Article Text
Abstract
Background Interest in monitoring health-related quality of life (HRQoL) in general populations has increased in the past 20 years, reinforced by population ageing and repeated economic crises. This study aims to identify temporal trends in HRQoL in France between 1995 and 2016 and to assess disparities according to demographic and socioeconomic characteristics.
Methods Data from repeated population-based cross-sectional surveys conducted in 1995, 2003 and 2016 were used. HRQoL was measured using the Medical Outcomes Study 36-item Short Form (SF-36) questionnaire.
Results A substantial decrease in score was observed between 1995 and 2016 for both genders in almost all subscales of the SF-36, with the largest decrease being in the mental health dimension for men. However, the age group 18–54 years were the most affected with persistent negative or even worsening trends in HRQoL. The largest decreases were among men aged 45–54 years and women aged 35–44 years in most dimensions, and among the age group 18–24 years in vitality. Conversely, an overall improvement was noted among the age group 65–84 years. People in employment were more affected than the unemployed by the decline in several HRQoL dimensions.
Conclusion A general decline in HRQoL was found between 1995 and 2016 in the French population, but with wide disparities in trends between age groups. Young and especially middle-aged, employed people exhibited persistent negative and worsening trends. Consistent with evidence from traditional mental health morbidity and mortality indicators, our findings raise questions about the potential influence of macro-socioeconomic factors, especially the 2008 crisis; these observations deserve special attention from health policy-makers.
- quality of life
- employment
- inequalities
- psychosocial factors
- self-rated health
Statistics from Altmetric.com
Introduction
‘Health-related quality of life’ (HRQoL) includes ‘those aspects of self-perceived well-being that are related to or affected by the presence of disease or treatment’1 and is now recognised as a meaningful indicator of health status.2 Self-rated health (SRH) is a measure of the perceived gap between one’s own health and the optimal health that could be expected, and similarly HRQoL includes the subject’s views on his/her health. Both are complementary to traditional biomedical approaches based on morbidity or mortality.
An individual’s perception of a decline health is recognised as a predictor of later poor health outcomes for that individual, as assessed by ‘hard’ measures such as hospitalisation or mortality.3 Demographic and socioeconomic factors can have large effects on HRQoL with significant disparities among subgroups.4 Monitoring HRQoL in the general population may thus be useful to predict subsequent health trends and help track health disparities. However, there have been few studies on the temporal evolution of HRQoL in populations. The latest general population-based surveys in Western countries suggest unfavourable trends in health perception5 6 and several studies show increasing inequalities in SRH.7 8 The possible determinants of these trends include socioeconomic factors such as unemployment and deindustrialisation, for which associations with HRQoL have been reported.9 It would therefore be interesting to measure the extent to which economic crises interact with mental health status and existing health inequalities.7 10 The current background of repeated economic crises and ageing populations in Western countries raises questions about the existence of disparities between age groups and the quality of life in the increasingly numerous members of the older age groups. More precisely, there are questions about the consequences of crises for quality of life, as could be mediated by employment strain or unemployment (which increased in France from 7% of the labour force in 2008 to 10% in 201611); there are also questions about whether the increase in duration of life, observed in many Western countries and notably in France in recent decades, is accompanied by changes in the ‘quality’ of life. The aim of this study was to assess trends in HRQoL since 1995 in the general French population and to identify differences according to demographic and socioeconomic characteristics. We used three population-based cross-sectional surveys, conducted in 1995, 2003 and 2016, assessing HRQoL using the Medical Outcomes Study 36-item Short Form (SF-36) questionnaire.
Materials and methods
Study population
The same validated French version of the SF-36 questionnaire was administered to three independent representative samples of the French population in 1995, 2003 and 2016, following a repeated cross-sectional design; these were the only surveys using the SF-36 questionnaire to have been conducted in France.
We adhered to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) in developing the database, analysis and presentation of the study (online supplementary appendix—GATHER statement).
Supplementary file 1
The 1995 survey used a two-stage stratified sampling method: a randomly selected household-based sample was assembled from the population and housing census, and one subject randomly chosen in each household selected was included. The sample included subjects aged 15 and over. Information was collected by post.
The 2003 survey used a multistage stratified sampling design: the households were randomly selected from the population and housing census, but unlike the 1995 survey, all subjects in the households selected were questioned. The sampling design was stratified on region and size of urban unit of residence. The sample included subjects of all ages. Information was collected using a combination of face-to-face interviews and self-administered questionnaires.
The 2016 survey used a quota-based sampling approach. The sample included subjects aged 18 and over. Information was collected by telephone, allowing the absence of missing data in the responses.
