Visit-to-Visit Variability of Blood Pressure Is Associated With Hospitalization and Mortality in an Unselected Adult Population

Visit-to-Visit Variability of Blood Pressure Is Associated With Hospitalization and Mortality in... Abstract BACKGROUND Blood pressure variability (BPV) has been associated with poor health outcomes in high-risk patients, but its association with more general populations is poorly understood. METHODS We analyzed outcomes from 240,622 otherwise unselected patients who had 10 or more outpatient blood pressure readings recorded over a 3-year period and were aged from 20 to 100 years. RESULTS Whether calculated as SD, average change, or greatest change and systolic or diastolic blood pressure, we found that higher outpatient BPV was associated with subsequent hospitalization and mortality. Systolic pressure average change exceeding 10–12 mm Hg or diastolic exceeding 8 mm Hg significantly increased risk of hospitalization and death (odds ratios [ORs] from 2.0 to 4.5). Variability in the highest decile increased risks even more dramatically, with propensity-matched ORs from 4.4 to 42. A systolic change exceeding 35 mm Hg increased the relative risk of death 4.5-fold. Similarly, a diastolic change greater than 23–24 mm Hg almost tripled the risks of hospitalization and death. Neither stratification for hypertension nor propensity matching for risk factors within the database affected these associations. CONCLUSIONS Systolic and diastolic variabilities were each associated with subsequent adverse outcomes. Physicians should pay special attention to patients with swings in blood pressure between clinic visits. Electronic medical records should flag such variability. blood pressure, hospitalization, hypertension, mortality, variability, variance Blood pressure variability (BPV) within 1 day was associated with increased risk for hypertensive target-organ damage decades ago.1,2 However, there has been increasing interest in chronic BPV as a risk for subsequent complications, spurred by computerized data from large clinical trials that facilitate such calculations.3–7 In various high-risk groups, BPV, generally defined as either the SD or coefficient of variation of sphygmomanometric readings over time, has been reported to increase risks of stroke,3,5 cardiovascular mortality,5,8 and diabetic vascular complications like renal impairment,9 although such associations are not always found.4,6,8,10–12 Such studies frequently rely upon retrospective analyses of trials in specifically defined populations. However, clinical trial subjects may differ from patients at large because of inclusion and exclusion criteria. Indeed, some hypertension trials specifically exclude episodic hypertension. Furthermore, while cardiovascular events and death have been examined in this connection, less attention has been paid to the all-cause risk of hospitalization, increasingly important as risk shifts from third-party payers to health care organizations. We tested the hypothesis that greater BPV is associated with increased risk of all-cause hospitalization and mortality among unselected patients from a large multihospital regional health care system. We defined BPV as the common SD as well as by 2 new techniques: average change and largest change. We calculated systolic and diastolic BPV in patients over age 20 years with at least 10 readings. During final data analysis, the American Heart Association redefined systolic pressures above 130 mm Hg as hypertensive and detailed procedures for blood pressure measurement.13 We used the previous guidelines here since these represented how patients’ blood pressures were measured and how patients were treated during the study period. Risks due to variability were tested for various subpopulations. METHODS Data An aggregate data file extracted information from records for all patients using the Sanford Health system from December 2013 to November 2016 (n = 1,143,028). Sanford Health is an integrated health system, the largest rural, not-for-profit health care system in the United States with 45 hospitals and 289 clinics. Blood pressures for 1,029,899 patients were available (90.1%); 240,622 had 10 or more readings during at least 6 months, no missing values, and were aged 20 to 100 years, fitting study criteria. Two binary outcomes were recorded: whether patients had on average at least 1 hospitalization per year (N = 35,587, 14.8%) and whether patients died (N = 9,556, 4.0%). Patients were categorized by amount and type of abnormal blood pressure readings (Supplementary Appendix A). Of the 240,622 patients studied, 41% had been diagnosed as hypertensive. Abnormal pressures were also measured using percent of a patient’s readings above or below specific values. The “urgent high” category included 6%, for whom ≥15% of systolic pressures exceeded 160 or diastolics exceeded 100. The “caution high” category contained 24%, and 70% were “normal,” with <15% abnormally high readings. For abnormal low pressures, 4% were “urgently low,” 28% were in “caution low,” and 68% were “normal,” with <15% abnormally low readings. Six variables regarding BPV were created for tests of association with the two events. We calculated the SD (SDV), the average change (ACV) among all of each patient’s readings (the values of each change from reading to reading were averaged over the number of readings), and the largest change (LCV) recorded between any two consecutive readings (regardless of time duration between readings). As these calculations change with the number of readings used, we evaluated how the means and SDs of the BPVs changed when we varied the inclusion requirement from 5 to 30 readings/patient over the 3-year period. At least 10 readings achieved stability (Supplementary Appendix B). Most patients (95%) had a year or more between the 10 readings. Propensity analysis Supplementary Appendix C details the propensity analysis. Patients were categorized as using or not using antihypertensives (acetylcholinesterase [ACE] inhibitors, beta blockers, or other cardiac medication). 19.4% used antihypertensives, and 14.4 % used cardiac medication other than ACE inhibitors or beta blockers. Approximately, 10% used ACE inhibitors, and 9% used beta blockers. We excluded medications used by 10 or fewer patients from analysis for statistical rigor. Age, sex, race, rural status, smoking, alcohol use, primary care visits, use of a patient-centered electronic communication portal, emergency department use, and frequency of blood pressure readings were correlated with outcomes and antihypertensive use, controlling for potential confounders. Descriptive statistics (Supplementary Appendix C1) were entered into a propensity analysis predicting antihypertensive use, creating an inverse weight that was applied in statistical analysis to control for these variables. Statistical analysis Average BPV measures of SDV, ACV, and LCV were compared for all patients, hypertensive, and nonhypertensive patients, patients with normal, caution, or urgent low pressure, and patients with normal, caution, or urgent high pressure using independent t-tests and one-way ANOVA. How ACV and LCV changed between different patient groups was compared with how SDV changed. Inverse weights from the propensity analysis were used in logistic regressions predicting hospitalization and death from each BPV measure while controlling for medication use and other variables. Odds ratios (ORs) from ACV and LCV models were compared with ORs from SDV models. Sensitivity analysis We reanalyzed subgroups to test whether their results differed. Logistic regression to identify the risk of hospitalization or death for patients with the highest 10% of BPV relative to those with the lowest 10% of that BPV. We also tested subgroups based on blood pressure values (extreme highs and lows or hypertension, Supplementary Table A1) or on numbers of pressures recorded (assuming those with higher numbers were hospitalized or sicker). Different cutoffs for the measures of change and how they affected likelihood of hospitalization or death were analyzed using receiver operating characteristic curves. Additional statistics included relative risk (RR), population attributable risk, sensitivity, specificity, positive predictive value, negative predictive value, population impact number, exposure impact number, and case impact number. SAS 9.4 and 14.2 (2017, SAS Institute, Cary, North Carolina) and NCSS 11 Statistical Software (2016, NCSS, Kaysville, Utah, ncss.com/software/ncss) were used. RESULTS Two hundred forty thousand, six hundred twenty-two patients aged 20 to 100 years had 10 or more blood pressure readings for at least 6 months during the 3-year period. Their average age was 54.9 years (SD 19.7), 63.6% were female, and 93.0% were white. Categories based on abnormal blood pressures are defined in Supplementary Table A1 of Supplementary Appendix A. Out of the entire sample, 98,335 (40.9%) had hypertension diagnoses. Based upon all patients’ recorded blood pressure readings, we retrospectively categorized 15,193 (6.3%) as having urgent high blood pressure and 57,306 (23.8%) as caution abnormal high blood pressure. Similarly, we retrospectively categorized 1,024 (4.2%) and 67,155 (27.9%) as having urgently low blood pressures and caution abnormal low blood pressure, respectively. Of the 240,622 patients, 19.3% took antihypertensive medications (Supplementary Appendix B); 35,587 patients (14.8%) had