Skip to main content

Predictive modeling for COVID-19 readmission risk using machine learning algorithms

Abstract

Introduction

The COVID-19 pandemic overwhelmed healthcare systems with severe shortages in hospital resources such as ICU beds, specialized doctors, and respiratory ventilators. In this situation, reducing COVID-19 readmissions could potentially maintain hospital capacity. By employing machine learning (ML), we can predict the likelihood of COVID-19 readmission risk, which can assist in the optimal allocation of restricted resources to seriously ill patients.

Methods

In this retrospective single-center study, the data of 1225 COVID-19 patients discharged between January 9, 2020, and October 20, 2021 were analyzed. First, the most important predictors were selected using the horse herd optimization algorithms. Then, three classical ML algorithms, including decision tree, support vector machine, and k-nearest neighbors, and a hybrid algorithm, namely water wave optimization (WWO) as a precise metaheuristic evolutionary algorithm combined with a neural network were used to construct predictive models for COVID-19 readmission. Finally, the performance of prediction models was measured, and the best-performing one was identified.

Results

The ML algorithms were trained using 17 validated features. Among the four selected ML algorithms, the WWO had the best average performance in tenfold cross-validation (accuracy: 0.9705, precision: 0.9729, recall: 0.9869, specificity: 0.9259, F-measure: 0.9795).

Conclusions

Our findings show that the WWO algorithm predicts the risk of readmission of COVID-19 patients more accurately than other ML algorithms. The models developed herein can inform frontline clinicians and healthcare policymakers to manage and optimally allocate limited hospital resources to seriously ill COVID-19 patients.

Peer Review reports

Introduction

The coronavirus disease 2019 (COVID-19) or acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a highly transmissible and widespread infection that, in its severe form, causes serious damage to the respiratory tract and in some individuals leads to pneumonia, multi-organ failure (MOF), and even death [1, 2]. The unknown clinical course and behavior of COVID-19 contributed to ambiguous discharge criteria for hospitalized patients [3]. Furthermore, the variability and dynamic nature of the virus and its new variants led to resistance to treatment and vaccinations [4,5,6]. According to reports, about 5% of definitive COVID-19 cases require hospitalization care services, and the rate of hospital readmission due to this disease varies from 2 and 10% in different studies [7, 8]. This rate varies depending on age, body mass index, underlying diseases, sex, vaccination, disease severity, and SARS-CoV-2 (COVID-19) variant types (Alpha, Beta, Delta, Omicron) [9,10,11]. After second-and third-dose vaccination, this rate considerably decreased [12].

Hospital readmission is defined as the admission of a patient to a hospital at a specific time within 30 to 60 days after discharge from the hospital. Readmissions represent important and costly events that impose a heavy burden on patients’ families and the healthcare systems [13, 14]. Hospital readmissions are mostly accountable for the reputation of the healthcare settings, causing notoriety and indicating clinicians' carelessness [15]. Hospital readmission has received increasing attention as the main performance indicator for evaluating the quality of care given to patients [16, 17]. Studies report that over 60% of hospital readmissions are potentially preventable. However, due to the varied and complex natures of factors causing disease recurrence and readmissions, caregivers cannot process all the information to precisely detect endangered patients [18]. Thus, increasing attention is being paid in the scientific community to this problem from a data analysis viewpoint [19].

Hospital readmission is known as a key indicator of the quality of service during the COVID-19 pandemic [20]. As the prevalence of COVID-19 increased and many communities became severely impacted, the healthcare systems of many countries failed to meet the growing needs of patients [21]. Many patients in such conditions were discharged after admission with relative recovery. Meanwhile, due to the unknown and aggressive nature of the disease, the readmission rate of patients increased [22, 23]. Readmission imposes additional costs on healthcare organizations and patients [22]. It also reduces the quality indicators of service delivery and raises the rate of serious complications and death during the pandemic [24, 25].

The use of clinical evaluation methods to predict disease re-infection and readmission is usually expensive, difficult, and lacks optimal predictive accuracy as it does not use cumulative patient data [26]. Scoring indices and conventional statistical models can only analyze simple and linear relationships between variables. Nevertheless, the unknown and multidimensional nature of COVID-19 requires innovative technologies such as artificial intelligence (AI) to analyze the nonlinear and complex relationships between variables [26,27,28,29,30,31,32,33,34,35]. Machine learning (ML), which is a major branch of AI, reveals new and practical patterns from huge raw datasets [36, 37]. ML algorithms diminish uncertainties and ambiguities related to new diseases such as COVID-19 by providing diagnostic and predictive models based on valid and scientific evidence [38, 39]. The multifaceted interaction between readmission and possible risk factors makes the precise prediction of readmission difficult. ML approaches can deal with high-dimensional clinical data to produce precise patient risk stratification models and shape healthcare decisions through the customization of care [36, 39].

Numerous studies have examined the application of ML and deep learning (DL) methods to predict the disease recurrence, reinfection, and patient deterioration among recovered COVID‐19 patients [40,41,42,43,44]. ML methods are more accurate than conventional statistical models for predicting hospital readmission in COVID-19 hospitalized patients [45,46,47]. Therefore, this study aimed to apply ML algorithms to predict the likelihood of hospital readmission of COVID-19 patients. The current study sought to answer two questions: What are the most important predictor variables affecting the readmission of COVID-19 patients? and Which ML model is more effective for predicting readmission in these patients?

Materials and methods

Study design

The current research was a retrospective study on the data of 2854 patients discharged from a 400-bed academic hospital in Abadan, Iran, from January 9, 2020 to October 20, 2021. The patient data were extracted from the COVID-19 hospital-based registry database. The implemented registry system is a comprehensive web-based application software that records patient data for clinical and research purposes in five main sections: demographic, diagnostic and therapeutic, paraclinical, and history and information. Patients aged less than 18 years, those who were admitted for non-COVID-19 conditions, died during hospitalization, were discharged against medical advice, or had incomplete case records with > 70% missing data were excluded from the study.

The study was conducted in three phases. In the first phase, the primary raw dataset was preprocessed. In the second phase, important features for predicting the risk of hospital readmission in COVID-19 patients were selected using meta-heuristic algorithms (MHAs). After identifying the most important features, three traditional ML algorithms and a meta-heuristic algorithm for water wave optimization using a neural network were trained. Finally, the developed models’ performances were compared, and the best algorithm was determined. The study protocol was approved by the Abadan University of Medical Science Ethics Board (ABADANUMS.REC.1400.136),https://ethics.research.ac.ir/ProposalCertificateEn.php?id=246118&Print=true&NoPrintHeader=true&NoPrintFooter=true&NoPrintPageBorder=true&LetterPrint=true).

Data preparation

We clustered certain classes to decrease the number of classes of these variables. Records with more than 70% of missing data were excluded from the analysis. For the remaining missing values, presuming that the missing data were distributed randomly, the imputation technique which is a common method to deal with missing values was adopted [19]. To manage noisy data, the normal range of each variable was first defined using the opinion of two infectious diseases specialists, a virology expert, and a hematology expert. Then, we specified all the values that were outside the defined range and filled them by referring to patient records or the responsible doctor. Because the p-value cut-off point was < 0.05 in this study, the median substitution was used instead of the mean for the missing values. In other words, we did not fill them with the mean values due to the uneven distribution of variables.

