Drowning in the Ripple Effect: Identifying a Syndemic Network of Health Experience with Modi�able Health Behaviours using the UK Biobank

A Syndemic model of health experience in severe mental illness (SMI) involving modi�able health behaviour contributors has been theorised but has not yet been investigated. Over the next 10 years mental ill-health and suicidal behaviours have been predicted to increase which will decrease health experience and increase hospitalisation and associated costs. This paper investigated a Syndemic model of health experience in people with SMI informed by physical activity levels, exposure to nature, personal resilience levels, addictive (tobacco smoking and alcohol consumption), and sleep behaviours. Results indicate partial evidence for a Syndemic model, with personal resilience being at its centre. Contrary to previous �ndings, addictive behaviours did not play an important role in the model. Implementing a Syndemic framework approach to current health care strategies could be bene�cial in the development of self-management strategies for people with SMI. This is the �rst paper using SEM analyses to investigate SMI under the Syndemic theory paradigm.


Full Text
Health experience is generally a combination of physical and mental health experience [1].Suffering from a serious illness impacts the happiness an individual experiences [2].Major depressive disorder (MDD), bipolar disorder, or psychotic disorder/schizophrenia are commonly combined under the banner of severe mental illness (SMI), and has been shown to impact health and health experience negatively to the point of premature and above average mortality rates [3].SMI in the UK is evidenced to disproportionately affect people in deprived areas compared to least deprived areas, as well as in people over the age of 35 [4].During the last decennia, global and local health issues have occurred one after the other.The occurrence of illnesses and natural disasters shifting into various epidemics will increase within the next ten years, which is estimated to increase mental ill-health, and suicidal behaviours [5][6][7].In turn, it is expected a decrease on individual health experience and increase hospitalisation numbers and associated cost increases.Research aims have therefore shifted from nding contributors to SMI development to nding preventive measures to reduce the prevalence and recurrence of SMI [i.e., by strengthening coping mechanisms in people with SMI [8]].Currently in mental health care, preventive strategies target risk factors and treat (sub-)clinical manifestations of mental illness to prevent deterioration, multimorbidity, and disability, and promote psychological well-being [8].However, self-management of mental and physical illnesses for people with SMI through behaviours is important in decreasing hospitalisation rates and improving clinical outcomes [9].Happiness is positively associated with mental and somatic health [10].Hence, investigating the relationship between health happiness and health satisfaction, affective states evidenced to affect psychological well-being [11], and engagement in health behaviours in a population of people with SMI could be bene cial in identifying modi able contributors to individuals' health experience [2].
Modi able health behaviours are intentional or unintentional actions that affect health and mortality in individuals, and are a contributor to mental wellbeing [12].Modi able health behaviours describe encouraged or supported behaviours that improve health experience with the potential of synergistically improving both quality of life and wellbeing [8].Engaging in one health behaviour is usually accompanied by engaging in other health behaviours [13], and a synergistic effect of health risk factors on negative (mental) health experience has been connected to engaging in multiple modi able health risk behaviours concurrently [14].Several health behaviours representative of current epidemics in Western cultures have also been linked to SMI and overall health experience; low levels of physical activity [15], addictive behaviour [16,17], personal resilience levels [18], exposure/access to green and blue spaces [19], and sleep behaviour [20,21].Personal resilience is in uenced by individual levels of physical activity [22], felt social support [23] and experiences loneliness [24], and personal strength engagement [25].Links between these health behaviours and SMI have already been explored individually, but the co-occurrence of them and therefore potential synergistic effect on health and health experience has yet to be investigated.
The Syndemic theory [26] provides an important theoretical background for studying these factors concurrently, as it suggests that a synergistic effect of co-occurring epidemics can exacerbate an individual's health outcome and/or experience due to Syndemic vulnerability.A Syndemic describes the effect of multiple co-occurring and simultaneously interacting epidemics to exacerbate each epidemic's effect on individuals [26].The stress of experiencing these co-occurring contributors can lead to excess burden of disease experience [27] and a decline in health experience due to Syndemic vulnerability.Syndemic vulnerability describes the extent to which individual experiences affect co-occurring social and health problems, morbidity, and mortality because of the ecopsychosocial context of an individual, which again can intensify these problems [28].Stress, poverty, discrimination, and other forms of social adversity are the primary route through which social factors have been found to contribute to negative health [29][30][31].A Syndemic model of contributors to health experience has been previously theorised [32], but has yet to be tested.
The main aim of this paper is to explore a Syndemic model of health experience in people with SMI under the Syndemic theory paradigm, informed by physical activity, addictive behaviours, personal resilience, exposure to nature, sleep behaviour, and deprivation.