on average 1 hospitalization per year; 9,556 (4.0%) were deceased. We examined 3 measures each of systolic and diastolic variability, comparing SDV with ACV and LCV. (Table 1) All 6 measures had distributions with a minor positive skew. Their means were compared between categories of blood pressure extremes (Supplementary Table A1). Patients with higher severity (hypertension or abnormal blood pressures) had significantly higher BPV measures (all P < 0.001). SDV and ACV values were nearly identical; LCV values were higher as they were based on largest changes. Diagnosed hypertensives had average systolic or diastolic SDV and ACV about 3 points or 1 point higher than nonhypertensives. The increase in LCV was about 9 or 3 points. Patients with abnormal high pressures showed a similar pattern with systolic BPV, but diastolic BPV had less decrease from urgent to caution categories. Systolic SDV decreased by nearly half, from 18.1 to 10.0, between urgent and not caution or urgent. Diastolic SDV decreased from 10.3 to 7.4. Patients with abnormally low blood pressure exhibited smaller changes of only a few points, similar to hypertension. Table 1. Measures of blood pressure variability for 240,622 patients All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability. View Large Table 1. Measures of blood pressure variability for 240,622 patients All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability. View Large Logistic regressions for SDV, ACV, and LCV tested how BPVs differed in association with hospitalization or death (Table 2), controlling for confounding variables using propensity weights (Supplementary Appendix B). The unadjusted hospitalization risk from BPV varied from OR = 1.03 (systolic LCV) to OR = 1.14 (diastolic ACV). Adjusted ORs slightly increased for all BPVs (1.04 to 1.21) except for diastolic SDV. The mortality risk was increased from hospitalization for all BPV, both unadjusted (OR ranged from 1.05 to 1.24) and adjusted (OR ranged from 1.04 to 1.25). Overall, the risk from ACV was nearly identical to the SDV, while the risk from LCV was 10 to 15% lower. Table 2. Unadjusted and propensity-adjusted ORs of blood pressure variance measures predicting hospitalization and death from logistic regressions Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Propensity scores controlled for demographics and health care of patients. See Supplementary Appendix C for details. Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Table 2. Unadjusted and propensity-adjusted ORs of blood pressure variance measures predicting hospitalization and death from logistic regressions Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Propensity scores controlled for demographics and health care of patients. See Supplementary Appendix C for details. Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Sensitivity of measures Patients in the highest 10%ile of BPVs were compared with those with the lowest 10%ile of the same BPV. Being in the highest, 10% greatly increased event risks. (Table 3) ACV showed the smallest increase with ORs of 4.9 systolic and 4.4 diastolic for hospitalization. SDV and LCV were near ORs of 7. Death risk increased dramatically for the top 10%. The OR for death for an SDV in the top 10% vs. the bottom 10% was 42. The smallest ORs were for diastolic measures with LCV OR = 8.0. Table 3. Risk of hospitalization and death in patients in the upper 10th percentile relative to those in the lower 10th percentile of BPV measures Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 ORs were from logistic regressions adjusting for demographics, health care, medication, and blood pressure. N ranged from 48,126 (diastolic SDV) to 53,006 (systolic largest change). Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Table 3. Risk of hospitalization and death in patients in the upper 10th percentile relative to those in the lower 10th percentile of BPV measures Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 ORs were from logistic regressions adjusting for demographics, health care, medication, and blood pressure. N ranged from 48,126 (diastolic SDV) to 53,006 (systolic largest change). Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Logistic regressions were repeated for various subpopulations (Supplementary Table A1) to examine the robustness and generalizability of our findings. (Supplementary Table D1). All ORs remained statistically significant (P < 0.05), and patterns of high and low findings were similar. (As with unadjusted and propensity-adjusted ORs, SDV and ACV ORs were similar (OR approximately 1.1) and LCV ORs were lower) We further excluded patient/dates with 3 or more consecutive pressures from the data set, assuming these were likely hospitalization days. Excluding such presumed inpatient readings did not change results from Tables 1 and 2 significantly (P > 0.05). In addition, systolic and diastolic SDV were similarly associated with outcomes in patients <50, 50–65, and >65 (not shown). Receiver operating characteristic analysis assessed optimal cutoff values of ACV and LCV. Supplementary Figures D1a and D1b showed few differences between SDV and ACV or LCV. The AUCs for hospitalization ranged from 63.5 to 68.3% and death from 67.1 to 74.2%. Cutoff values were chosen for slightly higher sensitivities from 61 to 72%. Using those cutoffs, RRs for hospitalization doubled or more the increase in risk, with LCVs more than 31 having an RR = 2.50 or higher (Table 4). A systolic change exceeding 35 mm Hg increased the RR of death 4.5-fold. Similarly, a diastolic change greater than 23–24 mm Hg almost tripled the risks of hospitalization and death. Supplementary Appendix D shows other statistics based on the cutoff values. Table 4. Sensitivity analysis of blood pressure variance cutoffs RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR: relative risk; there is a (1 − RR)% increase in the risk of outcome (hospitalization or death) due to BP change. PAR%: population attributable risk; among the population, PAR% of the total risk of event is due to BP change. Sens: sensitivity; ability to detect outcome when change is present; % of those with event who had BP change. Spec: specificity; ability to detect nonoutcome when no change; % of those with no event who had no BP change. PPV: positive predictive value; % of those with BP change who had outcome. NPV: negative predictive value; % of those with no BP change who had no outcome. PIN: population impact number; number in population whose outcome is attributable to BP change; for every PIN people, one outcome is attributable to BP change. EIN: exposure impact number; number of people with change where one with an outcome is due to BP change, for every EIN people with BP change, there is one with the outcome. CIN: case impact number; number of people with outcome where one outcome is from BP change, for every CIN people with outcomes, there is one attributable to BP change. Abbreviations: ACV, average change variability; LCV, largest change variability. View Large Table 4. Sensitivity analysis of blood pressure variance cutoffs RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR: relative risk; there is a (1 − RR)% increase in the risk of outcome (hospitalization or death) due to BP change. PAR%: population attributable risk; among the population, PAR% of the total risk of event is due to BP change. Sens: sensitivity; ability to detect outcome when change is present; % of those with event who had BP change. Spec: specificity; ability to detect nonoutcome when no change; % of those with no event who had no BP change. PPV: positive predictive value; % of those with BP change who had outcome. NPV: negative predictive value; % of those with no BP change who had no outcome. PIN: population impact number; number in population whose outcome is attributable to BP change; for every PIN people, one outcome is attributable to BP change. EIN: exposure impact number; number of people with change where one with an outcome is due to BP change, for every EIN people with BP change, there is one with the outcome. CIN: case impact number; number of people with outcome where one outcome is from BP change, for every CIN people with outcomes, there is one attributable to BP change. Abbreviations: ACV, average change variability; LCV, largest change variability. View Large DISCUSSION Even modestly heightened BPV posed a substantial risk for hospitalization and mortality in this general adult population. Systolic variance exceeding 10–12 mm Hg or diastolic variance exceeding 8 mm Hg notably worsened outcomes. Therefore, clinicians should not disregard apparently isolated elevated or low pressures despite other normal readings. BPV can be calculated from systolic, diastolic, or mean pressures, using coefficients of variation, SD, or variance. Systolic and diastolic variability yielded similar results. Others have emphasized blood pressure SD(5,14–16), which predicts similarly to the coefficient of the mean.17 Because SDs are not intuitively obvious, indeed, average and largest change were easier and approximately as effective as SD in predicting hospitalization and death after controlling for confounders. “Largest change” might be easiest for clinicians to assess from lists of measurements. Independent of the absolute magnitude of the patient’s blood pressure, specific changes showed an increased risk of 100 to 250% (Table 4). One might hypothesize that BPV simply reflects poorly