Short-Term and Long-Term Readmission Prediction in Uncontrolled

Diabetic Patients using Machine Learning Techniques

Monira Mahmoud

, Mohamed Bader

and James McNicholas

1,2

University of Portsmouth, Buckingham Building, Lion Terrace, Portsmouth PO1 3HE, U.K.

Queen Alexandra Hospital, Portsmouth Hospitals NHS Trust, U.K.

Keywords: Machine Learning, Data Mining, Diabetes, Uncontrolled Diabetes, Readmission.

Abstract: Diabetes is a chronic disease and major health problem which leads to many complications if not managed

probably. Hyperglycemia, or raised blood sugar, is a common effect of Uncontrolled diabetes that may leads

overtime to serious complications, especially in the nerves and blood vessels. As well as leads to repeated

hospital admission. The main purpose of this study is to help clinicians to improve healthcare of uncontrolled

diabetic patients through using machine learning as a tool in decision making, consequently this will improve

patient care and reduce the readmission which considered a medical quality measurement and cost reduction

objective. This study aims to predict the hospital readmission of the uncontrolled diabetic patient who is

considered more susceptible to developing life-threatening diabetes complications and based on the Diabetes

130-US hospitals dataset. Several machine learning employed to predict the short term (within 30 days), and

both short and long-term readmission (within or after 30 days) of uncontrolled diabetic patient. As expected,

the results are in line with other research in the literature. For the first scenario of whole readmission

prediction, our model achieved a better accuracy of 64.5 % with SVM and attribute selection and for the

second scenario, RF achieved the highest accuracy of 86.38 % which still come in context with other research

in the literature.

INTRODUCTION

Diabetes Mellitus (DM) is a major public health

problem, Worldwide, 415 million adults—or one in

every eleven—are projected to have diabetes. By

2040, there will likely be 642 million individuals

living with diabetes worldwide. (diabetes UK

organization). According to the world health

organization, the number of people with diabetes rose

from 108 million in 1980 to 422 million in 2014.

Also, within the UK, there are 3.5 million diabetics,

up from 1.4 million in 2000. Hospital readmission is

an episode when a patient who has been discharged

from the hospital is readmitted again within a

specified period. Indeed, the burdens of inpatient

diabetes is huge, growing, and expensive, and

readmission can greatly increase these burdens.

Hospital readmission is used as a measure of a

hospital’s ability to provide quality service and

patient care. Also, hospital readmission is often used

as a benchmark, since a high proportion of

readmission is likely to be preventable if the hospital

provided adequate care. Thus, the reduction of

readmission is a medical quality measurement and

cost reduction objective (Battineni et al, 2020). In

particular, uncontrolled diabetes implies high blood

sugar levels over a prolonged time even if the patient

on treatment, it is diagnosed when HbA1c is higher

than 6.5. According to Diabetes UK Organization,

HbA1c is one of the tests used to diagnose and

monitor the diabetic patient, known as glycated

hemoglobin, and refers to average blood glucose

levels for the last two to three months. For a diabetic

patient, an ideal HbA1c level is 48mmol/mol (6.5).

Uncontrolled diabetes can result in hyperglycemia,

which damages many of the body’s systems,

particularly the nerves and blood vessels, over time.

Nearly every organ in an uncontrolled diabetic

patient’s body can suffer a toll from diabetes,

including, the eyes, kidneys, nerves, heart, blood

vessels, gastrointestinal tract, teeth, and gum.

Interestingly, on a daily basis, hospitals generate a

great deal of data, but that information usually

remains as data that is not always converted into

knowledge. Through the application of ML

techniques, it is possible to uncover hidden

relationships or patterns among the data and convert

them into knowledge that can be used by healthcare

680

Mahmoud, M., Bader, M. and McNicholas, J.

Short-Term and Long-Term Readmission Prediction in Uncontrolled Diabetic Patients using Machine Learning Techniques.

DOI: 10.5220/0011926000003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 5: HEALTHINF, pages 680-688

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

professionals to make better decisions. Prediction of

readmission could play a role in early intervention for

the management of the uncontrolled diabetic patient

who is considered a host of complications if not

managed properly. Hence, this study aims to apply a

set of machine learning techniques to predict

uncontrolled patient readmission. Therefore,

predicting readmission will ultimately allow hospitals

to better calculate and assess the quality of care.

