Applications of machine learning for hemodialysis nursing cares based on a machine learning algorithm

Article Type : Original/Research Papers

Authors

1 Department of Immunology, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran

2 Department of Physiology, School of Medicine, Guilan University of Medical Sciences, Rasht, Iran

3 Student Research Committee, School of Nursing and Midwifery, Golestan University of Medical Sciences, Gorgan, Iran

4 Department of Physiology, School of Medicine, Cellular and the Molecular Research Center, Guilan University of Medical Sciences, Rasht, Iran

Abstract

Nursing care during dialysis involves managing symptoms and preventing complications among patients undergoing hemodialysis or peritoneal dialysis. In this regard, to improve the quality of nursing care during dialysis, several approaches were developed to enhance hemodialysis adequacy and prevent complications; however, machine learning (ML) emerged as a methodological approach for evaluating hemodialysis adequacy and complications. The current study aims to analyze ML approach in predicting and managing hemodialysis by R programming language analysis to provide a therapeutic concept for hemodialysis management in critical nursing care. An R programming language was used to perform the logical analysis of the data. ML algorithms based on usage rate included logistic regression (LR), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Complement Naive Bayes (CNB), Takagi-Sugeno-Kang fuzzy system (G-TSK-FS), k-nearest neighbors' classifier (KNN), Stochastic gradient descent (SGD), Linear Discriminant Analysis (LDA), and Multi-adaptive neural-fuzzy system (MANFIS). Also, the use of ML in nursing care during hemodialysis is categorized into three indications for predicting hemodialysis adequacy, complications, and vascular access performance. Using ML in hemodialysis nursing care is a growing research interest. The main application areas are the prediction of hemodialysis adequacy, complications, and vascular access performance. LR and SVM are practical ML algorithms for constructing AI tools to improve hemodialysis management.

Keywords

1 Introduction

Chronic kidney disease (CKD) is a progressive disease characterized by structural and functional changes in the kidney due to various factors [1]. CKD is generally a decline in renal function, an estimated glomerular filtration rate (eGFR) less than 60 mL/min per 1.73 m2, or evidence of kidney damage, such as albuminuria and hematuria. Based on epidemiological evidence, approximately 10% of adults worldwide suffer from some form of CKD, which results in 1.2 million deaths annually [2]. Furthermore, CKD is expected to become the fifth leading cause of death by 2040 [3]. To manage kidney failure, renal replacement therapy (RRT) (functional or anatomical) through chronic dialysis or kidney transplantation is considered a stable treatment. In contrast, dialysis is the dominant treatment option for most people with kidney failure due to the lack of kidney donors and the comorbidities that occur with aging and often prevent kidney transplantation [4]. Despite the recent technological advancements in dialysis equipment, the dialysis process may present various challenges for patients [5].

In nursing care, machine learning (ML) is a branch of artificial intelligence which increasingly used to determine diagnoses, complications, prognoses, and recurrences [6]. Unlike conventional statistical models, ML can actively learn complex relationships between data, overcoming the limitations of nonlinearity and maintaining stability even in high-dimensional datasets [7]. ML can provide a unique advantage in analyzing unstructured data, including pictures and other forms of information [8]. Despite this, several problems relating to model construction are still observed in many ML studies [7]. Notwithstanding the excellent performance of models on local datasets, many researchers have failed to consider their reproducibility in other clinical environments, limiting the further promotion of this powerful decision-support tool in clinical practice [9].

Nursing care during dialysis involves managing symptoms and preventing complications among patients undergoing hemodialysis or peritoneal dialysis [10]. The nurses are also responsible for assessing vital signs, fluid balance, electrolyte levels and monitoring the vascular access site for complications [11]. Additionally, nurses may administer medications and monitor the patient's condition for adverse reactions, such as bleeding or hypotension [12]. In this regard, to improve the quality of nursing care during dialysis, several approaches were developed to enhance hemodialysis adequacy and complication prevention; however, ML emerged as a methodological approach for evaluating hemodialysis adequacy and complications [13]. In addition, ML is useful for learning from data iteratively and identifying patterns while minimizing human intervention and user input bias during or after the hemodialysis procedure [14]. Based on the author's knowledge, former studies did not adequately describe ML-based prediction tasks involving hemodialysis nursing care. To summarize the ML applications for hemodialysis managements, an analysis of the model construction process's advantages and disadvantages must be analyzed. The current study aims to analyze the ML approach in predicting and managing hemodialysis by R programming language analysis to provide a therapeutic concept for hemodialysis management in critical nursing care.

