Multimorbidity of non-communicable diseases in low-income and middle-income countries: a systematic review and meta-analysis

Abstract

Introduction

Multimorbidity is a major public health challenge, with a rising prevalence in low/middle-income countries (LMICs). This review aims to systematically synthesise evidence on the prevalence, patterns and factors associated with multimorbidity of non-communicable diseases (NCDs) among adults residing in LMICs.

Methods

We conducted a systematic review and meta-analysis of articles reporting prevalence, determinants, patterns of multimorbidity of NCDs among adults aged >18 years in LMICs. For the PROSPERO registered review, we searched PubMed, EMBASE and Cochrane libraries for articles published from 2009 till 30 May 2020. Studies were included if they reported original research on multimorbidity of NCDs among adults in LMICs.

Results

The systematic search yielded 3272 articles; 39 articles were included, with a total of 1 220 309 participants. Most studies used self-reported data from health surveys. There was a large variation in the prevalence of multimorbidity; 0.7%–81.3% with a pooled prevalence of 36.4% (95% CI 32.2% to 40.6%). Prevalence of multimorbidity increased with age, and random effect meta-analyses showed that female sex, OR (95% CI): 1.48, 1.33 to 1.64, being well-off, 1.35 (1.02 to 1.80), and urban residence, 1.10 (1.01 to 1.20), respectively were associated with higher odds of NCD multimorbidity. The most common multimorbidity patterns included cardiometabolic and cardiorespiratory conditions.

Conclusion

Multimorbidity of NCDs is an important problem in LMICs with higher prevalence among the aged, women, people who are well-off and urban dwellers. There is the need for longitudinal data to access the true direction of multimorbidity and its determinants, establish causation and identify how trends and patterns change over time.

PROSPERO registration number

CRD42019133453.

Strengths and limitations of this study.

• Inclusion of most studies (14/36 articles) from the WHO Study on global AGEing and adult health (SAGE) ensured standardisation of methods of measurements and data collection.

• The included studies had large sample sizes, which ensured adequate statistical power to detect even a small effect of interest.

• Recall and self-declaration bias due to self-reported outcome may result in under/over estimation of the true prevalence of multimorbidity.

• Assessment of the determinants of multimorbidity did not take the heterogeneity and clusters of conditions into consideration.

• Involving patients with varied characteristics and from a wide range of settings may contribute to substantial heterogeneity.

Introduction

Although the burden of diseases in low/middle-income countries (LMICs) has classically been infectious, changes in demographic patterns as a result of the interplay between urbanisation, life-style and culture, has led to emerging non-communicable diseases (NCDs) in LMICs.^{1 2} The NCD burden is estimated to increase by 27% in the African region in the next 10 years, while Western Pacific and South-East Asia will account for the highest absolute number of deaths from NCDs.³

Coexistence of one or more chronic diseases in an individual is commonly denoted as multimorbidity.^{4 5} With the increasing prevalence of NCDs in LMICs,⁶ many of which share common risk factors, the prevalence of multimorbidity of NCDs will continue to rise. There is, however, a substantial difference in the burden of NCDs between LMICs and high-income countries (HICs) due to the difference in drivers, such as promotion of healthier lifestyles and providing equitable healthcare by instituting appropriate government policies.⁷ While research investigated common pathways on NCD multimorbidity in HICs, it is unclear if this is also valid for LIMCs.⁸ It is therefore important to identify common NCD multimorbidity patterns and pathways that are specific to LMICs.

Studies undertaken so far predominantly used self-reported measures and show multimorbidity to be associated with decreased quality of life, increased healthcare utilisation and costs in primary, secondary and tertiary healthcare settings,4 5 9–12 just as reported in HICs.^{13 14} There is also limited information on the distribution of patterns of multimorbidity, their size, their drivers and their risk factors in LMICs. There are a few studies indicating that multimorbidity in LMICs is more frequent in women and that it starts at an earlier age than in HICs, but these studies are scattered.^{15 16} In order to address and manage the increasing number of people with multimorbidity, it is important to assess the burden of multimorbidity as well as the combinations of NCDs and their patterns in LMICs. A recent scoping review of that summarised the prevalence and determinants of multimorbidity chronic NCDs in LMICs reported prevalence ranging from 3.2% to 90.5%.⁶ This review builds on the previous scoping review by adopting systematic methods and meta-analysis to synthesise the evidence on the prevalence, patterns and factors associated with multimorbidity of NCDs among adults residing in LMICs. We further showed the prevalence and patterns of multimorbidity of NCDs according to country’s income level classification by the World Bank.

Methods

Review framework and patient and public involvement

This systematic review and meta-analysis was reported according to the recommendations outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement¹⁷ (online supplemental file 1).

Patient and public involvement

This is a meta-analysis based on study-level data and no individual-level data were involved in the study or in defining the research question or outcome measures. It was not possible to involve patients or the public in the design, or conduct, or reporting, or dissemination plans of our research.

Search strategy

A structured search was done in the following databases: PubMed, EMBASE and Cochrane library for articles published in English from 2009 until April 2020. Keywords and Medical Subject Headings (MeSH) terms and their combinations used in the searches included “Multiple Chronic Conditions”, “Multimorbidity”, “Comorbidity”, “Non-Communicable Diseases”, “Developing Countries”, “Cardiovascular Diseases”, “Neoplasms”, “Lung Diseases, Obstructive”, “Diabetes Mellitus” and “Mental Disorders”, “Hypertension”. In addition, the reference lists and bibliographies of the included articles were examined to identify any other relevant article. The detailed search strategy is provided as online supplemental file 2.

