Skip to main content

Cardiac patients’ surgery outcome and associated factors in Ethiopia: application of machine learning



Cardiovascular diseases are a class of heart and blood vessel-related illnesses. In Sub-Saharan Africa, including Ethiopia, preventable heart disease continues to be a significant factor, contrasting with its presence in developed nations. Therefore, the objective of the study was to assess the prevalence of death due to cardiac disease and its risk factors among heart patients in Ethiopia.


The current investigation included all cardiac patients who had cardiac surgery in the country between 2012 and 2023. A total of 1520 individuals were participated in the study. Data collection took place between February 2022 and January 2023. The study design was a retrospective cohort since the study track back patients’ chart since 2012. Machine learning algorithms were applied for data analysis. For machine learning algorithms comparison, lift and AUC was applied.


From all possible algorithms, logistic algorithm at 90%/10% was the best fit since it produces the maximum AUC value. In addition, based on the lift value of 3.33, it can be concluded that the logistic regression algorithm was performing well and providing substantial improvement over random selection. From the logistic regression machine learning algorithms, age, saturated oxygen, ejection fraction, duration of cardiac center stays after surgery, waiting time to surgery, hemoglobin, and creatinine were significant predictors of death.


Some of the predictors for the death of cardiac disease patients are identified as such special attention should be given to aged patients, for patients waiting for long periods of time to get surgery, lower saturated oxygen, higher creatinine value, lower ejection fraction and for patients with lower hemoglobin values.

Peer Review reports


Based on the WHO definition, cardiovascular diseases are a group of disorders of the heart and blood vessels. Cardiovascular disease is one of the leading responsible causes of death. The global burden of disease studies reported that there were about 422.7 million cases of cardiovascular disease, and 17.92 million died since 2015. Over the last two decades, the global disease trend has shifted from communicable to non-communicable diseases [1, 2]. According to World Health Organization statistics, 17.6 million people have died from cardiovascular disease (CVD) globally in 2012, accounting for 31 per cent of global deaths [3].

In low-income countries, living in poverty is highly affected by cardiovascular disease(CVD) [4]. Unlike developed countries [5], it remains main cause of preventable cardiac disease in the Sub-Saharan Africa, due to educational, environmental, and economic disadvantages and reduced access to health care [6]. In North West Ethiopia, from all cardiovascular disease assessments, nearly half \(\left(49\text{\%}\right)\) of the study participants had valvular heart disease [7].

Current epidemiological data on valvular heart disease is critical for better understanding this evolving disease spectrum and allocating healthcare resources accordingly [8]. Epidemiological research on acquired and congenital valvular heart disease is limited. Changing the global trend of valvular heart disease is being examined due to ageing and advances in treatments, which is critical for advances in clinical practice and the preparation of health care policy. Because of ageing, the global prevalence of heart disease is increasing [9].

In Ethiopian, it is time for effective population-based, preventive health strategies for primary and secondary prevention of established cardiovascular disorders and risk factors [10]. Ethiopian population has seen major epidemiological shift. The main causes of death and illness change from infectious to non-communicable diseases [11]. One of the significant causes of morbidity and mortality is cardiovascular disease. However, there are no productive novel solutions to lower cardiovascular risk because there is a mistaken imagination that cardiovascular disease is solely a problem of the rich and developed world. There is also a tendency to associate cardiovascular mainly with hyper-nutrition [12, 13]. Cardiac disease is neglected by politicians and health related stakeholders [14].

Sufficient report on cardiovascular disorder is generally scarce in developing countries, including Ethiopia. According to a study in Tikur Anbessa and St.Paul’s Hospitals, cardiovascular diseases were one of the main causes of mortality and the most common reasons for hospitalization [15]. Evidence from cardiac center Ethiopia stated that cardiac disease patients increase dramatically from time to time. Another evidence from ministry of health shows, more than 40,000 cardiac disease patients have been waiting for cardiac surgery in Ethiopia since 2020. A report from cardiac center Ethiopia, around 7000 cardiac disease patients seek cardiac surgery at the waiting list [16].

