Data source
This study was designed based on secondary data from a nationally representative and unique survey, the Bangladesh Household Income and Expenditure Survey (HIES) 2016 (HIES2016). This survey was carried out by the Bangladesh Bureau of Statistics (BBS) from April 2016 to March 2017. The final report on this 16th round of HIES described the survey objectives, survey design, sampling technique, survey tools, measuring system, sample size calculation, quality control, and the questionnaire’s new modules [32]. The HIES2016 covered the ever highest number of 46,080 households selected from 2304 primary sampling units (PSUs), from 20 strata under 3 basic localities (rural, urban and city corporation): 8 rural, 8 urban, and 4 statistical metropolitan areas (Dhaka, Chattogram, Rajshahi and Khulna). It contains information on the household, disability, education, health, housing, and a wide range of socio-economic factors (e.g., family earnings, consumption and expenditures, assets, housing conditions, as well as data on demographic variables, education, employment) that has a strong role in the decision making process for the government.
The HIES 2016 followed a stratified two-stage cluster sampling design. At the first stage, PPS (probability proportional to size) systematic sampling technique was used to draw a total of 36 PSU’s from each district, where the number of households in each PSU being the measure of size. Enumeration Area, a cluster of around 110 households of Population Census 2011, was treated as PSU for this sample design. After selection of the PSU’s, a complete household listing in these selected PSU’s was done in the field. Thus, the total calculated sample size for the survey stands at 46,080 (2304 × 20) households. However, a total of 46,076 households were surveyed, with 32,096 from rural areas and 13,980 from urban areas. Among the selected households, a total of 186,076 individuals were interviewed, 130,435 from rural areas and the remaining 55,641 from urban areas. We considered both household and individual-level data in our analysis.
Study data
For calculating the wealth quintile using household characteristics, we first converted each of the household characteristics into a binary variable. Based on the binary variables, we computed the wealth quintile using a multivariate technique principal component (PC) analysis, where the first PC score was considered as the wealth quintile. Then these quintile values were allocated to all individuals based on households.
To finalize the study population for our study, we excluded some observations from the source data based on our inclusion/exclusion criteria (Fig. 2). Among 46,076 households, eight households were primarily excluded due to the lack of respondent’s information. Among the 186,083 individuals in the remaining 46,086 households, 152,347 were excluded as they did not suffer from chronic illness. Again, 968 individuals were excluded due to missing information on any of the socio-economic and demographic factors considered in our study. So far, we got 32,768 chronically ill individuals who constituted our Study Population to be analyzed in this study. For out-of-pocket expenditure analysis, data from 6572 chronically ill individuals were considered after excluding 26,164 individuals with missing OOPE (among them, 22,771 did not seek medical treatment and 3393 did not report OOPE) and 32 individuals with zero (0) OOPE from the study population of 32,768 individuals. We excluded zero values of OOPE from the analysis to avoid complexity in the normalization of data (e.g., the natural logarithm of zero is undefined). The whole process of inclusion/exclusion is depicted in the following Fig. 2.
Study variables
Outcome variables
For conceptualizing disability, an International Classification of Functioning (ICF), Disability, and Health (ICFDH) was established by the World Health Organization (WHO) (https://www.who.int/classifications/icf/en/). Implementation of the ICF-based technique for disabilities requires the improvement of new measurement models to conduct both surveys and censuses. A useful small set of six disability-related questions was developed and adopted by the Washington Group for national censuses and surveys [33]. The HIES2016 utilized these six disability-related questions to be consistent with the ICFDH [33] and they are on difficulty, (i) for seeing even if he/she is wearing glasses, (ii) in hearing even if he/she is wearing a hearing aid, (iii) for walking or climbing or any other physical movement, (iv) in remembering or concentrating, (v) in self-care such as washing all over or dressing, feeding, toileting, etc., and (vi) in communicating. In the HIES2016, each individual of the household was asked to provide information about the presence of any disability and the severity of the disability. Each question had four response levels: (1) No Difficulty, (2) Yes, Some Difficulty, (3) Yes, Severe Difficulty or (4) Yes, Can’t see/hear/walk/remember/self-care/communicate at all. For the convenience of analysis, we converted the presence of disability into two groups: ‘0’ denoted for “No difficulty” and ‘1’ for “Any level of difficulty” [33,34,35]. Moreover, an outcome variable was defined as ‘chronic illness with a disability’ based on disability status and chronic illness (chronic fever, injuries/disability, chronic heart disease, respiratory disease/asthma/bronchitis, diarrhea/dysentery, gastric or ulcer, blood pressure, arthritis/rheumatism, skin problem, diabetes, cancer, kidney diseases, liver diseases, mental health, paralysis, ear/ENT problem, eye problem, and others) with two values: 1 denoting “Presence of at least one of the above six disabilities with chronic illness” and 0 denoting “Absence of disability with chronic illness”. For the second burden measurement, OOPE for healthcare was created by adding up direct medical costs, including hospital outpatient fees, medicine, admission or registration fees, physician fees, diagnostic fees, and any other associated medical supplies and direct non-medical costs, including transportation and conveyance, lodging, tips, and other associated costs [32]. Indirect costs such as loss of opportunity, productivity, and other intangible costs, including pain and suffering, were not captured in the HIES2016 by the BBS. Therefore, we were limited to include only the direct healthcare costs in the calculation of OOPE. In Bangladesh, OOPE is measured in Bangladeshi Taka (BDT) equivalent to the U.S. $0.01163 (i.e., 1 USD = 86 BDT).
