Comparative Cost-Effectiveness of Clinic, Home, or Ambulatory Blood Pressure Measurement for Hypertension Diagnosis in US Adults
- This article is commented on by the following:
- Other version(s) of this article
You are viewing the most recent version of this article. Previous versions:
Previous cost-effectiveness models found ambulatory blood pressure monitoring (ABPM) to be a favorable strategy to diagnose hypertension; however, they mostly focused on older adults with a positive clinic blood pressure (BP) screen. We evaluated the cost-effectiveness of 3 methods of BP measurement for hypertension diagnosis in primary care settings among 14 age- and sex-stratified hypothetical cohorts (adults ≥21 years of age), accounting for the possibility of both false-positive (white-coat hypertension) and false-negative (masked hypertension) clinic measurements. We compared quality-adjusted life-years and lifetime costs ($US 2017 from the US healthcare perspective) associated with clinic BP measurement, home BP monitoring, and ABPM under 2 scenarios: positive and negative initial screen. Model parameters were obtained from published literature, publicly available data sources, and expert input. In the screen-positive scenario, ABPM was the dominant strategy among all age and sex groups. Compared with clinic BP measurement, ABPM was associated with cost-savings ranging from $77 (women 80 years of age) to $5013 (women 21 years of age). In the screen-negative scenario, ABPM was the dominant strategy in all men and women <80 years of age with cost-savings ranging from $128 (women 70 years of age) to $2794 (women 21 years of age). Sensitivity analyses showed that results were sensitive to test specificity and antihypertensive medication costs. ABPM is recommended as the diagnostic strategy of choice for most adults in primary care settings regardless of initial screening results.
See Editorial Commentary, pp 45–46
Three major strategies may be used to detect high blood pressure (BP) in primary care settings: clinic BP measurements (CBPM), eg, measurements taken in the clinic at biweekly intervals during 4 to 6 weeks; home BP monitoring (HBPM), eg, multiple measurements taken by the patient during a week; and ambulatory BP monitoring (ABPM), eg, measurements taken automatically during 24 hours. Among these strategies, CBPM is routinely used in practice; however, the limited diagnostic accuracy of this method may cause some individuals to be treated despite having normal true or usual BP (clinic false-positives [FPs] or white-coat hypertension) and some individuals to be left untreated despite an underlying problem of hypertension (clinic false-negatives [FNs] or masked hypertension).1 HBPM and ABPM are associated with better diagnostic accuracy, but these 2 modalities have higher cost or burden of implementation.2
In their comprehensive literature review on studies that examined the cost-effectiveness of different diagnostic modalities for hypertension, Wang et al1 found only 1 cost-effectiveness study that compared all 3 strategies. Most published cost-effectiveness models (CEMs) started their CEM only with individuals who had a positive office-based BP screening.1 This approach allows the detection of FP office-based BP measurements (ie, white-coat hypertension) but fails to model the FN cases (ie, masked hypertension) by excluding them from the analysis. Additionally, although the US Preventive Services Task Force endorses a recommendation to start BP screening for adults aged ≥18 years,3 most of existing studies limited their population to individuals aged ≥40 years. In this study, we aim to identify the incremental cost-effectiveness of HBPM versus ABPM versus CBPM by modeling lifelong costs and clinical outcomes of adults ≥21 years, accounting for both FN and FP clinic BP screen.
All data and supporting materials are available within the article except for the National Health and Nutrition Examination Survey data that are publicly available at https://wwwn.cdc.gov/nchs/nhanes/Default.aspx.
We developed a decision tree-Markov model hybrid design to compare costs and quality-adjusted life-years (QALYs) associated with 3 diagnostic strategies after a first clinic BP screening. The study population consisted of a hypothetical cohort of primary care patients aged ≥21 years with a cardiovascular risk factor profile equivalent to the US general population of the same age and sex. We ran the model separately for 14 age- and sex-stratified groups, including men and women 21, 30, 40, 50, 60, 70, and 80 years of age under 2 scenarios: (1) having a positive initial clinic-based screen and (2) having a negative initial clinic-based screen. Because limited data were available to estimate model parameters beyond the age of 85 years, the follow-up period ranged from 5 (for age group of 80 years) to 64 years (for age group of 21 years).
Members of each hypothetical cohort enter the model by having 1 initial clinic BP screening. Screen-negative individuals receive 1 of the 3 diagnostic strategies: (1) no confirmatory test (ie, do nothing) and rescreen (a CBPM) in the next period (ie, every 3 years for individuals between 18 and 40 years of age and every year for ≥40 years of age3), (2) HBPM, or (3) ABPM (Figure 1A). Individuals with a positive screen receive 1 of the 3 diagnostic strategies: (1) CBPM (2 confirmatory office-based measurements), (2) HBPM, or (3) ABPM (Figure 1B). Positive test results, including true-positive (TP) or FP, lead to antihypertensive treatment, whereas individuals with true-negative (TN) or FN results receive no antihypertensive treatment (See Table 1 for the definition of different hypertension phenotypes).
|Clinic-Based BP Measurement||Out-of-Office BP Measurement*||Phenotype|
|Elevated||Not elevated||FP (ie, white-coat hypertension)|
|Not elevated||Elevated||FN (ie, masked hypertension)|
|Not elevated||Not elevated||TN|
The members of each hypothetical cohort transitioned through the model’s different diagnostic (ie, TP, FP, TN, and FN) and health states (ie, stable angina [SA], unstable angina [UA], myocardial infarction [MI], stroke, and transient ischemic attack [TIA]) with age- and sex-specific transition probabilities and stayed in that health state for at least 1 cycle of the model (1 year) before transitioning to other states (Figure 1C). The model takes the perspective of a US payer, and as such, no indirect costs (eg, productivity losses or transportation costs) were taken into account. Future costs and QALYs gained were discounted at 3%.4 The parameters used in the model are presented in Table 2.
