Content uploaded by Simiao Chen
Author content
All content in this area was uploaded by Simiao Chen on Feb 27, 2020
Content may be subject to copyright.
By Nikkil Sudharsanan, Simiao Chen, Michael Garber, Till Bärnighausen, and Pascal Geldsetzer
The Effect Of Home-Based
Hypertension Screening On Blood
Pressure Change Over Time In
South Africa
ABSTRACT
There is considerable policy interest in home-based screening
campaigns for hypertension in many low- and middle-income countries.
However, it is unclear whether such efforts will result in long-term
population-level blood pressure improvements without more
comprehensive interventions that strengthen the entire hypertension care
continuum. Using multiple waves of the South African National Income
Dynamics Study and the regression discontinuity design, we evaluated the
impact of home-based hypertension screening on two-year change in
blood pressure. We found that the home-based screening intervention
resulted in important reductions in systolic blood pressure for women
and younger men. We did not find evidence of an effect on systolic blood
pressure for older men or on diastolic blood pressure for either sex. Our
results suggest that home-based hypertension screening may be a
promising strategy for reducing high blood pressure in low- and middle-
income countries, but additional research and policy efforts are needed to
ensure that such strategies have maximum reach and impact.
High blood pressure, or hyperten-
sion, is a main cause of stroke
and cardiovascular disease and
carries a substantial health and
economic burden globally.1–5It
is a growing problem in South Africa, where
more than 25 percent of adults older than age
thirty-five are hypertensive, and hypertension-
related causes of death are estimated to account
for three of the top ten causes of death.6–8If
hypertension is detected, diagnosed, and treated
effectively, its health and mortality consequenc-
es can be reduced substantially.9,10 Unfortunate-
ly, among South African adults with hyperten-
sion, only 28 percent are aware of their
condition, and just 9 percent have their blood
pressure under control.11
Home-based screening for hypertension has
the potential to result in large populationwide
improvements in blood pressure control in
South Africa and other low- and middle-income
countries. First, hypertension screening is a rel-
atively straightforward and low-cost process.
Second, home-based screening may result in
greater population coverage than health facili-
ty–based screening by capturing people who are
unlikely to seek preventive care or care for ill-
nesses perceived as minor at health facilities.
Despite the considerable enthusiasm for home-
based screening,12,13 broad community- and
home-based screening efforts might not result
in blood pressure improvements if people who
are screened at home and identified as potential-
ly hypertensive do not confirm their diagnosis at
a health facility, or if people who are aware that
they are hypertensive do not initiate and adhere
to treatment. To date, there is a dearth of evi-
dence on whether home-based hypertension
screening will result in long-term blood pressure
improvements without more comprehensive in-
terventions that strengthen the entire hyperten-
sion care continuum.
doi: 10.1377/hlthaff.2019.00585
HEALTH AFFAIRS 39,
NO. 1 (2020): 124–132
©2020 Project HOPE—
The People-to-People Health
Foundation, Inc.
Nikkil Sudharsanan (nikkil
.sudharsanan@uni-
heidelberg.de) is lead of the
Population Health and
Development research group
at the Heidelberg Institute of
Global Health, Heidelberg
University, in Germany.
Simiao Chen is head of the
research unit, Health and
Population Economics,
Heidelberg Institute of Global
Health, Heidelberg University.
Michael Garber is a PhD
candidateintheDepartment
of Epidemiology, Rollins
School of Public Health,
Emory University, in Atlanta,
Georgia.
Till Bärnighausen is the
Alexander von Humboldt
University Professor and
director of the Heidelberg
Institute of Global Health,
Heidelberg University. He is
also senior faculty at the
Africa Health Research
Institute, in Somkhele, South
Africa, and an adjunct
professor of global health at
the Harvard T. H. Chan School
of Public Health, in Boston,
Massachusetts.
Pascal Geldsetzer is an
instructor in the Division of
Primary Care and Population
Health, Department of
Medicine, Stanford University,
in California.
124 Health Affairs January 2020 39:1
Global Health Policy
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
In this study we evaluated the real-world im-
pact of home-based hypertension screening on
two-year change in blood pressure in a nationally
representative cohort of South African adults.
We employed a novel application of the regres-
sion discontinuity design that took advantage of
the fact that the activities of the fieldworker team
as part of this cohort study closely mimicked
those of a home-based hypertension screening
campaign. Our aim was to inform researchers
and policy makers seeking to identify effective
ways to reduce rising levels of cardiovascular
disease in South Africa and other low- and mid-
dle-income countries.
This study was preregistered on ClinicalTrials
.gov (No. NCT03762304) and was exempt from
Institutional Review Board approval because it
used publicly available, deidentified secondary
data.
Study Data And Methods
National Income Dynamics Study We used da-
ta from the 2008, 2010–11, 2012, 2014–15, and
2017 waves of the National Income Dynamics
Study.14 The study is a nationally representative
longitudinal survey of approximately 28,000
people from 7,300 households across South
Africa. The study data contain a wide array of
social, economic, demographic, and health in-
formation for both individuals and households.
