Multimorbidity and care for hypertension, diabetes and HIV among older adults in rural South Africa

Abstract Objective To examine how multimorbidity might affect progression along the continuum of care among older adults with hypertension, diabetes and human immunodeficiency virus (HIV) infection in rural South Africa. Methods We analysed data from 4447 people aged 40 years or older who were enrolled in a longitudinal study in Agincourt sub-district. Household-based interviews were completed between November 2014 and November 2015. For hypertension and diabetes (2813 and 512 people, respectively), we defined concordant conditions as other cardiometabolic conditions, and discordant conditions as mental disorders or HIV infection. For HIV infection (1027 people) we defined any other conditions as discordant. Regression models were fitted to assess the relationship between the type of multimorbidity and progression along the care continuum and the likelihood of patients being in each stage of care for the index condition (four stages from testing to treatment). Findings People with hypertension or diabetes plus other cardiometabolic conditions were more like to progress through the care continuum for the index condition than those without cardiometabolic conditions (relative risk, RR: 1.14, 95% confidence interval, CI: 1.09–1.20, and RR: 2.18, 95% CI: 1.52–3.26, respectively). Having discordant comorbidity was associated with greater progression in care for those with hypertension but not diabetes. Those with HIV infection plus cardiometabolic conditions had less progress in the stages of care compared with those without such conditions (RR: 0.86, 95% CI: 0.80–0.92). Conclusion Patients with concordant conditions were more likely to progress further along the care continuum, while those with discordant multimorbidity tended not to progress beyond diagnosis.


Introduction
Increases in ageing populations in low-and middle-income countries has contributed to a rising prevalence of multimorbidity, commonly defined as persons with more than one medical condition. 1 Previous studies have found that multimorbidity is associated with poorer clinical outcomes, 2 higher health expenditure and frequency of service use, [3][4][5][6] higher use of secondary than primary care, 7,8 and higher hospitalization rates among patients. 3,6,9 One limitation in the existing literature is that studies of multimorbidity often focus on simple counts of medical conditions. However, different combinations of diseases may affect a person's health and health care differently. To account for these differences, disease combinations can be categorized as either concordant (similar in risk profile and management) or discordant (not directly related in pathogenesis or management). 10 Theoretically, concordant conditions are more likely to be diagnosed and treated along with the index condition, because clinical guidelines often incorporate their interactions. For discordant conditions, however, the competing demands of dealing with different conditions may affect the quality of care provided. 11 Previous studies in high-income settings found that patients with diabetes 12,13 or hypertension 14,15 had higher odds of achieving testing and control goals when they had concordant conditions than discordant conditions. Diabetes patients with discordant conditions, on the other hand, had higher unplanned use of hospital services and specialized care than those with concordant conditions. 16 Little is known about the care of patients with human immunodeficiency virus (HIV) and multimorbidity, although studies in the United States of America found that patients with HIV received poorer care for their coexisting conditions than did those without HIV. [17][18][19] Much less is known about how the type of multimorbidity (concordant or discordant) affects a person's progression along the continuum of care in low-and middle-income countries. Our study aimed to fill this gap by studying the progression along the care continuum among people in South Africa with hypertension, diabetes or HIV infection, all prominent conditions contributing to the complex health transition underway in the country. Furthermore, this study assessed the effect of the type of multimorbidity on HIV care (and not on non-HIV comorbidities) among patients infected with HIV.

Study design
We analysed cross-sectional data from patients enrolled in the Health and Aging in Africa: a Longitudinal Study of an INDEPTH Community in South Africa. The main study is based the sub-district of Agincourt, in the Bushbuckridge area of Mpumalanga province in South Africa. 20 The study enrolled 5059 participants aged 40 years and older. Household-based interviews were completed between November 2014 and November 2015 using a primary survey instrument to collect data about respondents' demographic profile, medical conditions and economic status. More details on data collection are described elsewhere. 21 The study received ethical approvals from the University of the Witwatersrand human research ethics committee, the Mpumalanga province research and ethics committee, and the Harvard T.H. Chan School of Public Health office of human research administration.