Participants and Procedure
This is a secondary data analysis using the UK Biobank's (UKBB) baseline assessment data collected between 2006 and 2010.The UK Biobank ethical approval research committee approved the use of this data for this study (Ref.11/NW/0382).The UKBB dataset is a community-based cohort study comprising of over 500,000 volunteers aged between 40-69 years from across the United Kingdom.Participants have undergone routine, standardised measures, provided blood, urine and saliva samples, given detailed information about themselves and agreed to have their physical health followed.Full details of the study design and data collection processes have been previously published [33,34].
We used a sample of the UK Biobank based on SMI diagnosis.The population group was created by including data from participants diagnosed with nonaffective psychotic disorder including schizophrenia and schizophrenia-like conditions (ICD-10 disease classes F20-F29), and/or any bipolar disorder (ICD-10 disease classes F30-31), and/or any MDD (ICD-10 disease classes F32-39).

Measures
The UK Biobank questions and questionnaires are an accumulation of previously used questionnaires from observational studies, population surveys, and clinical trials to identify appropriate measures of exposure in the different areas [35].These stem from validated questionnaires, short scales, and clinical interviews.Validity of inclusion was discussed with a wide panel of international experts for each area of interest [35].Full details on the individual variables (i.e., questions asked to obtain data and response scales) are given in Online Resource 1.To provide a general overview presently, only essential information is listed.The hypothesised model is shown in Fig. 1.

Outcome
Health Experience.To assess health experience, data from two questions determining general health happiness and general health satisfaction using an 8-point Likert scale.

Modi able Health Behaviours
Physical Activity.Four questions assessing the level of physical activity the participants on average engage in a typical week were asked, requesting information on time spent doing moderate and vigorous physical activity, and days spent walking 10 or more minutes at a time as well as time spent walking for fun on an average day in the past 4 weeks.Response options were either numerical input between 0 and 7, or on a 9-point Likert scale categorising the time spent doing speci c types of physical activity.
Addictive Behaviours.Tobacco smoking frequency and number of smokers in household were recorded on a 4point Likert scale.Alcohol intake frequency was recorded on a 6-point Likert scale.Personal Resilience.We considered 9 questions addressing social support, friendship relations satisfaction, family relations satisfaction, nancial security, ability to con de, experienced loneliness, and leisure activities.
Responses were recorded on an 8-point Likert scale for part of the questions, a 3-point Likert scale (loneliness satisfaction), or as Yes/No.Exposure to Nature.Four variables assessed the participants' exposure levels to nature.We accessed data on green spaces (land coverage estimates for domestic gardens and natural vs built environments within 1000m of the participant's home location) and blue spaces (coastal proximity, access to bodies of water within 1000m of the participant's home location).
Sleep Behaviour.Average hours spent sleeping, requiring a numerical input between 1 and 23, and insomnia behaviour/sleeplessness on a 4-point Likert scale were included.

Syndemic Contributors
Townsend Deprivation Index.Additionally, the Townsend Deprivation Index (TDI) was included in the model as a Syndemic contributor to health experience.This is based on the preceding national census output areas and corresponds to the participant's postcode as baseline measurement.
Ethnicity.Ethnic background determined by a touchscreen questionnaire was included as a possible Syndemic contributor to health experience.

Data Analysis
Structural equation modelling (SEM) was used to investigate the relationships between the modi able health behaviours outlined above and how they relate to health experience.Relevant summary statistics were initially assessed using mean and standard deviations for continuous data and counts and percentages for categorical data.Spearman correlation coe cients were calculated to identify any multicollinearity that might reduce model validity and strength if kept [36].The analysis used standard indices to evaluate model t; the comparative t index (CFI), the Tucker-Lewis index (TLI), and the root mean square error (RMSEA).Threshold scores of 0.90 or higher for the CFI and TLI, as well as a RMSEA value of 0.08 or lower were used as indication of a good model t (Finch, 2020).All relevant statistics and model parameters of interest including con dence intervals are reported.Analyses were conducted in R (version 4.0.5)using RStudio and lavaan package [37,38].
We report the relevant summary statistics for each analysis (i.e., the averages and error surrounding measurements of interest and the composition and size of the sample) along with the model parameters of interest and the associated 95% con dence intervals of these estimates.