This study applies machine learning prediction

tools for a specific group of diabetic patients

(uncontrolled patients), based on UCI diabetes

dataset. Moreover, it considers different scenarios for

prediction (i.e. short-term or short- and long-term

readmission prediction) with feature selection. Six

supervised ML technique used for the prediction (RF,

NB, KNN, Ada-Boost, SVM, bagging, and NN) of

readmission. The study benefits from two scenarios.

The first scenario (i.e. using the first subset of data)

predicts the readmission event, while the second

scenario (using subset data two) predicts of the early

readmission (readmission within 30 days).

Experiments employed for both sub data sets with and

without attribute selection. Results shows that, in the

first scenario (all readmission events), SVM achieved

the highest accuracy of 64 % and NB achieved the

best AUROC area of 0.65. In the second scenario

(early readmission only), RF achieved the highest

accuracy of 86 % and the best AUROC area of 0.63.

Our goal within the healthcare prospective is to

use data mining, data analytical and ML to predict If

the uncontrolled diabetic patient will be readmitted at

any time point, as a first scenario or If will be

readmitted in a short-term time (within 30 days), as a

second scenario.

This research aims to develop a model that can

accurately predict the readmission of uncontrolled

diabetic patient. Also, to provide a better

understanding of the readmitted patient

characteristics through descriptive analysis.

Therefore, using ML for prediction of uncontrolled

diabetics readmission will boost early intervention,

and consequently lead to better disease management

and cost reduction

The paper is structured as follows: Section 2

presents the literature review on machine learning

(ML) and diabetes. Section 3 discusses the material

and methods. Section 4 provides the results, and

Section 5 provides conclusion.

LITERATURE SURVEY

A large and growing body of literature has

investigated the application of machine learning

algorithms in the healthcare domain ((Battineni et al,

2020), (Kumar et al, 2018), (Kohli et al, 2018)), (Ali

et al, 2020). In particular, a stream of research

examines the accuracy of machine learning

algorithms in predicting hospital readmission of

diabetic patients. The following section discuss the

existing literature related to this paper.

2.1 Diabetes and Machine Learning

Diabetes is linked to micro and macrovascular

diseases such as heart disease, kidney failure, eye

disease, and amputation, which also leads to a high

rate of repeated admission of diabetic patients.

Moreover, it is fast becoming a key instrument in

complicating other unrelated medical conditions like

infections, accidents, and surgery. For instance, the

United States (US) health system endures a

significant economic burden for diabetes care. This

cost reached about 327 billion dollars in 2017

(Kavakiotis et al 2017). Nevertheless, the cost of

diabetes is not directly related to the diagnosis and

management of diabetes itself but also costs

generated by long-term complications and their

economic and social consequences (Alamer et al,

2019).

Furthermore, uncontrolled diabetes, if not

managed properly, often leads to biochemical

imbalances that can cause acute life-threatening

events and hospitalizations. Evidently, the

uncontrolled diabetic patient is nine times higher risk

of admission (Boutayeb et al, 2004), three times more

susceptible to developing severe periodontitis (Hu et

al ,2019), much greater risk for presenting with later

stages of diabetic retinopathy, other rare diabetic

ocular complications, including glaucoma, cataract,

and dry eye disease (Eldarrat et al ,2011). Extent

research links uncontrolled diabetes with substantial

mortality and cardiovascular disease burden (Alamer

et al, 2019) and increases the risk of peroperative

complication (Threatt et al ,2013).

Therefore, predicting readmission will ultimately

allow hospitals to better calculate and assess the

quality of care they provide to their patients (Navarro-

Pérez et al ,2018). The readmission of an individual

with uncontrolled diabetes falls into the Potentially

Preventable Readmission (PPRs) category. Since

ambulatory care (outpatient care) plays an

important

role in diabetes management, most hospitalizations

with uncontrolled diabetes are a direct reflection of

the quality of primary health care received outside of

hospitals (Kim et al, 2010). Accordingly, the Agency

for Healthcare Research and Quality (AHRQ)

Short-Term and Long-Term Readmission Prediction in Uncontrolled Diabetic Patients using Machine Learning Techniques

681

selected uncontrolled diabetes as a prevention quality

indicator (PQI) where hospitalization would be

decreased through timely and appropriate ambulatory

care ((Pujianto et al ,2019), (Kim et al, 2007)).

