Patients with more comorbidities have better detection of chronic conditions, but poorer management and control: findings from six middle-income countries

We used cross-sectional data from the World Health Organisation Study of Global Ageing and Adult Health (WHO SAGE) Wave 1 (2007–10) which collected nationally representative samples of people aged 50+ years in China, Ghana, India, Mexico, Russia and South Africa, with a smaller sample of adults aged 18–49 years in each country for comparison [13]. SAGE contains information on sociodemographic characteristics, anthropometrics and biomarkers, NCDs, healthcare utilisation, quality of life and well-being, social cohesion, and impact on caregiving [13]. Face-to-face interviews were conducted in all countries, using a combination of computer-assisted personal interview, and paper and pencil [13].

The original total sample size of adults aged ≥18 years was 44,089 (China: 15,009, India: 12,198, Ghana: 5563, Russia: 4350, Mexico: 2744, South Africa: 4225). We excluded those who had missing values on outcome variables and covariates (5.7% of entire sample). Final sample size was 41,557 (China: 14,906, India: 11,159, Ghana: 5067, Russia: 4330, Mexico: 2618, South Africa: 3477).

Figure 4 in Appendix summarises predicting variables and outcomes.

We summarised subject characteristics by country with pooled data. For each NCD, we examined the prevalence of subjects with 0, 1, 2, 3, and 4+ comorbid conditions.

For the 6 NCDs that had symptom-based assessment, we presented the prevalence of undiagnosed subjects as their number of comorbidities increased from 0 to 1, 2, 3, and 4+. For all 9 NCDs, we presented prevalence of untreated subjects as the number of comorbidities increased from 0 to 1, 2, 3, and 4+. For the 6 NCDs that had symptom-based assessment, we presented the prevalence of uncontrolled subjects as the number of comorbidities increased from 0 to 1, 2, 3, and 4+. In addition, we presented the prevalence of being undiagnosed, untreated, and uncontrolled when each NCD co-occurred with only concordant NCDs, only discordant NCDs, and only depression.

We conducted a series of multivariable logistic regression analyses. For each NCD, we obtained adjusted odds ratios (AORs) of being undiagnosed, untreated and uncontrolled: Firstly, in association with greater number of comorbidities, and secondly, in association with the co-occurrence of only concordant NCDs, versus only discordant NCDs, and versus only depression. We adjusted for country fixed effects and covariates in all the regression models. The data analyses were weighted to account for the complex, multi-stage design of the SAGE survey. We performed statistical analyses using Stata 15∙1(StataCorp).

We presented subjects’ characteristics by country in Table 1. Median age was 58 (IQR = 51–68) years. The prevalence of subjects with 2 or more NCDs was overall 19%, and was 38.7% in Russia, 17.7% in South Africa, 17.2% in Ghana, 16.8% in Mexico, 14.8% in India, and 13.9% in China. Using pooled data, 43.0% were male, 22.5% were aged above 70 years, 61.3% had primary school education or less, 18.8% were from the lowest income quintile, 52.5% resided in rural areas, and 48.1% did not have insurance. Table 5 in Appendix displays the prevalence of subjects with 0, 1, 2, 3, and 4+ comorbidities for each NCD.

Table 2 shows the prevalence of subjects undiagnosed for each NCD, which ranged from 42.8% for undiagnosed arthritis to 62.2% for undiagnosed angina.

More comorbidities were associated with decreased odds of undiagnosed hypertension, angina and arthritis, but not for asthma, CLD, and depression (Fig. 1a). Comorbidity with concordant conditions was associated with decreased odds of undiagnosed hypertension and angina, but not for arthritis, asthma, CLD, and depression (Fig. 1b).

Table 2 shows the prevalence of subjects untreated for each NCD, with highest prevalence for CLD (91.2% ‘unT-last 2 weeks’, 84.8% ‘unT-last 12 months’), and lowest prevalence for diabetes (33.8% ‘unT-last 2 weeks’) and hypertension (17.4% ‘unT-last 12 months’).

More comorbidities were associated with decreased odds of untreated hypertension and angina, but not arthritis, asthma, CLD, diabetes, depression and stroke for ‘unT-last 12 months’ (Fig. 2a). Similar results were seen for untreated hypertension and angina for ‘unT-last 2 weeks’ (Figure 6a in Appendix). In contrast, comorbidity was associated with increased odds of untreated diabetes.

