Study population and design
This was a cross-sectional validation study using data from a prospective cohort study recruiting geriatric outpatients at the National Taiwan University Hospital (NTUH) in Taipei, Taiwan. Data was collected between June and December 2019. The inclusion criteria were age ≥ 65 years and having at least one of the following geriatric syndromes: fall or functional decline in recent one year, polypharmacy ≥ 5, urinary incontinence, history of osteoporosis or weight loss (≥ 5% in one month or 10% in 6 months). Our study excluded patients with severe dementia, severe hearing or visual impairment, severe functional impairment or contact precautions for multidrug resistant organisms in order to avoid communication or cooperation barriers. The study was approved by the Research Ethics Committee at NTUH. Written informed consent of the study participants was obtained before enrollment.
Sample size calculation
For inter-rater reliability, we assumed the minimum acceptable kappa was 0.2 and anticipated a substantial agreement (weighted Kappa = 0.61-0.80) between CFS-C of physicians and one research assistant. At least 48 participants were required for assuring a power of 80% and a significant level of 0.05 to detect a statistically significant kappa coefficient [26, 27]. For criterion validity, we assumed the minimum acceptable Kendall’s tau correlation was 0.2 and expected a high correlation (Kendall’s tau > 0.3) between CFS-C and Fried frailty phenotype. Thus, at least 211 participants were required for assuring a power of 80% and a significance level of 0.05 to detect a statistically significant Kendall’s tau coefficient [28]. Allowing 5-10% attrition rate for missing data, our study enrolled 226 geriatric outpatients. After exclusion of 5 participants who had no CFS-C assessment (n = 4) or no BabyBot vital data (n = 1), a total of 221 subjects were included for criterion validity and 52 of them were included for reliability analysis [see Additional file 1].
Data collection
A wide range of demographic and health data was collected on BabyBot vital data recording system (Netown Corporation, Taiwan) and comprehensive geriatric assessment (CGA). BabyBot included a 68-item self-reported questionnaire, bioelectrical impedance analysis (Tanita BC-418), and tests of hand grip, timed-up and go (TUG), and 6-meter (6-m) walk. CGA, comprised of Mini-Mental State Examination (MMSE) [29], Geriatric Depression Scale-15 (GDS-15) [30], Mini-Nutritional Assessment (MNA) [31], Barthel Index (BI) [32], and Instrumental Activities of Daily Living (IADL) [33], was evaluated by a trained research assistant. To measure comorbidity, six geriatricians scored the Cumulative Illness Rating Scale for Geriatrics (CIRS-G) [34].
Translation of the Clinical Frailty Scale into Chinese
With Dr. Rockwood’s permission, we undertook the translation process following Brislin’s translation model [35, 36]. To start, the English version CFS (referred to as the source CFS) was translated into traditional Chinese by one of the authors of this study, as well as by a bilingual translator working independently. The two translated CFS documents were evaluated and compared with the source CFS by a panel of experts (seven geriatricians and one nurse practitioner) to reach consensus. Afterwards, back translation was independently conducted by two bilingual primary care physicians who had never seen the source CFS. Lastly, three bilingual experts and a panel of geriatric experts were involved in group discussion to compare the two back translations with the source CFS. Minor discrepancies were resolved, and the expert reviewers agreed on the production of the final Chinese version of CFS (CFS-C, Fig. 1).
Assessment of frailty
The Chinese version of Clinical Frailty Scale (CFS-C)
The CFS-C was scored by the same trained research assistant after completing BabyBot and CGA. For the reliability group of 52 participants, CFS-C was scored independently and simultaneously by their geriatricians after reviewing the results of BabyBot and CGA. The results of CFS-C were blinded to each other. For criterion concurrent validity, CFS-C was categorised as robust (CFS-C = 1-2), prefrail (CFS-C = 3-4) and frail (CFS-C = 5-9) [15].
The Fried Frailty phenotype
Fried frailty phenotype was assessed by five criteria: exhaustion, weight loss, low activity, weakness, and slowness [2]. We assessed presence of exhaustion, weight loss or low activity by reporting of a “yes” answer to the following items in the self-reported questionnaire: “Feeling tired or fatigue in recent one month”, “weight loss of more than 3 kg or 5% in the previous year” and “low physical activity”, respectively. Weakness was determined by having low grip strength below established cut-off (< 28 kg in men, < 18 kg in women) [37]. Slowness was defined as gait speed < 1 m/s based on the 6-m walk or the participant was not able to walk [37]. From a 5-point scale, participants scored 0 were defined as non-frail, scored 1 or 2 as pre-frail, and scored ≥ 3 as frail.