Inclusion and exclusion criteria

Studies were included if they (i) reported original research on multimorbidity of NCDs, (ii) included adults aged 18 years and above and residing in LMICs, (iii) conducted in any of these study settings; community, residential care homes, primary care, secondary care, tertiary care and specialised care centres/institutions; or at the regional level using data from primary research, demographic and health surveys, or demographic and health surveillance systems. We defined LMICs according to the World Bank’s Country and Lending Group List.¹⁸ We excluded studies conducted in HICs. Studies published in languages other than English and studies on comorbidity (studies that recruited patients based on an index disease or primary disease of interest) were also excluded. However, we included comorbidity in the search strategy to enable us to capture and scrutinise studies that used the terms comorbidity and multimorbidity interchangeably or incorrectly.

Definition of terms/concepts

We defined multimorbidity of NCDs as co-occurrence of two or more chronic non-communicable health conditions in the same individual.⁸ Prevalence of multimorbidity of NCDs was defined as the proportion of people with two or more chronic NCDs in the study population.⁸ Patterns of multimorbidity NCDs were assessed by considering the frequencies and distributions of NCDs among individuals, regions and countries.

Data extraction

Two reviewers (OAA, AMF) extracted data from the included articles. In case of divergent opinions, KK-G and DB were consulted. Information extracted included author(s) name, year of publication and study country, survey/source of data, sample size, method of data collection, number of NCDs, multimorbidity definition, prevalence and factors associated with multimorbidity. The following summary measures were included: prevalence, odds ratio (OR), prevalence risk ratios and relative risk ratio with their 95% CI for the association between risk factors/determinants and NCD multimorbidity.

Quality assessment

The risk of bias in the included studies was assessed using the National Institute of Health Quality Assessment Tool for Observational Cohort and Cross-sectional Studies.¹⁹ This tool was used to appraise the reliability, validity, generalisability and overall quality of the included studies using 14 criteria. This included clearly stated research question and objective, clearly specified study population, adequate participation rate, similar subject selection/recruitment and uniform application of eligibility to all participants, sample size estimation, exposure measurement before outcome, sufficient time frame to detect an association, examination of different levels of exposure, multiply exposure measurement over time, valid outcome assessment, detection bias, loss to follow-up and adjustment of confounding variables. The tool provides general guidance to determine the overall quality of the studies and to grade their level of quality as good, fair or poor.

Data synthesis and analysis

Studies that provided sufficient data were used in the meta-analyses using Cochrane Review Manager (RevMan) software.²⁰ For multi-country studies with sufficient analysis of country level data, findings from individual countries were included separately in the meta-analyses. Findings of the remaining studies were presented in a narrative format. We pooled the OR (95% CI) for the association between sex, education, income, residence (rural/urban) and multimorbidity. A pooled OR of the association between age and multimorbidity was not estimated due to the variation in reference age categories whereas smoking, physical activity and alcohol consumption were not meta-analysed due to the limited number of studies that reported on them. The log OR and SEs were combined in RevMan using the generic inverse-variance.^{21 22} We performed a random effect analysis, and heterogeneity was assessed using the Cochrane’s Q and degree of inconsistency (I²).²³ The pooled prevalence of multimorbidity was estimated using Open Meta (analyst) software.²⁴ The pooled prevalence was further stratified according to different regions in LMICs. The robustness of the pooled estimates was assessed by conducting a leave-one-out sensitivity analysis.²⁵ All analyses were considered statistically significant at the two-sided 5% level (p<0.05).

Results

The electronic database and reference list search yielded 3272 articles, while 3134 articles remained after removal of duplicates. After the title and abstract screening, 68 articles were deemed potentially relevant. Twenty-nine articles were further excluded because they were conducted on communicable diseases (n=18), in non-LMICs (n=2), had poor quality (n=1) or based on other reasons such as presence of an index disease, or assessed multimorbidity in all ages without a separate report for adult above 18 years (n=8). We included 39 studies for the current review (figure 1). Some of the studies reported results from multiple countries, which were included individually in the analyses. For example, Bao et al²⁶ included analyses from seven countries; Cuba, Dominican Republic, Puerto Rico, Peru, Venezuela, Mexico, China. Agrawal and Agrawal,²⁷ Garin et al²⁸ and Christian et al²⁹ each reported findings from six countries; China, India, Mexico, Russia, South Africa and Ghana. Zhou et al³⁰ reported findings from India, China, and Bangladesh while Kunna et al³¹ assessed multimorbidity in China and India. Table 1 shows all the countries or regions where multimorbidity of NCDs were conducted.

Figure 1.

Flow chart for study inclusion and exclusion of studies.

Table 1.