|Prevalence of hypertension||4.15%–83.54%*||Calculated using combined 2011–2012 and 2013–2014 NHANES data5,6|
|Sensitivity||CBPM, 74.6% (95% CI, 60.7%–84.8%); HBPM, 85.7% (95% CI, 78.0%–91.0%); ABPM, 100.0%||Meta-analysis sensitivity analysis by Hodgkinson et al2; ABPM assumed to be the gold standard7|
|Specificity||CBPM, 74.6% (95% CI, 47.9%–90.4%); HBPM, 62.4% (95% CI, 48.0%–75.0%); ABPM, 100.0%||As above|
|Relative risk of hypertension treatment|
|Relative risk of CHD event on hypertension treatment||0.683 (95% CI, 0.633–0.717)||Calculated based on meta-analysis by Law et al8|
|Relative risk of a cerebrovascular event on hypertension treatment||0.633 (95% CI, 0.526–0.717)||As above|
|Transition probabilities to death|
|Probability of death eliminating major cardiovascular and cerebrovascular diseases||0.07%–12.6%*||Calculated based on US Life Tables Eliminating Certain Causes of Death, 1999–20019|
|SMR after myocardial infarction||2.68 (95% CI, 2.48–2.91)||Lovibond et al7|
|SMR after unstable angina||2.19 (95% CI, 2.05–2.33)||As above|
|SMR after stable angina||1.95 (95% CI, 1.65–2.31)||As above|
|SMR after stroke||2.27 (95% CI, 2.59–2.85)||As above|
|SMR after transient ischemic attack||1.4 (95% CI, 1.1–1.8)||As above|
|Transition probabilities to the health states|
|Probability of CHD event if normotensive||0.00%–0.26%*||Calculated using risk factor profile from combined 2011–2012 and 2013–2014 NHANES data5,6 and Framingham CHD and stroke risk equations10|
|Probability of CHD event if hypertensive||0.00%–0.48%*||As above|
|Probability of stroke event if normotensive||0.00%–0.13%*||As above|
|Probability of stroke event if hypertensive||0.00%–0.25%*||As above|
|CHD event distribution||Myocardial infarction, 14.3%–37.8%; unstable angina, 10.4%–20.9%; stable angina, 37.7%–62.9%; CHD death, 6.6%–17.8%*||Lovibond et al7|
|Stroke event distribution||Stroke, 51.7%–70.1%; transient ischemic attack, 13.4%–36.1%; stroke death, 12.2%–16.5%*||As above|
|Quality-of-life weights (utilities)|
|Baseline utilities for healthy population (ie, no cardiovascular event)||0.79 (SE, 0.005)–0.90 (SE, 0.003)*||Lubetkin et al11|
|Myocardial infarction||0.76 (SE, 0.018)||Lovibond et al,7 Ward et al12|
|Unstable angina||0.77 (SE, 0.038)||As above|
|Stable angina||0.81 (SE, 0.038)||As above|
|Stroke||0.63 (SE, 0.04)||As above|
|Transient ischemic attack||1||Assumption based on Lovibond et al7|
|On hypertension treatment||0.994||Hutchins et al13|
|Cost of diagnosis by method (2017 USD)|
|Cost of 1 clinic-based BP measurement||$52||Calculated based on consultation with clinical experts and CMS physician fee schedule, code 99213 (national facility price)14|
|Cost of diagnosis CBPM||$155||As above|
|Cost of diagnosis HBPM||$105||Calculated based on the approach by Lovibond et al,7 consultations with clinical experts, and CMS physician fee schedule (code 99213).14 For this calculation, we assumed a 5-y monitor lifetime and 40 uses per monitor per year.7|
|Cost of diagnosis ABPM||$158||Calculated based on consultation with clinical expert and CMS physician fee schedule, codes 93784 and 99213 (national facility price)14|
|Cost of health states (event costs)|
|Annual hypertension treatment cost||$444||Moran et al15|
|Costs of myocardial infarction||$21 439–$24 523*||Calculated based on HCUP,16 Lovibond et al,7 and Ohsfeldt et al17|
|Cost of stable angina||$2144–$2452*||As above|
|Cost of unstable angina||$12 863–$14 714*||As above|
|Cost of stroke||$15 749–$19 083*||Calculated based on HCUP,16 Ohsfeldt et al,17 and Luengo-Fernandez et al18|
|Cost of transient ischemic attack||$1575–$1908*||As above|
|Cost of health states (maintenance costs)|
|Costs of myocardial infarction||$4490||Calculated based on the cost of MI (above) and Lee et al19|
|Cost of unstable angina||$2694||As above|
|Cost of stable angina||$449||As above|
|Cost of stroke||$5401||Calculated based on the cost of stroke (above) and Moran et al15|
|Cost of transient ischemic attack||$540||As above|
Diagnostic States Probabilities
We assumed that ABPM has 100% sensitivity and 100% specificity; however, to account for ABPM not always working correctly, a 5% failure rate has been considered for which ABPM requires repeat testing.7 The sensitivity and specificity of HBPM and CBPM were extracted from published literature.2
In our model, we accounted for people who were not hypertensive on entering the model and became hypertensive over time via rechecking their BP using the same test at 3-year intervals before 40 years of age and every year for the age of ≥40 years.3,7 The transition probabilities of the diagnostic states were calculated based on the sensitivity and specificity of the diagnostic strategy, age- and sex-specific hypertension prevalence, and the probability of becoming hypertensive for normotensive individuals.
