Ethical approval

An instrument translation, evaluation and validation study were carried out from June 2019 to February 2020. The study instrument was approved by the Research Ethics Committee (REC) of Hong Kong Baptist University (HKBU) (Ref. No.: REC/18-19/0021). It was registered in ClinicalTrials.gov (ClinicalTrials.gov ID: NCT04491890) on July 26, 2020.

Participants

The CMPs for content validity were recruited from the outpatient clinics of the School of Chinese Medicine (SCM) of HKBU; the participants of evaluation and validation were recruited from the outpatient clinics of the SCM of HKBU and the community started from July 2019 to February 2020. Eligible participants were aged 65 and above HK residents, capable of consenting and agreed to participate. The potential participants who have serious diseases including serious mental or behavioral disorders were excluded.

Procedures

This study (Fig. 1) was comprised of four parts as recommended by the International Society for Quality of Life Assessment (IQOLA) project: (1) translation of the TCMECQ (First Version); (2) Content validation based on the First Version; (3) Form the second version of TCMECQ-C based on the first validation, reviewed by the TCMECQ-C developers (ZXB and YLW) and discussed in the expert panel’s meeting, and the third version of the TCMECQ-C was then formed by the panel; (4) Construct validation of the TCMECQ-C (Third Version) [18,19,20].

Fig. 1

The first phase of study was the translation of the TCMECQ into Cantonese version. Permission to translate and reword the TCMECQ-C with reference to the CCMQ was obtained from the author of the CCMQ (Cantonese version). The translation process followed the guidelines of Beaton et al.: (1) forward translation, (2) synthesis of translations, (3) back translation, (4) review by the expert committee, (5) pretesting and cognitive interviewing. Forward and backward translations were conducted in June 2019. One Chinese Medicine Practitioner (CMP) (YLW) and one social worker (SYW), both bilingual native Cantonese speakers, carried out the forward translation (Mandarin to Cantonese), while two independent bilingual translators of native Mandarin without medical background [one was a financial analyst (ZW); another was a PhD student from Communication Studies (LZ)] produced two back translations (Cantonese to Mandarin) of the questionnaire. The TCMECQ-C developers comprising the principal investigator (ZXB) and the CMP translator (YLW) produced the first version of TCMECQ-C. A volunteer group of 10 CMPs recruited from the HKBUSCM, 30 participants aged 65 and above from the community, were invited to evaluate the first version for content validity by cognitive debriefing interviews from July 2019 to August 2019 [12, 18, 19, 21, 22]. Then the TCMECQ-C developers (ZXB and YLW) reviewed the study results and produced the draft for the second version and submitted for the expert panel’s appraisal. An expert panel (see Additional file 1), including the original translator of the CCMQ, 3 experts on TCM basic theory, and 1 professional from the Faculty of Arts of HKBU, finalized the TCMECQ-C on 23 October 2019 as the Third Version [12].

In the second phase of the study, to investigate the construct validity of the TCMECQ-C (Third Version), 303 participants were recruited from December 2019 to February 2020. Each participant was assessed by a trained CMP for the participant demographic characteristics and the responses of TCMECQ, TCMECQ-C. During the first assessment, the CMP recorded the participant-provided demographics answers (e.g. gender, education) and participant’s chosen responses most reflecting their conditions. The CMP read each question of TCMECQ to the participant but did not intervene the participant’ choice for assisting in clarifying, explaining according to the original intentions of the questions to be assessed. Continuing the finished TCMECQ, without the reference to the previous answers of the responded TCMECQ, the participant answered each question of TCMECQ-C (Third Version) with also the CMP’s assistance. To evaluate test–retest reliability, the TCMECQ-C (Third Version) was again administered by the CMP to the same sample participants 2 weeks after the initial assessment [23,24,25]. 89.70% (270/301) participants completed the second visit for the test–retest assessment. With the accomplishment of the validation, suggested rewording of 4 questions of the TCMECQ-C was submitted to the panel of experts for discussion and finalization. Expert panel commented those questions deeply and endorsed the rewording of the 4 questions on 22 March 2020. The TCMECQ-C developers finalized the TCMECQ-C following the suggestions. In the third version of the TCMECQ-C, 16 questions had been reworded and 4 questions had been restructured into sub-questions respectively; in the validated endorsed third version (Final Version) of TCMECQ-C, 5 questions have been reworded (see Additional file 2).

