Our inclusion criteria were as follow:

Our exclusion criteria were:

The search was performed by B.R.-F in the databases PubMed/MedLine, Embase, Ovid MEDLINE, PsycINFO, Web of Science, Scopus, and CINAHL Complete on January, 2022. Google Scholar was used to supplement the main databases. This search was repeated in July 2022. The search strategy for this review was performed by combining three sets of keywords (multimorbidity, social-behavioral determinants, and patterns) to search for relevant literature using Boolean operators. The search strategy, which can be found in Supplementary Table 1, was adapted to each database. The references obtained from the search strategy were exported to Microsoft Excel 2019 for deduplication of the results and screening.

The quality of the studies was evaluated using Axis Tool quality assessment (41), a critical appraisal tool to systematically assess the methodological quality of selected studies. This tool consists of 20 questions with expanded explanations that are assessed with yes/no/don't know (for more information see the Supplementary Table 2). These questions evaluate general aspects of the research and reporting, such as the introduction, methodology, and quality of the discussion. The purpose of these questions is to guarantee the quality of the selected studies. The tool was implemented by two reviewers (BR-F, EO-M) that, in the case of disagreement, solved the possible discrepancies using the scores of a third reviewer. The three reviewers jointly analyzed the articles with the lowest methodological score (in our case, two articles with a score of 16/20). After a full evaluation, these articles were retained by the agreement of the reviewers.

First, titles and abstracts were examined by four reviewers (JC-B, EO-M, CP-M, BR-F). Subsequently, the full texts of the retrieved articles were reviewed according to the eligibility criteria. For data extraction from the selected articles, tables created for their organization were used. The information extracted from each publication was: author, title, year of publication, abstract, objective, conclusions, method of pattern extraction, multimorbidity patterns, social-behavioral factors, and population characteristics. Due to the breadth and variety of results the data were divided into different tables. The first table contained the identifying data of the articles (title, authors, and year) (Supplementary Table 3) as well as the summarized content of the article (objectives, topics, and conclusions). The second table compiled the pattern extraction methods and the results obtained, such as multimorbidity patterns, associated social-behavioral factors and the characteristics of the population studied (Supplementary Table 4). The third table combined the articles presenting each social-behavioral determinant according to the pattern extraction method (Supplementary Table 5). For the data analysis, a narrative synthesis and summary of the different methods of identifying patterns and their association with different social determinants were described in order the identify the work done so far and knowledge gaps in this field.

A broad set studies detected a stronger association of mental and musculoskeletal multimorbidity with female sex, including neurological diseases (64, 130), musculoskeletal and osteoarticular (42, 64, 67, 71, 93, 102, 126, 130, 131); and common mental disorders (42, 132), like anxiety (44, 67, 86, 93, 114) depression (43, 67, 86, 93, 114, 119), dementia (85) and emotional disorders (47). While cardiometabolic patterns (53, 71, 73, 78, 82, 90, 94, 114, 123), musculoskeletal (53, 96), digestive and genitourinary patterns (99, 110, 124, 131) were more frequent in men. As expected, a positive association between multimorbidity and age was found (67, 74, 105, 110, 132, 133). Although it was also possible to find multimorbidity among the young population, this group was usually classified in patterns called “relatively healthy” (44, 50, 60, 70, 78, 90, 125). The more prevalent multimorbidity patterns in older population were: cardiovascular (44, 58, 60, 71, 72, 81, 86), cardiometabolic (61, 73, 76, 94), musculoskeletal (81, 86, 94), cognitive impairment or dementia-related (60, 86), arthritis-cataracts (53, 58, 72, 87) and respiratory-mental-articular (56, 61). Mental health patterns linked to anxiety and depression were more frequent in younger people (44, 47, 50, 58, 86). However, multimorbidity patterns linked to dementia and cognitive impairment increase significantly with age (111, 128).

In several studies, being married was a protective factor against multimorbidity. In fact, those married were commonly classified in the relatively healthy class (55, 60, 66, 70, 72, 78, 97), while being divorced, separated, or cohabiting couples were more likely to be linked to mental health multimorbidity patterns such as the denominated “depression cluster” (66) and divorced people were more likely to be in the “psychiatric class” (97). In contrast, single people were more likely to belong to the complex multimorbidity pattern (42, 59). Widowers were more likely to belong to dementia, geriatric or neurodegenerative disease classes (78, 128), depression (66) and cardiorespiratory or musculoskeletal patterns (56).

