According to the World Health Organization, multimorbidity is characterized as the coexistence of 2 or more chronic conditions in a single individual. These chronic conditions are often long-term health conditions that require complex and ongoing care [1]. Chronic diseases include both mental and physical health conditions [2]. In the context of multimorbidity, no single chronic condition is necessarily more central than the others, as this could result in designating one condition as the index condition, where a more appropriate term would be comorbidity [2-4].

On a global scale, approximately one-third of adults, and over half of all adults with any chronic condition, already experience multimorbidity [5]. The occurrence of multimorbidity increases with age [5-7], with projections showing a doubling of individuals aged 60 years and older globally before 2050; the number of persons with multimorbidity is also expected to rise dramatically [5,8].

In Denmark, a report from 1 of the 5 regions responsible for delivering health care (Region Zealand) concludes that in 2022, there were approximately 1.2 million people with multimorbidity in Denmark, which corresponds to 26% of the population [9]. This number is expected to increase by 1.4% annually to 1.5 million in 2050 if no efforts are made to improve the population’s health [9].

Multimorbidity is linked to an increased likelihood of premature death [10,11]. This risk escalates with both the quantity of conditions and the particular combinations of chronic diseases [10,11]. Multimorbidity also lowers quality of life, which gets worse as you have more chronic diseases [11-14]. It disproportionately affects individuals with low socioeconomic status, contributing to increased health inequality [5,15-17], and is also associated with more mental symptoms and a greater perception of fragmented care [10,11]. Multimorbidity is also responsible for higher health care expenditure due to longer hospital stays and larger medication consumption [12,18].

There exist several risk factors associated with the development of multimorbidity including sociodemographic factors (eg, age, household income, and education) [5,14], biomedical risks (eg, overweight, high blood pressure, high cholesterol, and genetic predispositions), and health behaviors (eg, physical inactivity, poor nutrition, smoking, and alcohol consumption) [19,20]. Additionally, certain chronic conditions may share common risk factors, or one disease may be a risk factor for another [21].

However, in contrast to patients with single diseases, there is a complex relationship between risk factors and multimorbidity [22]. For example, according to a report from the Australian government [23], there is a correlation between the number of chronic conditions a person has and the likelihood of having multiple risk factors [23].

Nonetheless, without the information regarding the duration of an individual’s exposure to risk factors and the onset of their chronic condition, attributing their chronic conditions solely to the number of risk factors is not possible. This is due to the onset of some chronic conditions possibly motivating people to change their behavior for the better. For instance, when a person is diagnosed with chronic obstructive pulmonary disease (COPD), it may inspire them to quit smoking and thereby reduce the risk of exacerbating COPD and cancer. In contrast, a person diagnosed with COPD will not be motivated to engage in physical activity, which would then increase the risk of, for example, diabetes [19-21].

As a result of the complexity of the risk factors, clinical decisions for patients with multimorbidity are a complicated and challenging task since the health care system is primarily designed to manage patients with a single disease [24].

Understanding the patterns and factors associated with multimorbidity, particularly the modifiable factors, can help contribute to the prevention and treatment of multiple chronic diseases [24,25]. Predictive analytics, such as machine learning, has the potential to solve these kinds of problems [26]. It enables more advanced analysis, automation, and recognition of unidentified patterns [27]. Machine learning enables systems to learn and improve from experience without explicit programming. It involves the process of fitting one or more statistical models to a given data set that contains both explanatory variables and an outcome variable. The models provide coefficients that quantify the relationship between the explanatory variables and the outcome. These model coefficients are applied to predict the same outcome with the same explanatory variables but on new, unseen data. The model’s ability to predict the outcome depends on its capacity to recognize and apply statistical patterns found in the data while it is being trained [27,28].

In health care, machine learning has demonstrated improvements in various tasks, including drug discovery [29], histopathological diagnosis [30], brain magnetic resonance imaging segmentation [31], and disease prediction using electronic health records [32,33]. In the context of machine learning for multimorbidity, there are numerous options and choices when developing an algorithm that are relevant. Combinations of disease groups have the potential to uncover new mechanisms of diseases, develop treatments, develop multidisease clinical strategies, meet the patient’s needs, and manage polypharmacy [34,35]. For instance, Prados-Torres et al [36] have identified 3 common multimorbidity clusters: cardiovascular and metabolic diseases, mental health problems, and musculoskeletal disorders.