MeasuresTCMECQ and TCMECQ-C (first version)

The TCMECQ designed for the elderly in Beijing was translated to TCMECQ-C (First Version) targeting HK elderly aged 65 and above to identify their body constitution, formulate the health regimen strategy and facilitate the elderly constitution research and disease prevention. Both the TCMECQ and TCMECQ-C (First Version) comprises 33 items (including 4 questions (item 2, 4, 5 and 13) are shared between Qi-deficiency Constitution, Yang-deficiency Constitution, Stagnant Qi Constitution and Balanced Constitution) with the items identifying the respective nine body constitutions [5, 16, 26]. All items are calculated based on the five-point Likert scale scoring algorithm [5]. All the 8 biased constitutions were the sum scores of their corresponding items, with scores ≥ 11 indicating having the assessed constitution(s). The scoring of the Balanced Constitution: item 1 + item 2 (reverse scoring) + item 4 (reverse scoring) + item 5 (reverse scoring) + item 13 (reverse scoring) ≥ 17 and with all other eight biased constitution ≤ 8.

TCMECQ-C (second version)

The TCMECQ-C (Second Version) comprises 52 items (including 4 questions (item 2, 4, 5 and 13) are shared between Qi-deficiency Constitution, Yang-deficiency Constitution, Stagnant Qi Constitution and Balanced Constitution) with the items identifying the respective nine body constitutions, (1) for Yang-deficiency Constitution, the original item 12 and item 27 was split into 4 and 2 sub questions respectively, making it 4 more items were added to the TCMECQ; (2) for Yin-deficiency Constitution, the original item 10, item 20 and item 29 was split into 2 sub questions respectively, making it 3 more items were added to the TCMECQ; (3) for Phlegm-dampness Constitution, the original item 22 and item 32 was split into 4 and 2 sub questions respectively, making it 4 more items were added to the TCMECQ; (4) for Dampness-heat Constitution, the original item 24 was split into 2 sub questions, making it 1 more item was added to TCMECQ; (5) for Stagnant Blood Constitution, the original item 23 was split into 2 sub questions, making it 1 more item was added to TCMECQ; (6) for Stagnant Qi Constitution, the original item 8 was split into 2 sub questions, making it 1 more item was added to TCMECQ; (7) for Inherited Special Constitution, the original item 17 was split into 6 sub questions, making it 5 more items were added to TCMECQ, [5, 16, 26] 19 more items in total were thus added to form the TCMECQ-C (Second Version). Considering some measurements were carried out for the participants by the CMP (for example, item 20 the scratch test may done by the CMP), the order of the items was rearranged (for example, item 20 was rearranged to be item 30). All items are calculated based on the five-point Likert scale scoring algorithm which was consisted with TCMECQ. All the 8 biased constitutions were the sum scores of their corresponding items, with scores ≥ 11 indicating having the assessed constitution(s). The scoring of the Balanced Constitution: item 1 + item 2 (reverse scoring) + item 4 (reverse scoring) + item 5 (reverse scoring) + item 13 (reverse scoring) ≥ 17 and with all other eight biased constitution ≤ 8. The 4 biased constitutions including Yang-deficiency Constitution, Stagnant Blood, Stagnant Qi and Inherited Special Constitutions have the consistent scoring method and the cut-off value with TCMECQ, by counting the highest score from the composed items in the corresponding subscales to get one score to be added up with other items’ scores, such as in Stagnant Qi Constitution, the score was calculated as item 5 + item 6 + item 7 + item 8 (count the highest score from “8.1, 8.2” to get one score for item 8).