Data balancing

A major barrier to the use of ML algorithms is the problem of imbalanced data, which happens when classes are not categorized equally. In the selected dataset, the amount of data in outcome classes is significantly imbalanced and contains more samples related to the non-readmission class (1136 cases), while the readmission class is much smaller (only 89 cases). Accordingly, the developed models often deliver biased results towards the overriding class, and the ML models are much more likely to categorize new observations into the majority class. Herein, to handle class imbalance, the synthetic minority over-sampling technique (SMOTE) was employed in the Imbalanced-Learn toolbox to balance the dataset. We performed a Kolmogorov–Smirnov statistical test to check the normality and skewness of the data, the results of which showed that the data followed a normal distribution.

Predictor and outcome variables

Predictor variables

The data for analysis included six categories of predictor variables extracted from the hospital’s COVID-19 dataset. Sixty variables were categorized as demographic characteristics (six variables), clinical manifestation (14 variables), medical history and comorbidities (eight variables), laboratory results (28 variables), treatment (one variable), and radiological indicators (two variables).

Outcome variable

It calculated whether the patient was readmitted on the last visit within 30 days after being discharged from the hospital on the penultimate visit (coded 1) or not (coded 0). The detailed descriptions of all the variables are listed in Table 1.

Table 1 A list of variables and their corresponding category utilized in predicting COVID-19 readmission risk

Feature selection

Feature selection can be performed to enhance the prediction precision and reduce the algorithm's run time by selecting the most important variables, thereby alleviating the model’s computational intricacy [48]. In this study, the efficiency of several feature selection methods was compared to identify the best predictors. To this end, six well-known MHAs, including horse herd optimization algorithm (HOA), particle swarm optimization (PSO), genetic algorithm (GA), grey wolf optimization (GWO), differential evolution (DE), and Harris hawks optimization (HHO) were utilized for feature selection. In this phase, all the experiments were carried out using MATLAB 2019. To evaluate the performance of MHAs in identifying the most effective factors, three performance evaluation metrics of the mean fitness value, classification accuracy using k-nearest neighbors (KNN), and the number of selected features were calculated.

Model development

We trained four ML algorithms, namely KNN, water wave optimization (WWO), support vector machine (SVM), and decision tree (DT) in the WEKA application. Each method is described below.

SVM

The SVM is a supervised algorithm associated with datasets having data class labels. This algorithm can detect the pattern and assign the sample to specified output classes. With a high dimension of dataset, this algorithm has a proper classification potential. Contrary to artificial neural networks (ANNs), it is not stopped at the local maximum during the training process. This algorithm focuses on the line discriminating various class labels with high capability when there are complicated databases and patterns and enhancing the line. Generally, the SVM aims to find the hyperplane in categorizing the dataset sample to obtain the best classification performance in n-dimensional datasets. This capability of SVM contributes to its good performance compared to other approaches [49,50,51].

KNN

This algorithm, similar to the SVM algorithm, can be used for classification and regression. It is a supervised ML algorithm when considering an output class for the dataset. For a specific value of K, an object belongs to the classes according to its nearest samples. This algorithm does not need to assume the data pattern before classifying the objects. The KNN is classified as a lazy algorithm because the learning process is not concurrent with the algorithm training. In the training process, the data are stored and will be categorized when training the new data instances. Some advantages of this algorithm include its lack of training time because of being lazy, simple implementation with specified K and Euclidean distance, lost value imputing, and excellent performance thanks to its independence from new data instances [52,53,54].

DT

Decision trees are ML algorithms and have a potential structure for induction and interpretation in the ML process. This algorithm consists of three node types in their structural tress: roots, internal nodes, and external nodes named leaves. The root node in DT belongs to the dataset attribute with high capability in discriminating the output classes, i.e., the most crucial variable in the study. The internal nodes link the root to external nodes in trees; therefore, this structure can trace the tree from the root to leaves mediated by internal nodes to obtain the IF–THEN rules. The external nodes or leaves are places where the samples can be classified. In reality, the number of leaves constitutes the number of induction rules extracted from the tree. The benefits of this induction structure include simplicity for interpretation, easy implementation because of less complicated calculations, and less need for data normalization [55,56,57,58].

Proposed method

In this study, using a meta-heuristic algorithm for optimizing water waves, a model is presented for predicting the risk of readmission of COVID-19 patients. In the proposed model, the novel WWO algorithm was adopted to minimize the classification error. This algorithm cannot make predictions alone, so it is combined with the ANN algorithm. In other words, the proposed model uses the WWO evolutionary algorithm to promote the accuracy and effectiveness of predicting the readmission risk of COVID-19 patients. In optimization problems, modeling natural and biological phenomena is an effective method. This algorithm uses the existing relationships between water waves and their feedback to the environment to solve optimization problems. In the WWO algorithm, like any metaheuristic or evolutionary algorithm, sets of initial solutions are encoded in the form of a population. In this meta-heuristic algorithm, each problem solution is identified as a wave, and sets of waves are considered as the initial population of the problem. In a WWO algorithm, each solution to a problem or wave is encoded with properties such as wave height or wavelength. In the WWO algorithm, the solutions to the problem are first encoded as waves and several waves are randomly scattered in the problem search space.

In the proposed framework, a multilayer neural network is first created based on the training data set. Subsequently, the desired ANN is created as an array of weights and thresholds under the initial population of water waves. Afterward, a WWO algorithm is implemented on them to finally develop the best water wave or the corresponding ANN to predict the risk of readmission of COVID-19 patients. A multi-layered neural network with two hidden layers and five hidden nodes in each layer is randomly selected for initial training by 70% of the entire data. The desired ANN configuration is optimized by the WWO algorithm and implemented in MATLAB R2016a to select the best member of the neural network set. The performance of the proposed model was compared with other methods. To calculate the average error in the experiments, the number of experiments was considered to be 50, and the mean error in all these experiments was announced as the final result. Mean square error (MSE) and root mean square error (RMSE) were used as the objective function to reduce the error. In 50 experiments, values of 0.17 and 0.41 were respectively calculated. In the proposed method, an ANN is initially created by training, and several neural networks are encoded in the form of water waves. The waves are optimized, and then each of these waves (corresponding neural network) is evaluated by the objective function of the problem, and the best water wave or neural network is identified in this iteration of the algorithm. Any ANN or water wave that has a smaller classification error is considered to be better qualified. Figure 1 describes the steps of the proposed model.

Fig. 1
figure 1

The proposed method framework for predicting the risk of COVID-19 readmission

Models evaluation

To evaluate the performance of each algorithm, tenfold cross-validation was used to obtain reliable results for assessing prediction models or obtaining reliable results. The original training dataset was divided into 10 folds through stratified random sampling. For the ith iteration, fold i was considered as the test data, and the remaining nine folds were used to train the model. The model was assessed using the test data, and the procedure was repeated for 10 iterations. The evaluation results of 10 iterations were collected to compute the mean value and standard deviation.