Study setting
The Agincourt sub-district has six clinics and two health centres, and there are three district hospitals located 25-60 km from the study site. 20,22 Primary healthcare services are free of charge and most of out-of-pocket health expenditure for patients is incurred for transport, caregiver costs or private health care.
The Integrated Chronic Disease Management model was recently introduced in South Africa to address several elements of managing multimorbidity, including standardized clinical care based on national treatment protocols, and promotion of disease monitoring and management among patients. [23][24][25] In Agincourt, a patient with any symptom or disease arriving at a local clinic will be received by a nurse who is expected to address all the patient's needs. Those who visit the clinic primarily for HIV testing are directed to a nearby building staffed by health workers tasked solely with HIV testing. Patients are referred for the same management as other patients only if they are diagnosed as HIV positive.

Definitions
For this analysis, we studied three index conditions: (i) hypertension; (ii) diabe-tes; and (iii) HIV infection. We defined an index condition as a reference condition for which the continuum of care was evaluated, not as the time sequence in occurrence or diagnosis of multiple conditions. 26 For example, for an individual with hypertension plus other conditions, we assigned hypertension as the index condition and evaluated progression along the continuum of care for hypertension in relation to the presence of different types of either concordant or discordant multimorbidity. In addition to the three index conditions, we selected five others as concordant or discordant conditions: (i) dyslipidaemia; (ii) angina; (iii) depression; (iv) post-traumatic stress disorder; and (v) alcohol dependence. We ascertained the presence of the medical conditions based on the clinical diagnosis or clear clinical criteria (Box 1). We selected the medical conditions according to the data that were available in the main study, described in detail elsewhere. 31 We determined concordance and discordance based on the risk factors and multimorbidities for diagnosis and treatment in the South African national guidelines for hypertension and diabetes. [32][33][34] We found no definition of concordant diseases beyond opportunistic infections in the national HIV guidelines. For people with hypertension, we categorized other cardiometabolic conditions (dyslipidaemia, diabetes and angina) as concordant conditions, and mental disorders (depression, post-traumatic stress disorder and alcohol dependence) and HIV infection as discordant. Similarly, for people with diabetes, we classified other cardiometabolic conditions (hypertension, dyslipidaemia and angina) as concordant conditions, and mental disorders and HIV infection as discordant. For people with HIV, we considered any of the other conditions as discordant.
We defined the continuum of care for each index condition by four sequential stages of care for a patient: being tested for the disease (stage 1), knowing his or her diagnosis (stage 2), ever being initiated on treatment (stage 3) and currently being retained on treatment (stage 4). For hypertension and diabetes, the stage reached was determined from a patient's self-reporting. For HIV, we relied on both self-reported status and blood test results to determine progression. Patients with dried blood-spot results that showed exposure to antiretroviral therapy (ART) were considered to have reached the treatment stage and all preceding stages, even if they selfreported otherwise.

Statistical analyses
We first conducted descriptive analyses of the prevalence of the three index conditions as well as the prevalence of concordant and discordant conditions Box 1. Definitions of conditions in the study of multimorbidity and the care continuum in Agincourt sub-district, South Africa

Index conditions
Hypertension was defined as either mean systolic blood pressure ≥ 140 mmHg and mean diastolic blood pressure ≥ 90 mmHg or patients' self-report of receiving current treatment. Diabetes was defined as fasting blood glucose ≥ 126 mg/dL (defined as patients whose last meal was > 8 hours before specimen collection), non-fasting blood glucose ≥ 200 mg/dL or self-reported current treatment. Human immunodeficiency virus (HIV) status was ascertained either from collected dried blood spots that showed HIV infection or exposure to antiretroviral therapy or self-reported disease status.