Disease States Probabilities
For this study, hypertension is defined as systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg, taking antihypertensive medication or being told at least twice by a health professional that one has hypertension.20 Based on this definition, we used combined 2011 to 2012 and 2013 to 2014 National Health and Nutrition Examination Survey risk factor data to estimate the prevalence of cardiovascular risk factors (hypertension, diabetes mellitus, smoking, and cholesterol level).5,6 Incorporating the risk factor prevalence data in the relevant Framingham risk equation,10 the age- and sex-specific probability of coronary heart disease (CHD) and cerebrovascular disease (ie, stroke and transient ischemic attack) events were estimated. The probability of each health state was then calculated using the age- and sex-specific CHD and cerebrovascular disease event distributions.7,12
To estimate the corresponding probabilities for the TP group, 2 separate relative risk estimates were used for CHD events (SA, UA, and MI) and cerebrovascular diseases (stroke and TIA), assuming that antihypertensive treatment affects the probability of every disease state similarly across all age and sex groups. Relative risk reductions attributable to antihypertensive treatment were extracted from the peer-reviewed literature.8 We also assumed that hypertension treatment would not affect the probability of future cardiovascular events for individuals with white-coat hypertension (FP), and the treatment of patients with masked hypertension (FN) would reduce the probability of future cardiovascular events.7
We estimated the probability of death separately for (1) all causes except for the included disease states and (2) death attributable to the included disease states. The first component was estimated using US Life Tables Eliminating Certain Causes of Death,9 and the second component was calculated based on standardized mortality ratios extracted from the literature.7
We extracted the baseline age- and sex-stratified utilities based on the EQ-5D index from a nationally representative sample of the US noninstitutionalized civilian population11 (where 0 equals death, and 1 equals full health). To calculate each health state’s utility, quality-of-life multipliers for different health states were applied multiplicatively to the age and sex-stratified baseline utilities.7,12 The utility decrement of 0.006 was applied to the members of the hypothetical cohort who receive antihypertensive treatment (ie, TP and FP cases).13
The cost analyses for this study reflect the US healthcare payer perspective; thus, the analyses assess direct medical costs only. We defined event-related costs as all medical care costs for CHD or cerebrovascular diseases that happen within the cycle when the event occurs, whereas maintenance costs included the cost of medical care related to CHD or cerebrovascular diseases provided in the subsequent years when no event occurs.21 Average costs for hospitalizations with the primary diagnoses codes were retrieved from HCUPnet.16 Peer-reviewed literature was used to estimate the average event-related and maintenance cost of the disease states.22 All cost items were adjusted to 2017 USD.23 The costs of different diagnostic strategies of the model were calculated using the relevant Centers for Medicare and Medicaid Services physician fee schedule procedure codes14 and consultations with clinicians. Annual antihypertensive treatment cost was extracted from the published literature.15
To account for the impact of uncertainty, probabilistic analyses were performed for 1000 replications, separately for each model. The inputs that were tested include relative risk parameters (log-normal distribution), utility estimations (β-distribution), hypertension prevalence (β-distribution), specificity and sensitivity of HBPM and CBPM (β-distribution), and hypertension diagnosis, treatment, and disease states costs (γ-distribution).24 The CIs around incremental costs and incremental QALYs for ABPM and HBPM versus CBPM, as well as incremental cost-effectiveness planes, summarize the probabilistic analysis results. The models were constructed in Microsoft Excel 2016, and the probabilistic analyses were conducted using Crystal Ball (Oracle Corp, Pacific Shores, CA). Deterministic sensitivity analyses accounted for parameter uncertainty using a range of relevant values of sensitivity and specificity of CBPM, HBPM and ABPM, ABPM failure rate, hypertension prevalence, diagnostic costs, annual antihypertensive treatment cost, and utilities. Relevant parameter ranges were extracted from published literature.
The base-case analysis results for all age and sex groups are shown in Table 3 (for screen-positive models) and Table 4 (for screen-negative models). Under the screen-positive scenario, ABPM was the dominant strategy for hypertension diagnosis among all age and sex groups. Compared with CBPM, ABPM was associated with cost-savings ranging from $77 (women 80 years of age) to $5013 (women 21 years of age) and $147 (men 80 years of age) to $4671 (men 21 years of age) in women and men, respectively. Under the screen-negative scenario, ABPM was the dominant strategy in all adult men and women <80 years of age, for whom CBPM was estimated to be the most cost-effective diagnostic strategy. Compared with CBPM, ABPM was associated with cost-savings ranging from $128 (women 70 years of age) to $2794 (women 21 years of age) and $213 (men 70 years of age) to $2593 (men 21 years of age) for women and men, respectively.