The performance of models was measured using accuracy, precision, recall, specificity, and F-measure metrics. These evaluation criteria are commonly reported in the evaluation of models with ML [59], and their definitions are listed in Table 2. Furthermore, Friedman's statistical technique was adopted to compare the algorithms more precisely and select the algorithm with the highest efficiency. This test assigns a rank to each algorithm and the best algorithm has a lower rating. The null hypothesis states that all the algorithms are the same, while rejecting the null hypothesis shows that the compared algorithms significantly differ. In this paper, we set the significance level to α = 0.05.

Table 2 Definitions of evaluation metrics

Results

Sample characteristics

After applying the exclusion criteria, the records of 1225 discharged COVID-19 patients remained. Of these, 887 (72.40%) were male and 338 (27.60%) were women, and the median age of the participants was 57.25 years (interquartile 18–100). Of these, 89 patients had readmission, and 1136 patients had no readmission.

Feature selection

Given that MHAs are naturally random and the solutions may be slightly different in each independent execution, each algorithm was executed 20 times, and the average of the results was obtained after 20 independent executions. Furthermore, in all algorithms, the population size and the maximum number of iterations were set to 50 and 100, respectively. The mean fitness value of each algorithm, the accuracy of the KNN classifier based on the selected features, and the number of selected features are presented in Table 3.

Table 3 Comparison of algorithms in terms of different criteria in 20 runs

The numerical results show that the HOA algorithm is significantly superior to the other algorithms in terms of all three criteria [accuracy: 0.924 (95% CI 0.923 to 0.925)]. The most important variables to predict the readmission rate selected by HOA were age, sex, prior LOS, fever, dry coughs, cardiovascular disease, diabetes, hypertension, prior oxygen therapy, CRP, creatinine, ESR, D-dimer, ALT/ASP, absolute lymphocyte/ neutrophil count, pleural effusion and consolidation.

Model implementation

To select the best predictive performance, three traditional ML algorithms and a hybrid technique were trained, and their performance was compared according to the selected evaluation criteria. The steps of the proposed method (hybrid) for predicting the readmission risk of COVID-19 patients are as follows:

First, a multilayer artificial neural network with a specified number of hidden layers was trained by the COVID-19 dataset. Next, by training the desired ANN, the values and biases of the multilayer neural network were quantified, so several multilayer ANNs were developed with the same weights and thresholds and with relatively different values. Then, each of these neural networks created by the proposed coding was converted into several arrays or water waves, which constituted the initial population quantification step in the WWO algorithm. Each of the water waves or the initial population of the corresponding ANNs was delivered as an input to the wave optimization algorithm; then, each wave (the corresponding neural network) was evaluated by the objective function of the problem and the best water wave or the same neural network was detected in this iteration. The WWO algorithm was implemented on neural networks or water waves to extract the best wave or neural network to predict the re-admission risk in the last iteration. Finally, the efficiency of the proposed method was assessed based on model evaluation criteria.

Note that the performance of ML models on the initial dataset as well as the dataset after feature selection was implemented (trained) and compared separately (see Table 4).

Table 4 The performance of ML algorithms before and after preprocessing

Generally, the results in Table 4 reveal that the performance of ML algorithms in the prediction of readmission has improved significantly after preprocessing. The WWO classifier was introduced as the best algorithm for predicting the readmission risk of COVID-19 patients with a 0.9705 accuracy, 0.9729 precision, 0.9869 recall, 0.9259 specificity, and 0.9795 F-measure. The SVM with accuracy, precision, recall, specificity, and F-measure of 0.821, 0.743, 0.792, 0.921, and 0.767 had the poorest performance, respectively.

Given that the data in the outcome classes are unevenly distributed, the F1 score criterion is a more appropriate indicator than accuracy for model evaluation. Herein, due to the imbalance of readmission and non-readmission classes, according to Table 4, the F1 score criterion related to the proposed model was evaluated. With a value of 0.9795, the F2 index indicated the appropriate performance of the proposed model compared to other ML algorithms.

AUC is an effective technique to summarize the accuracy of predictive models. Its value ranges from 0 to 1, with the value of 0 indicating a completely incorrect test and 1 denoting a completely accurate diagnostic test. In general, an AUC of 0.5 does not indicate any discrimination, 0.7 to 0.8 is considered acceptable, 0.8 to 0.9 is considered excellent, and > 0.9 is regarded as prominent [60]. According to Fig. 2, the ACU of the proposed model in the test dataset was excellent.

Fig. 2
figure 2

ROC curve for the proposed model

Discussion

Accurately identifying the COVID-19 readmission risk can provide a practical solution for clinical decision-making to prevent disease reinfection and recurrent [31]. The present study retrospectively identified the most contributing factors in predicting the risk of hospital readmission in COVID-19 patients. The most important clinical variables were first selected and then leveraged as inputs for constructing ML models including KNN, SVM, WWO, and DT. Finally, the efficiency and performance of developed models were evaluated and compared.

Improving the quality of healthcare services and optimal management of hospital resources has given rise to the need to design predictive models to predict future disease behaviour and outcome [9, 10]. Using decision support systems to predict patient readmission and disease recurrence plays a crucial role in improving care quality and safety [26, 32]. The need to reduce the costs of early readmission up to 30 days after discharge and promote satisfaction during the pandemic has attracted the attention of many researchers [61].

Many studies on ML application to predict readmission have focused on chronic conditions such as cardiovascular diseases [62,63,64,65,66,67,68], stroke [69,70,71,72,73], and respiratory diseases [74,75,76,77,78]. Shang (2021) [79], Vosough (2021) [80], and Lin (2019) [81] assessed the performance of ML algorithms in disease recurrence and readmission prediction. Their results showed that ML methods provide a reasonable level of accuracy and certainty in predicting hospital readmission for chronic patients.

Several efforts are also made to apply ML algorithms for the prediction of readmission risk of COVID-19 patients. Mejia et al. concluded that the lack of a valid and scientific model for predicting readmission of COVID-19 patients influences the higher mortality due to disease recurrence [82]. Afrash et al. suggested the ML-based predictive models as useful for managing limited healthcare resources during the COVID-19 pandemic [83]. Donnely et al. also stated that the prediction of COVID-19 readmission is a challenging but important task in preventing the devastating effects of disease recurrence or reinfection [22]. Gavin et al. presented a predictive model to predict 30-day readmission in COVID-19 patients based on the simplified hospital score method for reducing patient readmission and directing resources toward high-risk cases [84]. Hebert et al., developed a risk score model for early prediction of the hospital readmission risk using multiple logistic regression techniques [85]. Rodriguez et al. also proposed a predictive model for readmission of COVID-19 patients based on statistical regression techniques with an AUC-ROC of 0.871 [86].

Eckert et al. reported that predictive modeling for patient readmission based on ML methods can identify high-risk groups of patients with high accuracy; in this way, unplanned readmission and severe complications of the disease will be reduced [87].

Accordingly, Cuong et al. concluded that ML techniques had a greater ability to predict patient readmission during COVID-19 than traditional statistical methods [88]. Davazdahemami et al. used the ML method to predict early or emergency readmission (less than 7 days) in COVID-19 patients. Their proposed model with an AUC of 0.883 showed good performance [33]. Raftarai et al. compared the performance of selected ML algorithms for predicting readmission among COVID-19 hospitalized patients [32]. Jia et al. also assessed the performance of some ML algorithms to predict future deterioration and readmission risk among discharged patients with COVID-19 [89]. Koteswari et al. utilized ML techniques to predict the readmission probability of various COVID-19 cases [15]. In other studies by Ryu [90] Zhao [91], Darabi [92], Chen [93], and Shah [94], ML algorithms were applied to predict the likelihood of readmission of COVID-19 patients.

