This paper is in the following e-collection/theme issue:

Public Health for Asian/Pacific Islands (6) Public (e)Health, Digital Epidemiology and Public Health Informatics (725) Longitudinal and Cohort Studies in Public Health (309) Decision Support for Health Professionals (1185) Artificial Intelligence, Machine Learning, and Natural Language Processing for Public Health (36) Machine Learning (49)

Published on in Vol 8 (2024)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/48378, first published .
A Random Forest Algorithm for Assessing Risk Factors Associated With Chronic Kidney Disease: Observational Study

A Random Forest Algorithm for Assessing Risk Factors Associated With Chronic Kidney Disease: Observational Study

A Random Forest Algorithm for Assessing Risk Factors Associated With Chronic Kidney Disease: Observational Study

Authors of this article:

Pei Liu1 Author Orcid Image ;   Yijun Liu2 Author Orcid Image ;   Hao Liu3 Author Orcid Image ;   Linping Xiong2 Author Orcid Image ;   Changlin Mei4 Author Orcid Image ;   Lei Yuan2 Author Orcid Image
Article Authors Cited by Tweetations Metrics

    Original Paper

    1Department of Mathematics and Physics, Second Military Medical University, Shanghai, China

    2Department of Health Management, Second Military Medical University, Shanghai, China

    3Faculty of Health Service, Second Military Medical University, Shanghai, China

    4Nephrology Department, Shanghai Changzheng Hospital, Shanghai, China

    *these authors contributed equally

    Corresponding Author:

    Lei Yuan, BS

    Department of Health Management

    Second Military Medical University

    No.800 Xiangyin Road, Yangpu District, Shanghai, China

    Shanghai, 200433

    China

    Phone: 86 15026929271

    Email: yuanleigz@163.com


    Background: The prevalence and mortality rate of chronic kidney disease (CKD) are increasing year by year, and it has become a global public health issue. The economic burden caused by CKD is increasing at a rate of 1% per year. CKD is highly prevalent and its treatment cost is high but unfortunately remains unknown. Therefore, early detection and intervention are vital means to mitigate the treatment burden on patients and decrease disease progression.

    Objective: In this study, we investigated the advantages of using the random forest (RF) algorithm for assessing risk factors associated with CKD.

    Methods: We included 40,686 people with complete screening records who underwent screening between January 1, 2015, and December 22, 2020, in Jing’an District, Shanghai, China. We grouped the participants into those with and those without CKD by staging based on the glomerular filtration rate staging and grouping based on albuminuria. Using a logistic regression model, we determined the relationship between CKD and risk factors. The RF machine learning algorithm was used to score the predictive variables and rank them based on their importance to construct a prediction model.

    Results: The logistic regression model revealed that gender, older age, obesity, abnormal index estimated glomerular filtration rate, retirement status, and participation in urban employee medical insurance were significantly associated with the risk of CKD. On RF algorithm–based screening, the top 4 factors influencing CKD were age, albuminuria, working status, and urinary albumin-creatinine ratio. The RF model predicted an area under the receiver operating characteristic curve of 93.15%.

    Conclusions: Our findings reveal that the RF algorithm has significant predictive value for assessing risk factors associated with CKD and allows the screening of individuals with risk factors. This has crucial implications for early intervention and prevention of CKD.

    Asian Pac Isl Nurs J 2024;8:e48378

    doi:10.2196/48378

    Keywords

    chronic kidney disease; random forest model; risk factors; assessment 


    Chronic kidney disease (CKD) is characterized by chronic structural and functional impairment of the kidney of >3 months, caused by various factors. CKD is diagnosed based on the presence of pathological injury for more than 3 months, abnormal glomerular filtration rate (GFR), abnormal blood or urine composition, abnormal imaging findings, or an index estimated GFR (eGFR) of <60 mL/minute/1.73 m2 [1]. CKD is a major global health concern. Between 1990 and 2015, the annual mortality rate attributed to CKD increased at an average rate of 3.4% per year, and the global prevalence rate of CKD increased to 14.3% [2]. The economic burden due to CKD accounts for 31.4% of the global annual burden of living with disability [3-6]. In China, the prevalence of CKD among patients aged 18 years and older is 10.8%, encompassing approximately 120 million patients, indicating that approximately 1 in 10 Chinese individuals have had CKD [1]. Nevertheless, the awareness rate of CKD is low, and only 12.5% of patients know about their illness. CKD is highly prevalent and its treatment cost is high but unfortunately remains unknown. Therefore, early detection and intervention can mitigate the treatment burden on patients and decrease disease progression.

    In recent years, risk factors including hypertension, diabetes, and obesity, which are associated with CKD, have gradually shown a trend toward affecting the younger population [7]. CKD is closely linked with an increased risk of all-cause mortality, cardiovascular disease (CVD), renal failure, and other adverse health outcomes, causing a serious disease burden [8-10]. CKD is a major health concern due to its high prevalence, low awareness rate, high treatment cost, increased risk of combined cardiovascular events, and early mortality. Early intervention, treatment, and controlling the risk factors of CKD can decelerate and decrease disease progression and consequently reduce overall morbidity and mortality. Hence, diagnosis and risk factor assessment for patients with early-stage CKD are of immense significance.