controlled hypertension because of severity or noncompliance. Indeed, most previous studies have examined high-risk patients, who are generally hypertensive. However, BPV was associated with more frequent cardiovascular events even in well-controlled hypertensives.18 Moreover, medication noncompliance explains only a small proportion of BPV.19 Other variability may signal autonomic instability that could precede development or worsening of vascular wall abnormalities,14,15,20 diabetes21 or brain lesions and/or dementia,22,23 or other even less well-understood disorders of homeostasis. Age, female gender, smoking, systolic or diastolic hypertension, peripheral vascular disease, diabetes, renal insufficiency, heart rate variability, widened pulse pressure, and angiotensin-converting enzyme therapy have each been correlated with BPV.16,24,25 Conversely, chlorthalidone or amlodipine may be associated with decreased BPV among treated hypertensives.25,26 We were unable to analyze all these factors in our administrative data, but propensity matching including (separately) use of ACE inhibitors, beta blockers, any blood pressure medication, and any cardiac medication did not invalidate our results.27 A recent meta-analysis suggested that BPV should be monitored as a prognostic indicator for mortality and cardiovascular complications in high-risk individuals.25 Our administrative data set was insufficiently granular to distinguish causes of hospitalization and mortality, but retrospective analyses of clinical trials in high-risk individuals suggest that the associated morbidity includes,5 but may not be limited to,(7,22) cardiovascular morbidity. Our results, taken together with these other observations, suggest that this is also likely true in normotensive individuals. Our results suggest that BPV is similarly predictive whether or not patients are diagnosed as hypertensive, take antihypertensives, or are actually hypertensive or hypotensive. Gosmanova28 associated systolic BPV with all-cause mortality in veterans who were not necessarily hypertensive, although likely many were. Some patients may have masked hypertension missed by office-based measurements. Whether “normotensive” patients with high BPV are more likely to have such masked hypertension awaits study, just as normotensive patients with “white coat syndrome” may exhibit elevated pressures from anxiety. However, since clinicians generally assess blood pressures in this fashion, this would not invalidate our suggestion that in-office normotensive patients with high in-office BPV should be considered high risk. Since we required at least 10 pressure readings over the 3-year study period to meaningfully calculate variability, different results might be obtained in people who do not require this many health care encounters. Gosmanova28 similarly required at least 8 measurements over 2 years. While we achieved similar results when we required 15 or 20 blood pressure readings to calculate BPV more precisely (not shown), decreasing the number of blood pressure readings to 5 substantially decreases the mathematical reliability of the calculations. (Supplementary Appendix A) However, Muntner24 associated systolic BPV and all-cause mortality in the NHANES study using only 3 measurements over 6 years, despite the increased statistical noise created by having fewer measurements. Only 33 patients had all their blood pressures within 30 days. Most (95%) took more than a year, and 79% took 2 or more years to produce 10 readings. Clustering of blood pressures within a short term over an acute event therefore seems unlikely to have substantially biased results. Operational purposes require a cutoff or “abnormal” value to trigger clinicians’ attention. These cutoffs trade sensitivity against specificity (Supplementary Figure D1), based upon system characteristics and resources and the trade-off between alarming patients unnecessarily and failing to warn patients appropriately. Like other scarce resources,29 physician time must be allocated on ethical as well as pragmatic grounds. However, choosing 75 and 50% as endpoints for sensitivity and specificity seems reasonable and yields population impact numbers of approximately 12 for hospitalization and 30–40 for death, which seems deserving of attention. The magnitude of the BPV effect resembles that of abnormal cholesterol levels, to which attention is already standard of care.30 It might seem counterintuitive that such small changes in BPV predict clinically significant differences in outcomes, but we similarly now understand that even a 2 mm Hg reduction in systolic pressure meaningfully reduces cardiovascular mortality.31 Data availability created limitations. Administrative data could not be validated by individual chart review, but there is no reason to assume that errors in the data would distribute differently across low- or high-BPV patients. We obtained time of blood pressure but not time of hospitalization or death, making any proportional hazard model impossible. Because we could not obtain records after 2 November 2016, we could not capture events after this time. Limiting the study to people with at least 6 months of blood pressure data removed any censoring at the beginning, such as excluding a person who died 1 month into the study. Further censoring would have required excluding anyone who had no event during the study. Including these patients may have introduced some time bias but offered a baseline group with no events in the study time frame. Furthermore, because we lacked time of event, we used a cross-sectional type approach to gathering BPV data both before and after an event (at any time in the 3-year period). Further research should incorporate time of event and exposure to adjust for changes over time and could use a proportional hazard model. Cases and controls could also be used in future analyses to estimate these results for specific populations, such as specific age groups or genders. In addition, although we accounted for medication use in our propensity matching, we lacked data regarding adherence to medication, as well as conventional measures of baseline cardiac risk. Primary care physicians encountering high BPV might seek such information about their patients, while future studies might investigate these variables’ effects on the associations we define here. In conclusion, BPV is a calculated variable not intuitively obvious to the clinician. These findings suggest an opportunity to use the electronic medical record to alert clinicians to an important but largely ignored risk factor. Clinicians should scan blood pressure listings over time for the largest consecutive changes, paying particular attention to patients with changes exceeding 34 systolic or 23 diastolic. Administrators of hospital systems should consider creating a computer-calculated variable, whether blood pressure SD or average change, that can be flagged like a calculated obese BMI. For instance, a systolic average change exceeding 12 or diastolic exceeding 8 merits attention. Although further studies are required to demonstrate the efficacy of targeting specific interventions to patients with high BPV, it would not seem unreasonable for clinicians encountering patients with high BPV taking ACE inhibitors to balance the known benefits of ACE inhibitors in diabetes and congestive heart failure against their potential adverse effects on BPV. Obesity and smoking increase BPV in some32 but not all33 studies and even if not are likely to be synergistic in effects. Medication adherence may also diminish BPV.34 Antihypertensive trials should address BPV in addition to blood pressure magnitude. BPV needs further study but is promising for the care of patients. SUPPLEMENTARY DATA Supplementary data are available at American Journal of Hypertension online. DISCLOSURE All authors participated in this research and in preparation of the manuscript. The authors declared no conflict of interest. ACKNOWLEDGMENTS We would like to acknowledge the generous access to data provided by the Sanford Health system under a Sanford research grant program, and the support and encouragement of David Pearce, Ph.D., Executive Vice President, Sanford Research. REFERENCES 1. Frattola A , Parati G , Cuspidi C , Albini F , Mancia G . Prognostic value of 24-hour blood pressure variability . J Hypertens 1993 ; 11 : 1133 – 1137 . Google Scholar Crossref Search ADS PubMed 2. Parati G , Pomidossi G , Albini F , Malaspina D , Mancia G . Relationship of 24-hour blood pressure mean and variability to severity of target-organ damage in hypertension . J Hypertens 1987 ; 5 : 93 – 98 . Google Scholar Crossref Search ADS PubMed 3. Dai H , Lu Y , Song L , Tang X , Li Y , Chen R , Luo A , Yuan H , Wu S . Visit-to-visit variability of blood pressure and risk of stroke: results of the Kailuan cohort study . Sci Rep 2017 ; 7 : 285 . Google Scholar Crossref Search ADS PubMed 4. Suchy-Dicey AM , Wallace ER , Mitchell SV , Aguilar M , Gottesman RF , Rice K , Kronmal R , Psaty BM , Longstreth WT , Jr . Blood pressure variability and the risk of all-cause mortality, incident myocardial infarction, and incident stroke in the cardiovascular health study . Am J Hypertens 2013 ; 26 : 1210 – 1217 . Google Scholar Crossref Search ADS PubMed 5. Vidal-Petiot E , Stebbins A , Chiswell K , Ardissino D , Aylward PE , Cannon CP , Ramos Corrales MA , Held C , López-Sendón JL , Stewart RAH , Wallentin L , White HD , Steg PG ; STABILITY