In our study, the results showed that the WWO algorithm with an accuracy of 0.9705, precision of 0.9729, recall of 0.9869, specificity of 0.9259, and F-measure of 0.9795 has the best capability for early prediction of the risk of readmission in discharged COVID-19 patients.

Selecting key variables affecting the COVID-19 readmission is critical to developing predictive models [9]. Using these variables as an input to ML models improves their performance [32]. Thus far, several studies have selected clinically important predictors for post-discharge COVID-19 recurrence and readmission risk. In Rodriguez's study, some variables (e.g., LDH, CRP, and ESR) were selected as the key factors in hospital readmission [86]. Mendito et al. also determined a number of clinical characteristics such as age, neutrophilia count, sequential organ failure assessment (SOFA), LDH, CRP, and D-dimer as highly contributing factors to the readmission of COVID-19 patients [95]. In the study by Duarte et al., polymerization, living in residential care homes, general malaise, thoracic pain, and hematologic symptoms along with headaches, depressive symptoms, nephrological manifestations, syncope or hypotension, and superinfection were selected as the most relevant factors in COVID-19 readmission [96]. In many studies, age, sex, BMI, length of stay (LOS), ICU hospitalization, and the presence of comorbidities were introduced as the most influencing factors on COVID-19 readmission [97]. In the study by Nematshahi et al., the increase in the time interval from discharge to readmission, age (over 60 years), sex (male), diabetes, elevated creatinine, and lung involvement were selected as influential factors in predicting the readmission of COVID-19 patients [98]. Similarly, in Jeon's research, age and sex were effective in increasing the risk of readmission of COVID-19 patients [99]. The presence of comorbidities, high BMI, adult age, and laboratory indicators such as CRP, creatinine, and ALT/ASP rate were also introduced as the major underlying factors for readmission in COVID-19 patients in Verna's study [100]. In a systematic review conducted by Akbari et al., it was concluded that male sex, white ethnicity, comorbid diseases, and old age affect COVID-19 readmission [101].

In our study, after comparing the performance of six MHAs for feature selection, the HOA method with a mean fitness value of 0.083 and a KNN accuracy of 0.924 achieved the best performance. A total of 17 highly correlated variables such as old age, high weight, dry coughs, fever, dyspnea, loss of smell, cardiovascular diseases, hypertension, CRP, ALT/ASP, SPO2, and leukocytosis were selected as the top predictors affecting COVID-19 readmission.

The proposed model can help healthcare providers in the timely detection of patient deterioration in order to reduce severe complications and the resulting mortalities. Although the current study presented an optimum performance in predicting the readmission risk of patients with COVID-19, it had several potential limitations and challenges. This was a retrospective and single-center dataset, which might have affected the quality, comprehensiveness, and generalizability of the data. In this situation, the existence of some non-integrated, incomplete, error-prone, and abnormal data fields could have negatively impacted prediction. Therefore, to improve the consistency of data, the normal range of each variable was defined using the opinion of two infectious diseases specialists, a virologist, and a hematologist. Then, all the values that were outside the defined range (noisy fields) were specified and completed by referring to patient records or the responsible physician. In addition, the records with more than 70% of empty fields were removed and imputed by median and mode values substitution for continuous and discrete variables, respectively. Moreover, we used only four (albeit well-known) ML algorithms for prediction analyses based on some clinical features. The accuracy and generalizability of our models can be enhanced if other ML techniques are tested on a larger, multicenter, and prospective dataset containing time-varying covariates to identify a more insightful set of longitudinal factors related to COVID-19 readmission. Besides, the external validation method should be used to confirm the results of the present study. Another possible limitation was that this study did not describe any causal relationship between the predictor and outcome variables. This was not the main purpose of this research, but it can be addressed in future studies. Overall, the integrity of predictive models based on ML algorithms depends on the comprehensiveness of the dataset. Since all analyses were based on a single-center dataset, the results of this study may not be generalizable enough for national use. In future research, by analyzing data from multiple COVID-19 care centers in different provinces of Iran, the comprehensiveness and generalizability of the proposed model can be improved.

Conclusions

Our models have a satisfactory potential in equipping physicians and healthcare policymakers with a practical and effective tool for the timely prediction of hospital readmission of COVID-19 patients. The insights provided by these predictive models may help better care delivery, lessen clinicians’ workload, and ultimately enhance both care quality and financial outcomes. In the present study, the proposed hybrid WWO algorithm yielded the best capability to predict COVID-19 hospital readmission based on influential features. In future studies, the proposed method can be applied to predict the risk of hospital readmissions for other chronic diseases. The MHA used in feature selection can also be improved.

Abbreviations

MOF:

Multi-organ failure

ICU:

Intensive care unit

ML:

Machine learning

DT:

Decision tree

SVM:

Support vector machine

MLP:

Multilayer perceptron

KNN:

K-nearest neighbors

HOAs:

Horse herd optimization

AI:

Artificial intelligence

LOS:

Length of stay

ALT:

Alanine aminotransferase

ASP:

Aspartate aminotransferase

ESR:

Erythrocyte sedimentation rate

CRP:

C-reactive protein

PSO:

Particle swarm optimization

GA:

Genetic algorithm

GWO:

Grey wolf optimization

DE:

Differential evolution

KNN:

K-nearest neighbors

ANNs:

Artificial neural

References

  1. Navik U, Bhatti J, Sheth V, Jawalekar S, Bhatti G, Kalra S Multi-organ failure in COVID-19 patients: a possible mechanistic approach. Authorea Preprints. 2020.

  2. Hu Y, Deng H, Huang L, Xia L, Zhou X. Analysis of characteristics in death patients with COVID-19 pneumonia without underlying diseases. Acad Radiol. 2020;27(5):752.

    PubMed  PubMed Central  Article  Google Scholar 

  3. Zhou F, Yu T, Du R, Fan G, Liu Y, Liu Z, et al. Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. The Lancet. 2020;395(10229):1054–62.

    CAS  Article  Google Scholar 

  4. McAuley AJ, Kuiper MJ, Durr PA, Bruce MP, Barr J, Todd S, et al. Experimental and in silico evidence suggests vaccines are unlikely to be affected by D614G mutation in SARS-CoV-2 spike protein. npj Vaccines. 2020. https://doi.org/10.1038/s41541-020-00246-8.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Emanuel EJ, Persad G, Upshur R, Thome B, Parker M, Glickman A, et al. Fair allocation of scarce medical resources in the time of Covid-19. N Engl J Med. 2020;382(21):2049–55.

    PubMed  Article  Google Scholar 

  6. Aleem A, Akbar Samad AB, Slenker AK. Emerging variants of SARS-CoV-2 and novel therapeutics against coronavirus (COVID-19). StatPearls. Treasure Island (FL): StatPearls Publishing Copyright © 2022, StatPearls Publishing LLC.; 2022.

