Neighbourhood socioeconomic status and pain among older adults—A cross‐sectional study

Pain is associated with falls, disability and a poor quality of life among older adults. It is highly prevalent in many societies, and studies have shown that pain could be preventable or managed more effectively at the population level. However, few studies have investigated who is at higher risk of pain in the general population, which is important for development of effective interventions. The purpose of this study was to investigate, by using nationally representative samples in Sweden, whether neighbourhood socioeconomic status (SES) is associated with pain among older adults after considering other important risk factors.

tion level.However, few studies have investigated who is at higher risk of pain in the general population, which is important for development of effective interventions.The purpose of this study was to investigate, by using nationally representative samples in Sweden, whether neighbourhood socioeconomic status (SES) is associated with pain among older adults after considering other important risk factors.

Methods:
The study used the Statistics on Income and Living Conditions (EU-SILC), which is a nationwide annual survey of the living conditions of residents in Sweden.We used the data of individuals who were over 65 years of age between 2008 and 2013.Multivariable logistic regression was conducted to investigate the association between neighbourhood SES and severe pain.
Results: Those who resided in low SES neighbourhoods had a 30% higher odds of having severe pain than those who resided in high SES neighbourhoods after controlling for individual risk factors, such as the sex, age, individual SES, smoking, exercise habits and body mass index.Exercise was protective against severe pain.Conclusion: Given the high prevalence of pain across populations, interventions targeting geographic areas (such as those in the current study) in combination with individual risk factors could be effective to reduce the burden of pain at the population level.
Significance: Those who reside in neighbourhoods with low SES may have higher risks of pain due to a lack of health-promoting resources as well as psychological stress.Further studies identifying the specific mechanisms behind the association between neighbourhood SES and pain would be useful in order to develop effective interventions.

| INTRODUCTION
Pain is a major health burden on individuals, health systems and societies.In Europe, the prevalence of pain among adults is approximately 20%-40%, and its financial cost to society is estimated to be €200 billion per year (Breivik et al., 2006;Todd et al., 2019;Tracey & Bushnell, 2009).The prevalence of pain among adults over 50 years of age has been increasing in most European countries, including Sweden (Zimmer et al., 2020).The Global Burden of Disease Study also highlights increases in the prevalence of pain while many other diseases, such as acute infectious diseases, have been decreasing over recent decades (Rice et al., 2016;Vos et al., 2015).This is partly due to an increase in the aging population as pain is more common and problematic among older age groups, where it is associated with falls, disability and poor quality of life, as well as greater costs within the healthcare system (Bernfort et al., 2015;Domenichiello & Ramsden, 2019;Leveille et al., 2009;Patel et al., 2013).
Several risk factors have been identified in epidemiological studies of pain, such as old age, female sex, low individual socioeconomic status (SES) and physical inactivity (van Hecke et al., 2013).Furthermore, some studies have shown associations between low neighbourhood SES and pain (Fuentes et al., 2007;Jordan et al., 2008).Associations between neighbourhood environments and other chronic conditions, such as obesity and cardiovascular diseases, have also been studied (Diez Roux et al., 2001).Even though the specific mechanisms are unclear, association studies can be useful in order to target high-risk groups and create health-promoting environments to reduce chronic ill-health.
However, few studies regarding neighbourhood environments and pain have been conducted among an older adult population; previous studies have rather been a mixture of individuals of working ages and those who are retired, and have often been based on samples from clinical settings.It is important to improve the understanding of how neighbourhoods may influence the prevalence of pain in older people because they may be more affected by neighbourhood environments after retiring since they often spend more time in their neighbourhood (Glass & Balfour, 2003;Rachele et al., 2019).Moreover, high age is one of the most significant risk factors for pain (Cohen et al., 2021).
To the best of our knowledge, no previous study has examined the association between neighbourhood SES and pain in a random sample of the entire older population, which will be done in the present study.
The primary aim of this study is to investigate whether neighbourhood SES is associated with pain aside from other individual factors in the older adult population.We hypothesize that older adults living in low SES neighbourhoods may have higher odds of having severe pain.A further aim is to study the trend of pain from 1988 to 2013 in the total sample, by sex, by age group (65-69, 70-74, 75-79 and 80-84 years of age) and by neighbourhood SES (low, medium and high).

