective language. The majority of the rest used only forward and backward translation. A few used a more thorough cross-cultural validation process, usually involving an expert panel review or a pilot study.

The 30 articles were further analyzed against five criteria suggested by Flaherty, et al. (Flaherty et al., 1988): content equivalence, semantic equivalence, technical equivalence, criterion equivalence, and conceptual equivalence. Every article met the criteria for content equivalence, which determined if the MBI-HSS was the appropriate tool for their study, and aimed to determine if burnout was present in the study population. Since the MBI-HSS was administered via questionnaire in all 30 articles, they met the technical equivalence requirement, which determined if the tool was being administered appropriately in the context.

Semantic equivalence was more difficult to determine. Semantic equivalence is defined by Flaherty et al. as “the meaning of each item is the same in each culture after translation into the language and idiom (written or oral) of each cultures” (Flaherty et al., 1988). A tool could be deemed semantically equivalent if it has been validated through a rigorous process. However, many of the articles did not mention how the translation was conducted nor indicated if they used a previously validated version of the MBI. Therefore, the team concluded that many of these articles did not meet the criteria for semantic equivalence. None of the articles met the definition for criterion equivalence, which evaluates if “the interpretation of the measurement of the variable remains the same when compared with the norm for each culture studied” (Flaherty et al., 1988). None of the articles met this criterion because it was not apparent if any of the studies met the norm for the culture being studied. Finally, it was nearly impossible to determine how authors of the studies addressed conceptual equivalence, which would measure “the same theoretical construct in each culture” (Flaherty et al., 1988).

2.3. RN4CAST translation approach and cross-cultural evaluation

For this paper, we examined the study’s pre-data collection evaluations of the cross-cultural relevance of the translated versions of the MBI-HSS survey questions produced by the RN4CAST country team members. This included eleven European countries with ten languages among them. They included Belgium (Dutch and French), Germany (German), Finland (Finnish), Greece (Greek), Ireland (Irish-English), the Netherlands (Dutch), Poland (Polish), Spain (Spanish), Sweden (Swedish), Switzerland (Swiss French, Swiss German, Swiss Italian), and England (British English). For baseline comparison, data from the United States (American English) were also included. At the time the study began, a valid Spanish translation approved by the company that owns the copyright of the MBI-HSS was available so Spain was the only country that used an official translation, and they did not participate in this pre-data collection evaluation exercise.

Prior to beginning the study, the RN4CAST team concluded that it could not use any of the existing versions of the translated instrument. Therefore, with the permission of the company that manages the MBI-HSS, the team conducted its own approach to translating and evaluating the cross-cultural relevance of the instrument. The translation process involved multiple steps and met the requirements for rigorous cross-cultural translation of established instruments, including an evaluation of content, context, conceptual, semantic, and technical equivalence. It involved two phases of reviews that included nurses with extensive experience in their field, the use of experienced translators who used a translation guide to facilitate translation, and a quantification of relevance of survey items to nursing in the country. By evaluating the “relevance” of survey items, we mean that the “expert” nurses reviewing the survey item determine if the question is appropriate for use with the population to be studied and in the case of this study, in the cultural context(s) of the country where is would be applied. We chose nurses as our study experts rather than occupational psychologists or psychiatrists to ensure that the translated instrument was validated by those intimately involved with the population who would be surveyed with the MBI-HSS. Their content and contextual knowledge were essential components of the expertise we sought in “expert” raters. Complete details of the approach to translation are found in Squires et al. (2013).