Frailty Index based on a comprehensive geriatric assessment (FI-CGA)
FI-CGA gathered information on ten standard domains from CGA and BabyBot, including cognition, emotion, communication, mobility, balance, bladder function, bowel function, nutrition, activities of daily living and social resources [38-40]. For each domain, “0” indicated no problem, “0.5” indicated a minor problem, and “1” indicated a major problem. Scores were summed up into an impairment index, ranging from 0 to 10. For co-morbidity index, CIRS-G was standardized to a range from 0 to 4, representing equivalence of 4 deficits. To construct FI-CGA, the sum of the impairment and co-morbidity index were further divided by 14 into a range from 0 to 1. The detailed scoring criteria were presented in Table 1. According to previous reported cutoffs, participants were categorised as robust (FI-CGA ≤ 0.08), prefrail (0.08 < FI-CGA ≤ 0.25) and frail (FI-CGA ≥ 0.25) [41].
Table 1
Frailty Index based on a Comprehensive Geriatric Assessment (FI-CGA)
Domains
|
|
Scoring methods
|
|
Data source
|
1. Cognition
|
|
0 - Normal MMSE
|
|
CGA
|
|
0.5 - Abnormal MMSEa and normal IADL and BI
|
|
|
1 - Abnormal MMSE and (IADL or BI)
|
|
2. Emotion
|
|
0 - GDS < 5
|
|
CGA
|
|
0.5–5 ≤ GDS < 10
|
|
|
1 - GDS ≥ 10
|
|
3. Communication
|
|
0 - No deficit in communication, hearing, vision
|
|
Questionnaire from Babybot
|
|
0.5–1 deficit in either communication, hearing, vision
|
|
|
1 - ≥2 deficits in either communication, hearing, vision
|
|
4. Mobility
|
|
0 - TUG < 10
|
|
TUG test from Babybot
|
|
0.5–10 ≤ TUG ≤ 19
|
|
|
1 - TUG > 19 or unable to walk
|
|
5. Balance
|
|
0 - No self-reported poor balance and no fall in previous year
|
|
Questionnaire from Babybot
|
|
0.5 - Report of either fall in previous year or poor balance
|
|
|
1 - Report fall and poor balance
|
|
6. Bladder
|
|
0 - Bladder control in BI = 10
|
|
CGA
|
|
0.5 - Bladder control in BI = 5
|
|
|
1 - Bladder control in BI = 0
|
|
7. Bowel
|
|
0 - Bowel control in BI = 10
|
|
CGA
|
|
0.5 - Bowel control in BI = 5
|
|
|
1 - Bowel control in BI = 0
|
|
8. Nutrition
|
|
0 - MNA = 12–14
|
|
CGA
|
|
0.5 - MNA = 8–11
|
|
|
1 - MNA = 0–7
|
|
9. ADL
|
|
0 - IADL = 8 and BI = 100
|
|
CGA
|
|
0.5 - IADL < 8 and BI = 100
|
|
|
1 - BI < 100
|
|
10. Social resources
|
|
0 - Not living alone and someone could help if needed
|
|
Questionnaire from Babybot
|
|
0.5 - (Living alone but someone could help if needed) or (not living alone but no one could help if needed)
|
|
|
1 - Living alone and no one could help if needed
|
|
Impairment Index = sum of deficits (numbers of deficits = 0–10)
|
Comorbidity Index = CIRS-G standardized to 0–4 (numbers of deficits = 0–4)
|
FI-CGA = (Impairment Index + Comorbidity Index)/14
|
aAbnormal MMSE was defined as MMSE ≤ 23 if years of education > 2 years or MMSE ≤ 13 if years of education ≤ 2 years. |
Statistical analysis
Descriptive analysis was presented as numbers (%) for categorical data, and mean ± standard deviation for continuous variables. Weighted kappa for agreement and Kendall’s tau for correlation were used to assess inter-rater reliability and validity tests. Inter-rater reliability was assessed between physicians and the research assistant. For criterion concurrent validity, CFS-C was compared with both Fried frailty phenotype and FI-CGA. Kendall’s tau was used to assess correlation between CFS-C and other geriatric assessments, including BI, IADL, MNA, MMSE, GDS, CIRS-G, 6-m gait speed, TUG, hand grip and appendicular skeletal muscle mass (ASM). Data was analyzed by using SAS version 9.4 (SAS Institute Inc., Cary, NC). A two-sided p < 0.05 was set as statistically significance.