|Sex and Age Group||Incremental Cost vs CBPM (95% CI)||Incremental QADs vs CBPM (95% CI)||Most CE Strategy*||Probability CE,* %||Probability CS, %|
|21 y||$773 ($604 to $936)||−$5013 (−$6045 to −$3875)||−4.1 (−4.1 to −4.0)||28.6 (27.6 to 29.3)||ABPM||100.0||100.0|
|30 y||$531 ($390 to $677)||−$3843 (−$4874 to −$2673)||−2.9 (−3.0 to −2.9)||23.3 (22.6 to 24.1)||ABPM||100.0||100.0|
|40 y||$213 ($74 to $337)||−$3028 (−$4058 to −$2068)||−1.4 (−1.5 to −1.4)||19.3 (18.1 to 20.4)||ABPM||100.0||100.0|
|50 y||$185 ($68 to $304)||−$2209 (−$2904 to −$1440)||−1.1 (−1.3 to −1.0)||13.9 (12.8 to 15.1)||ABPM||100.0||100.0|
|60 y||$134 ($44 to $228)||−$1083 (−$1459 to −$697)||−0.6 (−0.8 to −0.5)||6.7 (5.9 to 7.6)||ABPM||100.0||100.0|
|70 y||$111 ($35 to $196)||−$542 (−$778 to −$320)||−0.4 (−0.6 to −0.3)||3.4 (2.8 to 4.0)||ABPM||100.0||100.0|
|80 y||$77 ($7 to $152)||−$77 (−$147 to −$9)||−0.3 (−0.4 to −0.2)||0.3 (0.1 to 0.5)||ABPM||99.9||98.4|
|21 y||$744 ($575 to $915)||−$4671 (−$5707 to −$3548)||−3.9 (−4.0 to −3.8)||26.4 (25.6 to 27.2)||ABPM||100.0||100.0|
|30 y||$518 ($374 to $672)||−$3614 (−$4530 to −$2572)||−2.8 (−2.9 to −2.6)||21.8 (20.4 to 23.0)||ABPM||100.0||100.0|
|40 y||$207 ($80 to $330)||−$2859 (−$3762 to −$1923)||−1.3 (−1.4 to −1.2)||18.1 (16.7 to 19.5)||ABPM||100.0||100.0|
|50 y||$174 ($70 to $285)||−$1994 (−$2611 to −$1295)||−0.9 (−1.1 to −0.7)||12.7 (11.8 to 13.6)||ABPM||100.0||100.0|
|60 y||$122 ($22 to $213)||−$986 (−$1318 to −$640)||−0.3 (−0.6 to 0.0)||6.6 (5.7 to 7.4)||ABPM||100.0||100.0|
|70 y||$102 ($20 to $188)||−$563 (−$784 to −$365)||0.0 (−0.4 to 0.3)||4.3 (3.5 to 5.0)||ABPM||100.0||100.0|
|80 y||$75 (-$1 to $149)||−$147 (−$228 to −$67)||−0.2 (−0.3 to 0.0)||1.1 (0.7 to 1.3)||ABPM||100.0||99.9|
|Sex and Age Group||Incremental Cost vs CBPM (95% CI)||Incremental QADs vs CBPM (95% CI)||Most CE Strategy*||Probability CE,* %||Probability CS, %|
|21 y||$3180 ($2740 to $3653)||−$2794 (−$3723 to −$1733)||−15.2 (−15.3 to −15.1)||18.5 (18.0 to 18.9)||ABPM||100.0||100.0|
|30 y||$2428 ($2034 to $2799)||−$2345 (−$3334 to −$1381)||−11.8 (−11.9 to −11.7)||16.8 (16.5 to 17.2)||ABPM||100.0||100.0|
|40 y||$1493 ($1159 to $1813)||−$2412 (−$3404 to −$1282)||−7.6 (−7.8 to −7.4)||17.5 (16.9 to 18.1)||ABPM||100.0||100.0|
|50 y||$1368 ($1050 to $1639)||−$1735 (−$2560 to −$801)||−6.3 (−6.9 to −5.9)||13.3 (12.4 to 14.0)||ABPM||100.0||99.9|
|60 y||$1114 ($843 to $1361)||−$750 (−$1276 to −$181)||−4.0 (−5.1 to −3.2)||7.4 (6.1 to 8.2)||ABPM||100.0||99.8|
|70 y||$943 ($713 to $1141)||−$128 (−$451 to $204)||−2.9 (−4.0 to −2.0)||3.5 (2.3 to 4.4)||ABPM||99.4||77.2|
|80 y||$665 ($498 to $812)||$405 ($261 to $539)||−2.3 (−3.0 to −2.0)||−1.0 (−1.8 to −0.6)||CBPM||0.0||0.0|
|21 y||$3133 ($2712 to $3555)||−$2593 (−$3574 to −$1602)||−14.8 (−15.0 to −14.6)||17.2 (16.8 to 17.6)||ABPM||100.0||100.0|
|30 y||$2397 ($2030 to $2792)||−$2183 (−$3116 to −$1202)||−11.3 (−11.6 to −11.0)||15.9 (15.3 to 16.5)||ABPM||100.0||100.0|
|40 y||$1458 ($1101 to $1772)||−$2274 (−$3287 to −$1156)||−6.9 (−7.4 to −6.5)||16.8 (16.0 to 17.5)||ABPM||100.0||100.0|
|50 y||$1299 ($985 to $1585)||−$1571 (−$2357 to −$783)||−4.9 (−6.2 to −3.9)||13.0 (12.0 to 14.0)||ABPM||100.0||100.0|
|60 y||$1023 ($778 to $1263)||−$689 (−$1156 to −$185)||−1.3 (−3.7 to 0.9)||9.2 (6.8 to 11.3)||ABPM||100.0||99.7|
|70 y||$853 ($618 to $1094)||−$213 (−$572 to $114)||0.6 (−2.6 to 3.0)||7.2 (3.8 to 9.7)||ABPM||100.0||89.2|
|80 y||$602 ($436 to $763)||$246 ($83 to $411)||−1.0 (−2.2 to 0.0)||1.1 (−0.3 to 2.2)||CBPM||23.3||0.0|
Compared with CBPM, HBPM resulted in a loss in quality-adjusted days (QADs) in all age and sex groups, except for men 70 years of age, ranging from 0.2 QADs (men 80 years of age) to 4.1 QADs (women 21 years of age) in the screen-positive models and 1 QAD (men 80 years of age) to 15.2 QADs (women 21 years of age) in the screen-negative models. Compared with CBPM, HBPM was associated with higher costs across all age and sex groups under both scenarios, ranging from $75 (men 80 years of age) to $775 (women 21 years of age) in the screen-positive models and from $602 (men 80 years of age) to $3180 (women 21 years of age) in the screen-negative models. Compared with CBPM, men and women 21 years of age gained the maximum increase in QADs from receiving ABPM both in screen-positive (26.4 and 28.6, respectively) and screen-negative models (17.2 and 18.5, respectively).
The probabilistic analyses suggested low uncertainty within our base-case analysis (Tables 3 and 4). Resulting from the 1000 replications of the corresponding models, the last 2 columns in Tables 3 and 4 present estimated probabilities that ABPM has the highest net benefit (NB) at the ceiling ratio of $50 000 per QALY or is cost-saving, respectively. Consistent with the base-case results, ABPM showed the highest NB in 100% of model replications in the screen-positive models. In these models, compared with CBPM, ABPM was cost-saving the majority of the time (Table 3). In the screen-negative scenario, at the ceiling ratio of $50 000, ABPM was associated with the highest NB, 99% to 100%, of the time in all age and sex groups except for men and women 80 years of age (Table 4). Compared with CBPM, ABPM was cost-saving in the majority of the time in men and women <80 years of age.