Preventive treatment at an early stage of the disease is widely accepted to reduce morbidity and mortality. The benefits of early surgery should be balanced against the surgical risks and long-term outcomes. There is no adequate evidence on cardiac disease, specifically after cardiac surgery in Ethiopia. As such the study will serve as to deliver update information on cardiac patients’ status after surgery. In addition, this study aimed to compare the best machine learning algorithms for cardiac surgery outcomes and to identify the responsible predictors of death outcome.

Participants and methods

Study area and period

The study was done at two cardiac centers named as “Cardiac Center Ethiopia” and “Elouzeir Cardiac Center”. Those cardiac centers are located in Adis Ababa, the capital city of Ethiopia. Cardiac center Ethiopia was established in 2009 where as Elouzeir Cardiac Center was established in 2016. There are 15 cardiothoracic surgeons, from all, five can do open-heart surgery, one specialist might serve at more than one center. In cardiac center Ethiopia, more than 7000 patients were under waiting list to get the cardiac surgery service. Sine establishment, 1520 cardiac patients had surgery at those centers. Cardiac Center Ethiopia, currently provides the largest cardiac surgery in Ethiopia. This surgical center started to provide surgery service for patients with cardiac disease since 2009. A study was performed from 2012 to 2023 through utilizing patients’ follow-up charts. One of the study areas, called “cardiac center Ethiopia” (Fig. 1).

Fig. 1
figure 1

Cardiac center in Ethiopia

Study design:

The research was conducted using a retrospective cohort study design, focusing on a population of cardiac patients in Ethiopia who had surgery and follow-up appointments between 2012 and 2023.

Study population:

The study population was cardiac patients who had heart surgery from 2012 to 2023 in Ethiopia. In the study period, there are two cardiac centers in Ethiopia which serve as a fully functional cardiac surgery. Both cardiac centers were part of the investigation. In those centers, 1520 patients had cardiac surgery service. All 1520 patients were part of the study. Regarding inclusion criteria, all cardiac patients who have had surgery since 2012 was included.

Data Collection technique:

The information was gathered by utilizing structured questionnaires on an online data collection platform called Kobo Toolbox. The questionnaire was prepared in English language as of the chart form. The questionnaire was assessed in randomly selected patients’ charts; as such, pre-test was performed to check data quality. Health professionals were participated in the data collection journey. The data collection involved the participation of BSc nurses and medical doctors. In Ethiopia, there is a standardized patient chart form, developed according to the World Health Organization (WHO) standards, is universally used. Due to the absence of computerized data management systems, the data collectors manually retrieved patient information from the patients’ charts. The data collection process commenced after obtaining ethical clearance from Bahir Dar University Science College, and permission from the relevant healthcare provider institutions.

Outcome variables

Death status:

The outcome variable for a cardiac patient’s possible outcome (Death, Alive) after surgery.

The predictor variables:

The predictor variables are summarized in as the following (Fig. 2).

Fig. 2
figure 2

Conceptual framework

Data processing and analysis

Once the data collection process is complete, the collected data exported into Python and SPSS. Descriptive information characterises the study population. Categorical variables were expressed in terms of frequency and percentage whereas continuous variables were described in mean, range and standard deviation. Chi-square test of association was performed. Nine different machine learning algorithms were considered for comparison and analysis. Feature selection and analysis were performed using Python, IBM SPSS Modeler and IBM SPSS Statistics software.

Machine learning

Machine learning is the technology that allows computers to create algorithms that can mimic human intelligence. Ideas inspire it in artificial intelligence, probability and statistics, computer science, information theory, control theory, psychology, and philosophy [17, 18]. This technology is used in various fields, including biomedical and medical applications [19, 20]. The most important feature of machine learning algorithms is their unique opportunity to predict and to understand the environment from input data [17]]. Machine learning is classified into three types based on the nature of the data labelling: supervised, unsupervised, and semi-supervised. Supervised learning derives an unknown output and input, mapping from the known output and input samples, with the output labelled classification [21]. The study used nine supervised machine learning algorithms for analysis and comparison. The following are the algorithms:

Bayesian network

The Bayesian Network node allows you to create a probability model by combining observed and recorded evidence with real-world knowledge to determine the likelihood of events occurring. The node focuses on classification networks such as Tree Augmented Nave Bayes and Markov Blanket.


