Authors: Rodnell Busigó Torres (aLeni and Peter W. May Department of Orthopaedic Surgery, Icahn School of Medicine at Mount Sinai, New York, NY USA), Jennifer Yu (aLeni and Peter W. May Department of Orthopaedic Surgery, Icahn School of Medicine at Mount Sinai, New York, NY USA), Brett L. Hayden (aLeni and Peter W. May Department of Orthopaedic Surgery, Icahn School of Medicine at Mount Sinai, New York, NY USA), Lauren M. Shapiro (bDepartment of Orthopaedics, University of California San Francisco, 1500 Owens Street, San Francisco, CA USA), Brocha Z. Stern (aLeni and Peter W. May Department of Orthopaedic Surgery, Icahn School of Medicine at Mount Sinai, New York, NY USA; cInstitute for Health Care Delivery Science, Mount Sinai Health System, New York, NY USA; dDepartment of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY USA)
Categories: Article, Outcome Assessment, Health Care, Psychometrics, Limited English Proficiency, Total Hip Arthroplasty, Total Knee Arthroplasty, Osteoarthritis, Measurement Invariance
Source: The Journal of arthroplasty
Authors: Rodnell Busigó Torres, Jennifer Yu, Brett L. Hayden, Lauren M. Shapiro, Brocha Z. Stern
Versions of the Hip Disability/Knee Injury and Osteoarthritis Outcome Score (HOOS/KOOS) measures are widely used. However, inadequate translation and validation can lead to inaccurate information for non-English-speaking patients, exacerbating disparities. This systematic review examined the translation quality and measurement properties of non-English HOOS and KOOS versions.
We systematically reviewed peer-reviewed articles on linguistic or cultural adaptation and/or psychometric validation of HOOS or KOOS versions in adult knee or hip patients, focusing on the top five non-English languages spoken in the U.S. (Spanish, Chinese, Tagalog, Vietnamese, and Arabic). Translation quality and measurement properties were evaluated using Consensus-based Standards for the Selection of Health Measurement Instruments tools.
There were 18 articles HOOS (one Chinese), KOOS (two Arabic, seven Chinese, two Spanish, and one Tagalog), KOOS-12 (one Spanish), KOOS for Joint Replacement (JR; one Spanish), and KOOS-Patellofemoral (PF; two Arabic and one Spanish). Of 14 studies describing translation, six received a rating of at least adequate on ≥ 80% of translation criteria. There were nine studies that specifically described cultural adaptation, but none quantitatively assessed cross-cultural validity (i.e., measurement invariance) across language or culture. Only the Spanish KOOS-12 had high-quality evidence of sufficient structural validity and internal consistency. There was very low-quality to moderate-quality evidence of sufficient reliability for most measures and limited information about measurement error. There was sufficient construct validity for Spanish KOOS JR (high-quality evidence) and Spanish KOOS-PF (low-quality evidence), but moderate-quality to high-quality evidence of insufficient construct validity for several measures. There was sufficient responsiveness for most subscales of the Chinese KOOS (high-quality evidence), Spanish KOOS and KOOS-12 (moderate-quality evidence), and Chinese HOOS and Spanish KOOS-PF (low-quality evidence).
Limitations were found in the quality of available HOOS and KOOS translations. Until high-quality, validated translations are available, cautious use is necessary to prevent biased estimates of health status, which can lead to inappropriate treatment decisions and ultimately exacerbate healthcare disparities.
Patient-reported outcome measures (PROMs) play a critical role in decision-making and measuring and improving care quality [1, 2]. The Hip Disability and Osteoarthritis Outcome Score (HOOS) and Knee Injury and Osteoarthritis Outcome Score (KOOS) are multi-subscale PROMs that assess pain, other symptoms, activities of daily living, sports and recreation, and quality of life in individuals who have hip and knee conditions [3, 4]. These PROMs are commonly used for decision-making and treatment monitoring in individuals who have osteoarthritis, including those undergoing total hip or knee arthroplasty (THA or TKA) [4–6]. Multiple versions are available, including the HOOS/KOOS-12 [7, 8]; HOOS/KOOS-Physical Function Shortform [9, 10]; HOOS/KOOS for Joint Replacement (JR) [11, 12]; and KOOS-Patellofemoral subscale (PF) [13]. The HOOS/KOOS JR is included in a mandatory THA/TKA quality measure from the U.S. Centers for Medicare and Medicaid Services (CMS) [14].
