SlideShare a Scribd company logo
DOI: 10.5121/ijci.2019.8101 1
HYBRID MORTALITY PREDICTION USING
MULTIPLE SOURCE SYSTEMS
Isaac Mativo1
, Yelena Yesha 1
, Michael Grasso2
, Tim Oates 1
, Qian Zhu 3
1
Department of Computer Science, University of Maryland Baltimore County (UMBC)
2
School of Medicine, University of Maryland Baltimore (UMB)
3
Department of Information Systems, University of Maryland Baltimore County
(UMBC)
ABSTRACT
The use of artificial intelligence in clinical care to improve decision support systems is increasing. This is
not surprising since by its very nature, the practice of medicine consists of making decisions based on
observations from different systems both inside and outside the human body. In this paper, we combine
three general systems (ICU, diabetes, and comorbidities) and use them to make patient clinical predictions.
We use an artificial intelligence approach to show that we can improve mortality prediction of hospitalized
diabetic patients. We do this by utilizing a machine learning approach to select clinical input features that
are more likely to predict mortality. We then use these features to create a hybrid mortality prediction
model and compare our results to non artificial intelligence models. For simplicity, we limit our input
features to patient comorbidities and features derived from a well-known mortality measure, the Sequential
Organ Failure Assessment (SOFA).
KEYWORDS
Decision Support Systems; Artificial Intelligence; Hybrid Systems.
I. INTRODUCTION
Clinical decision support systems are driven by large amounts of data sourced from various
disparate systems. These may include data from lab instruments, wearable devices, prior hospital
records, patient reported data, etc. Often, the clinician is exposed to a lot of this data and is
expected to make decisions to improve clinical outcomes. Although decision support systems are
effective in logical tasks such as alerting the user when some parameters are outside defined
ranges, they are not designed to learn from the data they have. As such, they are limited to what
they have been programmed to do. This limitation becomes an opportunity to experiment with
artificial intelligence within the clinical setting [1]. Specifically, a hybrid system that combines
attributes from both traditional clinical decision support systems and attributes derived from
machine learning tools to create a prediction model becomes attractive.Hybrid systems have been
used in many fields to improve performance in areas such as managing humanitarian relief chains
[2] and multiple-criteria decision-making [3] . In healthcare, hybrid decision support systems
have been used to help clinicians make more informed decisions [4] [5]. For mortality prediction,
several hybrid approaches have been proposed such as the use of hybrid image features on neural
network for cancer patients [3][6].In this paper, we take a systems theory view of a clinical
setting whereby many different systems come together to inform the clinician on the best
decisions to make. In the next section, we will describe what we consider to be the three systems
that we consider and apply machine learning tools on.
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
2
2. THE THREE SYSTEMS
A. Diabetes
For this paper, we consider the chronic disease diabetes as one of our three systems. This is
because diabetes, being a disorder of metabolism, affects multiple systems within the human
body [7]. When blood sugar is not well controlled and therefore outside the normal range for
prolonged periods of time, the proper functioning of various organ systems is compromised.
Consequently, this manifests in various body conditions such as kidney disease, neuropathy, liver
disease, etc. [8] .From a cybernetics perspective, diabetes is interesting because it involves the
delicate control of blood sugar by both systems inside and outside of the body. Diabetic patients
often rely on external interventions such as medication ingestion, insulin injection, dietary
controls, and lifestyle changes such as exercise. When the sugar level in blood is found to be
outside the normal range based on information gathered, various actions are taken to bring it
under control.
B. Intensive Care Unit
We consider the Intensive Care Unit (ICU) in a hospital to be our next system. This is because the
ICU is a specialized facility with many controlled and interconnected systems that work together
to intervene in critically ill patients. In the ICU, information is availed through various channels
such as electronic medical records, clinical notes, instrument measurements, etc. [8] [9] All this
information is then used to provide a level of clarity in making clinical decisions. Various aspects
of this information such as temporal occurrence (e.g. did the blood pressure fall after or before
takings a specific medication?) dependency (e.g. eat solid foods before taking a certain
medication), are important. As such, in an ICU facility, the proper control of information from
various systems is critical [10] . In this paper therefore, we view the ICU facility as one
large system with many subsystems within it.
C. Comorbidities
Comorbidities are co-occurring disease conditions within a person. Often, diabetic patients have
other disease conditions as well. Some of these conditions are more commonly associated with
diabetes than others [11]. In this paper, we think of comorbidities as a system because they are
well-defined conditions within the human body that individually and collectively influence the
wellbeing of the entire body. Since comorbidities are an important piece of information
considered in patient treatment, we considered comorbidities to be relevant in informing our
hybrid prediction model.
With these three systems, we build a hybrid mortality prediction model and evaluated its
performance. It is a hybrid model because it combined both machine learning techniques and
non-machine learning techniques to select it features and perform its predictions. In the next
section, we discuss the materials we used to do our experiment.
3. MATERIALS
A. Medical Information Mart for Intensive Care (MIMIC) III
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
3
The data source we used to get our patient population is MIMIC-III [12] . MIMIC-III is a large
and single-center database comprising de identified health related information containing 46,520
patients of whom 20,399 are female and 26,121 are male. These patients stayed in critical care
units of the Beth Israel Deaconess Medical Center between 2001 and 2012. The database includes
information such as demographics, vital sign measurements, laboratory test results, procedures,
medications, caregiver notes, imaging reports, and mortality. In this study, we accessed MIMIC-
III to extract diabetes patients for mortality prediction.
B. Elixhauser comorbidity measure.
In order to get the comorbidity information for our diabetic patients, we used The Elixhauser
comorbidity index [13] . This index consists of 30 comorbidity measures that have been shown to
have an impact on patient hospital length of stay, cost, and mortality [14] . Classification in to this
index is based on conditions described by the ICD-9-CM (International Classification of
Diseases, Ninth Edition, Clinical Modifications) discharge records. The Elixhauser index assigns
a value of 1 for each of 30 predefined conditions, and computes the sum. Therefore, a score of 10
indicates a patient has 10 conditions present.
C. Mortality Measure
To compare our hybrid model prediction accuracy with a needed a tool that measured mortality.
For this, we chose the Sequential Organ Failure Assessment (SOFA) [15]. SOFA is developed
from a large sample of ICU patients from around the world and gives a score based on the
severity of on a patient. The SOFA score is computed based on six variables, with each variable
corresponding to an organ system. The organ systems included in SOFA are respiratory,
cardiovascular, hepatic, coagulation, renal and neurological systems. Each organ system is
assigned a value between 0 (good) and 4 (severe). The SOFA score is the sum of all the
individual organ system values.
4. METHODS
In this section, we describe the methods we used to retrieve the patient cohort, computer the
mortality and comorbidity measure, and to predict mortality. For this study, ethical approval was
not needed because the data used is de-identified t o conform with the Health Insurance
Portability and Accountability Act (HIPAA). Additionally, the data is publicly available to
researchers.
A. Patient Retrieval
From our MIMIC III PostgreSQL database, we retrieved the diabetes patient cohort from MIMIC
III database based on ICD-9 codes for diabetes. To avoid restricting diabetic patients with
comorbidities for further mortality prediction, we extracted the entire diabetes diagnoses for each
of the diabetic patients.
B. Mortality Measure Computation
We used the SOFA score as a mortality measure. To calculate the SOFA score for each patient,
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
4
we extracted the patient in the PostgreSQL database for calculating the the SOFA score
corresponding to the organ areas as shown in table 1. We extracted this data using query scripts in
PostgreSQL from the MIMIC III diabetic patients data based on the first day of each patient's’
ICU stay. For each patient, we used an SQL script to calculate the actual points for each organ
system, then added up the points for each organ system to obtain the SOFA score. For missing
values, we assigned zero points to that organ system. We did this for all collected diabetic
patients.
Table 1: Sofa Organ Areas and Points Thresholds
Organ/Points 1 2 3 4
Respiratory
PaO 2 /FiO 2
(mmHg)
<400 <300 <200 and
respiratory