The C5.0 node constructs a decision tree or a rule set. The model divides the sample based on the field that provides the greatest information gain at each level. The target field must be categorical in nature.


The CHAID node generates decision trees using chi-square statistics to identify optimal splits. In contrast to the C&R Tree and QUEST nodes, CHAID can generate non binary trees, which means that some divisions have more than two branches. The target and input fields can be either continuous numeric ranges or categorical. Exhaustive CHAID is a CHAID modification that examines all possible splits more thoroughly but takes longer to compute.


The QUEST shows a binary classifier method for building decision trees, which is intended to reduce the processing time required for large classification and regression tree assessments while reducing the tendency found in classification tree methods to favour inputs with more splits. The input fields can be continuous numerical ranges, but the target field must be categorical.

C&R tree

Regression and classification tree nodes generate a decision tree that can be used to predict or categorise future observations. The method divides the training records into segments by reducing impurity on every step, with a node in the tree considered “pure” if all of the cases in the node drop into a particular category of the target field. All splits are binary, with only two subgroups; target and input fields can be numeric ranges or categorical as nominal, ordinal, or flags.

Decision list

The Decision Node List identifies groups or segments that show more or less a given binary value compared to the population. A list of decision models consists of a list of rules, where each rule has an event and a consequence. The rules are applied sequentially, with the first matching rule determining the outcome.


Logistic regression is a statistical technique used to classify data based on the value of an input field. It is similar to linear regression but uses a categorical rather than numerical purpose. Logistic regression is an example of a model commonly used in classification problems. The model includes a maximum likelihood estimation technique to estimate the default.

Neural network

A neural network is a machine learning algorithm consisting of a set of nodes, the most important of which are the input layer [22], the hidden layer, and the output layer. It has a set of input neurons (X) that represent the data extracted from each feature in the data. Backpropagation is a technique used in neural networks, where the nonpredictive neural network is iterated before the weights of each neural connection are changed until there is less error [23, 24].


Discriminant testing creates a predictive model for group membership. The model has a discriminating function or discriminating function for more than two groups that provides the best discrimination between groups based on a combination of different predictors. This study consists of a sample of experienced members; these functions can be used for new situations with parameter values ​​, but group membership is unknown.

Variable selection:

In the research, a feature selection method was employed to determine the significance of inputs in relation to a specific target. Out of all the predictor variables, the model incorporated a total of 12 variables.

Model comparison

When compared to traditional models, machine learning models outperform them in classification problems. Raw data is rarely found in the form and shape required for machine learning algorithms to perform optimally. Because the algorithms used in machine learning are only numerical values, it is necessary to convert categorical variables to numerical values. As a result, the most important aspect of machine learning model applications is the pre-processing step [25,26,27]. The dataset’s categorical features are encoded to convert them into numerical values, and the continuous data in this study were normalised. For machine learning approaches, the dataset is randomly divided into two parts: a training dataset that trains the model and a test dataset that predicts the response variable and checks whether the predicted outcome is similar to the actual outcomes, and the validation dataset is used to estimate the parameters that will be used in the training models [28,29,30].

Influence of different training and testing ratios on the performance of the given ML models checked. The study divide the dataset into training and testing datasets for the performance assessment of models. Popular statistical indicators are employed to evaluate the predictive capability of the models under different training and testing ratios. After using feature selection, the study considered 90%/10%, 80%/20%, 70%/30%, and 60%/40% train/test partitions. The partition procedure shows in the following figure (Fig. 3). Comparison criteria are lift and AUC.

Fig. 3
figure 3

the graphical display of the analysis procedure


The study involved 1520 individuals, with an average age of 24 years and an average weight of 42 kg. Their average initial systolic and diastolic blood pressure measurements were 107 and 66 respectively. The average oxygen saturation (SPO2) level was also found to be 95%. The average ejection fraction was 59%. On average, patients had to wait 481 days for their surgical procedures (Table 1).

Table 1 Descriptive analysis for continuous variables

Among 1520 patients included in the study, 122 individuals (8% of the total) were died. Among the patients who underwent cardiac surgery, there were 842 females, 692 patients from rural areas, valvular heart disease cases (874), congenital cases (593), and other cases (53). There were 145 patients with a family history of heart disease, patients with heart failure (20), smokers (4), patients with normal sinus rhythm (49), patients who had undergone multiple surgeries (25), and 177 patients with joint heart conditions.