Given the large number of age and sex groups, we only present the cost-effectiveness planes (Figure 2) for men and women aged 40 years under both screen-positive and screen-negative scenarios. As presented in Figure 2, all simulated incremental cost-utility ratios for ABPM versus CBPM fell in the southeast quadrant of the cost-effectiveness plane (ie, more effective and less costly). The corresponding simulated incremental cost-utility ratios for HBPM versus CBPM fell in the northwest quadrant of the cost-effectiveness plane (ie, less effective and more costly). For each competing diagnostic strategy, the proportion of replications where that diagnostic strategy provides the greatest NB for a given willingness-to-pay threshold was calculated. For women and men aged 40 years under both screen-positive and screen-negative scenarios, ABPM provides the largest NB, 100%, of the time for all included thresholds ($0–$100 000).
Deterministic Sensitivity Analyses
Table 5 shows the results of deterministic sensitivity analyses for screen-negative women aged 40 years. ABPM remained the most cost-effective strategy in almost all analyses except for increasing the specificity of CBPM or HBPM or reducing ABPM specificity. For instance, setting HBPM specificity to 100% makes HBPM the most cost-effective diagnostic strategy, and setting the specificity of all diagnostic strategies to 100% would lead to CBPM being the most cost-effective strategy. Additionally, setting the utility decrement associated with taking antihypertensive medication to zero resulted in HBPM having the highest NB among all other options, despite ABPM remaining vastly cost-saving.
|Sensitivity Analyses (Deterministic)||Incremental costs|
(Other strategy vs CBPM)
|Incremental QAD Gained (Other Strategy vs CBPM)||Most CE Strategy*|
|Base case (deterministic)||−$2412||$1493||17.5||−7.6||ABPM|
|Base case (probabilistic)||−$2417||$1495||17.5||−7.6||ABPM|
|Sensitivity and specificity|
|Sensitivity and specificity ABPM=HBPM||$1535||$1493||−8.4||−7.6||CBPM|
|Sensitivity, 100% for all options||−$2447||$1471||17.6||−7.5||ABPM|
|Specificity, 100% for all options||$1064||$261||−0.6||−0.5||CBPM|
|HBPM specificity set to 70%||−$2412||$1326||17.5||−6.7||ABPM|
|HBPM specificity set to 80%||−$2412||$957||17.5||−4.7||ABPM|
|HBPM specificity set to 90%||−$2412||$118||17.5||−0.2||ABPM|
|HBPM specificity set to 100%||−$2412||−$3215||17.5||17.5||HBPM|
|HBPM sensitivity set to 90%||−$2412||$1498||17.5||−7.6||ABPM|
|HBPM sensitivity set to 95%||−$2412||$1502||17.5||−7.6||ABPM|
|HBPM sensitivity set to 100%||−$2412||$1506||17.5||−7.6||ABPM|
|ABPM sensitivity set to 95%||−$2285||$1493||16.4||−7.6||ABPM|
|ABPM sensitivity set to 90%||−$2166||$1493||15.4||−7.6||ABPM|
|ABPM sensitivity set to 85%||−$2057||$1493||14.5||−7.6||ABPM|
|ABPM specificity set to 95%||−$444||$1493||5.1||−7.6||ABPM|
|ABPM specificity set to 90%||$384||$1493||−0.3||−7.6||CBPM|
|ABPM specificity set to 85%||$821||$1493||−3.1||−7.6||CBPM|
|ABPM failure rate|
|Failure rate, 0%||−$2526||$1493||17.5||−7.6||ABPM|
|Failure rate, 10%||−$2299||$1493||17.5||−7.6||ABPM|
|Prevalence set to 10%||−$2804||$1603||20.1||−8.3||ABPM|
|Prevalence set to 20%||−$2544||$1530||18.4||−7.8||ABPM|
|Prevalence set to 30%||−$2262||$1452||16.5||−7.3||ABPM|
|Prevalence set to 40%||−$1957||$1366||14.4||−6.8||ABPM|
|ABPM cost increased by 20%||−$1937||$1493||17.5||−7.6||ABPM|
|CBPM cost decreased by 20%||−$2358||$1548||17.5||−7.6||ABPM|
|HBPM cost decreased by 20%||−$2412||$1456||17.5||−7.6||ABPM|
|Cost of hypertension treatment||ABPM|
|Cost of hypertension treatment increased by 20%||−$3260||$1867||17.5||−7.6||ABPM|
|Cost of hypertension treatment decreased by 20%||−$1565||$1120||17.5||−7.6||ABPM|
|Utility decrement associated with hypertension treatment, 0.01||−$2412||$1493||29.1||−12.8||ABPM|
|Utility decrement associated with hypertension treatment, 0||−$2412||$1493||0.0||0.3||HBPM|
Sensitivity to different input parameters varies among CEMs for different age/sex groups and positive/negative initial BP screen scenarios. In women with an initial negative screen, the results of the CEMs were highly sensitive to the annual antihypertensive medication cost and CBPM specificity for age groups ≤50 years. For example, setting CBPM specificity to 85% will result in CBPM becoming more cost-effective than ABPM in women >53, 62, 67, and 68 years of age, if the annual antihypertensive medication cost is set to $100, $150, $300, and $445 (the base-case value), respectively. For older age groups, hypertension prevalence became more important in determining the most cost-effective strategy than the annual antihypertensive medication cost and CBPM specificity. The same pattern was observed for men with an initial negative screen, with the exception of age groups ≥60 years, for whom the CEM results were the most sensitive to the relative risk reduction associated with antihypertensive treatment for CHD and cerebrovascular events. For women and men with an initial positive screen, in addition to the annual antihypertensive medication costs, the CEM results became increasingly sensitive to hypertension prevalence as the starting hypothetical cohort gets older.