Investigators . Visit-to-visit variability of blood pressure and cardiovascular outcomes in patients with stable coronary heart disease. Insights from the STABILITY trial . Eur Heart J 2017 ; 38 : 2813 – 2822 . Google Scholar Crossref Search ADS PubMed 6. Tully PJ , Debette S , Dartigues JF , Helmer C , Artero S , Tzourio C . Antihypertensive drug use, blood pressure variability, and incident stroke risk in older adults: three-city cohort study . Stroke 2016 ; 47 : 1194 – 1200 . Google Scholar Crossref Search ADS PubMed 7. Goyal A , Mezue K , Rangaswami J . Visit-to-visit systolic blood pressure variability predicts treatment-related adverse event of hyponatremia in SPRINT . Cardiovasc Ther 2017 ; 35 . doi:10.1111/1755-5922.12274 8. Pringle E , Phillips C , Thijs L , Davidson C , Staessen JA , de Leeuw PW , Jaaskivi M , Nachev C , Parati G , O’Brien ET , Tuomilehto J , Webster J , Bulpitt CJ , Fagard RH ; Syst-Eur investigators . Systolic blood pressure variability as a risk factor for stroke and cardiovascular mortality in the elderly hypertensive population . J Hypertens 2003 ; 21 : 2251 – 2257 . Google Scholar Crossref Search ADS PubMed 9. Yeh CH , Yu HC , Huang TY , Huang PF , Wang YC , Chen TP , Yin SY . The risk of diabetic renal function impairment in the first decade after diagnosed of diabetes mellitus is correlated with high variability of visit-to-visit systolic and diastolic blood pressure: a case control study . BMC Nephrol 2017 ; 18 : 99 . Google Scholar Crossref Search ADS PubMed 10. Hata Y , Muratani H , Kimura Y , Fukiyama K , Kawano Y , Ashida T , Yokouchi M , Imai Y , Ozawa T , Fujii J , Omae T . Office blood pressure variability as a predictor of acute myocardial infarction in elderly patients receiving antihypertensive therapy . J Hum Hypertens 2002 ; 16 : 141 – 146 . Google Scholar Crossref Search ADS PubMed 11. Asayama K , Kikuya M , Schutte R , Thijs L , Hosaka M , Satoh M , Hara A , Obara T , Inoue R , Metoki H , Hirose T , Ohkubo T , Staessen JA , Imai Y . Home blood pressure variability as cardiovascular risk factor in the population of Ohasama . Hypertension 2013 ; 61 : 61 – 69 . Google Scholar Crossref Search ADS PubMed 12. Schutte R , Thijs L , Liu YP , Asayama K , Jin Y , Odili A , Gu YM , Kuznetsova T , Jacobs L , Staessen JA . Within-subject blood pressure level—not variability—predicts fatal and nonfatal outcomes in a general population . Hypertension 2012 ; 60 : 1138 – 1147 . Google Scholar Crossref Search ADS PubMed 13. Whelton PK , Carey RM , Aronow WS , Casey DE , Jr , Collins KJ , Dennison Himmelfarb C , DePalma SM , Gidding S , Jamerson KA , Jones DW , MacLaughlin EJ , Muntner P , Ovbiagele B , Smith SC , Jr , Spencer CC , Stafford RS , Taler SJ , Thomas RJ , Williams KA , Sr , Williamson JD , Wright JT , Jr . 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 . Hypertension 2018 ; 71 : e13 – e115 . Google Scholar Crossref Search ADS PubMed 14. Aoyama R , Takano H , Suzuki K , Kubota Y , Inui K , Tokita Y , Shimizu W . The impact of blood pressure variability on coronary plaque vulnerability in stable angina: an analysis using optical coherence tomography . Coron Artery Dis 2017 ; 28 : 225 – 231 . Google Scholar Crossref Search ADS PubMed 15. Ribeiro AH , Lotufo PA , Fujita A , Goulart AC , Chor D , Mill JG , Bensenor IM , Santos IS . Association between short-term systolic blood pressure variability and carotid intima-media thickness in ELSA-Brasil baseline . Am J Hypertens 2017 ; 30 : 954 – 960 . Google Scholar Crossref Search ADS PubMed 16. Rothwell PM , Howard SC , Dolan E , O’Brien E , Dobson JE , Dahlöf B , Sever PS , Poulter NR . Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension . Lancet 2010 ; 375 : 895 – 905 . Google Scholar Crossref Search ADS PubMed 17. Hussein WF , Chang TI . Visit-to-visit variability of systolic blood pressure and cardiovascular disease . Curr Hypertens Rep 2015 ; 17 : 14 . Google Scholar Crossref Search ADS PubMed 18. Chia YC , Ching SM , Lim HM . Visit-to-visit SBP variability and cardiovascular disease in a multiethnic primary care setting: 10-year retrospective cohort study . J Hypertens 2017 ; 35 ( Suppl 1 ): S50 – S56 . Google Scholar Crossref Search ADS PubMed 19. Muntner P , Levitan EB , Joyce C , Holt E , Mann D , Oparil S , Krousel-Wood M . Association between antihypertensive medication adherence and visit-to-visit variability of blood pressure . J Clin Hypertens (Greenwich) 2013 ; 15 : 112 – 117 . Google Scholar Crossref Search ADS PubMed 20. Nagai M , Dote K , Kato M , Sasaki S , Oda N , Kagawa E , Nakano Y , Yamane A , Kubo Y , Higashihara T , Miyauchi S , Harada W , Masuda H . Visit-to-visit blood pressure variability, average BP level and carotid arterial stiffness in the elderly: a prospective study . J Hum Hypertens 2017 ; 31 : 292 – 298 . Google Scholar Crossref Search ADS PubMed 21. Takao T , Suka M , Yanagisawa H , Matsuyama Y , Iwamoto Y . Predictive ability of visit-to-visit variability in HbA1c and systolic blood pressure for the development of microalbuminuria and retinopathy in people with type 2 diabetes . Diabetes Res Clin Pract 2017 ; 128 : 15 – 23 . Google Scholar Crossref Search ADS PubMed 22. Nagai M , Dote K , Kato M , Sasaki S , Oda N , Kagawa E , Nakano Y , Yamane A , Higashihara T , Miyauchi S , Tsuchiya A . Visit-to-visit blood pressure variability and Alzheimer’s disease: links and risks . J Alzheimers Dis 2017 ; 59 : 515 – 526 . Google Scholar Crossref Search ADS PubMed 23. Havlik RJ , Foley DJ , Sayer B , Masaki K , White L , Launer LJ . Variability in midlife systolic blood pressure is related to late-life brain white matter lesions: the Honolulu-Asia Aging study . Stroke 2002 ; 33 : 26 – 30 . Google Scholar Crossref Search ADS PubMed 24. Muntner P , Shimbo D , Tonelli M , Reynolds K , Arnett DK , Oparil S . The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population: findings from NHANES III, 1988 to 1994 . Hypertension 2011 ; 57 : 160 – 166 . Google Scholar Crossref Search ADS PubMed 25. Wang JG , Zhou D , Jeffers BW . Predictors of visit-to-visit blood pressure variability in patients with hypertension: an analysis of trials with an amlodipine treatment arm . J Am Soc Hypertens 2017 ; 11 : 402 – 411 . Google Scholar Crossref Search ADS PubMed 26. Muntner P , Levitan EB , Lynch AI , Simpson LM , Whittle J , Davis BR , Kostis JB , Whelton PK , Oparil S . Effect of chlorthalidone, amlodipine, and lisinopril on visit-to-visit variability of blood pressure: results from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial . J Clin Hypertens (Greenwich) 2014 ; 16 : 323 – 330 . Google Scholar Crossref Search ADS PubMed 27. Yadav S , Cotlarciuc I , Munroe PB , Khan MS , Nalls MA , Bevan S , Cheng YC , Chen WM , Malik R , McCarthy NS , Holliday EG , Speed D , Hasan N , Pucek M , Rinne PE , Sever P , Stanton A , Shields DC , Maguire JM , McEvoy M , Scott RJ , Ferrucci L , Macleod MJ , Attia J , Markus HS , Sale MM , Worrall BB , Mitchell BD , Dichgans M , Sudlow C , Meschia JF , Rothwell PM , Caulfield M , Sharma P ; International Stroke Genetics Consortium . Genome-wide analysis of blood pressure variability and ischemic stroke . Stroke 2013 ; 44 : 2703 – 2709 . Google Scholar Crossref Search ADS PubMed 28. Gosmanova EO , Mikkelsen MK , Molnar MZ , Lu JL , Yessayan LT , Kalantar-Zadeh K , Kovesdy CP . Association of systolic blood pressure variability with mortality, coronary heart disease, stroke, and renal disease . J Am Coll Cardiol 2016 ; 68 : 1375 – 1386 . Google Scholar Crossref Search ADS PubMed 29. Basson MD . Choosing among candidates for scarce medical resources . J Med Philos 1979 ; 4 : 313 – 333 . Google Scholar Crossref Search ADS PubMed 30. Stevens SL , Wood S , Koshiaris C , Law K , Glasziou P , Stevens RJ , McManus RJ . Blood pressure variability and cardiovascular disease: systematic review and meta-analysis . BMJ 2016 ; 354 : i4098 . Google Scholar Crossref Search ADS PubMed 31. (NICE). NIfHaCE . Hypertension in adults: diagnosis and management. Clinical guideline [CG127] . <https://www.nice.org.uk/guidance/cg127> 2016 . Accessed 1 March 2018. 32. Qin X , Zhang Q , Yang S , Sun Z , Chen X , Huang H . Blood pressure variability and morning blood pressure surge in elderly Chinese hypertensive patients . J Clin Hypertens (Greenwich) 2014 ; 16 : 511 – 517 . Google Scholar PubMed 33. Diaz KM , Muntner P , Levitan EB , Brown MD , Babbitt DM , Shimbo D . The effects of weight loss and salt reduction on visit-to-visit blood pressure variability: results from a multicenter randomized controlled trial . J Hypertens 2014 ; 32 : 840 – 848 . Google Scholar Crossref Search ADS PubMed 34. Hong K , Muntner P , Kronish I , Shilane D , Chang TI . Medication adherence and visit-to-visit variability of systolic blood pressure in African Americans with chronic kidney disease in the AASK trial . J Hum Hypertens 2016 ; 30 : 73 – 78 . Google Scholar Crossref Search ADS PubMed © American Journal of Hypertension, Ltd 2018. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Hypertension Oxford University Press

Visit-to-Visit Variability of Blood Pressure Is Associated With Hospitalization and Mortality in an Unselected Adult Population

Loading next page...
 
/lp/ou_press/visit-to-visit-variability-of-blood-pressure-is-associated-with-K9pfecCMjg
Publisher
Oxford University Press
Copyright
© American Journal of Hypertension, Ltd 2018. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
0895-7061
eISSN
1941-7225
D.O.I.