Descriptive statistics and correlations
Out of the 500,000 participants in the UKBB database, a total of 8,014 participants had an SMI diagnosis and provided data on the variables of interest.Only data from participants who responded to all questions were included.On average, the participants were 55.61 years old (SD = 7.71 years), 70% were female and the majority were white (96.72%;Table 1).Table 1 gives an overview of all averages and standard deviations of the variables used in this model.The regression results depicting direct effects for all latent factors are outlined in Table 2. Within the latent factors of personal resilience, exposure to nature, physical activity level, and sleep behaviour the selected contributors were found to be signi cantly related to each other with few exceptions.In the latent factor of personal resilience, education pursued in leisure time (p = 0.212), engaging in religious activities in leisure time (p = 0.093), and going to the pub in leisure time (p = 0.205) were not found to be signi cantly related to the other representatives.
Similarly, for the latent factor addictive behaviour, the frequency of smoking was not signi cantly related to alcohol intake (p = 0.590), whereas the amount of smokers in the household was (p = 0.004).3 gives an overview of all main effects.All latent variables predicted health experience, apart from addictive behaviours (p = 0.742).This implies that the latent variables for physical activity, exposure to nature, personal resilience, and sleep behaviour, as well as the measure for TDI equally predict health experience.The higher the level of physical activity, personal resilience, and good sleep behaviours, the higher the health experience.Equally, the lower TDI, the higher health experience.Contrary to theoretical evidence, the lower the exposure to nature, the higher health experience seems to be.Addictive behaviours seem to not be directly related to health experience in the context of this data and model.
Figure 2 shows relationships between the variables of the model with effects and standardised path coe cients.
The standardised path coe cients between the latent variables have been excluded to improve readability.

Discussion
The present study aimed to explore a potential Syndemic model of health experience in people with SMI informed by physical activity levels, addictive behaviours, personal resilience, exposure to nature, sleep behaviour, and TDI.
Contrary to theoretical evidence, the relationship between addictive behaviours and health experience did not t into this novel Syndemic model [43].

Theoretical Implications
In support of previous publications, most investigated contributors signi cantly affect health experience.Evidence for the existence of a Syndemic model of intercorrelated contributors that affect health experience has not been published yet; this is the rst publication about a Syndemic model of health experience fed by modi able health behaviours as contributors.Though some model t indices did not reach the desired threshold of 0.9, the model aids in developing the theory of Syndemics in health experience in an SMI population.Ample theoretical evidence from both the theory of Syndemics as well as health behaviour contributors in mental health-focused research has thoroughly informed the development of both this model and theory.This study provides evidence in support of personal resilience building behaviours being intercorrelated to physical activity, sleep behaviour and exposure to nature, as well as being related to health experience in people with SMI; which is in support of a Syndemic model of health experience and resilience-related health behaviours in an SMI population.
Additionally, this is the rst publication of investigating the existence of a Syndemic model using SEM as an analysis tool.Though previously suggested as the way forward in investigating Syndemic models [44], recent research into Syndemic models has not yet implemented SEM to analyse the Syndemic relationships between contributors in one analytical model rather than using individual regression paths between outcome and possible contributors.
Though ample evidence supports addictive behaviours to contribute to health experience [45,46] Limitations and Future Research UK Biobank have stated that the sample collected for the dataset shows evidence of a "healthy volunteer" bias and is therefore not representative of the general population on information collected for lifestyle, sociodemographic, health-related, nor physical [34].Additionally, the sample is predominantly female (70%) and the mean age of 55 years suggests an older age group than the population mean [47].Due to the distribution of SMI within the sample, and 99.99% of the sample reporting an MDD diagnosis, the ndings can be mainly attributed to a population suffering from depression symptoms.Furthermore, the distribution of SMI diagnoses of bipolar disorder and schizophrenic spectrum disorder is below the estimated average in Britain [48].Therefore, these results are not generalisable to the wider population, and only to a population with SMI diagnoses.
Furthermore, a limit of the otherwise very thorough and vast UKBB dataset means that exposure to nature can only accurately be determined by measuring the proximity of participants to green or blue spaces.Arguably, proximity to green and blue spaces does not mean that someone is exposed to natural settings more often as well; participants living in areas that are deprived of green and blue spaces might still make the effort of seeking exposure to natural environments which is however not recorded in the dataset.Including questions that speci cally measure seeking exposure to natural environments like blue and green spaces could be a useful addition to the already substantial UKBB dataset.Comparably, addictive behaviour measurements were limited in their depth and breadth in the UKBB; probably due to the UK's successful anti-smoking campaigns [49], there is only a small percentage of smokers in the dataset, which makes measuring the extent of the effect of these behaviours less informative.