support
<100 and
respiratory
support
Cardiological
Mean arterial
pressure
MAP < 70
mm/Hg
dop <=
5 or
dob
dop > 5
OR epi <=
0.1 OR nor
<= 0.1
dop > 15 OR epi
> 0.1 OR nor >
0.1
Hepatic
Bilirubin (mg/dl)
[μmol/L]
1.2–1.9
[> 20-32]
2.0–5.9
[33-101]
6.0–11.9
[102-204]
> 12.0
[> 204]
Neurological
Glasgow coma
scale
13 - 14 10 - 12 6 - 9 <6
Coagulation
Platelets×103/μl
<150 <100 <50 <20
C. Comorbidity Measure and Weight Computation
We used the Elixhauser index to calculate the comorbidity measure for each diabetic patient in
MIMIC-III. For each diabetic patient, we extracted 29 different diagnosis excluding diabetes and
quantified them by using the Elixhauser index for comorbidities represented in the diabetic
population. In order to improve the efficiency of our classifier by reducing the number of
features, we identified the top five comorbidities that are most predictive of mortality across the
identified diabetic patients. To do this, we experimented using 5 different feature selection
algorithms, which are Gain Ratio [16] , Correlation [17] , Symmetrical Uncertainty [18] ,
Information Gain [19] , and Correlation Feature Selection (CFS) Subset Evaluator [20] . For each
algorithm run, we ranked the comorbidities based on the resulted predictive mortality. We then
averaged these results over the 5 algorithms to get our final top 5 most predictive comorbidities.
The final list of the comorbidities was cardiac arrhythmias, coagulopathy, metastatic cancer,
congestive heart failure, and fluid electrolyte.
D. Mortality Prediction
For mortality prediction, we created a hybrid mortality prediction model that took as input the 6
SOFA scores representing the 6 organ systems and the top 6 ranked comorbidities, for a total of
11 features. We then experimented with a Naive Bayes classifier and a Random Forest classifier
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
5
with 10-fold cross validation across our patient population. We compared the prediction accuracy
of our hybrid system with the accuracy of (i) a prediction model with only the SOFA scores and
(ii) a prediction model with only the comorbidity information. Our accuracy comparison was
based on the AUC on both the Naive Bayes and Random Forest classifiers on mortality.
5. RESULTS
A. Patient Retrieval
A total of 10,403 of diabetes patients were retrieved from 46,520 patients contained in MIMIC III
database. Of all the identified diabetic patients, 1,513 died while in ICU. Among the 10,403
diabetes patients, we identified 29 different comorbidities. B. Mortality Measure Computation
Table 2 shows the calculated SOFA scores and the corresponding mortality rates as computed
from our data. Our mortality rates were consistent with the expected SOFA mortality rates.
Table 2: Calculated Mortality Rates
C. Comorbidity Measure and Weights
Figure 1 shows the results of calculated comorbidity for the diabetic patients using the
Elixhauser index. As can be seen, most of the patients had comorbidities.
Figure 1: Comorbidities Across Patients
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
6
Table 3 shows results generated via the five feature extraction algorithms to determine what
comorbidities to be applied for prediction classifiers. These are the comorbidities that were used
as features in our model.
Table 3: Highly Weighed Comorbidities
D. Mortality Prediction
Table 4 shows the results generated by the Naive Bayes and Random Forest classifiers. The ROC
is lowest when using only the 5 comorbidity features are used, is better when using the 6 SOFA
scores are used, and best when both the 5 comorbidities and 6 SOFA scores are combined. Thus,
the hybrid model outperforms the natively SOFA or comorbidity only models.
Table 4: Mortality Prediction Results
6. DISCUSSION
In this study, we proposed viewing ICU clinical care through the lens of cybernetics. We
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
7
identified three systems; the ICU, diabetes, and patient comorbidity, as related and defined
systems that are important in supporting clinical decisions. We examined the mortality impact of
comorbidities on diabetic patients patients admitted in an ICU unit using machine learning
tools.We created a hybrid mortality prediction model and compare its results to non-hybrid
models.The results we obtained support our claim that mortality prediction can be improved by
incorporation key features from different systems (comorbidities in our case) and using artificial
intelligence to train our classifiers.
Another key observation from our results is that we can still get improved performance from our
prediction model without using all of the available input features. In our case, we used statistical
techniques to rank all our comorbidities based on their weight in impacting mortality. This
observation is important because within systems, we can optimize performance by carefully
selecting the components that affect our output. This careful selection results in dimension
reduction and consequently reduces computational resources needed to run algorithms while at
the same time producing improved results.
A logical extension of this research is further defining the subsystems within our system, or yet
still identifying other systems that are important in improving clinical decision support
systems.An improvement of our approach would be to take a closer look at the information flow
within and across our systems and evaluate its impact on the overall efficiency of an ICU
setting.Important makers, in this case, could include items such as readmission rates, adverse
events, hospital length of stay, central line infection, etc. We hope that with this study we have
demonstrated that it is possible to view clinical decision support systems from cybernetics
perspective and encouraged other researchers to consider this approach.
DISCLOSURE OF INTERESTS
The authors report no conflict of interest.
REFERENCES
1] R. Benbenishty and R. Treistman, ―The development and evaluation of a hybrid decision support
system for clinical decision making: The case of discharge from the military,‖ Soc.Work Res. , vol.
22, no. 4, pp. 195–204, Dec. 1998.
[2] N. Sahebjamnia, S. A. Torabi, and S. A. Mansouri, ―A hybrid decision support system for managing
humanitarian relief chains,‖ Decis. Support Syst. , vol. 95, pp. 12–26, Mar. 2017.
[3] E. Kazimieras Zavadskas et al. , ―HYBRID MULTIPLE CRITERIA DECISION MAKING
METHODS: A REVIEW OF APPLICATIONS IN ENGINEERING,‖ Scientia Iranica , vol.23, no. 1,
pp. 1–20, 2016.
[4] A. Mansoul, B. Atmani, and S. Benbelkacem, ―A hybrid decision support system :application on
healthcare,‖ arXiv [cs.AI] , 16-Nov-2013.
[5] O. W. Samuel, G. M. Asogbon, A. K. Sangaiah, P. Fang, and G. Li, ―An integrated decision support
system based on ANN and Fuzzy_AHP for heart failure risk prediction,‖ Expert Syst. Appl. , vol. 68,
pp. 163–172, 2017.
International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019
8
[6] P. Yan, H. Guo, G. Wang, R. De Man, and M. K. Kalra, ―Hybrid deep neural networks for all-cause
Mortality Prediction from LDCT Images,‖ arXiv [cs.CV] , 19-Oct-2018.
[7] F. Gomez, Blood Sugar Levels: Clear and Unbiased Facts About Diabetes . Lulu Press, Inc,2016.
[8] E. Bonora and R. A. DeFronzo, Diabetes Complications, Comorbidities and Related Disorders .
Springer, 2018.
[9] W. Koller, ―Flow of Information in the ICU — An Overview,‖ in Anaesthesia, Pain,Intensive Care
and Emergency Medicine — A.P.I.C.E. , 1996, pp. 883–891.
[10] S. L. Groves, ―Tools for ICU patient safety and quality improvement plans. Critical care safety:
Essentials for ICU patient care and technology. ECRI Institute Staff. Publisher:ECRI Institute; 146
pages, plus CD,‖ J. Healthc. Risk Manag. , vol. 27, no. 3, pp. 11–11,2007.
[11] P. Klimek, A. Kautzky-Willer, A. Chmiel, I. Schiller-Frühwirth, and S. Thurner,―Quantification of
diabetes comorbidity risks across life using nation-wide big claims data,‖PLoS Comput. Biol. , vol.
11, no. 4, p. e1004125, Apr. 2015.
[12] A. E. W. Johnson et al. , ―MIMIC-III, a freely accessible critical care database,‖ Sci Data ,vol. 3, p.
160035, May 2016.
[13] A. Elixhauser, C. Steiner, D. R. Harris, and R. M. Coffey, ―Comorbidity measures for use with
administrative data,‖ Med. Care , vol. 36, no. 1, pp. 8–27, Jan. 1998.
[14] M. E. Menendez, V. Neuhaus, C. N. van Dijk, and D. Ring, ―The Elixhauser comorbidity method
outperforms the Charlson index in predicting inpatient death after orthopaedic surgery,‖ Clin. Orthop.
Relat. Res. , vol. 472, no. 9, pp. 2878–2886, Sep. 2014.
[15] A. E. Jones, S. Trzeciak, and J. A. Kline, ―The Sequential Organ Failure Assessment score for
predicting outcome in patients with severe sepsis and evidence of hypoperfusion at the time of
emergency department presentation,‖ Crit. Care Med. , vol. 37, no. 5, pp.1649–1654, May 2009.
[16] A. G. Karegowda, A. S. Manjunath, and M. A. Jayaram, ―Comparative study of attribute selection
using gain ratio and correlation based feature selection,‖ International Journal of Information
Technology and Knowledge Management , vol. 2, no. 2, pp. 271–277, 2010.
[17] M. A. Hall and L. A. Smith, ―Feature Selection For Machine Learning: Comparing a Correlation-
based Filter Approach to the Wrapper,‖ 1999. [Online]. Available:
http://www.aaai.org/Library/FLAIRS/1999/flairs99-042.php . [Accessed: 17-Nov-2016].
[18] L. Yu and H. Liu, ―Feature selection for high-dimensional data: A fast correlation-based filter
solution,‖ in ICML , 2003, vol. 3, pp. 856–863.
[19] P. Jia, J.-H. Dai, Y.-H. Pan, and M.-L. Zhu, ―Novel algorithm for attribute reduction based on mutual-
information gain ratio,‖ JOURNAL-ZHEJIANG UNIVERSITY ENGINEERING SCIENCE , vol. 40,
no. 6, p. 1041, 2006.
[20] M. A. Hall and L. A. Smith, ―Practical feature subset selection for machine learning,‖ 1998

