Clinical trajectories of multimorbidity patterns over time

To date, around 31% of Dutch adults and 86% of Dutch elderly aged 75 or more are multimorbid ("Aantal mensen met multimorbiditeit bekend bij de huisarts," 2021). Thereby, multimorbidity is defined as having two or more chronic conditions, where each chronic condition is either a physical non-communicable or an infectious disease of long duration, or a mental health condition of long duration (The Academy of Medical Sciences, 2018). The high prevalence of multimorbidity challenges our healthcare system, as patients with multimorbidity have been indicated as high cost patients with a higher healthcare utilization and higher risk of poor health outcomes (Tanke et al., 2019; The Academy of Medical Sciences, 2018).
Especially the combination of somatic and mental health conditions leads to poor health outcomes and excess mortality, which might be related to the undertreatment of somatic conditions in patients with psychiatric disorders (Lawrence, Kisely, & Pais, 2010; Thornicroft et al., 2009; Viron & Stern, 2010). At the same time, this –often younger- group of patients with somatic-mental multimorbidity needs psychological support for a longer time than patients with only psychological complaints (Barnett et al., 2012; Bobo et al., 2016; Harber-Aschan, Hotopf, Brown, Henderson, & Hatch, 2019). In addition, multimorbidity has been linked to frailty, which in turn is associated with higher mortality independent of the mortality risk of multimorbidity (Hanlon et al., 2018; Theou, Rockwood, Mitnitski, & Rockwood, 2012).  
In addition, AF, which is the most common sustained cardia arrhythmia in clinical practice and has an estimated prevalence of 33.5 million affected individuals in 2010, is rarely without presence of other comorbidities (Andersson et al., 2013; Chamberlain et al., 2017; Chugh et al., 2014). Therefore, multimorbidity is very common among middle-aged to older AF patients, with about 80% of participant having at least one other co-morbid condition (Jani et al., 2018). Atrial fibrillation is associated with a higher risk of thromboembolic events, heart failure and death. Furthermore, it is, especially in combination with multimorbidity, associated with a high increased mortality risk and a significant impact on healthcare costs  (Blomstrom Lundqvist, Lip, & Kirchhof, 2011; Jani et al., 2018; Ruddox et al., 2017; Stewart, Murphy, Walker, McGuire, & McMurray, 2004; Wolf, Dawber, Thomas, & Kannel, 1978).
Part of the higher healthcare utilization and costs of multimorbid patients is due to the fragmented care that patients receive, resulting into more hospital admissions, general practitioner contacts, and inappropriate polypharmacy (Hopman, Heins, Rijken, & Schellevis, 2015; Tommelein et al., 2015). Therefore, a more patient-centred treatment for multimorbid patient has been recommended (NICE, 2016; van der Heide et al., 2015). To date, however, its implementation falls short as health care systems are mostly organised around single diseases or organ system (Whitty et al., 2020). 
To identify patients that are likely to profit most from a patient-centred care coordination, the importance of longitudinal cohort studies investigating multimorbidity profiles within a given population has been highlighted (Busija, Lim, Szoeke, Sanders, & McCabe, 2019; Prados-Torres, Calderón-Larrañaga, Hancco-Saavedra, Poblador-Plou, & van den Akker, 2014; Xu, Mishra, & Jones, 2017). Cross-sectional studies are not suitable to study disease progression and large long-term prospective studies are needed to study the prevalence, patterns, and risk factor of multimorbidity in-depth (France et al., 2012; Hassaine, Salimi-Khorshidi, Canoy, & Rahimi, 2020; Xu et al., 2017). In addition, the identification of patterns of accumulation of conditions and the change of prevalence of multimorbidity over time have been identified as a research priority (The Academy of Medical Sciences, 2018; US Department of Health Human Services, 2010). 