Concordant and discordant conditions
Dyslipidaemia was one of the following criteria: self-reported disease status; elevated total cholesterol (≥ 6.21 mmol/L); low high-density lipoprotein cholesterol (1.19 mmol/L); elevated low-density lipoprotein cholesterol (> 4.10 mmol/L); elevated triglycerides (> 2.25 mmol/L). Angina was diagnosed using the Rose chest pain questionnaire. 27 Depression was defined as three or more symptoms of depression on the Center for Epidemiological Studies depression scale 8-item questionnaire. 28 Post-traumatic stress disorder was diagnosed as four or more symptoms on a seven-symptom screening scale. 29 Alcohol dependence was defined using the CAGE questionnaire. 30 Research Multimorbidity and the care continuum, South Africa Angela Y Chang et al.
by key sociodemographic covariates. Next, we constructed a count variable for each index condition to signify how many stages each respondent with that index condition had advanced along the corresponding continuum of care for that index condition, with a minimum count of zero and maximum of four. We fitted quasi-Poisson regression models to analyse the relationship between the number of stages respondents reached in the continuum of care and the type of multimorbidity. We used a series of logistic regression models to estimate the odds ratio (OR) and 95% confidence interval (CI) for associations between either concordant or discordant multimorbidities and the odds of advancing to each stage of the care continuum, conditional on having reached the previous stage. In the case of diagnosis, the logistic regression modelled the unconditional odds. We adjusted all regression models for sociodemographic covariates, including age, sex, education, country of origin, marital status, household size, employment status, having limitations in activities of daily living and wealth (measured in quintiles based on household asset ownership) and synthesized these using standard methods. 35 All analyses were conducted in R software version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results
Complete data on disease prevalence and continuum of care were available for 4447 respondents (88% of the whole sample of 5059). We excluded 135 people due to missing data about disease status of at least one disease category and 477 people due to missing dried bloodspot samples. Table 1 shows the prevalence of hypertension (63%, 2813 people), diabetes (12%, 512 people) and HIV (23%, 1027 people) as well as the prevalence of concordant and discordant conditions by sociodemographic covariates. Among patients with hypertension, 1535 (55%) had one or more additional cardiometabolic condition, 615 (22%) had one or more mental disorder and 480 (17%) were HIV positive. Among those with diabetes, 465 (91%) patients had other cardiometabolic conditions,  presented with cardiometabolic conditions and 181 (18%) with mental disorders. Reflecting the wider population profile, people with HIV tended to be younger, poorer, in employment and separated from partners compared with those with hypertension and diabetes.

Hypertension
Looking more closely at each stage of the continuum, having discordant medical conditions was associated with a higher likelihood of being tested for hypertension. This was true both among the entire

Discussion
In line with theories and empirical findings from high-income settings, [12][13][14] we found that having concordant conditions was associated with a higher likelihood of progressing further along the continuum of care for hypertension and diabetes in our study population. This may be explained by the emphasis that the South African hypertension guidelines place on diabetes and dyslipidaemia as important comorbidities,

Variable
Index condition  Multimorbidity and the care continuum, South Africa Angela Y Chang et al. and the emphasis on hypertension and dyslipidaemia in the diabetes guidelines. 32,33 These guidelines do not give much emphasis to HIV, although both mention it, and neither mention mental disorders. Moreover, providers may be more inclined to treat concordant conditions urgently to reach the target treatment outcomes for the index condition. For example, treating dyslipidaemia in patients may lead to targeting blood pressure control, because of the benefits of preventing the progression of coronary artery diseases. 36 On the other hand, having discordant conditions was not associated with worse care progression for hypertension and diabetes, contrary to experience in high-income settings. 11,13 Although some studies have shown that mental disorders are associated with poorer progression in care for cardiometabolic conditions, 36 we did not find a significant effect. Negative findings were observed only among people with HIV, where the presence of cardiometabolic (discordant) conditions was associated with less progress in HIV care. This is a concerning finding given that both HIV infection and the use of ART have been associated with increased risk of coronary heart disease and myocardial infarction. 37,38 Previous studies found lower quality of care for non-HIV conditions among HIV patients. [17][18][19] Factors that may have contributed to those findings include the lack of specific guidelines for HIV patients for treating diseases other than opportunistic infections; prioritization of short-term health needs; and the difficulty of balancing the demands of caring for complex patients with other medical and psychosocial problems.