More Related Content

What's hot

Computer aided diagnosis for liver cancer using statistical model
Computer aided diagnosis for liver cancer using statistical modelComputer aided diagnosis for liver cancer using statistical model
Computer aided diagnosis for liver cancer using statistical model
eSAT Journals
 
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUESBLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
hiij
 
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
IRJET Journal
 
Automated health alerts using in home sensor data for embedded health assessment
Automated health alerts using in home sensor data for embedded health assessmentAutomated health alerts using in home sensor data for embedded health assessment
Automated health alerts using in home sensor data for embedded health assessment
I3E Technologies
 
How EMR can connect You and Your Doctor?
How EMR can connect You and Your Doctor?How EMR can connect You and Your Doctor?
How EMR can connect You and Your Doctor?
75health .com
 
Estimating Technical Efficiency of Turkish Hospitals
Estimating Technical Efficiency of Turkish HospitalsEstimating Technical Efficiency of Turkish Hospitals
Estimating Technical Efficiency of Turkish Hospitals
Mustafa Said YILDIZ
 
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
Apollo Hospitals Group and ATNF
 
A study on “impact of artificial intelligence in covid19 diagnosis”
A study on “impact of artificial intelligence in covid19 diagnosis”A study on “impact of artificial intelligence in covid19 diagnosis”
A study on “impact of artificial intelligence in covid19 diagnosis”
Dr. C.V. Suresh Babu
 
Medical Instrumentation
Medical InstrumentationMedical Instrumentation
Medical Instrumentation
Pongsak Sarapukdee
 
Insurance | Stratifying risk using wearable data | MunichRe
Insurance | Stratifying risk using wearable data | MunichReInsurance | Stratifying risk using wearable data | MunichRe
Insurance | Stratifying risk using wearable data | MunichRe
Galen Growth
 