Based on the chi-square association, there was a significant correlation found between the outcome status (Death or Alive) and predictor variables including heart disease type, family history, rheumatic disease status, heart failure status, and NYHA class (Table 2).

Table 2 The chi-square tests of association among categorical variables and death status among participants under investigation

logistic regression machine learning algorithm was preferable through AUC on 90/10 partition than others expressed from testing data. Finally, a logistic regression algorithm was implemented to analyze other than CHID, QUEST, Bayesian network, Neural network, C5, Decision list, Discriminant, and C&R machine learning algorithms. The model with an AUC value approach to one is preferable and reasonable to choose. As shown below, the AUC value of logistic regression under 90%/10% partition for cardiac patients was 0.961, greater than other algorithms (Table 3). In addition, based on the lift value of 3.33, it can be concluded that the logistic regression algorithm was performing well and providing substantial improvement over random selection. Lift value greater than 1(one) in general considered as a good fit. The higher lift value suggests that the model has a good ability to discriminate between the target and non-target outcomes, and it can be considered a crucial tool for making predictions and driving decision-making processes.

Table 3 area under the curve performance of different algorithms for cardiac patient’s surgery outcome

In addition to the global comparison method, the gain curve shows the curve approaches the best curve or the fastest to 100% is considered best (see Fig. 4A), and again logistics is a good fit for cardiac patients’ surgery outcomes since it is approved graphically too (Fig. 4(A-D)).

Fig. 4
figure 4

(A) 90/10 partition Gain graph. (B) 80/20 partition Gain graph. (C) 70/30 partition Gain graph. (D) 60/40 partition Gain graph

The importance of the logistic regression algorithm over other algorithms is in addition to predictor relative importance; it gives us a coefficient for interpretation. For cardiac surgery patients’ outcome prediction, from all predictors in the final model, the most important predictors were SPO2, creatinine value, age, waiting time for surgery, length of cardiac center stays after surgery, hemoglobin value, ejection fraction, and heart disease status. From logistic algorithm, SPO2 contributes approximately 21.23% to the model, age contributes 10.9%, time to surgery waiting time contributes approximately 18.12%, length of cardiac center stays after surgery values 11.14%, and creatinine value contributes 8.80%, heart disease type impacts 9.82%, ejection fraction 7.12% and hemoglobin value contributes approximately 6.52% (Fig. 5).

Fig. 5
figure 5

The relative predictor importance

From advanced modelling results, age, ejection fraction, SPO2, length of cardiac center leave day, waiting time, haemoglobin value, and creatinine value were significant predictors of death. When age increased by one year, the expected odds for the occurrence of death among cardiac patients was increased by 80.3%. For a single unit rise of SPO2, the expected odds for the occurrence of death among cardiac patients was decreased by 0.355 time as compare to the other groups, given other covariates constant. In another way, for a single unit rise of SPO2, the expected odds of occurrence of death were decreased by 64.5% given the other covariates constant. About ejection fraction, for a single unit increasement, the expected odds of occurrence of death were decreased by 27.5%. On the other hand, as patient’s waiting time after surgery increased by one day, the expected odds for occurrence of death was 2.3 times greater as compared with the previous day. For a single unit increasement of hemoglobin value, the expected odds of occurrence of death for a cardiac patient was decreased by 33%, given the other conditions constant. When creatinine value increased by a single unit, the expected odds of occurrence of death were also increased by 58% (Table 4).

Table 4 Multivariable binary logistic regression result


The objective of the research was to identify the factors that contribute to the likelihood of post-surgery death in cardiac patients. The study found several significant predictors of death outcome, including waiting time, hemoglobin value, creatinine value, ejection fraction, duration of cardiac center stays after surgery, saturated oxygen, and age.

The death prevalence due to cardiac disease among heart patients included in the study was 122(8%). The finding is consistent with other previously conducted researches [31,32,33,34].