Given the expansion of PROMs in outcomes research, quality measurement, and clinical care, equitable integration requires validated measures that are linguistically and culturally adapted [15–17]. Lack of high-quality non-English assessments can lead to missing, unreliable, or inaccurate data, potentially worsening disparities. Linguistic translation alone may be insufficient for measurement equivalence, with one study finding that patients who had comparable physical function responded differently to 44% of items on a PROM administered in Spanish versus English [18]. Specifically, colloquial and idiomatic descriptions (e.g., burning pain), non-standard units of measurement (e.g., walking a block), and country-specific or culture-specific activities (e.g., playing golf) can limit comprehensibility and relevance [19, 20]. Attention to language and culture is particularly relevant in the U.S., where approximately 20% of the population speaks a language other than English at home and approximately 25 million individuals have limited English proficiency [21]. Therefore, regulatory organizations have emphasized the need for translated, culturally adapted, and rigorously validated PROMs [22]. However, limitations in translation and validation have been identified for other orthopaedic PROMs [23, 24]. To guide equitable orthopaedic care, more information is needed on the quality of translated HOOS/KOOS versions.
To address this gap, this review aimed to synthesize research on the translation and cross-cultural adaptation of HOOS/KOOS versions for the five most commonly spoken non-English languages in the U.S. (Spanish, Chinese, Tagalog, Vietnamese, and Arabic), which together comprise about 75% of the population who speaks a non-English language at home [21]. Specifically, we examined the quality of translation and cross-cultural adaptation and the measurement properties (structural validity, internal consistency, cross-cultural validity, reliability, measurement error, construct validity, and responsiveness).
We followed guidance on conducting a systematic review of measurement properties from the Consensus-based Standards for the Selection of Health Measurement Instruments (COSMIN) [25–27] and registered the protocol on the International Prospective Register of Systematic Reviews (PROSPERO CRD42024544913). The review is reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [28]. This review was not considered human subjects research, and institutional review board approval was not sought.
The search was conducted in English on June 2, 2024, across PubMed, Ovid Embase, CINAHL, and Scopus. To identify articles that may have been missed in the databases, we also reviewed reference lists of reviews and eligible articles and Mapi Research Trust bibliographies of HOOS/KOOS translation studies [29–35]. The search strategy combined terms for (a) measure names, (b) translation, cross-cultural adaptation, and measurement properties [36], and (c) specific languages (Supplementary Material A). We included peer-reviewed articles that focused on linguistic or cultural adaptation and/or psychometric validation of any HOOS/KOOS version in Spanish, Chinese, Tagalog, Vietnamese, or Arabic in adults aged 18+ who had hip or knee conditions. We excluded conference abstracts, commentaries, reviews, letters to the editor that did not report original investigations, and study protocols. There were no restrictions on publication language or date. The retrieved articles were deduplicated in EndNote and then in Covidence. There were two independent reviewers (RBT/JY) who screened the titles and abstracts, followed by a full-text review using the above inclusion and exclusion criteria. All disagreements in screening or review were resolved by a third reviewer (BZS).
Information from eligible articles was extracted into a standardized data collection form. Elements included bibliographic information (author, publication year, journal), research aim, PROM version, language of translation, participant information (country/region, diagnosis/procedure, age, sex/gender), and analytic methods and results for measurement properties. Extractions were completed independently by two reviewers (JY/BZS), and all discrepancies were resolved via discussion.
