A Machine Learning Model to Predict Diuretic Resistance.
Journal
Kidney360
ISSN: 2641-7650
Titre abrégé: Kidney360
Pays: United States
ID NLM: 101766381
Informations de publication
Date de publication:
01 01 2023
01 01 2023
Historique:
received:
24
08
2022
accepted:
01
11
2022
entrez:
26
1
2023
pubmed:
27
1
2023
medline:
31
1
2023
Statut:
ppublish
Résumé
Volume overload is a common complication encountered in hospitalized patients, and the mainstay of therapy is diuresis. Unfortunately, the diuretic response in some individuals is inadequate despite a typical dose of loop diuretics, a phenomenon called diuretic resistance. An accurate prediction model that predicts diuretic resistance using predosing variables could inform the right diuretic dose for a prospective patient. Two large, deidentified, publicly available, and independent intensive care unit (ICU) databases from the United States were used-the Medical Information Mart for Intensive Care III (MIMIC) and the Philips eICU databases. Loop diuretic resistance was defined as <1400 ml of urine per 40 mg of diuretic dose in 24 hours. Using 24-hour windows throughout admission, commonly accessible variables were obtained and incorporated into the model. Data imputation was performed using a highly accurate machine learning method. Using XGBoost, several models were created using train and test datasets from the eICU database. These were then combined into an ensemble model optimized for increased specificity and then externally validated on the MIMIC database. The final ensemble model was composed of four separate models, each using 21 commonly available variables. The ensemble model outperformed individual models during validation. Higher serum creatinine, lower systolic blood pressure, lower serum chloride, higher age, and female sex were the most important predictors of diuretic resistance (in that order). The specificity of the model on external validation was 92%, yielding a positive likelihood ratio of 3.46 while maintaining overall discrimination (C-statistic 0.69). A diuretic resistance prediction model was created using machine learning and was externally validated in ICU populations. The model is easy to use, would provide actionable information at the bedside, and would be ready for implementation in existing electronic medical records. This study also provides a framework for the development of future machine learning models.
Sections du résumé
BACKGROUND
Volume overload is a common complication encountered in hospitalized patients, and the mainstay of therapy is diuresis. Unfortunately, the diuretic response in some individuals is inadequate despite a typical dose of loop diuretics, a phenomenon called diuretic resistance. An accurate prediction model that predicts diuretic resistance using predosing variables could inform the right diuretic dose for a prospective patient.
METHODS
Two large, deidentified, publicly available, and independent intensive care unit (ICU) databases from the United States were used-the Medical Information Mart for Intensive Care III (MIMIC) and the Philips eICU databases. Loop diuretic resistance was defined as <1400 ml of urine per 40 mg of diuretic dose in 24 hours. Using 24-hour windows throughout admission, commonly accessible variables were obtained and incorporated into the model. Data imputation was performed using a highly accurate machine learning method. Using XGBoost, several models were created using train and test datasets from the eICU database. These were then combined into an ensemble model optimized for increased specificity and then externally validated on the MIMIC database.
RESULTS
The final ensemble model was composed of four separate models, each using 21 commonly available variables. The ensemble model outperformed individual models during validation. Higher serum creatinine, lower systolic blood pressure, lower serum chloride, higher age, and female sex were the most important predictors of diuretic resistance (in that order). The specificity of the model on external validation was 92%, yielding a positive likelihood ratio of 3.46 while maintaining overall discrimination (C-statistic 0.69).
CONCLUSIONS
A diuretic resistance prediction model was created using machine learning and was externally validated in ICU populations. The model is easy to use, would provide actionable information at the bedside, and would be ready for implementation in existing electronic medical records. This study also provides a framework for the development of future machine learning models.
Identifiants
pubmed: 36700900
doi: 10.34067/KID.0005562022
pii: 02200512-202301000-00006
pmc: PMC10101605
doi:
Substances chimiques
Diuretics
0
Sodium Potassium Chloride Symporter Inhibitors
0
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
15-22Informations de copyright
Copyright © 2022 by the American Society of Nephrology.