 

2 Methods

2.1 Data selection

In the conducted study, relevant studies were extracted via the Preferred Reporting Items for Systematic Reviews and Meta-Analyzes (PRISMA) guideline [15]. For the period January 2023 through May 2023, a comprehensive search of the Scopus, PubMed, and Web of Science databases was conducted using relevant keywords, including " hemodialysis ", "Nursing," and "machine learning". Two researchers conducted searches independently using Boolean "AND" and "OR" operators to associate keywords. All types of dialysis cases were included. Also, the references of obtained studies were searched to prevent data loss, and 42 correlated cases were found. As the final step, papers "do not have ML applications" and cases written before 2010 were excluded. Ultimately, seven studies were extracted.

 

2.2 R Programming language plot

A plot was created using the R programming language.  As a programming language, R is an application for statistical computing and graphics supported by R Foundation for Statistical Computing and the R Core Team. The R programming language was created by statisticians Ross Ihaka and Robert Gentleman to analyze data and design statistical software [16].

 

2.3 Sankey plot

The "Sankey plot" was also used to illustrate ML's technical uses during hemodialysis nursing care. An indicator twice as broad represents twice the quantity visualized on the Sankey plot. Flow diagrams can demonstrate the flow of energy, materials, water, or costs. To illustrate directed flow, at least two nodes (processes) must be drawn on a Sankey chart. Consequently, the Sankey plot provides information about values, system structure, and distribution. Due to this, they are a great alternative to standard flow charts and bar charts [17].

 

3 Results

3.1 ML algorithm

ML algorithms based on usage rate included logistic regression (LR), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Complement Naive Bayes (CNB), Takagi-Sugeno-Kang fuzzy system (G-TSK-FS), k-nearest neighbors' classifier (KNN), Stochastic gradient descent (SGD), Linear Discriminant Analysis (LDA) and Multi-adaptive neural-fuzzy system (MANFIS) (Figure 1).

 

3.2 Application of ML in hemodialysis nursing cares

In general, the use of ML in nursing care during hemodialysis is categorized into three indications for predicting hemodialysis adequacy (determining Urea reduction ratio (URR) and KT/V) [18-20], predicting hemodialysis complications (Intradialytic hypotension and cerebral hemorrhage) [21-23], and predicting vascular access performance (detect access flow dysfunction in Arteriovenous Fistula) [24] (Table 1).

Figure 1. Sanky plot of extracted data, the first column indicates the type of ML algorithm, the second column shows the predicting indications, and the third column reveals the predicting criteria; note the thickness of the connection lines between the columns.

 

Table 1. The application of ML in hemodialysis nursing care.

Algorithm

Indication

Use

References

G-TSK-FS

Predicting HD adequacy

Predicting KT/V

[19]

LEM2

Predicting URR

[20]

MANFIS

[18]

SVM

Vascular Access

Detect access flow dysfunction in Arteriovenous Fistula

[24]

LR

Prediction of complications

Predict the risk of a cerebral hemorrhage.

[21]

CNB

KNN

SVM

XGB

LR

Prediction of Intradialytic Hypotension

[22]

RF

prediction of Intradialytic hypotension

[23]

XGB

LR

SGD

LDA

 

4 Discussion

ML has gradually spread across various disciplines due to the development of AI and computer technology. Additionally, ML was used as a diagnostic tool in several studies. Although ML was applied to a wide range of nursing topics, less research has been conducted. By analyzing "R" the results of the seven papers, we identified the most popular ML algorithm used in nursing management during dialysis, enabling us to develop high-quality predictive targets for future research.

The present study identified LR and SVM as the most commonly used ML algorithms for nursing care in hemodialysis management. A classification algorithm based on ML, LR predicts certain classes based on dependent variables [25]. In essence, the LR model calculates the logistic result based on the input features [26]. Also, LR is used to analyze binary and ordinal data in medical research. LR has also been shown to help estimate and evaluate the relative risks of rare events in cross-sectional and longitudinal studies [27]. Suzuki et al. have applied LR algorithms for cardiovascular prognosis prediction during hemodialysis and demonstrated the ability of ML to identify risk-predicting models with good predictive capability and good discrimination of the risk impact [28]. Additionally, KHAZAEI et al. used various ML algorithms, such as LR and neural networks, to predict mortality during hemodialysis and have shown that compared to other methods for predicting death among hemodialysis patients; LR had the best performance [29]. However, Montemayor et al. have shown that compared with LR, RF is a better method for generating mortality prediction models among hemodialysis patients [30].