TCMECQ-C (third version)

The TCMECQ-C (Third Version) comprises 43 items (including 4 questions are shared between Qi-deficiency Constitution, Yang-deficiency Constitution, Stagnant Qi Constitution and Balanced Constitution) with the items identifying the respective nine body constitutions, (1) for Yang-deficiency Constitution, the original item 12 was split into 4 sub questions, making it 3 more items were added to the TCMECQ; (2) for Stagnant Blood Constitution, the original item 24 was split into 2 sub questions, making it 1 more item was added to TCMECQ; (3) for Stagnant Qi Constitution, the original item 8 was split into 2 sub questions, making it 1 more item was added to TCMECQ; (4) for Inherited Special Constitution, the original item 17 was split into 6 sub questions, making it 5 more items were added to TCMECQ, [5, 16] 10 more items in total were thus added to form the TCMECQ-C (Third Version). All items are calculated based on the five-point Likert scale scoring algorithm which was consisted with TCMECQ. All the 8 biased constitutions were the sum scores of their corresponding items, with scores ≥ 11 indicating having the assessed constitution(s). The scoring of the Balanced Constitution: item 1 + item 2 (reverse scoring) + item 4 (reverse scoring) + item 5 (reverse scoring) + item 13 (reverse scoring) ≥ 17 and with all other eight biased constitution ≤ 8.

The 4 biased constitutions including Yang-deficiency Constitution, Stagnant Blood, Stagnant Qi and Inherited Special Constitutions have the consistent scoring method and the cut-off value with TCMECQ, to remain unchanged with the TCMECQ scoring method. Further, we count the highest score from the composed items in the corresponding subscales to get one score to be added up with other items’ scores, such as in Yang-deficiency Constitution, the score was calculated as item 11 + item 12 (count the highest score from “12.1, 12.2, 12.3 or 12.4” to get one score for item 12) + item 13 + item 29. The original item 12 (in Yang-deficiency Constitution), item 24 (in Stagnant Blood Constitution), item 8 (in Stagnant Qi Constitution) and item 17 (in Inherited Special Constitution) were split into sub questions respectively, aiming a more precise identification which considering important in TCM theory. For example, for Yang-deficiency Constitution, the original item 12 was split into 4 sub questions, aiming to more precisely identify the exact region of the body that afraid of cold.

Sample size

The recommended 10 CMPs and 30 participants were tested for the content validity [18, 27]. The sample size for the designated construct validation study was determined in terms of the subject to item ratio 5:1, with reference to two review articles [28, 29]. Therefore, a sample size of at least 215 participants (5 times 43 items of the questionnaire) was needed.

Statistical analysis

All statistical analyses were performed using SPSS (version 28) and AMOS (version 28) for Windows. Data are reported as mean \(\pm\) standard deviation (SD) for continuous variables, frequency and proportion for categorical variables. Independent sample t-test and Pearson correlation were used with the normal distribution of the data [30]. Two-sided p < 0.05 was considered significant.

Content validity index (CVI) (Evaluation Study Sample) examined content domain to ensure the item covered the intended TCMECQ-C (First Version) [30, 31]. Item-level content validity indices (I-CVIs) were calculated by dividing the number of CMPs and the participants that rated grade 3 (quite relevant/clear/appropriate) or 4 (highly relevant/clear/appropriate) on the four-point Likert scale of the first three questions in the cognitive debriefing table by the total number of assessments of the interview respectively [31,32,33]. An item-level content validity index (I-CVI) ≥ 0.78 was considered acceptable for at least nine experts [27, 31, 34].

Validity (validation study sample)

Scaling assumptions was used to determine whether the item scores of each scale can be summated, were tested by the summated rating method, including (1) item frequency distribution (to ascertain whether all of the response choices were used), (2) equivalence of item means and standard deviations (i.e. approximately equivalent within a scale, to check the floor and ceiling effects) [35]. Ceiling and floor effects were defined usually with ≥ 15% of participants in a sample achieving the best/maximum or the worst/minimum score, disabling the measure to discriminate between subjects at either extreme of the scale [36]. A critical ratio > 3 with p < 0.05 (calculated by independent sample t-test) indicating an item had good discriminability [16]. Explanatory Factor Analysis (EFA) was used to confirmed the dimensionality [37, 38]. Three major steps included a) assessment of the suitability of the data (recommended sample size i.e. ten cases for each item the study sample size (n = 270) met the requirement for factor analysis [37, 38]. Coefficient of correlation > 0.30 between the variables evidenced the suitable strength of the relationship among the items. Determinant score > 0.00001. Kaiser–Meyer–Olkin > 0.60 and Bartlett’s test of Sphericity < 0.05) [37, 38]; b) factor extraction (9 factors (according to the known original 9-constitution scale structure) were extracted using principal component analysis (PCA). Considering the TCM theory, the items were specifically selected indicating a factor and comprising an essential part of the factor, the item(s) were included even if they did not reach the suggested factor loading level [39]. The required cumulative variance percentage was ≥ 50% [37]. and c) factor rotation interpretation using orthogonal factor rotation approach with the related varimax method developed by Kaiser (1958), to minimize the significant cross loadings (i.e. many factors were correlated with many variables) usually obtained in the initial extraction phase. The factor rotation method solved the interpretation of the items loaded factors (i.e. avoided the number of items having high loadings on each factor). The diagonal anti-image correlation value of each item must be > 0.5 to inform the sampling adequacy. The factor loading > 0.30 indicating that the items represent the underlying factors [37, 40, 41].

