The association of health behavioral risk factors with quality of life in northern Sweden—A cross‐sectional survey

Abstract Background It is well known that behavioral risk factors such as obesity, smoking, physical activity, diet, and excessive alcohol are linked to general health in northern Sweden. This study aimed to explore the joint relationship between these risk factors and the quality of life (QoL). Methods Data were collected from Sweden's national public health survey between February and May 2014 in the four northern counties in Sweden. QoL was assessed using the EuroQol (EQ‐5D). Multivariable regression analysis was used to examine the relationship between five risk factors: BMI, physical activity, smoking status, fruit and vegetable intake, and alcohol consumption and QoL. Results Data from 17 138 complete questionnaires showed that individuals who were not obese, did at least 30 minutes of physical activity daily, consumed at least 3 portions of vegetable or fruits, were not smoking daily, and who did not report being drunk at least once every week were found to have better QoL (P < .005). The mean EQ‐5D score ranged from 0.85 to 0.79. Approximately, two thirds of the studied population reported being physically active for at least 30 minutes every day and two fifths of them had a normal BMI. Only around 7% of the sample reported that they were eating the recommended daily level of fruits and vegetables. Conclusions The results of the study suggest that QoL has a significant relationship with lifestyle behaviors. This finding would emphasize the role of interventions to improve population health.

Many of the survey questions originated from a "Survey of Living Conditions" that was done in Stockholm, Skåne, and northern counties and verified in a pilot study in 2003. The survey questions cover physical and mental health, consumption of pharmaceuticals, contact with healthcare services, dental health, living habits, financial conditions, work and occupation, work environment, safety, and social relationships. We followed the STROBE statement guidance in reporting the HLV survey results 20 .
A stratified random sampling (by age, sex, and municipality) was used in the survey, among all people with the age between 16 and 84 years who were residents in the four northmost counties of Sweden (Västerbotten, Norrbotten, Västernorrland, and Jämtland).
The questionnaires were sent out by regular mail, and the participants were asked to answer the questions and send their answers back to Statistics Sweden. The respondents were informed that the survey was voluntary; it was also possible to answer on the Internet.
They received login information in the form of a user name and a password and could then log in via Statistics Sweden's website. On the Web, there was also the possibility to answer the questions in English or Finnish. Checking that the right person has answered the questionnaire has been done by comparing answers to questions about birth year and gender with the corresponding record.

| Quality of Life (QoL)
EuroQol (EQ-5D) was used to measure QoL [21][22][23] . It provides a simple and generic measure of health, and it provides a descriptive profile and a single numeric value for the state of the individual health that can be used in different settings. It contains two parts, one is the descriptive system and the other is the visual analogue scale. The respondent is asked to self-rate his current health state as well as mobility, self-care, usual activities, pain/discomfort, and anxiety/depression.

| Body mass index (BMI)
The respondents self-reported their height and weight, and they were categorized in four groups according to their BMI: underweight, when the BMI <18.5 kg/m 2 ; normal weight, when BMI is between 18.5 and 25 kg/m 2 ; overweight, when it is between 25 and 30 kg/m 2 ; or obese if the value is over 30 kg/m 224 . The normal weight group was selected as the reference group in the analysis.

| Alcohol consumption
The participants were asked in the survey "How often during the past 12 months have you drunk so much alcohol that you have become drunk?" and the answers for that question were used to categorize the participants into two groups, the first group represent those who were consuming alcohol to the extent that they were feeling drunk at least once every week, and the other group involved the participants who report being drunk less frequently. These two groups were used in the study to analyze the association of binge drinking on the EQ-5D.

| Smoking habits
The participants were asked whether they were smoking and/or using snus daily or not, and the participants were grouped into two groups, the first group was those who replied that they were not smoking and were not using snus daily, and the other group was those who replied that they were smoking and/or using snus daily.

| Fruit and vegetable consumption
The participants were divided into three groups according to their total daily fruit and vegetable consumption: The first group was those who were eating <2 portions per day, the second group was those who were eating 2-4 portions per day, and the last group was those who were eating a total of more than 4 portions per day. The last group was used as the reference group in the analysis.