SVMs are used to classify linear and non-linear systems. The quality and complexity of SVM solutions are indirectly influenced by the dimension of the input space. The SVM has several characteristics. It is capable of working in high dimensions, efficient in memory usage, effective even if the number of samples is minor than dimensional spaces, and has a variety of kernels for making a decision, including one that can be customized (tricky). Several critical diseases, including CKD, can be diagnosed and predicted using SVM. Data analysis using SVM can replace the human weakness of finding hidden patterns in data [31]. Martínez et al., using the SVM algorithm, predicted the hemoglobin level in hemodialysis patients with acceptable accuracy [32]. Further, Radović et al., using SVM algorithms, realized the prediction of mortality rate during hemodialysis with an accuracy of up to 94.12% and up to 96.77% [33].

Furthermore, the present study applied ML to predict three indications: hemodialysis adequacy, hemodialysis complications, and predicting vascular access performance. The ultrafiltration rate (UFR) prediction is an area where ML is beneficial [34]. Hemodialysis adequacy is determined by UFR, which measures fluid removal during hemodialysis [34]. In order to develop predictive models for UFR for individual patients, ML algorithms can be trained on large datasets of patient information [35]. Predicting complications during hemodialysis is another area where ML can be helpful [36]. To identify patients at risk of complications from hemodialysis, ML algorithms analyze patient data, including lab results and vital signs [21].

Also, based on evidence, in order to identify early signs of vascular access complications, ML algorithms can be used to analyze real-time data from vascular access monitoring systems [37]. ML can also be challenging in predicting hemodialysis adequacy. Data quality and availability are major challenges. Hemodialysis data is often stored in disparate systems and is complex to collect. Errors in data can also lead to inaccurate predictions [38]. The complexity of ML algorithms is another challenge. ML models can be difficult to interpret, making it difficult for clinicians to understand what drives predictions [39].

 

4.1 Limitations

The present study is the first technical study on the application of ML for hemodialysis patients. Therefore, it is inevitable that the study will contain limitations. Firstly, only English-language publications since 2010 were considered, resulting in a publication bias. In addition, because the overall quality assessment of the studies was poor, the results of the review may have been somewhat biased. Finally, the present study focuses on the technical application of ML in hemodialysis analysis, but valuable information may also exist in other fields, such as clinical research.

 

4.2 Implications for nursing clinical practice

The ability to personalize treatment plans is one of the critical implications for clinical practice. ML can identify factors contributing to hemodialysis adequacy and vascular access performance by analyzing patient data, enabling healthcare providers to customize treatment plans according to each patient's needs. Based on a patient's medical history, laboratory values, and other data points, ML can help determine the optimal dialysis duration or UFR. As a result of this personalized approach, patient outcomes can be improved, and complications can be reduced. Identifying patients at risk of complications is another potential benefit. An ML algorithm can analyze patient data to identify factors that may increase the risk of vascular access complications or inadequate hemodialysis. Healthcare providers can reduce the risk of complications by identifying these patients early, such as adjusting dialysis parameters, initiating early treatment, or changing medication regimens.

 

4.3 Recommendations for future research

According to the collected data, ML could be applied to hemodialysis nursing care with promising results. For ML to qualify as a practical assistant for managing hemodialysis, further technology development is necessary. Furthermore, the following research objectives may be considered for future studies: 1) Identifying the most proper complementary algorithms to develop the ML-Based concept for hemodialysis nursing care; 2) Development of chemical-based AI tools for hemodialysis nursing care.

 

5 Conclusions

The application of ML for hemodialysis nursing care can be regarded as a research target of interest in recent years. The prediction of hemodialysis adequacy, complications, and vascular access performance are three significant fields for using ML in hemodialysis management. Furthermore, the current data suggest that the ML algorithm, including LR and SVM, can be a suitable platform for constructing AI tools to improve hemodialysis management among CKD patients. Nonetheless, data management, pre-processing, and model validation still require improvement in order to build practical models that can be applied to clinical approaches.

 

Acknowledgements

Not applicable.

 

Authors’ contributions

Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work: MZ, SR, YM, FA, VG, SK, MA; Drafting the work or revising it critically for important intellectual content: MZ, SR, YM, FA, VG, SK, MA; Final approval of the version to be published: MZ, SR, YM, FA, VG, SK, MA; Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: MZ, SR, YM, FA, VG, SK, MA.

 

Funding

Self-funded.

 

Ethics approval and consent to participate

Not applicable.

 

Competing interests

We do not have potential conflicts of interest with respect to the research, authorship, and publication of this article.

 

Availability of data and materials

The datasets used during the current study are available from the corresponding author on request.

