A companion commentary by Wang et al.
1
summarizes recent analyses at the Centers for Disease Control and Prevention of the
economic burden of cardiovascular disease (CVD) and of the cost effectiveness of interventions
targeting hypertension and CVD. The purpose of this editorial is to consider how researchers
and policy makers can use such findings to inform prevention efforts.
Cost-of-illness estimates are commonly generated to call attention to the economic
burden of various diseases and health risk factors (e.g., smoking) and the potential
economic gains from prevention.
2
Hypertension is both a disease and a risk factor for other CVDs. Even slight variations
in the methods used can produce different cost-of-illness estimates. The methodologic
choices include whether estimates are incidence-based or prevalence-based; to what
extent they control for covariates (e.g., comorbid conditions); the sources of data;
types of costs included; and the analytic perspective.
3
Most published cost-of-illness estimates are prevalence-based, meaning that they are
estimates of costs for the prevalent population affected by disease in a given year.
Such estimates allow one to project the potential cost savings under the counterfactual
scenario that the prevalent cases of disease had never occurred. In contrast, incidence-based
cost-of-illness estimates calculate the present value of attributable costs in both
current and future years for a cohort of incident cases, including costs associated
with disease complications and sequelae. The present value of costs of an incident
case is calculated by multiplying annual attributable costs for disease stratified
by age in years or years since onset by life table survival probabilities and discounting
to the present. Such estimates can be used to project cost savings from the prevention
of new cases of disease in cost-effectiveness analyses (CEAs) of prevention strategies.
All five original cost-of-illness analyses reported in this supplement are prevalence-based
4–8
and are appropriate for use in raising awareness of the economic burden of hypertension
and CVD.
Most cost-of-illness estimates for chronic disease are the outputs of regression models
that control for various covariates, including comorbidity. Deciding which comorbid
conditions should be controlled in a given analysis depends on the study question.
For analyses of disease-attributable cost excluding comorbid conditions that are downstream
in the causal pathway (i.e., conditions for which the disease is a risk factor) avoids
underestimating the economic burden. Thus, for example, an analysis of hypertension-related
costs among adults with diabetes by Wang and colleagues
6
treated stroke and heart disease as complications of hypertension rather than as comorbid
conditions to be controlled as covariates. In contrast, an analysis by Park et al.
5
of medical expenditures for people with hypertension included the number of comorbid
diagnoses as covariates because the purpose was to understand how costs among people
with hypertension vary based on the number of relevant diagnoses.
Almost all U.S. cost-of-illness analyses use expenditure data to estimate medical
costs. Insurance claims data can be used to estimate costs for specific payers (e.g.,
Medicaid or Medicare), or payer type (e.g., employer-sponsored health plans), and
population survey data, in particular the Medical Expenditure Panel Survey, is used
to estimate expenditures for all payers. All four original cost-of-illness analyses
of medical costs used Medical Expenditure Panel Survey data to estimate all-payer
mean expenditures.
5–8
A literature review by Chapel and colleagues
9
of cost-of-illness estimates for Medicaid-enrolled working-age adults with chronic
conditions found that most published studies used Medicaid claims data, with only
a few using Medical Expenditure Panel Survey data.
The cost-of-illness study by Joo et al.
4
used survey data from the Health and Retirement Study on reported hours of informal
caregiving time to estimate the economic burden of caregiving associated with falls
and stroke in the elderly. The economic value of informal caregiving time, mostly
provided by family members, is a large part of the economic cost of many chronic,
disabling conditions. Since 1996, U.S. guidelines have called for the inclusion of
informal caregiving costs as direct costs in societal-perspective CEAs.
10
There is no consensus, though, on how to estimate the economic value of caregiver
time. A replacement cost method using average wages of home aides, as was done by
Joo et al., can underestimate costs if it excludes payroll taxes and fringe benefits.
11
The opportunity cost approach, which values caregiver time based on age- and sex-specific
average earnings,
12
can yield either higher or lower estimates of costs depending on the earning potential
of the caregiver population.
Four articles address economic evaluations of interventions targeting hypertension
and CVD.
13–16
In a systematic review of CEAs of antihypertensive medications, Park and colleagues
14
found that cost-effectiveness ratios were consistently <$20,000 per quality-adjusted
life-year (QALY) saved; many were negative, meaning that their use may be cost saving
(i.e., lower total healthcare costs). Estimates of incremental cost-effectiveness
ratios across different classes of medications were also reviewed, but those estimates
are sensitive to assumptions about drug price, which vary across countries and years.
Zhang et al.
16
reviewed economic estimates of hypertension educational, self-monitoring, and screening
interventions and found consistent evidence of cost savings or low cost-effectiveness
ratios for educational interventions supporting medication adherence.
Chattopadhyay and colleagues
15
summarized estimates from Community Guide systematic reviews of specific strategies:
team-based care, reduced out-of-pocket costs, clinical decision support systems, value-based
insurance design, self-monitoring of blood pressure, and community health workers.
The independent Community Preventive Services Task Force categorizes interventions
that cost <$50,000 per QALY as cost effective. Using this threshold, the reviews found
evidence of cost effectiveness for team-based care and for self-monitoring of blood
pressure when accompanied by patient support or team-based care. It should be noted
that the U.S. DHHS does not endorse any particular cost-effectiveness threshold; it
is up to decision makers to decide whether an intervention’s return justifies the
cost of implementation given their specific objectives and resource constraints. The
Community Guide reviews noted that for some interventions there was mixed evidence
(conflicting findings) and for others there was no evidence relating to cost effectiveness.
The major contribution of the article by Chattopadhyay is its comprehensive summary
of the evidence gaps and challenges to demonstrating that public health interventions
improve health outcomes at an acceptable cost.
An original CEA by Joo et al.
13
modeled how the cost effectiveness of using intravenous recombinant tissue plasminogen
activator for treating acute ischemic stroke within 3-4.5 hours after onset of stroke
can vary by age group (18–44, 45–64, 65–80, and ≥81 years). The analysis took the
U.S. healthcare sector perspective in which only medical costs are considered. For
patients aged <65 years, the treatment was found to be cost saving; for older adults,
point estimates of the incremental cost-effectiveness ratio were <$50,000 per QALY.
These findings are consistent with previous non–age-stratified CEA estimates and also
with existing recommendations that all acute stroke patients be treated, regardless
of age. However, the cost-effectiveness conclusion is not robust for very elderly
stroke patients; the cost exceeded $50,000 per QALY in roughly half of the simulations
for the group aged >80 years, as well as exceeding $100,000 per QALY in one quarter
of simulations. In any case, these thresholds are arbitrary and are often exceeded
for widely used clinical practices.
17,18
Conducting a CEA can be very challenging because of gaps in evidence on model parameters,
and the assumptions adopted in this study, like others, can be questioned.
The scope of health economics extends beyond assessing economic costs associated with
conditions or calculating the costs avoided through prevention. Health economists
also use multiple quantitative research methods to evaluate health interventions and
policies. One article in this supplement falls in that category. Fang and colleagues
19
used survey data from 2006–2009 to 2011–2014 to assess trends in indicators of barriers
to access health care by young adults aged 19–25 years in general and by those with
hypertension. Previous studies have demonstrated reduced barriers to access by young
adults with the adoption of the dependent care provision of the Patient Protection
and Affordable Care Act.
20
This study reported results consistent with previous findings for young adults overall
and also observed similar findings for young adults with hypertension. Reducing barriers
to care is particularly important for people with chronic conditions.
20
We believe that the papers in this supplement provides researchers and policy makers
with critically important information about the costs of hypertension and CVD and
related interventions and that this information can help inform prevention efforts.