Care cascades for hypertension and diabetes: Cross-sectional evaluation of rural districts in Tanzania

The prevalence of noncommunicable diseases (NCDs), especially cardiovascular disease and diabetes, is rising sharply in sub-Saharan Africa (SSA) where the burden has increased 67% since 1990 [1]. It is estimated that over 125 million adults have hypertension, and 24 million adults have diabetes as of 2021, resulting in 2 million premature deaths every year in SSA [2,3].

This rapid epidemiological transition in SSA requires transformations in health systems. There is a need to strengthen primary care and address the long-standing constraints of underfunding and focus on episodic care for acute conditions [1,4]. While diabetes and hypertension can be effectively controlled with treatment, many health systems are currently underprepared to provide this care [5,6]. The persistently low levels of awareness and treatment for these conditions throughout SSA results in low levels of control, particularly in rural areas [7,8].

As effective chronic NCD treatment requires diagnosis and care continuity and coordination, a care cascade can be used to monitor the effective management of hypertension and diabetes [9]. The cascade-of-care, originally developed for chronic communicable diseases like HIV and tuberculosis, is a relevant framework for evaluating the health systems effectiveness for hypertension and diabetes care [10–12]. This method defines the proportion of people with a specific disease across steps of treatment, from screening to control. A cascade-of-care provides both a measure of disease control and identifies steps along the cascade with high proportion of loss to guide development of interventions and future policies.

In Tanzania, the Strategic and Action Plan for the Prevention and Control of NCDs was developed in 2016 to strengthen the health system capacity to provide appropriate chronic disease care [13]. However, current findings on the readiness to provide care have found low levels of diagnostic tools, training of healthcare providers, and availability of essential medicines for NCDs across all levels of the health system [14]. There are very low levels of hypertension and diabetes awareness and control across the country, with lower awareness in rural areas [15]. Poor adherence to care for those who have been diagnosed is driven by a combination of challenges including long waiting times, and poor point of care communication and patient understanding of the disease, availability of medication, and also high costs of care with regular visits, especially medication costs [16,17].

Of the 784 people in the sample, 771 had complete blood pressure measurements, and 707 had agreed to test their blood sugar. In our sample, 316 of the 771 who had their blood pressure measured (41%) screened positive, and 42 of the 707 screened for diabetes (6%) screened positive for diabetes, and 32 of the 699 who were screened for both (4%) were classified as having comorbid hypertension and diabetes (Table 1). When the age and gender profile of the survey was adjusted to fit the underlying Tanzanian population, the prevalence of hypertension in this population was 24%, and the age-adjusted prevalence of diabetes was considerably lower than national estimates, at 2% (See S1 Table for population-adjusted estimates). Respondent who screened positive for diabetes were more likely to be from Kilombero district, older, retired, and enrolled in a social health protection program compared to those who were not screened positive. Those screened positive for hypertension were older, more likely to be retired, enrolled in a social health protection program, and have lower educational attainment than those who were not screened positive for hypertension. Of the 42 who screened positive for diabetes and had their blood pressure measured, 32 had comorbid hypertension (78%), while that reflected only 11% of the 317 who screened positive for hypertension.

Fig 2 shows the percentage of those screened positive for hypertension or diabetes in each step in the care cascades. Of the 316 people screened positive for hypertension, 138 (43%, 95% CI: 38% to 49%) had no prior diagnosis of hypertension and 82 (25%, 95% CI 21% to 30%) were engaged in care. The next step, 66 of the 316 screened positive for hypertension were on treatment meaning they were on medication or retained in care and controlled without medicine (21%, 95% CI: 16% to 24%). At the far end of the cascade, 36 (11%, 95% CI: 8% to 15%) of those screened positive had their hypertension controlled at the point of testing. Of the 42 who screened positive for diabetes, 34 had prior diagnosis (81%, 95% CI: 69% to 93%) had no prior diabetes diagnosis while 28 (66%, 95% CI: 52% to 82%) engaged in care. All of those who were diagnosed and engaged in care were also retained in treatment. Of the 42 people screened positive for diabetes 50 (48%, 95% CI: 32% to 63%) had their diabetes controlled at the point of testing. The care cascades also differ across gender, age, and social health protection schemes, which can be used for further priority setting (see S1, S2, S3, and S4 Figs, demonstrating care cascade by these variables).