Given the increasing prevalence of hypertension and high associated disease burden, correct and timely diagnosis and treatment initiation for hypertension can improve population health and possibly save money by preventing future costly cardiovascular events. Currently, CBPM is widely used to diagnose hypertension and initiate treatment. However, the natural variability of BP and the phenomena of white-coat hypertension and masked hypertension challenge its diagnostic accuracy. ABPM and HBPM (also known as self-monitoring) are out-of-office BP measurement techniques that can improve the estimation of true BP.25 Use of these methods is also influenced by patient and provider acceptability. Moreover, because of the chronic nature of hypertension, patients often receive antihypertensive therapy for the rest of their lives, which further indicates the significance of accurate diagnosis and appropriate treatment initiation.2,3 To date, ABPM has not been widely used in clinical practice in the United States, partly because the costs of this diagnostic strategy were often not reimbursable by healthcare payers.26 Based on the geographic region, Medicare pays between $56 and $122 per 24-hour ABPM session only for suspected white-coat hypertension.27
Findings of our study suggest that ABPM is the most cost-effective strategy (at a willingness to pay $50 000 per QALY) and even has a high probability of being cost-saving for most adults regardless of their initial clinic BP screen results. ABPM remained the most cost-effective strategy across a wide range of varying assumptions in the deterministic sensitivity analysis. These results support the recent US Preventive Services Task Force recommendation to use ABPM as the reference standard to avoid misdiagnosis and overtreatment of hypertension.27,28 The results of the deterministic sensitivity analysis suggest that, especially in younger adults, better specificity along with high antihypertensive medication cost are 2 main factors that determine cost-effectiveness and cost-saving potential of strategies for hypertension diagnosis. This finding indicates that the ability to correctly diagnose white-coat hypertension (FP) makes ABPM the preferred strategy for hypertension screening especially in younger adults with lower hypertension prevalence. As the prevalence of hypertension increases with age, the CBPM positive predictive value increases (ie, the proportion of TP to FP increases); therefore, white-coat hypertension will not be as big of a problem. Compared with CBPM, HBPM was associated with higher cost and lower QALYs for initial hypertension diagnosis. The value of HBPM in settings other than the initial diagnosis of hypertension is beyond the scope of this article; however, numerous studies have shown the value of HBPM in self-monitoring and hypertension control, as well as antihypertensive drug adherence especially when paired with communication strategies, such as telemonitoring.29,30
Previous research conducted by Lovibond et al7 also concluded that using ABPM after an initial positive screen in the clinic would reduce misdiagnosis and save costs. Aligned with the results of our study, Lovibond et al found that ABPM had greater cost-saving in younger age groups. They also recommended the use of ABPM before the initiation of antihypertensive drugs for most patients. However, Lovibond et al started their model with a hypothetical cohort of patients that had an initial positive clinic-based BP screen. This assumption is restrictive in that it effectively hinders the model from taking cases of masked hypertension into account. Therefore, we developed our models under 2 scenarios of having a positive and having a negative initial clinic BP screen. Our model also expands on the previous work by including younger age groups (21–40 years). Our findings suggest that when taking into account the added testing (ie, rescreening) of people with an initial negative clinic BP screen, ABPM remains cost-effective and possibly cost-saving for adults <80 years of age. These savings might also be because of the prevention of cardiovascular events achieved by treating FNs (ie, masked hypertension). When comparing the results of this article to Lovibond et al CEM, the relative magnitude of main cost parameters needs to be considered. For example, in the base case, estimate by Lovibond et al of ABPM cost is almost 40% higher than CBPM cost, whereas in our model, ABPM is only slightly more costly. Moreover, in the model by Lovibond et al, the base-case annual antihypertensive treatment cost is slightly higher than the cost of ABPM, whereas in our model, the baseline cost of annual antihypertensive medication was estimated to be almost 3× the ABPM cost. These differences in the relative magnitude of cost parameters lead to an improved cost profile for ABPM diagnostic strategy relative to CBPM strategy.
We made several assumptions concerning the model structure and input parameters. Below are some of the central assumptions that can be a source of potential limitation. In this study, Framingham risk equations were used to estimate the age- and sex-specific probability of included cardiovascular and cerebrovascular events separately for hypertensive and normotensive individuals. Although we acknowledge the potential limitations of the Framingham risk formulas, for example, equations being calculated based on decades-old data, we used these equations because they are considered to be the standard of practice and have been commonly used in similar CEMs.7
We also made several assumptions about antihypertensive treatment effect, including (1) people who were incorrectly labeled as hypertensive (FP or white-coat hypertension) receive no treatment benefit from antihypertensive therapy. This assumption was made because the effect of antihypertensive therapy in this group is not well established, and previous models adopted the same assumption. (2) Our model does not account for potential side effects of antihypertensive therapy mainly because of the scarcity of appropriate data. Also, because accounting for the side effects would require making more assumptions and would increase the model complexity without decreasing the uncertainty surrounding the model estimates. Besides, including side effects in the model would benefit ABPM cost-effectiveness profile because of ABPM’s perfect specificity.7 (3) The model assumes that patients with masked hypertension receive the same benefit from antihypertensive treatment as regular hypertensive patients. Although being used in previous models, future studies including clinical trials of identifying and treating masked hypertension are needed to test this assumption. Another main assumption that was made is the homogeneity of antihypertensive treatment effect across all age, sex, and risk factor strata. (4) We also assume that antihypertensive treatment reduces the probability of all included cardiovascular events similarly. This assumption was made because we were unable to identify appropriate data to quantify heterogeneity of treatment effect and also because of inherent challenges of implementing differential treatment effect in the context of cohort simulation models.