Automated diagnosis of hepatitis b using multilayer mamdani
Automated diagnosis of hepatitis b using multilayer mamdaniAutomated diagnosis of hepatitis b using multilayer mamdani
Automated diagnosis of hepatitis b using multilayer mamdani
mga_pk
 
Introduction to biomedical instrumentation
Introduction to biomedical instrumentationIntroduction to biomedical instrumentation
Introduction to biomedical instrumentation
Sikkim Manipal Institute Of Technology
 
Care Kiosk افحص
Care Kiosk  افحصCare Kiosk  افحص
Care Kiosk افحص
Abdulrahman Jabour. PhD
 
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
iosrjce
 
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
IJECEIAES
 
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSISAN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
ijcsit
 
Bio Medical Engineering
Bio Medical Engineering Bio Medical Engineering
Bio Medical Engineering
Harindu Chathuranga Korala
 
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
IJTET Journal
 

What's hot (18)

Computer aided diagnosis for liver cancer using statistical model
Computer aided diagnosis for liver cancer using statistical modelComputer aided diagnosis for liver cancer using statistical model
Computer aided diagnosis for liver cancer using statistical model
 
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUESBLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
BLOOD TUMOR PREDICTION USING DATA MINING TECHNIQUES
 
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
IRJET -Improving the Accuracy of the Heart Disease Prediction using Hybrid Ma...
 
Automated health alerts using in home sensor data for embedded health assessment
Automated health alerts using in home sensor data for embedded health assessmentAutomated health alerts using in home sensor data for embedded health assessment
Automated health alerts using in home sensor data for embedded health assessment
 
How EMR can connect You and Your Doctor?
How EMR can connect You and Your Doctor?How EMR can connect You and Your Doctor?
How EMR can connect You and Your Doctor?
 
Estimating Technical Efficiency of Turkish Hospitals
Estimating Technical Efficiency of Turkish HospitalsEstimating Technical Efficiency of Turkish Hospitals
Estimating Technical Efficiency of Turkish Hospitals
 
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
Measuring long term effect of Telecare on Chronic Diseases by Masatugu Tsuji,
 
A study on “impact of artificial intelligence in covid19 diagnosis”
A study on “impact of artificial intelligence in covid19 diagnosis”A study on “impact of artificial intelligence in covid19 diagnosis”
A study on “impact of artificial intelligence in covid19 diagnosis”
 
Medical Instrumentation
Medical InstrumentationMedical Instrumentation
Medical Instrumentation
 
Insurance | Stratifying risk using wearable data | MunichRe
Insurance | Stratifying risk using wearable data | MunichReInsurance | Stratifying risk using wearable data | MunichRe
Insurance | Stratifying risk using wearable data | MunichRe
 
Automated diagnosis of hepatitis b using multilayer mamdani
Automated diagnosis of hepatitis b using multilayer mamdaniAutomated diagnosis of hepatitis b using multilayer mamdani
Automated diagnosis of hepatitis b using multilayer mamdani
 
Introduction to biomedical instrumentation
Introduction to biomedical instrumentationIntroduction to biomedical instrumentation
Introduction to biomedical instrumentation
 
Care Kiosk افحص
Care Kiosk  افحصCare Kiosk  افحص
Care Kiosk افحص
 
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
Application of Support Vector Machine and Fuzzy Logic for Detecting and Ident...
 
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network f...
 
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSISAN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
AN ALGORITHM FOR PREDICTIVE DATA MINING APPROACH IN MEDICAL DIAGNOSIS
 
Bio Medical Engineering
Bio Medical Engineering Bio Medical Engineering
Bio Medical Engineering
 
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
Summarization Techniques in Association Rule Data Mining For Risk Assessment ...
 

Similar to HYBRID MORTALITY PREDICTION USING MULTIPLE SOURCE SYSTEMS

A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
BASMAJUMAASALEHALMOH
 
Machine learning and operations research to find diabetics at risk for readmi...
Machine learning and operations research to find diabetics at risk for readmi...Machine learning and operations research to find diabetics at risk for readmi...
Machine learning and operations research to find diabetics at risk for readmi...
John Frias Morales, DrBA, MS
 
ICU MORTALITY PREDICTION
ICU MORTALITY PREDICTIONICU MORTALITY PREDICTION
ICU MORTALITY PREDICTION
IRJET Journal
 
IRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
IRJET- Survey on Risk Estimation of Chronic Disease using Machine LearningIRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
IRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
IRJET Journal
 
Application of artificial intelligence techniques in the intensive care unit
Application of artificial intelligence techniques in the intensive  care unitApplication of artificial intelligence techniques in the intensive  care unit
Application of artificial intelligence techniques in the intensive care unit
International Journal of Reconfigurable and Embedded Systems
 
Machine learning approach for predicting heart and diabetes diseases using da...
Machine learning approach for predicting heart and diabetes diseases using da...Machine learning approach for predicting heart and diabetes diseases using da...
Machine learning approach for predicting heart and diabetes diseases using da...
IAESIJAI
 
Proposed Model for Chest Disease Prediction using Data Analytics
Proposed Model for Chest Disease Prediction using Data AnalyticsProposed Model for Chest Disease Prediction using Data Analytics
Proposed Model for Chest Disease Prediction using Data Analytics
vivatechijri
 
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEMAN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
ijaia
 
Multiple Disease Prediction System
Multiple Disease Prediction SystemMultiple Disease Prediction System
Multiple Disease Prediction System
IRJET Journal
 
20411-38909-2-PB.pdf
20411-38909-2-PB.pdf20411-38909-2-PB.pdf
20411-38909-2-PB.pdf
IjictTeam
 
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUESPREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
IAEME Publication
 
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
TELKOMNIKA JOURNAL
 
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
IJERA Editor
 
Sepsis Prediction Using Machine Learning
Sepsis Prediction Using Machine LearningSepsis Prediction Using Machine Learning
Sepsis Prediction Using Machine Learning
IRJET Journal
 
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
IJARIIT
 
Cloud based Health Prediction System
Cloud based Health Prediction SystemCloud based Health Prediction System
Cloud based Health Prediction System
IRJET Journal
 
Multiple disease prediction using Machine Learning Algorithms
Multiple disease prediction using Machine Learning AlgorithmsMultiple disease prediction using Machine Learning Algorithms
Multiple disease prediction using Machine Learning Algorithms
IRJET Journal
 
Diabetes Prediction Using ML
Diabetes Prediction Using MLDiabetes Prediction Using ML
Diabetes Prediction Using ML
IRJET Journal
 
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
BRNSSPublicationHubI
 
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
ijdmtaiir
 

Similar to HYBRID MORTALITY PREDICTION USING MULTIPLE SOURCE SYSTEMS (20)

A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
A_Healthcare_Monitoring_System_for_the_Diagnosis_of_Heart_Disease_in_the_IoMT...
 
