Diabetologia: Using biomarkers and electronic patient data to derive and validate machine learning risk scores to predict the progression of diabetic nephropathy

Background introduction: Approximately one in four adults with type 2 diabetes has kidney disease (ie diabetic nephropathy [DKD])

The measurement of EGFR and urinary albumin-creatinine ratio (UACR) has been included in kidney disease: a risk stratification guideline for improving global prognosis (KDIGO)

Some blood biomarkers are related to the progression of DKD, the most notable of which are soluble TNF receptor 1/2 (TNFR1/2) and plasma kidney injury molecule-1 (Kim-1)

Soluble TNF receptor 1/2 (TNFR1/2) and plasma kidney injury molecule-1 (Kim-1) machine learning can combine biomarkers and EHR data to generate predictive risk scores

Method: This is a cohort study of two EHR-related organisms

^{- 1 - 2} _Event

Figure 1 Shapley additive interpretation (Shap) graph showing relative feature importance

Figure 1 Shapley additive interpretation (Shap) graph showing relative feature importance

Figure 2 (A) KidneyIntelX predicts the risk in the derived set, (B) KidneyIntelX predicts the risk in the validation set, and (C) predicts the distribution of KidneyIntelX scores in DKD patients based on the risk of the compound renal endpoint in the derived and validation set

Figure 2 (A) KidneyIntelX predicts the risk in the derived set, (B) KidneyIntelX predicts the risk in the validation set, and (C) predicts the distribution of KidneyIntelX scores in DKD patients based on the risk of the compound renal endpoint in the derived and validation set

Table KidneyIntelX test characteristics and comprehensive clinical model

Table KidneyIntelX test characteristics and comprehensive clinical model

Figure 3 The Kaplan-Meier curve of KidneyIntelX risk stratification shows that in the derived (A) and validated (B) sets, EGFR continues to decline by 40% or the end point of renal failure

Figure 3 The Kaplan-Meier curve of KidneyIntelX risk stratification shows that in the derived (A) and validated (B) sets, EGFR continues to decline by 40% or the end point of renal failure

Compared with KDIGO and clinical models, Kidneyintelx can better predict the renal prognosis of patients with early DKD

 Chan L, Nadkarni GN, Fleming F,et al, Derivation and validation of a machine learning risk score using biomarker and electronic patient data to predict progression of diabetic kidney disease.

Derivation and validation of a machine learning risk score using biomarker and electronic patient data to predict progression of diabetic kidney disease.

 

Leave a message here