For studies reporting translation, methodological quality was assessed using the COSMIN Translation Process Design Checklist [37]. It has 12 criteria (e.g., forward and back translation), with each criterion rated as “very good,” “adequate,” “doubtful,” or “inadequate” [37]. Since culture is not specifically mentioned in this checklist, we added a non-COSMIN criterion related to cultural equivalence, which extends beyond linguistic equivalence [38]. To fulfill this criterion, studies had to specifically describe attention to the cultural relevance of activities or concepts [39]. Translation quality was independently assessed by two reviewers (RBT/JY), and all discrepancies were resolved via discussion with a third reviewer (BZS). Since our study’s aim focused on linguistic and cultural adaptation, we did not evaluate content validity beyond examining whether the researchers completed a pilot study during translation.
Next, we applied the COSMIN Risk of Bias Checklists (version 3) and COSMIN Criteria for Good Measurement Properties (version 2) for each measurement property evaluated in each article [25]. These appraisal tools were applied independently by two reviewers (RBT/BZS), and all discrepancies were resolved through discussion. We examined structural validity (construct dimensionality), internal consistency (item interrelatedness), and cross-cultural validity (quantitative measurement invariance specifically using language or culture as the grouping variable) [40]. We also examined reliability (attribution of variance to true between-individual differences) and measurement error (presence of systematic and random error) [40]. Also, we evaluated construct validity (consistency of scores with hypothesized relationships with other PROM scores or known group differences) and responsiveness (ability to detect changes in the measured construct over time, per comparison with another instrument or before/after intervention) [40]. For risk of bias, each criterion was rated as “very good,” “adequate,” “doubtful,” or “inadequate,” and we applied the “worst score counts” principle (i.e., recording the lowest rating for each criterion) within each study [25, 41]. Each measurement property was rated as “sufficient,” “insufficient,” or “indeterminate,” with a separate rating for each HOOS/KOOS subscale when applicable, except for structural validity. For construct validity and responsiveness, we standardized evaluation across studies by creating our own hypotheses against which we assessed the study findings per COSMIN guidance [25]. Hypotheses were generated for construct validity based on expected associations with similar measures or score differences between known groups and were generated for responsiveness based on expected associations of scores with those on similar measures or change after an intervention (Tables S1 and S2 in Supplementary Material B). Next, we synthesized the measurement properties across all studies of the same measure version and language (e.g., Chinese KOOS, Arabic KOOS-PF). For construct validity and responsiveness, our synthesis examined whether 75% of the hypotheses were met across the studies [25]. For the other properties, we qualitatively synthesized across multiple studies by taking the majority value when consistent for ≥ 75% of results or noting the summarized rating as inconclusive [25]. Also, we rated the quality of evidence as “high,” “moderate,” “low,” or “very low” for sufficient, insufficient, and inconsistent ratings of measurement properties based on the number of studies and their risk of bias [25] (Table S3 in Supplementary Material B).
Of 400 records initially identified, 18 articles were included (Figure 1). These assessed HOOS (one Chinese), KOOS (two Arabic, seven Chinese, two Spanish, and one Tagalog), KOOS-12 (one Spanish), KOOS JR (one Spanish), and KOOS-PF (two Arabic and one Spanish). There were 15 articles aimed at translating or adapting and validating the measure, while three validated an existing translation (Table 1).
Of the 14 articles reporting translation, ≥ 80% received an adequate or very good rating on key criteria of language specification, forward and backward translation, forward translator expertise, backward translator naivete, independent translations, reconciliation of differences, and expert review (Table 2). The two most poorly rated criteria were writing a feedback report and conducting a pilot study. There were six articles (two Arabic KOOS, two Chinese KOOS, one Spanish KOOS JR, and one Spanish KOOS-PF) that had a very good or adequate rating on ≥ 80% of all translation criteria. There were nine articles that described attention to cultural adaptation beyond linguistic equivalence, such as specifying prayer positions or making terminology inclusive of multiple toilet types. There was no information on Spanish KOOS-12 translation quality.