Study characteristics of studies included in the systematic review

Author	Country/region	Inclusion criteria	Study design	Survey/source of data	Sampling characteristics	Field year	Sample size	Age range (years)	Data collection	No of NCDs	Multi-morbidity definition	Prevalence (%)	Quality of included studies
Afshar et al (27 LMICs)16	Africa, Central and South America, Eastern Europe and Central Asia, South Asia, South East Asia	Prevalence and determinants	Cross-sectional	WHS	Probabilistic	2001–2004	25 761	≥18	Self-report	6	≥2	Ranges from 1.7 to 15.2; Africa (3.6–11.2) Central and South America (5.7–13.4) Eastern Europe and Central Asia (7.6–15.0) South Asia (3.9–7.8) Southeast Asia (1.7–15.2) Mean global prevalence (95 CI) 7.8 (7.8 to 7.8)*	Good
Agrawal and Agrawal27	China	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	China (15048); (India 12198); Mexico (2725); Russia (4946); South Africa (4227); Ghana (5571)	≥18	Self-report+medication use+SBD	9	≥2	22.0–50.0: China (22.0); India (24.0); Mexico (27.0); Russia (50.0); South Africa (32.0); Ghana (23.0)	Good
Arokiasamy et al4	China, Ghana, India, Mexico, South Africa, Russia	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	42 236	≥18	Self-report+medication use+SBD	9	≥2	21.9	Good
Aye et al54	Myanmar	Prevalence, patterns and determinants	Cross-sectional	Household survey	Probabilistic	2016	4859	≥60	Self-report	14	≥2	33.2	Good
Bao et al 26	Cuba, Dominican Republic, Puerto Rico, Peru, Venezuela, Mexico, China	Prevalence	Population based Cohort	Household survey	NR	2003–2010	15 027	≥65	Self-report+physical examination	15	≥2	Ranges from 31.0 to 68.0; China (31.0); Peru (49.0) Cuba (58.0), Venezuela (60.0), Mexico (60.0), Dominican Republic (68.0)	Fair
Chen et al57	China	Prevalence and determinants	Cross-sectional	CHARLS	Probabilistic	2011–2012	3737	≥45	Self-report	16	≥2	46.0	GOOD
Christian et al29	China, Ghana, India, Mexico, South Africa, Russia	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	42 487	≥50	Self-report	8	≥2	Ranges from 8.8 to 50.2: Ghana (8.8), India (16), China (20.3), Mexico (20.8), South Africa (20.8%), Russia (50.2)	Good
Ebrahimoghli et al32	Iran	Prevalence	Retrospective cohort study	IHIO	All beneficiaries of IHIO	2013–2016	481 733	≥18	ATC CS	18	≥2	21.5	Fair
Garin et al28	China, Ghana, India, Mexico, South Africa, Russia	Prevalence determinants and patterns	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	China (13157), Ghana (4305), India (6560), Mexico (2301), South Africa (3763), Russia (3836)	≥50	Self-report +medication use +SBD	12	≥2	Ranges from 45.1 to 72.0: China (45.1), Ghana (48.3), India (57.9), Mexico (64.0), South Africa (63.4), Russia (72.0)
Hien et al33	Burkina Faso	Prevalence and determinants	Cross-sectional	Household survey	Probabilistic	2012	389	≥60	Self-report+clinical examination+medical record review	16	≥2	65.0	Good
Jawed et al55	Pakistan	Prevalence and determinants	Cross-sectional	The IMPACT study	Probabilistic	2015–2016	1500	≥30	Self-report+medication use+SBD	16	≥2	48.6	Good
Jerliu et al46	Kosovo	Prevalence and determinants	Cross-sectional	Community survey	Probabilistic	2011	2265	≥65	Self-report	7	≥2	45.0	Good
Jovic et al42	Serbian	Prevalence, patterns	Cross-sectional	NHS-Serbia	Probabilistic	2013	13 103	≥20	Self-report	12	≥2	26.9	Good
Jankovic et al43	Serbian	Prevalence and determinants	Cross-sectional	NHS-Serbia	Probabilistic	2013	13 765	≥20	Self-report	13	≥2	30.2	Good
Khan et al45	Bangladesh	Prevalence, patterns and determinants	Cross-sectional	Household survey	Probabilistic	2014–2016	12 338	≥35	Self-report+medication use+SBD	6	≥2	8.4	Good
Khanam et al48	Bangladesh	Prevalence and determinants	Cross-sectional	HDSS	Probabilistic	2003–2004	452	≥60	Self-report+physical examination+blood test	9	≥2	53.7	Good
Kumar et al47	India	Prevalence	Cross-sectional	Household survey	NR	2012–2013	58 590	≥20	Self-report	5	≥2	0.7	Fair
Kunna et al31	China, Ghana	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	China (11 814); Ghana (4050)	≥50	Self-report+SBD	7	≥2	China (29.7); Ghana (30.2)	Good
Koyanagi et al87	China, Ghana, India, Mexico, South Africa, Russia	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2007–2011	32 715	≥50	Self-report+SBD	10	≥2	49.8	Good
Lee et al9	China, Ghana, India, Mexico, South Africa, Russia	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	39 213	≥18	Self-report	9	≥2	Varies from 3.9 in Ghana-33.6 in Russia	Good
Mini and Thankappan50	India	Prevalence, patterns and determinants	Cross-sectional	UNFPA	Probabilistic	2011	9852	≥60	Self-report	12	≥2	30.7	Good
Nugraha et al41	Indonesia	Prevalence	Cross-sectional	Community survey	Probabilistic	2018	427	≥60	Self-report	15	≥2	60.7	Good
Nunes et al40	Brazil	Prevalence and patterns	Cross-sectional	Household survey	Probabilistic	2008	1593	≥60	Self-report	17	≥2	81.3	Good
Nunes et al34	Brazil	Prevalence, patterns and determinants	Cross-sectional	PNS	Probabilistic	2013	60 202	≥18	Self-report	22	≥2 or≥3	22 for ≥2 and 10.2 for ≥3	Good
Pati et al58	India	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007	10 973	≥18	Self-report	9	≥2	8.9	Good
Pati et al39	India	Prevalence and patterns	Cross-sectional	Primary healthcare	Probabilistic		1649	≥18	Self-report	21	≥2	28.3	Good
Pengpid and Peltzer36	Mekong	Prevalence, determinants	Cross-sectional	Primary healthcare	Probabilistic	NR	6236	≥18	Self-report	21	≥2	72.6 (28.6 had 2, 22.4 had 3 and 21.6 had ≥24 chronic conditions)	Good
Phaswana-Mafuya et al52	South Africa	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2008	3840	≥50	Self-report	8	≥2	22.5	Good
Price et al53	Malawi	Prevalence and determinants	Cross-sectional	Household survey	No sampling: all adults	2013–2016	28 891	≥18	Self-report+medication use+patient health record+clinical test	3	≥2	4.0	Good
Rehr et al51	Northern Jordan	Prevalence, patterns and determinants	Cross-sectional	UNHCR	Probabilistic	2016	8041	≥18	Self-report	6	≥2	44.7	Good
Rzewuska et al35	Brazil	Prevalence, patterns and determinants	Cross-sectional	PNS	Probabilistic	2013	60 202	≥18	Self-report+SBD	14	≥2	24.2	Good
Sum et al44	China, Ghana, India, Mexico, South Africa, Russia	Prevalence and patterns	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	41 557	≥18	Self-report+SBD	9	≥2	18.9	Good
Vadrevu56	India	Prevalence and determinants	Cross-sectional	Household survey	Probabilistic	2009	815	≥40	Self-report+SBD	6	≥2	44.1	Good
Vancampfort et al88	China, Ghana, India, Mexico, South Africa, Russia	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	34 129	≥50	Self-report+SBD	11	≥2	45.5	Good
Vancampfort et al89	China, Ghana, India, Mexico, South Africa, Russia	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	34 129	≥50	Self-report+SBD	11	≥2	45.5	Good
Vancampfort et al49	China, Ghana, India, Mexico, South Africa, Russia	Prevalence and determinants	Cross-sectional	WHO SAGE	Probabilistic	2007–2010	14 585	≥65	Self-report+SBD	11	≥2	60.2	Good
Waterhouse et al90	South Africa	Prevalence	Cross-sectional	WHO SAGE	Probabilistic	2007–2008	3055	≥50	Self-report	8	≥2	13.2	Good
Woldesemayat et al60	Ethiopia	Prevalence, patterns and determinants	Cross-sectional	Healthcare	NR	2016	411	≥18	Self-report+medical card	18	≥2	17.8	Good
Zhou et al30	Bangladesh, India, China	Prevalence	Cross-sectional	WHS	Probabilistic	2002–2004	Bangladesh (5507); India (9199); China (3990)	≥18	Self-report	9	≥2	Bangladesh (28.8); India (34.4); China (14.3)	Good