Fig 2. Care cascade for hypertension or diabetes, showing the percentage of all people who screened positive for the condition in each of the levels of the cascade: diagnosed, engaged in care, retained on treatment including lifestyle changes and pharmaceutical interventions, and controlled.

https://doi.org/10.1371/journal.pmed.1004140.g002

Fig 3 shows the mean prevalence in each step of the cascade as a share of the prior step. For people with hypertension, 45% (95% CI: 38% to 49%) had a previously been diagnosed with hypertension by a healthcare professional, 55% (95% CI: 47% to 64%) who had a prior diagnosis were engaged in care. Of those engaged in hypertensive care, 71% (95% CI: 61% to 81%) were actively retained in treatment, meaning they took medicine for their hypertension in the prior 7 days or they were controlled without medicine. Of those who were retained in hypertensive treatment, 43% (95% CI: 31% to 55%) were controlled. For those with diabetes, 86% (95% CI: 75% to 98%) had previously been told by a healthcare professional they had diabetes, 78% (95% CI: 63% to 93%) of people with a prior diagnosis were engaged in care, and 100% of those engaged in care were retained. Of the people engaged and retained in care, 69% (95% CI: 50% to 88%) had controlled diabetes at the point of testing.

Fig 3. Steps in the care cascade by percent of prior level for hypertension and diabetes.

This shows the share of people who were previously diagnosed with the condition if they screened positive, the percent who were engaged in care if previously diagnosed, the percent retained in treatment if they were engaged in care, and the percent that had their condition controlled if they were retained in treatment.

https://doi.org/10.1371/journal.pmed.1004140.g003

Table 2 provides the odds ratios from the multivariable regression analysis for each of the four steps of the hypertension care cascade. Males had much lower odds of being previously diagnosed for hypertension (OR 0.29; 95% CI [0.15 to 0.56] p < 0.005) compared to females. Increasing age was associated with increased odds of being previously diagnosed for hypertension (OR 1.05; 95% CI [1.02 to 1.07] p < 0.005), engaged in care (OR 1.04; 95% CI [1.01 to 1.07] p < 0.05), and retained in treatment (OR 1.03; 95% CI [1.01 to 1.05] p < 0.05). Completing primary school was also associated with higher odds of being diagnosed, engaged in care, and retained in treatment compared to not completing primary school. People with hypertension who had a prior diagnosis of diabetes had considerably higher odds of being engaged in care (OR 3.26; 95% CI [1.39 to 7.63] p < 0.005). Controlling for the demographic characteristics included in Table 2, and the prior diagnosis of diabetes, the NHIF enrollment was associated with higher odds of retention in treatment (OR 3.18; 95% CI [1.08 to 9.36] p < 0.05). Respondents with healthcare fee exemptions had higher odds of being previously diagnosed (OR 5.81; 95% CI [1.98 to 17.10] p < 0.005), engaged in care (OR 4.71; 95% CI [1.59 to 13.90] p < 0.005), and retained in treatment (OR 2.93; 95% CI [1.03 to 8.35] p < 0.05). Ps are derived from the Wald χ² test.

Narrow interventions focused only on single health system barriers to hypertension control are not enough, as indicated by the substantial drop-off at each step of the hypertension care cascade. As 59% of those screened hypertensive were unaware of their condition, screening efforts to improve awareness are potentially very important to improve control. However, screening efforts alone would be insufficient to improve health outcomes as other studies have found that linkage to care after a hypertension diagnosis is very low in Tanzania [24]. This is supported by our findings that 39% of those who were aware of their hypertension were not engaged in care, and 29% of those engaged in care were not on any treatment. The difference between engagement in care and on treatment provides a subtle but crucial distinction of people seeking care that are not accessing the treatments needed to control their condition, underscoring limitations in accessing medicines for hypertension. While other studies on antihypertensive drugs have reported high costs of medicines being a substantial barrier, 84% of the respondents in our survey who did not obtain all of their medications at the last health facility visit for their chronic condition reported that their prescribed medication was out of stock at the health facility (see S3 Table) [25]. This points to the importance of investments not only in medicines procurement but also in improving the efficiency of the whole medicines supply chain system. Additionally, even of those on treatment 57% are uncontrolled, meaning that less than 8% of people with hypertension have the condition under control.

Males had much lower odds of being previously diagnosed with hypertension, consistent with other research showing that men have lower engagement in the health system compared to females [26,27]. Older adults were more likely to be diagnosed, engaged in care, and retained in treatment, which may be related to perceptions of risk for cardiovascular events. While the 10-year cardiovascular disease risk is higher for older adults, younger people are still at risk for premature mortality. Risk also increases based on length of time that people are hypertensive without control, meaning that increasing diagnosis and treatment of people under 60 would further improve health outcomes. Despite the higher odds of treatment for older adults, they did not have higher levels of control, which is consistent with other research in SSA [28]. This might indicate poor quality of care received, especially regarding counseling, or need for additional caregiver support for older adults, as forgetfulness has been cited as a key barrier to medication adherence [25].