    With continuous advancements in artificial intelligence technology, many researchers have attempted to use machine learning models in the medical field. Many studies have reported that machine learning algorithms can improve the decision-making abilities of clinicians in different fields, including clinical prediction. A study published in The Lancet [11] developed a feasible and effective machine learning–based risk stratification model for predicting adverse events post hospital discharge in patients with acute coronary syndromes. The random forest (RF) algorithm was first proposed by Leo Breiman and Adele Cutler in the early 21st century [12]. In the last few years, the use of the RF algorithm for disease risk prediction has garnered increasing attention due to its high accuracy. Furthermore, some researchers have used econometric models based on logistic regression (LR) and RF to predict the risk of acute ovarian failure [13]. Additionally, Let et al [14] constructed an RF model to improve the early detection and prediction of the incidence of venous thromboembolism in patients with lung cancer.

    Some researchers have explored the application of machine learning algorithms in disease prediction, compared them with traditional statistical regression models, and reported the differences in the performance of various prediction models. While comparing conventional LR models with the RF algorithm, many studies reported that the RF algorithm is more advantageous than the LR model. A previous study investigated the predictability of the RF algorithm, the LR model, and deep neural network models and found that machine learning models, particularly deep neural network models, can improve the long-term prognosis prediction of patients with ischemic stroke [15]. Another study constructed an interpretable RF model to predict severe acute pancreatitis and found that the RF model showed better precision and diagnostic accuracy than the LR and Bedside Index Of Severity In Acute Pancreatitis models [16]. Some researchers used 5 machine learning algorithms separately to predict the malnutrition status of 5-year-old children in Bangladesh and found that the accuracy of the RF algorithm was 68.51%, which was greater than that of other algorithms [17]. Another study reported that the RF algorithm is a better predictive model for older patients with hip fractures and high-risk mortality within 1 year after surgery [18].

    A longitudinal study involving 143,043 patients with hypertension was performed to predict long-term CVD risk. The study reported that advanced machine learning algorithms using RF performed better than traditional LR [19]. A longitudinal cohort study compared clinical risk predictions among patients with CVD using 19 prediction techniques. The study also reported that excluding LR and commonly used machine learning algorithms from long-term risk prediction models underestimated the disease risk [20].

    Researchers have also investigated the advantages of using RF models in predicting kidney diseases. A previous study reported the performance of 4 prediction tools, namely deep learning, plain Bayesian, RF, and LR, for predicting all-cause mortality in patients with CKD. The study showed that Bayesian networks and LR showed superior prediction abilities [21]. However, another study reported that plain Bayesian, RF, and LR performed adequately well and showed high sensitivity for screening end-stage renal disease in patients with CKD, which is inconsistent with previous reports [22]. Another previous study constructed 3 algorithms, namely RF, plain Bayesian, and LR, to classify glomerular and tubular injury and found that RF showed the best performance in terms of accuracy, sensitivity, and specificity. These findings suggest that RF can facilitate early diagnosis of glomerular and tubular injury to mitigate CKD progression [23]. Therefore, previous studies on the viability of RF models have reported inconsistent conclusions due to differences in research perspectives and subjects.


    Data Source

    The data for this study were collected from the CKD screening population in Jing’an District from January 1, 2015, to December 22, 2020. Information obtained included demographic and sociological characteristics, height, weight, diastolic and systolic blood pressure, health insurance type, screening date, urinary protein and urinary albumin-creatinine ratio (UACR), blood creatinine, eGFR, and screening results. In total, 103,960 records were initially screened and CKD diagnoses were categorized based on ICD-10 (International Statistical Classification of Diseases, Tenth Revision) criteria. Records with incomplete or duplicate data were excluded, resulting in a final sample size of 40,686 cases for analysis. These data are considered credible and authentic.

    Definition of Grouping

    The participants were categorized based on dichotomous variables: 1 for the nonmanagement population (indicating the absence of CKD) and 2 for the management population (indicating the presence of CKD).

    Covariance

    We used the 11 factors identified in the univariate analysis as explanatory variables for the LR model. The grouping and assignment of the dependent and independent variables are listed in Table 1.

    Table 1. Grouping and assignment of dependent and independent variables.
    　NameVariableValue assignment
    CKDa screeningY1. Nonmanagement population; 2. Management population
    Genderx11. Male; 2. Female
    Agex21. <65 years; 2. 65-75 years; 3. ≥75 years
    BMIx31. Normal: 18.5-24; 2. Underweight: <18.5; 3. Overweight: 24-28; 4. Obesity: ≥28
    History of
    hypertension    		x41. No; 2. Yes
    Index blood
    creatinine    		x51. Normal; 2. Abnormal
    Index eGFRbx61. No; 2. Yes
    Index urinary proteinx71. Negative; 2. Positive
    Albuminuriax81. No; 2. Yes
    Urine albumin-
    creatinine ratio    		x91. <30; 2. 30-300; 3. ≥300
    Working statusx101. Retired staff; 2. Unemployed person; 3. Othersc
    Type of medical
    insurance    		x111. Urban employee medical insurance;
    2. Urban resident medical insurance; 3. Othersd

    aCKD: chronic kidney disease.

    beGFR: estimated glomerular filtration rate.

    cOthers include students, freelancers, and workers.

    dOthers include the poverty relief system, out-of-pocket insurance, new rural cooperative medical system (NRCMS), commercial medical insurance, and free medical service. The same as below.

    Statistical Model

    A database was established using Excel (Microsoft Corp) 2010, and SAS (version 9.4; SAS Institute Inc) statistical software was used for data analysis. The chi-square test was performed for 1-way analysis to select variables for inclusion in the model, with the threshold for statistical significance set at P<.05. Based on the GFR stage, albuminuria (Alb) grouping, and the distribution of data, the study categorized participants for CKD screening into management (suspected and diagnosed patients) and nonmanagement (healthy individuals) populations. The resulting dichotomous LR model was then used for subsequent analysis.