10.1093/ajh/hpy088
Publisher site
See Article on Publisher Site

Abstract

Abstract BACKGROUND Blood pressure variability (BPV) has been associated with poor health outcomes in high-risk patients, but its association with more general populations is poorly understood. METHODS We analyzed outcomes from 240,622 otherwise unselected patients who had 10 or more outpatient blood pressure readings recorded over a 3-year period and were aged from 20 to 100 years. RESULTS Whether calculated as SD, average change, or greatest change and systolic or diastolic blood pressure, we found that higher outpatient BPV was associated with subsequent hospitalization and mortality. Systolic pressure average change exceeding 10–12 mm Hg or diastolic exceeding 8 mm Hg significantly increased risk of hospitalization and death (odds ratios [ORs] from 2.0 to 4.5). Variability in the highest decile increased risks even more dramatically, with propensity-matched ORs from 4.4 to 42. A systolic change exceeding 35 mm Hg increased the relative risk of death 4.5-fold. Similarly, a diastolic change greater than 23–24 mm Hg almost tripled the risks of hospitalization and death. Neither stratification for hypertension nor propensity matching for risk factors within the database affected these associations. CONCLUSIONS Systolic and diastolic variabilities were each associated with subsequent adverse outcomes. Physicians should pay special attention to patients with swings in blood pressure between clinic visits. Electronic medical records should flag such variability. blood pressure, hospitalization, hypertension, mortality, variability, variance Blood pressure variability (BPV) within 1 day was associated with increased risk for hypertensive target-organ damage decades ago.1,2 However, there has been increasing interest in chronic BPV as a risk for subsequent complications, spurred by computerized data from large clinical trials that facilitate such calculations.3–7 In various high-risk groups, BPV, generally defined as either the SD or coefficient of variation of sphygmomanometric readings over time, has been reported to increase risks of stroke,3,5 cardiovascular mortality,5,8 and diabetic vascular complications like renal impairment,9 although such associations are not always found.4,6,8,10–12 Such studies frequently rely upon retrospective analyses of trials in specifically defined populations. However, clinical trial subjects may differ from patients at large because of inclusion and exclusion criteria. Indeed, some hypertension trials specifically exclude episodic hypertension. Furthermore, while cardiovascular events and death have been examined in this connection, less attention has been paid to the all-cause risk of hospitalization, increasingly important as risk shifts from third-party payers to health care organizations. We tested the hypothesis that greater BPV is associated with increased risk of all-cause hospitalization and mortality among unselected patients from a large multihospital regional health care system. We defined BPV as the common SD as well as by 2 new techniques: average change and largest change. We calculated systolic and diastolic BPV in patients over age 20 years with at least 10 readings. During final data analysis, the American Heart Association redefined systolic pressures above 130 mm Hg as hypertensive and detailed procedures for blood pressure measurement.13 We used the previous guidelines here since these represented how patients’ blood pressures were measured and how patients were treated during the study period. Risks due to variability were tested for various subpopulations. METHODS Data An aggregate data file extracted information from records for all patients using the Sanford Health system from December 2013 to November 2016 (n = 1,143,028). Sanford Health is an integrated health system, the largest rural, not-for-profit health care system in the United States with 45 hospitals and 289 clinics. Blood pressures for 1,029,899 patients were available (90.1%); 240,622 had 10 or more readings during at least 6 months, no missing values, and were aged 20 to 100 years, fitting study criteria. Two binary outcomes were recorded: whether patients had on average at least 1 hospitalization per year (N = 35,587, 14.8%) and whether patients died (N = 9,556, 4.0%). Patients were categorized by amount and type of abnormal blood pressure readings (Supplementary Appendix A). Of the 240,622 patients studied, 41% had been diagnosed as hypertensive. Abnormal pressures were also measured using percent of a patient’s readings above or below specific values. The “urgent high” category included 6%, for whom ≥15% of systolic pressures exceeded 160 or diastolics exceeded 100. The “caution high” category contained 24%, and 70% were “normal,” with <15% abnormally high readings. For abnormal low pressures, 4% were “urgently low,” 28% were in “caution low,” and 68% were “normal,” with <15% abnormally low readings. Six variables regarding BPV were created for tests of association with the two events. We calculated the SD (SDV), the average change (ACV) among all of each patient’s readings (the values of each change from reading to reading were averaged over the number of readings), and the largest change (LCV) recorded between any two consecutive readings (regardless of time duration between readings). As these calculations change with the number of readings used, we evaluated how the means and SDs of the BPVs changed when we varied the inclusion requirement from 5 to 30 readings/patient over the 3-year period. At least 10 readings achieved stability (Supplementary Appendix B). Most patients (95%) had a year or more between the 10 readings. Propensity analysis Supplementary Appendix C details the propensity analysis. Patients were categorized as using or not using antihypertensives (acetylcholinesterase [ACE] inhibitors, beta blockers, or other cardiac medication). 19.4% used antihypertensives, and 14.4 % used cardiac medication other than ACE inhibitors or beta blockers. Approximately, 10% used ACE inhibitors, and 9% used beta blockers. We excluded medications used by 10 or fewer patients from analysis for statistical rigor. Age, sex, race, rural status, smoking, alcohol use, primary care visits, use of a patient-centered electronic communication portal, emergency department use, and frequency of blood pressure readings were correlated with outcomes and antihypertensive use, controlling for potential confounders. Descriptive statistics (Supplementary Appendix C1) were entered into a propensity analysis predicting antihypertensive use, creating an inverse weight that was applied in statistical analysis to control for these variables. Statistical analysis Average BPV measures of SDV, ACV, and LCV were compared for all patients, hypertensive, and nonhypertensive patients, patients with normal, caution, or urgent low pressure, and patients with normal, caution, or urgent high pressure using independent t-tests and one-way ANOVA. How ACV and LCV changed between different patient groups was compared with how SDV changed. Inverse weights from the propensity analysis were used in logistic regressions predicting hospitalization and death from each BPV measure while controlling for medication use and other variables. Odds ratios (ORs) from ACV and LCV models were compared with ORs from SDV models. Sensitivity analysis We reanalyzed subgroups to test whether their results differed. Logistic regression to identify the risk of hospitalization or death for patients with the highest 10% of BPV relative to those with the lowest 10% of that BPV. We also tested subgroups based on blood pressure values (extreme highs and lows or hypertension, Supplementary Table A1) or on numbers of pressures recorded (assuming those with higher numbers were hospitalized or sicker). Different cutoffs for the measures of change and how they affected likelihood of hospitalization or death were analyzed using receiver operating characteristic curves. Additional statistics included relative risk (RR), population attributable risk, sensitivity, specificity, positive predictive value, negative predictive value, population impact number, exposure impact number, and case impact number. SAS 9.4 and 14.2 (2017, SAS Institute, Cary, North Carolina) and NCSS 11 Statistical Software (2016, NCSS, Kaysville, Utah, ncss.com/software/ncss) were used. RESULTS Two hundred forty thousand, six hundred twenty-two patients aged 20 to 100 years had 10 or more blood pressure readings for at least 6 months during the 3-year period. Their average age was 54.9 years (SD 19.7), 63.6% were female, and 93.0% were white. Categories based on abnormal blood pressures are defined in Supplementary Table A1 of Supplementary Appendix A. Out of the entire sample, 98,335 (40.9%) had hypertension diagnoses. Based upon all patients’ recorded blood pressure readings, we retrospectively categorized 15,193 (6.3%) as having urgent high blood pressure and 57,306 (23.8%) as caution abnormal high blood pressure. Similarly, we retrospectively categorized 1,024 (4.2%) and 67,155 (27.9%) as having urgently low blood pressures and caution abnormal low blood pressure, respectively. Of the 240,622 patients, 19.3% took antihypertensive medications (Supplementary Appendix B); 35,587 patients (14.8%) had on average 1 hospitalization per year; 9,556 (4.0%) were deceased. We examined 3 measures each of systolic and diastolic variability, comparing SDV with ACV and LCV. (Table 1) All 6 measures had distributions with a minor positive skew. Their means were compared between categories of blood pressure extremes (Supplementary Table A1). Patients with higher severity (hypertension or abnormal blood pressures) had significantly higher BPV measures (all P < 0.001). SDV and ACV values were nearly identical; LCV values were higher as they were based on largest changes. Diagnosed hypertensives had average systolic or diastolic SDV and ACV about 3 points or 1 point higher than nonhypertensives. The increase in LCV was about 9 or 3 points. Patients with