 

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (CC BY-NC 4.0).

© 2023 The Author(s).

  1. Webster AC, Nagler EV, Morton RL, Masson P. Chronic kidney disease. Lancet. 2017;389(10075):1238-1252.
  2. Kawada T. RE: epidemiology of chronic kidney disease of undetermined aetiology in Taiwanese farmers. Occup Environ Med. 2022;79(2):143.
  3. Foreman KJ, Marquez N, Dolgert A, Fukutaki K, Fullman N, McGaughey M, et al. Forecasting life expectancy, years of life lost, and all-cause and cause-specific mortality for 250 causes of death: reference and alternative scenarios for 2016–40 for 195 countries and territories. Lancet. 2018;392(10159):2052-2090.
  4. Chen Y-A, Ou S-M, Lin C-C. Influence of dialysis membranes on clinical outcomes: From history to innovation. Membranes. 2022;12(2):152.
  5. Reyes O. Effects of arterial needle placement in arteriovenous fistula on dialysis adequacy of end-stage renal disease patients undergoing maintenance hemodialysis. J Nur Healthcare. 2016;1(2):1-10.
  6. Zhou Y, Yang X, Ma S, Yuan Y, Yan M. A systematic review of predictive models for hospital‐acquired pressure injury using machine learning. Nurs Open. 2023;10(3):1234-1246.
  7. Mangold C, Zoretic S, Thallapureddy K, Moreira A, Chorath K, Moreira A. Machine learning models for predicting neonatal mortality: a systematic review. Neonatology. 2021;118(4):394-405.
  8. De Silva K, Mathews N, Teede H, Forbes A, Jönsson D, Demmer RT, et al. Clinical notes as prognostic markers of mortality associated with diabetes mellitus following critical care: a retrospective cohort analysis using machine learning and unstructured big data. Comput Biol Med. 2021;132:104305.
  9. Cabitza F, Campagner A. The need to separate the wheat from the chaff in medical informatics: Introducing a comprehensive checklist for the (self)-assessment of medical AI studies. Int J Med Inform. 2021;153:104510.
  10. F. Nunes A, editor. Multidisciplinary Experiences in Renal Replacement Therapy [Internet]. IntechOpen; 2022. Available from: http://dx.doi.org/10.5772/intechopen.90960.
  11. Elliott M, Coventry A. Critical care: the eight vital signs of patient monitoring. Br J Nurs. 2012;21(10):621-625.
  12. Rozich J, Haraden C, Resar R. Adverse drug event trigger tool: a practical methodology for measuring medication related harm. Qual Saf Health Care. 2003;12(3):194-200.
  13. Mezzatesta S, Torino C, De Meo P, Fiumara G, Vilasi A. A machine learning-based approach for predicting the outbreak of cardiovascular diseases in patients on dialysis. Comput Methods Programs Biomed. 2019;177:9-15.
  14. Akbilgic O, Obi Y, Potukuchi PK, Karabayir I, Nguyen DV, Soohoo M, et al. Machine learning to identify dialysis patients at high death risk. Kidney Int Rep. 2019;4(9):1219-1229.
  15. Moher D, Altman DG, Liberati A, Tetzlaff J. PRISMA statement. Epidemiology. 2011;22(1):128.
  16. Gardener M. Beginning R: the statistical programming language: John Wiley & Sons; 2012.
  17. Chen Q, Ning Z, Liu Z, Zhou Y, He O, Tian Y, et al. Textbook outcome as a measure of surgical quality assessment and prognosis in gastric neuroendocrine carcinoma: A large multicenter sample analysis. Chin J Cancer Res. 2021;33(4):433-446.
  18. Azar AT. Multi-adaptive neural-fuzzy system as a novel predictor of in-vivo blood side dialyser urea clearance. Int J Comput Appl Technol. 2013;46(2):77-90.
  19. Du A, Shi X, Guo X, Pei Q, Ding Y, Zhou W, et al. Assessing the adequacy of hemodialysis patients via the graph-based Takagi-Sugeno-Kang fuzzy system. Comput Math Methods Med. 2021;2021:1-11.
  20. Chen Y-S. Modeling hybrid rough set-based classification procedures to identify hemodialysis adequacy for end-stage renal disease patients. Comput Biol Med. 2013;43(10):1590-1605.
  21. Li F, Chen A, Li Z, Gu L, Pan Q, Wang P, et al. Machine learning-based prediction of cerebral hemorrhage in patients with hemodialysis: A multicenter, retrospective study. Front Neurol. 2023;14:1139096.
