RESUMEN
BACKGROUND: ML predictive models have shown their capability to improve risk prediction and assist medical decision-making, nevertheless, there is a lack of accuracy systems to early identify future rapid CKD progressors in Colombia and even in South America. OBJECTIVE: The purpose of this study was to develop a series of interpretable machine learning models that predict GFR at 6-months, 9-months, and 12-months. STUDY DESIGN AND SETTING: Over 29,000 CKD patients stage 1 to 3b (estimated GFR, <60 mL/min/1.73 m2) with an average of 3-year follow-up data were included. We used the machine learning extreme gradient boosting (XGBoost) to build three models to predict the next eGFR. Models were internally and externally validated. In addition, we included SHapley Additive exPlanation (SHAP) values to offer interpretable global and local prediction models. RESULTS: All models showed a good performance in development and external validation. However, the 6-months XGBoost prediction model showed the best performance in internal (MAE average = 6.07; RSME = 78.87), and in external validation (MAE average = 6.45, RSME = 18.94). The top 3 most influential features that pushed the predicted eGFR value to lower values were the interpolated values for eGFR and creatinine, and eGFR at baseline. CONCLUSION: In the current study we have developed and validated machine learning models to predict the next eGFR value at different intervals. Furthermore, we attempted to approach the need for prediction explanation by offering transparent predictions.
RESUMEN
Contexto: la enfermedad renal crónica (ERC) es un problema de salud pública mundial, con una prevalencia creciente y una carga sustancial sobre los sistemas de salud. Las características de los pacientes progresores rápidos junto con los factores que podrían predecir la progresión acelerada requieren una mayor exploración, sobre todo porque en Colombia no hay estudios publicados de este tipo. Objetivo: identificar los factores clínicos y sociodemográficos asociados con la progresión rápida de la ERC en una gran cohorte de pacientes de la costa norte colombiana. Metodología: este estudio observacional retrospectivo incluyó dos cohortes de pacientes de la costa norte colombiana con estadios de G1 a G4 sin diálisis. El seguimiento se realizó durante 18 meses (n = 14.420) y 24 meses (n = 10.042) y los pacientes se dividieron en progresores rápidos y progresores estables, según la pendiente de la tasa de filtración glomerular (TFGe) < -5 ml/min/1,73 m2/año o ≥ -5 mL/min/1,73 m2/año, respectivamente. Resultados: el 37,68 % de la cohorte se clasificó como progresión rápida a los 24 meses y el 28,41 % a los 18 meses. Se encontró un R2 = 0,77 en ambas cohortes para las mediciones de la TFGe, lo que indica una tendencia hacia una pérdida lineal. Además, la progresión acelerada se observó en pacientes más jóvenes y con valores más altos de tensión arterial diastólica. Conclusiones: en dos grandes cohortes de pacientes renales, aproximadamente 4 de cada 10 pacientes se clasificaron como progresores rápidos a los 24 meses de seguimiento y 3 de cada 10 a los 18 meses de seguimiento. Es importante identificar a los pacientes con mayor riesgo de progresión acelerada de la ERC en el corto plazo, para brindarles una atención personalizada y efectiva.
Background: Chronic Kidney Disease (CKD) is a global public health problem with increasing prevalence and a substantial burden on healthcare systems. The characteristics of rapid progressors and factors that may predict accelerated CKD progression require further exploration, especially since there are no published studies about this topic in Colombia. Purpose: To identify the clinical and sociodemographic factors associated with the rapid progression of CKD in a large cohort of Colombian patients. Methods: This retrospective observational study included two cohorts of G1 to G4 stage CKD patients without dialysis from the Colombian North Coast. Follow-up was conducted for 18 months (n=14,420) and 24 months (n=10,042). Patients were divided into rapid progressors and stable progressors based on the glomerular filtration rate (GFR) slope <−5 mL/min/1.73m2/year or ≥-5 mL/min/1.73 m2/year, respectively. Results: 37.68% of the cohort was classified as rapid progressors at 24 months and 28.41% at 18 months. An R2=0.77 was found in both cohorts for GFR measurements, indicating a linear trend toward loss. Furthermore, accelerated progression was observed in younger patients and those with higher diastolic blood pressure values. Conclusion: In two large cohorts of renal patients, approximately 4 out of 10 patients were classified as rapid progressors at 24 months of follow-up and 3 out of 10 at 18 months of follow-up. It is important to identify patients at higher risk of accelerated CKD progression in the short term to provide them with personalized and effective care.