| Study participants
The present study is based on the Statistics on Income and Living Conditions (EU-SILC), which is a nationwide annual survey of the living conditions of residents in Sweden.EU-SILC is conducted annually on a simple random sample from the national register of the total population aged 16-84 years.The survey is conducted and administered by Statistics Sweden and the details of the survey protocols have been published elsewhere (Statistics Sweden, 2022).The data can be requested from Statistics Sweden for research and other purposes.The individuals who are invited to participate in the EU-SILC survey are informed how their results will be used prior to their participation in it (Statistics Sweden, 2023).Participation in the survey is regarded as an indirect consent.For the primary aim of this study, we limited the individuals to those who were over 65 years of age between 2008 and 2013.The study individuals were interviewed over the phone by trained interviewers about their living conditions, including questions about pain, leisure time physical activity/exercise, health status, smoking, weight and height.In order to investigate the trend of pain for our secondary aim, data from 1988 to 2005 were also included.During the period between 1988 and 2005, surveys were conducted via faceto-face interviews and from 2006 onwards telephone surveys were used.Data from 2014 and onwards were not used in our study for several reasons.First, the questions regarding pain were changed after 2013.Second, exercise habits were not collected in the EU-SILC survey after 2013.Data for 2006 and 2007 were not used as those years included a transition from face-to-face interviews to telephone interviews.
Small area market statistics (SAMS) were used as neighbourhood units as has been done in previous studies (Sundquist, Malmström, et al., 2004;Sundquist, Winkleby, et al., 2004).Each SAMS unit has approximately 1000 residents, and there are 9617 units in all of Sweden.An index of neighbourhood SES was calculated by SAMS based on the socioeconomic conditions of each area (see details below) and linked with the study individuals based on their residential locations.

| Outcome
Pain was defined based on the answer 'Yes' to any of the following questions: 'Do you suffer from pain, ache in back or hip?; Do you suffer from pain, ache in neck or shoulders?;Do you suffer from pain, ache in arms, hands, legs or feet?' Severe pain was then defined based on the answer 'Severe' to the question: 'Would you say the ailments are severe or mild?'.In the primary analysis, pain was treated as a binary outcome, i.e.Severe versus Mild/ No pain because severe pain is more clinically relevant and is associated with psychological stress, which could be affected by low neighbourhood SES (Brooks Holliday et al., 2019;Davies et al., 2009;Fischer et al., 2016).We conducted a sensitivity analysis by treating pain differently, i.e., Severe/Mild pain versus No pain.

| Exposure
For each SAMS unit where individuals reside, a summary index of neighbourhood deprivation was used to define neighbourhood SES, which included the following four indicators for proportions of residents aged 25-64 years with low income; unemployment; low educational status; and social welfare (Winkleby et al., 2007).Each indicator was normalized based on the mean and standard deviation within each SAMS unit, which were later summed to create the index for each SAMS unit (Gilthorpe, 1995).The index was categorized into the following three groups: (1) Low neighbourhood deprivation (high neighbourhood SES): less than one standard deviation (SD) from the mean; (2) Moderate neighbourhood deprivation: within one SD of the mean; (3) High neighbourhood deprivation (low neighbourhood SES): more than one SD from the mean (Kawakami et al., 2011).We rephrased them for ease of interpretation as follows: (1) high SES = low neighbourhood deprivation, (2) medium SES = moderate neighbourhood deprivation and (3) low SES = high neighbourhood deprivation.

