Introduction
Unhealthy lifestyle behaviors, such as physical inactivity, unhealthy diet, a poor sleep pattern, and substance use, are prevalent among people with mental illness (MI) [Reference Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum and Galletly1,Reference Firth, Solmi, Wootton, Vancampfort, Schuch and Hoare2]. In recent years, these behaviors have gained more attention in mental health care due to their substantial role in the development of physical conditions, such as cardiovascular disease, obesity, and diabetes mellitus [Reference Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum and Galletly1,Reference Correll, Solmi, Veronese, Bortolato, Rosson and Santonastaso3,Reference Bendayan, Kraljevic, Shaari, Das-Munshi, Leipold and Chaturvedi4]. These physical conditions contribute significantly to the disability and mortality of people with MI, leading to a reduced life expectancy of up to 20 years compared to the general population [Reference Plana-Ripoll, Pedersen, Agerbo, Holtz, Erlangsen and Canudas-Romo5,Reference Chan, Correll, Wong, Chu, Fung and Wong6]. Despite extensive evidence and calls for action [Reference Gronholm, Chowdhary, Barbui, Das-Munshi, Kolappa and Thornicroft7,Reference Stubbs, Ma, Schuch, Mugisha, Rosenbaum and Firth8], the mortality gap persists. Moreover, the proportion of physical conditions appears to be increasing in people with MI, so promoting a healthier lifestyle is necessary and warrants additional investment [Reference Plana-Ripoll, Weye, Momen, Christensen, Iburg and Laursen9].
Lifestyle behaviors not only impact physical health but are also linked to the onset and persistence of mental disorders. Growing evidence supports the efficacy of lifestyle interventions in improving both physical and mental health [Reference Firth, Solmi, Wootton, Vancampfort, Schuch and Hoare2,Reference Stubbs, Vancampfort, Hallgren, Firth, Veronese and Solmi10–Reference Burrows, Teasdale, Rocks, Whatnall, Schindlmayr and Plain14]. Furthermore, a comprehensive meta-review investigated how various lifestyle behaviors individually affect the onset and treatment of mental disorders [Reference Firth, Solmi, Wootton, Vancampfort, Schuch and Hoare2]. However, it also highlights the predominant focus on the isolated effects of individual lifestyle behaviors. Since lifestyle behaviors do not occur in isolation, it is crucial to gain more understanding of their interrelations.
Research into lifestyle behaviors has primarily focused on physical activity (PA), which is strongly linked to other lifestyle behaviors [Reference Firth, Solmi, Wootton, Vancampfort, Schuch and Hoare2]. Regular PA has been shown to improve sleep quality [Reference Lederman, Ward, Firth, Maloney, Carney and Vancampfort15], while sleep deprivation can reduce motivation for exercise and lower overall activity levels [Reference Jurado-Fasoli, De-la-O, Molina-Hidalgo, Migueles, Castillo and Amaro-Gahete16]. Poor sleep quality can also lead to lowered mood and reduced impulse control, making it more difficult to maintain healthy behaviors [Reference Pilcher, Morris, Donnelly and Feigl17]. Additionally, PA also plays a role in cognitive functioning and executive planning, which can help with better meal planning and healthier food choices [Reference Luciano, Sampogna, Amore, Bertolino, Dell’Osso and Rossi18]. Conversely, sleep problems can increase dietary intake due to extended wakefulness and disrupted hormonal regulation, increasing cravings for unhealthy foods [Reference Godos, Grosso, Castellano, Galvano, Caraci and Ferri19]. Furthermore, lifestyle behaviors such as smoking complicate these relationships. While nicotine has a stimulant effect which reduces the quality of sleep [Reference Gordon20], smoking cessation may increase appetite, which may lead to weight gain. These examples illustrate the interconnected nature of lifestyle behaviors, influencing each other in ways that can either support or hinder mental and physical health outcomes. It is therefore crucial that we gain an understanding of how these behaviors are interrelated, to address multiple lifestyle behaviors simultaneously.