    The RF Algorithm

    RF is a classification algorithm that uses multiple decision trees to train and predict samples. Specifically, the algorithm samples the training data set N times with replacement and selects a random subset of training samples each time. The remaining undrawn samples are subsequently used to evaluate the prediction error of the model.

    Training Validation Split

    The data set of 40,686 participants was randomly split into the following 2 subsets using simple random sampling in Python 3.6: one for validation sample set A including 13,549 cases (or 33.3% of the total data set), and the other for then training sample set B including 27,139 cases (or 66.7% of the total data set). The first subset A constituted the external validation sample set with 3000 cases (accounting for 7.4% of the total data set). The RF algorithm was subsequently applied to the training sample set to evaluate the importance of each variable and construct a CKD risk factor model. This model was used to predict the test sample set, with a minimum prediction accuracy threshold of 70%.

    Parameters

    The mean number of feature selections was used for each random tree (mtry) in the model.

    For a set with predictors, a typical number is the rounded square root of mtry [12]. Only 11 features were used in this study. We did not use the square root method to calculate mtry. However, we randomly selected a certain number of features each time and fixed ntree to adjust mtry to determine the values that minimized generalization errors as the optimal value of mtry.

    The mean number of random trees was used in the RF algorithm (ntree) in the model. (1) Using bootstrap resampling, 20% of the B set was randomly split and was used as an internal validation set and 80% was used as the training set. (2) Assuming that the number of the decision tree was ntree, for each node, mtry features were randomly selected. These mtry features were used to divide the sample set, and the index Gini was used to determine the best partitioning method. (3) For determining the mean error of the test set, steps (1), (2), and (3) were repeated. With each iteration of step (2), the ntree was increased by 1. ntree gradually increased from 1 to 200. We obtained the set for average generalization error, and observed the variation in the average generalization error with ntree. When the optimal model was achieved, we obtained the number of ntrees.

    Variable Importance

    After establishing the RF model, it was used for prediction. Given the abundance of trees in the forest, determining which variables have the most significant impact on predictions can be challenging. Fortunately, an important method was used to assess the significance of variables in the model. Specifically, for each variable, in each decision tree of an RF, the decrease in the splitting criterion function (residual squared or Gini index) caused by that variable was measured. The decrease in magnitude for each decision tree was then averaged to determine the importance of the variable. The importance of each feature variable was ranked and plotted in order, resulting in a variable importance plot.

    Ethical Considerations

    The Institutional Review Committee Board at Shanghai Changzheng Hospital affiliated with the Naval Medical University approved this study with written consent (No.2016SL020). This observational study analyzed existing data sources, which did not contain any patient-identifiable information. This study did not involve the collection, use, or transmission of individually identifiable data.


    LR Model With 2 Classifications

    Results of Single Factor Analysis

    An LR model with 2 classifications (CKD and non-CKD) was used for analysis. As shown in Table 2, the results of the univariate analysis indicate a statistically significant distribution of differences in CKD status in the investigated population across 11 variables: gender, age, BMI, history of hypertension, index blood creatinine, index eGFR, index urinary protein, Alb, UACR, working status, and type of health insurance (P<.05).

    Table 2. Distribution and comparison of baseline characteristics among patients diagnosed with CKDa.
    Variable nameTotal participants, nManagement population, n (%)Chi-square (df)P value
    Gender47.43 (1)<.001

    		Male17,20516,052 (93.30)


    		Female23,48121,473 (91.45)

    Age (years)7811.50 (2)<.001

    		<6596386864 (71.22)


    		65-7520,15619,783 (98.15)


    		≥7510,89210,878 (99.87)

    BMI (kg/m2)220.31 (3)<.001

    		Normal (18.5-24)19,44417,545 (90.23)


    		Underweight (<18.5)1021936 (91.67)


    		Overweight (24-28)15,38714,457 (93.96)


    		Obesity (≥28)48344587 (94.89)

    History of hypertension8.62 (1).003

    		No37,51334,556 (92.12)


    		Yes31732969 (93.57)

    Index blood creatinine62.35 (1)<.001

    		Normal39,95936,798 (92.09)


    		Abnormal727727 (100)

    Index eGFRb1164.79 (1)<.001

    		Normal16,81714,603 (86.83)


    		Abnormal23,86922,922 (96.03)

    Urine protein indicators387.10 (1)<.001

    		Negative36,55733,396 (91.35)


    		Positive41294129 (100)

    Albuminuria519.68 (1)<.001

    		No35,32932,168 (91.05)


    		Yes53575357 (100)

    Urinary albumin-creatinine ratio580.49 (2)<.001

    		<3034,79331,632 (90.91)


    		30-30052075207 (100)


    		≥300686686 (100)

    Working status1471.67 (2)<.001

    		Retired staff37,40635,062 (93.73)


    		Unemployed person204142 (69.61)


    		Others30762321 (75.46)

    Type of medical insurance111.97 (2)<.001

    		Urban worker22,90921,405 (93.43)


    		Urban resident16,62615,055 (90.55)


    		Others11511065 (92.53)

    aCKD: chronic kidney disease.

    beGFR: estimated glomerular filtration rate.