| Covariates
The following variables were included in the analysis: year of survey, sex, age, marital status, family income, education, country of birth, smoking, physical activity and the body mass index (BMI).
Year of survey was categorized into three time points: 2008/2009, 2010/2011 and 2012/2013.Sex was defined as male or female based on data derived from the population register.
Age was categorized into four groups: 65-69, 70-74, 75-79 and 80-84 years.We created categories in order to identify age groups with the highest odds of pain.
Marital status was categorized into two groups: married/cohabiting and single living.
Education was categorized into two groups: <10 years (compulsory school or less) and ≥10 years (at least 1 year of high school).
Both family income and education were categorized in order to identify specific groups with the highest odds of pain.
Country of birth was categorized into two groups: Swedish-born and foreign-born.
Family income, education and country of birth were provided to us by Statistics Sweden (the Swedish Government-owned statistics bureau).
Smoking was categorized into two groups: never smokers and former or current smokers.
Exercise was categorized into two groups based on the answer to the survey question 'Do you exercise without interruption for at least 30 min?'.The answers were 'Basically never or a little now and then' and 'About once a week or more'.
BMI was calculated based on the self-reported weight and height and included as a continuous variable because including it as a categorical variable caused problems in the model fit.The linearity assumption between BMI and log odds was tested and found to be violated (p < 0.001, a j-shape curve was observed; see Figure S1) and thus a quadratic term for BMI was added to the analysis.

| Statistical analyses
A total of 11,685 individuals were included in the pooled analysis between 2008 and 2013.As the proportions of missing, non-response and outliers were small, these individuals were excluded (Figure 1).Descriptive statistics were run to show the distribution of individual characteristics by the total sample and by neighbourhood SES for the pooled data between 2008 and 2013.The prevalence of severe pain from 1988 to 2013 was calculated by sex, age and neighbourhood SES to describe a possible trend.As the method of data collection changed from face-to-face to telephone interviews between these years and they were not entirely comparable, the analyses and presentation of results were separated for 1988-2005 (shown in Table S1) and 2008-2013 (main analysis).Multivariable binary logistic regression was conducted to estimate the odds of severe pain by neighbourhood SES for the pooled data between 2008 and 2013.A model with a cluster robust sandwich estimator was conducted as a sensitivity analysis in order to account for potential correlations between individuals within the same neighbourhoods.There were 4946 neighbourhoods in our study and the average sample size within each neighbourhood was 2.4.The minimum and maximum number of individuals were 1 and 25, respectively.The sensitivity analysis was conducted for neighbourhoods with at least 2 individuals.The parameter estimates were similar between the logistic model and the model with cluster robust sandwich estimators.The first model was a crude association between each variable and severe pain.The second model was adjusted for year of survey, sex and age.The third model was adjusted also for the sociodemographic variables: individual SES, such as income and education, cohabiting status and country of birth.The full model was also adjusted for smoking, exercise and BMI, which may act as both confounders and potential mediators.We constructed our models in a stepwise manner to examine whether the estimates changed after including the potential confounders and mediators.Interaction tests between neighbourhood SES and all covariates, i.e., year of survey, sex, age, cohabiting status, family income, education, birth country, smoking, exercise and the BMI were conducted.None of them were found to be statistically significant (p value: 0.2-0.9).For sensitivity analyses, multivariable binary logistic regression was conducted by treating pain as a binary variable, i.e., mild/severe pain versus no pain (shown in Table S2).
The data were analysed using STATA (StataCorp, 2019).The presentation of the study was in alignment with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

| Descriptive statistics of individual variables
A total of 24.7% of the study sample reported severe pain (Table 1).That proportion was higher among those who lived in low SES neighbourhoods compared to those who lived in medium or high SES neighbourhoods (p < 0.001).The proportion of females was higher than males in the total study population, and that difference was larger in low SES than in medium or high SES neighbourhoods (p = 0.01).More than half of the individuals were below 74 years, and the proportion of individuals above 75 years was higher in low SES than in medium or high SES neighbourhoods (p < 0.001).Most individuals were married or cohabiting, and the proportion of those living alone was slightly higher in low SES than in medium or high SES neighbourhoods (p = 0.008).The proportion of those with low income, low education, and who were foreignborn was higher in low SES than in medium or high SES neighbourhoods (p < 0.001).The proportion of those who smoked and hardly exercised was higher in low SES than in medium or high SES neighbourhoods (p < 0.001).The mean BMI was higher in people living in low SES neighbourhoods (p < 0.001).T A B L E 1 (Continued) 3.2 | The trend of prevalence of severe pain The prevalence of severe pain was unchanged between 2008 and 2013 (Table 2).There were no statistically significant trends by sex, age group and neighbourhood SES in any of the time periods.