Machine learning and operations research to find diabetics at risk for readmi...
Machine learning and operations research to find diabetics at risk for readmi...Machine learning and operations research to find diabetics at risk for readmi...
Machine learning and operations research to find diabetics at risk for readmi...
 
ICU MORTALITY PREDICTION
ICU MORTALITY PREDICTIONICU MORTALITY PREDICTION
ICU MORTALITY PREDICTION
 
IRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
IRJET- Survey on Risk Estimation of Chronic Disease using Machine LearningIRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
IRJET- Survey on Risk Estimation of Chronic Disease using Machine Learning
 
Application of artificial intelligence techniques in the intensive care unit
Application of artificial intelligence techniques in the intensive  care unitApplication of artificial intelligence techniques in the intensive  care unit
Application of artificial intelligence techniques in the intensive care unit
 
Machine learning approach for predicting heart and diabetes diseases using da...
Machine learning approach for predicting heart and diabetes diseases using da...Machine learning approach for predicting heart and diabetes diseases using da...
Machine learning approach for predicting heart and diabetes diseases using da...
 
Proposed Model for Chest Disease Prediction using Data Analytics
Proposed Model for Chest Disease Prediction using Data AnalyticsProposed Model for Chest Disease Prediction using Data Analytics
Proposed Model for Chest Disease Prediction using Data Analytics
 
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEMAN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
AN IMPROVED MODEL FOR CLINICAL DECISION SUPPORT SYSTEM
 
Multiple Disease Prediction System
Multiple Disease Prediction SystemMultiple Disease Prediction System
Multiple Disease Prediction System
 
20411-38909-2-PB.pdf
20411-38909-2-PB.pdf20411-38909-2-PB.pdf
20411-38909-2-PB.pdf
 
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUESPREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
PREDICTION OF DIABETES MELLITUS USING MACHINE LEARNING TECHNIQUES
 
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
Therapeutic management of diseases based on fuzzy logic system- hypertriglyce...
 
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
An Intelligent Healthcare Serviceto Monitor Vital Signs in Daily Life – A Cas...
 
Sepsis Prediction Using Machine Learning
Sepsis Prediction Using Machine LearningSepsis Prediction Using Machine Learning
Sepsis Prediction Using Machine Learning
 
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
Diabetes Prediction by Supervised and Unsupervised Approaches with Feature Se...
 
Cloud based Health Prediction System
Cloud based Health Prediction SystemCloud based Health Prediction System
Cloud based Health Prediction System
 
Multiple disease prediction using Machine Learning Algorithms
Multiple disease prediction using Machine Learning AlgorithmsMultiple disease prediction using Machine Learning Algorithms
Multiple disease prediction using Machine Learning Algorithms
 
Diabetes Prediction Using ML
Diabetes Prediction Using MLDiabetes Prediction Using ML
Diabetes Prediction Using ML
 
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
Performance Evaluation of Data Mining Algorithm on Electronic Health Record o...
 
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
An Ill-identified Classification to Predict Cardiac Disease Using Data Cluste...
 

Recently uploaded

The basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptxThe basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptx
heathfieldcps1
 
A Independência da América Espanhola LAPBOOK.pdf
A Independência da América Espanhola LAPBOOK.pdfA Independência da América Espanhola LAPBOOK.pdf
A Independência da América Espanhola LAPBOOK.pdf
Jean Carlos Nunes Paixão
 
Executive Directors Chat Leveraging AI for Diversity, Equity, and Inclusion
Executive Directors Chat  Leveraging AI for Diversity, Equity, and InclusionExecutive Directors Chat  Leveraging AI for Diversity, Equity, and Inclusion
Executive Directors Chat Leveraging AI for Diversity, Equity, and Inclusion
TechSoup
 
How to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRMHow to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRM
Celine George
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
simonomuemu
 
MARY JANE WILSON, A “BOA MÃE” .
MARY JANE WILSON, A “BOA MÃE”           .MARY JANE WILSON, A “BOA MÃE”           .
MARY JANE WILSON, A “BOA MÃE” .
Colégio Santa Teresinha
 
How to Add Chatter in the odoo 17 ERP Module
How to Add Chatter in the odoo 17 ERP ModuleHow to Add Chatter in the odoo 17 ERP Module
How to Add Chatter in the odoo 17 ERP Module
Celine George
 
Chapter 4 - Islamic Financial Institutions in Malaysia.pptx
Chapter 4 - Islamic Financial Institutions in Malaysia.pptxChapter 4 - Islamic Financial Institutions in Malaysia.pptx
Chapter 4 - Islamic Financial Institutions in Malaysia.pptx
Mohd Adib Abd Muin, Senior Lecturer at Universiti Utara Malaysia
 
Introduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp NetworkIntroduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp Network
TechSoup
 
CACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdfCACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdf
camakaiclarkmusic
 
South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)
Academy of Science of South Africa
 
clinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdfclinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdf
Priyankaranawat4
 
Advanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docxAdvanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docx
adhitya5119
 
The simplified electron and muon model, Oscillating Spacetime: The Foundation...
The simplified electron and muon model, Oscillating Spacetime: The Foundation...The simplified electron and muon model, Oscillating Spacetime: The Foundation...
The simplified electron and muon model, Oscillating Spacetime: The Foundation...
RitikBhardwaj56
 
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdfবাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
eBook.com.bd (প্রয়োজনীয় বাংলা বই)
 
Types of Herbal Cosmetics its standardization.
Types of Herbal Cosmetics its standardization.Types of Herbal Cosmetics its standardization.
Types of Herbal Cosmetics its standardization.
Ashokrao Mane college of Pharmacy Peth-Vadgaon
 
Film vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movieFilm vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movie
Nicholas Montgomery
 
A Strategic Approach: GenAI in Education
A Strategic Approach: GenAI in EducationA Strategic Approach: GenAI in Education
A Strategic Approach: GenAI in Education
Peter Windle
 
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptxC1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
mulvey2
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
Jean Carlos Nunes Paixão
 

Recently uploaded (20)

The basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptxThe basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptx
 
A Independência da América Espanhola LAPBOOK.pdf
A Independência da América Espanhola LAPBOOK.pdfA Independência da América Espanhola LAPBOOK.pdf
A Independência da América Espanhola LAPBOOK.pdf
 
Executive Directors Chat Leveraging AI for Diversity, Equity, and Inclusion
Executive Directors Chat  Leveraging AI for Diversity, Equity, and InclusionExecutive Directors Chat  Leveraging AI for Diversity, Equity, and Inclusion
Executive Directors Chat Leveraging AI for Diversity, Equity, and Inclusion
 
How to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRMHow to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRM
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
 
MARY JANE WILSON, A “BOA MÃE” .
MARY JANE WILSON, A “BOA MÃE”           .MARY JANE WILSON, A “BOA MÃE”           .
MARY JANE WILSON, A “BOA MÃE” .
 