The network approach offers a powerful method for exploring these complex relationships [Reference Robinaugh, Hoekstra, Toner and Borsboom21,Reference Borsboom and Cramer22]. A psychological network consists of nodes representing observed variables, connected by edges representing statistical relationships [Reference Epskamp, Borsboom and Fried23]. For example, the Gaussian Graphical Model (GGM) estimates a network of partial correlation coefficients. These coefficients represent the strength of a relation between two variables after controlling for the other variables in the model [Reference Epskamp, Waldorp, Mõttus and Borsboom24]. Furthermore, by assessing network parameters like node strength, we can gain insight into which nodes are more strongly connected than others. Strongly connected nodes may signal symptoms that could potentially play an important role in stabilizing the network and may be investigated as treatment targets [Reference Jurado-Fasoli, De-la-O, Molina-Hidalgo, Migueles, Castillo and Amaro-Gahete16].
This study aims to explore the relationships among lifestyle behaviors and health outcomes and to identify the most central lifestyle behavior or health outcome in this network. In line with the exploratory nature of this study, there were no specific predictions about which behavior or health outcome was most central. Nevertheless, given the associations between lifestyle behaviors and mental and physical health, we hypothesized that these behaviors were interconnected rather than independent. Understanding these interconnections could inform treatment and guide future research to address the challenges people with MI face in improving their health.
Methods
Study design and setting
This study is based on cross-sectional data, collected as part of a larger trial evaluating the effectiveness and implementation of a lifestyle-focused approach for inpatients with MI (MULTI+) [Reference Gordon20]. The overarching trial was conducted at GGz Centraal, a mental healthcare facility in the Netherlands, comprising 45 inpatient wards grouped into three clusters with approximately 800 places of residence. During the trial, all clusters initially delivered care as usual (CAU), and every six months one cluster transitioned to MULTI+ until all clusters had switched. Measurements were collected at the start of the trial, and subsequently at a six-month interval (after 6, 12, and 18 months) across all clusters. For the present study, we used data collected prior to each cluster´s transition from CAU to MULTI+, thereby providing insights into lifestyle behavior and health outcomes of people with MI receiving CAU.
Study population
People were included if they were aged ≥16 years and had a treatment duration exceeding 10 days within one of the psychiatric wards. This time frame was pragmatically chosen to ensure that patients had sufficient exposure to treatment conditions. People were excluded if they had a limited understanding of the Dutch language or their (mental) health condition hindered informed consent.
Procedure
Data were collected during CAU, which includes pharmacological and psychological treatment, without structured lifestyle interventions. Instead, lifestyle-related activities varied between individuals or teams, depending on specific needs and available resources. Data were collected from routine screening and questionnaires. These questionnaires were administered as semi-structured interviews by trained research assistants (RAs), allowing for additional clarification when needed. We collaborated with staff across 45 wards to determine the optimal conditions for conducting the semi-structured interviews, including the best time of day and location. RAs were present for several days, approaching potential participants with support from staff. RAs received training and followed a standardized interview protocol, while weekly consensus meetings were held to ensure data quality. Participants provided verbal informed consent. This procedure was employed to visually communicate the study’s objectives and methodologies, enhancing comprehension for participants. A full description of the procedures can be found in den Bleijker et al. (2020) [Reference den Bleijker, van Schothorst, Hendriksen, Cahn, de Vries and van Harten25].
Outcomes
Demographic characteristics were obtained from the electronic patient file. Study measures and psychometric properties are outlined in Table 1, with a comprehensive description available in den Bleijker et al. (2020) [Reference den Bleijker, van Schothorst, Hendriksen, Cahn, de Vries and van Harten25]. Since lifestyle behaviors are central to our study, we included multiple nodes to capture their nuances, whereas for other variables, we used composite scores to reduce complexity while ensuring robust estimation.