We provide detailed information on the sam-
pling procedures and survey activities in online
appendix II.15 Briefly, the study used a two-stage
cluster probability sample, with the Statistics
South Africa primary sampling units as the first
stage and dwellings within each primary sam-
pling unit as the second stage. If there were mul-
tiple households in a dwelling, each household
was assigned a unique identifier. If a member of
a household agreed to be interviewed, the house-
hold was included in the sample and all individ-
uals in the household were interviewed. In all,
7,305 of 10,642 households agreed to participate
in the baseline survey, for a 69 percent baseline
response rate. All individuals identified in the
baseline survey were treated as panel respon-
dents, and efforts were made to locate and rein-
terview them in each of the subsequent waves.
New household members were interviewed in
subsequent waves but followed longitudinally
only if they were present in the household again
in the follow-up waves of data collection. Our
analysis longitudinally followed individuals for
only one pair of waves. For example, if a person
was interviewed in 2008, we needed information
on that person only from the 2010–11 wave. The
between-wave loss to follow-up was 26 percent.
(See appendix I for more details on missing data
and loss to follow-up.)15 Our overall sample con-
tained all age-eligible individuals with nonmiss-
ing blood pressure information in both the base-
line and follow-up waves of data.
Intervention The intervention we studied
was fieldworkers’informing people in the house-
hold that their blood pressure was high, that
high blood pressure can have adverse health
consequences if left uncontrolled, and that they
should seek further care. This intervention oc-
curred as part of routine data collection for the
National Income Dynamics Study. Specifically,
fieldworkers collected two blood pressure meas-
urements on each adult member of the house-
hold using an Omron digital blood pressure
monitor and entered these readings on a health
information sheet (see appendix XIII).15 If either
of the two readings had a systolic blood pressure
of 140 mmHg or more or a diastolic blood pres-
sure of 90 mmHg or more, the fieldworker
checked a box that read (in the participant’s na-
tive language): “Your blood pressure readings
are higher than normal. High blood pressure
is dangerous because it makes the heart work
too hard. High blood pressure increases the risk
of heart disease and stroke. High blood pressure
can also cause other problems, such as heart
failure, kidney disease, and blindness. You can
control high blood pressure by taking action.”
Based on the reading, additional boxes were
checked to suggest how soon the participant
should seek medical care. The fieldworkers then
orally conveyed this information to participants
and gave them a copy of the filled-out health
information sheet in their native language.
Outcome Our primary outcome of interest was
between-wave changes in blood pressure. For
example, for a person whose blood pressure was
measured in 2008 and again in 2010–11, we
would estimate the impact of the intervention
in 2008 on their change in systolic and diastolic
blood pressure (separately) between those two
measurements. We used the average of the two
blood pressure measurements recorded in each
wave of the data. Since we used information
from five waves of data with approximately two
years between each wave, the outcome corre-
sponded, on average, to a two-year change in
blood pressure.
Statistical Methods We used the regression
discontinuity design to evaluate the impact of
home-based hypertension screening on blood
pressure change over time. (Detailed informa-
tion on the study design and estimation is in
appendix III.)15 In comparison to other observa-
tional study designs, the regression discontinu-
ity design is thought to be particularly appropri-
ate for estimating causal effects because it relies
on relatively weak assumptions that can be par-
January 2020 39:1 Health Affairs 125
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
tially verified empirically.16–18 Indeed, recent
studies have found that estimates using this de-
sign come close to estimates from randomized
clinical trials.19,20
In this study we took advantage of the fact that
the screening intervention was administered to
people only if they had a measured systolic blood
pressure of at least 140 mmHg or a diastolic
blood pressure of at least 90 mmHg. Intuitively,
the main assumption of the design was that peo-
ple just above these blood pressure cutoffs were
comparable to those just below the cutoff on all
factors related to blood pressure change over
time. The only difference between these two
groups was that those above the 140 mmHg sys-
tolic or 90 mmHg diastolic cutoffs received the
intervention. Therefore, the effect of the inter-
vention was estimated by comparing the average
two-year change in blood pressure among people
whose systolic blood pressure was just above
140 mmHg or whose diastolic blood pressure was
just above 90 mmHg (who were administered the
intervention and thus formed the intervention
group) to the change among those whose blood
pressure was just below those cutoffs (who were
just shy of receiving the intervention). In prac-
tice, the regression discontinuity design was im-
plemented with slightly weaker assumptions,
which we discuss in detail in appendix III.15
The main assumption of the design was that
other characteristics related to our outcome—
two-year change in blood pressure—did not sub-
stantially change at the cutoff points used by
fieldworkers to determine which people should
receive the intervention. There are few reasons to
believe that other characteristics related to the
outcome differed substantially between people
with blood pressure just above and just below the
cutoffs. First, blood pressure monitors measure
blood pressure with a degree of random mea-
surement error, and blood pressure varies ran-
domly over time within individuals.21 Therefore,
whether people had blood pressure just above
or below one of the cutoffs at the time of the
survey was effectively random. Second, the field-
workers did not use the cutoffs to provide any
other interventions, so the effect of the interven-
tion could not be confounded with that of other
programs. Third, the cutoffs do not represent
an underlying pathophysiological phenomenon
that occurs at these precise levels of blood pres-
sure.22 Therefore, there are no reasons to believe
that people just above and just below the cutoffs
are biologically different in ways that would
also affect their blood pressure change over
time. Lastly, we empirically tested whether indi-
vidual characteristics were substantially differ-
ent among people just above and just below the
cutoffs (appendix VII),15 which is similar to the
balance test routinely done in clinical trials.23
We did not find consistent evidence of differenc-
es in any of the pre-intervention variables that
we tested at the cutoffs.