Age is one of the significant predictors [33,34,35,36,37]. This alignment further supports the investigation’s results. As patients grow older, the likelihood of experiencing complications and comorbidities increases. Moreover, the effectiveness of treatments tends to decrease with advancing age. The physiological aging of the heart plays a significant role in the development and occurrence of cardiac issues in older individuals, primarily due to heightened inflammation and oxidative stress [38].

Creatinine value is crucial in predicting death [36, 39, 40]. Given its significance, the role of creatinine as a powerful predictor has been incorporated into various mortality risk assessment tools currently used for patients undergoing cardiac surgery [41,42,43]. Creatinine test is conducted to evaluate the functioning of the kidneys, as they are responsible for removing creatinine from the body. If the kidneys are not functioning properly, the level of creatinine in the blood will increase due to the kidneys’ reduced ability to clear creatinine effectively. This elevated creatinine level indicates damaged kidney function.

Saturated oxygen were the main predictors for predicting death [44]. Low saturated oxygen levels have been associated with an increased risk of adverse cardiac events and mortality in heart patients. Insufficient oxygen supply to the heart can exacerbate existing cardiac conditions, contribute to the development of complications, and potentially lead to fatal outcomes.

The lower ejection fraction was the responsible predictor of death [37, 39, 45,46,47]. A low ejection fraction indicates that the heart is unable to pump enough blood, which can ultimately lead to heart failure. In addition, it may lead to shortness of breath. This could potentially contribute to increase the chance of death outcome.

Hemoglobin result was related with worse outcome [34]. Anemia occurs when the hemoglobin levels in the blood are low. This condition can negatively impact cardiac function in multiple ways. Anemia cases stress on the heart by causing increased heart rate and it can result in reduced blood flow to the kidneys.

The average waiting time for surgery was 481 days. According to a statement from the Ethiopian Ministry of Health, the figure presented underestimates the true situation regarding cardiac surgery patients and their waiting times. The average waiting time mentioned in the Ministry statement is seven years. However, the study results indicate that this estimation is lower than the actual waiting times. This is because the study only took into account the waiting time after referral for cardiac surgery admission. In reality, patients may have to wait for a considerable amount of time for free surgery services, and if they become too ill, they may choose to opt for private surgery service, which is provided immediately. The urgency level might also affect the figure of waiting time. These factors undermine the accuracy of the waiting time results. The duration of waiting time to get surgery was found to be highly correlated with and served as the most significant predictor of mortality among cardiac patients after surgery [48]. However, there are some papers that have suggested that the outcome of the surgery is not influenced by the waiting time [49, 50]. The difference might arise due to getting the availability of cardiac surgery service in a short period as compared to Ethiopia. It is important to note that Ethiopia has a limited number of cardiac centers, which serving a population of more than 126 million. This disparity in healthcare infrastructure could be a contributing factor.

Conclusions and recommendations

To sum up, logistic regression machine learning was preferable to other algorithms. Ejection fraction, age, saturated oxygen, duration of cardiac center stays after surgery, waiting time, hemoglobin value, and creatinine value were responsible predictors of death outcome.

According to worldometr figure, Ethiopia has about more than 126 million people, making it the 11th most populated country globally and the 2nd most populous in Africa [51], a lot of health investment in the field of cardiac is a future task. Serving this total population in a very resource limited centers might be also unrealistic.

For ministry of health, Health Bureaus, health institutes, national and local Governments, Universities, local and international organizations, NGO’s, lack of cardiac surgery service in Ethiopia is a serious concern as such, to reduce the waiting time, the study suggest to increase health facilities like cardiac centers, hospitals or institutes which serve as cardiac surgery in Ethiopia for at least 10 regions in a short period. So, the allocation of required budget and supplies through collaboration with other partners or stakeholders are timely. Those stakeholders are also required to design or formulate sustainable income generating activities of those centers. In addition, universities are expected to graduate the required number of cardiac specialists too. For health professionals and health facilities, the study recommended to give special attention for cardiac patients with lower saturated oxygen, higher age, patients with elongated waiting time to surgery, lower ejection fraction value, lower hemoglobin value and higher creatinine value. The study also recommends Ministry of Health and Federal Government to established and realized “Cardiovascular Institute”.