Similar to the model by Lovibond et al,7 we considered ABPM as the reference standard for diagnosing hypertension with 100% sensitivity and specificity. We acknowledge that assuming perfect performance for ABPM might be a strong assumption; therefore, we allowed 5% failure rate for ABPM to address the practical issues that might happen. Multiple scenarios of lower diagnostic accuracy for ABPM have been considered in the deterministic sensitivity analysis.
In this study, we did not include use of automated office BP (AOBP) measurement (eg, by devices such as the Omron HEM907XL or the BpTru) for the initial BP screen in the clinic. AOBP measurement is a method for diagnosing hypertension in physician offices, which entails measurement of BP using a fully automated sphygmomanometer that takes multiple readings with the patient resting quietly alone.31 However, based on the results of our sensitivity analysis, we would expect ABPM to be more cost-effective than AOBP even under conservative assumptions, such as AOBP specificity of 87%32 and AOBP costs being equal to CBPM. These assumptions are conservative given that the costs of these devices are higher than a typical oscillometric device, and the total cost (eg, added time) of performing AOBP would be higher, in addition, 87% is on the higher end of AOBP specificity estimates.
Overall, the findings of this study suggest that using ABPM is the strategy of choice for hypertension diagnosis and treatment initiation for most adults in primary care settings. We predict that correctly diagnosing white-coat and masked hypertension by ABPM reduces the overall cost of treatment for hypertension and future cardiovascular disease events.
The findings of this study suggest that using ABPM is the strategy of choice for hypertension diagnosis and treatment initiation for most adults in primary care settings in the United States. We predict that correctly diagnosing white-coat and masked hypertension by ABPM reduces the overall cost of treatment for hypertension and future cardiovascular and cerebrovascular disease events. Our findings have policy significance in the sense that expanding the reimbursement of ABPM as a diagnostic strategy for hypertension in primary care settings may reduce healthcare costs and improve health outcomes by getting hypertension treatment to the correct patients.
We thank Sally C. Stearns, PhD, for her expert advice on designing and implementing the study models and Andrea K. Biddle, PhD, for her consultation and input in the planning of this study.
Sources of Funding
A.J. Viera has received funding from the National Heart, Lung, and Blood Institute to study ambulatory blood pressure monitoring and home blood pressure monitoring. The other authors report no conflicts.
Wang YC, Koval AM, Nakamura M, Newman JD, Schwartz JE, Stone PW. Cost-effectiveness of secondary screening modalities for hypertension.Blood Press Monit. 2013; 18:1–7. doi: 10.1097/MBP.0b013e32835d0fd3CrossrefMedlineGoogle Scholar
Hodgkinson J, Mant J, Martin U, Guo B, Hobbs FD, Deeks JJ, Heneghan C, Roberts N, McManus RJ. Relative effectiveness of clinic and home blood pressure monitoring compared with ambulatory blood pressure monitoring in diagnosis of hypertension: systematic review.BMJ. 2011; 342:d3621.CrossrefMedlineGoogle Scholar
Sheridan S, Pignone M, Donahue K. Screening for high blood pressure: a review of the evidence for the U.S. Preventive Services Task Force.Am J Prev Med. 2003; 25:151–158.CrossrefMedlineGoogle Scholar
Sanders GD, Neumann PJ, Basu A,. Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: second panel on cost-effectiveness in health and medicine.JAMA. 2016; 316:1093–1103. doi: 10.1001/jama.2016.12195CrossrefMedlineGoogle Scholar
- 5. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 2011; 2012. https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2011. Accessed November 15, 2015.Google Scholar
- 6. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 2013; 2014. https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2013. Accessed December 15, 2016.Google Scholar
Lovibond K, Jowett S, Barton P, Caulfield M, Heneghan C, Hobbs FD, Hodgkinson J, Mant J, Martin U, Williams B, Wonderling D, McManus RJ. Cost-effectiveness of options for the diagnosis of high blood pressure in primary care: a modelling study.Lancet. 2011; 378:1219–1230. doi: 10.1016/S0140-6736(11)61184-7CrossrefMedlineGoogle Scholar
Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs in the prevention of cardiovascular disease: meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies.BMJ. 2009; 338:b1665.CrossrefMedlineGoogle Scholar
Arias E, Heron M, Tefada-Vera M. United States life tables eliminating certain causes of death, 1999–2001.Natl Vital Stat Rep. 2013; 61:1–128.MedlineGoogle Scholar
Anderson KM, Odell PM, Wilson PW, Kannel WB. Cardiovascular disease risk profiles.Am Heart J. 1991; 121(1 pt 2):293–298.CrossrefMedlineGoogle Scholar
Lubetkin EI, Jia H, Franks P, Gold MR. Relationship among sociodemographic factors, clinical conditions, and health-related quality of life: examining the EQ-5D in the U.S. general population.Qual Life Res. 2005; 14:2187–2196. doi: 10.1007/s11136-005-8028-5CrossrefMedlineGoogle Scholar
Ward S, Lloyd Jones M, Pandor A, Holmes M, Ara R, Ryan A, Yeo W, Payne N. A systematic review and economic evaluation of statins for the prevention of coronary events.Health Technol Assess. 2007; 11:1–160.CrossrefGoogle Scholar
Hutchins R, Pignone MP, Sheridan SL, Viera AJ. Quantifying the utility of taking pills for preventing adverse health outcomes: a cross-sectional survey.BMJ Open. 2015; 5:e006505. doi: 10.1136/bmjopen-2014-006505CrossrefMedlineGoogle Scholar