Previous literature using longitudinal data has focused on common risk factors for the development of multimorbidity.  Xu, Mishra, Dobson, and Jones (2018) identified obesity, hypertension, physical inactivity, smoking and having other chronic conditions as risk factors for multimorbidity. Similarly, Shang, Peng, Wu, He, and Zhang (2020) found that obesity, smoking, but also old age, lower socio-economic status and diets high in red meat and chicken were risk factors for the development of multimorbidity. Furthermore, other longitudinal data analyses calculated the incidence of multimorbidity to study the development of the second or third chronic condition in the chronic condition dyads and triads, respectively.  They emphasize, however, the need for studies to understand how additional chronic conditions accumulate over time in order to prevent or delay the onset of a third or fourth condition (Nicholson et al., 2019; St Sauver et al., 2015). A description of such longitudinal patterns can contribute to a learning healthcare system which supports both the management and prevention of multimorbidity. 
Nicholson et al. (2019) described the unique permutations of multimorbidity, being “the same chronic conditions with a specific sequence”. Among females the most common permutation was anxiety/depression followed by obesity and for males musculoskeletal problems followed by obesity. Furthermore, Ashworth et al. (2019) aimed to define sociodemographic and cardiovascular risk factors associated with the acquisition of multimorbidity and to determine the acquisition sequence of multimorbidity. They identified diabetes and depression as most common starting conditions, with coronary heart disease being more common as starting condition than depression in the least socio-economically deprived quintile. Furthermore, serious mental illness and depression where more common as starting condition in those aged under 65 followed by chronic pain, morbid obesity and diabetes as common second or third conditions. 
In contrast to Ashworth et al. (2019) and Nicholson et al. (2019), two recent studies by Vetrano et al. (2020) and Violán et al. (2020) analysed the change of multimorbidity clusters over time instead of the accumulation of multimorbidity over time.  Violán et al. (2020) employed Hidden Markov Models to analyse five-year trajectories of multimorbidity patterns and found that most participants stayed in the same cluster and the majority of patients belonged to a “non-specific pattern”. Vetrano et al. (2020) used a fuzzy c-means clustering algorithm and compared resulting clusters from baseline with clusters from the six- and twelve-year follow-ups. Similarly, the majority of patients belonged to an “unspecific cluster”. In addition, they identified that these patients moved mostly to, amongst others, the heart and vascular disease clusters at the 6- and 12-year follow ups. These clusters, in return, were associated with a higher six-year mortality at both follow-ups.  
Nevertheless, both Violán et al. (2020) and Vetrano et al. (2020) point out that there is scarce evidence on how multimorbidity clusters change over time. The emphasize that the repetition of similar analyses in a broader population with a longer follow-up is needed to better characterize the change of multimorbidity patterns over time. Thereby, the longitudinal analyses of multimorbidity patterns can help to predict patterns of patients, to understand the complexity of clusters, and support the clinical and organisational management of this complex population. Furthermore, the analyses of multimorbidity clusters over time can identify patients “at a high risk of progressing to severe disease clusters with worse prognosis” (Vetrano et al., 2020, p. 7). 
Therefore, the aim of this research is to analyse clinical trajectories of multimorbidity over time in a large Dutch population-based cohort study. Thereby, we want to build on the work by Violán et al. (2020) and Vetrano et al. (2020) and further characterize the change of multimorbidity patterns over time. We want to include multimorbid adult patients of all ages instead of solely focusing on elderly and aim to compare multimorbidity clusters at the two follow-up assessment of the LifeLines cohort study. 
Insights into the development of multimorbidity trajectories over time can inform the organisation of theme-based outpatient clinics or can identify the need of multidisciplinary consultation at an early stage to prevent or delay the transition to disease clusters with worse prognosis.  Furthermore, knowledge about the association between multimorbidity patterns with psychopathology, and changes in psychological well-being over time can help to determine if and when psychological care for patients with both somatic and mental diseases should be offered. The identification of patterns that are associated with functional decline can, moreover,  help to determine when a more holistic care approach should be initiated to prevent worse health outcomes in the future. Further, as AF patients with higher amounts of co-morbidities have a great increased risk of mortality, more insight into patterns of multimorbidity in patients with (prevalent) AF is needed to optimize treatment of AF and its associated conditions.