We included in the present study only subjects aged between 18 and 84 years old who had completed at least one subscale of the eight SF-36 subscale scores: consequently we studied samples of 3582 (including an oversample of 339 aged over 65 years), 22 743 (including oversamples for Paris, North, Eastern Parisian Basin and Mediterranean Basin regions) and 1494 subjects for the years 1995, 2003 and 2016, respectively.
More details regarding the methodology of the three surveys are available in the online supplementary appendix—Survey methodology.
Supplementary file 2
HRQoL measurement
The SF-36 is a validated, generic, self-reported questionnaire designed to measure perceived HRQoL. The SF-36 includes 36 items measuring eight health concepts: Physical Functioning, Role limitations relating to physical health, Bodily Pain, General Health, Vitality, Social Functioning, Role limitations relating to mental health and Mental Health. The eight SF-36 subscale scores were constructed using the sum of Likert-type items secondarily transformed into 0–100 scales to create eight crude scores. A score of zero indicates the worst possible perceived health and a score of 100 indicates the best possible perceived health. Each subscale score was standardised for gender and age group and can be expressed in SDs such that the 1995 survey is the normative reference. The same French version of the SF-36 questionnaire (version 1.3) was used in our three samples.12
Dependent and explanatory variables
The eight standardised SF-36 subscale scores (age and sex-adjusted) expressed in SDs12 were studied in relation to the year of survey, personal variables (age group, education level, occupational status, matrimonial status and chronic conditions) and contextual variables (size of urban unit of residence and geographical area). The year of survey was considered as a continuous variable to take into account the different time intervals between the three surveys. All the other variables are categorical. To match the three surveys—the original published analyses used different approaches to categorise variables—we used category groups defined in the same way for all three surveys. More details about the explanatory variables are available in the online supplementary appendix—Explanatory variables.
Statistical methods
Preliminary analysis
Continuous data are presented as means (SDs) and categorical data are presented as percentages. We used a one-way analysis of variance and χ2 tests to compare variables among the three surveys. The oversample of 339 subjects older than 65 years in the 1995 survey were excluded from these descriptive analyses to provide representative summary statistics.
Modelling analysis
To study changes over time in standardised SF-36 subscale scores and to detect differences in these changes according to sociodemographic characteristics, we used polynomial regression models; these models were used to estimate the effect of explanatory variables on age–sex standardised SF-36 subscale scores controlling for individual and contextual variables. The modelling strategy consists of four successive steps. First, we entered the year of survey and the year² of survey (second-degree polynomial) with no additional covariate (crude model). The squared term was used to search for non-linear relationships between scores and time. Second, we entered individual variables: age group, education level, occupational status, matrimonial status and selected chronic conditions (cancer, diabetes, hypertension and heart disease) for which data were available. Third, we entered contextual variables: size of urban unit of residence and geographical area. Finally, interactions were tested between the year of survey (linear and square terms) and the covariates one by one according to their input order. These four steps generated the ‘final adjusted models’. These analyses were performed for men and women separately. Additional methodological details of the modelling analysis are available in online supplementary appendix—Modelling analysis.
Patient involvement
No patients were involved in setting the research question or the outcome measures, nor were they involved in developing plans for design or implementation of the study. No patients were asked to advise on interpretation or writing up of results. There are no plans to involve patients in the dissemination of the results.
Results
Preliminary analysis
The main sociodemographic characteristics of the three survey samples are presented in table 1. Estimates were within comparable ranges in the three samples, except for those concerning education level and occupational status.
The largest decrease between 1995 and 2016 was in the mental health score for men. Most of the standardised SF-36 scores decreased significantly over time in all age groups, with subjects aged 18–54 years being particularly affected, and subjects aged 65–84 years least, or even not, affected by this negative trend. The mental health subscale showed a significant decline in almost all age groups, more markedly in men (online supplementary appendix table 1).
Supplementary file 3
Mentally oriented raw SF-36 scores for subjects aged 65–84 years in 2016 were in many cases higher than those for subjects under 65 years (online supplementary appendix table 2).
Supplementary file 4
Modelling analysis
Polynomial regression models, both crude models (see online supplementary appendix table 3) and adjusted models (online supplementary appendix tables 4 and 5), showed changes over time in standardised SF-36 subscale scores.
Supplementary file 5
Year of survey
In crude models, almost all scores decreased substantially over time. Nonetheless, four scores (physical functioning and role limitations relating to mental health for men, physical functioning and vitality for women) declined substantially and then subsequently increased, possibly indicating convex relationships between these four scores and time. Conversely, the mental health score for men worsened substantially with a faster decline between 2003 and 2016, possibly indicating a concave relationship between this score and time.