All articles assessed internal consistency, but only six assessed structural validity, and none assessed measurement invariance specific to language or culture (Table 3). There was high-quality evidence of sufficient structural validity and internal consistency for Spanish KOOS-12. Additionally, there was low- to very low-quality evidence of insufficient structural validity for Arabic KOOS and Arabic KOOS-PF, such that the number of statistically identified factors did not align with the number of subscales in that KOOS version. Because evidence was lacking for structural validity, there was indeterminate information for internal consistency for all measures except Spanish KOOS-12 (Table 4).
All articles assessed reliability, but only four had at least an adequate quality rating (one Chinese KOOS, one Spanish KOOS, Spanish KOOS-12, and one Arabic KOOS-PF; Table 3). There was evidence for sufficient reliability across all measures except for the Spanish KOOS-12 Pain and Function subscales, but the quality of evidence was often low to very low (Table 5). Studies of measurement error were sparse, with only one having at least an adequate quality rating (Table 3). There was indeterminate information on measurement error for Chinese HOOS, Arabic KOOS, Spanish KOOS-12, and Arabic KOOS-PF because of limited information on the minimal clinically important difference. There was very low-quality evidence of sufficient measurement error for Spanish KOOS-PF and moderate-quality evidence of negative or inconsistent findings for Chinese KOOS (Table 5).
All articles assessed construct validity, but several had inadequate or doubtful quality ratings (Table 3). Information for construct validity was indeterminate for Tagalog KOOS, given reporting weaknesses. There was sufficient construct validity for Spanish KOOS JR (high-quality evidence) and Spanish KOOS-PF (low-quality evidence), but there was moderate- to high-quality evidence of insufficient construct validity for Arabic KOOS-PF and several subscales of other versions (Table 6). There were eight articles that assessed responsiveness, with only three studies with at least an adequate quality rating (two Chinese KOOS and Spanish KOOS-12; Table 3). There was sufficient responsiveness for most subscales of the Chinese KOOS (high-quality evidence), Spanish KOOS and KOOS-12 (moderate-quality evidence), and Chinese HOOS and Spanish KOOS-PF (low-quality evidence; Table 6).
This systematic review identified gaps in available non-English translations of HOOS and KOOS versions with published psychometric properties, including the absence of Vietnamese measures and only one translation of any HOOS version. Additionally, no measures had evidence of sufficient ratings across all measurement properties. Furthermore, some translations lacked rigorous cross-cultural adaptation, and no studies assessed measurement invariance, limiting the potential accuracy and applicability of the PROMs in diverse populations.
There was variability in the quality of translation and cultural adaptation, with limited specification of cultural adaptation processes (e.g., expert review, cognitive interviewing) in many articles. While the poorly rated translation criterion of writing a feedback report may not actually reflect poor translation quality, the other poorly rated criterion of conducting a pilot study is vital to ensure content validity of the translated PROM, including clinical and cultural relevance and comprehensibility. Other reviews synthesizing the translation quality of Spanish PROMs in other orthopaedic populations have also identified inconsistent use of pilot studies [23, 24]. Additionally, all the studies in this review were conducted outside of the U.S., and such translations may not be fully relevant for those in the U.S. who speak that language. Specifically, for the CMS-mandated HOOS/KOOS JR measures [14], there were no published HOOS JR translations that were identified for this review, and the only Spanish KOOS JR translation was developed in Chile [56]. Of note, after completing the synthesis for this review, we identified publication of a pilot evaluation of translated and culturally-adapted versions of the HOOS JR, HOOS-12, KOOS JR, and KOOS-12 in Colombian Spanish [60]. The English version of this Spanish-language article was not available when the search for this investigation was run and thus was not captured in this review. While the authors of that pilot study describe a detailed translation process, they report the need for a formal validation study before using the measures, as the only measurement property assessed in the pilot was construct validity [60]. Generally, speaking a non-English language is associated with decreased response to PROMs [61, 62]. Since hospitals will be financially penalized for reporting less than 50% of paired pre- and postoperative HOOS/KOOS JR scores [14], the limited availability of translated measures may lead to disproportionate penalties in hospitals that treat larger proportions of non-English-speaking patients. Therefore, future research should prioritize translations in samples of non-English-speaking individuals in the U.S., with attention to factors such as cultural expressions and relevance of activities.