| Sociodemographic characteristics
Sex, age, marital/civil status, educational level, occupation and income, and long-term health problems were collected for the study participants to rule out potential confounders in multivariable analyses.
For education, the participants were divided into three groups depending on their education level. The first group was the low education-level group which involved the participants who complete <3 years in upper-secondary education, and the second group was the medium education-level group and it involved the participants who complete at least 3 years in upper-secondary education but <3 years in postsecondary education. The final group was the high education-level group which involved those who complete at least 3 years in postsecondary education. The education-level variable was treated as a categorical variable, and the participants with high education level were used as the reference group in the analysis.
For occupation, the participants were divided into two groups depending on the type of their occupation which was retrieved from register data, the blue-collar worker group and white-collar worker group. The blue-collar worker consists of both the unskilled and the skilled manual workers, while the white collar consists of assistant nonmanual, intermediate nonmanual, professionals, and self-employed. The white-collar group was used as the reference group in the analysis.
For income, the participants were divided into three groups depending on their personal disposable income level, the first group included participants with the lowest third of income, the second group included participants with the middle third of income, and the last group included participants with the highest third of income.
Income was treated as a categorical variable, and the group that included participants with the highest third of income was used as the reference group. The data of income were collected from the income and taxation register.

| Statistical analysis
The statistical analyses were performed with Stata/MP 13.0 software. Multivariable linear regression was used to investigate the relationship between the EQ-5D utility score (dependent variable) and health behavioral risk factors (BMI, alcohol consumption, smoking status, fruit and vegetable intake, and physical activity) and the potential confounders (age, sex, civil status, education level, occupation class, income, and the presence of any chronic health problem).

| RE SULTS
With a response rate of 50%, we had 25 667 returned questionnaires. Questionnaires with missing data were dropped to ensure complete analysis. We analyzed 17 138 (67%) completed questionnaires. Binge drinking and the occupation were the most missed items. Sensitivity analysis by removing binge drinking and occupation variables and repeating the multiple regression analysis did not show a significant difference in results.
Approximately, two thirds of the individuals reported being physically active for at least 30 minutes every day and two fifths of them had a normal BMI (Table 1). Only around 7% of the sample reported that they were eating the recommended daily level of fruits and vegetables. Regarding the tobacco use and alcohol consumption, only around 3% of the sample reported being drunk at least once every week and about one quarter of them were using a daily tobacco.
Women tended to be more in normal weight group, while men tended to be more in the overweight group (Table 2). With regard to physical activity, there was no significant difference between men and women. While more than half of the men tended to eat less than two portions of fruits and vegetables daily, more than half of the women tended to eat between 2 and 4 portions from fruits or vegetables daily. Regarding the daily tobacco use, women reported daily tobacco use more than men however men reported being drunk at least once weekly more than women. TA B L E 1 Descriptive statistics for the EQ-5D index score across health behavioral risk factors and socioeconomic characteristics (n = 17 138) The P values were calculated using the t test for the binary variables and one-way ANOVA for the variables with multiple categories.
The mean EQ-5D utility index score for the total study population was 0.811 with standard deviation of 0.204. Respondents who reported being active at least 30 minutes every day, in normal BMI group, eating the recommended daily level of fruits and vegetables, not smoking or using snus daily, and not report being drunk at least once every week had a significantly higher mean EQ-5D compared with the other group in each category (P < .001).
Comparing the mean EQ-5D across the socioeconomic and sociodemographic variables, it was found to be significantly higher in individuals who were men, those who were within the age group 16-30 years, those who were married or cohabitating, had higher education, had higher income level, had a white-collar occupation, and did not have any long-term health (P < .001). The association of daily physical activity and tobacco use with the EQ-5D was significant (P < .001) across all the three models (Table 3). Being drunk at least once every week was correlated with EQ-5D in Model 2 and Model 3 (P < .05). Obesity was strongly correlated with EQ-5D across all the three models (P < .001). However, there was not a significant difference in the EQ-5D scores in the underweight and overweight groups. Consumption of fruits or vegetables daily was significantly associated with EQ-5D in Model 2 and Model 3.
When linear regression was stratified by sex (Tables 4 and 5), the daily tobacco use, low level of daily physical activity, and being obese remain associated with low EQ-5D (P < .001). Being overweight was significantly associated with low EQ-5D score in women in Model 1 and Model 2, but not in Model 3, while in men overweight was significantly associated with low EQ-5D only in Model 1. Fruit and vegetable consumption was not found to have any association with the EQ-5D in men.

