Family functionality and its association with non-communicable diseases among urban adults in Selangor, Malaysia: a cross-sectional study

Abstract

Objectives

Non-communicable diseases (NCDs) are rapidly escalating in developing countries and social factors such as the dynamics of the family play an important part in the lifestyle choices that lead to the onset and maintenance of chronic illness. There remains a gap in Malaysia as the majority of the studies were focused on the normal population rather than directly towards persons having NCDs. This study aimed to examine emerging risk factors such as family functionality and its association with NCD.

Design

A cross-sectional survey was conducted using a multistage random sampling method.

Setting

Urban residential areas in Selangor, Malaysia.

Participants

A total of 2542 adults residing in urban areas of Selangor were recruited.

Primary and secondary outcome measures

Family functionality was measured using the APGAR (Adaptation, Participation, Gain or Growth, Affection and Resources) scale and multiple logistic regression was performed to measure the association between emerging risk factors and NCD.

Results

The prevalence of diabetes mellitus and hypertension was 10.8% and 6.1%, respectively. Widowed/separated status (adjusted OR (AOR) 41.53, 95% CI 19.06 to 90.48, p value=0.001) was reported to be a predictor of diabetes. As for hypertension, familial functionality (AOR 4.2, 95% CI 1.11 to 14.50, p value<0.05) was found to be a significant predictor.

Conclusions

There is a growing concern that family functionality is an emerging risk factor for NCDs. Future family-centred health promotion programmes should be incorporated to improve self-management behaviours and health outcomes.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• A multistage random sampling design was adopted with a substantial number of participants (n=2542) to increase statistical power and to ensure that the sample represents the target population.

• The validated family APGAR scale, which assesses five domains (Adaptability, Partnership, Growth, Affection and Resolve), was used to evaluate family functionality through a standardised scoring system.

• Multiple logistic regression was conducted, which allows for a rigorous analysis of the associations between independent variables and the outcome while effectively controlling for potential confounders.

• Self-administered questionnaire may introduce response bias, though it may have been minimised through clear instructions, validated questionnaire items and anonymity.

• The nature of cross-sectional studies limits the ability to observe changes or trends over time; therefore, causality cannot be established

Introduction

Non-communicable disease (NCD) is a non-infectious health condition that is not spread through personal contact, biological vectors, vehicles or airborne transmission.¹ Approximately 75% of the total global individual NCD cases were recorded in developing countries, and it is believed that this is because of industrialisation such as sedentary lifestyles, poor nutrition, cigarette smoking and risky alcohol consumption.² In Malaysia, NCDs are the leading cause of morbidity and premature mortality, with ischaemic heart disease (17.0%), pneumonia (11.4%) and cerebrovascular disease (8.3%) as the leading causes of mortality among Malaysians in 2020.³ According to the National Health and Morbidity Survey (NHMS),⁴ the prevalence of diabetes increased by 4.9% from 2015 to 2019 and the national prevalence of hypertension among Malaysian adults was 30.3% with rates increasing with age.⁵

Social factors such as the dynamics of the family and health-seeking behaviour (HSB) play an important part in the lifestyle choices that lead to the onset and maintenance of chronic illness.⁶ This is because families foster motivation, effectiveness and social support to address behavioural risk factors for NCDs (eg, physical activity, nutrition, smoking cessation, alcohol and stress management). Family function refers to the social and structural properties of the global family environment, which includes cohesion (ie, the degree of commitment and support that family members provide for one another), expressiveness (ie, the extent to which family members are encouraged to express their feelings) and conflict (ie, amount of openly expressed anger and conflict among family members).⁷ Individual health develops in the context of family whereby the characteristics of the family and household setting are the place where many health behaviours, either positive or negative, are developed.⁸ Based on a previous study, individuals living in a dysfunctional living environment and a single parent structure develop poor and unhealthy eating habits that could lead to becoming overweight and obese, which could influence the onset and development of arterial hypertension and other obesity-related diseases like diabetes.⁹ Moreover, persistent and negative social exchanges within the family unit have harmful consequences to health such as an increased risk for chronic illnesses (eg, cardiovascular disease), while those with a well-functioning family are associated with better physical health.¹⁰ This was supported in a previous study conducted in an Asian population, whereby poorer family function was linked with severe nicotine dependence and frequent alcohol consumption which is a recognised modifiable risk factor for many NCDs.¹¹

