Table 4.

Negative and positive predictive values for the ACT model and outcome of severe AKI within 72 hours^a

Cohort	Performance Goal	Patients Above/Below Threshold, n	Proportion of Patients with Severe AKI, n (%)	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
Derivation	Maximizing NPV	35/714	62/749 (8)	0.47 (0.35 to 0.58)	0.99 (0.98 to 0.99)	0.83 (0.71 to 0.94)	0.95 (0.94 to 0.96)
	Maximizing PPV	21/728	62/749 (8)	0.32 (0.21 to 0.44)	0.99 (0.99 to 1.00)	0.95 (0.86 to 1.00)	0.94 (0.93 to 0.95)
Internal validation	Maximizing NPV	16/310	26/326 (8)	0.46 (0.27 to 0.65)	0.99 (0.97 to 0.99)	0.75 (0.55 to 0.93)	0.95 (0.94 to 0.97)
	Maximizing PPV	9/317	26/326 (8)	0.27 (0.11 to 0.42)	0.99 (0.98 to 1.00)	0.80 (0.50 to 1.00)	0.94 (0.93 to 0.95)
External validation	Maximizing NPV	11/245	15/262 (6)	0.40 (0.20 to 0.67)	0.98 (0.96 to 0.99)	0.55 (0.29 to 0.86)	0.96 (0.95 to 0.98)
	Maximizing PPV	2/254	15/262 (6)	0.07 (0.00 to 0.20)	0.99 (0.99 to 1.00)	0.50 (0.00 to 1.00)	0.94 (0.94 to 0.95)

PPV is the probability of those with a predicted probability above the threshold having stage 2 or 3 AKI. NPV is the probability of those with a predicted probability below the threshold not having stage 2 or 3 AKI. ACT, age, cirrhosis, and sTNFR-1; PPV, positive prediction value; 95% CI, 95% confidence interval; NPV negative prediction value; sTNFR-1, soluble TNR receptor-1.

^aACT predicted probability for an individual is obtained from model using age, cirrhosis, and sTNFR-1 fit to the derivation dataset: (probability of severe AKI within 72 hours after biomarker measurement)=expit [−35.7–0.04×age −2.59×I (cirrhosis) +8.51×log₁₀ (TNFR-1 concentration)], where expit(x) = exp(x)/[1+exp(x)].