The findings of this study will be helpful for Ministry of Health, clinicians, and policymakers to develop policies to reduce the burden of cardiac disease in the country. Therefore, this study provided current situation of cardiac patients’ burden after surgery.

Limitation of current study:

This study limitations is that there was limited information about the current burden status of cardiac patients in each cardiac center in Ethiopia. Hence, the current study recommends future researches to conduct research including important information related to the burden of cardiac patients after surgery.

Data availability

The data used in the current study is available under the corresponding author and can be attached on request.



New York Health Association


Oxygen Saturated


International Normalised Ratio


Area Under the Curve


Chi square Automatic Interaction Detection


Quaternion Estimator


Classification and Regression


Machine Learning


  1. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, Abraham J, Adair T, Aggarwal R. Ahn SYJTl: global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the global burden of Disease Study 2010. 2012, 380(9859):2095–128.

  2. Yin M, Augustin B, Fu Z, Yan M, Fu A, Yin, PJSr. Geographic distributions in hypertension diagnosis, measurement, prevalence, awareness, treatment and control rates among middle-aged and older adults in China. 2016, 6(1):1–11.

  3. McAloon CJ, Boylan LM, Hamborg T, Stallard N, Osman F, Lim PB. Hayat SAJIjoc: The changing face of cardiovascular disease 2000–2012: An analysis of the world health organisation global health estimates data. 2016, 224:256–264.

  4. Yeates K, Lohfeld L, Sleeth J, Morales F, Rajkotia Y, Ogedegbe O. A global perspective on cardiovascular disease in vulnerable populations. Can J Cardiol. 2015;31(9):1081–93.

    Article  PubMed  Google Scholar 

  5. Remenyi B, Carapetis J, Wyber R, Taubert K, Mayosi BM. Position statement of the World Heart Federation on the prevention and control of rheumatic heart disease. Nat Reviews Cardiol. 2013;10(5):284.

    Article  CAS  Google Scholar 

  6. Lawrence JG, Carapetis JR, Griffiths K, Edwards K, Condon JR. Acute rheumatic fever and rheumatic heart disease: incidence and progression in the Northern Territory of Australia, 1997 to 2010. Circulation. 2013;128(5):492–501.

    Article  PubMed  Google Scholar 

  7. Haddish HT, Hailu T, Berhane S, Hailu A, Luel A. Meles HGJEAJoHS: patterns and Factors Associated with Cardiovascular disorders at Ayder Comprehensive Specialized Hospital, Tigrai. North Ethiopia. 2019;1(2):149–59.

    Google Scholar 

  8. Andell P, Li X, Martinsson A, Andersson C, Stagmo M, Zöller B, Sundquist K, Smith JGJH. Epidemiology of valvular heart disease in a Swedish nationwide hospital-based register study. 2017, 103(21):1696–703.

  9. Coffey S, Roberts-Thomson R, Brown A, Carapetis J, Chen M, Enriquez-Sarano M, Zühlke L, Prendergast BDJNRC. Global epidemiology of valvular heart disease. 2021:1–12.

  10. Abdissa SG, Oli K, Feleke Y, Goshu DY, Begna DM, Tafese AJE. Spectrum of cardiovascular diseases among Ethiopian patients at tikur anbessa specialized university teaching hospital. Addis ababa. 2014;52(1):9–17.

    Google Scholar 

  11. Defo BKJGha. Demographic, epidemiological, and health transitions: are they relevant to population health patterns in Africa? 2014, 7(1):22443.

  12. Kothari, SSJAopc. Of history, half-truths, and rheumatic fever. 2013, 6(2):117.

  13. Mocumbi AOJCA. Rheumatic heart disease in Africa: is there a role for genetic studies? 2015, 26(2 H3Africa Suppl):S21.

  14. d’Arcy J, Prendergast B, Chambers JB, Ray S, Bridgewater B. Valvular heart disease: the next cardiac epidemic. In.: BMJ Publishing Group Ltd; 2011.

  15. Adem A, Demis T, Feleke Y. Trend of diabetic admissions in Tikur Anbessa and St. Paul’s University Teaching Hospitals from January 2005-December 2009, Addis Ababa, Ethiopia. Ethiop Med J. 2011;49(3):231–8.