Machine learning (ML) is a subclass of artificial

intelligence technology, where algorithms process

large data sets to detect patterns, learn from them, and

execute tasks autonomously without being instructed

on exactly how to address the problem. There is

ample evidence on the rapid increase in Machine

learning applications in disease prediction and

diagnosis ((Battineni et al,2020), (Kumar et al,2018),

(Kohli et al, 2018) and (Ali et al,2020)). Thus, using

machine learning to predict the readmission of

diabetic patients will play a role in improving the

healthcare system by decreasing the negative

consequences related to diabetes readmission.

In the context of diabetes, ML methods have been

used to detect, predict, and diagnose i.e. bio-marker

Prediction and Diagnosis in DM (Farajollahi etal

,2021), Diabetic Complications (Dagliati et al,2018),

Drugs and Therapies (Donsa et al, 2015), Genetic

Background and Environment (Urban et al, 2018),

and Health Care Management which includes the

readmission prediction (Sharma et al ,2019). An

example, Chaki et. al (2020) surveyed 107 papers that

addressed the application of machine learning and

artificial intelligence techniques in DM detection,

diagnosis, and self-management. Likewise, Dagliati

et. al (Dagliati et al., 2018) provides empirical

evidence on the importance of ML in predicting the

complications of diabetes.

2.2 Related Work

Several studies used the UCI diabetes dataset for the

purpose of diabetic patient readmission prediction.

However, the results are mixed due to the variation in

the data prepossessing and the used ML algorisms.

Bhuvan et. al studies both short-term and long-term

readmission as two scenarios (Bhuvan et. al,2016).

The first scenario considered the short-term

readmission versus all readmission cases. The second

scenario combined all the readmission cases versus

non-readmitted cases. They found that RF was

optimal for this task, compared to NB, Ada-Boost,

and NN. Moreover, they employed an ablation study

to identify risk factors and association rule mining to

identify the association across critical risk factors.

They found that the number of inpatient visits,

discharge disposition, and admission type are the

most important for identifying the high risk patient.

Proposing an ensemble model and cluster

analysis, Pham et al (Pham et al,2019), investigate the

whole readmission events. the final ensemble model

was created using the five best models, which were

chosen from a pool of 15 models. The final ensemble

reaches a 56 % sensitivity while maintaining a 63.5

% accuracy. Using cluster analysis, they identified

four unique patient groupings. Their results suggest

that patients who have had previous in-patient visits

or who received a large amount of treatment during

their most recent visit were shown to be more likely

to be readmitted.

Addressing short-term readmission, Al-Ars et al

(Al-Ars et al, 2022), Farajollahi et al (Farajollahi et

al,2021), Sharma et al (Sharma et al ,2019) and Neto

et al (Neto et al,2021) explored the accuracy of

alternative predictors and the attributes selections for

predicting rea mission of diabetic patients.

Sharma et al (Sharma et al ,2019) investigate the

prediction of short-term readmission using RF, LR,

XGBoost, Adaboost and DT. They concluded that

random forest achieved t highest accuracy of 94. They

also pointed out the most important 10 attributes

which contribute mostly to the hospital readmission

of a diabetes patient in case of using RF an DT

algorisms, However, the handling out of the

prediction attribute not defined.

Furthermore, Al-Ars et al (Al-Ars et al, 2022)

studies prediction of the short-term readmission

based on the measurement of HbA1c and the primary

diagnosis using LR, NB, J8 and comparing the results

with and without using t discretization step. They

found that the discretization o numerical attributes

step improves the performance of N into 93.51.

Applying principal component analysis (PCA) for

feature selection, Farajollahi et al (Farajollahi et

al,2021) identified three scenarios of attribute

selection. In This paper, they employed RF, DT,

XGBoost, KNN, AdaBoost, and Deep learning to

predict the short-term readmission and found that dee

learning achieved the highest accuracy of 86.8%.

However, the handling out of the prediction attribute

not defined. the study showed that a machine learning

model’s effectiveness depends on the choice of the

prediction model, the numb of selected features, and

the number” k” for k-fold validation.