Li mi tati on s Our study had several limita-
tions. First, we were unable to identify whether
fieldworkers administered the intervention or
not. Therefore, our results correspond to an
intention-to-treat estimate. This issue is not
unique to our study, however, and intention-
to-treat estimates are commonly used in clinical
trials in which participants’adherence to a treat-
ment or intervention cannot be ensured. Indeed,
the intention-to-treat estimate is a better mea-
sure of the real-life impact of an intervention
than the estimate of effectiveness under condi-
tions of perfect fieldworker adherence and inter-
vention fidelity.24
Second, 31 percent of households that were
selected for the National Income Dynamics
Study did not provide a response to the survey.
These households were more likely to be white
and located in urban areas.25
Third, 26 percent of the people in our sample
were lost to follow-up between waves. In appen-
dix IX we compare differences in baseline char-
acteristics between those who were and were not
lost to follow-up.15 We found that men and wom-
en lost to follow-up were more likely to report
having fair or poor self-rated health at baseline,
and that women lost to follow-up were more
likely to have more than secondary schooling.
However, we found no change to our conclusions
after we reestimated our main effects with in-
verse probability weights to adjust for these ob-
served differences. Inverse probability weight-
ing, however, cannot adjust for loss to follow-
up due to unobserved characteristics.
Fourth, 15 percent of age-eligible people were
dropped because of missing blood pressure data.
As a result of the foregoing limitations, our
results might not represent the effect that would
be observed in the overall South African popula-
tion if the intervention had a different impact on
individuals in households who did not respond,
were lost to follow-up, or were dropped due to
missing blood pressure data, compared to the
Theimpactofthe
intervention was fairly
large but still short of
clinical goals.
Global Health Policy
126 Health Affairs January 2020 39:1
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
impact on individuals included in the analysis.
A broader limitation of the regression discon-
tinuity design is that it estimates local effects (in
this case, only among people with blood pressure
near the cutoffs) and might not be generalizable
to the entire distribution (here, of blood pres-
sure). This limitation is especially important to
consider when interpreting our findings, if the
goal of screening policies is to identify high-risk
people who may have blood pressure far above
the cutoffs. Appendix III presents a full discus-
sion of all the limitations, validity checks, and
sensitivity analyses.15
Study Results
Sample Characteristics Our analytic samples
consisted of individuals in the overall samples
whose blood pressure was within the windows
around cutoffs that were used to form the inter-
vention and control groups. These windows ex-
tend both above and below the cutoffs and are
the range in which the analysis is conducted. The
size of the window is estimated empirically and is
therefore different for each sex and outcome.
There were few important differences between
people in the overall and analytic systolic blood
pressure samples. Members of the analytic sys-
tolic blood pressure sample were older for both
men (49.5 years versus 46.8 years) and women
(54.2 years versus 47.6 years), compared to their
respective overall systolic blood pressure sam-
ples (exhibit 1). Additionally, women in the ana-
lytic systolic blood pressure sample were less
likely to have more than secondary schooling
(10 percent versus 14 percent) and slightly more
likely to report having fair or poor self-rated
health (25 percent versus 20 percent). In con-
trast, we did not find any meaningful differences
between the diastolic blood pressure samples for
either men or women.
Baseline Maximum Blood Pressure And
Two-Year Change For men, there was little vi-
sual evidence of a discontinuity at the cutoff for
systolic blood pressure (exhibit 2). In contrast,
for women, there was evidence of a downward
jump at the cutoff, which suggests that the inter-
vention had an impact on their systolic blood
pressure change over time.
For diastolic blood pressure, we did not ob-
serve evidence of a potential intervention effect
among either sex (exhibit 3).
Impact Of The Intervention On Two-Year
Change In Blood P ressure Appendix table 2
Exhibit 1
Characteristics of the overall and analytic samples for South African adults ages 30 and older, 2008–17
Men Women
Characteristics Overall sample Analytic sample Overall sample Analytic sample
Systolic blood pressure
Number 6,163 2,265 11,396 2,801
Mean age, years (SD) 46.8 (13.1) 49.5 (14.0) 47.6 (13.5) 54.2 (13.5)
Urban 52% 50% 47% 44%
More than secondary schooling 16 14 14 10
Fair or poor self-rated health 16 17 20 25
Prior stroke 1 1 2 2
Prior diabetes 5 6 7 10
Prior heart attack 2 2 4 5
Smoker 42 41 8 9
Has health insurance 13 13 10 8
Diastolic blood pressure
Number 6,405 2,699 12,753 8,045
Mean age, years (SD) 45.3 (12.3) 44.8 (11.7) 46.1 (12.5) 46.6 (12.2)
Urban 53% 55% 48% 48%
More than secondary schooling 17 17 14 14
Fair or poor self-rated health 15 13 19 19
Prior stroke 1 1 2 2
Prior diabetes 4 4 6 7
Prior heart attack 2 2 4 4
Smoker 42 40 8 9
Has health insurance 13 15 10 10
SOURCE Authors’analysis of data from the 2008, 2010–11, 2012, 2014–15, and 2017 waves of the National Income Dynamics Study.