Références
Zhang L, Chen Z, Diao Y, Yang Y, Fu P. Associations of fluid overload with mortality and kidney recovery in patients with acute kidney injury: a systematic review and meta-analysis. J Crit Care. 2015;30(4):860.e7–860.e13. doi: 10.1016/j.jcrc.2015.03.025
doi: 10.1016/j.jcrc.2015.03.025
Wang N, Jiang L, Zhu B, Wen Y, Xi XM. Fluid balance and mortality in critically ill patients with acute kidney injury: a multicenter prospective epidemiological study. Crit Care. 2015;19(1):371. doi: 10.1186/s13054-015-1085-4
doi: 10.1186/s13054-015-1085-4
Felker GM, Ellison DH, Mullens W, Cox ZL, Testani JM. Diuretic therapy for patients with heart failure: JACC state-of-the-art review. J Am Coll Cardiol. 2020;75(10):1178–1195. doi: 10.1016/j.jacc.2019.12.059
doi: 10.1016/j.jacc.2019.12.059
Gupta R, Testani J, Collins S. Diuretic resistance in heart failure. Curr Heart Fail Rep. 2019;16(2):57–66. doi: 10.1007/s11897-019-0424-1
doi: 10.1007/s11897-019-0424-1
Neuberg GW, Miller AB, O'Connor CM, et al. Diuretic resistance predicts mortality in patients with advanced heart failure. Am Heart J. 2002;144(1):31–38. doi: 10.1067/mhj.2002.123144
doi: 10.1067/mhj.2002.123144
Valente MAE, Voors AA, Damman K, et al. Diuretic response in acute heart failure: clinical characteristics and prognostic significance. Eur Heart J. 2014;35(19):1284–1293. doi: 10.1093/eurheartj/ehu065
doi: 10.1093/eurheartj/ehu065
Testani JM, Brisco MA, Turner JM, et al. Loop diuretic efficiency: a metric of diuretic responsiveness with prognostic importance in acute decompensated heart failure. Circ Heart Fail. 2014;7(2):261–270. doi: 10.1161/circheartfailure.113.000895
doi: 10.1161/circheartfailure.113.000895
Johnson AEW, Pollard TJ, Shen L, et al. MIMIC-III, a freely accessible critical care database. Sci Data. 2016;3(1):160035. doi: 10.1038/sdata.2016.35
doi: 10.1038/sdata.2016.35
Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Mark RG, Badawi O. The eICU collaborative research database, a freely available multi-center database for critical care research. Sci Data. 2018;5(1):180178. doi: 10.1038/sdata.2018.178
doi: 10.1038/sdata.2018.178
Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613–619. doi: 10.1016/0895-4356(92)90133-8
doi: 10.1016/0895-4356(92)90133-8
Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–1139. doi: 10.1097/01.mlr.0000182534.19832.83
doi: 10.1097/01.mlr.0000182534.19832.83
Zhang X, Yan C, Gao C, Malin BA, Chen Y. Predicting missing values in medical data via XGBoost regression. J Healthc Inform Res. 2020;4(4):383–394. doi: 10.1007/s41666-020-00077-1
doi: 10.1007/s41666-020-00077-1
Ter Maaten JM, Dunning AM, Valente MAE, et al. Diuretic response in acute heart failure—an analysis from ASCEND-HF. Am Heart J. 2015;170(2):313–321.e4. doi: 10.1016/j.ahj.2015.05.003
doi: 10.1016/j.ahj.2015.05.003
Aronson D, Burger AJ. Diuretic response: clinical and hemodynamic predictors and relation to clinical outcome. J Card Fail. 2016;22(3):193–200. doi: 10.1016/j.cardfail.2015.07.006
doi: 10.1016/j.cardfail.2015.07.006
Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 2016.
Nogueira F. Bayesian Optimization: Open Source Constrained Global Optimization Tool for Python. Accessed February 2021. https://github.com/fmfn/BayesianOptimization .
Lundberg SM, Allen PG, Lee S-I. A unified approach to interpreting model predictions. In: 31st Conference on Neural Information Processing Systems (NIPS). Long Beach, CA; 2017.