Apart from family function, HSBs (eg, initiating care at the right time, with the right provider; maintaining regularity of care seeking) play significant roles in decision-making and have an impact on the preferences and practices of patients, which decide whether individuals will make the best use of medical treatments when they are accessible.¹² According to a systematic review, people having NCDs with strong family support had demonstrated good HSB such as initiating early screening, treatment and follow-up which could potentially reduce the risk of premature death.¹³ Finally, psychological distress had also been associated with NCDs; however, people living with NCDs may not be diagnosed with any mental disorder but may still be having symptoms of psychological distress. Previous evidence has shown that the prevalence of hypertension and diabetes is relatively high among people with psychological distress, which suggests a call for action to improve psychiatric screening in patients with NCDs.¹⁴

Although previous evidence has reported the association between family function and NCDs, there remains a gap in Malaysia as the majority of the studies were focused on the normal population rather than directly towards persons having NCDs.¹⁵ Furthermore, how family functions act as a precursor in the chronic illness population is still not fully well understood in developing countries, but is important to ensuring the best possible outcome for people with NCDs.¹⁶ Therefore, understanding the association between family function and other emerging factors is critical to developing an impactful family-centred intervention for people with NCDs. This population-based study aims to fill a gap in the literature by determining the emerging social, psychological and behavioural factors associated with NCD in an urban setting.

Methods

Study design, setting and population

This study was a cross-sectional design that was conducted among residents of Section 36, Shah Alam. Section 36 of Shah Alam is a diverse neighbourhood with residents belonging to the B40 (the bottom 40%, households earning below Malaysian ringgit RM 5249), M40 (middle 40%, households earning between RM5250 and RM11 819) and T20 (top 20%, households earning above RM11 819 monthly) income levels. Compared with the overall Malaysian adult population, Section 36 of Shah Alam has a similar ethnic composition and age distribution. Inclusion criteria included Malaysians or permanent residents who had resided in the area over the past 1 year, aged 18 years old and above and were able to read and communicate in Malay or English. Individuals who were bedridden or physically dependent were excluded from the study. Data was collected from 8 August 2022 to 15 August 2022.

Sample size determination and sampling method

A minimum sample size was statistically determined using the Epi Info software by inserting the following parameters; an estimated population size of 50 000 based on prior local surveys and municipal planning documents, as precise section-level census data were unavailable, 90% power, a precision level of 0.05, CI of 95%, design effect of 2 and the national prevalence of hypertension (30.3%).⁵ The final total minimum required sample size for this study was 2478 respondents, which includes an additional 30% non-response rate. The study participants were chosen by employing the multistage random selection method. During the first phase of sampling, two locations within section 36 of Shah Alam were chosen through the random sampling method. During the second stage, the selection of every housing unit was conducted by simple random sampling which was facilitated by a computer-based random number generator. Households were visited during the daytime on both weekdays and weekends to increase the chance of reaching eligible participants. Individuals within the selected household who fulfilled the inclusion criteria were provided with a description of the objectives of the study and invited to provide their written consent to continue with the completion of the online self-administered questionnaire using Google Forms that was automatically recorded in Google Spreadsheets. For single-member households, the head of household completed the questionnaire. If multiple members (eg, four to five individuals) met the criteria, all were included in the survey and answered the questionnaire individually to minimise bias. The data cleaning process was carried out carefully, as it could substantially affect the study’s final statistical results. Before beginning the data cleaning procedure, a copy of the data was made, and when the data were transferred from Excel file to the SPSS V.27 software, labels and values were added. Then, the data was cleaned and inspected for missing, duplicated or inaccurate values. Descriptive statistics were used to find the deviating value and filter the data for missing values. These values included frequency distribution, minimum and maximum values. Eventually, the data set were accurate and ready to be analysed.