    Multivariate Analysis

    On univariate analysis, variables with statistically significant differences were subjected to multivariate analysis as explanatory variables in binary LR to establish a regression model. The variables were screened using the input method with a significance level of α=.05. The results of the multivariate analysis are presented in Table 3. The risk of CKD was lower in women than in men (odds ratio [OR] 0.909, 95% CI 0.829-0.997). Furthermore, the risk of CKD gradually increased with an increase in age, with people aged 75 years and older (OR 256.759, 95% CI 151.115-436.259) and those aged 65-75 years (OR 20.471, 95% CI 18.209-23.013) being at higher risk than those younger than 65 years. Moreover, individuals with a BMI above the normal range were at a higher risk of CKD. People with a BMI of ≥28 (OR 2.024, 95% CI 1.426-1.733) and those with a BMI of 24-28 (OR 1.572, 95% CI 1.426-1.733) were at a higher risk of CKD than those with a normal BMI. Similarly, people with an abnormal eGFR index were at a higher risk of CKD (OR 1.397, 95% CI 1.271-1.537) than those with a normal eGFR. Compared with other participants, retirees (OR 2.432, 95% CI 2.162-2.736) and people with medical insurance for urban employees (OR 1.769, 95% CI 1.319-2.372) were at higher risk of CKD.

    Table 4 shows that in the test sample, a high proportion of records (98.9%) was accurately predicted. Specifically, the prediction model correctly identified all management population records, whereas only 6.4% of nonmanagement population records were accurately predicted.

    Although dichotomous LR offers notable advantages including fast training, easy understanding, and high interpretability, its limitations should be acknowledged. First, its effectiveness may be hampered when managing imbalanced data sets, as observed in this study where indicators including urine routine proteins (PROs) exhibited excessive ORs because of the higher proportion of abnormal values within the management population. Second, similar to the accuracy rates of linear models, the accuracy rates of LR models may not be optimal because the latter can experience difficulty in fitting the true data distribution. Herein, imbalanced data sets in the regression model led to statistically insignificant urine test results. Thus, to overcome these limitations, we considered using a machine learning approach.

    Table 3. Logistic regression analysis of factors affecting chronic kidney disease in people with different characteristics.
    Variable nameβWald chi-square (df)P valueOdds ratio (95% CI)
    Female gender (reference: male)–0.0950.0474.103 (1).040.909 (0.829-0.997)
    Age (years; reference: ≤65 years)

    		65-753.0190.0602555.045 (1)<.00120.471 (18.209-23.013)

    		≥755.5480.270420.803 (1)<.001256.759 (151.115-436.259)
    BMI (kg/m2; reference: normal [18.5-24 kg/m2])

    		Underweight (<18.5)–0.2860.1483.737 (1).050.751 (0.562-1.004)

    		Overweight (24-28)0.4520.05082.521 (1)<.0011.572 (1.426-1.733)

    		Obesity (≥28)0.7050.08176.341 (1)<.0012.024 (1.728-2.370)
    Having a history of hypertension (reference: no)0.1270.0892.031 (1).151.135 (0.953-1.352)
    Abnormal index blood creatinine (reference: normal index blood creatinine)16.4071054.2000.000 (1).991.33×107 (0.000-0.000)
    Abnormal index eGFRa (reference: normal index eGFR)0.3350.04847.630 (1)<.0011.397 (1.271-1.537)
    Positive urine protein indicators (reference: negative urine protein indicators)15.990436.5340.001 (1).978.80×106 (0.000-0.000)
    Having albuminuria (not having albuminuria)17.360403.3170.002 (1).973.46×107 (0.000-0.000)
    Urine albumin-creatinine ratio (reference: <30)

    		30-30017.435440.6540.002 (1).973.73×107 (0.000-0.000)

    		≥30015.8241063.960<0.001 (1).997.45×106 (0.000)
    Working status (reference: other)

    		Retired staff0.8890.060218.852 (1)<.0012.432 (2.162-2.736)

    		Unemployed person–0.0320.2030.026 (1).870.968 (0.651-1.441)
    Type of medical insurance (reference: other)

    		Urban employee
    medical insurance    		0.5700.15014.504 (1)<.0011.769 (1.319-2.372)

    		Urban resident
    medical insurance    		–0.1590.1511.116 (1).290.853 (0.634-1.146)

    aeGFR: estimated glomerular filtration rate.

    Table 4. Classification of model predictions.
    Real testPrediction of chronic kidney disease statusPercentage of accurate predictions, %

    		Nonmanagement population, nManagement population, n
    Chronic kidney disease

    		Non-management target population3446.4

    		Management target population03818100.00
    Total percentage98.90

    Machine Learning: RF Algorithm

    Modeling

    The data set was split into 66.7% of samples, which corresponded to 27,139 records, randomly selected without replacement. The control method was applied by fixing the ntree (number of means of random trees in the RF algorithm) constant and debugging the mtry (mean number of feature selections used for each random tree) parameter. In each iteration, a certain number of features were randomly selected, and the average generalization error value was computed for 11 trials. The change in the error rate of the model, with respect to mtry, is depicted in Figure 1. The error rate decreased significantly when the number of features changed from 1 to 2, followed by an increase close to the minimum value, which was achieved when mtry=4. Next, the mtry value was set to 4, and the ntree value was adjusted accordingly. In total, 200 random trials were conducted to gauge the average generalization error of the test set (Figure 2). The generalization error rate decreased rapidly from 1 to 10, decreased slowly from 10 to 25, and thereafter flattened and stabilized. Thus, the optimal model was identified when the ntree value was 166.

    ‎
    Figure 1. The effect of mtry on the error rate of random forest algorithm.
    ‎
    Figure 2. The effect of ntree on the error rate of the random forest (RF) algorithm.
    Analysis of the Results of the RF Algorithm

    The RF algorithm was trained on a test data set comprising 27,139 records, with ntree=166 and mtry=4. Using these parameters, the algorithm was applied to classify the test set data, and the importance ranking of each feature was determined (Multimedia Appendix 1). The 4 most important features identified were age, Alb, working status, and UACR. These features were further selected for the prediction study, which yield a final classification accuracy rate of 92.67%.

