A back propagation neural network approach to estimate the glomerular filtration rate in an older population

BMC Geriatrics

Table 3 Pairwise comparisons between BPNN and other GFR estimation equations

Pairwise comparisons	Difference in bias^a	p	Difference in P30^b	p
Pairwise comparisons	Median (95% CI)	p	Median (95% CI)	p
BPNN–CKD-EPI	2.19 (1.05 to 3.56)	0.31	13.52 (5.61 to 21.44)	< 0.001
BPNN–EKFC	-1.41 (-0.09 to 1.01)	0.26	9.84 (2.04 to 17.63)	0.008
BPNN–BIS1	0.64 (-0.09 to 1.01)	0.99	4.10 (-3.45 to 11.64)	0.98
BPNN–LMR	-2.42 (-2.86 to -1.93)	0.03	11.07 (3.22 to 18.91)	0.001
BPNN–Asian modified CKD-EPI	4.75 (3.78 to 6.32)	0.02	15.98 (8.0 to 23.97)	< 0.001
BPNN–MDRD	1.11 (-0.03 to 2.67)	0.45	14.34 (6.4 to 22.29)	< 0.001

GFR, glomerular filtration rate; CI, confidence interval; CKD-EPI, Chronic Kidney Disease-Epidemiology equation; EKFC, European Kidney Function Consortium equation; BIS1, Berlin Initiative Study-1 equation; LMR, Lund-Malmö Revised equation; BPNN, Back propagation neural network model; MDRD, Modification of Diet in Renal Disease equation. The 95% CIs were calculated by a bootstrap method (1000 bootstraps) over all measure differences
^aMedian difference in bias is the difference between equation biases (measured GFR minus estimated GFR), and a positive median difference means that, on average, the second equation provides higher values of estimated GFR than the first one
^bP30 represents the percentage of estimated GFR within 30% of measured GFR; a positive median difference means that, on average, more values from the first equation are close to measured GFR.

ISSN: 1471-2318