Comorbidity with concordant conditions was associated with decreased odds of untreated hypertension and angina, but not arthritis, asthma, CLD, diabetes, depression and stroke (Fig. 2b, Figure 6b in Appendix).

Table 2 shows the prevalence of subjects uncontrolled for each NCD, with highest prevalence for angina (83.6% ‘unC-T-last 2 weeks’; 77.8% ‘unC-T-last 12 months’), and lowest prevalence for depression (50.3% ‘unC-T-last 2 weeks’; 62.2% ‘unC-T-last 12 months’).

More comorbidities were associated with increased odds of uncontrolled hypertension, angina, arthritis, and asthma, but not CLD and depression for ‘unC-T-last 12 months’ (Fig. 3a). Similar results were seen for ‘unC-T-last 2 weeks’ (Figure 7a in Appendix).

Comorbidity with concordant conditions was not associated with decreased nor increased odds of being uncontrolled for all NCDs (Fig. 3b, Figure 7b in Appendix).

Tables 6-15 in Appendix show prevalence of being undiagnosed, untreated and uncontrolled for each NCD, in association with greater comorbidity, and with having only concordant NCDs, only discordant NCDs, and only depression.

More comorbidities were associated with better diagnosis of hypertension, angina and arthritis, and better odds of having treatment for hypertension and angina. However, more comorbidities were associated with worse control of hypertension, angina, arthritis, and asthma.

Comorbidity with concordant conditions was associated with decreased odds of undiagnosed and untreated hypertension and angina. Comorbidity with concordant conditions was not associated with decreased nor increased odds of being uncontrolled for NCDs.

Our finding on the positive effect of comorbidities on diagnosis is consistent with the small number of existing articles. Subjects with more comorbidities likely resulted in more frequent visits to, and interactions with multiple health providers. Healthcare professionals were more likely to detect previously undiagnosed co-occurring conditions [25, 26], such as hypertension, angina, and arthritis in this particular study. More comorbidities and increased frequency of healthcare visits were likely associated with a greater tendency for routinely taken blood pressure measurements to indicate hypertension, and for patients to self-report “chest pains” and arthritis symptoms [27‐29]. For other conditions such as CLD and depression that require non-routine checks and are less self-reported, they may still be undiagnosed despite more patients having more comorbidities and more frequent healthcare visits [30‐33].

Our finding on more comorbidities being associated with decreased odds of untreated hypertension and angina was not consistent with the small amount of existing literature, which mostly showed that comorbidity was associated with increased odds of untreated conditions. Existing studies showed that patients had difficulty coping with complex treatment regiments from polypharmacy, and had poor adherence from adverse drug events and high out-of-pocket expenditures on medicines [6, 34‐36]. The difference in our findings and current literature is likely explained by our study examining only whether subjects were taking treatment or not, and did not assess if treatment was adequate, in terms of adherence to medicines (i.e. dose, frequency, duration, administration (e.g. techniques for subcutaneous insulin injections or inhaler use), route (oral, parenteral)), and to lifestyle modifications [37]. Our study showed that with more comorbidities, subjects had higher odds of taking treatment, but we expect that in reality, with more comorbidities, odds of treatment adherence and having adequate treatment would decline.

Our finding on more comorbidities being associated with poorer control is also consistent with the little amount of existing literature. The difficulty controlling NCDs tend to be exacerbated with more co-occurring conditions, and patients with comorbidities were less likely to have certain NCDs addressed [7, 9, 10]. Our study also showed that concordant comorbidities were associated with decreased odds of both undiagnosed and untreated hypertension and angina. Previously undiagnosed concordant NCDs may have a higher tendency to be simultaneously diagnosed during consultations with physicians for the primary NCD of interest, and concordant NCDs are more likely to have better co-ordination of care [6, 9, 38].

Our study’s finding that discordant conditions were not associated with poorer control of NCDs is generally in contrast to the literature. The small number of existing studies have revealed that discordant comorbidities, which have different pathophysiology and management plans, compromise the quality of care of the patient [7, 9, 39]. For example, a study in the United States on hypertension found that patients with discordant conditions were less likely to have controlled hypertension [9]. The existing literature is from high-income countries with likely higher adherence to treatment [40‐42], and in contrast, adherence to treatment may be lower in MICs due to financial constraints and lower health literacy [36, 43]. Hence this non-adherence in MICs may explain the lack of difference between concordant versus discordant comorbidities in the association with control of NCDs.