NOTES The overall sample is explained in the text. The analytic sample is the people in the overall sample whose blood pressure was
within the windows around the blood pressure cutoffs (explained in the text) used to estimate the effect of the intervention on two-
year change in blood pressure. Appendix IVcontains an extended version of this exhibit (see note 15 in text). SD is standard deviation.
January 2020 39:1 Health Affairs 127
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
presents the regression-discontinuity estimates
of the impact of the intervention on two-year
change in blood pressure.15 These estimates con-
firm the visual evidence presented in exhibits 2
and 3.We found that the intervention resulted in
a systolic blood pressure reduction of 4.7 mmHg
for women (95% confidence interval: −12.6,
−2.1; p¼0:006). In contrast, we did not find
evidence that the intervention lowered diastolic
blood pressure for women or either blood pres-
sure outcome for men.
Heterogeneity Analyses Exhibit 4 plots the
effect of the intervention on two-year change
in blood pressure separately by sex, age, and
schooling. We found evidence that the impact
of the intervention on systolic blood pressure
was more pronounced among adults ages 30–
45 than among older age groups. For men
ages 30–45, the intervention resulted in a reduc-
tion in systolic blood pressure of 7.0 mmHg
(p¼0:022), compared to essentially null effects
for men in the other age groups. Similarly, for
women, the intervention had the largest impact
on systolic blood pressure for those ages 30–45: a
reduction of 9.1 mmHg. However, this effect was
estimated with a very wide confidence interval.
In contrast, we did not find evidence of hetero-
geneity in the impact of the intervention on sys-
tolic blood pressure by schooling groups or on
diastolic blood pressure for any of the groups.
Robustness And Validity We tested the as-
sumption that there were no significant discon-
tinuities in other variables that could also influ-
ence the outcome at the cutoffs for a number
of baseline pre-intervention variables (appen-
dix VII).15 For women in the systolic blood pres-
sure sample, we found no evidence of significant
changes at the cutoff for any of the pretreatment
variables. For men in the systolic blood pressure
sample, we found a small increase in age at
the cutoff, which suggests that the intervention
group was slightly older than the control group.
Next, we tested whether fieldworkers may
have deliberately underreported respondents’
baseline blood pressure measurement to avoid
having to administer the intervention. We did
this by examining the density of baseline blood
pressure to check whether there was a bunching
of individuals just below the cutoffs. We did not
find evidence of bunching suggestive of manip-
ulation (appendix VI).15 The results presented
here were also robust to the size of the window
around the cutoffs used to form the treatment
and control groups (appendix VIII)15 and to po-
tential selection bias introduced by loss to fol-
low-up between waves (appendix X).15 Lastly, our
results were consistent when we split the sample
by pairs of waves instead of pooling all five waves
of data (appendix XI).15
Discussion
We found that home-based hypertension screen-
ing resulted in an important 4.7 mmHg reduc-
tion in systolic blood pressure for South African
women.While there was no evidence of an impact
of hypertension screening among men overall,
the intervention did result in a reduction in sys-
tolic blood pressure of 7.0 mmHg for younger
men (those ages 30–45). The age-variation find-
ings were from a subgroup analysis that we did
not specify in our preregistration analysis plan;
however, we found a consistent advantage for
younger men when we examined each pair of
waves separately (appendix XI).15 This provides
some evidence of a consistent and nonspurious
effect. In contrast to systolic blood pressure, we
found no evidence that the intervention reduced
diastolic blood pressure among either men or
women. These results were consistent across
multiple robustness checks.
Our finding of a beneficial impact on systolic
Exhibit 2
Two-year change in systolic blood pressure (BP) among South African adults ages 30 and
older, by sex and baseline systolic BP, 2008–17
SOURCE Authors’analysis of data from the 2008, 2010–11, 2012, 2014–15, and 2017 waves of the
National Income Dynamics Study. NOTES Theverticallineat140mmHgisthecutoffpointafterwhich
the intervention was administered. Each point in the exhibit is the average two-year change for each
single-unit BP value, and the lines represent local linear fits separately on each side of the cutoff
point. Women have fewer values than men do because the range around the intervention cutoff was
determined empirically, and this resulted in a wider range for men than women.