Study instrument

Our data for the study was collected through a set of online questionnaires. The questionnaires were adapted from numerous validated questionnaires which consisted of six components (part A to part F). Part A concentrated on the sociodemographic information, part B assessed physical activity in the last 6 months,¹⁷ part C measured the nutrition status which included self-reported information on the respondents’ weight, height and waist circumference and part D assessed family function using the Adaptation, Participation, Gain or Growth, Affection and Resources (APGAR) scale and HSB. The psychometric properties of the family APGAR scale are well-established and a validated instrument for assessing perceived family functioning. Factor analysis studies consistently support a unidimensional structure, and internal consistency has been reported as good to excellent, with Cronbach’s alpha ranging from 0.81 to 0.99 across diverse adult populations. Where reported, test–retest reliability has also been high, supporting the stability of the scale over time.1218,20 The family APGAR scale consisted of a five-question evaluation tool encompassing adaptation, partnership, growth, affection and resolve. The total score ranged from 0 to 10. A good family function had a score of 7~10, moderate dysfunction of 4~6 and severe dysfunction of 0~3.²¹ The HSB questionnaire assessed whether the participant had ever attempted to treat his illness. Each item had options for answers ranging from 2 to 5, and they were scored accordingly. The total score for this domain was 15. Scores were summed up, and participants were given scores based on percentages. A score of 12 or more, which is equivalent to 80% or more, indicated appropriate treatment-seeking behaviour (TSB), and a score of less than 12, or equivalent to less than 80%, indicated inappropriate.¹⁵ Part E included smoking history and alcohol consumption,²² and lastly, part F, which entails the General M-health questionnaire (GHQ-12) that measured psychological distress and is categorised as low mental distress and high mental distress.²³ All measurement tools used in this study were administered in English and have established validity. The family APGAR and GHQ-12 have been widely validated in English, including in Malaysian populations. The Global Adult Tobacco Survey uses a standardised, internationally validated questionnaire, and the English version was used with translation and pretesting procedures ensuring appropriateness. The health-seeking behaviour items adapted from Arumugam et al were validated in English through expert review and pilot testing.1218,23

Patient and public involvement

Participants were not involved in the design, recruitment and conduct of the study. The results were disseminated to study participants via the city council representatives.

Statistical methods

Data entry and analysis were performed using the SPSS V.27.0. Descriptive analysis was used to analyse the data, with frequency and percentage reported for categorical data and mean and SD reported for numerical data. In the inferential analysis, simple and multiple logistic regression analyses were conducted to examine the crude and adjusted relationship between independent and dependent variables. Then, independent variables with a p value of ≤0.3 simple logistic regression were included into adjusted logistic regression analysis to find out the predictors of the outcome, which were represented using the adjusted OR (AOR) and 95% CI. Hosmer-Lemeshow goodness-of-fit test was checked for the model fitness and a p value of <0.05 is taken as an indication of poor fit. In addition, multicollinearity was tested using variance inflation factors (VIF). At a 95% CI, a p value of 0.05 was regarded as significant.

Results

A total of 2542 participants completed the questionnaire on NCD and their associated factors. The prevalence of self-reported NCDs in our study population was 10.8% for self-reported diabetes mellitus and 6.1% for hypertension. The figures are different with national figures (18% and 30%) due to differences in population characteristics. The current study consists of younger respondents, and both conditions increase significantly with age. The sample also came from a single urban area with relatively better healthcare access, which may contribute to lower disease burden.

Table 1 shows the characteristics of the study participants.

Table 1. The characteristics of study participants (n=2542).