  7. Naghavi S, Kavosh A, Adibi I, Shaygannejad V, Arabi S, Rahimi M, et al. COVID-19 infection and hospitalization rate in Iranian multiple sclerosis patients: what we know by May 2021. Mult Scler Relat Disord. 2022;57:103335. https://doi.org/10.1016/j.msard.2021.103335.

    Article  PubMed  Google Scholar 

  8. Szente Fonseca SN, de Queiroz SA, Wolkoff AG, Moreira MS, Pinto BC, Valente Takeda CF, et al. Risk of hospitalization for Covid-19 outpatients treated with various drug regimens in Brazil: comparative analysis. Travel Med Infect Dis. 2020;38: 101906.

    PubMed  PubMed Central  Article  Google Scholar 

  9. Verna EC, Landis C, Brown RS, Mospan AR, Crawford JM, Hildebrand JS, et al. Factors associated with readmission in the US following hospitalization with COVID-19. Clin Infect Dis. 2021. https://doi.org/10.1093/cid/ciab464.

    Article  PubMed  Google Scholar 

  10. Jeon WH, Seon JY, Park SY, In-Hwan O. Analysis of risk factors on readmission cases of COVID-19 in the Republic of Korea: using nationwide health claims data. Int J Environ Res Public Health. 2020;17(16):5844. https://doi.org/10.3390/ijerph17165844.

    CAS  Article  PubMed Central  Google Scholar 

  11. Akbari A, Fathabadi A, Razmi M, Zarifian A, Amiri M, Ghodsi A, et al. Characteristics, risk factors, and outcomes associated with readmission in COVID-19 patients: a systematic review and meta-analysis. Am J Emerg Med. 2022;52:166–73.

    PubMed  Article  Google Scholar 

  12. Rahmani K, Shavaleh R, Forouhi M, Disfani HF, Kamandi M, Dezfuli AAZ, et al. Effectiveness of COVID-19 vaccines and post-vaccination SARS-COV 2 infection, hospitalization, and mortality: a systematic review and meta-analysis of observational studies. medRxiv. 2021;21(11):100331.

    Google Scholar 

  13. Wallmann R, Llorca J, Gómez-Acebo I, Ortega ÁC, Roldan FR, Dierssen-Sotos T. Prediction of 30-day cardiac-related-emergency-readmissions using simple administrative hospital data. Int J Cardiol. 2013;164(2):193–200.

    PubMed  Article  Google Scholar 

  14. Dharmarajan K, Hsieh AF, Lin Z, Bueno H, Ross JS, Horwitz LI, et al. Diagnoses and timing of 30-day readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia. JAMA. 2013;309(4):355–63.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  15. Koteswari MJL, Balaji M, Sainadh K, Kavya KCS, Ch K. Reducing Covid-19 readmissions using machine learning. Turkish J Physiother Rehabilit. 2021;32:2.

    Google Scholar 

  16. Baillie CA, VanZandbergen C, Tait G, Hanish A, Leas B, French B, et al. The readmission risk flag: using the electronic health record to automatically identify patients at risk for 30-day readmission. J Hosp Med. 2013;8(12):689–95.

    PubMed  Article  Google Scholar 

  17. Jamei M, Nisnevich A, Wetchler E, Sudat S, Liu E. Predicting all-cause risk of 30-day hospital readmission using artificial neural networks. PLoS ONE. 2017;12(7): e0181173.

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  18. Allaudeen N, Schnipper JL, Orav EJ, Wachter RM, Vidyarthi AR. Inability of providers to predict unplanned readmissions. J Gen Intern Med. 2011;26(7):771–6.

    PubMed  PubMed Central  Article  Google Scholar 

  19. Lo YT, Liao JC, Chen MH, Chang CM, Li CT. Predictive modeling for 14-day unplanned hospital readmission risk by using machine learning algorithms. BMC Med Inf Decis Mak. 2021;21(1):1–11.

    Article  Google Scholar 

  20. Sotoodeh Ghorbani S, Taherpour N, Bayat S, Ghajari H, Mohseni P, Hashemi Nazari SS. Epidemiologic characteristics of cases with reinfection, recurrence, and hospital readmission due to COVID-19: a systematic review and meta-analysis. J Med Virol. 2021;94(1):44–53.

    PubMed  Article  CAS  Google Scholar 

  21. Rosted E, Thomsen TG, Krogsgaard M, Busk H, Geisler A, Thestrup Hansen S, et al. On the frontline treating COVID-19: a pendulum experience—from meaningful to overwhelming—for Danish healthcare professionals. J Clin Nurs. 2021;30(23–24):3448–55.

    PubMed  Article  Google Scholar 

  22. Donnelly JP, Wang XQ, Iwashyna TJ, Prescott HC. Readmission and death after initial hospital discharge among patients with COVID-19 in a large multihospital system. JAMA. 2021;325(3):304–6.

    CAS  PubMed  Article  Google Scholar 

  23. Yeo I, Baek S, Kim J, Elshakh H, Voronina A, Lou MS, et al. Assessment of thirty-day readmission rate, timing, causes and predictors after hospitalization with COVID-19. J Intern Med. 2021;290(1):157–65.

    CAS  PubMed  Article  Google Scholar 

  24. Alanli R, Kucukay MB, Yalcin KS. Readmission rates of patients with COVID-19 after hospital discharge. Rev Assoc Méd Bras. 2021;67(11):1610–5. https://doi.org/10.1590/1806-9282.20210675.

    Article  PubMed  Google Scholar 

  25. Leijte WT, Wagemaker NMM, van Kraaij TDA, de Kruif MD, Mostard GJM, Leers MPG, et al. [Mortality and re-admission after hospitalization with COVID-19]. Nederlands tijdschrift voor geneeskunde. 2020;164.

  26. Rodriguez VA, Bhave S, Chen R, Pang C, Hripcsak G, Sengupta S, et al. Development and validation of prediction models for mechanical ventilation, renal replacement therapy, and readmission in COVID-19 patients. J Am Med Inform Assoc. 2021;28(7):1480–8.

    PubMed  PubMed Central  Article  Google Scholar 

  27. Ullah SMA, Islam M, Mahmud S, Nooruddin S, Raju S, Haque M. Scalable telehealth services to combat novel coronavirus (COVID-19) pandemic. SN Comput Sci. 2021;2(1):1–8.

    Article  Google Scholar 

  28. Islam M, Mahmud S, Muhammad L, Nooruddin S, Ayon SI. Wearable technology to assist the patients infected with novel coronavirus (COVID-19). SN Comput Sci. 2020;1(6):1–9.

    Article  Google Scholar 

  29. Islam M, Ullah SMA, Mahmud S, Raju S. Breathing aid devices to support novel coronavirus (COVID-19) infected patients. SN Comput Sci. 2020;1(5):1–8.

    Google Scholar 

  30. Rahman MM, Manik MMH, Islam MM, Mahmud S, Kim J-H, editors. An automated system to limit COVID-19 using facial mask detection in smart city network. 2020 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS); 2020: IEEE.

  31. Citu IM, Citu C, Gorun F, Neamtu R, Motoc A, Burlea B, et al. Using the NYHA classification as forecasting tool for hospital readmission and mortality in heart failure patients with COVID-19. J Clin Med. 2022;11(5):1382.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  32. Raftarai A, Mahounaki RR, Harouni M, Karimi M, Olghoran SK. Predictive models of hospital readmission rate using the improved adaboost in COVID-19. In: Saba T, Khan AR, editors. Intelligent computing applications for COVID-19: predictions, diagnosis, and prevention. Boca Raton: CRC Press; 2021. p. 67–86. https://doi.org/10.1201/9781003141105-4.

