Does High Systolic Blood Pressure Truly Increase Medical Expenditure?—An Empirical Analysis of the New 2017 ACC/AHA Hypertension Guideline

Kazumitsu Nawata^1,2, Moriyo Kimura³

¹Graduate School of Engineering, University of Tokyo, Tokyo, Japan.
²Research Institute of Economy, Trade and Industry (RIETI), Tokyo, Japan.
³The Public Health Institute, Tokyo, Japan.

DOI: 10.4236/health.2018.108079                

Abstract

Background: High blood pressure (BP) or hypertension is considered one of the top global disease burden risk factors. In November 2017, the ACC/AHA and other organizations announced a new hypertension guideline of 130/80 mmHg. Data and Methods: We evaluate the effects of BP on increases in medical expenditures using transformation tobit models and a dataset containing 175,123 medical checkups and 6,312,125 receipts from 88,211 individuals in three health insurance societies. The sample period was April 2013 to March 2016. We first created a database of combined checkup results and medical expenditures. The power transformation tobit model was then used to remove the effects of other variables, and we investigated the relation between medical expenditures and BP, especially systolic BP (SBP). Results: We observed negative effects of SBP on medical expenditures. The results raise uncertainty about the reliability of the new guideline, at least for SBP. Although the simple correlation coefficient of medical expenditures and SBP was positive, the sign of the SBP estimate became negative when a variable representing obesity was included. In terms of other medical checkup items, while LDL is considered the “bad” cholesterol, it reduced medical expenditures. Conclusion: Our results did not support the new 2017 ACC/AHA guideline for SBP. A wide and careful range of reviews not only for heart diseases but also for other disease types will be absolutely necessary. New studies to verify the guideline should also be conducted. Limitations: The dataset was observatory, the sample period only 3 years, and we could not complete a time-series analysis of individuals.

Keywords

Blood Pressure, Hypertension, 2017 ACC/AHA Hypertension Guideline, Health and Medical Checkup

1. Introduction

The World Health Organization (WHO) [1] reports that, “Five of the top 10 selected global disease burden risk factors identified by World Health Report 2002: reducing risks, promoting healthy life―obesity, high blood pressure, high cholesterol, alcohol and tobacco.” In terms of blood pressure (BP), the WHO [2] states: “Worldwide, raised blood pressure is estimated to cause 7.5 million deaths, about 12.8% of the total of all deaths. This accounts for 57 million disability adjusted life years (DALYS) or 3.7% of total DALYS.” It also specified “complication of hypertension accounted for 9.4 million deaths in each year” [3] . Lim et al. [4] also found that hypertension was one of the three leading risk factors for the global disease burden (GDB) in 2010. Guidelines stated individuals should be treated for high BP or hypertension when their systolic BP (SBP; during the heart beat) is 140 mmHg or more, or their diastolic BP (DBP, when the heart is at rest) is 90 mmHg or more [5] . (For problems with BP measurement errors due to the white-coat effect and other factors, see [6] [7] [8] [9] ).

The National Heart, Lung and Blood Institute [10] classified hypertension as Stage 1 (SBP of 140 - 159 mmHg; DBP of 90 - 99 mmHg) and Stage 2 (SBP of 160 mmHg or more; DBP of 100 mmHg). It has been reported that the risk of cardiac diseases increased with BP level in all age groups [11] [12] . WHO and the International Society of Hypertension (ISH) [13] provide guidelines for controlling hypertension WHO-ISH classified hypertension into three categories: Grade 1 (mild) - SBP of 140 - 159 mmHg, DBP of 90 - 99 mmHg; Grade 2 (moderate) - SBP of 160 - 179 mmHg, DBP of 100-109 mmHg; and Grade 3 (severe) - SBP of 180 mmHg or more, DBP of 110 mmHg or more. Based on BP and other risk factors, they concluded that the risk of a major cardiovascular event over a 10-year period was about 20% - 30% in high-risk and 30% or more in high- and very-high-risk groups.

The Prospective Studies Collaboration [14] performed a meta-analysis of individual data for 1,000,000 adults in 61 prospective analyses. They studied 12.7 million person-years at risk. They identified about 56,000 cases of vascular death, including 12,000 stroke, 34,000 ischaemic heart disease (IHD), and other vascular deaths. They found that IHD mortality increased in all age cohorts (from 40 - 49 to 80 - 89) as SBP and DBP increased. Moreover, hypertension lowered quality of life (QOL) [15] [16] , and the true cost including indirect cost such as effects on QOL is considered to be much higher than the direct cost. It is estimated that hypertension caused 4.5% of the current GDB [17] , suboptimal BP cost $370 (US $) billion globally in 2001 [18] , and hypertension cost the U.S. about $51.2 billion per year in 2012-2013 [19] [20] . The Centers for Disease Control and Prevention (CDC) [21] estimated high BP costs the U.S. $48.6 billion per year. The medical expenditure for hypertension and related diseases reached as high as 1.85 trillion yen in fiscal year 2015 in Japan [22] . Almost all studies agree that hypertension is a major health risk factor and costly disease.

Race, genetic and environmental factors, and health administrative activities are important factors for BP [7] [23] [24] . For example, Rose [25] compared the distribution of SBP in Kenyan nomads and London civil servants, and found that the former was much lower than the latter. Nawata et al. [26] reported that various factors such as age, gender, health conditions and lifestyle strongly affected BP levels in Japan. They found that SBP increased about 5 mmHg with 10 years of increased age, and was 4 mmHg higher in males. These facts raised questions about the results of previous studies. For example, the Prospective Studies Collaboration study [14] selection criteria for the analyses did not clarify the reasons for the 10-year age cohort interval, or the methods for removing effects of individual characteristics other than BP.

Many hypertension patients have few or no subjective symptoms. The distribution of BP is close to the normal distribution, and even becomes a statistical example of normal distribution [27] , while the definition of hypertension is artificial and may be altered in the future by the accumulation of new medical knowledge.