    Next, 100 random trials were conducted to ensure the reliability of our results. The generalization error plot is presented in Figure 3. The error was concentrated around 0.0735, with a small fluctuation and an average error of 7.371%. Our results indicate a good generalization ability of the model, suggesting its reliability in classification tasks.

    ‎
    Figure 3. The generalization error rate of the random forest algorithm was estimated by conducting 100 randomized trials.

    Comparison of the Sensitivity and Specificity of RF Models

    The area under the receiver operating characteristic curve (AUC) of the RF model based on the training and test sets was 93.15% (Figure 4). The RF algorithm outputs voting results (0s and 1s), whereas the receiver operating characteristic curve requires voting probability data. Converting probabilities to voting results can lead to error because of extreme probabilities, such as 0.01515526 and 0.98484474. Therefore, we calculated the AUC to assess model performance and the classification prediction rate to indicate the accuracy of the model. Herein, the RF algorithm achieved an accuracy rate of 92.67%, with some degree of error. These results suggest that the model exhibited good predictive power and accurately classified new data samples.

    ‎
    Figure 4. Receiver operating characteristic (ROC) curve of chronic kidney disease prediction by the random forest algorithm. AUC: area under the receiver operating characteristic curve.

    Confusion Matrix

    Four possible predicted results were as follows: true positives, false positives, true negatives, and false negatives. Table 5 shows the confusion matrix of the RF model. The precision, recall, and F1-score were 0.951, 0.984, and 0.967, respectively.

    Table 5. Confusion matrix of the random forest algorithm model.

    		Predicted values (=1)Predicted values (=0)
    Actual values (=1)True positive: 12,505False negative: 209
    Actual values (=0)False positive: 640True negative: 195

    Principal Findings

    A risk assessment model for CKD was developed in this study using dichotomous LR and RF models. Our results indicate that gender, older age, BMI beyond the normal range, abnormal index eGFR, retirement status, and urban employee medical insurance were significantly associated with a higher risk of CKD. By leveraging the RF model, the most important factors for CKD development were older age, abnormal urinary test results (eg, Alb, UACR, and index PRO indicators), and high BMI.

    In China, the number of studies on the assessment of risk factors for CKD and the investigation of methods for risk prediction is increasing and LR analysis is commonly being performed. Feng et al [24] used an adjusted LR model to investigate CKD prevalence and related risk factors in 38 megacities across China. Liu et al [25] and Yang et al [26] performed cross-sectional studies to analyze risk factors for diabetic nephropathy in Shanghai, whereas a community-based, 7-year-long cohort study from Tianjin used LR to examine the association between the high triglyceride waist phenotype and risk of CKD development [27]. Yan et al [28] performed LR analysis to assess the correlation between residual cholesterol levels and CKD, and identify other significant risk factors affecting middle-aged and older individuals residing within a city. Gradual advancements in machine learning models have prompted further scrutiny of the divergent performance and inherent limitations of the conventional LR approach. To distinguish this study from previous studies that followed the LR approach for exploratory purposes, we used the RF algorithm to rank risk factors that were subjected to single-factor analysis according to their relevance and consequently evaluated comparative predictive precision by performing LR analysis using training samples. Our results reveal that both the RF and LR models achieved an overall accuracy rate exceeding 90% in the prediction test set. Conversely, the dichotomous LR model exhibited a marginally superior predictive performance than the RF model. Nevertheless, one should pay attention to the tendency of LR to result in excessive ORs when imbalanced data are used. Although LR exhibits excellent predictive abilities and desirable attributes such as high accuracy and stability, and ease of operation with a minimal possibility of overlearning during classification prediction, RF has the ability to assess the importance of variables when classifying data into suitable categories while compensating for errors in imbalanced sets of categorical data.

    Our results indicate that age was the primary significant factor in the RF model, and LR analysis confirmed that higher age was significantly associated with CKD. Compared to participants aged ≤64 years, those aged 65-75 years and older were at a significantly higher risk of CKD, which is in line with previous results [29,30]. The risk of CKD increases with age; thus, early screening and risk prediction for CKD are crucial for middle-aged and older people.

    A cross-sectional study published in The Lancet [31], using a nationally representative sample of Chinese adults also identified independent factors associated with kidney damage, which included age and gender. Age and gender are independent CKD risk factors [32]. Many studies worldwide have shown that women are at a higher risk of CKD [33,34], and similar observations have been reported in China [24,30]. This correlation may be attributed to differences in the prevalence of primary diseases and the availability of medical resources across genders [35]. However, our results show that females in the survey population were at a lower risk of CKD than were males, which is inconsistent with the majority of previous results. Our data include information regarding the registered population in a district of Shanghai. The exclusion of samples with incomplete information and regional differences, as well as the presence of unregistered patients, may have led to bias, ultimately yielding inconsistent results.

    Next, this study shows that people with a higher-than-normal BMI were at a higher risk of CKD, similar to a time-series study that investigated risk factors regarding CKD burden in China from 1991 to 2011 and identified the correlation between high BMI and CKD [36]. Obesity is an important risk factor for CKD worldwide [24,25,37-39]. Potential obesity-associated factors that may lead to or aggravate CKD include hemodynamic disorder and renal tissue hypoxia [40,41]. However, weight loss through diet and regular exercise can reverse kidney damage; hence, maintaining a healthy lifestyle and controlling body weight could prevent or decelerate CKD progression to a certain extent [42]. Additionally, this study shows that CKD risk was higher in people who had urban employee medical insurance. These people were employed and had relatively better economic conditions; however, health risk factors such as work stress and unhealthy lifestyles probably contribute to an increased CKD risk [43].