    Chapter  Google Scholar 

  33. Davazdahemami B, Zolbanin HM, Delen D. An explanatory machine learning framework for studying pandemics: The case of COVID-19 emergency department readmissions. Decis Support Syst. 2022. https://doi.org/10.1016/j.dss.2022.113730.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Amritphale A, Chatterjee R, Chatterjee S, Amritphale N, Rahnavard A, Awan GM, et al. Predictors of 30-day unplanned readmission after carotid artery stenting using artificial intelligence. Adv Ther. 2021;38(6):2954–72.

    PubMed  PubMed Central  Article  Google Scholar 

  35. Hogan AH, Brimacombe M, Mosha M, Flores G. Comparing artificial intelligence and traditional methods to identify factors associated with pediatric asthma readmission. Acad Pediatr. 2022;22(1):55–61. https://doi.org/10.1016/j.acap.2021.07.015.

    Article  PubMed  Google Scholar 

  36. Aljouie AF, Almazroa A, Bokhari Y, Alawad M, Mahmoud E, Alawad E, et al. Early prediction of COVID-19 ventilation requirement and mortality from routinely collected baseline chest radiographs, laboratory, and clinical data with machine learning. J Multidiscip Healthc. 2021;14:2017–33.

    PubMed  PubMed Central  Article  Google Scholar 

  37. Bolourani S, Brenner M, Wang P, McGinn T, Hirsch JS, Barnaby D, et al. A machine learning prediction model of respiratory failure within 48 hours of patient admission for COVID-19: model development and validation. J Med Internet Res. 2021;23(2): e24246.

    PubMed  PubMed Central  Article  Google Scholar 

  38. Abdullah AA, Hafidz SA, Khairunizam W. Performance comparison of machine learning algorithms for classification of chronic kidney disease (CKD). J Phys Conf Ser. 2020;1529(5):052077. https://doi.org/10.1088/1742-6596/1529/5/052077.

    Article  Google Scholar 

  39. Arvind V, Kim JS, Cho BH, Geng E, Cho SK. Development of a machine learning algorithm to predict intubation among hospitalized patients with COVID-19. J Crit Care. 2021;62:25–30.

    CAS  PubMed  Article  Google Scholar 

  40. Asraf A, Islam M, Haque M. Deep learning applications to combat novel coronavirus (COVID-19) pandemic. SN Comput Sci. 2020;1(6):1–7.

    Article  Google Scholar 

  41. Islam MM, Karray F, Alhajj R, Zeng J. A review on deep learning techniques for the diagnosis of novel coronavirus (COVID-19). IEEE Access. 2021;9:30551–72.

    PubMed  Article  Google Scholar 

  42. Islam MZ, Islam MM, Asraf A. A combined deep CNN-LSTM network for the detection of novel coronavirus (COVID-19) using X-ray images. Inf Med Unlocked. 2020;20: 100412.

    Article  Google Scholar 

  43. Saha P, Sadi MS, Islam MM. EMCNet: automated COVID-19 diagnosis from X-ray images using convolutional neural network and ensemble of machine learning classifiers. Inf Med Unlocked. 2021;22: 100505.

    Article  Google Scholar 

  44. Rahman MM, Islam M, Manik M, Hossen M, Al-Rakhami MS. Machine learning approaches for tackling novel coronavirus (COVID-19) pandemic. Sn Comput Sci. 2021;2(5):1–10.

    Google Scholar 

  45. Shanbehzadeh M, Orooji A, Kazemi-Arpanahi H. Comparing of data mining techniques for predicting in-hospital mortality among patients with covid-19. J Biostat Epidemiol. 2021;7(2):154–73.

    Google Scholar 

  46. Dan T, Li Y, Zhu Z, Chen X, Quan W, Hu Y, et al., editors. Machine Learning to Predict ICU Admission, ICU Mortality and Survivors’ Length of Stay among COVID-19 Patients: Toward Optimal Allocation of ICU Resources. 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM); 2020: IEEE.

  47. Lorenzen SS, Nielsen M, Jimenez-Solem E, Petersen TS, Perner A, Thorsen-Meyer H-C, et al. Using machine learning for predicting intensive care unit resource use during the COVID-19 pandemic in Denmark. Sci Rep. 2021;11(1):1–10.

    Article  CAS  Google Scholar 

  48. Sun L, Mo Z, Yan F, Xia L, Shan F, Ding Z, et al. Adaptive feature selection guided deep forest for COVID-19 classification with chest CT. IEEE J Biomed Health Inform. 2020;24(10):2798–805.

    PubMed  Article  Google Scholar 

  49. Suthaharan S. Support vector machine. Machine learning models and algorithms for big data classification: Springer; 2016. p. 207–35.

    Google Scholar 

  50. Pisner DA, Schnyer DM. Support vector machine. In: Machine learning. Elsevier; 2020. p. 101–21. https://doi.org/10.1016/B978-0-12-815739-8.00006-7.

    Chapter  Google Scholar 

  51. Pradhan A. Support vector machine-a survey. Int J Emerg Technol Adv Eng. 2012;2(8):82–5.

    Google Scholar 

  52. Deng Z, Zhu X, Cheng D, Zong M, Zhang S. Efficient kNN classification algorithm for big data. Neurocomputing. 2016;195:143–8.

    Article  Google Scholar 

  53. Zhang S, Cheng D, Deng Z, Zong M, Deng X. A novel kNN algorithm with data-driven k parameter computation. Pattern Recogn Lett. 2018;109:44–54.

    Article  Google Scholar 

  54. Cheng D, Zhang S, Deng Z, Zhu Y, Zong M. kNN algorithm with data-driven k value. In: Luo X, Yu JX, Li Z, editors. Advanced data mining and applications. Cham: Springer International Publishing; 2014. p. 499–512. https://doi.org/10.1007/978-3-319-14717-8_39.

    Chapter  Google Scholar 

  55. Sharma H, Kumar S. A survey on decision tree algorithms of classification in data mining. Int J Sci Res. 2016;5(4):2094–7.

    Google Scholar 

  56. Navada A, Ansari AN, Patil S, Sonkamble BA, editors. Overview of use of decision tree algorithms in machine learning. 2011 IEEE control and system graduate research colloquium; 2011: IEEE.

  57. Gupta B, Rawat A, Jain A, Arora A, Dhami N. Analysis of various decision tree algorithms for classification in data mining. Int J Comput Appl. 2017;163(8):15–9.

    Google Scholar 

  58. Song Y-Y, Ying L. Decision tree methods: applications for classification and prediction. Shanghai Arch Psychiatry. 2015;27(2):130.

    PubMed  PubMed Central  Google Scholar 

  59. Hossin M, Sulaiman MN. A review on evaluation metrics for data classification evaluations. Int J Data Min Knowl Manag Process. 2015;5(2):1.