Nohara Y, Matsumoto K, Soejima H, Nakashima N. Explanation of machine learning models using shapley additive explanation and application for real data in hospital. Comput Methods Programs Biomed. 2022;214:106584. doi: 10.1016/j.cmpb.2021.106584
doi: 10.1016/j.cmpb.2021.106584
Li D, Liu Z, Armaghani DJ, Xiao P, Zhou J. Novel ensemble intelligence methodologies for rockburst assessment in complex and variable environments. Sci Rep. 2022;12(1):1844. doi: 10.1038/s41598-022-05594-0
doi: 10.1038/s41598-022-05594-0
Mohamadlou H, Lynn-Palevsky A, Barton C, et al. Prediction of acute kidney injury with a machine learning algorithm using electronic Health record data. Can J Kidney Health Dis. 2018;5:205435811877632. doi: 10.1177/2054358118776326
doi: 10.1177/2054358118776326
Shimabukuro DW, Barton CW, Feldman MD, Mataraso SJ, Das R. Effect of a machine learning-based severe sepsis prediction algorithm on patient survival and hospital length of stay: a randomised clinical trial. BMJ Open Respir Res. 2017;4(1):e000234. doi: 10.1136/bmjresp-2017-000234
doi: 10.1136/bmjresp-2017-000234
McCoy A, Das R. Reducing patient mortality, length of stay and readmissions through machine learning-based sepsis prediction in the emergency department, intensive care unit and hospital floor units. BMJ Open Qual. 2017;6(2):e000158. doi: 10.1136/bmjoq-2017-000158
doi: 10.1136/bmjoq-2017-000158
Voors AA, Davison BA, Teerlink JR, et al. Diuretic response in patients with acute decompensated heart failure: characteristics and clinical outcome—an analysis from RELAX-AHF. Eur J Heart Fail. 2014;16(11):1230–1240. doi: 10.1002/ejhf.170
doi: 10.1002/ejhf.170
Feola M, Rossi A, Testa M, et al. Six-month predictive value of diuretic resistance formulas in discharged heart failure patients after an acute decompensation. J Clin Med. 2020;9(9):2932. doi: 10.3390/jcm9092932
doi: 10.3390/jcm9092932
Doering A, Jenkins CA, Storrow AB, et al. Markers of diuretic resistance in emergency department patients with acute heart failure. Int J Emerg Med. 2017;10(1):17. doi: 10.1186/s12245-017-0143-x
doi: 10.1186/s12245-017-0143-x
Rao VS, Ivey-Miranda JB, Cox ZL, et al. Natriuretic equation to predict loop diuretic response in patients with heart failure. J Am Coll Cardiol. 2021;77(6):695–708. doi: 10.1016/j.jacc.2020.12.022
doi: 10.1016/j.jacc.2020.12.022
Ter Maaten JM, Damman K, Hanberg JS, et al. Hypochloremia, diuretic resistance, and outcome in patients with acute heart failure. Circ Heart Fail. 2016;9(8):e003109. doi: 10.1161/circheartfailure.116.003109
doi: 10.1161/circheartfailure.116.003109
Testani JM, Hanberg JS, Arroyo JP, et al. Hypochloraemia is strongly and independently associated with mortality in patients with chronic heart failure. Eur J Heart Fail. 2016;18(6):660–668. doi: 10.1002/ejhf.477
doi: 10.1002/ejhf.477
Broughton JS, Hanberg JS, Rao VS, et al. Hypochloremia and diuretic resistance in heart failure: mechanistic insights. J Card Fail. 2016;22(8):S8. doi: 10.1016/j.cardfail.2016.06.035
doi: 10.1016/j.cardfail.2016.06.035
Mangoni AA, Jackson SHD. Age-related changes in pharmacokinetics and pharmacodynamics: basic principles and practical applications. Br J Clin Pharmacol. 2003;57(1):6–14. doi: 10.1046/j.1365-2125.2003.02007.x
doi: 10.1046/j.1365-2125.2003.02007.x