How to Add Chatter in the odoo 17 ERP Module
How to Add Chatter in the odoo 17 ERP ModuleHow to Add Chatter in the odoo 17 ERP Module
How to Add Chatter in the odoo 17 ERP Module
 
Chapter 4 - Islamic Financial Institutions in Malaysia.pptx
Chapter 4 - Islamic Financial Institutions in Malaysia.pptxChapter 4 - Islamic Financial Institutions in Malaysia.pptx
Chapter 4 - Islamic Financial Institutions in Malaysia.pptx
 
Introduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp NetworkIntroduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp Network
 
CACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdfCACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdf
 
South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)
 
clinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdfclinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdf
 
Advanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docxAdvanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docx
 
The simplified electron and muon model, Oscillating Spacetime: The Foundation...
The simplified electron and muon model, Oscillating Spacetime: The Foundation...The simplified electron and muon model, Oscillating Spacetime: The Foundation...
The simplified electron and muon model, Oscillating Spacetime: The Foundation...
 
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdfবাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
বাংলাদেশ অর্থনৈতিক সমীক্ষা (Economic Review) ২০২৪ UJS App.pdf
 
Types of Herbal Cosmetics its standardization.
Types of Herbal Cosmetics its standardization.Types of Herbal Cosmetics its standardization.
Types of Herbal Cosmetics its standardization.
 
Film vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movieFilm vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movie
 
A Strategic Approach: GenAI in Education
A Strategic Approach: GenAI in EducationA Strategic Approach: GenAI in Education
A Strategic Approach: GenAI in Education
 
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptxC1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
C1 Rubenstein AP HuG xxxxxxxxxxxxxx.pptx
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
 