More recently, the American College of Cardiology (ACC), American Heart Association (AHA), and nine other organizations [28] published the “2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults” (hereafter, 2017 ACC/AHA guideline). Under the new guideline, the threshold for hypertension, requiring treatment with lifestyle changes and medication, is 130/80 mmHg. The 2017 ACC/AHA guideline replaces the term “prehypertension” with “elevated BP” (SBP 120 - 129 mmHg and DBP below 80 mm Hg) and “stage 1 hypertension” (SBP 130 - 139 mmHg or DBP 80 - 89 mmHg). Stage 2 hypertension is defined as SBP of 140 mmHg or more, or DBP of 90 mmHg or more (replacing BP of 160/100 mmHg or more [28] [29] [30] ). However, the American Academy of Family Physicians (AAFP), one organization asked to join the publication, announced on December 12, 2017, that “AAFP Decides to Not Endorse AHA/ACC Hypertension Guideline, Academy Continues to Endorse JNC8 Guideline” [31] .

As a reason for withholding its endorsement, the AAFP stated that the new guideline lacked a systematic review, and would classify 46% of the U.S. adult population as having hypertension, a 32% increase. Dr. O’Gurek, chair of the AAFP’s Commission on Health of the Public and Science added: “although the guideline’s recommendations were given an evidence quality grade, they weren’t grounded in an assessment of the background resources. Finally, substantial weight was given to the Systolic Blood Pressure Intervention Trial (SPRINT), but other trials were minimized.” Actually, the word “SPRINT” was used many times in the main text of the 2017 ACC/AHA guideline. In the SPRINT [32] , 9361 individuals with a SBP of 130 mmHg or higher and an increased cardiovascular risk, but without diabetes, were randomly assigned to a SBP target of less than 120 mmHg (intensive treatment) or a target of less than 140 mmHg (standard treatment). The trial period was from November 2010 to March 2013, and was stopped early after a median follow-up of 3.26 years owing to a significantly lower rate in the intesive treatment group.

As Nawata, Sekizawa and Kimura [33] point out, however, biases such as a publication [34] [35] [36] , conflict of interest [37] [38] [39] , and termination (or endpoint) biases [40] might exist in randomized clinical trails (CRT). The change of the hypertension guideline will affect the medical system worldwide. As suggested by the AAFP [31] , careful and wide-ranging reviews are needed to determine the hypertension guideline. The Action to Cardiovascular Risk in Diabetes (ACCORD) Study Group [41] did a similar trial for 4733 participants with type 2 diabetes. In that study, 491 were recruited from January-June 2001, and an additional 4242 were recruited from January 2003 through October 2005. The researchers concluded that targeting SBP below 120 mmHg did not reduce the rate of composite outcome of fatal and major nonfatal major cardiovascular events compared to below 140 mmHg. The SPRINT research group [32] themselves admitted: “Rates of serious adverse events of hypertension, ... and acute kidney injury, … were higher in the intensive-treatment group than in the standard-treatment group.” In other words, further investigation as to whether the new guideline is appropriate or not are absolutely necessary. Such research must include not only cardiovascular diseases but also all other (serious) diseases. Since the influence of the AHA/ACC Guideline is so vast, these studies must be done as soon as possible.

In this paper, we analyze the effects of BP (especially SBP) on annual medical expenditures using a dataset containing 175,123 medical checkups and 6,312,125 receipts obtained from 88,211 individuals. The power transformation tobit model [42] is used in the analysis, because the distribution of medical expenditure shows a heavy tail on the right side and many “zeroes” are observed. Although we did not directly analyze heart diseases, medical expenditure is a very important indicator representing the health conditions of an individual. The sample period is from fiscal year 2013 to 2015 (i.e., April 2013 to March 2016).

2. Data and Methods

In this paper, we used an anonymized dataset combining medical checkups and receipts. First, we compared the distributions of medical expenditures for each fiscal year. Various characteristics and health conditions, including BP, affect medical expenditures. To measure the effects of BP correctly, it is necessary to remove the effects of other variables; we therefore employed a regression-type analysis. However, there are problems, as previously identified by Gregori et al. [43] . One is that medical expenditures take many zero values (about 20%). The other is that the distribution is asymmetric and has a very heavy tail, and the variance becomes very large. It may not be best to use the ordinary least squares methods in such cases. We therefore used the power transformation tobit model for the analysis.

2.1. Data

Japan employs a public health insurance system for the whole nation , and everyone must join some type of public health insurance organization. Corporations form heath insurance societies for employees and their family members. Most employees 40 or older are required to have a medical checkup once a year by the Industrial Safety and Health Act [44] , and family members can voluntarily have such checkups. The dataset was created with the cooperation of three health insurance societies (Societies 1 - 3).

Society 1 was formed at a large Japanese corporation with offices and operational centers throughout Japan. Societies 2 and 3 were formed by groups of corporations. The dataset contained information regarding 175,123 medical checkups from 88,211 individuals between fiscal years 2013-2015 (i.e., April 2013 to March 2016). The monthly reports of medical treatments and payments, called “receipts”, are sent from medical institutes to the health insurance associations. Payments are made to the medical institutes after checking the receipts. (According to the Health Insurance Claims Review & Reimbursement Services [45] , nearly 99% of receipts were paid as requested in March 2018.) Receipts were classified into five categories: dental; inpatients of DPC hospitals; outpatients and inpatients of non-DPC hospitals; and pharmacies. Of these, we used the sum of DPC, outpatient & non-DPC hospital, and pharmacy receipts as the medical expenditure.

Japan measures medical expenditures in points, paying 10 yen per point to medical institutes. Moreover, the same points are allotted for the same treatments and medicines determined by the government regardless of region and medical institution, with a few exceptions [46] . This means that medical expenditure is a good indicator of the health condition of an individual in Japan. A total of 6,312,125 receipts were summed, and medical expenditures in each fiscal year were calculated. In the analysis, we used the dataset containing 175,123 observations for which both the results of checkups and medical expenditures were available in the same fiscal year.