Variable	Frequency	%
Gender
Male	1102	43.4
Female	1440	56.6
Age*
Young adults (18–35 years)	1449	57.0
Middle age (36–55 years)	672	26.4
Older adulthood (≥56 years)	421	16.6
Marital status
Single	1133	44.6
Married	1299	51.1
Widow, separate, others	110	4.3
Race
Malay	2334	91.8
Other	208	8.2
Employment
Student	578	22.7
Unemployed	679	26.7
Employed	1285	50.6
Education level
SPTM and less	1407	55.4
Bachelor and more	1135	44.6
Income (n=1074)†
Low: B40 (≤RM4850)	649	60.4
Middle: M40 (≤RM10 970)	276	25.7
High: T20 (>RM10 970)	149	13.8
Number of people living with you (n=2538)
1	90	3.5
2–4	715	28.1
≥5	1733	68.2
Body mass index (n=2520)
Underweight (<18.5)	278	11.0
Normal (18.5–24.99)	1081	42.9
Overweight (25–29.99)	660	26.2
Obese (≥30)	501	19.9
Physical activity in the last 6 months (n=2237)
No	453	20.3
2–3 times	1101	49.2
>3 times	683	30.5
Smoking status (n=2535)
No	2017	79.6
Yes	351	13.8
Former smoker	167	6.6
Alcohol consumption
No	2488	97.9
Occasional (once per month)	26	1.0
Regular (more than once per month)	28	1.1
Adaptation family function (APGAR)
Dysfunctional (0–3)	177	7.0
Moderately dysfunctional (4–7)	550	21.6
Highly functional (8–10)	1815	71.4
General Health Questionnaire (GHQ-12)
Low mental distress (≤13)	1427	56.1
High mental distress (≥14)	1115	43.9
Health-seeking behaviour (n=1962)
inappropriate (0–12)	1267	64.6
Appropriate (≥13)	695	53.4
Waist circumference (cm)‡
Mean and SD	72.6, SD=24.13
Minimum and maximum	21–170

^*Reference: Petry48

^†Department of Statistics’ (DOSM)49

^‡Optimal cut-off values of WCs for men and women are different. However, in this study, we used the mean value to represent the WCs in both genders.

APGAR, Adaptation, Participation, Gain or Growth, Affection and Resources; DOSM, Department of Statistics Malaysia; RM, Malaysian ringgit; STPM, Sijil Tinggi Persekolahan Malaysia (Malaysian Higher School Certificate); WC, waist circumference.

Table 2 shows that four factors are found to be the predictors of diabetes, namely, gender, marital status, employment and waist circumference. VIF is less than five for all variables in the model, which means that there is no interaction among the included variables. Regarding gender, males are two times more at risk of diabetes in comparison to females (OR 2.3, 95% CI 1.50 to 3.65, p value<0.001). In addition, married participants and widow/separated participants are six times (OR 6.16, 95% CI 3.15 to 12.02, p value<0.001) and 15 times (OR 15.4, 95% CI 6.77 to 35.35, p value<0.001) at higher risk of diabetes, respectively, in comparison to unmarried respondents. Also, as the waist circumference increased, the probability of being diabetic increased (OR 1.01, 95% CI 1.002 to 1.019, p value<0.05).

Table 2. Determinants of diabetes mellitus: multivariate logistic regression (n=2542).

Variable*	P value	AOR	95% CI
			Lower	Upper
Gender
Male	<0.001†	2.347	1.508	3.653
Female (Ref.)
Marital status
Single (Ref)
Married	<0.001	6.161	3.156	12.027
Widow, separate, others	<0.001	15.471	6.770	35.355
Employment
Student	0.080	0.263	0.059	1.175
Unemployed	<0.001†	4.294	2.975	6.198
Employed (Ref.)
Waist circumference (cm)‡	0.014	1.011	1.002	1.019

Assumptions of logistic regression have been met and the Hosmer-Lemeshow goodness-of-fit test indicated good fit (p=0.709); VIFs are less than five for all variables included in this model.

^*Adjusted for age, race, education, smoking status, physical activity, alcohol consumption, general health and family function.

^†Significant at p<0.05.

^‡Optimal cut-off values of WCs for men and women are different. However, in this study, we used the mean value to represent the WCs in both genders.

AOR, adjusted ORs; VIF, variance inflation factor; WC, waist circumference.

Table 3 shows that three factors are found to be the predictors of hypertension, namely, age, physical activity and family function. VIF is less than five for all variables in the model, which means that there is no interaction among the included variables. Regarding age, older adults aged ≥56 years are four times more at risk of hypertension in comparison to young adults (OR 3.80, 95% CI 1.92 to 7.51, p value<0.001). On the other hand, those who never exercised in the last 6 months are two times more at risk of hypertension compared with those who exercised more than three times per week in the last 6 months (OR 2.02, 95% CI 1.01 to 4.06, p value<0.05). Regarding adaptation of family function, those with dysfunctional families are four times more likely to have hypertension compared with those with highly functional families (OR 4.2, 95% CI 1.11 to 14.50, p value<0.05).