Confirmatory Factor Analysis (CFA) model fit was performed on the overall scale and subscales using AMOS. The model fit of the overall scale excluding four items in the Balanced Constitution subscale due to the reverse scores of Q2, Q4 (both shared with Qi-deficiency Constitution), Q5 (shared with Stagnant Qi Constitution) and Q13 (shared with Yang-deficiency Constitution). Statistical parameters evaluated in the CFA process including (1) the absolute fitting statistics: goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR); (2) the incremental fitting statistics: normed fit index (NFI), incremental fit index (IFI), Tucker–Lewis index (TLI), comparative fit index (CFI) and relative fit index (RFI); and (3) the parsimony fitting statistics: the chi-squared degree-of-freedom ratio (χ2/DF). GFI, AGFI, NFI, IFI, TLI, CFI and RFI ≥ 0.90, RMSEA and SRMR ≤ 0.08 and χ2/DF between 1.00 and 3.00, implied excellent CFA model fit [16]. Pearson correlation, i.e. Pearson correlation coefficient (r) was investigated for the construct validity (convergent validity and divergent validity) among the nine subscales of TCMECQ-C [35, 42]. Sufficient convergent validity between the subscales were represented by (1) moderate to strong correlation: 0.30 < r < 0.70, (2) strong correlation: r ≥ 0.70; sufficient divergent validity between the subscales: r ≤ 0.30) [35]. Assuming moderate to strong correlations among the three constitutions (Qi-deficiency Constitution, Yang-deficiency Constitution and Yin-deficiency Constitution) as they are measuring the deficiency construct and other five constitutions (Phlegm-dampness Constitution, Dampness-heat Constitution, Stagnant Blood Constitution, Stagnant Qi Constitution, Inherited Special Constitution) as they are measuring the excessive construct. Assuming low correlations between deficiency subscales and excessive subscales as the hypotheses for the divergent validation.

Reliability (validation study sample)

Internal consistency was measured by Cronbach’s alpha [43]. A rule-of-thumb good scale consistency was defined with the Cronbach’s alpha coefficient between 0.50 and 0.70 based on the scale dimensionality [44]. Hinton et al. suggested four cut-off points for a more concrete reliability acceptable level including excellent reliability (0.90 and above), high reliability (0.70–0.90), moderate reliability (0.50–0.70) and low reliability (0.50 and below), was adopted to reveal the true reliability of the scale [45]. A lower Cronbach’s alpha was considered sufficient in indicating consistency for scales with less than 10 items [12, 44,45,46]. The corrected item-total correlation defined the association of the item with the total score on the other items [44, 47, 48]. Corrected item-total correlation using a correlation coefficient ≥ 0.40 as the cut-off for adequate correlation [45, 49]. Moreover, a correlation higher than 0.20 suggested that each item has a good correlation with the scale [44]. Items having item-to-total correlation less than 0.20 were retained if Cronbach’s alpha did not increase upon deleting these items [44, 47, 48]. Test–retest reliability was evaluated by Intra-class correlation coefficients (ICC) [50, 51], in which using the single-measurement (comparing TCMECQ-C (Third Version) with TCMECQ), absolute-agreement, the two-way mixed effects model, the ICC results based on the 95% confident interval of the ICC estimate were then reported [51]. Values < 0.50 were indicative of poor reliability, values between 0.5 and 0.75 indicated moderate reliability, values between 0.75 and 0.9 indicated good reliability, and values > 0.90 indicated excellent reliability [51].