abnormal high pressures showed a similar pattern with systolic BPV, but diastolic BPV had less decrease from urgent to caution categories. Systolic SDV decreased by nearly half, from 18.1 to 10.0, between urgent and not caution or urgent. Diastolic SDV decreased from 10.3 to 7.4. Patients with abnormally low blood pressure exhibited smaller changes of only a few points, similar to hypertension. Table 1. Measures of blood pressure variability for 240,622 patients All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability. View Large Table 1. Measures of blood pressure variability for 240,622 patients All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 All (N = 240,622) Hypertension diagnosis Hypertensive (N = 98,335) Not hypertensive (N = 142,287) Mean SD Mean SD Mean SD Systolic SDV 11.195 3.869 12.899 4.268 10.018 3.055  ACV 11.073 3.923 12.600 4.293 10.017 3.249  LCV 32.517 13.834 37.983 15.364 28.740 11.208 Diastolic SDV 7.877 2.160 8.332 2.280 7.562 2.013  ACV 7.892 2.349 8.320 2.408 7.596 2.260  LCV 22.841 8.569 24.574 9.164 21.644 7.913 Abnormal high blood pressure % Urgent (N = 15,193) % Caution (N = 57,306) Not caution or urgent (N = 168,123) Systolic SDV 18.072 4.868 12.887 3.499 9.997 2.902  ACV 17.107 5.291 12.639 3.747 9.993 3.069  LCV 49.870 17.332 37.525 13.840 29.242 11.619 Diastolic SDV 10.351 2.920 8.529 2.174 7.431 1.843  ACV 9.999 3.067 8.511 2.412 7.490 2.091  LCV 29.210 11.174 24.943 9.038 21.549 7.689 Abnormal low blood pressure %Urgent (N = 1,024) % Caution (N = 67,155) Not caution or urgent (N = 163,263) Systolic SDV 12.804 4.729 10.511 3.513 11.300 3.848  ACV 12.524 4.755 10.475 3.646 11.159 3.889  LCV 37.740 16.705 30.796 13.124 32.664 13.648 Diastolic SDV 9.002 2.547 7.296 1.789 7.989 2.186  ACV 8.918 2.745 7.322 2.022 8.009 2.372  LCV 26.658 10.236 21.324 7.610 23.048 8.598 Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability. View Large Logistic regressions for SDV, ACV, and LCV tested how BPVs differed in association with hospitalization or death (Table 2), controlling for confounding variables using propensity weights (Supplementary Appendix B). The unadjusted hospitalization risk from BPV varied from OR = 1.03 (systolic LCV) to OR = 1.14 (diastolic ACV). Adjusted ORs slightly increased for all BPVs (1.04 to 1.21) except for diastolic SDV. The mortality risk was increased from hospitalization for all BPV, both unadjusted (OR ranged from 1.05 to 1.24) and adjusted (OR ranged from 1.04 to 1.25). Overall, the risk from ACV was nearly identical to the SDV, while the risk from LCV was 10 to 15% lower. Table 2. Unadjusted and propensity-adjusted ORs of blood pressure variance measures predicting hospitalization and death from logistic regressions Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Propensity scores controlled for demographics and health care of patients. See Supplementary Appendix C for details. Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Table 2. Unadjusted and propensity-adjusted ORs of blood pressure variance measures predicting hospitalization and death from logistic regressions Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Unadjusted Propensity adjusted OR LL UL OR LL UL Predicting hospitalization  Systolic SDV 1.103 1.100 1.105 1.115 1.113 1.117   ACV 1.086 1.083 1.088 1.101 1.099 1.103   LCV 1.034 1.034 1.035 1.038 1.038 1.039  Diastolic SDV 1.210 1.205 1.215 1.208 1.205 1.211   ACV 1.144 1.140 1.148 1.158 1.155 1.161   LCV 1.061 1.060 1.062 1.062 1.061 1.063 Predicting death  Systolic SDV 1.210 1.205 1.215 1.184 1.181 1.187   ACV 1.178 1.174 1.183 1.157 1.154 1.160   LCV 1.048 1.047 1.049 1.043 1.042 1.043  Diastolic SDV 1.315 1.305 1.325 1.246 1.240 1.253   ACV 1.243 1.234 1.252 1.199 1.193 1.205   LCV 1.061 1.059 1.063 1.051 1.049 1.052 Propensity scores controlled for demographics and health care of patients. See Supplementary Appendix C for details. Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Sensitivity of measures Patients in the highest 10%ile of BPVs were compared with those with the lowest 10%ile of the same BPV. Being in the highest, 10% greatly increased event risks. (Table 3) ACV showed the smallest increase with ORs of 4.9 systolic and 4.4 diastolic for hospitalization. SDV and LCV were near ORs of 7. Death risk increased dramatically for the top 10%. The OR for death for an SDV in the top 10% vs. the bottom 10% was 42. The smallest ORs were for diastolic measures with LCV OR = 8.0. Table 3. Risk of hospitalization and death in patients in the upper 10th percentile relative to those in the lower 10th percentile of BPV measures Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 ORs were from logistic regressions adjusting for demographics, health care, medication, and blood pressure. N ranged from 48,126 (diastolic SDV) to 53,006 (systolic largest change). Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Table 3. Risk of hospitalization and death in patients in the upper 10th percentile relative to those in the lower 10th percentile of BPV measures Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 Hospitalization Death OR LL UL OR LL UL Systolic SDV 6.836 6.573 7.110 42.066 36.736 48.170  ACV 4.868 4.693 5.050 17.524 15.646 19.628  LCV 7.980 7.700 8.270 15.559 14.069 17.207 Diastolic SDV 6.406 6.192 6.627 12.939 11.782 14.210  ACV 4.434 4.293 4.581 8.268 7.562 9.039  LCV 7.548 7.291 7.813 8.083 7.382 8.851 ORs were from logistic regressions adjusting for demographics, health care, medication, and blood pressure. N ranged from 48,126 (diastolic SDV) to 53,006 (systolic largest change). Abbreviations: ACV, average change variability; LCV, largest change variability; SDV, SD variability; OR, odds ratio; LL, lower limit; UL, upper limit. View Large Logistic regressions were repeated for various subpopulations (Supplementary Table A1) to examine the robustness and generalizability of our findings. (Supplementary Table D1). All ORs remained statistically significant (P < 0.05), and patterns of high and low findings were similar. (As with unadjusted and propensity-adjusted ORs, SDV and ACV ORs were similar (OR approximately 1.1) and LCV ORs were lower) We further excluded patient/dates with 3 or more consecutive pressures from the data set, assuming these were likely hospitalization days. Excluding such presumed inpatient readings did not change results from Tables 1 and 2 significantly (P > 0.05). In addition, systolic and diastolic SDV were similarly associated with outcomes in patients <50, 50–65, and >65 (not shown). Receiver operating characteristic analysis assessed optimal cutoff values of ACV and LCV. Supplementary Figures D1a and D1b showed few differences between SDV and ACV or LCV. The AUCs for hospitalization ranged from 63.5 to 68.3% and death from 67.1 to 74.2%. Cutoff values were chosen for slightly higher sensitivities from 61 to 72%. Using those cutoffs, RRs for hospitalization doubled or more the increase in risk, with LCVs more than 31 having an RR = 2.50 or higher (Table 4). A systolic change exceeding 35 mm Hg increased the RR of death 4.5-fold. Similarly, a diastolic change greater than 23–24 mm Hg almost tripled the risks of hospitalization and death. Supplementary Appendix D shows other statistics based on the cutoff values. Table 4. Sensitivity analysis of blood pressure variance cutoffs RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR: relative risk; there is a (1 − RR)% increase in the risk of outcome (hospitalization or death) due to BP change. PAR%: population attributable risk; among the population, PAR% of the total risk of event is due to BP change. Sens: sensitivity; ability to detect outcome when change is present; % of those with event who had BP change. Spec: specificity; ability to detect nonoutcome when no change; % of those with no event who had no BP change. PPV: positive predictive value; % of those with BP change who had outcome. NPV: negative predictive value; % of those with no BP change who had no outcome. PIN: population impact number; number in population whose outcome is attributable to BP change; for every PIN people, one outcome is attributable to BP change. EIN: exposure impact number; number of people with change where one with an outcome is due to BP change, for every EIN people with BP change, there is one with the outcome. CIN: case impact number; number of people with outcome where one outcome is from BP change, for every CIN people with outcomes, there is one attributable to BP change. Abbreviations: ACV, average change variability; LCV, largest change variability. View Large Table 4. Sensitivity analysis of blood pressure variance cutoffs RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR PAR% Sens Spec PPV NPV PIN EIN CIN Hospitalization  Systolic ACV ≥ 10 2.00 35.71 71.51 47.05 18.99 90.49 18.93 10.54 2.80   LCV ≥ 31 2.50 41.02 68.42 57.35 21.78 91.28 16.48 7.66 2.44  Diastolic ACV ≥ 8 1.96 29.94 61.14 58.34 20.30 89.64 22.59 10.06 3.34   LCV ≥ 23 2.61 41.11 66.65 60.68 22.73 91.29 16.45 7.13 2.43 Death  Systolic ACV ≥ 12 3.92 49.76 66.78 67.47 7.82 98.00 50.61 17.15 2.01   LCV ≥ 35 4.48 54.95 70.75 66.41 8.01 98.21 45.82 16.07 1.82  Diastolic ACV ≥ 8 2.70 43.07 68.43 56.45 6.10 97.74 58.46 26.04 2.32   LCV ≥ 24 3.00 44.84 67.20 60.56 6.58 97.81 56.16 22.77 2.23 RR: relative risk; there is a (1 − RR)% increase in the risk of outcome (hospitalization or death) due to BP change. PAR%: population attributable risk; among the population, PAR% of the total risk of event is due to BP change. Sens: sensitivity; ability to detect outcome when change is present; % of those with event who had BP change. Spec: specificity; ability to detect nonoutcome when no change; % of those with no event who had no BP change. PPV: positive predictive value; % of those with BP change who had outcome. NPV: negative predictive value; % of those with no BP change who had no outcome. PIN: population impact number; number in population whose outcome is attributable to BP change; for every PIN people, one outcome is attributable to BP change. EIN: exposure impact number; number of people