Table 3. Determinants of hypertension: multivariate logistic regression (n=2542).

Variable*	P value	AOR	95% CI
			Lower	Upper
Age
Young adults (18–35 years) (Ref.)
Middle age (36–55 years)	0.232	1.517	0.766	3.003
Older adulthood (≥56 years)	0.000†	3.802	1.923	7.517
Physical activity/week in the last 6 months (n=2237)
No	0.046†	2.028	1.013	4.063
2–3 times	0.262	1.418	0.771	2.609
>3 times (ref)
Adaptation family function (APGAR)
Dysfunctional (0–3)	0.033†	4.024	1.116	14.505
Moderately dysfunctional (4–7)	0.927	1.033	0.511	2.088
Highly functional (8–10) (Ref.)

Assumptions of logistic regression have been met and the Hosmer-Lemeshow goodness-of-fit test indicated good fit (p>0.05); VIFs are less than five for all variables included in this model.

^*Adjusted for gender, marital status, income, GH and HSB.

^†Significant at p<0.05.

AOR, adjusted OR; APGAR, Adaptation, Participation, Gain or Growth, Affection and Resources; GH, general health; HSB, health-seeking behaviour; VIF, variance inflation factor.

Discussion

The prevalence of self-reported NCDs in our study population was 10.8% for self-reported diabetes mellitus and 6.1% for hypertension. In contrast, according to national data in 2019, the prevalence of hypertension was 30.0% and diabetes was 18.3%.²⁴ The figures are different with national figures national figures due to differences in population characteristics. The current study consists of younger respondents, and both conditions increase significantly with age. The sample also came from a single urban area with relatively better healthcare access, which may contribute to lower disease burden. The prevalence of obesity in the present study (19.9%) was comparable to the national estimate of 19.7% reported in the NHMS 2019. However, the proportion of overweight individuals (26.2%) was slightly lower than the national prevalence of approximately 30–30.4% reported during the same period, when using the body mass index (BMI cut-off of 25–29.9 kg/m². When combined, the proportion of overweight and obese individuals in this study (46.1%) was marginally lower than the national figures of 50.1% in 2019 and 54.4% in 2023. In contrast, the proportion of respondents classified as underweight or normal weight (53.9%) was slightly higher compared with the national estimate of about 49.9%, suggesting a somewhat lower burden of excess weight in the study population compared with the general Malaysian adult population.

Family function is one of the emerging factors related to NCD and the significant association of this factor is observed in this study. Family dynamics refers to the patterns of interactions among relatives, their roles and relationships and the various factors that shape their interactions. Family members rely on each other for emotional, physical and economic support.²⁵ A dysfunctional family may have a negative relationship that can cause stress, impact mental health and even cause physical symptoms. Poor family relationships may lead to stress that is known to predispose a person to hypertension, as shown in our study. Dysfunctional families compared with highly functional ones were four times more likely to report hypertension. Not only may dysfunctional families predispose one to hypertension, subsequently, a study in China found that one with existing hypertension would attain a higher health-related quality of life with good family function.²⁶ A dysfunctional family may not be able to provide a good family support as a study has reported that family support is needed for disease education and management of a chronic disease. A functional family may provide a good family support that motivates patients to seek treatments and manage their diseases better.²⁷

The symptoms affecting chronic illness patients not only affect the patient but also the family members, especially their spouses.²⁸ Hence, these findings shed light on the importance of creating awareness among caregivers of the importance of family support, especially in the care of patients with chronic diseases such as hypertension and diabetes. Part of ensuring a normal life for a chronically ill patient is to get their spouses’ involvement in their day-to-day activities.²⁹