Regarding measurement properties, very few studies assessed structural validity. Despite the widespread use of these measures, structural validity should not be assumed, as recent studies have identified limitations for English-language HOOS/KOOS versions [63–65]. Additionally, no studies examined cross-cultural validity specific to language or culture, which is essential to identify if score differences are due to health status versus item interpretation based on language or culture. A lack of cross-cultural validity (i.e., where scores vary based on language or culture versus the degree of function or symptom) has been identified for various PROMs [18, 66, 67]. Not accounting for language- or culture-based score differences can bias the use of scores from diverse populations in decision-making, including in decision aids based on PROM scores [68, 69]. Limited cross-cultural validity testing has also been noted in other reviews of orthopaedic PROM translations [23, 24]. For future research, large multilingual samples are needed to assess measurement invariance based on language and culture. Such studies may be increasingly feasible with the expansion of PROMs in international registries [70].
While there was overall sufficient evidence of reliability across all measures except for the Spanish KOOS-12, there was no high-quality evidence for this measurement property. Additionally, information on measurement error was limited, primarily secondary to missing information on meaningful change. Although COSMIN guidance for good measurement properties allows for comparison with meaningful change estimates from another study in a similar population, meaningful change and normative thresholds may vary by language or culture [71–73]. Additionally, for construct validity, some hypotheses were not met because of associations that were too large (e.g., between pain and physical function), which may reflect limitations in structural validity and overlapping subscales. Furthermore, there was no information on responsiveness for the Spanish KOOS JR, which is important for TKA quality measurement. Future research priorities include understanding variation in meaningful change based on language and/or culture to improve score interpretation and identifying the responsiveness of measures to common interventions, including THA or TKA.
Several potential limitations should be acknowledged in this systematic review. Our synthesis was limited to published peer-reviewed articles to examine evidence of translation and validation, and translations available in the gray literature or unpublished sources were not considered. For example, Mapi Research Trust reports available translations in Vietnamese that were not included in this review, as they have not been published in the peer-reviewed literature [31, 34]. Our search strategy only included English-language terms, and we may have missed translation studies in other languages. Additionally, this review focused exclusively on synthesizing the measurement properties within specific language versions and is not intended to fully reflect the measurement properties of the PROMs across all translations. Regarding construct validity and responsiveness, the hypotheses were developed by the team following COSMIN guidance [25], and subjectivity was involved in this evaluation process. Furthermore, the “worst score counts” methodology, while standard practice [41], may have led to a more stringent risk of bias evaluation. Additionally, some studies did not provide detailed methodological reporting, and we did not contact study authors to clarify any missing information. As a result, some aspects of translation or cross-cultural adaptation processes may not have been fully captured. Furthermore, heterogeneity of study populations, including diagnoses, interventions, and timing of measurement, challenged synthesis across studies and comparisons across languages.
This review identified gaps in the availability and quality of HOOS and KOOS translations in the peer-reviewed literature for the five most common non-English languages in the U.S. Additionally, there were limitations in both linguistic and cultural adaptation processes, and no measures had evidence of sufficient ratings across all measurement properties. Clinicians and researchers should be cautious when using the currently available non-English HOOS/KOOS versions since biased estimates of health status can contribute to inappropriate treatment decisions and exacerbate health disparities. When high-quality, validated translations are unavailable, PROMs should be used as adjuncts to other assessment methods, such as clinician interviews. Given the growing cultural and linguistic diversity of the U.S. population, there is a critical need for high-quality translation, cross-cultural adaptation, and validation studies of PROMs to ensure accurate, equitable, patient-centered outcome measurement.