with change where one with an outcome is due to BP change, for every EIN people with BP change, there is one with the outcome. CIN: case impact number; number of people with outcome where one outcome is from BP change, for every CIN people with outcomes, there is one attributable to BP change. Abbreviations: ACV, average change variability; LCV, largest change variability. View Large DISCUSSION Even modestly heightened BPV posed a substantial risk for hospitalization and mortality in this general adult population. Systolic variance exceeding 10–12 mm Hg or diastolic variance exceeding 8 mm Hg notably worsened outcomes. Therefore, clinicians should not disregard apparently isolated elevated or low pressures despite other normal readings. BPV can be calculated from systolic, diastolic, or mean pressures, using coefficients of variation, SD, or variance. Systolic and diastolic variability yielded similar results. Others have emphasized blood pressure SD(5,14–16), which predicts similarly to the coefficient of the mean.17 Because SDs are not intuitively obvious, indeed, average and largest change were easier and approximately as effective as SD in predicting hospitalization and death after controlling for confounders. “Largest change” might be easiest for clinicians to assess from lists of measurements. Independent of the absolute magnitude of the patient’s blood pressure, specific changes showed an increased risk of 100 to 250% (Table 4). One might hypothesize that BPV simply reflects poorly controlled hypertension because of severity or noncompliance. Indeed, most previous studies have examined high-risk patients, who are generally hypertensive. However, BPV was associated with more frequent cardiovascular events even in well-controlled hypertensives.18 Moreover, medication noncompliance explains only a small proportion of BPV.19 Other variability may signal autonomic instability that could precede development or worsening of vascular wall abnormalities,14,15,20 diabetes21 or brain lesions and/or dementia,22,23 or other even less well-understood disorders of homeostasis. Age, female gender, smoking, systolic or diastolic hypertension, peripheral vascular disease, diabetes, renal insufficiency, heart rate variability, widened pulse pressure, and angiotensin-converting enzyme therapy have each been correlated with BPV.16,24,25 Conversely, chlorthalidone or amlodipine may be associated with decreased BPV among treated hypertensives.25,26 We were unable to analyze all these factors in our administrative data, but propensity matching including (separately) use of ACE inhibitors, beta blockers, any blood pressure medication, and any cardiac medication did not invalidate our results.27 A recent meta-analysis suggested that BPV should be monitored as a prognostic indicator for mortality and cardiovascular complications in high-risk individuals.25 Our administrative data set was insufficiently granular to distinguish causes of hospitalization and mortality, but retrospective analyses of clinical trials in high-risk individuals suggest that the associated morbidity includes,5 but may not be limited to,(7,22) cardiovascular morbidity. Our results, taken together with these other observations, suggest that this is also likely true in normotensive individuals. Our results suggest that BPV is similarly predictive whether or not patients are diagnosed as hypertensive, take antihypertensives, or are actually hypertensive or hypotensive. Gosmanova28 associated systolic BPV with all-cause mortality in veterans who were not necessarily hypertensive, although likely many were. Some patients may have masked hypertension missed by office-based measurements. Whether “normotensive” patients with high BPV are more likely to have such masked hypertension awaits study, just as normotensive patients with “white coat syndrome” may exhibit elevated pressures from anxiety. However, since clinicians generally assess blood pressures in this fashion, this would not invalidate our suggestion that in-office normotensive patients with high in-office BPV should be considered high risk. Since we required at least 10 pressure readings over the 3-year study period to meaningfully calculate variability, different results might be obtained in people who do not require this many health care encounters. Gosmanova28 similarly required at least 8 measurements over 2 years. While we achieved similar results when we required 15 or 20 blood pressure readings to calculate BPV more precisely (not shown), decreasing the number of blood pressure readings to 5 substantially decreases the mathematical reliability of the calculations. (Supplementary Appendix A) However, Muntner24 associated systolic BPV and all-cause mortality in the NHANES study using only 3 measurements over 6 years, despite the increased statistical noise created by having fewer measurements. Only 33 patients had all their blood pressures within 30 days. Most (95%) took more than a year, and 79% took 2 or more years to produce 10 readings. Clustering of blood pressures within a short term over an acute event therefore seems unlikely to have substantially biased results. Operational purposes require a cutoff or “abnormal” value to trigger clinicians’ attention. These cutoffs trade sensitivity against specificity (Supplementary Figure D1), based upon system characteristics and resources and the trade-off between alarming patients unnecessarily and failing to warn patients appropriately. Like other scarce resources,29 physician time must be allocated on ethical as well as pragmatic grounds. However, choosing 75 and 50% as endpoints for sensitivity and specificity seems reasonable and yields population impact numbers of approximately 12 for hospitalization and 30–40 for death, which seems deserving of attention. The magnitude of the BPV effect resembles that of abnormal cholesterol levels, to which attention is already standard of care.30 It might seem counterintuitive that such small changes in BPV predict clinically significant differences in outcomes, but we similarly now understand that even a 2 mm Hg reduction in systolic pressure meaningfully reduces cardiovascular mortality.31 Data availability created limitations. Administrative data could not be validated by individual chart review, but there is no reason to assume that errors in the data would distribute differently across low- or high-BPV patients. We obtained time of blood pressure but not time of hospitalization or death, making any proportional hazard model impossible. Because we could not obtain records after 2 November 2016, we could not capture events after this time. Limiting the study to people with at least 6 months of blood pressure data removed any censoring at the beginning, such as excluding a person who died 1 month into the study. Further censoring would have required excluding anyone who had no event during the study. Including these patients may have introduced some time bias but offered a baseline group with no events in the study time frame. Furthermore, because we lacked time of event, we used a cross-sectional type approach to gathering BPV data both before and after an event (at any time in the 3-year period). Further research should incorporate time of event and exposure to adjust for changes over time and could use a proportional hazard model. Cases and controls could also be used in future analyses to estimate these results for specific populations, such as specific age groups or genders. In addition, although we accounted for medication use in our propensity matching, we lacked data regarding adherence to medication, as well as conventional measures of baseline cardiac risk. Primary care physicians encountering high BPV might seek such information about their patients, while future studies might investigate these variables’ effects on the associations we define here. In conclusion, BPV is a calculated variable not intuitively obvious to the clinician. These findings suggest an opportunity to use the electronic medical record to alert clinicians to an important but largely ignored risk factor. Clinicians should scan blood pressure listings over time for the largest consecutive changes, paying particular attention to patients with changes exceeding 34 systolic or 23 diastolic. Administrators of hospital systems should consider creating a computer-calculated variable, whether blood pressure SD or average change, that can be flagged like a calculated obese BMI. For instance, a systolic average change exceeding 12 or diastolic exceeding 8 merits attention. Although further studies are required to demonstrate the efficacy of targeting specific interventions to patients with high BPV, it would not seem unreasonable for clinicians encountering patients with high BPV taking ACE inhibitors to balance the known benefits of ACE inhibitors in diabetes and congestive heart failure against their potential adverse effects on BPV. Obesity and smoking increase BPV in some32 but not all33 studies and even if not are likely to be synergistic in effects. Medication adherence may also diminish BPV.34 Antihypertensive trials should address BPV in addition to blood pressure magnitude. BPV needs further study but is promising for the care of patients. SUPPLEMENTARY DATA Supplementary data are available at American Journal of Hypertension online. DISCLOSURE All authors participated in this research and in preparation of the manuscript. The authors declared no conflict of interest. ACKNOWLEDGMENTS We would like to acknowledge the generous access to data provided by the Sanford Health system under a Sanford research grant program, and the support and encouragement of David Pearce, Ph.D., Executive Vice President, Sanford Research. REFERENCES 1. Frattola A , Parati G , Cuspidi C , Albini F , Mancia G . Prognostic value of 24-hour blood pressure variability . J Hypertens 1993 ; 11 : 1133 – 1137 . Google Scholar Crossref Search ADS PubMed 2. Parati G , Pomidossi G , Albini F , Malaspina D , Mancia G . Relationship of 24-hour blood pressure mean and variability to