Studies have shown that social relationships can profoundly influence well-being across the life course.^{21 30} Being married, especially happily married, is associated with better mental and physical health.^{31 32} Interestingly, in this study, marital status is not a significant factor for hypertension, whereas marital status is an influential factor for diabetes. Specifically, this study depicts that being married had 15 times higher odds of diabetes than the unmarried respondents. Although dysfunctional families may increase stress, stress affects blood sugar more directly than blood pressure by changing hormone levels and reducing the body’s ability to use insulin properly.33,36 Similarly, marital status may influence diabetes through social support, health-seeking behaviour and adherence to preventive care, while its effect on blood pressure is less consistent.³⁷ These mechanisms may explain the differential associations observed in our study. The higher prevalence of diabetes among married respondents may reflect shared lifestyle behaviours within Malaysian households. Married adults in Malaysia tend to share eating patterns, daily routines and household environments that influence metabolic risk, resulting in stronger spousal concordance for diabetes than for hypertension.^{38 39} Nevertheless, the odds for diabetes were 41 times greater for those being divorced and widowed in comparison to those being unmarried. This shows that the married ones were more predisposed to attaining diabetes than the unmarried ones, yet for those who became widowed or separated, their odds of getting diabetes were higher than the married ones. There are studies that correlate psychological distress with divorce and separation.^{40 41} The ones found to be divorced in this study were significantly 41 times higher to develop diabetes, and this finding is akin to a study in Brazil which concludes that those who remain in their marriage were less likely to develop diabetes than those who were divorced.⁴¹ This is consistent with the Poland study that showed a significantly higher percentage of patients was observed in non-married men (unmarried, divorced, widowed) as compared with married men.⁴²

In the light of social causation, those who are married have less stress and anxiety as marriages accentuate health-protective effects and promote healthy behaviour, whereas those unmarried or widowed are prone to adverse health effects,^{43 44} which may also explain the results we obtain in this study. On the contrary, a longitudinal study conducted in Iran found a decreased risk in type 2 diabetes (T2D) among the widowers in comparison to the married ones.⁴⁵ It was interesting to note, as they further discuss the possible reason for these findings may be due to family functional factors whereby married women usually sacrifice themselves to care for their husband as they grow older, which has health implications on themselves. Thus, a consequence of a divorce or death of their husband may alleviate their stress burden,³⁵ resulting in possible findings of lower T2D in that Iran study, which nevertheless needs further study to acquire the means of such association.⁴⁶ Similarly, further research is needed to depict the real mechanism for the association of family function in regards to possible cause attributing to 41 higher odds of developing diabetes among the unmarried and widowers in our study. Furthermore, unemployment has a significant influential role in diabetes. This may be linked to physical inactivity and overeating among unemployed respondents. A study revealed that retirement will increase weight and BMI, especially among men. This is because the transition to retirement tends to reduce the probability of engaging in vigorous activities and the tendency to eat more than they did before they retired. Being unemployed may have similarities with a retiree due to these unhealthy risk behaviours, thus increasing the prevalence of NCDs.

Regarding gender, men had higher odds of having diabetes compared with women. This is following a study in Poland among adult residents, where the prevalence of diabetes showed a higher percentage in men (6.6%) compared with women (3.5%).³⁷ On the other hand, individuals in older age groups (≥56 years old) had higher odds of having hypertension. This shows that individuals in older age groups are more commonly to be diagnosed with NCDs due to the presence of risk factors and advancing age. Furthermore, the prevalence of having major risk factors of NCDs was also greater among older persons.^{47 47}

There are several limitations to the study. The findings were confined to urban residents and may not be fully generalisable to adults living in rural areas. The diagnosis of NCD was self-reported and unable to check the validity of the diagnosis. Recall bias is inevitable while collecting the primary data. Further research is recommended to gain insight into the family dynamics mechanisms and their influence on the development of NCD. The family-based intervention targeting shared behaviours like physical activity and eating patterns is recommended for the prevention and management of chronic diseases. These modifiable lifestyle behaviours are shaped by the family dynamics where a family member may influence the action of other family members.

Conclusion

According to this study, there is a growing concern that family functionality is an emerging risk factor for NCDs. As a result, future implementation of health promotion programmes with a multisectoral approach should incorporate family function as one of the fundamental concepts that may strengthen patients’ and families’ ability to deal with problems and maintain self-management of chronic conditions.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.