*Include prevalence from 27 LMICs and 1 high-income country.

ATC CS, Anatomical Therapeutic Chemical Classification System; BDHS, Bangladesh Demographic and Health Survey; CHARLS, China Health and Population Fund; HDSS, Health and Demographic Surveillance System; IHIO, Iranian Health Insurance Organization; LMIC, low/middle-income country; NCDs, non-communicable diseases; PNS, Pesquisa Nacional de Saude (Brazillian National Health Survey); SA-NIDS, South Africa National Income Dynamics Study; SBD, symptom based diagnosis; UNHCR, United Nations High Commission for Refugees; WHO-SAGE, WHO Study on Global AGEing and adults health; WHS, World Health Survey.

All included articles were cross-sectional except two studies that were cohort studies.^{26 32} A total of 1 220 309 individuals were included and the sample size ranged from 389³³ to 60 202.^{34 35} NCDs were assessed through self-report in all included studies, or in combination with a health insurance database,³² or medication use and clinical test. One study assessed based on the Anatomical Therapeutic Chemical Classification System. The number of self-reported NCDs ranged from 3 to 22. All studies defined multimorbidity as coexistence of two or more chronic NCDs, except one study which defined multimorbidity as a count of 21 chronic health conditions³⁶ (table 1).

Most of the studies were of good quality. Three included articles were judged to be of fair quality.^{26 32 37} One study was excluded because of having a small sample size, and a lack of data or non-robust methods.³⁸ Twenty-one of the included studies did not give information about missing data handling4 26 29–31 33 34 36 39–52 (online supplemental file 3).

