Skip to main content

Advertisement

Table 1 Median chance-corrected concordance (%), cause-specific mortality fraction accuracy for 6 methods across 500 splits by age and health care experience

From: Using verbal autopsy to measure causes of death: the comparative performance of existing methods

  Tariff SSP
CCC CSMF accuracy CCC CSMF accuracy
Median 95% CI Median 95% CI Median 95% CI Median 95% CI
Adult No HCE 37.8 (37.6, 37.9) 0.717 (0.711, 0.721) 41.4 (41.1, 41.6) 0.715 (0.709, 0.720)
HCE 50.5 (50.2, 50.7) 0.77 (0.766, 0.775) 54.1 (53.9, 54.3) 0.764 (0.760, 0.769)
Child No HCE 44.6 (44.2, 45.0) 0.744 (0.736, 0.751) 44.9 (44.5, 45.2) 0.74 (0.735, 0.746)
HCE 52.5 (52.1, 53.0) 0.783 (0.776, 0.786) 52.1 (51.7, 52.4) 0.768 (0.762, 0.774)
Neonate No HCE 44.6 (44.2, 44.9) 0.809 (0.801, 0.817) 48.2 (47.9, 48.6) 0.778 (0.768, 0.787)
HCE 47.3 (46.9, 47.7) 0.817 (0.809, 0.824) 50.3 (50.1, 50.7) 0.79 (0.781, 0.801)
Cont’d
  RF PCVA
CCC CSMF accuracy CCC CSMF accuracy
Median 95% CI Median 95% CI Median 95% CI Median 95% CI
Adult No HCE 37.2 (37.0, 37.4) 0.708 (0.705, 0.712) 29.1 (28.9, 29.3) 0.638 (0.632, 0.644)
HCE 49.2 (49.0, 49.4) 0.769 (0.766, 0.774) 42.2 (41.8, 42.5) 0.68 (0.673, 0.687)
Child No HCE 44.6 (44.1, 44.9) 0.715 (0.706, 0.721) 33.5 (33.2, 33.7) 0.63 (0.615, 0.637)
HCE 50.2 (49.7, 50.5) 0.739 (0.729, 0.748) 44.5 (44.1, 45.2) 0.678 (0.671, 0.685)
Neonate No HCE 47.2 (46.9, 47.5) 0.769 (0.759, 0.779) 25.5 (25.2, 26.0) 0.692 (0.682, 0.701)
HCE 50 (49.8, 50.4) 0.793 (0.779, 0.802) 29.3 (28.7, 29.7) 0.719 (0.707, 0.734)
Cont’d
  KL InterVA
CCC CSMF accuracy CCC CSMF accuracy
Median 95% CI Median 95% CI Median 95% CI Median 95% CI
Adult No HCE    0.672 (0.667, 0.676) 23.4 (23.3, 23.6) 0.611 (0.605, 0.620)
HCE    0.688 (0.682, 0.692) 23.8 (23.6, 24.0) 0.625 (0.617, 0.632)
Child No HCE    0.674 (0.668, 0.680) 29.6 (29.3, 29.9) 0.514 (0.506, 0.526)
HCE    0.69 (0.683, 0.696) 30.3 (30.0, 30.6) 0.52 (0.510, 0.529)
Neonate No HCE    0.808 (0.796, 0.817) 20 (19.8, 20.4) 0.627 (0.602, 0.641)
HCE    0.81 (0.799, 0.819) 19.4 (19.2, 19.8) 0.629 (0.606, 0.648)
King-Lu (KL) does not estimate individual causes so chance-corrected concordance and Cohen's kappa cannot be calculated.
  1. Table 1 shows chance-corrected concordance (CCC), and cause-specific mortality fraction (CSMF) accuracy for all methods across 500 splits by age and health care experience for adults, children, and neonates. CI, confidence interval; KL, King-Lu; PCVA, physician-certified VA; RF, Random Forest; SSP, Simplified Symptom Pattern; VA, verbal autopsy.