Hypertension Diabetes HIV infection
Comparing across each stage in the continuum of care, both hypertension and diabetes patients with concordant or discordant conditions had a higher likelihood of reaching the first stages of care. This may be due to the lower opportunity costs involved for health-care providers and patients in relation to testing and diagnosis, versus those related to initiation and

Research
Multimorbidity and the care continuum, South Africa Angela Y Chang et al.
adherence to treatment. Testing and diagnosing hypertension involve simple procedures with relatively little effort required by providers, and thus the presence of any type of multimorbidity may increase the chance that the patient will be tested. However, the positive effect of discordant diseases may recede as the opportunity cost increases, as is the case for being initiated on and supported to adhere to treatment. More effort is required on the part of the practitioner to determine the right regimen, initiate the treatment, provide counselling on adherence and follow-up regularly to ensure the desired outcomes are met. Patients who have non-diabetes cardiometabolic conditions may be tested for diabetes, given the overlap in the risk factors, pathophysiological pathways and treatment guidelines. We did not see this positive effect of multimorbidity among people who were HIV-infected, perhaps due to stigma, practitioners' lower awareness of HIV among older people and the fact HIV testing requires more complex laboratory-based assessment than measuring blood pressure. Furthermore, we suggest that the negative association between HIV care and having cardiometabolic diseases may relate in part to how the clinics in Agincourt are organized. The separate procedure for HIV testing may explain why people with only HIV and no other conditions were more likely to be diagnosed with HIV conditional on being tested since they likely entered the clinic solely for receiving HIV care.
The findings also imply that the objective of the South Africa's Integrated Chronic Disease Management model may not yet be realized. While not examined empirically in our study,   barriers such as long waiting times, staff shortages and drug stock-outs may have negatively impacted the implementation of the management model and resulted in fewer visits made by the patients and shorter consultation times with providers. 25 The nurses may not be trained to diagnose or manage all diseases, and, given time constraints, they are often only able to address the patient's chief complaint and, in some cases, the concordant diseases that are listed in the guidelines. 25 For nationwide implementation of the integrated chronic disease management model, our findings suggest the need for improvements in leveraging one programme (such as the HIV programme) for scaling-up services for another condition (such as noncommunicable disease services), for example by putting more effort into ensuring patient engagement in stages with higher opportunity costs. There may be potential for benefits through the introduction of programmes, such as the Sustainable East Africa Research in Community Health's campaign and the United States President's Emergency Plan for AIDS Relief. [39][40][41] Implementing such joint programmes would make cardiometabolic disease management available alongside HIV services to bring populations with different types of multimorbidity into care. Our study is subject to several limitations. First, we assessed whether the presence of a concordant or dis-cordant condition was associated with progression in the care continuum, not whether being in care for one disease leads to being in care for another. Due to the cross-sectional nature of this study, we could not determine the time sequencing of the conditions or the care progression. We were also unable to assess causality on which type of multimorbidity affects care progression. Second, while the prevalence of the three index conditions and the concordant and discordant conditions were based on clinical criteria, data on the stages to which people progressed were selfreported, and our results therefore may have over-or underestimated coverage of different services. As the excluded samples were most commonly due to missing HIV measurements (due to patients' refusal to be tested), it is likely that we have underestimated the prevalence of HIV infection. The HIV prevalence within the sample is similar to the prevalence level found earlier in Agincourt. 42 Third, all conditions within the cardiometabolic and mental conditions were weighted equally, whereas it is plausible that specific combinations of diseases are associated with higher likelihood of progressing further along the care continuum. Finally, the study's comparability with existing studies and generalizability to settings with low HIV prevalence may be limited.
We conclude that the presence of any type of multimorbidity is associ-ated with a higher likelihood of being in stages of care with lower opportunity costs, while the presence of concordant conditions is associated with higher likelihood of being in stages with higher opportunity costs. Our findings from a relatively typical setting in rural South Africa have policy implications for enhancing access to testing and treatment services to improve service coverage and population health in the country. While we could not corroborate causality, further research, informed by forthcoming waves of the main study, will improve our understanding of the impact of different types of multimorbidity on health outcomes and the use of health services. ■