Furthermore, using six different scenarios based

on attribute selection, Neto et al (Neto et al,2021)

considered the short-term readmission, using RF, J48,

NB, IBK, and MLP algorithms. Comparing

alternative scenarios, they documented that the best

performance is for the RF with an accuracy of 0,898

in the case of the scenarios with the highest number

of attributes.

CCH 2023 - Special Session on Machine Learning and Deep Learning for Preventive Healthcare and Clinical Decision Support

682

MATERIALS AND METHODS

3.1

Materials

This study is based on a dataset obtained from the

UCI machine learning repository about diabetic

patients (Dua and Graff,2019). The data set contains

about 100,000 instances and it includes 55 features

from 130 hospitals in the United States for10 years

(1999-2008). the attributes describing the diabetic

encounters, including demographics, diagnoses,

diabetic medications, number of visits in the year

preceding the encounter, and payer information.

3.2 Methodology

This research will follow The CRoss Industry

Standard Process for Data Mining (CRISP-DM)

methodology. The CRISP-DM steps will be

described in details next. For the data preparation

phase of this study, Excel used for data preparation

and WEKA for the Modelling and Evaluation.

Excel’s usability and the number of classifiers

available by WEKA made it the ideal tool for this

analysis.

A. Business Understanding

As a measure of a hospital’s ability to provide quality

service and care, readmissions are often used as a

benchmark since most readmissions can be prevented

if patients receive adequate treatment. In addition to

being a quality indicator of healthcare systems,

readmissions are also a financial burden, about 3.3

million readmissions were reported in the United

States after 30 days, according to the Agency for

Healthcare Research and Quality (AHRQ). The

burden of inpatient diabetes is huge, growing, and

expensive, and readmission can greatly increase this

burden. Nevertheless, reducing readmission rates for

diabetics could significantly reduce medical costs

while improving care outcomes. The reduction of

readmission is a medical quality measurement and

cost reduction objective (Battineni et al,2020). As

well as the uncontrolled diabetic patent is considered

a host of diabetes complications which is considered

a cost burden as well. As a result, predicting cases of

uncontrolled diabetes patients who are likely to have

hospital readmission is the project’s commercial goal

in order to help decrease the readmission rate.

This graph shows the summary of this research

methodology.

Figure 1: Methodology.

B. Data Understanding

This study is based on a data set obtained from the

UCI machine learning repository about diabetic

patients (Dua and Graff,2019). The dataset contains

about 100,000 instances and it includes 50 features

from 130 hospitals in the United States for 10 years

(1999-2008). the attributes describing the diabetic

encounters, including demographics, diagnoses,

diabetic medications, number of visits in the year

preceding the encounter, and payer information. The

full list of the features and their description is

provided in Table1 (

Strack

et al, 2014).

C. Data Preparation

To ensure that the data is suitable to be used in the

various models, the following data prepossessing

methods are applied. Figure 2 shows a summary of

the prepossessing steps.

Figure 2: Summary of the prepossessing steps.

Short-Term and Long-Term Readmission Prediction in Uncontrolled Diabetic Patients using Machine Learning Techniques

683

Table 1: Data description.

Feature name Type Description and Values %

missing

Encounter ID Numeric Unique identifier of an encounter 0%

Patient number Numeric Unique identifier of a patient 0%

Race Nominal Values: Caucasian, Asian, African American, Hispanic, and other 2%

Gender Nominal Values: male, female, and unknown/invalid 0%

Age Nominal Grouped in 10-year intervals: 0, 10), 10, 20), …, 90, 100) 0%

Weight Numeric Weight in pounds. 97%

Admission type Nominal Integer identifier corresponding to 9 distinct values, for example, emergency, urgent,

elective, newborn, and not available

Discharge disposition Nominal Integer identifier corresponding to 29 distinct values, for example, discharged to

home, expired, and not available

Admission source Nominal Integer identifier corresponding to 21 distinct values, for example, physician referral,

emergency room, and transfer from a hospital

Time in hospital Numeric Integer number of days between admission and discharge 0%

Payer code Nominal Integer identifier corresponding to 23 distinct values, for example,Blue Cross/Blue

