Data for this study were derived from a large cross-sectional survey on informal caregiving, health, and healthcare among caregivers (≥ 18 years) of older adults aged 50 years or more. These informal caregivers resided in 13 districts, made up of 18 rural and 21 urban communities, in the Ashanti Region of Ghana (see Fig. 1). Communities and participants in the survey were recruited through probability (simple random and cluster sampling) and non-probability sampling techniques (snowball sampling) respectively. In all, the analytical sample of the survey was 1853 informal caregivers. Details of the sample size calculation and sampling process are found in Agyemang-Duah and Rosenberg [38]. The main data collection instrument was interviewer-administered questionnaire. All the questions captured in the questionnaire were later transferred to Qualtrics, an e-survey tool, to digitally record the responses of the participants. Previous research on the methods of the survey has been reported elsewhere [38].

Study area location. A Shows the study area covered by the selected districts, and B shows the study area in the context of Ghana

First, before the fieldwork, we obtained approval to the study site from the Ashanti Regional Health Directorate under the Ghana Health Service (Ref: GHS/ASH/RES/V.2). Second, ethical approval was obtained from the Queen's University General Research Ethics Board (GREB), Kingston, Canada (Ref: GGEOPL-344-22) and the Committee on Human Research Publication and Ethics (CHRPE), School of Medical Sciences, College of Health Sciences, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana (Ref: CHRPE/AP/182/22). Third, both verbal and written informed consents were obtained from the study participants. We declare that the procedure for obtaining the verbal informed consent was approved by the ethics committee/institutional review board. Whereas informed consent for the literate participants was obtained from themselves, that of the illiterate participants was obtained from their legal guardians which was approved by the institutional review board. The ethics protocol emphasized that participation was voluntary and that participants had the right to withdraw from the survey before, during and even after the completion of the survey.

In this study, health-related quality of life represents our dependent variable. The Health Survey scale of the health-related quality of life which consists of eight items (SF-8 Health Survey scale), is the shorter version of the 36-item Health Survey scale measuring health-related quality of life [39,40,41,42,43], was employed in this study. The SF-8 Health Survey scale is a general multipurpose short-form health-related quality of life instrument which was developed by the RAND Corporation and the Medical Outcomes Study (MOS) [44]. The SF-8 Health Survey scale has widely been used for health-related quality of life assessment in international health research [39,40,41, 45,46,47].

The SF-8 Health Survey scale focuses on 8 domains/dimensions to measure health-related quality of life [48, 49]. These dimensions are physical functioning, physical role functioning, bodily pain, general health, vitality, social functioning, emotional role functioning and mental health [39, 41, 42, 49]. Each domain of the health-related quality of life is linked to a single item on the SF-8 Health Survey scale [40, 41]. These items are as follows: (a) Overall, how would you rate your health during the past 4 weeks? The responses are as follows: (1) very poor, (2) poor, (3) fair, (4) good, (5) very good, and (6) excellent; (b) during the past 4 weeks, how much did physical health problems limit your usual physical activities? The responses are as follows: (1) could not do physical activities, (2) quite a lot, (3) somewhat, (4) very little, and (5) not at all; (c) during the past 4 weeks, how much difficulty did you have in doing your daily work both at home and away from home because of your physical health? The responses are as follows: (1) could not do daily work, (2) quite a lot, (3) somewhat, (4) a little bit, and (5) none at all; (d) how much bodily pain have you had during the past 4 weeks? The responses are as follows: (1) very severe, (2) severe, (3) moderate, (4) mild, (5) very mild, and (6) none; (e) During the past 4 weeks, how much energy did you have? The responses are as follows: (1) none, (2) a little, (3) some, (4) quite a lot, and (5) very much; (f) during the past 4 weeks, how much did your physical health or emotional problems limit your usual social activities with family or friends? The answers are as follows: (1) could not do social activities, (2) quite a lot, (3) somewhat, (4) very little, (5) not at all; (g) during the past 4 weeks, how much have you been bothered by emotional problems? The responses are as follows: (1) extremely, (2) quite a lot, (3) moderately, (4) slightly, and (5) not at all; (h) during the past 4 weeks, how much did personal or emotional problems keep you from doing your usual work, school, or other daily activities? The responses are  as follows: (1) could not do daily activities, (2) quite a lot, (3) somewhat, (4) very little, and (5) not at all. The scale ranges from 8 to 42 with higher score demonstrating higher health-related quality of life.