In order to produce the quantifiable measure of cross-cultural relevance and translation quality, the research team opted to use an approach normally advocated for initial instrument development: Content Validity Indexing (CVI) with corrections for chance agreement. During the CVI process, expert raters provide scored feedback to determine if the question or statement, e.g., in a survey instrument, is relevant to the population being studied and if the format of the question is appropriate (Polit, Beck, & Owen, 2007). This assessment of a question’s “relevance” allows an expert to determine the content, context, and conceptual equivalence of a translated survey question through a single rating. The experts used the following standard CVI rating scale to evaluate a question’s relevance: 1 = Not relevant; 2 = Somewhat relevant; 3 = Very relevant; 4 = Highly relevant. A calculation of chance correction, via a modified kappa score, adjusts the CVI score to indicate agreement over and above chance, thereby increasing the rigor of the CVI process (Polit et al, 2007). Only scores of 3 or 4 are included in the calculation process.

The CVI cannot evaluate the semantic and technical equivalence of a translated question in its rating process. Therefore, the team added to the CVI process a simple “yes” or “no” rating to evaluate the semantic and technical equivalence of the translations. It was the team’s intention to keep this aspect of the rating process as simple as possible to bolster confidence in the cross-cultural translation process. Comments sections for raters allow for recommended corrections to translations, as needed.

Both scoring systems use the same formula for calculating item-level and scale-level scores with modified kappa statistics for comparison. For the CVI process, I-CVI represents the item level score where the minimum acceptable result is 0.78 or higher (without correction for chance agreement among raters) to include an item in a survey. The modified kappa score equivalent to the 0.78 CVI score is 0.74, but offers a comparison of scores that would be considered “good” (0.60–0.73) vs. “excellent” (0.74 or higher) (Polit et al., 2007). A modified kappa score of 0.59 or lower would mean the item was not acceptable for inclusion in the survey or that the content or phrasing of the question might need modification (Squires et al., 2013). The scale level CVI score (S-CVI) is an average of all ratings for all items in the instrument.

The team then had to agree on how to manage low scores from raters on individual items. Since the MBI-HSS is an established instrument, the team could not discard any items from the survey. Therefore, the team agreed to set minimum levels for scores in order to “approve” the translation for use in the larger study. Survey questions that received low scores from the raters, a modified kappa score of 0.59 or lower, were identified as potentially problematic items. Each team would note the potentially problematic items and then work to aggregate the pre-data collection data to modify the translation when needed. In the larger study, results keyed to potentially problematic items receive closer scrutiny for trends that might suggest an issue with cross-cultural equivalence.

A total of 120 raters from the twelve countries involved were invited to participate. Expert raters from each country completed the evaluation process through an online survey. Raters were asked to report their educational level and current role: Practitioner (meaning working in a front line patient care role), educator (involved in health education in a hospital or educational institution), or administrator (hospital or other organization). Upon completion of the rating process, the survey company emailed the results to the project manager in a spreadsheet format. Preprogrammed formulas in Microsoft Excel calculated the item and survey level data using the CVI with chance correction method described by Polit et al. (2007).

2.4. Results of the RN4CAST cross-cultural CVI analysis

A total of 106 raters out of 120 participated in the CVI with the translation evaluation process and their basic demographics are found in Table 3. With one exception, Germany (n = 5), each country had 7 or more raters with a maximum of 11. Germany was only able to obtain five raters to participate in the process because there was a significant lack of bilingual German-English speaking nurses in the country. Complete results of the MBI-HSS CVI rating process with chance correction are found in Table 4, illustrating the S-CVI scores. Generally, the scale-level MBI-HSS scores varied widely and ranged from 0.49 (US and Belgian French) to 0.93 (Germany). Out of the ten languages and their country variants involved in the validation process, half received acceptable scale level scores, while half did not. Fig. 1 illustrates 5 out of 7 items on the MBI-HSS that received extremely low scores by 60% or more of participating countries and these items were identified as likely to produce “problematic” responses in the larger RN4CAST study.

Table 3

Expert rater demographicsa (n = 106).

	% of Respondents
Education
Bachelors	63
Masters	37
Role
Practitioner	50
Educator	29
Administrator	21

Gender identity of raters was not formally collected for this exercise. Almost all of the raters were female, reflective of the gender dominance within the nursing profession globally.