Global Health Policy
128 Health Affairs January 2020 39:1
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
blood pressure among women overall but not
among men overall is consistent with a large
literature on chronic diseases in low- and middle-
income countries that generally has found great-
er levels of health-seeking behavior and treat-
ment adherence among women compared to
men. For example, the benefits of antiretroviral
treatment scale-up in South Africa—and sub-
Saharan Africa more broadly—has dispropor-
tionately benefited women,26 who tend to have
better medication adherence, care retention, and
health outcomes than men do.27 Qualitative stud-
ies suggest that this might be because men tend
to view health care facilities as being designated
for women and children,28,29 and gender norms
expect men to endure bad health instead of seek-
ing help.30,31 The restricted hours when health
care facilities are open may also make it more
difficult for men to seek care since in many pop-
ulations they are more likely than women to
work outside of their communities during the
day.32 Similar factors are likely to affect men’s
care seeking for cardiovascular disease risk fac-
tors. Indeed, the few large-scale studies to date
on care seeking for hypertension in low- and
middle-income countries have found higher
rates of awareness, treatment, and control of
hypertension among women than men.5,11,33,34
One unexpected finding is that the interven-
tion resulted in reductions in systolic, but not
diastolic, blood pressure. This pattern may be
because antihypertensive medicines often result
in greater reductions in systolic compared to
diastolic blood pressure, with this difference be-
coming larger with increasing age.35 Greater re-
ductions in systolic blood pressure are especially
pronounced when people are treated with thia-
zide diuretics,36 which are the recommended
first-line antihypertensive medications in South
Africa.36,37 Importantly, lowering systolic blood
pressure is the more relevant target for the pre-
vention of cardiovascular events and mortali-
ty.38–41 This is especially so in the context of aging
populations such as South Africa’s, since systolic
blood pressure continues to rise in older age
while diastolic blood pressure tends to level off
in midlife.22
The impact of the intervention was fairly large
but still short of clinical goals. For reference,
most clinical protocols suggest that people near
the systolic blood pressure cutoff of 140 mmHg
should aim for a target blood pressure below
130 mmHg.42 The improvements we observed
relative to clinical goals might reflect losses at
any of multiple steps of the care cascade. First,
people who are screened and identified as poten-
tially hypertensive might not seek further care.
This hypothesis is consistent with descriptive
studies from African countries that generally
have found low levels of health care linkage
following a home-based screening.43–45 Second,
people who are diagnosed and prescribed treat-
ment might not initiate treatment or might not
adhere to treatment after initiation. Studies on
the cascade of care for hypertension from Afri-
can countries have found low levels of treatment
and control among people diagnosed with hyper-
tension.5,46 However, the contribution of treat-
ment initiation and adherence to the number of
hypertensive people with suboptimal blood pres-
sure is small when compared to the substantial
share of people with hypertension who do not
make contact with the health system and are not
formally diagnosed.11,33 This suggests that low
levels of health-seeking behavior following a
positive home-based screening for hypertension
may be the most important contributor to the
low effect of screening on blood pressure reduc-
tions over time found in this study.
Policy Implications
Population aging in South Africa is expected to
result in an additional 9–12 million people in
need of care for hypertension by 2050.47 South
Africa’s health system is ill prepared for provid-
ing this level of care and will need to develop new
systems to achieve widespread blood pressure
control.13 Controlling blood pressure at the pop-
Exhibit 3
Two-year change in diastolic blood pressure (BP) among South African adults ages 30 and
older, by sex and baseline diastolic BP, 2008–17
SOURCE Authors’analysis of data from the 2008, 2010–11, 2012, 2014–15, and 2017 waves of the
National Income Dynamics Study. NOTES The vertical line at 90 mmHg is the cutoff point after which
the intervention was administered.The points and lines are explained in the notes to exhibit 2, as is
thedifferenceinvaluesbetweenmenandwomen.
January 2020 39:1 Health Affairs 129
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
ulation level is the result of several sequential
steps, starting from identifying people with hy-
pertension, moving through treatment initia-
tion and adherence, and ultimately leading to
controlled blood pressure. Since a substantial
share of individuals are lost at each step of this
care cascade in both high-income countries and
low- and middle-income countries,5,48 interven-
ing at each step has the potential to improve
populationwide blood pressure control.49
Improving the detection of hypertension is
potential low-hanging fruit for achieving popu-
lationwide blood pressure improvements be-
cause hypertension screening is comparatively
easier and more affordable than interventions
targeted at other steps of the care cascade—such
as interventions to improve linkage to care fol-
lowing a positive screening or improving treat-
ment initiation and adherence. The main contri-
bution of our study is that we examined whether
home-based hypertension screening alone could
result in meaningful improvements in blood
pressure control without devoting additional
resources to addressing the more complex steps
of the care cascade.
We found that home-based screening may
need to be combined with interventions that ad-
dress other care cascade steps to result in cost-
effective and populationwide improvements in
blood pressure control. For example, decentral-
izing hypertension care to community health
workers may be a promising strategy for improv-
ing linkage to health facilities and treatment
initiation and adherence following a positive
household hypertension screening.
At the health facility level, improving the qual-
ity of hypertension care may ensure that people
understand the importance of blood pressure
control, how to control their blood pressure,
and how often they need to have follow-up visits.