2.2. Power Transformation Tobit Model

Many studies of medical expenditures use various types of regression analyses [43] [47] . Medical expenditures do not become negative values (left censored), and there exist many zero values. Moreover, the distribution is asymmetric, having a very heavy tail on the right side, and variance is very large. Therefore, analysis by a standard regression model and ordinary least squares methods may not be the best in this case. We therefore used the power transformation model in this analysis. We briefly explain the power transformation tobit model.

1) Tobit model

The standard tobit model (or censored regression model) is given by

$y_i^{*}={x^{\prime }}_i\beta +u_i,u_i~N\left(0,{\sigma }^2\right),i=1,\cdots ,n$ (1)

$y_i=\{\begin{array}{l}{y}_i^{*}\text{if}{y}_i^{*}>0\\ 0\text{if}{y}_i^{*}\le 0\end{array}$

where $y_i^{*}$ is a latent variable and its value is not observable when it is negative, $x_i$ is a vector of explanatory variables, $\beta$ is a vector of unknown parameters, and $u_i$ is an error term following the normal distribution with mean 0 and variance ${\sigma }^2$ . For more detail, see Amamya [48] .

2) Power transformation

In the tobit model, normality of the error term is assumed. However, the medical expenditures follow the distribution with a heavy tail on the right side, and do not follow normal distribution. Since the medical expenditure takes zero values, we cannot use the log transformation. In this study, we use the power transformation to make the error term cross to the normal distribution. Gregori et al. [43] considered the Box-Cox transformation [49] , including both log and power transformations. However, the log transformation is not included because of zero values; the power transformation is sufficient in this study. The power transformation is given by:

$y=M^{\alpha },0<\alpha \le 1$ (2)

where M is the medical expenditure and $\alpha$ is the transformation parameter.

3) Power transformation tobit model

In this paper, we use the power transformation tobit model that combines the tobit model and power transformation. Here,

$\frac{\text{d}y}{\text{d}M}=\alpha M^{\alpha -1}$ . (3)

Therefore, the likelihood function to be maximized is obtained by

$\begin{array}{c}\log L\left(\theta \right)={\displaystyle {\sum }_{M>0}\left\{-\frac{1}{2}\log \left(\text{2π}\right)-\log \left(\sigma \right)-\frac{{\left(y-x^{\prime }\beta \right)}^2}{2{\sigma }^2}+\log \left(\alpha \right)+\left(\alpha -1\right)\log \left(M\right)\right\}}\\ +{\displaystyle {\sum }_{M=0}\log \left\{1-\Phi \left(\frac{x^{\prime }\beta }{\sigma }\right)\right\}}\end{array}$ (4)

where $\Phi$ is the standard normal distribution function. We consider several different models, and they are explained with the estimation results. Note that when $\alpha$ is given, $\beta$ and ${\sigma }^2$ are obtained by the standard tobit model. The maximum likelihood estimator (MLE) is calculated by the following scanning method [50] :

1) Choose ${\alpha }_1<{\alpha }_2<\cdots <{\alpha }_m$ from 0.01-1.0 at intervals of 0.01.

2) Calculate $\hat{\beta }$ and ${\hat{\sigma }}^2$ for each by the tobit maximum likelihood method.

3) Choose $\hat{\alpha }$ that maximizes the BC likelihood function.

4) Choose ${\alpha }_i$ in the neighborhood of $\hat{\alpha }$ with an interval of 0.0001, and repeat steps (2) and (3).

5) Determine the final estimator.

3. Distributions of Medical Expenditures and Blood Pressures

3.1. Medical Expenditures

Figure 1 shows the distribution of medical expenditures. The distribution is skewed and has a very heavy tail on the right side. The basic statistics (points) are as follows: mean: 13,356, median: 4061, standard deviation (SD): 39,241, skewness: 11.0, kurtosis: 174.0, and maximum: 1,212,291. A total of 20.2% of all observations of medical expenditures are zero. On the other hand, 1.9, 0.4, and 0.16% used more than 100,000, 300,000 and 500,000 points, and their medical expenditures accounted for 30.3, 14.3 and 7.8%, respectively, of total medical expenditures.

3.2. SBP and DBP

Figures 2-4 present the distributions of SBP and DBP, respectively. Excluding observations of BP that are too high (SBP > 300 or DBP > 200) or too low (SBP, DBP < 30), the basic statistics of SBP and DBP of 175,083 observations are given in Table 1. Under the 140/90 criterion, 22.8% are diagnosed with hypertension. Under the new guideline of 130/80, this value jumps up to 51.1%, more than a half of observations, suggesting the effect of changing the criterion is quite large.

3.3. Relation between SBP and Medical Expenditures

Figure 5 shows the relation of SBP to average medical expenditures. Average medical expenditures are averages of SBP at intervals of 5 mmHg (i.e., for a SBP

Figure 1. Distribution of medical expenditures.

Figure 2. Distribution of SBP.

Figure 3. Distribution of DBP

Table 1. Summaries of SBP and DBP.

Figure 4. Relationship of SBP and DBP, and 130/80 and 140/90 criteria

Figure 5. Relationship of average medical expenditure and SBP.

value of 130 mmHg, average medical expenditures of observations between 127.5 - 132.5 mmHg). The figure shows an upward trend, and the correlation coefficient between SBP and average medical expenditure between 80 - 180 mmHg is 0.843. This result seems to support the new guideline. But the question is whether this relation is a true or spurious one.

As already mentioned, various factors affect medical expenditures and BP. Figure 6 shows the relationships between the average medical expenditures at intervals of 5 years by gender. As an individual ages, medical expenditure increases, and there is a difference between males and females. Nawata et al. [26] pointed out that BP is strongly affected by age and gender. SBP increases by 5 mmHg over a 10-year aging period, and the SBP of males is about 4 mmHg higher than that of females. BP becomes higher as an individual grows older. In the next section, we conduct the analysis by the transformation tobit model.

Figure 6. Medical expenditures by age and gender.

4. Results of Analysis by the Power Transformation Model

Models and Explanatory Variables

Regression models are used to remove the effects of various factors. We first consider the following the power transformation tobit model in Equation (1).