Description of outcome measures and their psychometric properties
Abbreviations: SIMPAQ, Simple Physical Activity Questionnaire; SCOPA, SLEEP Scales for Outcomes in Parkinson’s disease Sleep; DP, Diastolic blood pressure; SP, Systolic blood pressure; BSI, Brief Symptom Inventory; WHOQoL-BREF, World Health Organization Quality of Life; DDD, Daily Defined Dose; ATC, Anatomical Therapeutic Chemical Classification System.
a Answering options differ between questions, such as from very poor to very good, or from not at all to extremely.
Lifestyle behaviors
Physical activity was measured with the Simple Physical Activity Questionnaire (SIMPAQ; [Reference Rosenbaum, Morell, Abdel-Baki, Ahmadpanah, Anilkumar and Baie26]), a reliable and valid tool for assessing physical activity in people with severe MI. Sleep problems were measured with the validated Scales for Outcomes in Parkinson’s disease Sleep (SCOPA SLEEP; [Reference Marinus, Visser, van Hilten, Lammers and Stiggelbout27]). We categorized smoking behavior according to the categorization of the QRISK3 algorithm [Reference Hippisley-Cox, Coupland and Brindle28], in line with the primary outcome measure of the MULTI+ trial. We used the 24-hour recall (24HR) method to measure dietary intake quality, in which foods and beverages consumed over the past 24 hours are assessed. We evaluated this according to the National Food-based Dietary Guidelines (FBDG). The “Wheel of Five” (WoF) is part of the FBDG and includes food groups associated with a reduced risk for chronic diseases [Reference Kromhout, Spaaij, De Goede and Weggemans29]. Each recalled food item was classified within or outside the WoF and ranked on a 1–3 scale (1 = below guideline, 2 = meets guideline, 3 = exceeds guideline). This (classification) method is not validated, but was reviewed by a dietitian and consensus meetings were held to improve consistency.
Physical health
We used body mass index (BMI), cholesterol ratio, and mean arterial pressure (MAP) to assess physical health. Additionally, we incorporated the Physical Quality of Life (QoL) scale from the validated World Health Organization Quality of Life-BREF (WHOQoL-BREF; [30]) to include a subjective perspective to our assessment of physical health.
Mental health
We used the Global Severity Index (GSI) from the Brief Symptom Inventory (BSI; [Reference Derogatis and Melisaratos31]) to measure symptom severity. The BSI is a validated and shorter questionnaire, which measures symptoms of psychopathology[Reference Derogatis and Melisaratos31]. To measure different domains of quality of life (QoL), the Environmental, Psychological and Social scales of the WHOQoL-BREF were included [30].
Medication
Medication use was obtained from the pharmacy’s electronic system. Prescriptions are converted into Daily Defined Dose (DDD) according to the Anatomical Therapeutic Chemical Classification System (ATC) from the World Health Organization (WHO). The DDD is a standardized unit for statistical purposes and represents the presumed average daily maintenance dosage of a drug when prescribed for its main indication [32]. For this study, we calculated the DDD for ATC codes N05A (antipsychotics) and N06A (antidepressants).
Statistical analysis
Questionnaires were processed according to their manuals. Routine screening data were checked for entry errors, which were removed. Any extreme values that were not due to errors were retained to maintain a representative view of the population.
Network construction
We estimated a Gaussian Graphical Model (GGM) incorporating all measures outlined in Table 1 as continuous variables [Reference Isvoranu and Epskamp33]. We used LASSO regularization because the number of included variables was relatively high compared to the number of observations. We opted for a hyper-tuning parameter of 0, resulting in a more lenient inclusion of edges, as our study aim is exploratory [Reference Blanken, Isvoranu and Epskamp34]. Since many variables were skewed, we used Spearman’s rank-correlation and pairwise complete observations to handle missing data [Reference Isvoranu and Epskamp33].
Visualization We used the Fruchterman–Reingold algorithm for the layout of our network [Reference Fruchterman and Reingold35]. This algorithm positions nodes with high strength and/or more connections closer to each other, and closer to the center of the network. The thickness and saturation of edges are proportional to the strength of the conditional association. Blue edges indicate a positive conditional association, while red edges indicate a negative conditional association [Reference Jones, Mair and McNally36].
