Content area
Abstract
In order to assess the overall discriminative ability of SBP thresholds in predicting 7-day mortality, we made receiver operating characteristic (ROC) curves and calculated area under the ROC curves (AUROC). Because mortality increased in both ends of the SBP scale and we aimed to identify the lower SBP threshold, we restricted the ROC curve to patients with SBPs below the value where the ROC curve crossed the diagonal reference line. Ambulance MECU P value † All patients, n 112,727 15,862 12,270 Male, n (%) 59,523 (53) 7,778 (49) 7,028 (57) <0.001 Essential hypertension, n (%) 26,592 (24) 7,057 (46) 5,292 (43) <0.001 Charlson Comorbidity Index, n (%) <0.001 0 83,068 (74) 8,204 (52) 6,595 (54) 1 - 2 22,379 (20) 4,949 (31) 3,655 (30) >2 7,280 (7) 2,709 (17) 2,020 (17) Age, median [25th-75th percentile] 49 [30-69] 64 [44-78] 60 [42-74] <0.001 Systolic blood pressure, mean ± s.d. 148 ± 28 145 ± 28 141 ± 31 <0.001 Diastolic blood pressure, mean ± s.d. 85 ± 16 84 ± 18 84 ± 20 <0.001 Heart rate, mean ± s.d. 84 ± 19 87 ± 20 92 ± 25 <0.001 7-day mortality, n (%) 2,065 (1.8) 356 (2.2) 703 (5.7) <0.001 30-day mortality, n (%) 3,503 (3.1) 789 (5.0) 1,063 (8.7) <0.001 *Values expressed as total number (fraction), medians [25th percentile to 75th percentile], or mean ± standard deviation (s.d.), as appropriate. †Chi-squared test for categorical variables and Kruskal-Wallis test for continuous variables. [ Table Omitted - see PDF ] Table 2 Summary statistics of possible systolic blood pressure thresholds in predicting 7-day mortality for the three cohorts Sensitivity Specificity PPV NPV LR+ LR- FP FN FP/FN trade-off * Emergency department <90 mmHg 12 (11-14) 99 (99-99) 22 (20-25) 98 (98-98) 15 (14-18) 0.89 (0.87-0.9) 871 1815 1 (ref) <100 mmHg 19 (17-20) 98 (98-98) 15 (14-16) 98 (98-99) 9.4 (8.5-10) 0.83 (0.81-0.85) 2193 1679 10 <110 mmHg 27 (25-29) 94 (94-95) 8 (8-9) 99 (99-99) 4.8 (4.5-5.2) 0.78 (0.76-0.8) 6116 1512 17 <120 mmHg 36 (33-38) 87 (87-87) 5 (5-5) 99 (99-99) 2.7 (2.6-2.9) 0.74 (0.72-0.76) 14329 1331 28 Ambulance <90 mmHg 10 (7-13) 99 (98-99) 13 (9-18) 98 (98-98) 6.7 (4.8-9.4) 0.92 (0.88-0.95) 228 321 1 (ref) <100 mmHg 18 (14-22) 97 (96-97) 11 (8-13) 98 (98-98) 5.2 (4.1-6.5) 0.85 (0.81-0.89) 540 292 11 <110 mmHg 27 (22-32) 92 (92-93) 7 (6-9) 98 (98-98) 3.4 (2.9-4.1) 0.79 (0.74-0.84) 1214 260 16 <120 mmHg 37 (32-43) 84 (83-84) 5 (4-6) 98 (98-99) 2.3 (2-2.6) 0.75 (0.69-0.81) 2540 223 24 MECU <90 mmHg 16 (14-19) 96 (96-97) 21 (17-24) 95 (95-95) 4.3 (3.6-5.2) 0.87 (0.84-0.9) 440 588 1 (ref) <100 mmHg 25 (22-29) 93 (92-93) 18 (16-21) 95 (95-96) 3.6 (3.1-4.2) 0.8 (0.77-0.84) 813 524 6 <110 mmHg 34 (31-38) 88 (87-88) 14 (13-16) 96 (95-96) 2.7 (2.5-3.1) 0.75 (0.71-0.79) 1443 462 8 <120 mmHg 43 (40-47) 79 (78-80) 11 (10-12) 96 (95-96) 2.1 (1.9-2.3) 0.72 (0.67-0.76) 2417 398 10 *The FP/FN trade-off expresses the additional number of false positive results for one less false negative result that comes with increasing the systolic blood pressure threshold from 90 mmHg. [...]as the negative likelihood ratios range between 0.72 and 0.92, 7-day mortality cannot with confidence be ruled out on the basis of any of the suggested SBP thresholds. [...]SBP thresholds alone cannot, regardless of whether the threshold is low or high, rule out risk of death.