Model A:

$\begin{array}{c}{y}_i^{*}={\beta }_1+{\beta }_{2}Age+{\beta }_{3}Female+{\beta }_{4}Height+{\beta }_{5}BMI+{\beta }_{6}SBP+{\beta }_{7}DBP\\ +{\beta }_{10}HDL+{\beta }_{9}LDL+{\beta }_{10}Triglyceride+{\beta }_{11}GGP+{\beta }_{12}AST+{\beta }_{13}ALT\\ +{\beta }_{14}Boold\_Sugar+{\beta }_{14}Urine\text{\_}sugar+{\beta }_{15}Urin\text{\_}protein\\ +{\beta }_{16}F\text{\_}year14+{\beta }_{17}F\text{\_}year15+{\beta }_{18}Society2+{\beta }_{19}Socitey3+u\end{array}$ (5)

Besides SBP and DBP (mmHg), the following explanatory variables are used. Age, Female (1: if female, 0: otherwise), Height (cm), BMI ( = height (m)/weight (kg)²), HDL (high density lipoprotein cholesterol blood, mg/dL), LDL (low-density lipoprotein cholesterol, mg/DL), Triglyceride (mg/dL), GGP (γ-glutamyl transferase, U/L), AST (aspartate aminotransferase, U/L), ALT (alanine aminotransferase, U/L), Blood_sugar (mg/dL), Urine_sugar (integers of 1-5, sugar in urine increasing with number; 1 is normal, 5 is worst), Urine_protein (same as Urine_sugar), F_year 14 (1: fiscal year 2014, 0: otherwise), F_year 2015 (1: fiscal year 2015), Society 2 (1: Society 2, 0: otherwise) and Society3 (1: Society 3, 0: otherwise) where U/L is units per liter.

For all explanatory variables, objectively measured values could be obtained from medical checkup data. This model did not include variables related to anamnesis, currently treated diseases, or individual lifestyles. For example, hypertension is an important risk factor of diabetes [51] . Suppose that the relation may be “hypertension = > diabetes = > medical expenditure”. In this case, if a variable representing diabetes is included, the relation of “hypertension = > medical expenditure” could not be observed. In econometric terms, we used the reduced form so as not to miss any possible effects of BP.

Age, Female and Height represent basic individual characteristics; BMI represents obesity; while HDL, LDL and Triglyceride represent lipid concentration in the blood. If lipid concentration is abnormal (too high or too low), an individual is diagnosed as dyslipidemia. Lipoproteins are proteins that carry cholesterol through the blood. LDL cholesterol makes up most of the body’s cholesterol, and HDL cholesterol absorbs cholesterol and carries it back to the liver [52] . Triglyceride is the most common type of fat in the body, and stores excess energy [53] . Although our bodies need lipids to build cells, too much could be a problem [54] .

Currently, dyslipidemia is mainly hyperlipidemia, where the lipid concentration is too high. WHO [55] warned: “Raised cholesterol increases the risks of heart disease and stroke. Globally, a third of ischemic heart disease is attributable to high cholesterol. Overall, raised cholesterol is estimated to cause 2.6 million deaths (4.5% of total) and 29.7 million disability adjusted life years (DALYS), or 2.0% of total DALYS.” LDL and HDL cholesterols are classified as “bad” and “good”. LDL (bad) cholesterol contributes to fatty buildups in arteries, and raises the risk factor for chronic coronary heart disease, heart attack and stroke. On the other hand, HDL (good) cholesterol removes LDL cholesterol from the arteries [52] [53] . GGP, AST and ALT are mainly related to liver functions; Blood_sugar and Urine_sugar are important indicators of diabetes; and Urine_protein represents the condition of the kidneys [56] .

We first excluded observations with missing values in explanatory variables. We then excluded the following observations: BMI too high (over 100); SBP too high (over 300) or too low (under 30); DBP too high (over 200) or too low (under 30); SBP-DBP becomes zero or negative; HDL too high (over 500); LDL too high (over 500); Triglyceride too high (over 1000); GGT too high (over 1000); ALT too high (over 500); AST too high (over 500); and Blood_sugar too high (500). Excluding observations with missing values in explanatory variables, we used 173,498 (M > 0: 138,407, and M = 0: 35,091) observations for the estimation of the model. Among these observations, 20.2% of medical expenditures were zero, and 79.8% were positive values.

Model A assumes that the effects of BP are continuous. However, it is possible that BP affects health conditions only if it becomes higher than certain threshold values (hereafter, threshold value hypothesis; criteria such as 140/90 and 130/80 are obviously based on this hypothesis). Therefore, we consider the model using dummy variables of SBP. Note that we analyzed only SBP as the SPRINT.

Model B:

$\begin{array}{c}{y}_i^{*}={\beta }_1+{\beta }_{2}Age+{\beta }_{3}Female+{\beta }_{4}Height+{\beta }_{5}BMI+{\beta }_{6}SBP130\\ +{\beta }_{7}SBP140+{\beta }_{8}SBP160+{\beta }_{9}SBP180+{\beta }_{10}DBP+{\beta }_{11}HDL\\ +{\beta }_{12}LDL+{\beta }_{13}Triglyceride+{\beta }_{14}GGP+{\beta }_{15}AST+{\beta }_{16}ALT\\ +{\beta }_{17}Boold\_Sugar+{\beta }_{18}Urine\text{\_}sugar+{\beta }_{19}Urin\text{\_}protein\\ +{\beta }_{20}F\text{\_}year14+{\beta }_{21}F\text{\_}year15+{\beta }_{22}Society2+{\beta }_{23}Socitey3+u\end{array}$ (6)

SBP130 (1: if SPB ≥ 130, 0:otherwise), SBP140 (1: if SPB ≥ 140, 0:otherwise), SBP160 (1: if SPB ≥ 160, 0:otherwise) and SBP180 (1: if SPB ≥ 180, 0:otherwise) are dummy variables representing threshold values. Table 2 presents a summary of the explanatory variables.