Centrality analysis
We calculated strength centrality to quantify how strongly nodes were connected to other nodes in the network. Node strength is calculated by summing the absolute weighted number and strength of all edges of a node and comparing it to those of all other nodes in the network [Reference Hevey37].
Network accuracy
Before interpreting the network, we evaluated the accuracy and stability of the estimated network. We followed the bootstrap procedures as described in Epskamp et al. (2018) [Reference Epskamp, Waldorp, Mõttus and Borsboom24]. First, we examined the stability of strength centrality using a case-dropping bootstrap based on 1000 samples (re-estimating the network with a different number of observations). This method quantifies the stability of the order of strength centrality with the correlation stability coefficient (CS-coefficient). A CS coefficient of 0.7 is considered reliable. Second, we evaluated the accuracy of the edge weights. We used non-parametric bootstrapping based on 1000 samples (observations are resampled with replacement, creating new datasets). Third, we performed bootstrapped difference tests between the edge weights and the strength indices to test if these differed significantly from each other.
Statistical packages
The analyses were performed in R Statistical Software [38]. For network estimation, we used the estimateNetwork function in the bootnet R package version 1.5.3 [Reference Epskamp, Borsboom and Fried23]. Furthermore, methods for accuracy analyses are implemented in this package [Reference Epskamp, Waldorp, Mõttus and Borsboom24]. We used the qgraph R package version 1.9.5 to visualize our network [Reference Epskamp, Cramer, Waldorp, Schmittmann and Borsboom39].
Results
Patient characteristics
The study included 423 patients, of whom 42% were female and 41% had a diagnosis of schizophrenia or another psychotic disorder. The mean age was 55.5 (SD = 17.6, range=19-91), and more than half of the participants were hospitalized for more than a year. Demographic characteristics are described in Table 2. Analyses were conducted with and without extreme values. Because the results showed no substantial differences, the results including extreme values are presented.
Patient characteristics
a Item frequency varies across variables due to missing values resulting from low screening rates, and because not all patients could complete all questionnaires due to illness severity or cognitive deficits.
b Diagnoses in this category are: personality disorder, n = 22; neurocognitive disorder, n = 11; anxiety disorder, n = 7; trauma and stressor-related disorder, n = 7; somatic symptom disorder, n = 4; other, n = 5; missing, n = 5.
c The defined daily doses (DDDs) of the three most frequently prescribed antipsychotics and antidepressants are noted.
d Other antipsychotics prescribed, in order of prevalence, are: haloperidol, n = 38; aripiprazole, n = 32; risperidone, n = 28; zuclopenthixol, n = 20; amisulpride, n = 14; flupentixol, n = 12; pipamperone, n = 9; penfluridol, n = 8, paliperidone, n = 5, chlorpromazine, n = 4; pimozide, n = 4; sulpiride, n = 2.
e Other antidepressants prescribed, in order of prevalence, are: clomipramine, n = 14; paroxetine, n = 14; venlafaxine, n = 12; mirtazapine, n = 11; tranylcypromine, n = 11; fluoxetine, n = 9; sertraline, n = 8; bupropion, n = 8; fluvoxamine, n = 7; amitriptyline, n = 3; imipramine, n = 1; dusolepin, n = 1; trazodone, n = 1.
Network analysis
The network structure in Figure 1 illustrates the conditional associations among lifestyle behaviors, physical health, and mental health outcomes. Each node represents a symptom or behavior, while each edge depicts a bidirectional partial correlation between the nodes, considering all other associations in the network. The accompanying strength centrality indices are presented in Figure 2.
Graphical representation of the estimated network model, including lifestyle behaviors, physical health, and mental health, differentiated by colors. Blue edges indicate a positive conditional association, and red edges indicate a negative conditional association. The thickness and saturation of edges are proportional to the strength of the conditional association. Higher scores on overall sleep quality mean more overall sleep problems.