  22. Lin C-J, Chen C-Y, Wu P-C, Pan C-F, Shih H-M, Huang M-Y, et al. Intelligent system to predict intradialytic hypotension in chronic hemodialysis. J Formos Med Assoc. 2018;117(10):888-893.
  23. Hong D, Chang H, He X, Zhan Y, Tong R, Wu X, et al. Construction of an Early Alert System for Intradialytic Hypotension Before Initiating Hemodialysis Based on Artificial Intelligence. Available at SSRN 4019260.
  24. Chen C-H, Tao T-H, Chou Y-H, Chuang Y-W, Chen T-B. Arteriovenous fistula flow dysfunction surveillance: early detection using pulse radar sensor and machine learning classification. Biosensors. 2021;11(9):297.
  25. Stephan D, Lucila O-M. Logistic regression and artificial neural network classification models: a methodology review. J Biomed Inform. 2002;35(5-6):352-359.
  26. Bender R, Grouven U. Ordinal logistic regression in medical research. J R Coll Physicians Lond. 1997;31(5):546-551.
  27. Lumley T, Kronmal R, Ma S. Relative risk regression in medical research: models, contrasts, estimators, and algorithms. 2006. UW Biostatistics Working Paper Series. Working Paper 293.
  28. Suzuki S, Yamashita T, Sakama T, Arita T, Yagi N, Otsuka T, et al. Comparison of risk models for mortality and cardiovascular events between machine learning and conventional logistic regression analysis. PLoS One. 2019;14(9):e0221911.
  29. Khazaei S, Najafi-Gh OS, Ramezani-Doroh V. Construction data mining methods in the prediction of death in hemodialysis patients using support vector machine, neural network, logistic regression and decision tree. J Prev Med Hyg. 2021;62(1):e222-e230.
  30. Garcia-Montemayor V, Martin-Malo A, Barbieri C, Bellocchio F, Soriano S, Pendon-Ruiz de Mier V, et al. Predicting mortality in hemodialysis patients using machine learning analysis. Clin Kidney J. 2020;14(5):1388-1395.
  31. Cheng-Hong Y, Novaliendry D, Jin-Bor C, Renyaan AS, Lizar Y, Guci A, et al. Prediction of mortality in the hemodialysis patient with diabetes using support vector machine. Rev Argentina de Clin Psicol. 2020;29(4):219-232.
  32. Martínez-Martínez JM, Escandell-Montero P, Barbieri C, Soria-Olivas E, Mari F, Martínez-Sober M, et al. Prediction of the hemoglobin level in hemodialysis patients using machine learning techniques. Comput Methods Programs Biomed. 2014;117(2):208-217.
  33. Radović N, Prelević V, Erceg M, Antunović T. Machine learning approach in mortality rate prediction for hemodialysis patients. Comput Methods Biomech Biomed Engin. 2022;25(1):111-122.
  34. Kim HW, Heo S-J, Kim JY, Kim A, Nam C-M, Kim BS. Dialysis adequacy predictions using a machine learning method. Sci Rep. 2021;11(1):15417.
  35. Liu Y-S, Yang C-Y, Chiu P-F, Lin H-C, Lo C-C, Lai AS-H, et al. Machine learning analysis of time-dependent features for predicting adverse events during hemodialysis therapy: model development and validation study. J Med Internet Res. 2021;23(9):e27098.
  36. Othman M, Elbasha AM, Naga YS, Moussa ND. Early prediction of hemodialysis complications employing ensemble techniques. Biomed Eng Online. 2022;21(1):74.
  37. Khavanin Zadeh M, Rezapour M, Sepehri MM. Data mining performance in identifying the Risk Factors of early arteriovenous fistula failure in Hemodialysis Patients. Int J Hosp Res. 2013;2(1):49-54.
  38. Bica I, Alaa AM, Lambert C, Van Der Schaar M. From real‐world patient data to individualized treatment effects using machine learning: current and future methods to address underlying challenges. Clin Pharmacol Ther. 2021;109(1):87-100.
  39. Gómez-Pulido JA, Gómez-Pulido JM, Rodríguez-Puyol D, Polo-Luque M-L, Vargas-Lombardo M. Predicting the Appearance of Hypotension during Hemodialysis Sessions Using Machine Learning Classifiers. Int J Environ Res Public Health. 2021;18(5):2364.
Volume 1, Issue 1
April 2023
Pages 4-9
  • Receive Date: 04 May 2023
  • Revise Date: 24 May 2023
  • Accept Date: 29 May 2023
  • First Publish Date: 30 May 2023
  • Publish Date: 30 May 2023