Exhibit 4
Two-year changes in systolic and diastolic blood pressure (BP) among South African adults ages 30 and older, by sex, age,
and schooling, 2008–17
SOURCE Authors’analysis of data from the 2008, 2010–11, 2012, 2014–015, and 2017 waves of the National Income Dynamics Study.
NOTES The error bars represent 95 percent confidence intervals. The estimates use robust standard errors that are clustered at the
individual level. The estimate for “more than secondary schooling”for women in the systolic blood pressure sample was omitted
because of a small sample size.
Global Health Policy
130 Health Affairs January 2020 39:1
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
This may result in a greater share of people initi-
ating treatment and ultimately achieving blood
pressure control.
Lastly, achieving blood pressure control for
hypertensive patients requires several repeated
visits to ensure that medicines are appropriately
dosed and to monitor blood pressure over time.
Health information systems have the potential to
alleviate the burden of needing multiple visits for
both patients and providers by tracking which
patients need to see a provider, updating physi-
cians about patients’current drug regimens, and
automatically reminding patients not to miss
visits.
Conclusion
Home-based hypertension screening may be a
promising strategy for reducing raised blood
pressure in low- and middle-income countries.
However, further work is needed to ensure that
such strategies have maximum reach and im-
pact. Developing and testing interventions to
maximize the proportion of individuals who
achieve hypertension control following a home-
based screening is a critical next step for both
research and policy. ▪
An earlier version of this article was
presented at the Annual Meeting of the
Society for Epidemiologic Research in
Minneapolis, Minnesota, June 20, 2019.
Michael Garber was supported by a
grant from the National Heart, Lung, and
Blood Institute (Grant No.
F31HL143900). The content is solely
the responsibility of the authors and
does not necessarily represent the
official views of the National Institutes
of Health. Till Bärnighausen was
supported by the Alexander von
Humboldt University Professor Award.
Nikkil Sudharsanan and Simiao Chen
made equal contributions and are joint
first authors.
NOTES
1Gajalakshmi V, Lacey B, Kanimozhi
V, Sherliker P, Peto R, Lewington S.
Body-mass index, blood pressure,
and cause-specific mortality in India:
a prospective cohort study of 500
810 adults. Lancet Glob Health.
2018;6(7):e787–94.
2Sudharsanan N. Population-level
mortality benefits of improved blood
pressure control in Indonesia: a
modelling study. Int J Epidemiol.
2019;48(3):954–65.
3Bloom DE, Cafiero ET, Jané-Llopis
E, Abrahams-Gessel S, Bloom LR,
Fathima S, et al. The global eco-
nomic burden of non-communicable
diseases [Internet]. Geneva: World
Economic Forum; 2011 Sep [cited
2019 Nov 13]. Available from: http://
www3.weforum.org/docs/WEF_
Harvard_HE_GlobalEconomic
BurdenNonCommunicable
Diseases_2011.pdf
4Kontis V, Mathers CD, Rehm J,
Stevens GA, Shield KD, Bonita R,
et al. Contribution of six risk factors
to achieving the 25x25 non-
communicable disease mortality re-
duction target: a modelling study.
Lancet. 2014;384(9941):427–37.
5Geldsetzer P, Manne-Goehler J,
Marcus ME, Ebert C, Zhumadilov Z,
Wesseh CS, et al. The state of hy-
pertension care in 44 low-income
and middle-income countries: a
cross-sectional study of nationally
representative individual-level data
from 1.1 million adults. Lancet.
2019;394(10199):652–62.
6Cois A, Ehrlich R. Antihypertensive
treatment and blood pressure trends
among South African adults: a re-
peated cross-sectional analysis of a
population panel survey. PLoS One.
2018;13(8):e0200606.
7Statistics South Africa. Mortality and
causes of death in South Africa,
2016: findings from death notifica-
tion [Internet]. Pretoria: Statistics
South Africa; 2018 Mar 27 [cited
2019 Nov 13]. (Statistical Release
No. P0309.3). Available from:
http://www.statssa.gov.za/
publications/P03093/P030932016
.pdf
8Osetinsky B, Hontelez JAC, Lurie
MN, McGarvey ST, Bloomfield GS,
Pastakia SD, et al. Epidemiological
and health systems implications of
evolving HIV and hypertension in
South Africa and Kenya. Health Aff
(Millwood). 2019;38(7):1173–81.
9Collins R, Peto R, MacMahon S,
Hebert P, Fiebach NH, Eberlein KA,
et al. Blood pressure, stroke, and
coronary heart disease. Part 2, short-
term reductions in blood pressure:
overview of randomised drug trials
in their epidemiological context.
Lancet. 1990;335(8693):827–38.
10 MacMahon S, Peto R, Cutler J,
Collins R, Sorlie P, Neaton J, et al.
Blood pressure, stroke, and coronary
heart disease. Part 1, prolonged dif-
ferences in blood pressure: pro-
spective observational studies cor-
rected for the regression dilution
bias. Lancet. 1990;335(8692):
765–74.
11 Berry KM, Parker WA, Mchiza ZJ,
Sewpaul R, Labadarios D, Rosen S,
et al. Quantifying unmet need for
hypertension care in South Africa
through a care cascade: evidence
from the SANHANES, 2011–2012.