Table 3 lists the result of estimations for Model A. Figure 7 shows the distribution of the medical expenditures after the power transformation ( $y=M^{0.4088}$ ). The distribution is much closer to the normal distribution, suggesting usefulness of the model for analyzing this dataset. Since the sample size was quite large, all variables except F_year 14 were significant at the 1% level. The estimates of Age, Female, Height, BMI, Triglyceride, GGT, AST, ALT, Blood_suger, Urine_suger, Urine_protein, andF_year15 were positive, with these variables making medical

Figure 7. Distribution medical expenditures after the transformation $\left(y=M^{0.4088}\right)$ .

Table 2. Explanatory variables.

SD: Standard deviation.

Table 3. Result of estimation (Model A).

SE: standard error, **: significant at 1% level.

expenditures higher. The effects of most of these variables were as expected. On the other side, the estimates of LDL, HDL, Society 2 and Society 3 were negative. Although LDL Cholesterol is called “bad” and HDL “good” [53] , higher levels of both LDL and HDL cholesterols reduced medical expenditures in our study. Hence further studies are necessary to determine the roles and functions of cholesterols, especially LDL cholesterol. This is one important finding of this study.

The medical expenditures of Societies 2 and 3 were lower than those of Society 1. Society 1 was formed by one large corporation, while Societies 2 and 3 were formed by groups of smaller corporations. Although the reason why cannot be elucidated, it might be necessary to check and revise the healthcare system in Society 1. Although the sign of DBP was positive and significant, the estimate of SBP was −0.0566, and its t-values were −10.42 and significant at any reasonable significance at any reasonable level. This means that higher SBP reduced medical expenditures.

Table 4 presents the estimation results of Model B, which contained the threshold value dummies for SBP. The values of estimations for variables other than BP were very similar to those of Model A. For the SBP dummy variables, SBP130, SBP140 and SBP180 dummies were not significant even at the 5% level, despite the fact that the sample size was quite large. Only the SBP160 dummy was significant at the 1% level, but the estimated value was negative. Although it was not significant at the 5% level, the estimate of DBP becomes a negative value in this model. These findings do not support the threshold value hypothesis, at least for SBP.

5. Discussion

The effects of BP on medical expenditures are mixed. Higher DBP makes them higher, but higher SBP makes them lower. We evaluated the relations between medical expenditures and high SBP or SBP hypertension. As shown in Figure 5, there exists an upward trend between SBP and average medical expenditures.

We consider a simple regression model of Equation (1) that is given by:

Model C:

$y_i^{*}={\beta }_1+{\beta }_{2}SBP+u$ (7)

Then we get (standard errors are in parentheses),

$\begin{array}{l}{y}_i^{*}=13.264+0.1308SBP,\hat{\alpha }=0.4094\left(0.0007\right)\\ \left(0.466\right)\left(0.00367\right)\end{array}$ (8)

The estimate of SBP is positive, and its t-value is 35.70 and significant at any reasonable significance level. As shown in Figure 6, age and gender might affect medical expenditures. We add Age and Female, and consider the model:

Model D:

$y_i^{*}={\beta }_1+{\beta }_{2}Age+{\beta }_{3}Female+{\beta }_{4}SBP+u$ (9)

The estimation results of this model are given by:

$\begin{array}{l}\hat{\alpha }=0.4088\left(0.0007\right)\\ y_i^{*}=-24.016+0.8508Age+5.8150Female+0.0747SBP\\ \left(0.5665\right)\left(0.0100\right)\left(0.15468\right)\left(0.00359\right)\end{array}$ (10)

Although the size is almost half that of the previous case, the estimate of SBP is still positive, and the t-value is 20.86 and significant at any reasonable level.

We then add BMI, representing obesity, and consider the model,

Model E:

$y_i^{*}={\beta }_1+{\beta }_{2}Age+{\beta }_{3}Female+{\beta }_{4}SBP+{\beta }_{5}BMI+u$

Table 4. Result of estimation (Model B).

SE: standard error, **: significant at 1% level.

The estimation result is given by:

$\begin{array}{l}\hat{\alpha }=0.4085\left(0.0007\right)\\ y_i^{*}=-58.963+1.0416Age+9.0088Female-0.0250SBP+1.7028BMI\\ \left(0.6973\right)\left(0.0101\right)\left(0.1554\right)\left(0.0035\right)\left(0.0186\right)\end{array}$ (11)

In Model E, the coefficient of SBP becomes negative and significant at the 1% level. Muntner et al. [57] analyzed data from the US National Health and Nutrition Examination Survey (NHANES). They pooled data from the 2011-2012 and 2013-2014 NHANES cycles of adult participants, 20 years of age and older (n = 10,907). They declared that, “Implementation of the 2017 ACC/AHA hypertension guideline has the potential to increase the prevalence of hypertension and use of antihypertensive medication among U.S. adults. This should translate into a reduction in CVD events.” Although age, gender, race, smoking, total and HDL-cholesterol, and diabetes were included, “obesity” was not considered in their analysis. The correlation coefficient of SBP and BMI is 0.307 in this study. The relation between obesity and hypertension has been recognized for more than a half century [58] , and many studies have been conducted. For details, see the review works of Kotchen [59] , Jiang et al. [60] and Leggio et al. [61] . Jiang et al. [60] declared that: “The mechanisms underlying obesity-associated hypertension or other associated metabolic diseases remains to be adequately investigated.” They furthermore contended that, “There is no single cause to explain all the cases of obesity worldwide.” The relation between BP and obesity should be carefully studied.

The results of this study suggest that the risks of hypertension might be spurious, and other factors such as obesity might be affecting health condition. Moreover, BP has been found to affect not only heart diseases but also various other health conditions such as kidney diseases [32] [62] . The influences of the new guideline of 130/80 are so large that careful reviews of various studies including analyses of various factors and diseases (not only heart diseases) affected by BP levels are absolutely necessary to determine whether or not the new guideline is appropriate.