The overall prevalence of multimorbidity of NCD varied from 0.7% (in a population aged ≥20 years in a rural community in Western India) to 81.3% (in an elderly population aged ≥60 years in Southern Brazil).^{40 47} A study that assessed prevalence of multimorbidity among adults ≥18 years in 27 LMICs using the World Health Surveys reported a mean prevalence ranging from 1.7% (95% CI 1.4 to 2.0) in Myanmar to 15.2% (95% CI 14.3 to 16.0) in Nepal.¹⁶ In studies that combined self-reported diseases with symptom based diagnosis, medication use/medical card review, prevalence varied between 4.0% and 72% in people ≥18 years.^{36 53} The overall prevalence of multimorbidity was 36.4% (95% CI 32.2% to 40.6%) as shown in figure 2. In a subgroup analysis, the pooled prevalence according to the countries’ income levels was 39.3% (95% CI 34.5% to 44.1%) for upper middle-income countries (MICs) (online supplemental figure 1a) and 29.2% (95% CI 23.0% to 35.4%) for lower MICs (online supplemental figure 1b). We did not pool the prevalence for low-income countries (LICs) because there were only three studies, with prevalence ranging from as low as 4.0% in Malawi to 65.0% in Burkina Faso. Subgroup analysis according to the World Bank regions of LMICs was 26.2% (95% CI 18.9% to 33.5%) for sub-Saharan Africa (SSA); 29.5% (95% CI 20.9% to 38.1%) for Asia; 31.8% (95% CI 25.7% to 37.8%) for East Asia; 33.1% (95% CI 10.4% to 55.8%) for Middle-East and North Africa (MENA); for Europe and Central Asia (excluding high income) 44% (95% CI 32.7% to 55.3%) and 50.4% (95% CI 35.6% to 65.2%) for Latin America and the Caribbean (LAC). According to the leave-one-out sensitivity analysis, no single study had a substantial influence on the overall prevalence of NCD multimorbidity (online supplemental figure 2).

Figure 2.

Forest plot of pooled prevalence of multimorbidity in low/middle-income countries.

Age, sex, education, wealth/income, urban/rural setting and marital status were the most studied factors associated with multimorbidity of NCDs (online supplemental table 1). ORs for the association between major predictors and multimorbidity are shown in online supplemental table 2. Age was positively associated with multimorbidity of NCDs in 22 studies, whereas 3 studies^{28 48 49} found no association.

Figure 3 shows a forest plot of pooled OR for the association between major predictors and multimorbidity; details of the meta-analysis for the individual predictors are shown in online supplemental figure 3a–d). Women had significantly higher odds of multimorbidity compared with men in 11 studies,4 9 28 34–36 48 50 54–56 whereas 8 studies showed a non-significant association28 31 33 46 49 51 57 58 (online supplemental table 2). Fourteen studies (one study included six different country level results) were meta-analysed and the pooled OR for female sex and NCD multimorbidity was 1.48 (95% CI 1.33 to 1.64) (figure 3, online supplemental figure 3a). The association between education and multimorbidity was assessed in 31 studies. In most studies, the risk of multimorbidity was higher among those with a lower educational status,4 16 28 34–36 43 51 while four studies reported a lower risk of lower education statu.45 50 53 57 A meta-analysis of 13 studies (one study included six different country level results; one study included results for males and females) showed an OR of 1.22 (95% CI 1.00 to 1.49) for those with no formal education or lower educational attainment (figure 3, online supplemental figure 3b).

Figure 3.

Forest plot of pooled ORs of factors associated with multimorbidity in low/middle-income countries.

The association between socioeconomic status (income/wealth) and multimorbidity was determined in 14 studies; 7 studies found an association with higher odds/risk/prevalence of multimorbidity for people in the most well-off class,9 28 31 45 48 50 53 while in 3 studies the odds/prevalence of multimorbidity was higher for people considered to be poor.^{4 28 31} The pooled OR from 10 studies (one study included six different country level results; one study included results for males and females) showed increased odds of NCD multimorbidity among people who are well-off, OR 1.35 (95% CI 1.02 to 1.80) (figure 3, online supplemental figure 3c). There were significantly higher odds/risk for multimorbidity of NCDs for urban areas.9 34 49 53 54 59 A meta-analysis of 10 studies (one study included six different country level results; two included results for males and females) showed a pooled OR of 1.10 (95% CI 1.01 to 1.20) for urban residence (figure 3, online supplemental figure 3d). There was a high degree of heterogeneity as depicted by high I² >90% in the various meta-analyses conducted.

Three of the seven studies that assessed the association between multimorbidity of NCDs and physical activity/exercise showed significantly higher odds for those that do little or no physical activity,^{4 31 49} while the other five showed no significant relationship.31 36 45 53 60 Eight studies examined the relationship between obesity and multimorbidity; five articles found higher a positive association between multimorbidity of NCDs and obesity.4 27 31 45 56 In the WHO SAGE study among five LMICs, obese individuals were 2.3 times (95% CI 2.0 to 2.52) more likely to have multimorbidity compared with the non-obese when multimorbidity was compared with no disease.⁴ Eight studies assessed the association between smoking and multimorbidity, with a study conducted among the elderly from seven Indian urban and rural states reporting a positive association⁵⁰ when compared with no NCD (OR: 1.22, 95% CI 1.08 to 1.37). Alcohol consumption was associated with higher odds of NCD multimorbidity.^{50 53}

The patterns of reported NCD multimorbidity are shown in table 2. Seventeen studies assessed patterns of multimorbidity of NCDs using factor analysis,28 34 35 42 54 cluster analysis⁵⁰ or descriptive methods.39 40 44 45 51 60 Sixteen out of the 17 studies that reported on patterns of multimorbidity were conducted in MICs, while only one study was conducted in LIC. Cardiometabolic and cardiorespiratory conditions were the most identified patterns seen in MICs, while cardiovascular, musculoskeletal system diseases and endocrine system diseases were observed in the only one study in LMICs (table 2). The highest prevalence of cardiometabolic pattern was 70.3% and 60.7% among males and females aged 20–40 years, respectively in MICs. Cardiometabolic, mental and respiratory conditions were present in both men and women in two MICs studies that stratified by sex.^{35 42} Mental disorder was also reported to cluster with other conditions such as cardiometabolic, respiratory and musculoskeletal conditions in studies conducted in Brazil, Serbia and a multi-country study in South Africa, Ghana, Mexico, Russia, Bangladesh, India and China.28 34 35 39 42 44

Table 2.