Different outcomes have also been addressed in the literature, such as predicting a new disease for patients with multimorbidity [25,37,38], identifying factors for multimorbidity [39,40], predicting rehospitalization [41], and developing a multimorbidity frailty index [42]. Furthermore, a variety of different machine learning models have been used, for example, logistic regression [37,39,40], random forest [25,41,42], neural network [37,40], and network model [38]. It is also evident that a wide range of features have been applied in developing these machine learning algorithms. These features include sociodemographic data [25,41,42], electronic health record data [39,41], self-reported data [40], and medical data [38,39,42], among others.

The literature indicates ongoing research and a growing interest in the field of multimorbidity. However, the type and amount of evidence remain unclear as well as potential gaps in the literature [25]. Therefore, a scoping review was chosen to identify and map available evidence within this field.

The objective of this scoping review is to explore the existing literature regarding the use of machine learning in the context of multimorbidity. Specifically, it seeks to identify the machine learning models that have been used, the patient groups that have been the focus in various studies, the features, types of input data, the maturity of the algorithms, and outcomes that have been in focus. In doing so, any potential gaps in research or knowledge can be identified.

Ethics approval is not required, as this scoping review does not involve primary data collection from human participants.

The development of this protocol has followed the guidelines from Joanna Briggs Institute [43]. The conduct and reporting of this scoping review will follow the guidelines of the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) [44]. Furthermore, the review will follow the methodological framework for scoping reviews outlined by Peters et al [45], which provides guidance in the development and execution of a scoping review.

Multimorbidity is a global challenge for the health care system since it is associated with increased mortality, reduced quality of life, and increased health care costs. With the increasing prevalence, the problem is only growing. Machine learning has the potential to address these challenges, as these methods can understand the patterns and factors associated with multimorbidity. This scoping review aims to explore and synthesize the current literature addressing the use of machine learning in the context of patients with multimorbidity.

Following the definition of a scoping study established by Arksey and O’Malley [47], this scoping review serves 2 primary objectives. It aims to compile and communicate research findings to policy makers, practitioners, and consumers. Besides that, it will also pinpoint any existing research gaps within the literature [47]. The available literature indicates ongoing research into the application of machine learning within the context of multimorbidity. Unfortunately, there is uncertainty regarding the type and quantity of evidence as well as potential knowledge gaps. To our knowledge, this may be the first scoping review in this field to uncover this.

The insights gained from this scoping review can be valuable for both researchers and health care practitioners. For researchers, understanding the current landscape of machine learning methods can provide a foundation for future studies by stimulating new research questions and help to identify research gaps. Health care practitioners can gather awareness of the current state of machine learning use in multimorbidity research. Overall, the findings from this scoping review can contribute to more precise interventions, ultimately improving patient outcomes.

Nevertheless, the scoping review may have some limitations regarding the search strategy. The scoping review follows the PCC framework, where including the last block (context) may potentially constrain the search outcomes. A potential solution is only to use the first 2 blocks (participants and concept), although this approach may introduce more noise into the search results. However, this might not pose an issue, as the last block (context) is comprehensive due to the included search words within the block.

A particular challenge in this field is the interchangeable use of the terms “multimorbidity” and “comorbidity” in the literature. To address this issue, the participant search block in the PCC framework contains synonyms for both terms. This ensures the retrieval of relevant studies and accounts for potential variations in terminology across the literature. Besides that, it also captures a broader spectrum of research related to patients with multiple chronic conditions. A disadvantage of this approach is the potential for more noise and irrelevant results.

Whether there is a need for quality assessment of the included studies can be discussed. Typically, this is not done in a scoping review, as the objective is to provide an overview of the evidence rather than a critically appraised and synthesized response to a specific question [48-50]. On the other hand, using tools such as the Prediction Model Risk of Bias Assessment, which is designed for systematic reviews, could provide insights into the quality of prediction models [51]. Being aware of this, this study will follow the typical approach for conducting scoping reviews, and thereby not conduct a quality appraisal of the included studies.

The primary objective of this scoping review is to investigate the existing literature on machine learning in patients with multiple chronic conditions and potentially highlight research gaps. To our knowledge, this scoping review marks the initial exploration into the application of machine learning in multimorbidity research.

The outcome of the searches will be presented following the PRISMA-ScR checklist and flow diagram, which will provide a comprehensive overview of the search process, including the number of identified studies, duplicates removed, screened studies, excluded studies, and studies included in the review. A conclusive summary based on the results of the scoping review will be presented, and recommendations for future research will be suggested if any possible knowledge gaps are identified. The potential findings of the scoping review will be reported in an academic paper, which will be submitted to a peer-reviewed journal and presented at a national conference.