- 14. Centers for Medicare and Medicaid Service. Physician Fee Schedule.https://www.cms.gov/apps/physician-fee-schedule/search/search-criteria.aspx. Accessed April 15, 2018.Google Scholar
Moran AE, Odden MC, Thanataveerat A, Tzong KY, Rasmussen PW, Guzman D, Williams L, Bibbins-Domingo K, Coxson PG, Goldman L. Cost-effectiveness of hypertension therapy according to 2014 guidelines.N Engl J Med. 2015; 372:447–455. doi: 10.1056/NEJMsa1406751CrossrefMedlineGoogle Scholar
- 16. Agency for Healthcare Research and Quality. National and Regional Estimates on Hospital Use for all Patients From the HCUP National Inpatient Sample (NIS).https://hcupnet.ahrq.gov/#setup. Accessed April 15, 2017.Google Scholar
Ohsfeldt RL, Gandhi SK, Fox KM, Bullano MF, Davidson M. Medical and cost burden of atherosclerosis among patients treated in routine clinical practice.J Med Econ. 2010; 13:500–507. doi: 10.3111/13696998.2010.506348CrossrefMedlineGoogle Scholar
Luengo-Fernandez R, Gray AM, Rothwell PM. Costs of stroke using patient-level data: a critical review of the literature.Stroke. 2009; 40:e18–e23. doi: 10.1161/STROKEAHA.108.529776LinkGoogle Scholar
Lee S, Anglade MW, Pham D, Pisacane R, Kluger J, Coleman CI. Cost-effectiveness of rivaroxaban compared to warfarin for stroke prevention in atrial fibrillation.Am J Cardiol. 2012; 110:845–851. doi: 10.1016/j.amjcard.2012.05.011CrossrefMedlineGoogle Scholar
Carretero OA, Oparil S. Essential hypertension. Part I: definition and etiology.Circulation. 2000; 101:329–335.LinkGoogle Scholar
Russell MW, Huse DM, Drowns S, Hamel EC, Hartz SC. Direct medical costs of coronary artery disease in the United States.Am J Cardiol. 1998; 81:1110–1115.CrossrefMedlineGoogle Scholar
Go AS, Mozaffarian D, Roger VL,; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics–2014 update: a report from the American Heart Association.Circulation. 2014; 129:e28–e292. doi: 10.1161/01.cir.0000441139.02102.80LinkGoogle Scholar
- 23. US Bureau of Labor Statistics. Archived Consumer Price Index Detailed Reports.https://www.bls.gov/cpi/data.htm. Accessed March 15, 2018.Google Scholar
Weinstein MC, Siegel JE, Gold MR, Kamlet MS, Russell LB. Recommendations of the panel on cost-effectiveness in health and medicine.JAMA. 1996; 15:1253–1258.CrossrefGoogle Scholar
Siu AL; U.S. Preventive Services Task Force. Screening for high blood pressure in adults: US preventive services task force recommendation statement.Ann Intern Med. 2015; 163:778–786. doi: 10.7326/M15-2223CrossrefMedlineGoogle Scholar
Pickering TG, Shimbo D, Haas D. Ambulatory blood-pressure monitoring.N Engl J Med. 2006; 354:2368–2374. doi: 10.1056/NEJMra060433CrossrefMedlineGoogle Scholar
Krakoff LR. Cost-effectiveness of ambulatory blood pressure: a reanalysis.Hypertension. 2006; 47:29–34. doi: 10.1161/01.HYP.0000197195.84725.66LinkGoogle Scholar
Piper MA, Evans CV, Burda BU, Margolis KL, O’Connor E, Whitlock EP. Diagnostic and predictive accuracy of blood pressure screening methods with consideration of rescreening intervals: a systematic review for the U.S. Preventive Services Task Force.Ann Intern Med. 2015; 162:192–204. doi: 10.7326/M14-1539CrossrefMedlineGoogle Scholar
Fletcher BR, Hinton L, Hartmann-Boyce J, Roberts NW, Bobrovitz N, McManus RJ. Self-monitoring blood pressure in hypertension, patient and provider perspectives: a systematic review and thematic synthesis.Patient Educ Couns. 2016; 99:210–219. doi: 10.1016/j.pec.2015.08.026CrossrefMedlineGoogle Scholar
Parati G, Omboni S, Albini F, Piantoni L, Giuliano A, Revera M, Illyes M, Mancia G; TeleBPCare Study Group. Home blood pressure telemonitoring improves hypertension control in general practice. The TeleBPCare study.J Hypertens. 2009; 27:198–203.CrossrefMedlineGoogle Scholar
Armstrong D, Matangi M, Brouillard D, Myers MG. Automated office blood pressure - being alone and not location is what matters most.Blood Press Monit. 2015; 20:204–208. doi: 10.1097/MBP.0000000000000133CrossrefMedlineGoogle Scholar
Choi TY, Rhee M, Kim JH, Namgung J, Lee SY, Cho DK, Kim SY, Kim JY, Park JS, Park SM. Multiple office blood pressure measurement with an automated device is superior to blood pressure measured by the doctor in the diagnosis of hypertension: a prospective multicenter study.J Am Coll Cardiol. 2015; 65(suppl 10):A1429.CrossrefGoogle Scholar
Novelty and Significance
What Is New?
Compared with home-based and office-based blood pressure measurement, ambulatory blood pressure monitoring reduces the likelihood of the wrong diagnosis and is associated with cost-savings and better health outcomes in most adults regardless of their office blood pressure screen results.
This study provides the first cost-effectiveness analysis that compares all 3 main diagnostic strategies for hypertension screening from US healthcare payer perspective that accounts for younger population groups and considers the possibility of both white-coat hypertension and masked hypertension.
What Is Relevant?
The US Preventive Services Task Force endorses a recommendation to start blood pressure screening for adults ≥18 years of age.
Prior cost-effectiveness models of different strategies to diagnose hypertension are limited in their scope with regard to younger age groups and screen-negative individuals.
The findings of this study suggest that using ambulatory blood pressure monitoring is the strategy of choice for hypertension diagnosis and treatment initiation for most adults in primary care settings. Reimbursing ambulatory blood pressure monitoring as a diagnostic strategy for hypertension in primary care settings may reduce healthcare costs and improve health outcomes by getting hypertension treatment to the correct patients.