6. Conclusions

In this study, we mainly evaluated the effects of BP on medical expenditures by the transformation tobit models using a dataset containing 175,123 medical checkups and 6,312,125 receipts obtained from 88,211 individuals obtained from three health insurance societies. Medical expenditure is a very good indicator of an individual’s health condition, because under the current Japanese national health insurance system, most medical institutes receive the same amount for the same treatments and medicines, independent of region. We first considered a model that included various heath information factors for individuals obtained in yearly medical examinations. Although the estimate of DBP had a positive value, that of SBP became negative and the absolute t-value was larger than 10, suggesting that the new guideline for SBP was not supported.

We then theorized that threshold values and BP might affect health condition only if BP exceeded those values (threshold value hypothesis). We used SBP dummies to check the threshold value hypothesis, but the results did not support the hypothesis for SBP. While the estimates of most other variables had expected signs, LDL cholesterol, considered “bad”, showed the opposite result. It is likely we will need additional studies for the evaluation of cholesterols.

We then evaluated the relation between medical expenditures and SBP. Medical expenditures and SBP were positively correlated, and if the simple model only contained SBP, the estimate became a positive value. Although the size of the coefficient was almost cut in half, the sign did not change if age and gender variables were considered in the model. However, when BMI, representing obesity, was added, the estimate of SBP became negative and significant at the 1% level.

It is possible that the relation between SBP and medical expenditures might be spurious, and the correlation of SBP and BMI might affect the result. The relation between BP and obesity should be carefully studied. Moreover, the effect of the new 2017 ACC/AHA guideline, the first comprehensive hypertension clinical practice guideline since 2003 [30] , is so large that a careful and wide range of reviews of various studies, not only of heart diseases but for other types of diseases as well, are absolutely necessary. New studies verifying the guideline should also be conducted.

In this paper, we evaluated medical expenditures, not the risks of BP on heart diseases. Evaluation of the effects of BP on heart diseases and other important diseases is needed. It will also be necessary to analyze a larger and longer time-range dataset from various insurance societies to make the analysis more precise. These are subjects to be studied in future.

Acknowledgements

This study was supported by a Grant-in-Aid for Scientific Research, “Analyses of Medical Checkup Data and Possibility of Controlling Medical Expenses (Grant Number: 17H22509),” from the Japan Society of Science, and by a research grant, “Exploring Inhibition of Medical Expenditure Expansion and Health-oriented Business Management Based on Evidence-based Medicine” from the Research Institute of Economics, Trade and Industry (RIETI). The dataset was anonymized at the health insurance societies. This study was approved by the Institutional Review Boards of the University of Tokyo (number: KE17-30). The authors would like to thank the health insurance societies for their sincere cooperation in providing us the data. We would also like to thank an anonymous referee for his/her helpful comments and suggestions.

