Ethics and permission for data collection
Ethics approval for this study (Project Number: PS/2018/1/11) was granted by the University Research Center, Shahjalal University of Science & Technology, Bangladesh and formal permission for data collection was sought from Sylhet Civil Surgeon Office, Bangladesh.
Study design
A cross-sectional study design, involving a person-centered general health assessment and a self-administered survey, was employed in this research.
Setting and participants
This study was conducted from January 2018 to December 2019 in the north-eastern region (i.e. Sylhet City Corporation) of Bangladesh. In Bangladesh, while elderly cut-off age is 60 years and the national average of life expectancy is 72.05 years, literature notes that people who live in Sylhet have a lower life expectancy than the national average [29]. It is evident in literature that physical fitness or health condition of a person starts decline at the age of 55 years and present signs of aging (wrinkles, dullness of skin, weakness, dry skin, blotchiness and age spots) [30–34]. In Sylhet, the health literacy is low and the people are not aware of their healthcare needs until their problems are manifested [35]. As a result, we approached the people aged 55 years and above through Sylhet City Corporation. The inclusion criteria were: (i) aged 55 years and older; (ii) living in Sylhet City Corporation; and (iii) agreed to participate in a general health assessment and an interview. We used the single-stage cluster sampling, and this resulted in a random sample size of 400 participants from 13 administrative wards of the City Corporation. The administrative wards were selected through random number generated by R-Programming language.
Data collection
A multi-indicator survey design was employed to explore diverse health issues and conditions of the participants. In accordance with the standards of Helsinki Declaration of 2000 (revised version), written informed consent was obtained from each of the participants. After formal consent, the health profile of each participant was collected, through a general health assessment and a structured questionnaire, including self-reported health problems, biomarkers, performance of daily activities and socio-demographic information. Necessary medical equipment was provided to a trained public health assistant to assess and collect anthropometric data.
Outcome measures
General health assessment of height, weight, BMI, BP and RBS
The following Table 1 presents the measurement system of basic health indicators.
Frailty Index (FI30) measurement and coding of variables
Developing a list of variables is important to compute frailty, while there is no uniform list and researchers used different variables in calculating the Frailty Index. We consulted with clinicians and an epidemiologist, included 30 categorical variables specific to the older adults’ frailty and relevant to our socioeconomic contexts, following the works of Gobbens et al., 2010; Searle et al., 2008; and Rockwood et al., 2018 (Table 2) [36,37]. Categorical variables were coded using the convention that '0' indicated the absence of the deficit, and '1' the presence of a deficit [36–38]. For some categorical variables (e.g. self-rated health, BMI etc.) that comprised one or more intermediate responses (e.g. 'average' or 'frequently'), we considered the additional value of ‘0.25’, '0.5' and ‘0.75’. We used the short form of Geriatric Depression Scale and a list of 15 questions in a separate questionnaire to collect data [39]. The mathematical formula used for calculating FI30 in this study was as follows:
See formula 1 in the supplementary files.
There was no missing value in the datasets and the calculated FI30 was a continuous score. Obviously, FI30 lies between 0 and 1, however, it was categorized, according to Clegg et al. (2016): FI30 ≤ 0.12 (No Frailty or Fit/Good Health); 0.12 < FI30 ≤ 0.24 (Mild Frailty/Slightly Poor Health); 0.24 < FI30 ≤ 0.36 (Moderate Frailty/Poor Health); and FI30 > 0.36 (Severe Frailty/Very Poor Health) for analysis purposes [40]. We further classified the FI30 into: No Frailty/Fit (Fairly Healthy, FI ≤ 0.24); and Frailty (Medical Conditions, FI > 0.24) to perform the multivariable binary logistic regression.
Data analysis
We employed frequency distribution to understand the participants’ demographics. The relationships between the participants’ frailty (physio-psychosocial health) and socio-demographic factors were tested using bivariate analysis. The Pearson chi-square test was performed to observe the significant association between FI30 and socio-demographic variables. Student t-test of Frailty Index was employed to assess the mean difference between male and female older adults. The degree of associated risk factors was assessed by univariable [unadjusted odd ratio (unadjusted OR)] and multivariable odds ratios [adjusted odds ratio (adjusted OR)] obtained from the binary logistic regression model [41], where FI30 was a binary dependent variable, and physio-psychosocial variables were independent variables. This project’s data management and statistical analyses were carried out through IBM SPSS Statistics 20.0.
To ensure reliability and validity of the study results, we used a number of techniques: (i) employ SPSS 20.0 in data analysis; (ii) explore descriptive statistics of the sample (socio-demographic items) and their means and standard deviations; (iii) reliability analyses of the data used in calculating FI30 (Cronbach’s alpha, the reliability coefficient values for the variables used in FI30 is found 0.70); and (iv) use of two statistical techniques, namely chi-square test and logistic regression [42].