HYBRID MORTALITY PREDICTION USING MULTIPLE SOURCE SYSTEMS

  • 1. DOI: 10.5121/ijci.2019.8101 1 HYBRID MORTALITY PREDICTION USING MULTIPLE SOURCE SYSTEMS Isaac Mativo1 , Yelena Yesha 1 , Michael Grasso2 , Tim Oates 1 , Qian Zhu 3 1 Department of Computer Science, University of Maryland Baltimore County (UMBC) 2 School of Medicine, University of Maryland Baltimore (UMB) 3 Department of Information Systems, University of Maryland Baltimore County (UMBC) ABSTRACT The use of artificial intelligence in clinical care to improve decision support systems is increasing. This is not surprising since by its very nature, the practice of medicine consists of making decisions based on observations from different systems both inside and outside the human body. In this paper, we combine three general systems (ICU, diabetes, and comorbidities) and use them to make patient clinical predictions. We use an artificial intelligence approach to show that we can improve mortality prediction of hospitalized diabetic patients. We do this by utilizing a machine learning approach to select clinical input features that are more likely to predict mortality. We then use these features to create a hybrid mortality prediction model and compare our results to non artificial intelligence models. For simplicity, we limit our input features to patient comorbidities and features derived from a well-known mortality measure, the Sequential Organ Failure Assessment (SOFA). KEYWORDS Decision Support Systems; Artificial Intelligence; Hybrid Systems. I. INTRODUCTION Clinical decision support systems are driven by large amounts of data sourced from various disparate systems. These may include data from lab instruments, wearable devices, prior hospital records, patient reported data, etc. Often, the clinician is exposed to a lot of this data and is expected to make decisions to improve clinical outcomes. Although decision support systems are effective in logical tasks such as alerting the user when some parameters are outside defined ranges, they are not designed to learn from the data they have. As such, they are limited to what they have been programmed to do. This limitation becomes an opportunity to experiment with artificial intelligence within the clinical setting [1]. Specifically, a hybrid system that combines attributes from both traditional clinical decision support systems and attributes derived from machine learning tools to create a prediction model becomes attractive.Hybrid systems have been used in many fields to improve performance in areas such as managing humanitarian relief chains [2] and multiple-criteria decision-making [3] . In healthcare, hybrid decision support systems have been used to help clinicians make more informed decisions [4] [5]. For mortality prediction, several hybrid approaches have been proposed such as the use of hybrid image features on neural network for cancer patients [3][6].In this paper, we take a systems theory view of a clinical setting whereby many different systems come together to inform the clinician on the best decisions to make. In the next section, we will describe what we consider to be the three systems that we consider and apply machine learning tools on.
  • 2. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 2 2. THE THREE SYSTEMS A. Diabetes For this paper, we consider the chronic disease diabetes as one of our three systems. This is because diabetes, being a disorder of metabolism, affects multiple systems within the human body [7]. When blood sugar is not well controlled and therefore outside the normal range for prolonged periods of time, the proper functioning of various organ systems is compromised. Consequently, this manifests in various body conditions such as kidney disease, neuropathy, liver disease, etc. [8] .From a cybernetics perspective, diabetes is interesting because it involves the delicate control of blood sugar by both systems inside and outside of the body. Diabetic patients often rely on external interventions such as medication ingestion, insulin injection, dietary controls, and lifestyle changes such as exercise. When the sugar level in blood is found to be outside the normal range based on information gathered, various actions are taken to bring it under control. B. Intensive Care Unit We consider the Intensive Care Unit (ICU) in a hospital to be our next system. This is because the ICU is a specialized facility with many controlled and interconnected systems that work together to intervene in critically ill patients. In the ICU, information is availed through various channels such as electronic medical records, clinical notes, instrument measurements, etc. [8] [9] All this information is then used to provide a level of clarity in making clinical decisions. Various aspects of this information such as temporal occurrence (e.g. did the blood pressure fall after or before takings a specific medication?) dependency (e.g. eat solid foods before taking a certain medication), are important. As such, in an ICU facility, the proper control of information from various systems is critical [10] . In this paper therefore, we view the ICU facility as one large system with many subsystems within it. C. Comorbidities Comorbidities are co-occurring disease conditions within a person. Often, diabetic patients have other disease conditions as well. Some of these conditions are more commonly associated with diabetes than others [11]. In this paper, we think of comorbidities as a system because they are well-defined conditions within the human body that individually and collectively influence the wellbeing of the entire body. Since comorbidities are an important piece of information considered in patient treatment, we considered comorbidities to be relevant in informing our hybrid prediction model. With these three systems, we build a hybrid mortality prediction model and evaluated its performance. It is a hybrid model because it combined both machine learning techniques and non-machine learning techniques to select it features and perform its predictions. In the next section, we discuss the materials we used to do our experiment. 3. MATERIALS A. Medical Information Mart for Intensive Care (MIMIC) III
  • 3. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 3 The data source we used to get our patient population is MIMIC-III [12] . MIMIC-III is a large and single-center database comprising de identified health related information containing 46,520 patients of whom 20,399 are female and 26,121 are male. These patients stayed in critical care units of the Beth Israel Deaconess Medical Center between 2001 and 2012. The database includes information such as demographics, vital sign measurements, laboratory test results, procedures, medications, caregiver notes, imaging reports, and mortality. In this study, we accessed MIMIC- III to extract diabetes patients for mortality prediction. B. Elixhauser comorbidity measure. In order to get the comorbidity information for our diabetic patients, we used The Elixhauser comorbidity index [13] . This index consists of 30 comorbidity measures that have been shown to have an impact on patient hospital length of stay, cost, and mortality [14] . Classification in to this index is based on conditions described by the ICD-9-CM (International Classification of Diseases, Ninth Edition, Clinical Modifications) discharge records. The Elixhauser index assigns a value of 1 for each of 30 predefined conditions, and computes the sum. Therefore, a score of 10 indicates a patient has 10 conditions present. C. Mortality Measure To compare our hybrid model prediction accuracy with a needed a tool that measured mortality. For this, we chose the Sequential Organ Failure Assessment (SOFA) [15]. SOFA is developed from a large sample of ICU patients from around the world and gives a score based on the severity of on a patient. The SOFA score is computed based on six variables, with each variable corresponding to an organ system. The organ systems included in SOFA are respiratory, cardiovascular, hepatic, coagulation, renal and neurological systems. Each organ system is assigned a value between 0 (good) and 4 (severe). The SOFA score is the sum of all the individual organ system values. 4. METHODS In this section, we describe the methods we used to retrieve the patient cohort, computer the mortality and comorbidity measure, and to predict mortality. For this study, ethical approval was not needed because the data used is de-identified t o conform with the Health Insurance Portability and Accountability Act (HIPAA). Additionally, the data is publicly available to researchers. A. Patient Retrieval From our MIMIC III PostgreSQL database, we retrieved the diabetes patient cohort from MIMIC III database based on ICD-9 codes for diabetes. To avoid restricting diabetic patients with comorbidities for further mortality prediction, we extracted the entire diabetes diagnoses for each of the diabetic patients. B. Mortality Measure Computation We used the SOFA score as a mortality measure. To calculate the SOFA score for each patient,
