Predicting organ donation outcome using network-based machine learning algorithms

Abstract

Thousands of people die every year waiting for organs. Therefore, it is now more important than ever before to study the factors associated with organ donation, which will potentially help in making data-driven strategies to increase the consent for deceased organ donation. This paper uses machine learning algorithms to accurately predict organ donation outcome (consent: yes/no from family). In this study, 6 years patients' data from an OPO located in New York city has been used to build the consent prediction model. The predictions models are compared and the best models can be used by Organ Procurement Organizations (OPO) to come up with strategies to optimize the consent rate thereby saving more lives. The experimental results show that our approach outperforms for detection because we combined networks (see in Figure 1) and machine learning algorithms. The proposed approach can be used as a decision support system to help increase the consent rate for organ donation. Figure 1. Donor Networks based on relations. According to the comparison results in the research, the proposed detection framework outperforms traditional machine learning methods (Logistic Regression, Decision Tree, Support Vector Machine and Naive Bayes) combined with network metrics used for organ donation outcome in most of the evaluation criteria. For the other evaluation metrics (e.g., accuracy, sensitivity and specificity), the framework performs mostly better than the other methods. Hence, by including the interaction between different event activities we can make a more precise decision for organ donation outcome. In conclusion, detecting the organ donation outcome (consent: yes/no from family) is critical for effective healthcare performance by combining networks and predictive models. Therefore, it is very significant to focus on the organ donation features. The proposed predictive model for family consent for organ donation has been built to predict if the family will give consent provided all the related factors. Therefore, the model can be used as a decision support system to help increase the consent rate for organ donation thereby abridging the gap between organ demand and supply [1]. The predictions models are compared and the best models can be used by Organ Procurement Organizations (OPO) to come up with strategies to optimize the consent rate thereby saving more lives. For future studies, we will combine network science and machine learning together to build the even stronger model to predict organ donation outcome from family. Based on the networks, we will capture interactions for better donor prediction. 1. Chon, W. J., et al. When the living and the deceased cannot agree on organ donation: a survey of US organ procurement organizations (OPOs).

Full text

Predicting organ donation outcome using network-based machine learning algorithms
MD Ehsan Khan1*, Salih Tutun1
1Systems Science and Industrial Engineering, Binghamton University, Binghamton, NY, 13850, USA
*stutun1@binghamton.edu
Thousands of people die every year waiting for organs. Therefore, it is now more important than ever before to
study the factors associated with organ donation, which will potentially help in making data-driven strategies to
increase the consent for deceased organ donation. This paper uses machine learning algorithms to accurately predict
organ donation outcome (consent: yes/no from family). In this study, 6 years patients' data from an OPO located in
New York city has been used to build the consent prediction model. The predictions models are compared and the
best models can be used by Organ Procurement Organizations (OPO) to come up with strategies to optimize the
consent rate thereby saving more lives. The experimental results show that our approach outperforms for detection
because we combined networks (see in Figure 1) and machine learning algorithms. The proposed approach can be
used as a decision support system to help increase the consent rate for organ donation.
Figure 1. Donor Networks based on relations.
According to the comparison results in the research, the proposed detection framework outperforms traditional
machine learning methods (Logistic Regression, Decision Tree, Support Vector Machine and Naive Bayes)
combined with network metrics used for organ donation outcome in most of the evaluation criteria. For the other
evaluation metrics (e.g., accuracy, sensitivity and specificity), the framework performs mostly better than the other
methods. Hence, by including the interaction between different event activities we can make a more precise decision
for organ donation outcome.
In conclusion, detecting the organ donation outcome (consent: yes/no from family) is critical for effective healthcare
performance by combining networks and predictive models. Therefore, it is very significant to focus on the organ
donation features. The proposed predictive model for family consent for organ donation has been built to predict if
the family will give consent provided all the related factors. Therefore, the model can be used as a decision support
system to help increase the consent rate for organ donation thereby abridging the gap between organ demand and
supply [1]. The predictions models are compared and the best models can be used by Organ Procurement
Organizations (OPO) to come up with strategies to optimize the consent rate thereby saving more lives. For future
studies, we will combine network science and machine learning together to build the even stronger model to predict
organ donation outcome from family. Based on the networks, we will capture interactions for better donor
prediction.
1. Chon, W. J., et al. When the living and the deceased cannot agree on organ donation: a survey of US
organ procurement organizations (OPOs). 2014 Vol: 14 American Journal of Transplantation 172-177.