Based on previous studies, the psychometric properties of the scale such as Cronbach alpha are generally very high [39, 42, 43, 46, 50, 51]. For instance, Wirtz et al. [39] reported a Cronbach alpha value of 0.918 for all items under the SF-8 Health Survey scale. In this study, the Cronbach alpha for all the items was 0.932 showing strong internal consistency.

The focal independent variable of this study is caregiver burden. The World Health Organization Impact of Caregiving Scale [World Health Organization cited in 24] was used to assess caregiver burden. The scale is made up of 10 items with a 5-point Likert scale ranging from: 1 = None, 2 = Mild, 3 = Moderate, 4 = Severe, 5 = Extreme. The 10 items scale sought to find out from informal caregivers who have provided care for at least a year [24], if caregiving results in the following: 1 = “difficulty getting enough sleep”, 2 = “problem getting enough food to eat”, 3 = “not enough energy for extra work”, 4 = “do not know the correct care to provide for health problems of care recipients”, 5 = “cannot take care of health, ailment/chronic condition”, 6 = “unable to pay for medication/treatment for ailment/chronic condition alone”, 7 = “cannot visit friends and relatives as much as before”, 8 = “cannot share feelings about caregiving responsibility with others”, 9 = “experienced financial problems due to loss of income”, 10 = “experienced stigma or problems as a result of the care recipient's illness or death”. Considering this, we built a composite score which ranged from 10 to 50 with higher score showing higher caregiver burden. A Cronbach alpha value of 0.881 was reported indicating strong internal consistency.

Socio-ecological (which is considered here as socio-economic), cultural, demographic, and healthcare factors were controlled for. Specifically, age (years) (0 = 18–24, 1 = 25–34, 2 = 35–44, 3 = 45–54, 4 = 55–64, 5 = 65 or above), gender (0 = male, 1 = female), ethnicity (0 = Akan, 1 = non-Akan), religion (0 = Christianity, 1 = Islam, 2 = African traditional religion, 3 = no religion), living arrangement (that is whether the participants are living with the care recipient or not) (0 = no, 1 = yes), health insurance enrollment (0 = no, 1 = yes), place of residence (0 = rural, 1 = urban), marital status (0 = never married, 1 = currently married, 2 = separated/widowed/divorced), monthly income level (GH¢) (0 = less than 1000 [US$99.50 as at the time of the field survey, September 2022], 1 = 1000–1999, 2 = 2000 or above), education level (0 = no formal education, 1 = primary education, 2 = junior high school education, 3 = senior high school education, 4 = tertiary education), employment status (0 = unemployed, 1 = employed) and healthcare (0 = none, 1 = once, 2 = 2 times, 3 = 3 times, 4 = 4 times 5 = 5 or more times) of caregivers. All the control variables were considered as categorical. When we performed multi-collinearity analysis of the independent and control variables, we had a variance inflation factor less than 1.5 which suggests no strong multi-collinearity.

With the aid of the SPSS software version 28 (IBM, Armonk, NY), we employed descriptive (such as mean, standard deviation, percentages, and frequencies) and inferential analytical techniques (such as Generalized Linear Models) to analyze our data. The sample characteristics of the participants, caregiver burden and health-related quality of life were analyzed using frequencies, percentages, mean and standard deviations. Generalized Linear Models were employed to establish association between caregiver burden, other control variables (such as age, gender, ethnicity, religion, health insurance enrollment, place of residence, marital status, income level, education level, employment status and healthcare utilization) and health-related quality of life. We employed Generalized Linear Models because the dependent variable (that is health-related quality of life) was continuous. In all, five models were fitted. Model 1 determined the association between caregiver burden and health-related quality of life. Model 2 included variable in Model 1 plus socio-economic variables. Model 3 added cultural variables plus all variables in Model 2. Model 4 included demographic factors and all variables in Model 3. The final Model (Model 5) consisted of healthcare variable and all variables in Model 4. We reported beta values and standard errors with a significance level of 0.05 or less.