  • 4. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 4 we extracted the patient in the PostgreSQL database for calculating the the SOFA score corresponding to the organ areas as shown in table 1. We extracted this data using query scripts in PostgreSQL from the MIMIC III diabetic patients data based on the first day of each patient's’ ICU stay. For each patient, we used an SQL script to calculate the actual points for each organ system, then added up the points for each organ system to obtain the SOFA score. For missing values, we assigned zero points to that organ system. We did this for all collected diabetic patients. Table 1: Sofa Organ Areas and Points Thresholds Organ/Points 1 2 3 4 Respiratory PaO 2 /FiO 2 (mmHg) <400 <300 <200 and respiratory support <100 and respiratory support Cardiological Mean arterial pressure MAP < 70 mm/Hg dop <= 5 or dob dop > 5 OR epi <= 0.1 OR nor <= 0.1 dop > 15 OR epi > 0.1 OR nor > 0.1 Hepatic Bilirubin (mg/dl) [μmol/L] 1.2–1.9 [> 20-32] 2.0–5.9 [33-101] 6.0–11.9 [102-204] > 12.0 [> 204] Neurological Glasgow coma scale 13 - 14 10 - 12 6 - 9 <6 Coagulation Platelets×103/μl <150 <100 <50 <20 C. Comorbidity Measure and Weight Computation We used the Elixhauser index to calculate the comorbidity measure for each diabetic patient in MIMIC-III. For each diabetic patient, we extracted 29 different diagnosis excluding diabetes and quantified them by using the Elixhauser index for comorbidities represented in the diabetic population. In order to improve the efficiency of our classifier by reducing the number of features, we identified the top five comorbidities that are most predictive of mortality across the identified diabetic patients. To do this, we experimented using 5 different feature selection algorithms, which are Gain Ratio [16] , Correlation [17] , Symmetrical Uncertainty [18] , Information Gain [19] , and Correlation Feature Selection (CFS) Subset Evaluator [20] . For each algorithm run, we ranked the comorbidities based on the resulted predictive mortality. We then averaged these results over the 5 algorithms to get our final top 5 most predictive comorbidities. The final list of the comorbidities was cardiac arrhythmias, coagulopathy, metastatic cancer, congestive heart failure, and fluid electrolyte. D. Mortality Prediction For mortality prediction, we created a hybrid mortality prediction model that took as input the 6 SOFA scores representing the 6 organ systems and the top 6 ranked comorbidities, for a total of 11 features. We then experimented with a Naive Bayes classifier and a Random Forest classifier
  • 5. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 5 with 10-fold cross validation across our patient population. We compared the prediction accuracy of our hybrid system with the accuracy of (i) a prediction model with only the SOFA scores and (ii) a prediction model with only the comorbidity information. Our accuracy comparison was based on the AUC on both the Naive Bayes and Random Forest classifiers on mortality. 5. RESULTS A. Patient Retrieval A total of 10,403 of diabetes patients were retrieved from 46,520 patients contained in MIMIC III database. Of all the identified diabetic patients, 1,513 died while in ICU. Among the 10,403 diabetes patients, we identified 29 different comorbidities. B. Mortality Measure Computation Table 2 shows the calculated SOFA scores and the corresponding mortality rates as computed from our data. Our mortality rates were consistent with the expected SOFA mortality rates. Table 2: Calculated Mortality Rates C. Comorbidity Measure and Weights Figure 1 shows the results of calculated comorbidity for the diabetic patients using the Elixhauser index. As can be seen, most of the patients had comorbidities. Figure 1: Comorbidities Across Patients
  • 6. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 6 Table 3 shows results generated via the five feature extraction algorithms to determine what comorbidities to be applied for prediction classifiers. These are the comorbidities that were used as features in our model. Table 3: Highly Weighed Comorbidities D. Mortality Prediction Table 4 shows the results generated by the Naive Bayes and Random Forest classifiers. The ROC is lowest when using only the 5 comorbidity features are used, is better when using the 6 SOFA scores are used, and best when both the 5 comorbidities and 6 SOFA scores are combined. Thus, the hybrid model outperforms the natively SOFA or comorbidity only models. Table 4: Mortality Prediction Results 6. DISCUSSION In this study, we proposed viewing ICU clinical care through the lens of cybernetics. We
  • 7. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 7 identified three systems; the ICU, diabetes, and patient comorbidity, as related and defined systems that are important in supporting clinical decisions. We examined the mortality impact of comorbidities on diabetic patients patients admitted in an ICU unit using machine learning tools.We created a hybrid mortality prediction model and compare its results to non-hybrid models.The results we obtained support our claim that mortality prediction can be improved by incorporation key features from different systems (comorbidities in our case) and using artificial intelligence to train our classifiers. Another key observation from our results is that we can still get improved performance from our prediction model without using all of the available input features. In our case, we used statistical techniques to rank all our comorbidities based on their weight in impacting mortality. This observation is important because within systems, we can optimize performance by carefully selecting the components that affect our output. This careful selection results in dimension reduction and consequently reduces computational resources needed to run algorithms while at the same time producing improved results. A logical extension of this research is further defining the subsystems within our system, or yet still identifying other systems that are important in improving clinical decision support systems.An improvement of our approach would be to take a closer look at the information flow within and across our systems and evaluate its impact on the overall efficiency of an ICU setting.Important makers, in this case, could include items such as readmission rates, adverse events, hospital length of stay, central line infection, etc. We hope that with this study we have demonstrated that it is possible to view clinical decision support systems from cybernetics perspective and encouraged other researchers to consider this approach. DISCLOSURE OF INTERESTS The authors report no conflict of interest. REFERENCES 1] R. Benbenishty and R. Treistman, ―The development and evaluation of a hybrid decision support system for clinical decision making: The case of discharge from the military,‖ Soc.Work Res. , vol. 22, no. 4, pp. 195–204, Dec. 1998. [2] N. Sahebjamnia, S. A. Torabi, and S. A. Mansouri, ―A hybrid decision support system for managing humanitarian relief chains,‖ Decis. Support Syst. , vol. 95, pp. 12–26, Mar. 2017. [3] E. Kazimieras Zavadskas et al. , ―HYBRID MULTIPLE CRITERIA DECISION MAKING METHODS: A REVIEW OF APPLICATIONS IN ENGINEERING,‖ Scientia Iranica , vol.23, no. 1, pp. 1–20, 2016. [4] A. Mansoul, B. Atmani, and S. Benbelkacem, ―A hybrid decision support system :application on healthcare,‖ arXiv [cs.AI] , 16-Nov-2013. [5] O. W. Samuel, G. M. Asogbon, A. K. Sangaiah, P. Fang, and G. Li, ―An integrated decision support system based on ANN and Fuzzy_AHP for heart failure risk prediction,‖ Expert Syst. Appl. , vol. 68, pp. 163–172, 2017.
  • 8. International Journal on Cybernetics & Informatics (IJCI) Vol. 8, No.1, February 2019 8 [6] P. Yan, H. Guo, G. Wang, R. De Man, and M. K. Kalra, ―Hybrid deep neural networks for all-cause Mortality Prediction from LDCT Images,‖ arXiv [cs.CV] , 19-Oct-2018. [7] F. Gomez, Blood Sugar Levels: Clear and Unbiased Facts About Diabetes . Lulu Press, Inc,2016. [8] E. Bonora and R. A. DeFronzo, Diabetes Complications, Comorbidities and Related Disorders . Springer, 2018. [9] W. Koller, ―Flow of Information in the ICU — An Overview,‖ in Anaesthesia, Pain,Intensive Care and Emergency Medicine — A.P.I.C.E. , 1996, pp. 883–891. [10] S. L. Groves, ―Tools for ICU patient safety and quality improvement plans. Critical care safety: Essentials for ICU patient care and technology. ECRI Institute Staff. Publisher:ECRI Institute; 146 pages, plus CD,‖ J. Healthc. Risk Manag. , vol. 27, no. 3, pp. 11–11,2007. [11] P. Klimek, A. Kautzky-Willer, A. Chmiel, I. Schiller-Frühwirth, and S. Thurner,―Quantification of diabetes comorbidity risks across life using nation-wide big claims data,‖PLoS Comput. Biol. , vol. 11, no. 4, p. e1004125, Apr. 2015. [12] A. E. W. Johnson et al. , ―MIMIC-III, a freely accessible critical care database,‖ Sci Data ,vol. 3, p. 160035, May 2016. [13] A. Elixhauser, C. Steiner, D. R. Harris, and R. M. Coffey, ―Comorbidity measures for use with administrative data,‖ Med. Care , vol. 36, no. 1, pp. 8–27, Jan. 1998. [14] M. E. Menendez, V. Neuhaus, C. N. van Dijk, and D. Ring, ―The Elixhauser comorbidity method outperforms the Charlson index in predicting inpatient death after orthopaedic surgery,‖ Clin. Orthop. Relat. Res. , vol. 472, no. 9, pp. 2878–2886, Sep. 2014. [15] A. E. Jones, S. Trzeciak, and J. A. Kline, ―The Sequential Organ Failure Assessment score for predicting outcome in patients with severe sepsis and evidence of hypoperfusion at the time of emergency department presentation,‖ Crit. Care Med. , vol. 37, no. 5, pp.1649–1654, May 2009. [16] A. G. Karegowda, A. S. Manjunath, and M. A. Jayaram, ―Comparative study of attribute selection using gain ratio and correlation based feature selection,‖ International Journal of Information Technology and Knowledge Management , vol. 2, no. 2, pp. 271–277, 2010. [17] M. A. Hall and L. A. Smith, ―Feature Selection For Machine Learning: Comparing a Correlation- based Filter Approach to the Wrapper,‖ 1999. [Online]. Available: http://www.aaai.org/Library/FLAIRS/1999/flairs99-042.php . [Accessed: 17-Nov-2016]. [18] L. Yu and H. Liu, ―Feature selection for high-dimensional data: A fast correlation-based filter solution,‖ in ICML , 2003, vol. 3, pp. 856–863. [19] P. Jia, J.-H. Dai, Y.-H. Pan, and M.-L. Zhu, ―Novel algorithm for attribute reduction based on mutual- information gain ratio,‖ JOURNAL-ZHEJIANG UNIVERSITY ENGINEERING SCIENCE , vol. 40, no. 6, p. 1041, 2006. [20] M. A. Hall and L. A. Smith, ―Practical feature subset selection for machine learning,‖ 1998