| D ISCUSS I ON
This study analyzed the association between the QoL measured by EQ-5D-3L score and the following five lifestyle risk factors: BMI, physical activity, fruit and vegetable intake, tobacco use, and binge drinking, in 16-to 85-year-old individuals from northern Sweden.
The study results showed that there was a significant association between the EQ-5D score and the study variables (BMI groups, 30 minutes of daily physical activity, daily fruit and vegetable consumption, daily tobacco use, and at least being drunk once a week).
The study found that individuals who were not obese, did at least 30 minutes of physical activity daily, consumed at least 3 portions of vegetable or fruits, were not smoking or using snus daily, and who did not report being drunk at least once every week were found to report better EQ-5D.
The mean EQ-5D utility index score for the study population The study showed that among lifestyle behavioral factors, obesity and low physical activity had the strongest association with low QoL in the participants. The inverse relationship between these two lifestyle factors and the QoL found in this study were consistent with the findings found in previous studies from different parts of the world such as in China, Spain, and the United States 26-28 .
The study found that the association between overweight and low EQ-5D is less than what it was with obesity, and it seemed to be confounded by other variables. These findings were similar to what was found in a study previously done in the general population in the UK where it showed that underweight was not associated with a significant reduction in EQ-5D score and the overweight association with low EQ-5D was becoming nonsignificant when the regression was adjusted to other sociodemographic factors 3 .
As expected, daily tobacco use and the heavy episodic drinking of alcohol were found to be associated with a significant decrease in the QoL. These results are supporting findings in a previous study that tobacco use and heavy alcohol drinking associated with significant reduction in QoL 3,4,11,18 . We opted to use a more rare but severe outcome of a harmful pattern of alcohol use rather than a general measure of total alcohol consumption, since we assessed this to be the most relevant outcome in relation to QoL. The study showed that 2.43% of the surveyed population reported being drunk at least once every week. Further studies are, of course, needed to clarify this correlation.
The study showed that while low daily fruit and vegetable consumption was related to low QoL, it had the lower association compared with the other lifestyle characteristics. This finding was comparable to other studies 3,11,17,29 . Therefore, it is possible that the benefits of "5-a-day" could well be closely related to these other health behaviors.
In the Swedish context, a study has shown that increasing weight above the normal range had a negative effect on the QoL even after the control of the major health risk factors and the sociodemographic characteristics 30 . Another study also conducted in Sweden suggests that obesity, not consuming adequate vegetables, and smoking among young people have independent associations with QoL 8 .
It was noteworthy to find that approximately one in three participants were inactive and that this had a similar independent association with lower EQ-5D than being obese. Given that low levels of physical activity and high BMI, a common finding in national surveys, interventions to promote physical activity, and weight management are urgently needed. A minimally important difference in EQ-5D utility score has previously been estimated at 0.0744; findings of reductions in EQ-5D utility scores of greater than this for obesity and physical inactivity suggest that the estimated differences are clinically important.
The study also found that there are some differences between men and women (Tables 4 and 5). Obesity had a more significant association with low EQ-5D in women compared with men (P < .001).
This could be explained that weight perception and impact on the QoL are different between men and women 31 .

TA B L E 3
Linear regression with EQ-5D regressed on lifestyle risk factors and sociodemographic and socioeconomic characteristics (n = 17 138)   Although this survey provided a detailed assessment of health behavioral risk factors, apart from BMI, the measures were self-reported and therefore open to biases as a consequence of potential under-or overreporting. The response rate to the survey was around 50%, and we excluded 8529 (33%) individuals because of incomplete data. However, sensitivity analysis by removing the most dropped variables and repeating the multiple regression analysis did not show a significant difference in results. The cross-sectional design of the study does not assess the temporal relationship between exposure and outcome. The recall period for most of the survey was 12 months made the recall bias a possible limitation to the study.
The study used registration-linked data to the Swedish personal number that allowed accurate matching of socioeconomic and sociodemographic characteristics. The consistency between our findings and approach to those of comparable studies adds credibility to our study 3,11,14,31 .
The strength of this study was its use of a large cross-sectional public survey. Although the study is based on the 2014 HLY survey, it represented a good representation of the northern Sweden population. Future studies would be recommended to analyze subsequent HLV surveys.
In summary, the results of the study show that QoL measured by EQ-5D-3L has a significant correlation with lifestyle behaviors.
The findings suggest that behavioral risk factors namely obesity, less than the recommended daily level of physical activity, low daily consumption of fruits and vegetables, daily use of tobacco, and the heavy episodic drinking of alcohol may have a negative association with QoL. Our study highlights the importance of tackling these be-

ACK N OWLED G M ENTS
The authors would like to thank all the participants in the health survey.

CO N FL I C T O F INTER E S T S
None.

AUTH O R CO NTR I B UTI O N S
All authors contributed to research design, interpreted the data, and reviewed the drafts of the paper. AA performed the statistical analyses.

E TH I C A L A PPROVA L
The use of the "Health on Equal Terms" survey in the present study was reviewed and approved by the ethical committee at the Regional Ethical Review Board in Umeå (2015/134-31Ö). The participants had received information letters about the survey. Individuals' identifying details have been removed for confidentiality. The necessary informed consents were obtained for the study when required.