    Article  Google Scholar 

  60. Hosmer DW, Lemeshow S, Sturdivant RX. Applied logistic regression. Wiley; 2013.

    Book  Google Scholar 

  61. Yeo I, Baek S, Kim J, Elshakh H, Voronina A, Lou M, et al. Assessment of thirty-day readmission rate, timing, causes and predictors after hospitalization with COVID-19. J Intern Med. 2021;290(1):157–65.

    CAS  PubMed  Article  Google Scholar 

  62. Kerexeta J, Artetxe A, Escolar V, Lozano A, Larburu N, editors. Predicting 30-day readmission in heart failure using machine learning techniques. HEALTHINF 2018 - 11th International Conference on Health Informatics, Proceedings; Part of 11th International Joint Conference on Biomedical Engineering Systems and Technologies, BIOSTEC 2018; 2018.

  63. Awan SE, Bennamoun M, Sohel F, Sanfilippo FM, Dwivedi G. Machine learning-based prediction of heart failure readmission or death: implications of choosing the right model and the right metrics. ESC Heart Failure. 2019;6(2):428–35.

    PubMed  PubMed Central  Article  Google Scholar 

  64. Mahajan SM, Ghani R. Using ensemble machine learning methods for predicting risk of readmission for heart failure. Studies in Health Technology and Informatics2019. p. 243–7.

  65. Najafi-Vosough R, Faradmal J, Hosseini SK, Moghimbeigi A, Mahjub H. Predicting hospital readmission in heart failure patients in Iran: a comparison of various machine learning methods. Healthc Inform Res. 2021;27(4):307–14.

    PubMed  PubMed Central  Article  Google Scholar 

  66. Sampedro-Gómez J, Higuero-Saavedra A, Lorenzo-Martín ÁL, Ramírez-Hernández P, Valenzuela-Serrano M, Sánchez PL. Prediction of in-hospital mortality and 30-day readmission in heart failure using machine learning. REC: CardioClinics. 2021.

  67. Sarijaloo F, Park J, Zhong X, Wokhlu A. Predicting 90 day acute heart failure readmission and death using machine learning-supported decision analysis. Clin Cardiol. 2021;44(2):230–7.

    PubMed  Article  Google Scholar 

  68. Shin S, Austin PC, Ross HJ, Abdel‐Qadir H, Freitas C, Tomlinson G. Machine learning vs. conventional statistical models for predicting heart failure readmission and mortality. ESC Heart Failure. 2021;8(1):106–15. https://doi.org/10.1002/ehf2.13073.

    Article  PubMed  Google Scholar 

  69. Hung LC, Sung SF, Ya-Han H. A machine learning approach to predicting readmission or mortality in patients hospitalized for stroke or transient ischemic attack. Appl Sci. 2020;10(18):6337. https://doi.org/10.3390/app10186337.

    CAS  Article  Google Scholar 

  70. Darabi N, Hosseinichimeh N, Noto A, Zand R, Abedi V. Machine learning-enabled 30-day readmission model for stroke patients. Front Neurol. 2021. https://doi.org/10.3389/fneur.2021.638267.

    Article  PubMed  PubMed Central  Google Scholar 

  71. Lineback CM, Garg R, Elissa O, Naidech AM, Holl JL, Prabhakaran S. Prediction of 30-day readmission after stroke using machine learning and natural language processing. Front Neurol. 2021. https://doi.org/10.3389/fneur.2021.649521.

    Article  PubMed  PubMed Central  Google Scholar 

  72. Chen Y-C, Chung J-H, Yeh Y-J, Lin H-F, Lin C-H, Hsien H-H, et al. Machine learning algorithms to predict 30-day readmission in patients with stroke: a prospective cohort study. 2020.

  73. Kommina L, Theerthagiri P, Payyavula Y, Vemula PS, Reddy GD. Post-Stroke readmission prediction model using machine learning algorithms. In: Rajeev Mathur CP, Gupta VK, Jat DS, Yadav N, editors. Emerging trends in data driven computing and communications: proceedings of DDCIoT 2021. Singapore: Springer; 2021. p. 53–65. https://doi.org/10.1007/978-981-16-3915-9_4.

    Chapter  Google Scholar 

  74. Goto T, Jo T, Matsui H, Fushimi K, Hayashi H, Yasunaga H. Machine learning-based prediction models for 30-day readmission after hospitalization for chronic obstructive pulmonary disease. COPD J Chronic Obstr Pulm Dis. 2019;16(5–6):338–43. https://doi.org/10.1080/15412555.2019.1688278.

    Article  Google Scholar 

  75. Min X, Bin Y, Wang F. Predictive modeling of the hospital readmission risk from patients’ claims data using machine learning: a case study on COPD. Sci Rep. 2019. https://doi.org/10.1038/s41598-019-39071-y.

    Article  PubMed  PubMed Central  Google Scholar 

  76. Verma VK, Lin WY, editors. A Machine Learning-Based Predictive Model for 30-Day Hospital Readmission Prediction for COPD Patients. Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics; 2020.

  77. Kaskovich S, Hemmrich M, Venable L, Carey K, Churpek M, Press V. Matching patients with chronic obstructive pulmonary disease (COPD) to personalized care: a novel machine learning tool to predict cause of 90-Day readmission. D102 OPTIMIZING OUTCOMES IN COPD: American Thoracic Society; 2019. p. A7119-A.

  78. Hemmrich M, Kaskovich S, Venable L, Carey K, Churpek M, Press V. Accuracy comparison of a machine learning readmission prediction model with hospital and pearl scores for chronic obstructive pulmonary disease (COPD) inpatients. D102 OPTIMIZING OUTCOMES IN COPD: American Thoracic Society; 2019. p. A7118-A.

  79. Shang Y, Jiang K, Wang L, Zhang Z, Zhou S, Liu Y, et al. The 30-days hospital readmission risk in diabetic patients: predictive modeling with machine learning classifiers. BMC Med Inform Decis Mak. 2021;21(Suppl 2):57.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  80. Najafi-Vosough R, Faradmal J, Hosseini SK, Moghimbeigi A, Mahjub H. Predicting hospital readmission in heart failure patients in Iran: A Comparison of Various Machine Learning methods. Healthc Inform Res. 2021;27(4):307–14.

    PubMed  PubMed Central  Article  Google Scholar 

  81. Lin Y, Wu JY, Lin K, Hu YH, Kong GL. Prediction of intensive care unit readmission for critically ill patients based on ensemble learning Beijing da. J Peking Univ Health Sci. 2021;53(3):566–72.

    CAS  Google Scholar 

  82. Mejia OAV, Borgomoni GB, Silveira LMV, Guerreiro GP, Falcão Filho ATG, Goncharov M, et al. The arrival of COVID-19 in Brazil and the impact on coronary artery bypass surgery. J Card Surg. 2021;36(9):3070–7.

    PubMed  Article  Google Scholar 

  83. Afrash MR, Kazemi-Arpanahi H, Shanbehzadeh M, Nopour R, Mirbagheri E. Predicting hospital readmission risk in patients with COVID-19: a machine learning approach. Inform Med Unlocked. 2022;30:100908. https://doi.org/10.1016/j.imu.2022.100908.

    Article  PubMed  PubMed Central  Google Scholar 

  84. Gavin W, Rager J, Russ J, Subramoney K, Kara A. Accuracy of the Simplified hospital score in predicting COVID-19 readmissions-exploring outcomes from a hospital-at-home program. J Healthc Manag Am Coll Healthc Executives. 2021;67(1):54–62.

