An artificial neural network can select patients at high risk of developing progressive IgA nephropathy more accurately than experienced nephrologists

Nephrol Dial Transplant. 1998 Jan;13(1):67-71. doi: 10.1093/ndt/13.1.67.


Background: The object of the study was to develop an artificial neural network (ANN) to identify patients with IgA nephropathy (IgAN) with a poor prognosis and to compare the predictions of the ANN with the predictions of six experienced nephrologists.

Methods: The following data from the time of renal biopsy were retrieved from the records of 54 patients with IgAN: age, sex, systolic and diastolic blood pressure, number of prescribed antihypertensive drugs, 24-h urine protein excretion, and serum creatinine. Patients aged less than 14 years, or who had serum creatinine > 350 mumol/l at presentation, liver disease or concomitant kidney disease were excluded. Outcome was assigned as 'stable' if serum creatinine was < 150 mumol/l after 7 years and 'non-stable' if serum creatinine was > or = 150 mumol/l. The ANN was trained and tested using a 'jack-knife' sampling technique and performance evaluated in terms of number of correct predictions, sensitivity and specificity. The six nephrologists were asked to predict outcome at 7 years for each patient using the same data as the ANN and their performance was assessed in the same manner.

Results: The ANN assigned the correct outcome to 47/54 (87.0%) patients: sensitivity 19/22 (86.4%), specificity 28/32 (87.5%). The mean score for nephrologists was 37.5/54 (69.4%, range 35-40), mean sensitivity 72% and mean specificity 66%.

Conclusions: An ANN trained using routine clinical information obtained at the time of diagnosis can potentially predict 7-year outcome for renal function in IgAN more accurately than experienced nephrologists, and can therefore identify a group of high-risk patients requiring close follow-up.

MeSH terms

  • Adolescent
  • Adult
  • Aged
  • Female
  • Glomerulonephritis, IGA / diagnosis*
  • Humans
  • Male
  • Middle Aged
  • Nephrology
  • Neural Networks, Computer*
  • Sensitivity and Specificity