    PubMed  Google Scholar 

  16. Abera E. Analysis: with 7000 patients on waiting list Ethiopia’s Cardiac Center fights to stay open. Addis Standard. Addis Ababa: Addis Standard; 2019.

    Google Scholar 

  17. Alpaydin E. Introduction to machine learning, 3rd editio. ed. In.: The MIT Press; 2014.

  18. Bishop CM, Nasrabadi NM. Pattern recognition and machine learning. Volume 4. Springer; 2006.

  19. Cleophas TJ, Zwinderman AH, Cleophas-Allers HI. Machine learning in medicine. Volume 9. Springer; 2013.

  20. Malley JD, Malley KG, Pajevic S. Statistical learning for biomedical data. Cambridge University Press; 2011.

  21. El Naqa I, Murphy MJ. What is machine learning? Springer; 2015.

  22. Liang N-Y, Huang G-B, Saratchandran P, Sundararajan N. A fast and accurate online sequential learning algorithm for feedforward networks. IEEE Trans Neural Networks. 2006;17(6):1411–23.

    Article  PubMed  Google Scholar 

  23. Hecht-Nielsen R. Theory of the backpropagation neural network. Neural networks for perception. edn.: Elsevier; 1992. pp. 65–93.

  24. Abdelhafiz D, Yang C, Ammar R, Nabavi S. Deep convolutional neural networks for mammography: advances, challenges and applications. BMC Bioinformatics. 2019;20:1–20.

    Article  Google Scholar 

  25. Deo RC. Machine learning in medicine. Circulation. 2015;132(20):1920–30.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Kotsiantis SB, Zaharakis I, Pintelas P. Supervised machine learning: a review of classification techniques. Emerg Artif Intell Appl Comput Eng. 2007;160(1):3–24.

    Google Scholar 

  27. Raschka S. Python machine learning: Packt publishing ltd; 2015.

  28. Quinlan JR. Induction of decision trees. Mach Learn. 1986;1:81–106.

    Article  Google Scholar 

  29. Gareth J, Daniela W, Trevor H, Robert T. An introduction to statistical learning: with applications in R: Spinger; 2013.

  30. Molina M, Garip F. Machine learning for sociology. Ann Rev Sociol. 2019;45:27–45.

    Article  Google Scholar 

  31. Jamieson WE, Germann E, Ye J, Chan F, Cheung A, MacNab JS, Fradet GJ, Stanford EA, Bryson LA. Lichtenstein SVJTAots: Effect of prosthesis-patient mismatch on long-term survival with mitral valve replacement: assessment to 15 years. 2009, 87(4):1135–42.

  32. Appelbaum A, Kouchoukos NT, Blackstone EH, Kirklin JWJTAJC. Early risks of open heart surgery for mitral valve disease. 1976, 37(2):201–9.

  33. Murana G, Alfonsi J, Savini C, Mariani C, Coppola G, Lo Coco V, Pilato E, Pacini D, Di Bartolomeo RJIC. On-X mitral valve replacement: a single-centre experience in 318 patients. Surg T. 2018;27(6):836–41.

    Google Scholar 

  34. Moreira JL, Barletta PHAAS, Baucia JAJBJCS. Morbidity and mortality in patients undergoing mitral valve replacement at a cardiovascular surgery referral service: a retrospective analysis. 2021, 36:183–91.

  35. Wang B, Xu Z-y, Han L, Zhang G-x, Lu F-l. Song Z-gJEJoC-TS: impact of preoperative atrial fibrillation on mortality and cardiovascular outcomes of mechanical mitral valve replacement for rheumatic mitral valve disease. 2013, 43(3):513–9.

  36. Roques F, Nashef S, Michel, PJTJohvd. Risk factors for early mortality after valve surgery in Europe in the 1990s: lessons from the EuroSCORE pilot program. 2001, 10(5):572–7; discussion 577.

  37. Remenyi B, ElGuindy A, Smith SC, Yacoub M, Holmes DRJTL. Valvular aspects of rheumatic heart disease. 2016, 387(10025):1335–46.