Positive associations were found between household income and multimorbidity (107, 133). In particular, people with lower income presented a higher risk of being in the depression-arthritis (58), depression-chronic pain (48), respiratory (107), and mental health patterns (138). Overall, they were relatively healthy (52, 74) even though having multimorbidity. Conversely, having a low income was associated with a higher probability of belonging to the cardiometabolic pattern (53, 60, 65, 76), suffering musculoskeletal disease (sometimes combined with mental health conditions) (53, 107), and respiratory patterns (53, 76). The proportion of patients with multimorbidity who had physical-mental comorbidities increased substantially with greater socioeconomic deprivation (132, 138). In particular, high-income groups were less likely to present comorbidities such as depression-arthritis (58), depression-chronic pain (48), rhinitis-allergies (107), heart-related diseases or belonging to severe (or complex) multimorbidity classes (44). In general, the higher their income, the more likely they were to belong to the “relatively healthy” class (52, 74). Conversely, having a low income was found to be associated with a higher likelihood of belonging to the cardiometabolic pattern (60), depression-insomnia and musculoskeletal-mental-functional (107), cardiopulmonary and cerebrovascular (65), arthritis, asthma, allergy, depression, and thyroid-related diseases (76), cardiorespiratory-arthritis-cataracts and metabolic classes (63). They were also associated with classes linked to alcohol and substance abuse problems (43). In general, having a low income increased the likelihood of belonging to the classes with the highest rate of disease and mortality (44, 48, 49). In parallel, chronic disease in adulthood tended to increase the risk of poverty (104). Material deprivation was a risk factor for all multimorbidity patterns (59).

Different studies show that low educational levels increase the risk of suffering multimorbidity (42, 70, 75). Low education is associated with cardiometabolic (42, 60), musculoskeletal (75), respiratory (42) and mental health multimorbidity patterns (58), which are commonly organized around depressive disorders. On the other hand, high education was found to be a protective factor among the severely impaired population, metabolic and joint-COPD-ulcer patterns in the study by Zacarías-Pons et al. (59). However, a high educational level also presented a positive association with mental health patterns (specifically, psychiatric disorders and depression) (58). Similarly, being a woman with good education and high income was mainly related to the anxiety-related classes (44). Patterns were also found to be associated with low socioeconomic status (i.e., low education level and low income), such as cardiometabolic (76, 104), metabolic (74), or respiratory-mental-musculoskeletal ones (103). Low educational level and difficulty managing income were also significantly associated with a higher likelihood of having psychosomatic and musculoskeletal patterns (108).

Concerning occupational status, never working increased the risk of belonging to the cardiometabolic class (108) and being unemployed of being classified in multimorbidity patterns linked to substance use (i.e., alcohol, tobacco, and other drugs) (77). Similarly, non-working (42), being unemployed, or pre-retired (70) increased the odds of belonging to any multimorbidity class except the respiratory ones (i.e., characterized by asthma-allergy conditions). Civil servants presented a higher prevalence of musculoskeletal patterns (75) similar to manual workers, which even had a higher probability of belonging to the cognitive and sensory impairment-related classes (97). Also, the vascular-inflammatory class was associated with a low educational level and unemployment (74).

In terms of ethnicity, most studies assessed were focused on Caucasian population (44, 47, 48, 52, 60, 86, 123, 128). Race or ethnicity was found to be associated with certain chronic diseases. Van Cleave et al. (65) found a higher prevalence of cardiopulmonary and cerebrovascular patterns among black men, while Bisquera et al. (86) found that pain and liver disease conditions were also most probably linked with this population group. On the other hand, Janssen et al. (128) found associations with liver patterns and other physical injuries among Native Americans. Kalgotra et al. (123) found associations between endocrine, nutritional, immune, injury, and poisoning-related diseases. The Asian population was linked to cerebrovascular patterns and gastrointestinal cancer (128), and the Caucasian population was commonly associated with mental and musculoskeletal patterns (123).