Independent variables
To take into account the minority problem and the coastal climate crisis, we defined and categorized ‘Religion’ as ‘Muslim/Non-Muslim’ and ‘Region’ as ‘Exposed Coast/Non-Exposed’. The ‘Non-Exposed’ region consists of interior coast and non-coastal areas. In Bangladesh, the Muslims are the majority and the non-Muslims (Hindus, Buddhis, Christians and others) led by Hindus are the minority. A detailed delineation of the exposed coast and non-exposed areas are provided in Uddin and Kaudstaal [36], and Bahauddin et al. [37]. The coastal regions of Bangladesh, with 19 districts containing 147 Upazilas, cover/occupy 32% of the country’s total geographic area, wherein 28% of the country’s total population live. To be more focused, we recognized a further division between the coastal areas as the exposed coast and the interior coast; where the former areas with 48 Upazilas are exposed to the sea and/or lower estuaries, and the later areas with the remaining 99 Upazilas located behind the exposed coastal were added to the non-exposed category in our study. So, the non-exposed areas consist of coastal interior and non-coastal areas. We depicted the exposed coast and non-exposed areas in Fig. 3 using freely available QGIS (version 2.8.5-Wien) software (https://qgis.org/en/site/ or https://qgis.org/downloads/).
The explanatory variables, religion and region by adjusting other control variables, demographic (such as age, sex, education, and marital status), healthcare provider (being categorized as public, private, pharmacy/dispensary, traditional, and other), wealth quintile for economic status (grouped into poorest: lowest 20%, poorer: 2nd quintile, middle: 3rd quintile, richer: 4th quintile, and richest: upper 20%), and employment/earning status (Yes/No) were used for predicting two burdens: chronic illness with a disability and OOP healthcare expenditure. Respondent’s age was categorized as childhood (≤ 19 years), young adulthood (20–39 years), middle-aged (40–64 years), senior-aged (65–84 years), and old senior-aged (≥ 85 years) [38]. We divided marital status into three groups: married, unmarried, and others (widowed, divorced, or separated). Likewise, Educational level was grouped as no education, primary, secondary, higher secondary, and higher education. The list of predictors of OOPE included one extra variable that describes whether respondents were currently enrolled or received any assistance from any SSNP.
Statistical analysis
Data on the different variables were summarized using descriptive statistics. We also tabulated chronic illness by disability status in frequency count and percentage. Normality of the OOPE was checked using the Kolmogorov-Smirnov test before commencing the analyses of OOPE. Since the distribution of OOPE was non-normal (right-skewed), we summarized OOPE by the median and interquartile range (IQR), and the Wilcoxon-Mann-Whitney test was used to compare OOPE between two groups. We normalized the distribution of right-skewed OOPE data by applying a natural logarithmic transformation to OOPE. Bi-directionally connected two burdens – health and financial – were linked in box plots. Multiple logistic regression analysis was performed to evaluate the association of disability status with religion and region controlling for other explanatory variables, and to calculate odds ratios (ORs) for determining the odds of developing disability among minority group and people from the exposed coast compared to their respective counterparts. Multiple linear regression model was used with the normalized OOPE as the dependent variable to assess the influence of explanatory variables on OOPE. Data processing (Data cleaning, derivation of required variables and creating analysis dataset), validation, and all statistical analyses were performed using the SAS 9.4 software (SAS Institute, Inc., Cary, North Carolina, USA).
(Note: This is an unedited and auto-generated story from Syndicated News feed. PriyaSaha.Com Staff may not have modified or edited the content body. Please visit the Source Website that deserves the credit and responsibility for creating this content.)