Age categories
In adjusted models, significant interactions between the year of survey and age, and/or between year² and age, were found for several subscale scores (physical functioning, role limitations relating to physical health, bodily pain, general health, vitality and social functioning for men; physical functioning, bodily pain, general health, vitality and social functioning for women). The significance of those interaction terms indicates that age groups did not change similarly over time (see figures 1 and 2).
Supplementary file 6
Supplementary file 7
The age group 18–54 years was the most affected by the decline in HRQoL, with the decline continuing or even worsening between 2003 and 2016, especially in the bodily pain, general health, vitality and social functioning dimensions. The groups most markedly affected by a faster decline between 2003 and 2016 were: men aged 18–54 years, and women aged 18–24 and 35–54 years for bodily pain; men aged 25–44 years and women aged 18–54 years for general health; men aged 18–44 years and women aged 18–24 years for vitality; and women aged 18–24 and 45–54 years for social functioning. Those worsening trends were up to −0.65 SD between 2003 and 2016. Otherwise, in most dimensions, the largest decreases over the study period were for men aged 45–54 years (they clearly stand out from the other age groups for physical functioning, bodily pain and social functioning) and for women aged 35–44 years (they clearly stand out from the other age groups for bodily pain, general health and social functioning), with up to −0.80 SD declines between 1995 and 2016. Also, vitality decreased substantially in the age group 18–24 years, especially in men, with a fall of −1.00 SD since 1995 (the sharpest decline in any age group for any dimension).
Conversely, values for the age group 65–74 years and 75–84 years increased between 2003 and 2016. These age groups clearly differ from the other age groups in most dimensions, with increases of up to +0.80 SD between 2003 and 2016. For both genders, the largest divergences between the age groups were for vitality, with widening disparities between the young and middle-aged and the elderly.
Gender
Demographic trends were similar in men and women: an emerging gap between the young and middle-aged and the elderly for several dimensions, and no significant interaction between the year of survey and age for role limitations relating to mental health and mental health. However, there were some differences; for example, the decrease of mental health was much larger in men than in women (about −0.60 SD vs −0.30 SD, respectively, between 1995 and 2016). Almost all raw scores for women were lower than those for men, although the differential changes in the mental health dimension tended to reduce gender disparities.
Work status
We restricted each model to the age group 25–64 years (an age range within which most people are professionally active in France) to assess more specifically the effect of unemployment. This identified significant interactions between year and occupational status and/or between year² and occupational status in several subscale scores: bodily pain, general health and vitality for men, and physical functioning, bodily pain, general health, vitality, social functioning and mental health for women (figure 3). Scores for the unemployed were generally below those for the employed. However, the employed showed larger decreases in HRQoL than the unemployed over the study, with worsening trends between 2003 and 2016 in the bodily pain and general health dimensions for both genders.
Discussion
We report a substantial decrease in almost all dimensions of HRQoL in France between 1995 and 2016 in both genders. Mental health was the most affected dimension with a decline in almost all age groups; men were more severely affected than women, and the decline was faster between 2003 and 2016 than between 1995 and 2003. For most dimensions, age groups did not evolve similarly: the largest declines in HRQoL were for the age group 18–54 years and these trends continued or worsened between 2003 and 2016, especially in the bodily pain, general health, vitality and social functioning dimensions. The largest decreases in most dimensions over the last 20 years were for men aged 45–54 years and for women aged 35–44 years; subjects aged between 18 and 24 years (especially men) showed the biggest falls in vitality. The widening disparities between the young and middle-aged and the elderly are particularly evident in the vitality dimension. Indeed, an overall improvement of HRQoL was observed among the age group 65–84 years between 2003 and 2016; in 2016, mentally oriented raw scores for the elderly were in many cases higher than those of subjects under 65. For several dimensions, those who were employed were more affected by the decline in HRQoL than the unemployed, with larger decreases between 2003 and 2016 in the bodily pain and general health dimensions.