Table 4

Scale level modified kappa scores by country for the MBI-HSS.

Country	k Score
Belgium (Dutch)	0.84
Belgium (French)	0.49
Finland	0.84
Germany	0.91
Greece	0.93
Ireland	0.51
Netherlands	0.77
Poland	0.6
Sweden	0.81
Switzerland (French)	0.55
Switzerland (German)	0.68
Switzerland (Italian)	0.88
United Kingdom	0.57
United States	0.49
Study Average	0.70

Fig. 1

Potentially problematic items, with average modified k of 0.60 or lower.

Table 5 illustrates the problematic item rating trends across the languages and regions involved in the study using 5 items from Fig. 1. Similar languages are placed side-by-side in the table for comparison. While the sample size limited our ability to test for significant differences between the groups, some trends are evident and are worthy of analysis. To begin, Greece and the Italian region of Switzerland had very high relevance scores with all items on the MBI-HSS, suggesting that the MBI-HSS could be a useful tool in assessing burnout for nurses or that expressing burnout is culturally acceptable in these regions. Some countries that spoke the same language (such as Belgian French and Swiss French) gave very similar individual question scores. Swiss German and German, on the contrary, were exactly opposite in their ratings of all questions. Belgian Dutch and the Netherlands’ Dutch were mixed.

Table 5

Potentially problematic items – a comparison between similar languages.

Country (language)	Belgium(Dutch)		Netherlands(Dutch)		Belgium(French)		Switzerland(French)		Germany(German)		Switzerland(German)		Ireland(English)		UK(English)		US(English)
Problem itemQuestion	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No
..............impersonal objects...................		✓	✓		✓		✓			✓	✓		✓		✓		✓
....................a strain.............................		✓	✓		✓		✓			✓	✓		✓		✓		✓
..................hardening me..............	✓		✓		✓		✓			✓	✓		✓			✓	✓
................................don’t really care.....................		✓		✓	✓		✓			✓	✓		✓		✓		✓
................................stress......................		✓		✓	✓		✓			✓	✓		✓		✓		✓
.............................end of my rope......................		✓		✓	✓		✓			✓	✓		✓		✓		✓
.................................blame me...............	✓		✓		✓		✓			✓	✓		✓		✓		✓

Core meaning for factor tables used with permission of the publisher, Mind Garden, Inc. www.mindgarden.com. MBI-Human Services Survey: Copyright ©1981 Christina Maslach & Susan E. Jackson. All rights reserved in all media.

3. Discussion

The notable lack of attention paid to translation processes in the studies evaluated for the integrative review and the variability of the results and our own subsequent evaluations suggest that researchers and journal editors should pay more attention to how authors have conducted translation processes for the MBI-HSS. Statistical analyses of survey results to analyze their validity will not capture problems related to translation, which occurs before data collection. Our findings also suggest that outliers in study results may be the result of an issue with the translation process. Factor loading on factor analyses may also deviate and affect subscale composition and thus, the consistency of the scale across contexts and cultures.

Some I-CVI scores on the MBI-HSS were lower than anticipated. A possible explanation for the lower MBI-HSS scores, in the case of this paper, concerns whether or not the concept of or language for “burnout” is actually present among the nursing workforce vernacular in a country. Nurses may very well possess the signs and symptoms of burnout, but they may not yet have a name for what they are experiencing and may only report high levels of emotional exhaustion. For some cultures, even the idea of feeling “burnt out” in a job may not align with their cultural norms and values; therefore, nurses and other healthcare workers might ignore or suppress symptoms of burnout.

The remainder of the discussion proceeds on a country-by-country or language “case” basis. To begin, several explanations might shed light on the results from the English-languages versions rating process. The first concerns the use of American English slang in the MBI-HSS. Comments from the Irish and English raters suggested that while they understood the intent of the question, the language describing the concept was not expressed in a way that was common in their home country. This underscores the need to cross-culturally validate even English language instruments when used in another country.