Regarding migration status, Diaz et al. (109) found that the patterns among Norwegians and immigrants from Western Europe and North America were similar. However, multimorbidity patterns were not the same when compared with other immigrant groups. Janssen et al. (128) found that cardiovascular, cancer and geriatric patterns were more prevalent among local US residents, while urological, gastrointestinal, or respiratory disease patterns were more frequent among non-residents.

Risk behaviors such as alcohol consumption or smoking increased the risk of belonging to all multimorbidity classes, especially in men (54, 79, 108), and also in complex mental health patterns in young men (138). However, several articles report that smoking is also a risk factor for the cardiometabolic class (60, 73), which is associated with other determinants such as less education, lower income and related overweight (60) or higher waist circumference, physical activity, and a history of asthma or heart attack (73). In the study by Barile et al. (68), the mental health class was the most likely to smoke currently. In fact, 46% of people aged 18–44 in the depression class were regular smokers (43). Smoking was a risk factor only for the respiratory-mental-articular class (61). Other studies found that smoking increased the risk for the musculoskeletal class (75, 108). And, as expected, being a current or former smoker also increases the risk of belonging to respiratory multimorbidity patterns (108). Simoes et al. (67) found a direct association between smoking and obesity in the respiratory class. In the cardiorespiratory-arthritis-cataract pattern, the probability of membership increased with increasing age and being a smoker (53).

The relationship between alcohol consumption and multimorbidity was mediated by the age groups and related chronic conditions. Hernández et al. (49) found variations by the country for the relationship between alcohol and multimorbidity patterns. In the USA, the association was negative for all patterns. In contrast, in Canada and England, it was harmful to the patterns of high disease probability and metabolic and arthritis. Finally, positive associations were found in certain patterns such as osteoporosis, arthritis, hypertension, and metabolic in Ireland. According to the study by Jackson et al. (108), people who did not drink alcohol were more likely to belong to the cardiometabolic pattern. Specifically, 50% of people with complex cardiometabolic and cognitive impairment patterns were abstainers. This figure increased in the respiratory patterns where the number of moderate-strong drinkers was highest, while the younger heaviest drinking group were also the healthy subjects (60).

Physical activity was a protective factor for the respiratory-mental-articular class (61). However, physical activity significantly decreased in several groups having certain chronic conditions such as diabetes (75), respiratory and vascular-inflammatory (74). Likewise, some studies (58, 59) found more frequent intense physical activity among relatively healthy classes. On the contrary, lower physical activity was more characteristic in population groups with cardiovascular, musculoskeletal and mental health conditions (68, 97).

Although we did not find studies that analyzed the impact and/or relationship of diet with multimorbidity patterns, a clear association was observed between BMI and the subsequent occurrence of chronic comorbidities. For this reason, we used this indicator as a proxy variable for diet. In particular, obesity increases the risk of belonging to all multimorbidity classes (67, 75). Overweight and obesity were found to be risk factors for musculoskeletal or severe impairment (44) and cardiometabolic (108). Similarly, patterns of physical and mental conditions (68), complex metabolic, respiratory, and age-related conditions (60) were also associated with an increased risk of obesity. Depression was strongly associated with overweight or obesity among females (67).

Regarding the type of area of residence, we found that multimorbidity was more prevalent in urban areas (98). For instance, the cardiometabolic burden was lower in rural areas. Territorial characteristics such as the level of crime and violence in the area were also found to be associated with a pattern of chronic pain and respiratory diseases in men (101). In these areas of higher economic deprivation and lower socioeconomic status (i.e., lower household income and lower education levels), cardiometabolic, respiratory, mental and musculoskeletal patterns were more prevalent (104). Consequently, the possible effect of the area of residence over multimorbidity patterns was linked to specific socioeconomic conditions of the groups. In any case, the existing literature does not provide a clear characterization of either the social mechanisms or the ecological effects involved in the association between the local area of residence and multimorbidity patterns.