Patterns of multimorbidity reported in included studies

Pattern	Study	Economy status	Diseases	Prevalence % (95% CI)
Cardiometabolic	Garin et al28 (China)	MIC	Diabetes, obesity, hypertension, angina, stroke, cataract	NR
	Garin et al28 (Ghana, India, Mexico)	MIC	Diabetes, obesity, hypertension	NR
	Garin et al28 (Russia)	MIC	Diabetes, obesity, hypertension, angina, stroke, cataract, arthritis, edentulism, depression	NR
	Garin et al28 (South Africa)	MIC	Diabetes, obesity, hypertension, angina, stroke, arthritis, edentulism	NR
	Jovic et al42	MIC	Male (age 20–44 years): Cardiometabolic Age 45–64 years: Cardiometabolic Age 65+ years: Cardiometabolic	70.3 39.2 29.5
		MIC	Female (Age 20–44 years): Cardiometabolic Age 45–64 years: Cardiometabolic Age 65+ years: Cardiometabolic	60.7 53.2 33.2
	Khan et al45	MIC	Hypertension, diabetes, CVD; Hypertension, diabetes, stroke; Hypertension, diabetes, cancer; Hypertension, CVD, stroke Diabetes, CVD, stroke Hypertension, diabetes, CVD, stroke	0.6 0.4 0.0 0.3 0.3 0.6
	Mini and Thankappan50	MIC	High blood pressure, diabetes	4.7
	Nunes et al34	MIC	High blood pressure, heart attack, angina, heart failure, stroke, hypercholesterolaemia, diabetes, arthritis/rheumatism	NR
	Rehr et al51	MIC	Diabetes and hypertension; Diabetes, hypertension and CVD; Hypertension and CVD; Diabetes, hypertension and thyroid disease	17.6 (15.9 to 19.5) 8.1 (6.9 to 9.7) 7.1 (5.9 to 8.4) 1.3 (0.9 to 2.0)
	Rzewuska et al35	MIC	Male and female: diabetes, stroke, cardiovascular disorders; high blood cholesterol, hypertension	NR
Cardiovascular	Aye et al54	MIC	Coronary heart disease, Heart failure	NR
	Jovic et al42	MIC	Male (age 45–64 years): cardiovascular Age >65 years: cardiovascular Female (45–64 years): cardiovascular Age >65 years: cardiovascular	22.8 28.7 29.6 18.9
	Rehr et al51	MIC	Hypertension and CVD	7.1 (5.9 to 8.4)
	Woldesemayat et al60	LIC	Cardiovascular and endocrine system diseases	2.4
Cardiorespiratory	Aye et al54	MIC	Asthma, COPD, hypertension, diabetes, stroke	NR
	Garin et al28 (China)	MIC	Angina, asthma, COPD, depression, arthritis, cataract	NR
	Garin et al28 (Ghana)	MIC	Angina, asthma, COPD	NR
	Garin et al28 (India)	MIC	Angina, asthma, COPD, depression	NR
	Garin et al28 (Mexico)	MIC	Angina, asthma, COPD, stroke, depression, arthritis, cataract	NR
	Garin et al28 (South Africa)	MIC	Angina, asthma, COPD, stroke, depression, arthritis	NR
	Rehr et al51	MIC	Hypertension and chronic respiratory condition	1.3 (0.8 to 1.9)
	Khan et al45	MIC	Hypertension, diabetes, COPD	0.1
Mental	Aye et al54	MIC	Depression, mental illness	NR
Respiratory	Garin et al28 (Russia)	MIC	Asthma, COPD, cataract	NR
	Jovic et al42	MIC	Male (age 45–64): respiratory Age 65+ years: respiratory Female (age 45–64 years): respiratory 16.8 Age 65+ years: respiratory	13.7 16.0 14.5
	Rzewuska et al35	MIC	Male and female: Asthma, chronic obstructive pulmonary disease	NR
Musculoskeletal	Aye et al54	MIC	Arthritis, osteoporosis	NR
Ocular+musculoskeletal +cardiorespiratory	Aye et al54	MIC	Asthma, COPD, cataract, arthritis, osteoporosis, asthma, COPD, hypertension, diabetes, stroke	NR
	Mini and Thankappan50	MIC	Arthritis, hypertension; Arthritis, cataract	7.5 5.3
	Nunes et al40	MIC	HBP, heart problem, eyesight problem, spinal column disease, rheumatism	10.6 to 5.5
	Pati et al39	MIC	Hypertension+APD+diabetes, Hypertension+APD+CBA; Hypertension+arthritis+diabetes; Hypertension+arthritis+CBA; APD+visual impairment+CBA; APD+visual impairment+arthritis Athritis+CBA+CLD; Athritis+CBA+visual impairment APD+hypertension/visual impairment/CBA/arthritis/diabetes/CLD/deafness Hypertension+visual impairment/CBA/visual impairment/arthritis/deafness Arthritis+visual impairment/CBA/diabetes	NR
	Sum et al44	MIC	Age 18–49 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD, Age 50–64 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD Age >64 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD, Hypertension+cataract	4.99 4.13 1.90 19.08 17.08 9.79 33.77 29.73 16.44 15.27
Mental+musculoskeletal	Garin et al28 (China)	MIC	Arthritis, depression, stroke, cataract	NR
	Garin et al28 (Ghana)	MIC	Arthritis, depression	NR
	Garin et al28 (India)	MIC	Arthritis, depression, cataract, angina	NR
	Jovic et al42	MIC	Male age ≥65 years: mechanical/mental/metabolic Female age ≥65 years: mechanical/mental/metabolic	25.8 32.3
	Nunes et al34	MIC	Arthritis/rheumatism, spinal column problem, spinal column problem, asthma/wheezy bronchitis, COPD, work-related muscle-skeletal disorders, depression, bipolar disorder, kidney problem	NR
	Rzewuska et al35	MIC	Male and female: Arthritis or rheumatism, high blood cholesterol, MSK-D related to work, any chronic back problem, chronic renal insufficiency, schizophrenia, bipolar, obsessive-compulsive disorder, depression	NR
	Sum et al44	MIC	Age: 18–49 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD, Hypertension+depression	4.99 4.13 1.90 1.67
Cardio metabolic+musculoskeletal	Mini and Thankappan50	MIC	Arthritis, hypertension	7.5
	Nunes et al40	MIC	HBP, rheumatism, spinal column disease; HBP, heart problem, spinal column disease; HBP, heart problem, cognitive impairment; HBP, spinal column, falls	10.6 to 5.7
	Pati et al39	MIC	Hypertension, arthritis, diabetes/CBA
	Woldesemayat et al60	LIC	Cardiovascular and musculoskeletal system diseases	1.2
Cardio metabolic+musculoskeletal +mental	Nunes et al40	MIC	HBP, heart problem, cognitive impairment, depression	10.6 to 5.2
	Jovic et al42	MIC	Male: Age 45–64 years: Aggregate pattern, such as degenerative joint disease/arthrosis, depression, cardiovascular, kidney disease, stroke and malignancy Female: Aged 20–44 years: non-communicable pattern such as degenerative joint disease/arthrosis, depression, cardiovascular and malignancy	24.3 13.3
Cardio+metabolic+respiratory +musculoskeletal+mental	Jovic et al42	MIC	Male: Age 20–44 years: non-communicable pattern such as degenerative joint disease/arthrosis, depression, cardiovascular, respiratory, kidney disease and malignancy	29.7
	Pati et al39	MIC	Arthritis+CBA/visual impairment/chronic lung disease	NR
	Sum et al44	MIC	Age 18–49 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD, Hypertension+depression Age 50–64 years: Hypertension+arthritis, Hypertension+angina, Hypertension+CLD	4.99 4.13 1.90 1.67 19.08 17.08 9.79