    Google Scholar 

  85. Hebert KJ, Linder BJ, Gettman MT, Ubl D, Habermann EB, Lyon TD, et al. A contemporary analysis of ureteral reconstruction 30-day morbidity utilizing the national surgical quality improvement program database: comparison of minimally invasive vs open approaches. J Endourol. 2022;36(2):209–15.

    PubMed  Article  Google Scholar 

  86. Rodriguez VA, Bhave S, Chen R, Pang C, Hripcsak G, Sengupta S, et al. Development and validation of prediction models for mechanical ventilation, renal replacement therapy, and readmission in COVID-19 patients. J Am Med Inform Assoc. 2021;28(7):1480–8. https://doi.org/10.1093/jamia/ocab029.

    Article  PubMed  PubMed Central  Google Scholar 

  87. Eckert C, Nieves-Robbins N, Spieker E, Louwers T, Hazel D, Marquardt J, et al. Development and prospective validation of a machine learning-based risk of readmission model in a large military hospital. Appl Clin Inform. 2019;10(2):316–25.

    PubMed  PubMed Central  Article  Google Scholar 

  88. Cuong L, Wang D. A comparison of machine learning methods to predict hospital readmission of diabetic patient. Estudios De Economia Aplicada. 2021.

  89. Jia L, Wei Z, Zhang H, Wang J, Jia R, Zhou M, et al. An interpretable machine learning model based on a quick pre-screening system enables accurate deterioration risk prediction for COVID-19. Sci Rep. 2021;11(1):23127.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  90. Ryu B, Yoo S, Kim S, Choi J. Thirty-day hospital readmission prediction model based on common data model with weather and air quality data. Sci Rep. 2021;11(1):1–9.

    Article  CAS  Google Scholar 

  91. Zhao P, Yoo I, Naqvi SH. Early prediction of unplanned 30-day hospital readmission: model development and retrospective data analysis. JMIR Med Inform. 2021;9(3): e16306.

    PubMed  PubMed Central  Article  Google Scholar 

  92. Darabi N, Hosseinichimeh N, Noto A, Zand R, Abedi V. Machine learning-enabled 30-day readmission model for stroke patients. Front Neurol. 2021;12:425.

    Article  Google Scholar 

  93. Chen L, Chen S. Prediction of readmission in patients with acute exacerbation of chronic obstructive pulmonary disease within one year after treatment and discharge. BMC Pulm Med. 2021;21(1):1–17.

    CAS  Article  Google Scholar 

  94. Shah AA, Devana SK, Lee C, Bugarin A, Lord EL, Shamie AN, et al. Prediction of major complications and readmission after lumbar spinal fusion: a machine learning-driven approach. World Neurosurg. 2021;152:e227–34.

    PubMed  Article  Google Scholar 

  95. Menditto VG, Fulgenzi F, Bonifazi M, Gnudi U, Gennarini S, Mei F, et al. Predictors of readmission requiring hospitalization after discharge from emergency departments in patients with COVID-19. Am J Emerg Med. 2021;46:146–9.

    PubMed  PubMed Central  Article  Google Scholar 

  96. Romero-Duarte Á, Rivera-Izquierdo M, Láinez-Ramos-Bossini AJ, Redruello-Guerrero P, Cárdenas-Cruz A. Factors associated with readmission to the Emergency Department in a cohort of COVID-19 hospitalized patients. 2021.

  97. Green H, Yahav D, Eliakim-Raz N, Karny-Epstein N, Kushnir S, Shochat T, et al. Risk-factors for re-admission and outcome of patients hospitalized with confirmed COVID-19. Sci Rep. 2021;11(1):1–8.

    Article  CAS  Google Scholar 

  98. Nematshahi M, Soroosh D, Neamatshahi M, Attarian F, Rahimi F. Factors predicting readmission in patients with COVID-19. BMC Res Notes. 2021;14(1):1–6.

    Article  CAS  Google Scholar 

  99. Jeon W-H, Seon JY, Park S-Y, Oh I-H. Analysis of risk factors on readmission cases of COVID-19 in the Republic of Korea: using nationwide health claims data. Int J Environ Res Public Health. 2020;17(16):5844.

    CAS  PubMed Central  Article  Google Scholar 

  100. Verna EC, Landis C, Brown RS Jr, Mospan AR, Crawford JM, Hildebrand JS, et al. Factors associated with readmission in the us following hospitalization with COVID-19. Clin Infect Dis. 2021;5(20):2021.

    Google Scholar 

  101. Akbari A, Fathabadi A, Razmi M, Zarifian A, Amiri M, Ghodsi A, et al. Characteristics, risk factors, and outcomes associated with readmission in COVID-19 patients: a systematic review and meta-analysis. A J Emerg Med. 2022;52:166–73. https://doi.org/10.1016/j.ajem.2021.12.012.

    Article  Google Scholar 

Download references

Acknowledgements

We thank the Research Deputy of the Abadan University of Medical Sciences for financially supporting this project. We also would like to thank all experts who participated in this study.

Funding

There was no funding for this research project.

Author information

Authors and Affiliations

Authors

Contributions

HKA, and MShanbehzadeh: Project administration; Resources; Supervision; Roles/Writing—original draft. MShanbehzadeh, AY, HKA and MShafiee: Conceptualization; Formal analysis; Investigation; Roles/Writing—original draft; Funding acquisition; Methodology; Writing—review and editing. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Hadi Kazemi-Arpanahi.

Ethics declarations

Ethics approval and consent to participate

This article is extracted from a research project supported by the Abadan University of Medical Sciences (IR.ABADANUMS.REC.1400.136). The study was approved by the ethical committee of the Abadan Faculty of Medical Sciences. All methods of the present study were performed in accordance with the relevant guidelines and regulations. Participation was voluntary, the consent was verbal, but all participants responded via email or text message to approve their participation. Participants had the right to withdraw from the study at any time without prejudice. All participants were required to sign a privacy agreement and study participation consent form before joining the expert panel. They were cognizant of the objectives of the study. Choose or not participate with them. If you do not participate in all the services received in this center, you will continue and nothing will change. We assured participants that participation in this study is not a risk for them. The authors told the participants that if they are interested in collaborating, please allow us in writing to access the required documents and information. On the other hand, the purposes of the study and the type of study were explained to the participants. All of the above is stated in the informed consent form.

Consent for publication

Not applicable.

Availability of data and material

All data generated and analyzed during the current study are not publicly available but are available from the corresponding author upon reasonable request and the Abadan University of Medical Sciences’ approval.

Competing interests

We declare that we have no significant competition for financial, professional, or personal interests that might have influenced the performance or presentation of the work described in this manuscript. We have described our potential competition for financial, professional, and/or personal interests in the space.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Shanbehzadeh, M., Yazdani, A., Shafiee, M. et al. Predictive modeling for COVID-19 readmission risk using machine learning algorithms. BMC Med Inform Decis Mak 22, 139 (2022). https://doi.org/10.1186/s12911-022-01880-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12911-022-01880-z

Keywords

  • COVID-19
  • Machine learning
  • Data mining
  • Patient readmission