BMJ Glob Health. 2017;2(3):
e000348.
12 Olsen MH, Angell SY, Asma S,
Boutouyrie P, Burger D, Chirinos JA,
et al. A call to action and a lifecourse
strategy to address the global burden
of raised blood pressure on current
and future generations: the Lancet
Commission on Hypertension. Lan-
cet. 2016;388(10060):2665–712.
13 Republic of South Africa Department
of Health. Strategic Plan for the
Prevention and Control of Non-
Communicable Diseases 2013–17
[Internet]. Pretoria: Department of
Health; [cited 2019 Nov 13]. Avail-
able from: https://extranet.who.int/
ncdccs/Data/ZAF_B3_NCDs_
STRAT_PLAN_1_29_1_3%5B2%5D
.pdf
14 University of Cape Town, Southern
Africa Labour and Development Re-
search Unit. South Africa—National
Income Dynamics Study Wave 1,
2008 [Internet]. Cape Town: Uni-
versity of Cape Town; 2019 Aug 30
[cited 2019 Nov 27]. Available from:
https://www.datafirst.uct.ac.za/
dataportal/index.php/catalog/451/
study-description
15 To access the appendix, click on the
Details tab of the article online.
16 Bor J, Moscoe E, Mutevedzi P,
Newell ML, Bärnighausen T. Re-
gression discontinuity designs in
epidemiology: causal inference
without randomized trials. Epide-
miology. 2014;25(5):729–37.
17 Oldenburg CE, Moscoe E,
Bärnighausen T. Regression discon-
tinuity for causal effect estimation in
epidemiology. Curr Epidemiol Rep.
2016;3:233–41.
18 Moscoe E, Bor J, Bärnighausen T.
Regression discontinuity designs are
underutilized in medicine, epidemi-
ology, and public health: a review of
current and best practice. J Clin
Epidemiol. 2015;68(2):122–33.
19 van Leeuwen N, Lingsma HF,
January 2020 39:1 Health Affairs 131
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
Mooijaart SP, Nieboer D, Trompet S,
Steyerberg EW. Regression discon-
tinuity was a valid design for di-
chotomous outcomes in three ran-
domized trials. J Clin Epidemiol.
2018;98:70–9.
20 Maas IL, Nolte S, Walter OB, Berger
T, Hautzinger M, Hohagen F, et al.
The regression discontinuity design
showed to be a valid alternative to a
randomized controlled trial for es-
timating treatment effects. J Clin
Epidemiol. 2017;82:94–102.
21 Kallioinen N, Hill A, Horswill MS,
Ward HE, Watson MO. Sources of
inaccuracy in the measurement of
adult patients’resting blood pres-
sure in clinical settings: a systematic
review. J Hypertens. 2017;35(3):
421–41.
22 Lewington S, Clarke R, Qizilbash N,
Peto R, Collins R. Age-specific rele-
vance of usual blood pressure to
vascular mortality: a meta-analysis of
individual data for one million adults
in 61 prospective studies. Lancet.
2002;360(9349):1903–13. Erratum
in: Lancet. 2003;361(9362):1060.
23 Roberts C, Torgerson DJ. Under-
standing controlled trials: baseline
imbalance in randomised controlled
trials. BMJ. 1999;319(7203):185.
24 DeMets DL, Cook T. Challenges of
non-intention-to-treat analyses.
JAMA. 2019;321(2):145–6.
25 Woolard I, Leibbrandt MV, De
Villiers L. Design and methodologi-
cal issues. Studies in Economics and
Econometrics. 2010;34(3):7–24.
26 Bor J, Rosen S, Chimbindi N, Haber
N, Herbst K, Mutevedzi T, et al. Mass
HIV treatment and sex disparities in
life expectancy: demographic sur-
veillance in rural South Africa. PLoS
Med. 2015;12(11):e1001905, discus-
sion e1001905.
27 Druyts E, Dybul M, Kanters S,
Nachega J, Birungi J, Ford N, et al.
Male sex and the risk of mortality
among individuals enrolled in anti-
retroviral therapy programs in
Africa: a systematic review and meta-
analysis. AIDS. 2013;27(3):417–25.
28 Skovdal M, Campbell C, Madanhire
C, Mupambireyi Z, Nyamukapa C,
Gregson S. Masculinity as a barrier
to men’s use of HIV services in
Zimbabwe. Global Health. 2011;7:13.
29 Katirayi L, Chadambuka A,
Muchedzi A, Ahimbisibwe A,
Musarandega R, Woelk G, et al.
Echoes of old HIV paradigms: re-
assessing the problem of engaging
men in HIV testing and treatment
through women’s perspectives. Re-
prod Health. 2017;14(1):124.
30 Courtenay WH. Constructions of
masculinity and their influence on
men’s well-being: a theory of gender
and health. Soc Sci Med. 2000;
50(10):1385–401.
31 Jewkes R, Morrell R. Gender and
sexuality: emerging perspectives
from the heterosexual epidemic in
South Africa and implications for
HIV risk and prevention. J Int AIDS
Soc. 2010;13:6.