Centrality plot illustrating the strength of the nodes in the network depicted in Figure 1. Nodes are ordered from the node with the highest strength to the node with the lowest strength. Node strength quantifies how strongly a node is directly connected to other nodes (summing the absolute value of the edges to each node). All values are standardized, with higher values indicating more centrality.
Generally, we observe a network structure in which all nodes are connected to at least one other node in the network. The nodes with the highest strength centrality are psychological QoL (15), physical QoL (12), nighttime sleep problems (2), and overall sleep quality (1). Supplementary Figure 3 in the supplement provides an overview of the (non)significant differences between strength centrality indices.
When investigating the strength of the nodes related to lifestyle behavior, nighttime sleep problems (2) were stronger than almost half of the nodes in the network. Overall sleep quality (1) cannot be shown to be significantly different from many other nodes (see Supplementary Figure 3). A strong positive connection existed between overall sleep quality and nighttime sleep problems (1–2). Furthermore, sleep was strongly associated with physical QoL, with associations between both overall sleep quality and physical QoL (1–12) and nighttime sleep problems and physical QoL (2–12). In terms of strength, psychological QoL (15) and physical QoL (12) were statistically stronger than most of the other nodes (see Supplementary Figure 3). All QoL nodes (12, 14, 15, 16) are positively associated, indicating that higher QoL in one domain is associated with higher QoL in other domains.
Additionally, we observed strong negative associations between psychological QoL and both the daily dose of antidepressants (15–18) and Global Severity Index (15–13). This suggests that psychological QoL is probably lower when people take higher doses of antidepressants or when they experience more severe symptoms (and vice versa). Other strong associations in the network include the negative association between the percentage of healthy food intake and cholesterol ratio (5–10) and the positive association between daily doses of antipsychotics and length of hospital stay (17–20). No clear pattern of relationships emerged among other lifestyle behaviors or physical health outcomes.
Network accuracy
Results of the accuracy analyses are available in the supplement. We quantified the stability of node strength with the CS–coefficient, which indicated that node strength stability is good and that 75% of the sample can be dropped to still maintain a correlation of 0.7 with the original strength metrics as computed on the entire sample (S(cor = 0.7) = 0.75; Supplementary Figure 1). Thus, the order of the variables as indexed by strength can be interpreted. Supplementary Figure 2 shows that the edges between the strongest nodes (e.g., 1–2, 12–15, 1–12, and 2–12) were present in all of the bootstrapped samples, and differed from approximately half of the other edge weights (Supplementary Figure 4).
Sensitivity analyses
We estimated a post-hoc network excluding antipsychotic medication use (given its impact on lifestyle behavior and health outcomes) and conducted subgroup analyses for individuals aged 65 and younger, and those with schizophrenia and other psychotic disorders. Visualizations show that most of the links are similar across networks. Additionally, the correlation between edge-weight matrices is high (r = 0.81–0.93), indicating that results remain consistent across subgroups. Results are provided in Appendix 2 of the supplement. These findings support the robustness of our original findings.
Discussion
This study applied a network approach to explore the complex interrelations among lifestyle behaviors and physical and mental health outcomes in people with MI. Sleep and QoL emerged as the most central nodes, based on strength centrality. Constructing this exploratory network provides valuable insights into the importance of lifestyle behaviors, health outcomes, and their interconnectedness. This complements current evidence in which such relationships were mainly analyzed in isolation.
Sleep emerged as the most strongly connected lifestyle behavior, and results indicate that sleep and QoL are related (i.e., people with more sleep problems may have a lower QoL and vice versa). The well-established association between sleep disturbances and reduced QoL is particularly relevant for people with MI, who often experience sleep problems, affecting their physical and mental health [Reference Stafford, Oduola and Reeve40]. Furthermore, evidence is increasing for the causal role of sleep in both the onset and treatment of various mental disorders [Reference Scott, Webb, Martyn-St James, Rowse and Weich11]. Despite this, sleep is often perceived as a consequence of MI, rather than as a symptom to address. Sleep problems are often treated pharmacologically, which helps with sleep duration but negatively affects sleep quality and hinders daytime activity in the long term due to its sedative nature [Reference Gee, Orchard, Clarke, Joy, Clarke and Reynolds41]. Our findings underscore the importance of addressing sleep problems because improving sleep quality has the potential to impact other health-related outcomes in people with MI, especially QoL [Reference Freeman, Sheaves, Waite, Harvey and Harrison42].