Being insured at the medical level presented important variations depending on the different countries and the respective health systems, so no clear association pattern was detected (74, 125). For instance, in the US, people with Medicaid and an additional private insurance were more likely to be found in the oncology and neurological patterns (45). Regarding the use of health services, in the study by Craig et al. (74), all classes of multimorbidity were significantly associated with higher use of health services. Similarly, patterns associated with cardiometabolic complications in pregnancy, vascular and cancer were directly related to unplanned readmissions (121). More specifically, Buja et al. (57) found that patterns such as cardiac, metabolic-ischaemic, cardiorespiratory, or cancer impairment were more at risk of hospitalization than the neurological class. Patterns linked to substance use were also associated with having more than three emergency department visits (77). Egan et al. (79) associated more hospitalizations in congestive heart failure (35%), neurological (27%), and chronic kidney-related disease patterns (24%).

In the study of 2020 by Marengoni et al. (97), the patterns of cognitive and sensory impairment had the highest percentage of people with a poor (or reduced) social network and inadequate physical activity levels. In addition, a large network of friends and having social support were found to reduce the risk of presenting patterns of degenerative-digestive diseases while increasing adherence to medications (62). Moreover, in the same way that social support has been identified as a protective factor against chronicity, the suicide of a family member has also been identified as a risk factor in the development and consolidation of severe patterns of mental multimorbidity (138). On the other hand, the cardiometabolic and respiratory-mental-articular patterns between the ages of 50 and 64 years were associated with a greater presence of feelings of loneliness, limitations in activities of daily living, and a worse state of health (61). On the other hand, the probability of suffering from degenerative-digestive diseases seemed to increase in patients with poor sleep quality and an unbalanced diet, despite having a large family network and a medium-high income (62).

Poor quality of life was associated with belonging to the complex cardiometabolic and respiratory classes (42), and activities of daily living were also affected to a greater extent by specific patterns such as cardiovascular, respiratory, mental and articular (56, 61).

With the intention of reducing the diversity of multimorbidity patterns that we have found in the different studies analyzed up to this point, the most prevalent and relevant profiles were synthesize around nine global classes (in decreasing order): (1) Cardiovascular; (2) Musculoskeletal; (3) Mental; (4) Respiratory; (5) Complex (characterized by many simultaneous chronic diseases or complicated multimorbidity); (6) Cancer; (7) Metabolic; (8) Neurological; and (9) Relatively Healthy (i.e., those who present minor or less serious chronic conditions). Figure 2 describes the number of relationships identified between the most frequent patterns and the respective social determinants (negative values denote that an inverse statistical association was identified).

Although one of our initial objectives was to be able to concisely compare the multimorbidity patterns identified in each investigation, the high heterogeneity of the data and classification techniques used, in addition to the fact that there is no general nomenclature to define the patterns, has not allowed us to synthesize the findings in a generic way, nor to estimate the prevalence of the different patterns. Due to the diversity of articles on multimorbidity and the variety of ways of measuring this health condition and its associated terms, it was difficult to identify an integrative definition of the concept that covered the full spectrum and complexity of mental and physical chronic conditions. Consequently, the pattern extraction methodology was diverse among the selected articles. In the grouping into five categories, the most frequent method was the LCA technique, and the least frequent was expert opinion, which seems to be a classification criterion that is progressively disappearing as new machine learning methods appear that enable the use of different levels of variable measurement, the use of data with high dimensionality and, consequently, facilitate the classification of complex multimorbidity patterns (139). In any case, despite this limitation, we consider this finding to be beneficial, since it highlights the need to work on future research that will make it possible to create a clear conceptual framework and an interoperable taxonomy of multimorbidity patterns in the field, as well as new methods that will allow us to better measure and compare chronicity profiles (146).

The present review has some advantages over previous work. On the one hand, this is the first time that a comprehensive review has been carried out that synthesizes the different classification methods used for the extraction of co-joint chronic disease patterns, together with a detailed description of the different typologies of multimorbidity described in the literature and their respective social and behavioral determinants (including individual sociodemographic and socioeconomic factors, lifestyles and risk behaviors, ethnic group, as well as contextual aspects such as social connectivity, place of residence, or use of health services). On the other hand, despite the difficulties in synthesizing the results of studies based on different populations, chronic pathologies and classification methods, the breadth of our work has allowed us to identify gaps in knowledge on topics that may be fundamental for future approaches to multimorbidity, such as chronicity patterns in children and young people or, for example, the influence of relationships and social networks on the emergence and evolution of chronic pluri-pathology.