  38. Curtis AB, Karki R, Hattoum A, Sharma UC. Arrhythmias in patients ≥ 80 years of age: pathophysiology, management, and outcomes. J Am Coll Cardiol. 2018;71(18):2041–57.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Pieri M, Belletti A, Monaco F, Pisano A, Musu M, Dalessandro V, Monti G, Finco G, Zangrillo A, Landoni G. Outcome of cardiac surgery in patients with low preoperative ejection fraction. BMC Anesthesiol. 2016;16(1):1–10.

    Article  Google Scholar 

  40. Khan MF, Khan MS, Bawany FI, Dar MI, Hussain M, Farhan S, Fatima K, Hamid K, Arshad MH. Aziz MJGJoHS: predictors of mortality in patients undergoing mitral valve replacement. 2016, 8(3):37.

  41. Nilsson J, Algotsson L, Höglund P, Lührs C, Brandt J. Comparison of 19 pre-operative risk stratification models in open-heart surgery. Eur Heart J. 2006;27(7):867–74.

    Article  PubMed  Google Scholar 

  42. Geissler HJ, Hölzl P, Marohl S, Kuhn-Régnier F, Mehlhorn U, SuÈdkamp M, de Vivie ER. Risk stratification in heart surgery: comparison of six score systems. Eur J Cardiothorac Surg. 2000;17(4):400–6.

    Article  CAS  PubMed  Google Scholar 

  43. Badreldin AM, Doerr F, Kroener A, Wahlers T, Hekmat K. Preoperative risk stratification models fail to predict hospital cost of cardiac surgery patients. J Cardiothorac Surg. 2013;8:1–7.

    Article  Google Scholar 

  44. Zhong Z, Yuan X, Liu S, Yang Y, Liu F. Machine learning prediction models for prognosis of critically ill patients after open-heart surgery. Sci Rep. 2021;11(1):1–10.

    Article  CAS  Google Scholar 

  45. Lam B-K, Chan V, Hendry P, Ruel M, Masters R, Bedard P, Goldstein B, Rubens F. Mesana TJTJot, surgery c: the impact of patient–prosthesis mismatch on late outcomes after mitral valve replacement. 2007, 133(6):1464–73. e1463.

  46. Rankin JS, Hammill BG, Ferguson TB Jr, Glower DD, O’Brien SM, DeLong ER, Peterson ED. Edwards FHJTJot, surgery c: determinants of operative mortality in valvular heart surgery. 2006, 131(3):547–57.

  47. Rahimtoola SHJJotACoC. Choice of prosthetic heart valve in adults: an update. 2010, 55(22):2413–26.

  48. Tadege M, Tegegne AS, Dessie ZG. Post-surgery survival and associated factors for cardiac patients in Ethiopia: applications of machine learning, semi-parametric and parametric modelling. BMC Med Inf Decis Mak. 2024;24(1):91.

    Article  Google Scholar 

  49. Badakhshan A, Arab M, Gholipour M, Behnampour N, Saleki S. Heart surgery waiting time: assessing the effectiveness of an action. Iran Red Crescent Med J 2015, 17(8).

  50. Ray AA, Buth KJ, Sullivan JA, Johnstone DE, Hirsch GM. Waiting for cardiac surgery: results of a risk-stratified queuing process. Circulation. 2001;104(suppl1):I–92.

    Google Scholar 

  51. World population by country. [].

Download references


We need to forward our acknowledgement for all staff at the two Cardiac centers. Data collectors shall also owe our heartfelt thanks for their cooperation for the success of this research work.


There is no agent/institution that funded this research.

Author information

Authors and Affiliations



All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Corresponding author

Correspondence to Melaku Tadege.

Ethics declarations

Ethics approval and consent to participate

Ethical approval to undertake the studies was obtained from the Bahir Dar University science college ethical committee with reference number (Ref.No PRCSVD/289/2014) on 16 February 2022, and permission from the health institutions was taken.Informed consent was waived by Bahir Dar University science college ethical committee due to secondary data is collected from patients’ chart as such patients’ data is often anonymized before being accessed for research purposes.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

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 The Creative Commons Public Domain Dedication waiver ( 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

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Tadege, M., Tegegne, A.S. & Dessie, Z.G. Cardiac patients’ surgery outcome and associated factors in Ethiopia: application of machine learning. BMC Pediatr 24, 395 (2024).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: