Structure of the simulation model. The projection equations in each submodel are based on the first National Health and Nutrition Examination Survey Epidemiologic Followup Study (see “Methods” section). Estimates are weighted before summing to reflect the person’s share of the Third National Health and Nutrition Examination Survey population aged 25 to 74 years.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Russell LB, Valiyeva E, Carson JL. Effects of Prehypertension on Admissions and Deaths: A Simulation. Arch Intern Med. 2004;164(19):2119–2124. doi:10.1001/archinte.164.19.2119
The Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure recently released new clinical practice guidelines that target systolic blood pressure and identify persons with “prehypertension” (systolic blood pressure, 120-139 mm Hg), previously considered normal, as being at elevated risk and in need of intervention.
We used a simulation model, fitted to longitudinal data from the first NHANES (National Health and Nutrition Examination Survey) Epidemiologic Followup Study, to estimate the effects of prehypertension and residual hypertension (systolic blood pressure, ≥140 mm Hg). The term residual hypertension recognizes that many people with hypertension have lowered their pressures through treatment, but not to less than 140 mm Hg. We applied the model to a representative sample of US adults aged 25 to 74 years from NHANES III.
Except for women aged 25 to 44 years, more than a third of each age group in NHANES III had prehypertension. Approximately two thirds of persons aged 45 to 64 years and 80% of persons aged 65 to 74 years had prehypertension or residual hypertension. Together, prehypertension and residual hypertension accounted for 4.8% of hospital admissions per 10 000 adults aged 25 to 74 years, 9.9% of nursing home admissions, and 13.0% of deaths. Prehypertension alone accounted for 3.4% of hospitalizations, 6.2% of nursing home stays, and 8.5% of deaths. Numbers of events attributable to prehypertension are greatest for men aged 45 to 64 years and persons aged 65 to 74 years.
Our results confirm the substantial public health consequences of prehypertension. If prehypertension were eliminated, hospitalizations, nursing home admissions, and premature deaths could decline substantially.
On May 14, 2003, the National Heart, Lung, and Blood Institute released new clinical practice guidelines for high blood pressure. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) was published a week later in The Journal of the American Medical Association.1 The guidelines caught the attention of clinicians and the media with 2 changes in particular: they emphasized that systolic blood pressure (SBP), not diastolic BP, was the treatment target, and they defined a new category, “prehypertension,” which includes persons whose SBPs are 120 to 139 mm Hg; frank hypertension begins at 140 mm Hg. Systolic blood pressures in this range were previously considered normal. The new category includes 22% of American adults (approximately 45 million persons).2
Then director of the National Heart, Lung, and Blood Institute, Claude Lenfant, MD, explained that
[w]e also now know that damage to arteries begins at fairly low blood pressure levels—those formerly considered normal and optimal. In fact, studies show that the risk of death from heart disease and stroke begins to rise at blood pressures as low as 115 over 75, and . . . doubles for each 20 over 10 millimeters of mercury increase. So the harm starts long before people get treatment.2
Studies show that risk rises continuously as SBP increases and is independent of other risk factors.1
The harm from SBPs in this range causes hospitalizations, nursing home admissions, and premature deaths. What share of admissions and deaths can be attributed to BPs in the prehypertension range? To contribute to a fuller picture of its public health consequences, we present estimates of hospital and nursing home admissions and deaths attributable to prehypertension in NHANES III (third National Health and Nutrition Examination Survey) adults aged 25 to 74 years.
The NHANES, large national surveys conducted by the National Center for Health Statistics, collect information about risk factors and health outcomes for representative samples of Americans. Our estimates are based on NHANES I and III. Fielded in 1971 through 1975, NHANES I is unique in having a longitudinal follow-up study (NHANES I Epidemiologic Followup Study [NHEFS]) that tracked hospitalizations, nursing home admissions, and deaths of participants for 2 decades, allowing baseline risk factors to be linked to subsequent outcomes. The NHEFS is the basis for our simulation model. The simulation model is applied to adults from NHANES III, fielded in 1988 through 1994, because the prevalence of risk factors in NHANES III, for example, more obesity and fewer smokers, better represents today’s adults. We used the simulation model to estimate the effect of SBPs in the prehypertension range.
The model relates hospitalizations, nursing home admissions, and deaths to baseline risk factors and has 4 components (Figure): the cohort data set, which contains baseline information for NHANES III adults aged 25 to 74 years; the mortality submodel, which projects all-cause mortality; and the hospitalization and nursing home submodels, which project all-cause admissions. The submodels use separate regressions, estimated with NHEFS data, for 6 age-sex groups: men aged 25 to 44, 45 to 64, and 65 to 74 years and women aged 25 to 44, 45 to 64, and 65 to 74 years; the NHEFS excluded older persons.
The regressions relate hospitalizations, nursing home admissions, and mortality to all clinical risk factors that have been shown to be statistically significantly related to disease or death in multiple studies: age, race, smoking, SBP, overweight and underweight, laboratory test results (serum albumin, serum cholesterol), exercise, alcohol consumption, diet (fiber, fish/shellfish, and fruits/vegetables), and 9 groups of chronic conditions. These risk factors are more fully described elsewhere.3-5 Thus, the estimates of the effects of SBP are adjusted for other risk factors.
Earlier published estimates4-6 from the model used regressions fitted to the NHEFS through 1987 and were validated by in-sample and out-of-sample tests. For this study, the model was updated with data for 1988 through 1992, the final follow-up. Systolic BP is a statistically significant determinant of all 3 outcomes.3,4,7
The cohort data set includes all NHANES III respondents aged 25 to 74 years who were examined by a physician (N = 12 841). Because 85% of respondents had complete information, 8% were missing data for only 1 risk factor, and only 24 individuals (0.19%) lacked data for more than 5 risk factors, we retained all sample persons and replaced missing values with the mean of that risk factor for the age-sex group. The NHANES I/NHEFS measured SBP only once. Thus, to match the projection equations, we used the first measurement of SBP in NHANES III.
The mortality and hospital submodels are described in detail elsewhere.4,5 (The regressions are available on request from the authors.) The mortality submodel uses Weibull hazard functions to relate survival to the baseline risk factors. The hospital and nursing home submodels use negative binomial regressions to relate annual admissions to the same risk factors. The admissions submodels first estimate hospital (or nursing home) admissions for each individual and then adjust the estimates for the individual’s probability of survival to the year in question (from the mortality submodel), as indicated in the Figure. Each sample person represents many people with similar characteristics in the US population, and this calculation is equivalent to estimating admissions for members of the group who are still alive in the specified year.
The NHANES III was a stratified cluster sample. We weighted the estimates for each participant to reflect his or her share of the 1988-1994 US population of noninstitutionalized adults. Of several statistical weights available for the NHANES III, we used the “examined sample final weight,” because all individuals in our cohort underwent medical examination.
We calculated baseline estimates of hospitalizations, nursing home admissions, and deaths by entering observed baseline risk factors for the 12 841 adults in the NHANES III cohort into the model regressions and running the model. Attributable-risk estimates were then calculated for 2 scenarios: (1) all persons with SBPs of 140 mm Hg or higher had their pressures reduced to 139 mm Hg (residual hypertension eliminated), and (2) all persons with SBPs of 120 mm Hg or higher had their pressures reduced to 119 mm Hg (residual hypertension and prehypertension eliminated). The term residual hypertension recognizes that many people with hypertension have lowered their pressures through treatment, although they have not achieved SBPs of less than 140 mm Hg. To estimate scenario 1, we identified all adults in the NHANES III cohort with recorded SBPs of 140 mm Hg or higher and set their pressures at 139 mm Hg. The model was run with these new values for SBP; all other risk factors remained at observed baseline levels for these individuals, and all risk factors, including BP, remained at baseline levels for individuals with observed SBPs of less than 140 mm Hg. The procedure was repeated for scenario 2 by identifying all adults with SBPs of 120 mm Hg or higher and setting their pressures at 119 mm Hg. The differences between the baseline and attributable-risk estimates represent hospital admissions, nursing home admissions, and deaths attributable to residual hypertension (baseline minus scenario 1) and to residual hypertension and prehypertension combined (baseline minus scenario 2). The differences between scenarios 1 and 2 represent events attributable to prehypertension alone.
Table 1 gives the percentages of NHANES III adults with SBPs of 140 mm Hg or higher (residual hypertension), 120 to 139 mm Hg (prehypertension), and 120 mm Hg or higher (residual hypertension and prehypertension), by age and sex. The group with residual hypertension, which includes people treated for hypertension but whose SBPs remained at 140 mm Hg or greater, accounts for a small percentage of persons aged 25 to 44 years, about a fifth of persons aged 45 to 64 years, and more than 40% of persons aged 65 to 74 years. The percentages with prehypertension are large—more than a third of every age-sex group except women aged 25 to 44 years. Together, persons with residual hypertension and prehypertension account for two thirds of persons aged 45 to 64 years and 80% of persons aged 65 to 74 years.
Table 2 and Table 3 give hospital admissions, nursing home admissions, and deaths per 10 000 persons in the first year of the projection period for baseline BP levels (“baseline”) and after the elimination of residual hypertension and prehypertension. Table 2 includes percentage reductions from baseline, and Table 3 includes absolute reductions in levels.
If all NHANES III adults had SBPs of less than 140 mm Hg, hospital admissions per 10 000 persons would be 1.4% lower; nursing home admissions, 3.7% lower; and deaths, 4.5% lower in the first year (Table 2). By the fifth year of the scenario, in which all adults have their SBPs reduced to less than 140 mm Hg (data not shown), the reductions would be slightly less: hospital admissions per 10 000 persons would be 1.2% lower; nursing home admissions, 2.6% lower; and deaths, 3.9% lower.
Women aged 65 to 74 years benefit most from elimination of residual hypertension. Their hospital admission rate declines 3.6% (Table 2), or by 70 admissions per 10 000 persons (Table 3). Their percentage reductions in nursing home admissions and deaths are not the largest, but their higher event rates mean, again, that more events are avoided—approximately 1.6 nursing home admission and 3.8 deaths per 10 000 persons.
Eliminating both prehypertension and residual hypertension would require further reductions in the SBPs of hypertensive persons and reductions in the SBPs of persons with prehypertension. If all NHANES III adults had SBPs of less than 120 mm Hg, hospital admissions per 10 000 persons would be 4.8% lower; nursing home admissions, 9.9% lower; and deaths, 13.0% lower (Table 2). Again, reductions after 5 years are similar but slightly smaller (data not shown).
The gains from eliminating prehypertension and residual hypertension, compared with eliminating residual hypertension alone, are substantial. Hospitalizations per 10 000 persons decline 4.8%, compared with 1.4% for residual hypertension alone; the difference of 3.4% is the additional gain from eliminating prehypertension (its attributable risk). For nursing home admissions and deaths, the gains from eliminating prehypertension are additional declines of 6.2% and 8.5%, respectively.
Table 3 indicates that numbers of events avoided are greatest for women aged 65 to 74 years. Their hospital admissions decline by 186 per 10 000 persons compared with baseline, an additional decline of 116 admissions per 10 000 persons compared with eliminating residual hypertension alone. Eliminating prehypertension also avoids 2.7 more nursing home admissions and almost 5.7 more deaths per 10 000 persons. Men aged 45 to 64 and 65 to 74 years show substantial reductions as well. For example, when prehypertension is eliminated, hospital admissions decline another 75 admissions per 10 000 men aged 45 to 64 years and 47 admissions per 10 000 men aged 65 to 74 years compared with eliminating only residual hypertension.
The authors of the JNC 71 state that “the ultimate public health goal of antihypertensive therapy is the reduction of cardiovascular and renal morbidity and mortality.” Our estimates show that if both residual hypertension and prehypertension could be eliminated, hospital admissions per 10 000 persons would be 4.8% lower, nursing home admissions per 10 000 persons would be 9.9% lower, and deaths per 10 000 persons would be 13.0% lower. The effect of prehypertension is substantial: it accounts for 3.4, 6.2, and 8.5 percentage points, respectively, of these totals.
Our estimates are for adults aged 25 to 74 years in NHANES III, a large, nationally representative survey conducted between 1988 and 1994 that provided data for the JNC 7’s deliberations.1 In the editorial that accompanied the guidelines, Kottke et al8 observe that hypertension awareness and control have not changed in the past decade (although a study9 published more recently shows modest improvement in control). Thus, current conditions are similar to those in NHANES III.
The new prehypertension category defined by the JNC 7 (SBP, 120-139 mm Hg) means that many adults who were previously told that their pressures were normal will now be told that they are at heightened risk of cardiovascular disease. In each age-sex group considered herein, except women aged 25 to 44 years, more than one third fall in the prehypertension range. Together with those classified as hypertensive (SBP, ≥140 mm Hg), almost two thirds of men and women aged 45 to 64 years and 80% of men and women aged 65 to 74 years are considered to be at risk.
The JNC 7 emphasizes the risk to the elderly, noting not only that the prevalence of hypertension is highest in this group but also that rates of control are lowest.1 Our estimates show that the potential gains from eliminating residual hypertension and prehypertension in the elderly are large. Although their percentage reductions are sometimes lower than those for younger groups, numbers of events avoided are the same or higher because of their high event rates. For example, the 9.6% reduction in hospitalizations per 10 000 persons for women aged 65 to 74 years represents 186 fewer admissions per 10 000 persons, 116 of them attributable to prehypertension. Numbers of events avoided are greatest for women and men in this age group and for men aged 45 to 64 years.
The goal of treatment of hypertensive persons is to reduce their BP to 140/90 mm Hg or, if they have diabetes mellitus or kidney disease, to 130/80 mm Hg; most patients need to take 2 or more antihypertensive medications to achieve their goal.1 The JNC 7 recommended lifestyle changes for those with prehypertension but did not specify a treatment target. Although 120 mm Hg is the lower boundary for the category, there is no recommendation that prehypertensive persons be treated to achieve SBPs of less than 120 mm Hg. Thus, even if the new guidelines are enthusiastically received, they are unlikely to yield gains of the magnitude we estimated. Our estimates serve, however, to confirm that prehypertension is an important health problem.
Several limitations of our model should be noted. The diagnosis of hypertension or prehypertension in individuals is based on the average of 2 or more readings.1 Our model is based on a single BP reading because only 1 was taken in the NHEFS, which is the basis for the projection equations. Our estimates of NHANES III adults with residual hypertension and prehypertension are thus higher than if the determinations were based on multiple readings. Nonetheless, our estimates for NHANES III adults of the outcomes attributable to residual hypertension and prehypertension are accurate, because our estimating equations, based on the NHEFS, self-correct for regression to the mean. The NHEFS followed up individuals for 2 decades and recorded their hospitalizations, nursing home admissions, and deaths. Individuals whose single baseline BPs were unusually high would have outcomes during follow-up in line with their true lower pressures. The equations linking outcomes to baseline BPs thus correct for regression to the mean, because persons whose BPs were overstated by a single reading contributed fewer events during follow-up than those with genuinely elevated BPs. When the equations were applied to the single BP readings for NHANES III adults, they again estimated outcomes corrected for regression to the mean.
We limited our projections to the first and fifth years of the projection period. Thus, the projections are primarily based on the first 5 years of the NHEFS and should model the natural history of hypertension/prehypertension well. In the 1970s, the efficacy of drug treatment for hypertension was being documented. Not until 1980 did the JNC publish its first recommendations advocating treatment for people with moderate or severe hypertension, and not until 1984 did it extend its recommendations to people with SBPs as low as 140 mm Hg.
Nonetheless, our estimates understate the gains from controlling hypertension, because by 1988-1994, when NHANES III was conducted, hypertension was often treated aggressively, so SBPs observed in NHANES III were lower than they would have been naturally. Thus, although the projection equations reflect the natural history of hypertension, the population to which they are applied, NHANES III adults, included many individuals whose SBPs had already been lowered through treatment. For that reason, we referred to these estimates as showing the effects of residual hypertension.
This limitation does not apply to individuals with prehypertension, who, until the JNC 7, were considered normal and were not treated. Thus, our estimates for prehypertension reflect the full effects of this condition—its attributable risk. However, to fully validate the concept of prehypertension, it will be necessary to demonstrate through clinical trials that treatment with diet or medication reduces mortality, morbidity, or hospitalizations. Very large trials are needed to establish efficacy of treatment for prehypertension and to determine the extent to which treatment can reduce risk.
Finally, national trends during the NHEFS suggested that our hospitalization estimates might be high. Hospitalization rates measured by the National Hospital Discharge Survey (NHDS) increased between 1970 and the early 1980s, then fell until the early 1990s; they continued to decline through 2000 for persons aged 45 to 64 years, but they rose again for persons 65 years or older.10 The NHDS rates should exceed those for NHANES III (and NHANES I) adults, because NHANES participants were drawn from the noninstitutionalized population. In addition, because they were able to travel to an examination site, they may have been even healthier than the general noninstitutionalized population. To check that our estimates are reasonable, we compared year 1 hospitalization rates with 1990-2000 NHDS rates for the same age-sex groups. As they should be, our estimates were uniformly lower (60%-90% of NHDS rates).
By introducing the category of prehypertension, the JNC 71 has focused attention on lifestyle and its effects. The report1 notes that
lifestyle modifications . . . [are] . . . critical for the prevention of high BP and an indispensable part of the management of those with hypertension.
The accompanying editorial8 states
The JNC 7 report is about more than hypertension. For many, high blood pressure is just one manifestation of what may be termed the lifestyle syndrome, which is a cluster of conditions and diseases that result from . . . too many calories; ingesting too much saturated fat, sodium, and alcohol; not expending enough calories; and using tobacco or being exposed to tobacco smoke.
The reductions in hospitalizations, nursing home admissions, and deaths that might result from eliminating prehypertension and residual hypertension are thus only part of the potential impact of lifestyle change. The total effect could be much larger.
Our estimates show that if means can be found to prevent prehypertension in addition to hypertension, the reductions in hospital and nursing home admissions and deaths would be substantial. Residual hypertension and prehypertension together accounted for 4.8% of hospital admissions per 10 000 US adults aged 25 to 74 years, 9.9% of nursing home admissions, and 13.0% of deaths. In each case, prehypertension by itself was responsible for approximately two thirds of the excess admissions and deaths.
Correspondence: Louise B. Russell, PhD, Institute for Health, Health Care Policy, and Aging Research, Rutgers University, 30 College Ave, New Brunswick, NJ 08901 (firstname.lastname@example.org).
Accepted for Publication: February 27, 2004.
Financial Disclosure: None.
Funding/Support: Development of the simulation model used for this study was supported in part by grants HS07002 and HS11477 from the Agency for Healthcare Research and Quality, Rockville, Md.
This article was corrected on August 8, 2005.