APD, acid peptic disease; CBA, chronic back pain; CLD, chronic lung disease; COPD, chronic obstructive pulmonary disease; CRD, cardiorespiratory disease; CVD, cardiovascular disease; LIC, low-income country; MIC, middle-income country; MSK-D, musculoskeletal disorder.

Discussion

This systematic review with meta-analyses of 39 studies shows that the overall prevalence of NCD multimorbidity in LMICs was 36% with substantial variation between studies. Prevalence differed by region and was observed to be lowest in SSA and highest in LAC region. According to income levels of countries, the prevalence of NCD multimorbidity was higher among upper MICs and as compared with lower-middle income countries. Older age, female sex, higher income and urban residence increased the odds of having NCD multimorbidity. Cardiometabolic and cardiorespiratory patterns of multimorbidity of NCDs were most common; in addition, multimorbidity of mental disorders with respiratory, musculoskeletal and cardiometabolic conditions was observed.

An important finding from our review is the large variation in the estimates of prevalence of multimorbidity of NCDs in LMICs. This may be explained by differences in definition/measurement of multimorbidity, study populations, demographics, study settings, self-reported diseases and the number of NCDs included. Similar variation was seen in reviews that focused on South Asia⁶¹ and HICs.^{62 63} A recent scoping review of multimorbidity of chronic NCDs in LMICs also found a wide variation in the prevalence of multimorbidity in LMICs (3.2%–90.5%), depending on population age and the number of conditions considered.⁶ Since prevalence estimates depend on the number and the type of chronic conditions included in the measurement of multimorbidity, there might be underreporting due to lack of data or undiagnosed conditions. To date, there is no valid standard measurement of multimorbidity indicating a need for a uniform definition and a reporting system for multimorbidity, as suggested by the Academy of Medical Science.⁸

The positive association of multimorbidity with age and female sex is consistent with a study comparing 27 LMICs and 1 HIC using the World Health Survey,¹⁶ other reviews on multimorbidity in South Asia and LIMCs^{6 39} as well as reviews from HICs.^{62 63} The meta-analyses showed higher odds of multimorbidity among women compared with men. While the association between these factors and multimorbidity is inconsistently reported, the sex-related differences in multimorbidity could be related to context related proxy for behavioural characteristics such as care seeking, that might influence the detection of multimorbidity.⁸ Women are more likely to have frequent healthcare consultations than men^{64 65} and might be able to self-report their health status than men. In addition, sex differences in socioeconomic status could also account for the discrepancy observed. Socioeconomic status affects general health functioning, including mental and physical health. Research show that women, in general, have lower socioeconomic status than men, which is in part related to gender inequality and could negatively affect health outcomes.⁶⁶