Conclusion
The results of these analyses provide evidence for the existence of a Syndemic network of health behaviours as contributors to health experience in an SMI population.Previously established contributors to detrimental health experience in people with SMI not just singularly affect health experience, but also interact and syndemically exacerbate health experience.Health care policies and practices should therefore move from a strategy of challenging singular contributors one at a time to tackling multiple contributors simultaneously.Implementing policies and strategies using the Syndemic framework approach could therefore improve the effectiveness of current health care strategies and provide better health care options for a population with SMI, including evidencebased self-management strategies.Future research into the Syndemic network of contributors to health experience and how health care strategies could improve it would be bene cial for populations nationally and ultimately, globally.

Figures
Figure 1 SEM of the presented model.

Table 1
Means and standard deviations for all variables included in the model, grouped by SMI diagnosis.
Spearman correlation coe cients indicated a high correlation between the two Syndemic contributors, ethnicity and TDI.As TDI is a continuous variable based on externally veri ed census methods, and ethnicity is a categorical self-input variable, we elected to eliminate ethnicity from the model to improve model stability.Primary analyses indicated high correlations between several variables (days engaging in vigorous physical activity and practising a sport during leisure activities; engaging in religious activities in leisure time and engaging in group activities in leisure time; and alcohol intake frequency and pub visits in leisure time), which were allowed to be estimated in the model.The measurement model analysis con rmed that the selected variables show appropriate ts (Leisure time -pub and Leisure time -education for the Personal Resilience construct, and smoke frequency for the Addictive Behaviours construct).Sensitivity analyses on the effect of removing these contributors from the model resulted in comparable model indices, so to provide a broader insight into the model, we elected to keep the contributors in the nal SEM.The results of the SEM indicated a reasonably good t of the tested model.Although χ² was signi cant (χ²= 6035.766,p < 0.001, df = 281), the TLI and CFI indicators demonstrate a reasonably good model t (TLI = 0.786, CFI = 0.815).The RMSEA indicated a very good t of the model (RMSEA = 0.051).This implies that the model provides a reasonably good t to the data, however, the model could probably be improved.A closer investigation of the individual relationships within the model for future iterations is therefore necessary.

Table 3 SEM
Regression Analysis.

Table 4
shows the correlations between the latent factors of health behaviours.The latent factors of exposure to nature, physical activity, and sleep behaviour were signi cantly related to personal resilience (p ≤ 0.001).This implies that individual levels of personal resilience affect individual levels of exposure to nature, physical activity, and sleep behaviour.Beyond these signi cant relationships, no other latent factors showed signi cant interrelatedness.

Table 4
Correlation between the Latent Factors.
, they did not come forward as a signi cant predictor in our model.Despite suggestive that addictive behaviours may play a different role other than the other modi able health behaviours tested (i.e., physical activity, sleep behaviour, and personal resilience), further studies are still necessary.For example, addictive behaviour may operate on as mediator or moderator of the effects of other behaviours on the health experience.In addition, the variables available to construct the addictive behaviours latent variable did not include any dependence-related measure.Therefore, we were not able to test whether adding substance-related problems or addiction diagnosis would change the observed results.There is therefore scope for investigating which model of Syndemic contributors to health experience addictive behaviours t into; the current Syndemic model is focused on modi able health behaviours, but addictive behaviours measured this way may be more closely correlated to a different Syndemic layer health experience.