Conflicts of Interest

The authors declare no conflicts of interest regarding the publication of this paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]	World Health Organization (WHO) (2013) Obesity, High Blood Pressure, High Cholesterol, Alcohol and Tobacco: The World Health Organization’s Response. World Health Report. http://www.who.int/whr/media_centre/factsheet3/en/
[2]	WHO (2016) Raised Blood Pressure. http://www.who.int/gho/ncd/risk_factors/blood_pressure_prevalence_text/en/
[3]	WHO (2013) A Global Brief on Hypertension. http://www.who.int/cardiovascular_diseases/publications/global_brief_hypertension/en/
[4]	Lim, S.S., Vos, T., Flaxman, A.D., et al. (2012) A Comparative Risk Assessment of Burden of Disease and Injury Attributable to 67 Risk Factors and Risk Factor Clusters in 21 Regions, 1990-2010: A Systematic Analysis for the Global Burden of Disease Study 2010. Lancet, 380, 2224-2260. https://doi.org/10.1016/S0140-6736(12)61766-8
[5]	Blood Pressure Association, UK (2008) Blood Pressure Chart for Adults. http://www.bloodpressureuk.org/BloodPressureandyou/Thebasics/Bloodpressurechart
[6]	Wohlfahrt, P., Cifková, R., Movsisyan, N., et al. (2016) Threshold for Diagnosing Hypertension by Automated Office Blood Pressure Using Random Population. Journal of Hypertension, 2180-2186. https://doi.org/10.1097/HJH.0000000000001076
[7]	Omiboni, S., Aristizabal, D., De la Sierra, A., et al. (2016) Hypertension Types Defined by Clinic and ambulatory Blood Pressure in 14142 Patients Referred to Hypertension Clinics Worldwide. Data from ARTEMIS Study. Journal of Hypertension, 34, 2187-2198. https://doi.org/10.1097/HJH.0000000000001074
[8]	AACN Practice Alert (2016) Obtaining Accurate Noninvasive Blood Pressure Measurements in Adults. Critical Care Nurse, 36, e12-e16. https://doi.org/10.4037/ccn2016590
[9]	Nawata, K. and Kimura, M. (2017) Reliability of Blood Pressure Measurements: An Analysis of the White Coat Effect and Its Fluctuations. Health, 9, 506-519. https://doi.org/10.4236/health.2017.93036
[10]	National Heart, Lung and Blood Institute, US Department of Health & Human Services (2015) Description of High Blood Pressure.
[11]	Vasan, S.R., Larson, M.G., Leip, E.P., et al. (2001) Impact of High-Normal Blood Pressure on the Risk of Cardiovascular Disease. New England Journal of Medicine, 345, 1291-1297. https://doi.org/10.1056/NEJMoa003417
[12]	Fujiyoshi, A., Ohkubo, T., Miura, K., et al. (2012) Blood Pressure Categories and Log-Term Risk of Cardiovascular Disease According to Age Group in Japanese Men and Women. Hypertension Research, 35, 947-953. https://doi.org/10.1038/hr.2012.87
[13]	Guidelines Subcommittee (1999) 1999 World Health Organization - International Society of Hypertension Guidelines for Management of Hypertension. Journal of Hypertension, 17, 151-183.
[14]	Prospective Studies Collaboration (2002) Age-Specific Relevance of Usual Blood Pressure to Vascular Mortality: A Meta-Analysis of Individual Data for One Million Adults in 61 Prospective Studies. Lancet, 360, 1903-1913. https://doi.org/10.1016/S0140-6736(02)11911-8
[15]	Blanchflower, D.G. and Oswald, A.J. (2008) Hypertension and Happiness across Nations. Journal of Health Economics, 27, 218-233. https://doi.org/10.1016/j.jhealeco.2007.06.002
[16]	Mojon-Azzi, S. and Sousa-Pazo, A. (2011) Hypertension and Life Satisfaction: An Analysis Using Data from the Survey of Health, Aging and Retirement in Europe. Applied Economics Letters, 18, 183-187. https://doi.org/10.1080/13504850903508291
[17]	World Health Organization, International Society of Hypertension Writing Group (2003) 2003 World Health Organization (WHO)/International Society of Hypertension (ISH) Statement on Management of Hypertension. Journal of Hypertension, 11, 1983-1992.
[18]	Gaziano, T.A., Bitton, A., Anand, S., et al. (2009) The Global Cost of Nonoptimal Blood Pressure. Journal of Hypertension, 27, 1472-1477. https://doi.org/10.1097/HJH.0b013e32832a9ba3
[19]	Benjamin, E.J., Blaha, M.J., Chiuve, S.E., et al. (2017) Heart Disease and Stroke Statistics—2017 Updates: A Report from the American Heart Association. Circulation, 135, e146-e603. https://doi.org/10.1161/CIR.0000000000000485
[20]	Wang, G., Grosse, G.S.D. and Schooley, M.W. (2017) Conducting Research on the Economics of Hypertension to Improve Cardiovascular Health. American Journal of Preventive Medicine, 53, S115-S117. https://doi.org/10.1016/j.amepre.2017.08.005
[21]	Centers for Disease Control and Prevention (CDC) (2016) High Blood Pressure Fact Sheet. https://www.cdc.gov/dhdsp/data_statistics/fact_sheets/fs_bloodpressure.htm
[22]	Ministry of Health, Welfare, and Labor (2017) Heisei 27 nendo kokumin iryouhi no gayou (Summary of National Medical Expenditures). (In Japanese)
[23]	Ibrahim, M.M. and Damasceno, A. (2012) A Hypertension in Developing Countries. The Lancet, 380, 611-619. https://doi.org/10.1016/S0140-6736(12)60861-7
[24]	Gorman, B.K. and Porter, J.R. (2011) Social Networks and Support, Gender, and Radical/Ethic Disparities in Hypertension among Older Adults. Population Research & Policy Review, 20, 885-911. https://doi.org/10.1007/s11113-011-9215-4
[25]	Rose, G. (2001) Sick Individuals and Sick Populations. International Journal of Epidemiology, 30, 427-432. https://doi.org/10.1093/ije/30.3.427
[26]	Nawata, K., Matsumoto, A., Kajihara, K., et al. (2017) Evaluation of the Distribution of and Factors Affecting Blood Pressure Using Medical Checkup Data in Japan. Health, 9, 124-137. https://doi.org/10.4236/health.2017.91009
[27]	Surfstat.australia: An Online text in Introductory Statistics (2018) Summarising and Presenting Data, Normal Distributions. https://surfstat.anu.edu.au/surfstat-home/1-3-1.html
[28]	Whelton, P.K., Carey, R.M., Arrow, W.S., et al. (2018) 2017 ACC/AHA/AAPA/ABC /ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology, 71, e13-e115.
[29]	Whelton, P.K., Carey, R.M., Arrow, W.S., et al. (2018) 2017 Guild Line for Blood Pressure in Adults. American College of Cardiology. http://www.acc.org/latest-in-cardiology/ten-points-to-remember/2017/11/09/11/41/2017-guideline-for-high-blood-pressure-in-adults
[30]	Carey, R.M. and Whelton, P.K. for the 2017 ACC/AHA Hypertension Guideline Writing Committee (2018) Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: Synopsis of the 2017 American College of Cardiology/American Heart Association Hypertension Guideline. Annals of Internal Medicine, 168, 351. https://doi.org/10.7326/M17-3203