In LMICs, people who are well-off in terms of income seem to be most affected by multimorbidity, in contrast with evidence from HIC⁸ that shows an inverse association. Few studies from HIC have, however, reported higher prevalence among people who are well-off.^{9 16 59} Contextually, people who are well-off in LMICs are generally less physically active and consume more fats, salt and processed food which could partly explain the higher prevalence of NCD multimorbidity.⁶⁷ Further, they might be better educated, informed and have greater access to medical care and are more likely to receive disease diagnosis. The significantly higher odds for multimorbidity of NCDs seen in the urban areas may be due to under-reporting in rural areas as a result of poorer access to healthcare and healthcare insurance.⁶⁸ In most LMICs, healthcare services are paid out of pocket for every inpatient and outpatient visit.⁹ People living in rural areas are less likely to have long-term healthcare insurance and also less likely to be provided with adequate healthcare.⁶⁹ Furthermore, regional differences in lifestyle could also explain higher odds of multimorbidity of NCDs in people living in urban areas as residence in urban areas is associated with unfavourable diets and lower physical activity levels.^{70 71}

This review identified various patterns of NCD multimorbidity across different regions in LMICs. Cardiometabolic and cardiorespiratory patterns of multimorbidity were most common and share major pathophysiological pathways and common risk factors such as smoking,^{72 73} partly explaining their clustering together. The frequent co-occurrence of cardiometabolic conditions and mental disorders among studies in LMICs as shown in this review is consistent with findings from HICs^{62 74 75} and highlights the importance of prevention and management policies addressing environmental and living conditions.⁷⁶

Current evidence suggests a poorer health-related quality of life, worse clinical outcomes and an increased risk of premature mortality among patients with concurrent physical and mental health conditions than those who have physical conditions alone.^77–79 Individuals with concurrent physical and mental health conditions are also found to have challenges with medication adherence, compromised self-management,⁸⁰ high risk of adverse drug events,⁸¹ higher rates of healthcare utilisation. They are however at a risk of receiving suboptimal care for coexisting health conditions, leading to poorer health outcomes and increased mortality.⁸²

Strength and limitations

A strength of this review is that most of the included studies from the database search were from the WHO Study on global AGEing and adult health (SAGE), which ensured standardisation of methods of measurements and data collection. This review provides worldwide prevalence rates and predictors for multimorbidity. The standardised methods and large sample sizes of the underlying studies ensure a high qualitative standard of the report.

A main limitation of this review is that all studies included self-reported measures for data collection of multimorbidity, and very few collected physical or biochemical data. Self-reported disease is fairly accurate, and may be subject to recall and self-declaration bias, under or over reporting of outcome of interest.^{83 84} This may result in under/over estimation of the true prevalence of multimorbidity. The restriction of inclusion criteria to only studies conducted in English might have also led to studies from other LMICs, especially South America where Spanish dominates, leading to potential bias in the estimates. Generally, studies that assessed determinants of multimorbidity did not take the heterogeneity and clusters of conditions into consideration. The observational studies summarised involved patients with varied characteristics and from a wide range of settings contributing to substantial heterogeneity, which could affect the reliability of the findings. The use of cross-sectional design in almost all studies limits the ability to assess the outcome over a longer period and therefore makes it impossible to draw a causal relationship between the various determinants and multimorbidity.⁸⁵ In the absence of intervention studies, the meta-analysis of the observational studies provides insight into the direction and strength of the association between the various risk factors and NCD multimorbidity. We did not include MeSH terms related to metabolic diseases such as obesity/overweight, metabolic syndrome and osteoarthritis mainly because they are risk factors of major NCDs. We believe, however, that our search strategy was able to cover these risk factors since most of the major NCDs are assessed together with these in most multimorbidity studies.

Implications of findings

The rising burden of multimorbidity in LMICs indicates the urgent need to strengthen the healthcare system to accommodate for the diagnosis and management of multiple chronic conditions. Available evidence shows that patients with multimorbidity have significantly higher mean outpatient and inpatient visits, resulting in higher out-of-pocket expenditure.^{9 43 58} Increased healthcare utilisation among patients with multimorbidity poses challenges to the patients, health providers and the healthcare system.

Evidence from HIC shows diverse challenges when dealing with patients with multimorbidity, including the complexity of multiple guidelines which focus on the management of single conditions and challenges in delivering patient-centred care.⁸⁶ This emphasises the need to develop context-specific guidelines on how to diagnose and deal with multiple chronic conditions and to ensure better health service provision, health management and resource deployment to manage the increasing number of people with multimorbidity. Exploring the economic burden of multimorbidity across different settings and populations in LMICs will be crucial in informing policy decisions about service provision and resource allocation.

Despite the clear rise of multimorbidity in LMICs, there is a challenge in explaining the factors behind this rising burden given inconsistencies in findings. This is partly due to the lack of longitudinal studies providing strong evidence on the determinants and the differences in patterns of multimorbidity among different age groups as well as factors that influence variation in clusters of multimorbidity. The acceptance of a standard definition of multimorbidity will provide more clarity on the burden and epidemiology of multimorbidity.

Conclusion

In conclusion, this review shows a high burden of multimorbidity in LMICs, especially among women, the people who are well-off, and people residing in urban areas, with cardiometabolic and cardiorespiratory profiles being the most prevalent patterns of multimorbidity. There are however major gaps in epidemiological research on this topic, including the need for longitudinal data to access the true direction of the multimorbidity and its determinants, to establish causation and to identify how trends and patterns change over time.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information. All data relevant to the study are included in the article or uploaded as supplementary information. Extra data are available by emailing the corresponding author (d.boateng-2@umcutrecht.nl).

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study does not involve human participants.