Low numbers of people with health insurance means it is hard to draw any conclusions regarding the associations between insurance on hypertensive care cascade (number of events for each covariate along the hypertensive care cascade is shown in S2 Table). NHIF coverage was associated with higher odds of being on treatment, but not associated with awareness, engagement in care, or control. One possible driver of this could be the comprehensive benefits of the NHIF including purchasing medications at private pharmacies. This would reduce the barrier of pharmacy stockouts and the higher prices of private pharmacies. Those with healthcare fee exemptions had higher odds of being diagnosed, engaged in care, and on treatment, suggesting that they work to improve access to care. Further research into the policies around healthcare fee exemptions, the population that receives them, and the other benefits associated with these exemptions would provide useful insight to direct policy changes to maximize the effect of these fee waivers. The respondents in our sample had higher levels of diagnosis, treatment, and control compared to other estimates of the hypertensive care cascades in SSA and were comparable to the mean level of diagnosis and control for other low- and middle-income countries (LMICs) [4]. However, this is still below the so-called “rule of halves” used to conceptualize the large proportion of people with undiagnosed, untreated, and uncontrolled hypertension, and very far below the WHO Global Action Plan for the Prevention and Control of NCDs target of 50% of eligible people receive drug therapy and counselling to prevent heart attacks and strokes [29,30]. Engagement in the care cascade for diabetes was much higher. Only 14% of those who screened positive for diabetes were unaware, and only 12% of those who had previously been diagnosed were not engaged in care. All of those engaged in care for diabetes were also on treatment, which is very different from those engaged in care for hypertension. Of those in care and on treatment, only 29% were uncontrolled, resulting in an overall control of diabetes of 54%. The differences of the care cascade for diabetes and hypertension may come from differences in patient demand, as hypertension is generally asymptomatic and recent health literacy campaigns in Tanzania have emphasized the threat of diabetes, which may result in more people seeking care and then adhering to treatment plans [31]. This possible driver is also suggested in research that found a common barrier in awareness and management of hypertension in Tanzania was the poor understanding of the chronicity of the diseases and associated long-term treatment [17]. The benefits of the increased engagement with the health system among people with a prior diabetes diagnosis in our sample also carried over to those with comorbid hypertension. When examining the multivariable analysis for the hypertensive care cascade, a diagnosis of comorbid diabetes is also strongly associated with higher odds of awareness, engagement, treatment, and control.

While the number that screened positive for diabetes is considerably smaller than for hypertension, the high levels of awareness, treatment, and control for diabetes suggests that improvements along the entire cascade are also possible for hypertensive treatment in this same setting with targeted interventions. Compared to other LMICs, especially in SSA, the respondents in our sample were more likely to have a prior diagnosis, be on treatment, and have control of their diabetes [32]. Our sample also had considerably lower prevalence of diabetes compared to other samples in Tanzania [21,32,33]. We relied primarily on a random blood glucose test as out of the 707 people who had their blood sugar tested, only 39 were in a fasted state. As a result, there may be people with diabetes in the population that were not included in this analysis, thereby biasing the sample to those with more extreme cases of diabetes or who were already in care. The low prevalence in our sample may be reflective of regional differences in prevalence. In a sample of three sites in eastern Africa, the prevalence varied significantly from 16% in rural Uganda to 8% in urban Tanzania [34]. Taking into account these limitations and the small sample of people with diabetes in our study, we must interpret the findings on diabetes care with caution.

This study has some other limitations that are important to consider in the interpretation of the results. We are reliant on patient self-report that a healthcare provider had given them a prior diagnosis of an NCD, and the timing of the most recent healthcare visits for a chronic condition, and current treatment use. Patient records from health facilities might have provided more detailed records of diagnosis or care engagement, but that would not be possible nor comprehensive in this setting. We also estimated the spacing of follow-up visits to classify engagement in care, rather than follow-up schedules tailored to individual patients. Our definition of hypertension and diabetes is also limited as it was measured at a single time point. While this is standard protocol for screening, actual diagnosis in health facilities relies on several separate measurements at different visits. Diabetes diagnosis, in particular, would generally require an additional fasting blood sugar measurement or hemoglobin A1C. Our reliance on sampling the self-identified household head who was available on the day the data collectors were in the town may bias the sample towards older respondents and may be the cause of the imbalance of the genders. Therefore, as this study was conducted in two districts in Tanzania that are mainly rural and overrepresent women and older people compared to the underlying population, the generalizability of the results is limited.