[31]	American Academy of Family Physicians (AAFP) (2017) AAFP Decides to Not Endorse AHA/ACC Hypertension Guideline, Academy Continues to Endorse JNC8 Guideline. AAFP News. https://www.aafp.org/news/health-of-the-public/20171212notendorseaha-accgdlne.html
[32]	Wright, J.T., Williamson, J.D., Whelton, P.K., et al. (2015) A Randomized Trial of Intensive versus Standard Blood-Pressure Control. New England Journal of Medicine, 373, 2103-2116. https://doi.org/10.1056/NEJMoa1511939
[33]	Nawata, K., Sekizawa, Y. and Kimura, M. (2018) Evaluation of Blood Pressure Control Medicines Using Health and Medical Checkup Data in Japan: Alternative Methods for Randomized Controlled Trials. Health, 10, 587-603. https://doi.org/10.4236/health.2018.105047
[34]	Dickersin, K., Chan, S.S, Chalmersx, T.C., et al. (1987) Publication Bias and Clinical Trials. Controlled Clinical Trials, 8, 343-353. https://doi.org/10.1016/0197-2456(87)90155-3
[35]	Hopewell, S., Loudon, K., Clarke, M.J., et al. (2009) Publication Bias in Clinical Trials Due to Statistical Significance or Direction of Trial Results. Cochrane Database of Systematic Reviews, No. 1, MR000006. http://sverigesradio.se/diverse/appdata/isidor/files/83/5775.pdf
[36]	Turner, E.H., Matthews, A.M., Linardatos, E., et al. (2008) Selective Publication of Antidepressant Trials and Its Influence on Apparent Efficacy. New England Journal of Medicine, 358, 252-260. https://doi.org/10.1056/NEJMsa065779
[37]	Lexchin, J., Bero, L.A., Djulbegovic, B.B., et al. (2003) Pharmaceutical Industry Sponsorship and Research Outcome and Quality: Systematic Review. BMJ, 326, 1167. https://doi.org/10.1136/bmj.326.7400.1167
[38]	Perlis, R.H., Perlis, C.S., Wu, Y., et al. (2005) Industry Sponsorship and Financial Conflict of Interest in the Reporting of Clinical Trials in Psychiatry. American Journal of Psychiatry, 162, 1957-1960. https://doi.org/10.1176/appi.ajp.162.10.1957
[39]	Lundh, A., Lexchin, J., Mintzes, B., et al. (2017) Industry Sponsorship and Research Outcome. Cochrane Database of Systematic Reviews, No. 12, MR000033. http://cochranelibrary-wiley.com/doi/10.1002/14651858.MR000033.pub3/full https://doi.org/10.1002/14651858.MR000033.pub3
[40]	Park, T.R., Rodriguez, D.M. and Roth, F.P. (2015) Why Clinical Trials Are Terminated. https://www.biorxiv.org/content/biorxiv/early/2015/07/02/021543.full.pdf
[41]	The ACCORD Study Group (2010) Effect of Intensive Blood-Pressure Control in Type 2 Diabetes Mellitus. New England Journal of Medicine, 362, 1575-1585. https://doi.org/10.1056/NEJMoa1001286
[42]	Nawata, K. and Kimura, M. (2017) Evaluation of Medical Costs of Kidney Diseases and Risk Factors in Japan. Health, 9, 1734-1749. https://doi.org/10.4236/health.2017.913127
[43]	Gregori, D., Petrinco, M., Simona, B., et al. (2011) Regression Models for Analyzing Costs and Their Determinants in Health Care: An Introductory Review. International Journal for Quality in Health Care, 23, 331-341. https://doi.org/10.1093/intqhc/mzr010
[44]	Japanese Law Translation (2009) Industrial Safety and Health Act. http://www.japaneselawtranslation.go.jp/law/detail/?id=1926&vm=&re
[45]	Health Insurance Claims Review & Reimbursement Services (2018) Sinsa joukyou heisei 30 nen 3 gatsu (Results of Checking, March 2018). (In Japanese)
[46]	Nawata, K., Ii, M., Toyama, H., et al. (2009) Evaluation of the Inclusive Payment System Based on the Diagnosis Procedure Combination with Respect to Cataract Operations in Japan. Health, 1, 93-103. https://doi.org/10.4236/health.2009.12016
[47]	Deng, L., Lou, W. and Mitsakakis, N. (2018) Modeling Right-Censored Medical Cost Data in Regression and the Effects of Covariates. Statistical Methods & Applications, 1-13. https://doi.org/10.1007/s10260-018-0428-0
[48]	Amemiya, T. (1985) Advanced Econometrics. Harvard University Press, Cambridge.
[49]	Box, G.E.P. and Cox, D.R. (1964) An Analysis of Transformation. Journal of the Royal Statistical Society B, 26, 211-252.
[50]	Nawata, K. (1994) Estimation of Sample Selection Bias Models by the Maximum Likelihood Estimator and Heckman’s Two-Step Estimator. Economic Letters, 45, 33-40. https://doi.org/10.1016/0165-1765(94)90053-1
[51]	Nawata, K. and Kimura, M. (2017) An Analysis of the Medical Costs of and Factors Affecting Diabetes Using the Medical Checkup and Payment Dataset in Japan: Can We Reduce the Prevalence of Diabetes? Health, 9, 1113-1127. https://doi.org/10.4236/health.2017.98081
[52]	CDC (2017) LDL and HDL Cholesterol: “Bad” and “Good” Cholesterol. https://www.cdc.gov/cholesterol/ldl_hdl.htm
[53]	American Heart Association (AEA) (2017) HDL (Good), LDL (Bad) Cholesterol and Triglycerides. http://www.heart.org/HEARTORG/Conditions/Cholesterol/HDLLDLTriglycerides/HDL-Good-LDL-Bad-Cholesterol-and-Triglycerides_UCM_305561_Article.jsp#.WxUKUkjRCUk
[54]	AEA (2017) About Cholesterol. http://www.heart.org/HEARTORG/Conditions/Cholesterol/AboutCholesterol/About-Cholesterol_UCM_001220_Article.jsp#.Wtrw2ojFKUk
[55]	WHO (2018) Raised Cholesterol. http://www.who.int/gho/ncd/risk_factors/cholesterol_text/en/
[56]	U.S. National Library of Medicine (2018) Medical Encyclopedia. https://medlineplus.gov/encyclopedia.html
[57]	Muntner, P., Carey, R.M., Gidding, S., et al. (2018) Potential U.S. Population Impact of the 2017 ACC/AHA High Blood Pressure Guideline. Journal of the American College of Cardiology, 71, 109-118. https://doi.org/10.1016/j.jacc.2017.10.073
[58]	Kennel, W.B., Brand, N., Skinner, J.J., et al. (1967) The Relation of Adiposity to Blood Pressure and Development of Hypertension: The Framingham Study. Annals of Internal Medicine, 67, 48-59. https://doi.org/10.7326/0003-4819-67-1-48
[59]	Kotchen, T.A. (2010) Obesity-Related Hypertension: Epidemiology, Pathophysiology, and Clinical Management. American Journal of Hypertension, 23, 1170-1178. https://doi.org/10.1038/ajh.2010.172
[60]	Jiang, S.-Z., Lu, W., Zong, X.-F., et al. (2016) Obesity and Hypertension. Experimental and Therapeutic Medicine, 12, 2395-2399. https://doi.org/10.3892/etm.2016.3667
[61]	Loggi, M., Lombardi, M. and Caldarone, E. (2017) The Relationship between Obesity and Hypertension: An Comprehensive Overviews on Twins. Hypertension Research, 40, 947-963. https://doi.org/10.1038/hr.2017.75
[62]	The National Institute of Diabetes and Digestive and Kidney Diseases, Health Information Center (2014) Blood Pressure & Kidney Disease, What Is High Blood Pressure? https://www.niddk.nih.gov/health-information/kidney-disease/high-blood-pressure