    Moreover, people with abnormal urine test results (Alb, UACR, and PRO indicators) were at a higher CKD risk, which is consistent with previous results reported worldwide [36,44,45]. Similarly, a Chinese study using 4 machine learning models, comprising 19,270 adult samples, showed that UACR, Alb, age, and gender were important CKD risk factors [44]. Urine tests can serve as an early warning system for CKD detection. Similarly, our risk prediction model could guide decision-making regarding early CKD screening.

    Limitations

    Herein, we effectively assessed the risk of CKD by combining internal data for model construction and testing. However, this study has some limitations. First, the generalization ability of the model remains unknown because the study did not include external data for external validation. Second, owing to the bias in data collection, our results were inconsistent with those of the previous studies. Finally, more prospective studies are required to verify the predictive power and practical utility of our model. Thus, health care professionals should routinely evaluate the level of agreement within and between models before reaching any clinical decision on the basis of the present limitations and previous findings [46].

    Conclusions

    In conclusion, the RF model has significant predictive value for assessing risk factors associated with CKD and is capable of correcting errors in imbalanced categorical data sets. It can be used to screen individuals with risk factors, which is of great significance for early intervention and prevention of CKD.

    For the prevention and treatment of CKD, early intervention can involve a low-protein diet, regular physical examination, actively promoting urine examination, and screening of high-risk groups to achieve early detection, early treatment, early diagnosis, and early intervention of CKD, and to reduce the social and personal losses caused by diseases and improve people’s quality of life.

    Acknowledgments

    We are grateful for the enthusiastic cooperation of the nephrology department of Shanghai Changzheng Hospital, Shanghai. We thank Bullet Edits for providing language editing support. This study was funded by the Shanghai 3-Year Action Plan for Public Health System Construction (SCREENING STUDY GWIV-18). The funder had no role in the study’s design, data collection, or analysis, the decision to publish, or the preparation of the manuscript.

    Data Availability

    The data sets used or analyzed in this study are available from the first author upon reasonable request.

    Authors' Contributions

    LX and CM obtained the funding. PL, YL, HL, LX, CM, and LY conceived and designed the experiments. PL, YL, HL, and LY performed the experiments, analyzed the data, and contributed reagents, materials, and analysis tools. PL drafted the manuscript. All authors participated in the discussion, revision, and approval of the final manuscript.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Importance ranking of each indicator in the random forest algorithm.