severity of target-organ damage in hypertension . J Hypertens 1987 ; 5 : 93 – 98 . Google Scholar Crossref Search ADS PubMed 3. Dai H , Lu Y , Song L , Tang X , Li Y , Chen R , Luo A , Yuan H , Wu S . Visit-to-visit variability of blood pressure and risk of stroke: results of the Kailuan cohort study . Sci Rep 2017 ; 7 : 285 . Google Scholar Crossref Search ADS PubMed 4. Suchy-Dicey AM , Wallace ER , Mitchell SV , Aguilar M , Gottesman RF , Rice K , Kronmal R , Psaty BM , Longstreth WT , Jr . Blood pressure variability and the risk of all-cause mortality, incident myocardial infarction, and incident stroke in the cardiovascular health study . Am J Hypertens 2013 ; 26 : 1210 – 1217 . Google Scholar Crossref Search ADS PubMed 5. Vidal-Petiot E , Stebbins A , Chiswell K , Ardissino D , Aylward PE , Cannon CP , Ramos Corrales MA , Held C , López-Sendón JL , Stewart RAH , Wallentin L , White HD , Steg PG ; STABILITY Investigators . Visit-to-visit variability of blood pressure and cardiovascular outcomes in patients with stable coronary heart disease. Insights from the STABILITY trial . Eur Heart J 2017 ; 38 : 2813 – 2822 . Google Scholar Crossref Search ADS PubMed 6. Tully PJ , Debette S , Dartigues JF , Helmer C , Artero S , Tzourio C . Antihypertensive drug use, blood pressure variability, and incident stroke risk in older adults: three-city cohort study . Stroke 2016 ; 47 : 1194 – 1200 . Google Scholar Crossref Search ADS PubMed 7. Goyal A , Mezue K , Rangaswami J . Visit-to-visit systolic blood pressure variability predicts treatment-related adverse event of hyponatremia in SPRINT . Cardiovasc Ther 2017 ; 35 . doi:10.1111/1755-5922.12274 8. Pringle E , Phillips C , Thijs L , Davidson C , Staessen JA , de Leeuw PW , Jaaskivi M , Nachev C , Parati G , O’Brien ET , Tuomilehto J , Webster J , Bulpitt CJ , Fagard RH ; Syst-Eur investigators . Systolic blood pressure variability as a risk factor for stroke and cardiovascular mortality in the elderly hypertensive population . J Hypertens 2003 ; 21 : 2251 – 2257 . Google Scholar Crossref Search ADS PubMed 9. Yeh CH , Yu HC , Huang TY , Huang PF , Wang YC , Chen TP , Yin SY . The risk of diabetic renal function impairment in the first decade after diagnosed of diabetes mellitus is correlated with high variability of visit-to-visit systolic and diastolic blood pressure: a case control study . BMC Nephrol 2017 ; 18 : 99 . Google Scholar Crossref Search ADS PubMed 10. Hata Y , Muratani H , Kimura Y , Fukiyama K , Kawano Y , Ashida T , Yokouchi M , Imai Y , Ozawa T , Fujii J , Omae T . Office blood pressure variability as a predictor of acute myocardial infarction in elderly patients receiving antihypertensive therapy . J Hum Hypertens 2002 ; 16 : 141 – 146 . Google Scholar Crossref Search ADS PubMed 11. Asayama K , Kikuya M , Schutte R , Thijs L , Hosaka M , Satoh M , Hara A , Obara T , Inoue R , Metoki H , Hirose T , Ohkubo T , Staessen JA , Imai Y . Home blood pressure variability as cardiovascular risk factor in the population of Ohasama . Hypertension 2013 ; 61 : 61 – 69 . Google Scholar Crossref Search ADS PubMed 12. Schutte R , Thijs L , Liu YP , Asayama K , Jin Y , Odili A , Gu YM , Kuznetsova T , Jacobs L , Staessen JA . Within-subject blood pressure level—not variability—predicts fatal and nonfatal outcomes in a general population . Hypertension 2012 ; 60 : 1138 – 1147 . Google Scholar Crossref Search ADS PubMed 13. Whelton PK , Carey RM , Aronow WS , Casey DE , Jr , Collins KJ , Dennison Himmelfarb C , DePalma SM , Gidding S , Jamerson KA , Jones DW , MacLaughlin EJ , Muntner P , Ovbiagele B , Smith SC , Jr , Spencer CC , Stafford RS , Taler SJ , Thomas RJ , Williams KA , Sr , Williamson JD , Wright JT , Jr . 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 . Hypertension 2018 ; 71 : e13 – e115 . Google Scholar Crossref Search ADS PubMed 14. Aoyama R , Takano H , Suzuki K , Kubota Y , Inui K , Tokita Y , Shimizu W . The impact of blood pressure variability on coronary plaque vulnerability in stable angina: an analysis using optical coherence tomography . Coron Artery Dis 2017 ; 28 : 225 – 231 . Google Scholar Crossref Search ADS PubMed 15. Ribeiro AH , Lotufo PA , Fujita A , Goulart AC , Chor D , Mill JG , Bensenor IM , Santos IS . Association between short-term systolic blood pressure variability and carotid intima-media thickness in ELSA-Brasil baseline . Am J Hypertens 2017 ; 30 : 954 – 960 . Google Scholar Crossref Search ADS PubMed 16. Rothwell PM , Howard SC , Dolan E , O’Brien E , Dobson JE , Dahlöf B , Sever PS , Poulter NR . Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension . Lancet 2010 ; 375 : 895 – 905 . Google Scholar Crossref Search ADS PubMed 17. Hussein WF , Chang TI . Visit-to-visit variability of systolic blood pressure and cardiovascular disease . Curr Hypertens Rep 2015 ; 17 : 14 . Google Scholar Crossref Search ADS PubMed 18. Chia YC , Ching SM , Lim HM . Visit-to-visit SBP variability and cardiovascular disease in a multiethnic primary care setting: 10-year retrospective cohort study . J Hypertens 2017 ; 35 ( Suppl 1 ): S50 – S56 . Google Scholar Crossref Search ADS PubMed 19. Muntner P , Levitan EB , Joyce C , Holt E , Mann D , Oparil S , Krousel-Wood M . Association between antihypertensive medication adherence and visit-to-visit variability of blood pressure . J Clin Hypertens (Greenwich) 2013 ; 15 : 112 – 117 . Google Scholar Crossref Search ADS PubMed 20. Nagai M , Dote K , Kato M , Sasaki S , Oda N , Kagawa E , Nakano Y , Yamane A , Kubo Y , Higashihara T , Miyauchi S , Harada W , Masuda H . Visit-to-visit blood pressure variability, average BP level and carotid arterial stiffness in the elderly: a prospective study . J Hum Hypertens 2017 ; 31 : 292 – 298 . Google Scholar Crossref Search ADS PubMed 21. Takao T , Suka M , Yanagisawa H , Matsuyama Y , Iwamoto Y . Predictive ability of visit-to-visit variability in HbA1c and systolic blood pressure for the development of microalbuminuria and retinopathy in people with type 2 diabetes . Diabetes Res Clin Pract 2017 ; 128 : 15 – 23 . Google Scholar Crossref Search ADS PubMed 22. Nagai M , Dote K , Kato M , Sasaki S , Oda N , Kagawa E , Nakano Y , Yamane A , Higashihara T , Miyauchi S , Tsuchiya A . Visit-to-visit blood pressure variability and Alzheimer’s disease: links and risks . J Alzheimers Dis 2017 ; 59 : 515 – 526 . Google Scholar Crossref Search ADS PubMed 23. Havlik RJ , Foley DJ , Sayer B , Masaki K , White L , Launer LJ . Variability in midlife systolic blood pressure is related to late-life brain white matter lesions: the Honolulu-Asia Aging study . Stroke 2002 ; 33 : 26 – 30 . Google Scholar Crossref Search ADS PubMed 24. Muntner P , Shimbo D , Tonelli M , Reynolds K , Arnett DK , Oparil S . The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population: findings from NHANES III, 1988 to 1994 . Hypertension 2011 ; 57 : 160 – 166 . Google Scholar Crossref Search ADS PubMed 25. Wang JG , Zhou D , Jeffers BW . Predictors of visit-to-visit blood pressure variability in patients with hypertension: an analysis of trials with an amlodipine treatment arm . J Am Soc Hypertens 2017 ; 11 : 402 – 411 . Google Scholar Crossref Search ADS PubMed 26. Muntner P , Levitan EB , Lynch AI , Simpson LM , Whittle J , Davis BR , Kostis JB , Whelton PK , Oparil S . Effect of chlorthalidone, amlodipine, and lisinopril on visit-to-visit variability of blood pressure: results from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial . J Clin Hypertens (Greenwich) 2014 ; 16 : 323 – 330 . Google Scholar Crossref Search ADS PubMed 27. Yadav S , Cotlarciuc I , Munroe PB , Khan MS , Nalls MA , Bevan S , Cheng YC , Chen WM , Malik R , McCarthy NS , Holliday EG , Speed D , Hasan N , Pucek M , Rinne PE , Sever P , Stanton A , Shields DC , Maguire JM , McEvoy M , Scott RJ , Ferrucci L , Macleod MJ , Attia J , Markus HS , Sale MM , Worrall BB , Mitchell BD , Dichgans M , Sudlow C , Meschia JF , Rothwell PM , Caulfield M , Sharma P ; International Stroke Genetics Consortium . Genome-wide analysis of blood pressure variability and ischemic stroke . Stroke 2013 ; 44 : 2703 – 2709 . Google Scholar Crossref Search ADS PubMed 28. Gosmanova EO , Mikkelsen MK , Molnar MZ , Lu JL , Yessayan LT , Kalantar-Zadeh K , Kovesdy CP . Association of systolic blood pressure variability with mortality, coronary heart disease, stroke, and renal disease . J Am Coll Cardiol 2016 ; 68 : 1375 – 1386 . Google Scholar Crossref Search ADS PubMed 29. Basson MD . Choosing among candidates for scarce medical resources . J Med Philos 1979 ; 4 : 313 – 333 . Google Scholar Crossref Search ADS PubMed 30. Stevens SL , Wood S , Koshiaris C , Law K , Glasziou P , Stevens RJ , McManus RJ . Blood pressure variability and cardiovascular disease: systematic review and meta-analysis . BMJ 2016 ; 354 : i4098 . Google Scholar Crossref Search ADS PubMed 31. (NICE). NIfHaCE . Hypertension in adults: diagnosis and management. Clinical guideline [CG127] . <https://www.nice.org.uk/guidance/cg127> 2016 . Accessed 1 March 2018. 32. Qin X , Zhang Q , Yang S , Sun Z , Chen X , Huang H . Blood pressure variability and morning blood pressure surge in elderly Chinese hypertensive patients . J Clin Hypertens (Greenwich) 2014 ; 16 : 511 – 517 . Google Scholar PubMed 33. Diaz KM , Muntner P , Levitan EB , Brown MD , Babbitt DM , Shimbo D . The effects of weight loss and salt reduction on visit-to-visit blood pressure variability: results from a multicenter randomized controlled trial . J Hypertens 2014 ; 32 : 840 – 848 . Google Scholar Crossref Search ADS PubMed 34. Hong K , Muntner P , Kronish I , Shilane D , Chang TI . Medication adherence and visit-to-visit variability of systolic blood pressure in African Americans with chronic kidney disease in the AASK trial . J Hum Hypertens 2016 ; 30 : 73 – 78 . Google Scholar Crossref Search ADS PubMed © American Journal of Hypertension, Ltd 2018. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

American Journal of HypertensionOxford University Press

Published: Sep 11, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off