| Association between severe pain and neighbourhood SES
The odds of having severe pain were 1.76 times higher in low SES neighbourhoods than in high SES neighbourhoods (Table 3: Crude model).After adjusting for year of survey, sex and age in the second model and individual SES, i.e., income and education, cohabiting status and birth country in the third model, the odds decreased but remained statistically significant (OR = 1.70, 95% CI = 1.49-1.95and OR = 1.43, 95% CI = 1.24-1.64, in the second and third model, respectively).In the full model, after also adjusting for smoking, exercise and the BMI, the odds decreased further but remained statistically significant (OR = 1.30, 95% CI = 1.12-1.50).Engaging in exercise decreased the odds of pain.Similar to the trend between 2008 and 2013, the prevalence of severe pain was unchanged between 1988 and 2005 (Table S1).There were no statistically significant trends by sex, age group and neighbourhood SES in any of the time periods.When the outcome was treated as a binary variable of mild/severe pain compared to no pain, the association with neighbourhood SES was consistent with that of severe pain (Table S2).The estimates in the model accounting for the potential correlations between individuals in the same neighbourhoods were similar to those of the logistic model.

| DISCUSSION
This study shows a significant association between neighbourhood SES and severe pain among older adults in a random sample of the Swedish population.The potential effect of neighbourhood SES remained after adjusting for individual SES as well as for exercise, smoking and the BMI, which are important risk factors for pain.
The findings are consistent with previous studies investigating mid-old adults and older adults (Fuentes et al., 2007;Jordan et al., 2008).Possible mechanisms in the association between neighbourhood SES and pain are that living in deprived neighbourhoods may lead to psychological stress, which in turn can increase the risk of pain.Daily psychological stress is known to affect plasma cortisol and, subsequently, the severity of pain (Fischer et al., 2016).Davies et al. (2009) found that the association between neighbourhood SES and pain was explained by psychological factors, i.e., depression and distress among adults in England.Brooks Holliday et al. ( 2019) found that psychological stress was a significant mediator between neighbourhood SES and pain among African American adults.Those findings suggest that living in a low SES neighbourhood increases psychological stress levels, possibly due to concerns about safety and poor access to public services, and may, as a result, increase the risk of pain (Everson-Rose et al., 2011;Hill et al., 2005;Pickett & Pearl, 2001;Wen et al., 2006).
Additionally, there may be indirect paths between living in low SES neighbourhoods and pain through some behavioural factors, such as physical inactivity.Physical activity is known to be a protective behavioural factor against pain (Geneen et al., 2017).The walkability of the neighbourhoods, measured as access to parks and sidewalks, was found to be associated with pain (Okabe et al., 2019), i.e. poor access increased the risk of pain.The mechanism could be that living in a neighbourhood with poor access to facilities or infrastructure promoting physical activity may foster physical inactivity leading to pain.In many Western settings, residents of low SES neighbourhoods tended to have poorer access to structures and facilities promoting physical activity, such as parks, sidewalks and gyms (Powell et al., 2006).However, our data did not differentiate between types of physical activity such as leisure, transportation, occupation and household.Future studies should include various types of physical activity to elucidate the specific mechanisms between neighbourhood SES and pain.Furthermore, accessibility to healthcare could be a factor lying behind the association between neighbourhood SES and pain.Accessibility to healthcare is important for older adults in order to receive proper pharmacological and non-pharmacological treatment for, e.g., pain (Hayden et al., 2005;Hoeldtke & Hoeldtke, 2000).
Interventions to reduce psychological stress in neighbourhoods may be difficult to achieve since it is a subjective feeling that may differ between individuals.It may also be difficult to modify neighbourhood SES per se as it may emerge gradually over long periods of time (Macintyre & Ellaway, 2003).However, efforts to improve neighbourhood conditions, for example, reduce the crime rates, improve access to physical activity and healthcare facilities may have a positive impact on pain (Brooks Holliday et al., 2019).An important caveat is that we do not know which neighbourhood conditions that are the sources of psychological stress or lack of physical activity, which may lead to pain.Any interventions should be planned carefully and started with a smaller pilot study.