    PNG File , 16 KB
    1. Expert Group on Early Detection, Diagnosis and Treatment System Construction of Chronic Kidney Disease in Shanghai, Gao X, Mei C. Guideline for screening, diagnosis, prevention and treatment of chronic kidney disease [Article in Chinese]. Chin J Pract Int Med. 2017;37(01):28-34. [CrossRef]
    2. Danial M, Hassali MA, Meng OL, Kin YC, Khan AH. Development of a mortality score to assess risk of adverse drug reactions among hospitalized patients with moderate to severe chronic kidney disease. BMC Pharmacol Toxicol. Jul 08, 2019;20(1):41. [FREE Full text] [CrossRef] [Medline]
    3. Thomas B, Matsushita K, Abate KH, Al-Aly Z, Ärnlöv J, Asayama K, Global Burden of Disease 2013 GFR Collaborators, et al. Global Burden of Disease Genitourinary Expert Group. Global cardiovascular and renal outcomes of reduced GFR. J Am Soc Nephrol. Jul 2017;28(7):2167-2179. [FREE Full text] [CrossRef] [Medline]
    4. GBD 2015 MortalityCauses of Death Collaborators. Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. Oct 08, 2016;388(10053):1459-1544. [FREE Full text] [CrossRef] [Medline]
    5. Kerr PG, Tran HTB, Ha Phan H, Liew A, Hooi LS, Johnson DW, et al. OSEA Regional Board. Nephrology in the Oceania-South East Asia region: perspectives and challenges. Kidney Int. Sep 2018;94(3):465-470. [FREE Full text] [CrossRef] [Medline]
    6. Luo K, Bian J, Wang Q, Wang J, Chen F, Li H, et al. Association of obesity with chronic kidney disease in elderly patients with nonalcoholic fatty liver disease. Turk J Gastroenterol. Jul 2019;30(7):611-615. [FREE Full text] [CrossRef] [Medline]
    7. Kuma A, Kato A. Lifestyle-related risk factors for the incidence and progression of chronic kidney disease in the healthy young and middle-aged population. Nutrients. Sep 14, 2022;14(18). [FREE Full text] [CrossRef] [Medline]
    8. Nugent RA, Fathima SF, Feigl AB, Chyung D. The burden of chronic kidney disease on developing nations: a 21st century challenge in global health. Nephron Clin Pract. 2011;118(3):c269-c277. [FREE Full text] [CrossRef] [Medline]
    9. A. US Renal Data System 2019 Annual Data Report: epidemiology of kidney disease in the United States. Am J Kidney Dis. Oct 31, 2019;75(1):S1-S64. [CrossRef] [Medline]
    10. Chronic Kidney Disease Prognosis Consortium, Matsushita K, van der Velde M, Astor BC, Woodward M, Levey AS, et al. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet. Jun 12, 2010;375(9731):2073-2081. [FREE Full text] [CrossRef] [Medline]
    11. D'Ascenzo F, De Filippo O, Gallone G, Mittone G, Deriu MA, Iannaccone M, et al. Machine learning-based prediction of adverse events following an acute coronary syndrome (PRAISE): a modelling study of pooled datasets. The Lancet. Jan 2021;397(10270):199-207. [CrossRef]
    12. Breiman L. Random forests. Mach Learn. 2001;45:5-32. [CrossRef]
    13. Clark RA, Mostoufi-Moab S, Yasui Y, Vu NK, Sklar CA, Motan T, et al. Predicting acute ovarian failure in female survivors of childhood cancer: a cohort study in the Childhood Cancer Survivor Study (CCSS) and the St Jude Lifetime Cohort (SJLIFE). Lancet Oncol. Mar 2020;21(3):436-445. [FREE Full text] [CrossRef] [Medline]
    14. Lei H, Zhang M, Wu Z, Liu C, Li X, Zhou W, et al. Development and validation of a risk prediction model for venous thromboembolism in lung cancer patients using machine learning. Front Cardiovasc Med. 2022;9:845210. [FREE Full text] [CrossRef] [Medline]
    15. Heo J, Yoon JG, Park H, Kim YD, Nam HS, Heo JH. Machine learning-based model for prediction of outcomes in acute stroke. Stroke. May 2019;50(5):1263-1265. [CrossRef] [Medline]
    16. Hong W, Lu Y, Zhou X, Jin S, Pan J, Lin Q, et al. Usefulness of random forest algorithm in predicting severe acute pancreatitis. Front Cell Infect Microbiol. 2022;12:893294. [FREE Full text] [CrossRef] [Medline]
    17. Talukder A, Ahammed B. Machine learning algorithms for predicting malnutrition among under-five children in Bangladesh. Nutrition. Oct 2020;78:110861. [CrossRef] [Medline]
    18. Xing F, Luo R, Liu M, Zhou Z, Xiang Z, Duan X. A new random forest algorithm-based prediction model of post-operative mortality in geriatric patients with hip fractures. Front Med (Lausanne). May 11, 2022;9:829977. [FREE Full text] [CrossRef] [Medline]
    19. Xi Y, Wang H, Sun N. Machine learning outperforms traditional logistic regression and offers new possibilities for cardiovascular risk prediction: A study involving 143,043 Chinese patients with hypertension. Front Cardiovasc Med. 2022;9:1025705. [FREE Full text] [CrossRef] [Medline]
    20. Li Y, Sperrin M, Ashcroft DM, van Staa TP. Consistency of variety of machine learning and statistical models in predicting clinical risks of individual patients: longitudinal cohort study using cardiovascular disease as exemplar. BMJ. Nov 04, 2020;371:m3919. [FREE Full text] [CrossRef] [Medline]
    21. Tran NTD, Balezeaux M, Granal M, Fouque D, Ducher M, Fauvel J. Prediction of all-cause mortality for chronic kidney disease patients using four models of machine learning. Nephrol Dial Transplant. Jun 30, 2023;38(7):1691-1699. [CrossRef] [Medline]
    22. Bai Q, Su C, Tang W, Li Y. Machine learning to predict end stage kidney disease in chronic kidney disease. Sci Rep. May 19, 2022;12(1):8377. [FREE Full text] [CrossRef] [Medline]
    23. Song W, Zhou X, Duan Q, Wang Q, Li Y, Li A, et al. Using random forest algorithm for glomerular and tubular injury diagnosis. Front Med (Lausanne). 2022;9:911737. [FREE Full text] [CrossRef] [Medline]
    24. Feng T, Xu Y, Zheng J, Wang X, Li Y, Wang Y, et al. Prevalence of and risk factors for chronic kidney disease in ten metropolitan areas of China: a cross-sectional study using three kidney damage markers. Ren Fail. Dec 2023;45(1):2170243. [FREE Full text] [CrossRef] [Medline]
    25. Liu W, Du J, Ge X, Jiang X, Peng W, Zhao N, et al. The analysis of risk factors for diabetic kidney disease progression: a single-centre and cross-sectional experiment in Shanghai. BMJ Open. Jun 28, 2022;12(6):e060238. [FREE Full text] [CrossRef] [Medline]