This is the first study on high-population MICs that investigates the associations between comorbidity and the odds of being undiagnosed, untreated, and uncontrolled for NCDs. Self-reported diagnosis of NCDs may be under-reported and symptom-based assessment of NCDs may not correlate with true medical status [13, 44, 45]. Additionally, stigma could be a reason for under-reporting of depression in MICs [32, 46]. These limitations may have implications on under-estimations of associations in the study. However, previous work suggests this may not be a substantial problem as SAGE incorporated measures to minimise these issues [45]. There may be differential survey responses across countries. However, the survey methodology included strategies to detect and correct for systematic reporting biases in health interview surveys, such as vignette methods and objective performance tests [44]. Strategies were used to improve data comparability, such as utilising common definitions of concepts, common data collection methods and translations, rigorous sample design, and post-hoc harmonisation [44].

Additionally, this survey only asked if subjects were taking treatment (medicines, lifestyle changes), but did not measure self-reported treatment adherence (i.e. dosage, frequency, duration, etc) [37]. There are limitations of the Rose questionnaire for assessing angina, including it being short in length, and that subjects with mental health disorders may be more likely to have false positive results [47]. However, large cohort studies showed that the Rose questionnaire had good predictive ability, and false positives from individuals with mental health conditions may be minimal [47‐50]. This study did not adjust for the number of visits to a general practitioner in clinics to avoid possible over-adjustment bias. The associations between more comorbidities with better detection of previously undiagnosed NCDs [51], could be mediated by more clinic visits and interactions with healthcare professionals from having more comorbidities [12, 25, 26, 52]. The study was based on 9 NCDs, so future work could examine more conditions. For example, a large-scale Scotland study included 40 NCDs [53]. The study’s cross-sectional design does not allow for causal interpretations, and studies that use prospective cohort designs could examine how comorbidities cause treatment and control of NCDs in subjects that are followed-up prospectively, such as over a few years [6].

Our study revealed that subjects with more comorbidity had better detection of NCDs, but control was worse with more comorbidity. There are three principle possible explanations. First, it may be related to poor access to care, whereby patients with multiple chronic conditions are getting treatment for their conditions (perhaps from a pharmacy) but there is a lack of access to care from the primary care system [54‐57]. It could also relate to cost, such as a lack of comprehensive universal coverage fee at the point of care [36, 58‐60]. Second, it could be from suboptimal adherence to medications [59, 60]. It is well documented that adherence drops as polypharmacy rises [34, 36, 61]. The lack of intentional ongoing monitoring by physicians and allied health professionals might compound this problem, as regular follow-up could include checks on adherence and reinforce the importance of taking the prescribed medications. Third, it could be due to a lack of effectiveness of medication prescribed for each condition in patients with multiple chronic conditions [62]. For most treatments of individual NCDs, the evidence is based on randomised controlled trials which exclude patients with multiple NCDs [62‐64]. Thus what works in a patient with a single condition, may not work in a patient with the same NCD with comorbidities [65].

In reality, the problem may be a combination of these three possibilities. Primary healthcare clinicians need to improve the follow up on patients, in order to assess possible discontinuation of certain treatments due to adverse drug events, and the financial constraints that limit patients’ ability to go for follow up check-ups and refilling prescriptions [36, 56, 57]. Another clinical implication is on prioritising NCDs, whereby clinicians should assess NCD profiles of patients, prioritise treatment for patients who would have the greatest clinical benefit of better management and control, and consider personal preferences of patients for managing various NCDs [6, 66‐68].

Policies could prioritise NCDs that have higher prevalence or burden, such as focusing on lowering out-of-pocket expenditures for follow-up visits to healthcare providers and medicines [36, 69]. Policies also need to address continual accessibility to healthcare after first diagnosis. In MICs, it is not uncommon for patients to travel from rural residences to the urban areas to seek medical treatment, limiting their ability to regularly visit healthcare services and continue treatment, which results in suboptimal control of NCDs [52, 70, 71]. In addition, there are complexities of insurance policies. Health insurance from employers may be limited to certain districts or urban areas in MICs, and it is possible that healthcare for different conditions are covered by health providers in different physical locations [72, 73]. This problem aggravates the already existing fragmentation of care from utilising multiple health providers and services [53, 74].

Fragmented healthcare is a challenge for patients, policymakers, and clinicians. Reorganising healthcare structures particularly primary care, would benefit individuals with multiple chronic conditions who often have difficulty managing treatment due to polypharmacy, have multiple health providers and medical appointments, high financial burden, and suboptimal control of conditions [6, 11, 53, 74].