This study was unable to investigate underlying mechanisms, yet we can hypothesize on potential mechanisms based on previous studies.For example, in Sweden, nationwide geographical analysis revealed that low SES neighbourhoods had higher accessibility to healthpromoting facilities (Kawakami et al., 2011), which does not support that the lack of health-promoting facilities lies behind the association between low neighbourhood SES and pain.There may be other unknown direct or indirect pathways between neighbourhood SES and pain.In fact, in our analysis, after adjusting for potential  confounders as as mediating such as physical activity and the BMI, the effect of neighbourhood SES remained; this suggests that additional pathways exist.These pathways could represent residents' perceptions, pain tolerance or management (Brooks Holliday et al., 2019).If poor management of pain is an important factor, telemedicine and eHealth that support individuals in low SES neighbourhoods could perhaps be a feasible intervention (Eccleston et al., 2020).
The study also found that women and individuals who do not engage in exercise had greater odds of having pain than their counterparts.These results are consistent with previous studies (Bergman et al., 2001;Hamano et al., 2014;Jordan et al., 2008;Kamada et al., 2014;McCarthy et al., 2009) and are now confirmed within a nationally representative sample of an older adult population.Together with the findings of high odds of pain among those in low SES neighbourhoods, this could be useful information for clinicians active in such neighbourhoods.
The findings concerning trends in the prevalence of pain over 20 years did not support causality.For  example, if SES would have a causal effect on trajectories in prevalence rate of pain over time would differ across neighbourhood SES.In other words, we would have observed an improvement regarding the prevalence of pain in medium and high SES neighbourhoods compared to low SES neighbourhoods.While such a trend was not observed, the proportion of the prevalence of pain was consistently higher among those in low SES neighbourhoods.This supports the main findings as well as previous study findings that older adults in low SES neighbourhoods are consistently at a high risk of pain, which is useful information for public health strategies.
There are several limitations to our study.The crosssectional study design does not prove causality.However, the findings of this study provide useful population-level perspectives on the management of pain among older adults.This was achieved by using nationally representative samples with access to individual data on a number of characteristics.In addition, the prevalence trends over time were estimated by making use of serial crosssectional data over 20 years.Generalizability may also be limited due to geographical variations, yet our study involved random samples of older individuals in the whole country.There are several unknown factors that may have biased the results, such as the occupation before retirement, alcohol consumption and diet.Taking into account those variables may be informative to minimize confounding effects.Missing data from a larger proportion of females, older age groups, foreign-born and obese individuals, which might have higher odds of pain, could have influenced the effect of neighbourhood SES.This was similar to other variables, such as family income, education and exercise.Neighbourhood SES in this study was based on proportions of residents with low income, unemployment, low educational status and social welfare.Other studies in the United States used various indicators such as walkability, crime rates, proportions of minority groups (Brooks Holliday et al., 2019;Fuentes et al., 2007).This makes it difficult to directly compare the findings internationally.Meanwhile, using these indicators is consistent with previous studies in Sweden, which found associations between neighbourhood SES and other health outcomes; this may enable discussions on potential mechanisms between neighbourhood SES and pain (Kawakami et al., 2011;Sundquist, Winkleby, et al., 2004).Finally, we were unable to assess the potential chronicity of the pain.Chronic pain, which lasts for more than 3 months, is known to be associated with psychosocial factors to a higher extent than acute pain (Cohen et al., 2021;Treede et al., 2019).We were also unable to consider the management of the pain as well as other pain locations.

| CONCLUSION
Older adults living in low SES neighbourhoods have higher odds of pain, and prevention and treatment may be particularly important in this population group.However, in order to develop effective population-based interventions for a certain environment, the mechanisms that underlie the association between neighbourhood SES and pain in older adults should be further investigated, given the previous and current evidence.

F
Abbreviation: BMI, body mass index.a Chi-square test.b ANOVA test.

T A B L E 2
The trend of prevalence of severe pain from 2008/2009 to 2012/2013.

Crude model Sex and age adjusted
a Goodness-of-fit was tested by the Hosmer-Lemeshow test.