    26. Yang Y, Wang N, Jiang Y, Zhao Q, Chen Y, Ying X, et al. The prevalence of diabetes mellitus with chronic kidney disease in adults and associated factors in Songjiang District, Shanghai. Ann Palliat Med. Jul 2021;10(7):7214-7224. [FREE Full text] [CrossRef] [Medline]
    27. Chen R, Sun G, Liu R, Sun A, Cao Y, Zhou X, et al. Hypertriglyceridemic waist phenotype and risk of chronic kidney disease in community-dwelling adults aged 60 years and older in Tianjin, China: a 7-year cohort study. BMC Nephrol. May 19, 2021;22(1):182. [FREE Full text] [CrossRef] [Medline]
    28. Yan P, Xu Y, Miao Y, Bai X, Wu Y, Tang Q, et al. Association of remnant cholesterol with chronic kidney disease in middle-aged and elderly Chinese: a population-based study. Acta Diabetol. Dec 2021;58(12):1615-1625. [CrossRef] [Medline]
    29. Li Y, Ning Y, Shen B, Shi Y, Song N, Fang Y, et al. Temporal trends in prevalence and mortality for chronic kidney disease in China from 1990 to 2019: an analysis of the Global Burden of Disease Study 2019. Clin Kidney J. Feb 2023;16(2):312-321. [FREE Full text] [CrossRef] [Medline]
    30. Zhuang Z, Tong M, Clarke R, Wang B, Huang T, Li L. Probability of chronic kidney disease and associated risk factors in Chinese adults: a cross-sectional study of 9 million Chinese adults in the Meinian Onehealth screening survey. Clin Kidney J. Dec 2022;15(12):2228-2236. [FREE Full text] [CrossRef] [Medline]
    31. Zhang L, Wang F, Wang L, Wang W, Liu B, Liu J, et al. Prevalence of chronic kidney disease in China: a cross-sectional survey. Lancet. Mar 03, 2012;379(9818):815-822. [CrossRef] [Medline]
    32. Deng Y, Li N, Wu Y, Wang M, Yang S, Zheng Y, et al. Global, regional, and national burden of diabetes-related chronic kidney disease from 1990 to 2019. Front Endocrinol (Lausanne). 2021;12:672350. [FREE Full text] [CrossRef] [Medline]
    33. Brar A, Markell M. Impact of gender and gender disparities in patients with kidney disease. Curr Opin Nephrol Hypertens. Mar 2019;28(2):178-182. [CrossRef] [Medline]
    34. Forni Ogna V, Ogna A, Ponte B, Gabutti L, Binet I, Conen D, et al. Prevalence and determinants of chronic kidney disease in the Swiss population. Swiss Med Wkly. 2016;146:w14313. [FREE Full text] [CrossRef] [Medline]
    35. Carrero JJ, Hecking M, Chesnaye NC, Jager KJ. Sex and gender disparities in the epidemiology and outcomes of chronic kidney disease. Nat Rev Nephrol. Mar 22, 2018;14(3):151-164. [CrossRef] [Medline]
    36. Li P, Yang M, Hang D, Wei Y, Di H, Shen H, et al. Risk assessment for longitudinal trajectories of modifiable lifestyle factors on chronic kidney disease burden in China: a population-based study. J Epidemiol. Oct 05, 2022;32(10):449-455. [FREE Full text] [CrossRef] [Medline]
    37. Duan J, Duan G, Wang C, Liu D, Qiao Y, Pan S, et al. Prevalence and risk factors of chronic kidney disease and diabetic kidney disease in a central Chinese urban population: a cross-sectional survey. BMC Nephrol. Apr 03, 2020;21(1):115. [FREE Full text] [CrossRef] [Medline]
    38. Wang L, Xu X, Zhang M, Hu C, Zhang X, Li C, et al. Prevalence of chronic kidney disease in China: results from the sixth China Chronic Disease and Risk Factor Surveillance. JAMA Intern Med. Apr 01, 2023;183(4):298-310. [FREE Full text] [CrossRef] [Medline]
    39. Betzler BK, Sultana R, Banu R, Tham YC, Lim CC, Wang YX, et al. Association between body mass index and chronic kidney disease in Asian populations: a participant-level meta-analysis. Maturitas. Dec 2021;154:46-54. [CrossRef] [Medline]
    40. Shen W, Chen H, Chen H, Xu F, Li L, Liu Z. Obesity-related glomerulopathy: body mass index and proteinuria. Clin J Am Soc Nephrol. Aug 2010;5(8):1401-1409. [FREE Full text] [CrossRef] [Medline]
    41. Redon J, Lurbe E. The kidney in obesity. Curr Hypertens Rep. Jun 2015;17(6):555. [CrossRef] [Medline]
    42. Jiang Z, Wang Y, Zhao X, Cui H, Han M, Ren X, et al. Obesity and chronic kidney disease. Am J Physiol Endocrinol Metab. Jan 01, 2023;324(1):E24-E41. [FREE Full text] [CrossRef] [Medline]
    43. Wang X, Shi KX, Yu CQ, Lyu J, Guo Y, Pei P, et al. [Prevalence of chronic kidney disease and its association with lifestyle factors in adults from 10 regions of China]. Zhonghua Liu Xing Bing Xue Za Zhi. Mar 10, 2023;44(3):386-392. [CrossRef] [Medline]
    44. Shih C, Lu C, Chen G, Chang C. Risk prediction for early chronic kidney disease: results from an adult health examination program of 19,270 individuals. Int J Environ Res Public Health. Jul 10, 2020;17(14). [FREE Full text] [CrossRef] [Medline]
    45. Murton M, Goff-Leggett D, Bobrowska A, Garcia Sanchez JJ, James G, Wittbrodt E, et al. Burden of chronic kidney disease by KDIGO categories of glomerular filtration rate and albuminuria: a systematic review. Adv Ther. Jan 2021;38(1):180-200. [FREE Full text] [CrossRef] [Medline]
    46. Bikbov B, Perico N, Remuzzi G, on behalf of the GBD Genitourinary Diseases Expert Group. Disparities in chronic kidney disease prevalence among males and females in 195 countries: analysis of the Global Burden of Disease 2016 Study. Nephron. 2018;139(4):313-318. [FREE Full text] [CrossRef] [Medline]

    ‎
    Alb: albuminuria
    CKD: chronic kidney disease
    CVD: cardiovascular disease
    eGFR: estimated glomerular filtration rate
    GFR: glomerular filtration rate
    ICD-10: International Statistical Classification of Diseases, Tenth Revision
    LR: logistic regression
    OR: odds ratio
    PRO: urine routine protein
    RF: random forest
    UACR: urinary albumin-creatinine ratio

    Edited by H Ahn; submitted 21.04.23; peer-reviewed by M Singh, N Trehan, M Gasmi , Y Zhang; comments to author 04.01.24; revised version received 02.02.24; accepted 16.04.24; published 03.06.24.

    Copyright

    ©Pei Liu, Yijun Liu, Hao Liu, Linping Xiong, Changlin Mei, Lei Yuan. Originally published in the Asian/Pacific Island Nursing Journal (https://apinj.jmir.org), 03.06.2024.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Asian/Pacific Island Nursing Journal, is properly cited. The complete bibliographic information, a link to the original publication on https://apinj.jmir.org, as well as this copyright and license information must be included.