English speaking countries also scored the MBI-HSS relatively low on the relevance of “burnout” to nursing practice in their home countries. This was a surprising finding for the team. One explanation for these scores included expert sampling bias, where “experts” worked in organizations with supportive organizational cultures for nursing practice. Therefore, they did not perceive the “burnout” questions as relevant to practice compared to a nurse who might work in a less supportive organization. System-wide reforms in both English speaking countries might also contribute to changed perceptions about the relevance of burnout. Language and phrasing of the items may also have differed sufficiently for the experts to score the items as less relevant. The number of internationally educated nurses in both the UK and Ireland results in a variety of “other Englishes” as well, thereby presenting additional challenges of harmonizing English language phrasing between healthcare workers from different countries that speak English. Future research may need to better account for issues of rater identity during this process.

In the case of Poland, many questions had modified kappa scores below 0.60. Consequently, the RN4CAST team recommended that the Polish team pilot test the entire instrument before using it in the larger study (Brzostek et al., in press). The relative “newness” of nursing research in Poland may also explain some of the low CVI scores from that country. The Polish case highlights the importance of a rigorous cross-cultural adaptation process when using the MBI-HSS in a new country.

The breadth of scores in Switzerland is puzzling to the outside researcher. The Swiss team, however, was not surprised by the variation among the scores between the three linguistic regions of the country as they commonly find variations in research results when conducting national workforce studies. The Swiss team, therefore, made note of the potentially problematic items from their rating process (beyond the 7 identified by the entire study). They concluded that if outlier responses in the Swiss burnout results were noted a problem with the contextual applicability of the item might be indicated. The Belgian team adopted a similar tactic with the Belgian French version of the MBI-HSS to address its low score. The findings from both countries suggest that the linguistic expression of burnout has a cultural component that researchers need to anticipate in comparative national studies of healthcare workers and within country comparisons when more than one official language exists. Additional analyses to explore this phenomenon are planned.

3.1. Limitations

With twelve countries and ten languages, limitations for this study relate mostly to the individuals conducting the CVI rating process. Germany, notably, had only five raters, which skewed its results toward higher scores even though comments provided by raters suggested some issues related to the translation of terms found in the seven problematic items. It is possible that with more raters their results may have looked more like the Swiss-German version. Additionally, language abilities and educational levels of the raters varied across the countries and were not always formally verifiable through English language test scores or other measures of language competence. This likely affected scoring processes due to widely varying bilingual capabilities among nurses. The use of only raters with nursing experience may have also biased the results. We recommend that researchers seeking to replicate these methods find a way to account for the linguistic capabilities of their expert raters. For the integrative review, due to our linguistic limitations for the search results and the possibility that a national journal (which might contain a good translation of the MBI-HSS) was not listed in the search databases, we might not have been able to capture all of the available studies.

3.2. Conclusions

To the best of our knowledge this is the first time a mixed-methods evaluation of the translatability of the MBI-HSS occurred through a structured research approach. The RN4CAST team plans to conduct additional analyses confirming the validity and reliability of the instrument across the different national contexts. Results of this study have lead us to encourage individual country-level validation processes for research teams seeking to use translated instruments for research projects, even if the country is linguistically similar the instrument’s country of origin.

We also note that as health services research spreads across the globe, researchers using survey instruments in their studies should be cautioned to carefully and systematically translate and evaluate the cross cultural relevance of the instrument prior to data collection. Our literature reviews shows that this is often not done. Most survey instruments were developed for use in the country of origin of the researcher who created it. There is no guarantee that the concepts measured by the instrument will apply in the same way in another country and language context, especially when health systems are organized in different ways. Studies that fail to pre-evaluate the cross-cultural applicability of a survey instrument may not produce reliable and valid results. Using systematic approaches can help to reduce the risk for those kinds of errors.