QoL was another central node, particularly the psychological and physical domains. These domains address the intrinsic experiences of individuals, unlike the social and environmental dimensions of QoL. The strength of these nodes emphasizes the importance of internal experiences of well-being. This aligns with research recognizing the value of such patient-reported outcomes, as they provide direct insights into individuals’ perceptions of their own health and quality of life [Reference Pape, Adriaanse, Kol, van Straten and van Meijel43]. Furthermore, the strong association between psychological and physical QoL aligns with the well-documented comorbidity between physical and mental health, yet physical health is often neglected in treatment [Reference Bendayan, Kraljevic, Shaari, Das-Munshi, Leipold and Chaturvedi4]. While clinical guidelines emphasize monitoring and managing physical health risks of people with MI, adherence in clinical practice remains poor [Reference Noortman, de Winter, van Voorst, Cahn and Deenik44]. Our results highlight the importance of perceived psychological and physical health and its potential impact on other health-related outcomes.
Contrary to prior research on the relationship between lifestyle behaviors and health outcomes, physical activity, nutrition, and smoking did not emerge as central nodes in our network. One possible explanation lies in methodological factors: the distribution of physical activity was highly skewed, potentially limiting its role in the network; smoking was categorized as a five-level variable, reducing variability; and nutrition was measured using a non-validated method, which may have introduced measurement errors. However, another relevant possibility is that sleep simply plays a more dominant role in this network than other lifestyle behaviors. Sleep is known to affect mood, cognition, and self-regulation, all of which are crucial for maintaining other healthy behaviors [Reference Fortier-Brochu, Beaulieu-Bonneau, Ivers and Morin45–Reference Tomaso, Johnson and Nelson47]. This suggests that sleep may be a key factor in improving other lifestyle behaviors, rather than these behaviors independently driving health outcomes. In the context of network analysis, this does not necessarily imply that physical activity, nutrition, or smoking are unimportant, but rather that sleep plays a more central role.
Beyond the centrality of sleep and QoL, several other noteworthy associations were observed. A positive association was found between the percentage of healthy food intake and cholesterol ratio, aligning with existing research in the general population [Reference Schoeneck and Iggman48]. However, research on this relation remains limited in people with MI, and disrupted cholesterol levels can also be influenced by hereditary factors and psychotropic medication [Reference Douglas and Nasrallah49]. While our findings suggest a potential link between healthier dietary intake and cholesterol ratio, this estimate was unstable, and more research is needed to investigate this link. Further, the association between the use of antipsychotics and the duration of admission may be explained by the higher illness severity in people with psychotic disorders, who are more frequently and longer hospitalized compared to other psychiatric populations [Reference Carranza Navarro, Álvarez Villalobos, Contreras Muñoz, Guerrero Medrano, Tamayo Rodríguez and Saucedo Uribe50]. However, medication effects are complex, and more in-depth analyses of the underlying mechanisms of medication effects were beyond the scope of this analysis. It would be a valuable direction for future research to further explore these interdependencies, providing a more comprehensive understanding of the role of medication in an interconnected network of health behaviors.
Limitations
Several limitations affect the interpretation of our results. First, when two nodes are strongly connected, they may measure the same underlying construct (topological overlap), with the risk of misinterpretation of the network structure [Reference Fried and Cramer51]. In our network, this concern arises in the association between psychological QoL and physical QoL, as well as between quality of sleep and nighttime sleepiness, as they originate from the same questionnaire. However, these constructs represent distinct domains within a validated questionnaire. Furthermore, results showed that the association between these domains was stable. Another limitation is missing data. The use of routine screening data helped reduce participant burden but also resulted in missing values due to low screening rates. Additionally, not all participants could complete all questionnaires due to illness severity or cognitive deficits. To account for missing values, we used the pairwise complete observations integrated in the Bootnet package to estimate a GGM. Finally, skewed variables could have affected the stability of our results.
