The study was conducted at the ED, Universiti Kebangsaan Malaysia Medical Centre (UKMMC) in Kuala Lumpur, Malaysia, between June 2020 and December 2020. The study was conducted in three phases. Phase one includes the translation and cultural adaptation of the English version of the KCL. In phase two, content validation and pilot testing were performed. In phase three, we performed field testing to determine the reliability and factor structure of the Malay-translated KCL (Fig. 1). The study received ethical approval from the UKMMC research ethic committee (FF-2020-349).
Phase one: translation process
The permission to translate the English version of KCL to the Malay language was obtained from the original publisher. Two independent certified professional translators performed a forward translation of the English version of KCL to the Malay language; it was followed by backward translation by two different independent translators. The consensus was made on the translation of words, phrases, and items of the Malay version between researchers and translators; the Malay version was then compared with the original KCL. After some slight revisions, the final version of the Malay-translated KCL was prepared and ready to undergo the validation process.
Phase two: content validation and pilot testing
The content validity of the tool was conducted based on the quantitative methods described by Lynn (18). Three expert panels, which consist of a geriatrician, emergency physician, and family medicine specialist with an area of interest in geriatric medicine, were invited to review the Malay-translated KCL. The panels rated each item according to a four-point rating scale: 1 = not relevant, 2 = item needs major revision, 3 = relevant but needs minor revision, and 4 = very relevant; comments that carefully assessed and reworded.
As noted by Lynn (18), the content validity index (CVI) score was estimated both at the item level (I-CVI) and scale level (S-CVI). The I-CVI was calculated by dividing the number of experts that rated an acceptable grade (rating 3 or 4) by the total number of expert panels. The S-CVI was calculated based on two different methods: the average method (S-CVI/Ave) and the universal agreement method (S-CVI/UA). The S-CVI/Ave is the average of the I-CVI score for all items on the scale. Meanwhile, the S-CVI/UA is the proportion of the relevant items on the scale (rating 3 or 4) by all experts. Based on standard recommendations for content analysis, the item with I-CVI > 0.79 and scale with S-CVI ≥ 0.9 was considered to have excellent content validity (19). Additionally, a modified Kappa statistic (K) was computed to adjust I-CVI for chance agreement using the formula K = (CVI-Pc)/(1-Pc) (20). The probability of chance occurrence (Pc) was computed using the formula Pc = [N!/A!(N-A)!]0.5N, where N is the number of experts, A is the number agreeing on good relevance, and ! is a mathematical symbol for the product of all positive interfere less than or equal to N. The K of 0.75 and above is considered to have the excellent agreement of relevance (21).
To confirm for face validity, we conducted a qualitative approach using cognitive interviews (22, 23). A pilot study was performed, in which the instrument was administered to 15 elderly patients who presented in the ED. The respondents provided feedback if they have difficulties answering the items and if any of the items were confusing or containing difficult vocabulary. The feedback was carefully assessed and reworded to give the final version of the Malay-translated KCL.
Phase three: field testing
Over 4 months between September 2020 and December 2020, a convenient sample of elderly patients aged 60 and above who presented in the ED of UKMMC were invited for the recruitment of the subject. Patients must be able to read and speak in the Malay language. We excluded patients with preexisting cognitive impairment and underlying psychiatric illness and who are critically ill. All included subjects provided written consent. Recruitment was conducted without interfering with clinical evaluation, investigations, treatments, and interventions by the attending clinician. The following data were documented: age, sex, race, education level, marital status, living support, comorbidity, diagnosis, and disposition.
Frailty assessment
The Malay-translated KCL was self-administered by the patients or surrogates for patients with physical difficulties. No time limit was given during the administration. In addition to the Malay KCL, we used the CFS, a clinical judgment-based screening tool that evaluates specific frailty domains including comorbidity, function, and cognition, to assess frailty. The CFS has been validated and proven to be reliable to identify frailty in EDs (24). The CFS has a total score of 9, ranging from 1 (very fit) to 9 (terminally ill), and a total score of 5 or more is considered frail (25). Measurements using either instrument represent trait or baseline at two weeks before assessment as suggested from the previous study (26).
Data analysis and sample size
Statistical analysis was performed using the IBM Statistical Package for Social Sciences (SPSS) Statistics for Windows software and SPSS Amos (version 26.0; IBM Corp., Armonk, NY, USA). A descriptive analysis of study variables was summarized using frequency and percentages for categorical variables and mean ± SD or median (interquartile range) for continuous variables.
Confirmatory factor analysis (CFA) was performed to evaluate the factor structure of the Malay KCL. Three prior hypothesized factorial models of KCL were chosen. The first model was a seven-correlated factor model based on the original KCL (15), and the dimensions are as follows: the activity of daily living (items 1–5), physical strength (items 6–10), nutrition (items 11 and 12), oral condition (items 13–15), socialization (items 16 and 17), memory (items 18–20), and mood/cognition (items 21–25). The second model was a two-correlated factor structure based on the approach adopted by Fukutomi et al. (27). In this model, items 1–20 were classified under the lifestyle domain, and items 21–25 were classified under the mood or depression domain. The third factor was the one-factor structure where all the KCL items load on a single factor of frailty as proposed by Satake et al. (14). Model fit for each structure was assessed using the following fit statistics: χ2 goodness-of-fit test, comparative fit index (CFI), Tucker–Lewis Index (TLI), and root mean square error of approximation (RMSEA). The insignificant (p-value > 0.05) model χ2 goodness-of-fit test indicates model fit. The χ2 goodness-of-fit is sensitive to the sample size. Therefore, for the well-fitting model, other fit indices, namely, CFI, TLI, and RMSEA, must be ≥ 0.93, ≥ 0.92, and ≤ 0.08, respectively (28). The models were compared by examining the Akaike information criterion (AIC), Bayesian information criterion (BIC), and expected cross-validation index (ECVI), where smaller values indicated a better fit (29, 30).
The correlation of the KCL scores with the CFS was determined using Pearson, r correlation analysis. Internal consistency was evaluated using the Cronbach’s alpha with values α > 0.7 considered acceptable reliability (31). For all analysis, p < 0.05 was considered significant.
In performing factor analysis, we determined the sample size according to the number of items, whereby five subjects are required per item of the questionnaire (32). With 25 items in the questionnaire, the minimum subjects were 225.