Contrasted trends across gender and age categories
Our finding of a general decrease in HRQoL in France is consistent with recent reports of similar deteriorations in perceived health in Western populations.5–7 Our work indicates that the consideration of overall tendencies in whole populations may fail to identify large disparities between subgroups, and particularly between age groups. Several studies have reported possibly widening disparities between the young and middle aged and the elderly. A 2001–2011 US study found a declining HRQoL in younger age groups, whereas the older age groups tended to report higher HRQoL.13 Similar trends have been observed in Estonia, Lithuania and Finland,14 and other studies found unexpected improvements of mentally oriented scales with increasing age.8 15 Various factors may cause such disparities between age groups in both HRQoL and its temporal trends. They include differences in expressed needs and expectations of health between older subjects and their younger counterparts, and the possible easing of stress relating to employment and the working environment as indicated by the positive relationship reported between retirement and mental health.16 Our findings are also of interest to the current debate over whether or not the steady increase in life expectancy in Western countries, notably in France, is accompanied by an increase in time without disability. Our findings also converge with a large population-based study covering a period matching our study period (1991–2011) which reported significant increases in years lived with low dependency after age 65: this suggests that the gain in life expectancy may at least partially contribute to an improvement in health status and HRQoL at older ages.17
Socioeconomic determinants and contrasted trends across work status
Our findings raise questions about the extent to which socioeconomic determinants have contributed to health trends. In particular, it is plausible that the 2008 economic crisis negatively affected the health of the population. In the short term, mental health may be more strongly affected than physical health by an economic crisis.10 Consistent with our findings, other recent studies report that men have been more affected than women by the adverse consequences on mental health of the economic crisis.18 19
The apparently paradoxical findings for occupational status—although the HRQoL of the unemployed declined, that of the employed declined more substantially—agree with other studies which found that populations usually considered to be less vulnerable, such as those who remained employed, also suffer from the consequences of economic crisis.14 19 20 This suggests that employment status alone is insufficient to explain how the recession may influence mental health: job insecurity may result in adverse effects on mental health even if employment status is unchanged.21 Job insecurity and adverse psychosocial conditions at work during or following the 2008 recession may help explain our findings for the employed, and indeed work-related stress effects on health have been widely documented.22
The 2008 economic crisis has severely affected working populations. A recent European study shows a deterioration of the working conditions in most countries.23 In France between 2006 and 2010, there was an increase in exposure to psychosocial risks at work affecting all categories of workers, especially the middle-aged.24 Working conditions are major determinants of workers’ health. Consistent with our findings, a deterioration in self-reported health for both genders, increases in smoking for both genders, excessive alcohol consumption among women, sleep problems among men and insufficient sleep duration for both genders have been described in the French working population between 2006 and 2010.25 Work–family conflict, resulting from juggling work and family tasks, has been increasing and is negatively associated with general and mental health.26 It is thus not surprising that, in our study and in other European countries,7 the middle-aged appear to be the most severely affected by the decline in perceived health.
We report the apparently paradoxical result that the unemployed show a smaller decline in HRQoL than the employed, thus reducing perceived health differences, and there have been similar findings in other European countries. Indeed, some health inequalities between the unemployed and employed might be reduced during periods of widespread unemployment, especially in the ‘crisis countries’.14 27 28 Unemployment becomes a smaller deviation from the social norm and an individual’s unemployment can be attributed to external factors. Consequently, the social stigma associated with job loss tends to be diminished during such periods.29 The suicide rate among the unemployed decreases during economic recession and increases during recovery.30 Social protection for the unemployed can moderate negative health-related consequences of unemployment, as shown in some European countries27; it seems plausible that the French social protection system contributed to limiting disparities between the unemployed and the employed during the recession.31
In line with other surveys, our findings suggest a negative association between the 2008 economic crisis and HRQoL not only for the unemployed but also, and most importantly, for those who remained employed, regardless of occupational status. The populations most at-risk in the French labour market, including the middle-aged, are not those generally considered to be the most vulnerable in other contexts.
Self-perceived health versus objective health indicators
Some studies suggest that the use of HRQoL or SRH may lead to improperly estimating the true magnitude of health inequalities between socioeconomic groups.32 Lang et al 2 indicate that perceived health varies with culture and time and a decrease in perceived health does not necessarily reflect a decrease in health status; it may also reflect an increase in health expectations. Health perception is a major determinant of access to healthcare and health-related behaviours,33 but individual perceptions do not systematically align with physician observations or ‘objective’ indicators.34 It is therefore necessary to take into account an individual’s perception when adapting health promotion programmes. A decrease in SRH can be observed before the diagnosis of a disease leading to death suggesting that SRH can detect pathological changes before ‘objective’ indicators.35 Even though observed differences according to age and occupational status may not accurately reflect the exact magnitude of health inequalities between categories, our study documents a growing feeling of unease since 1995 among the young and middle-aged and the employed. These findings converge with several traditional objective health indicators for France, and especially with the suicide rate. The suicide rate is an imperfect but sensitive indicator or proxy for mental health disorders and social well-being in a country: between 1995 and 2014, the smallest relative decrease of the suicide rate in France was for the age group 45–54 years (both sexes)36; between 2004 and 2011, women aged 40–49 and men aged 35–39 years had the highest rate of hospitalisation for suicide attempt37; and between 2005 and 2010, the prevalence of major depressive episodes increased only for men aged 35–54 years.38
Strengths and limitations
This study has several strengths. First, our samples are representative of the French population, are large and provide more than 20 years of hindsight. Second, the same validated French version of the SF-36 questionnaire was used in all three surveys. The property of invariance of the SF-36, ensuring temporal stability, has been demonstrated.39 Third, the modelling analysis adjusted for many sociodemographic factors influencing HRQoL. Consequently, we are confident that the changes in HRQoL observed between the three surveys are genuine.