32 Sharma M, Barnabas RV, Celum C.
Community-based strategies to
strengthen men’s engagement in the
HIV care cascade in sub-Saharan
Africa. PLoS Med. 2017;14(4):
e1002262.
33 Chow CK, Teo KK, Rangarajan S,
Islam S, Gupta R, Avezum A, et al.
Prevalence, awareness, treatment,
and control of hypertension in rural
and urban communities in high-,
middle-, and low-income countries.
JAMA. 2013;310(9):959–68.
34 Basu S, Millett C. Social epidemiol-
ogy of hypertension in middle-
income countries: determinants of
prevalence, diagnosis, treatment,
and control in the WHO SAGE study.
Hypertension. 2013;62(1):18–26.
35 Wang JG, Staessen JA, Franklin SS,
Fagard R, Gueyffier F. Systolic and
diastolic blood pressure lowering as
determinants of cardiovascular out-
come. Hypertension. 2005;45(5):
907–13.
36 Musini VM, Nazer M, Bassett K,
Wright JM. Blood pressure–lowering
efficacy of monotherapy with thia-
zide diuretics for primary hyperten-
sion. Cochrane Database Syst Rev.
2014;(5):CD003824.
37 Seedat YK, Rayner BL, Veriava Y.
South African hypertension practice
guideline 2014. Cardiovasc J Afr.
2014;25(6):288–94.
38 Pastor-Barriuso R, Banegas JR,
Damián J, Appel LJ, Guallar E. Sys-
tolic blood pressure, diastolic blood
pressure, and pulse pressure: an
evaluation of their joint effect on
mortality. Ann Intern Med. 2003;
139(9):731–9.
39 Stamler J, Stamler R, Neaton JD.
Blood pressure, systolic and dia-
stolic, and cardiovascular risks. US
population data. Arch Intern Med.
1993;153(5):598–615.
40 Ettehad D, Emdin CA, Kiran A,
Anderson SG, Callender T,
Emberson J, et al. Blood pressure
lowering for prevention of cardio-
vascular disease and death: a sys-
tematic review and meta-analysis.
Lancet. 2016;387(10022):957–67.
41 Bundy JD, Li C, Stuchlik P, Bu X,
Kelly TN, Mills KT, et al. Systolic
blood pressure reduction and risk of
cardiovascular disease and mortality:
a systematic review and network
meta-analysis. JAMA Cardiol. 2017;
2(7):775–81.
42 Carey RM, Whelton PK. Prevention,
detection, evaluation, and manage-
ment of high blood pressure in
adults: synopsis of the 2017 Ameri-
can College of Cardiology/American
Heart Association hypertension
guideline. Ann Intern Med. 2018;
168(5):351–8.
43 Bovet P, Gervasoni JP, Mkamba M,
Balampama M, Lengeler C, Paccaud
F. Low utilization of health care
services following screening for hy-
pertension in Dar es Salaam
(Tanzania): a prospective popula-
tion-based study. BMC Public
Health. 2008;8:407.
44 Siedner MJ, Baisley K, Orne-
Gliemann J, Pillay D, Koole O, Wong
EB, et al. Linkage to primary care
after home-based blood pressure
screening in rural KwaZulu-Natal,
South Africa: a population-based
cohort study. BMJ Open. 2018;
8(12):e023369.
45 Doulougou B, Kouanda S,
Ouédraogo GH, Meda BI, Bado A,
Zunzunegui MV. Awareness, treat-
ment, control of hypertension, and
utilization of health care services
following screening in the North-
central region of Burkina Faso. Pan
Afr Med J. 2014;19:259.
46 Bovet P, Burnier M, Madeleine G,
Waeber B, Paccaud F. Monitoring
one-year compliance to antihyper-
tension medication in the Sey-
chelles. Bull World Health Organ.
2002;80(1):33–9.
47 Sudharsanan N, Geldsetzer P. Im-
pact of coming demographic
changes on the number of adults in
need of care for hypertension in
Brazil, China, India, Indonesia,
Mexico, and South Africa. Hyper-
tension. 2019;73(4):770–6.
48 Zhou B, Danaei G, Stevens GA, Bixby
H, Taddei C, Carrillo-Larco RM, et al.
Long-term and recent trends in hy-
pertension awareness, treatment,
and control in 12 high-income
countries: an analysis of 123 na-
tionally representative surveys. Lan-
cet. 2019;394(10199):639–51.
49 Chen S, Sudharsanan N, Huang F,
Liu Y, Geldsetzer P, Bärnighausen T.
Impact of community based screen-
ing for hypertension on blood pres-
sure after two years: regression dis-
continuity analysis in a national
cohort of older adults in China. BMJ.
2019;366:I4064.
Global Health Policy
132 Health Affairs January 2020 39:1
Downloaded from HealthAffairs.org on February 27, 2020.
Copyright Project HOPE—The People-to-People Health Foundation, Inc.
For personal use only. All rights reserved. Reuse permissions at HealthAffairs.org.