Clinical implications
Given the central role of sleep, addressing sleep disturbances in treatment may not only improve sleep quality but also positively impact QoL. This can be done through Cognitive Behavioral Therapy for Insomnia, an effective first-line treatment for people with MI that has demonstrated beneficial effects [Reference Hertenstein, Trinca, Wunderlin, Schneider, Züst and Fehér52]. Furthermore, the centrality of physical QoL underscores the need for better physical health management, especially given the health disparities of people with MI. Likewise, the central role of psychological and physical QoL emphasizes their importance in the health status of people with MI. While this study is cross-sectional, it underscores the need to prioritize sleep and QoL in both clinical practice and research.
Conclusion and future research
This study provides a novel perspective on the interplay between lifestyle behaviors and physical and mental health outcomes in people with MI. Our findings highlight the central role of sleep and QoL in this network, suggesting that sleep disturbances are important to address in treatment. Building on these results, future research could focus on testing specific (causal) pathways through methods such as mediation analysis or network intervention analysis. For instance, by exploring whether improving sleep as a key lifestyle behavior could enhance quality of life and activate other health outcomes. These approaches would offer a deeper understanding of the mechanisms at play, which was beyond the scope of the current study. Additionally, our findings show the importance of internal experiences of QoL. Given their interconnected nature, we advocate for a holistic therapeutic approach, taking the reciprocal influence of lifestyle behavior and physical and mental health into account to improve the treatment of people with MI.
Abbreviations
- 24HR
-
24-hour recall
- ATC
-
Anatomical Therapeutic Chemical Classification System
- bCI
-
bootstrapped Confidence Intervals
- BMI
-
Body Mass Index
- BSI
-
Brief Symptom Inventory
- CS-coefficient
-
Correlation Stability coefficient
- DDD
-
Daily Defined Dose
- FBDG
-
the National food-based dietary guidelines
- GGM
-
Gaussian Graphical Model
- GSI
-
Global Severity Index
- LASSO
-
Least absolute shrinkage and selection operator
- MAP
-
Mean Arterial Pressure
- MI
-
Mental illness
- MULTI+
-
multidisciplinary lifestyle focused approach in the treatment of inpatients with mental illness
- QoL
-
Quality of Life
- SCOPA SLEEP
-
Scales for Outcomes in Parkinson’s Disease Sleep
- SIMPAQ
-
Simple Physical Activity Questionnaire
- WHOQoL-BREF
-
World Health Organization Quality of Life-BREF
- WoF
-
Wheel of Five
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1192/j.eurpsy.2025.2442.
Data availability statement
Due to the strict regulations and its sensitive nature, supporting data cannot be made openly available.
Acknowledgements
The authors are grateful to all patients and mental healthcare professionals of GGz Centraal for their time and efforts dedicated to participating in this study. We would also like to thank our research assistants for their contribution to data collection. Additionally, as non-native English speakers, we used Large Language Models (LLM) for AI assisted copy editing (AI-assisted improvements to human-generated texts for readability and style, and to ensure that the texts are free of errors in grammar, spelling, punctuation, and tone). The authors take full accountability for the work presented.
Financial support
This study received funding from “ Stichting tot Steun Vereniging tot Christelijke Verzorging van Geestes- en Zenuwzieken”, grant number 277. The funding organization had no involvement in the study’s design, data collection, analysis, interpretation, or manuscript preparation.
Competing interests
None of the authors have any competing interests
Trial registration
ClinicalTrials.gov registration. Identifier: NCT04922749. Retrospectively registered 3rd of June 2021.
Open access
Comments
No Comments have been published for this article.