This study also has limitations. First, sampling and questionnaire administration methods differed between the three surveys. However, there was no evidence for differential selection bias between age groups or gender that could explain the contrasting trends observed across age categories or gender groups. Information was collected by telephone in 2016 versus self-administered questionnaire in 1995 and 2003. HRQoL scores recorded by telephone have been shown to be more positive than those obtained by self-administered questionnaire.40 Had this been the case in our study, the participants in the 2016 survey would have adjusted their responses upward; the magnitude of the resulting potential information bias is unknown. However, any such effect cannot explain the deterioration in HRQoL observed among the young and middle-aged. Second, only three time points were studied, preventing the analysis of fluctuations in the time between surveys. Third, residual confounding cannot be ruled out because data for some potentially influent covariates were not available in the three surveys, notably household income, physical activity, dietary habits and other health behaviours and risk factors. The observed deterioration among the young and middle-aged was of such magnitude, in many cases exceeding −0.50 SD (up to −1.00 SD), that we believe it unlikely that it is the consequence of residual confounding. Finally, the study design, an observational study with repeated cross-sectional surveys, does not allow us to draw causal inferences.
Conclusion
There has been a general decline in HRQoL between 1995 and 2016 in the French population, with differences between subpopulations. There were differences between age groups with an improvement for the elderly between 2003 and 2016, and continuing negative or worsening trends for the young and middle-aged over the same period. Similar to this, HRQoL deteriorated more for the employed group than for the unemployed group. Thus in both cases, the decline in HRQoL was greater for the group generally considered to be less vulnerable than in the group considered to be more vulnerable. It is plausible that these findings could in part be consequences of the 2008 crisis, and macro-socioeconomic factors more generally. Our findings are coherent with the trends in HRQoL observed in other European and Western countries and with trends in several traditional health indicators based on morbidity or mortality in France. HRQoL may predict the ‘objective’ health status of populations or subpopulations, so the deterioration of HRQoL among those who are employed can be expected to be followed by increased morbidity and health spending. These disturbing observations deserve special attention from health policy-makers.
What is already known on this subject
There have been reports of overall declines in health-related quality of life (HRQoL) or in self-rated health over recent decades in Western populations, and of the possible involvement of deindustrialisation and unemployment. However, the extent to which economic crises interact with HRQoL and existing health inequalities is unknown.
What this study adds
The decline in HRQoL over the last two decades in France, mainly affecting young and especially middle-aged, employed subjects, raises questions about the potential influence of macro-socioeconomic factors, especially the 2008 crisis.
This decline is consistent with trends in some traditional mental health indicators based on morbidity or mortality and also with observations in populations in other Western countries.
Health policy-makers should take note of these findings as it is likely they indicate subsequent increased morbidity and health spending in those populations.
Acknowledgments
We thank Serge Briançon, Francis Guillemin, Thierry Lang for useful comments on the manuscript.
References
Footnotes
Contributors JC and EA planned the study. A-CC-V did the statistical analyses and drafted the manuscript. A-CC-V, EA, AL and JC contributed to the analysis and reviewed the manuscript. All authors had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. JC and A-CC-V are the guarantors.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Ethics approval The 1995 study was performed by SOFRES as part of the IQOLA project. All of the participants gave written informed consent before data collection. The 2003 study was performed by the National Institute for Statistics and Economic Studies. All of the participants gave written informed consent before data collection. The data were made available in the public research domain without any identification of personal information. The 2016 study were performed by IPSOS as part of a European study designed to test the psychometric properties of the PROMIS-29 questionnaire. All of the participants gave written informed consent before data collection. The study was conducted in conformity with the Declaration of Helsinki (https://www.wma.net/policies-post/wma-declaration-of-helsinki-ethical-principles-for-medical-research-involving-human-subjects) and with the French law on privacy.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data available.