Shield, Medicare, and self-pay

52%

Medical specialty Nominal Integer identifier of a specialty of the admitting physician, corresponding to 84

distinct values, for example, cardiology, internal medicine, family/general practice,

and surgeon

53%

Number of lab procedures Numeric Number of lab tests performed during the encounter 0%

Number of procedures Numeric Number of procedures (other than lab tests) performed during the encounter 0%

Number of medications Numeric Number of distinct generic names administered during the encounter 0%

Number of outpatient visits Numeric Number of outpatient visits of the patient in the year preceding the encounter 0%

Number of emergency visits Numeric Number of emergency visits of the patient in the year preceding the encounter 0%

Number of inpatient visits Numeric Number of inpatient visits of the patient in the year preceding the encounter 0%

Diagnosis 1 Nominal The primary diagnosis (coded as first three digits of ICD9); 848 distinct values 0%

Diagnosis 2 Nominal Secondary diagnosis (coded as first three digits of ICD9); 923 distinct values 0%

Diagnosis 3 Nominal Additional secondary diagnosis (coded as first three digits of ICD9); 954 distinct

values

Number of diagnoses Numeric Number of diagnoses entered to the system

Glucose serum test result Nominal Indicates the range of the result or if the test was not taken. Values: “>200,” “>300,”

“normal,” and “none” if not measured

A1c test result Nominal Indicates the range of the result or if the test was not taken. Values: “>8” if the result

was greater than 8%, “>7” if the result was greater than 7% but less than 8%,

“normal” if the result was less than 7%, and “none” if not measured.

Change of medications Nominal Indicates if there was a change in diabetic medications (either dosage or generic

name). Values: “change” and “no change”

Diabetes medications Nominal Indicates if there was any diabetic medication prescribed.Values: “yes” and “no” 0%

24 features for medications Nominal For the generic names: metformin, repaglinide, nateglinide, chlorpropamide,

glimepiride, acetohexamide, glipizide, glyburide, tolbutamide, pioglitazone,

rosiglitazone, acarbose, miglitol, troglitazone, tolazamide, examide, sitagliptin,

insulin, glyburide-metformin, glipizide-metformin, glimepiride-pioglitazone,

metformin-rosiglitazone, and metformin-pioglitazone, the feature indicates whether

the drug was prescribed or there was a change in the dosage. Values: “up” if the

dosage was increased during the encounter “down” if the dosage was decreased,

“steady” if the dosage did not change, and “no” if the drug was not prescribed

Readmitted Nominal Days to inpatient readmission. Values: “<30” if the patient was readmitted in less than

30 days, “>30” if the patient was readmitted in more than 30 days, and “No” for no

record of readmission.

•Missing Data:

weight attribute (97 % missing) was considered to be

too sparse and it was not included in further analysis.

Furthermore, the payer code attribute is considered

irrelevant to the outcome as well as it has a high

percentage of missing values so it is excluded too.

“medical specialty” refers to the specialty of

CCH 2023 - Special Session on Machine Learning and Deep Learning for Preventive Healthcare and Clinical Decision Support

684

attending physician which has some missing data so

we fill “Missing” in the missing place as this is an

important feature for analysis.

•Zero Variance Attributes:

Troglitazone, acetohexamide,citoglipton,glimepirie

pioglitazone,metformin pioglitazone, and examide

were excluded as no patients on these drugs.

•Near Zero Variance Attributes:

metformin rosiglitazone, glipizide-metformin,

tolazamide, tolbutamide, chlorpropamide, and

miglitol were excluded as there are only very few

cases with steady doses (less than 10 instances).

•Transformation of Skewed Variables:

Age attribute is categorised into 3 distinct groups

based on trends proposed by Beata Strack et al (Strack

et al,2014). Admission Type id, admission source,

and discharge disposition id attribute are categorised

with similar categories merged.

•Discretization:

The three diagnosis results are given in icd-9 coding

discretized into 9 groups. As well as discretization

applied to the numerical attributes (time in hospital,

number medications, number lab procedures, number

procedures, number outpatient, number emergency

and number inpatient) discretized in to 5 pins, using

unsupervised splitting technique based on a specified

number of bins.

•Class Imbalance:

SMOTE is used to balance the prediction variable

classes. For the purpose of uncontrolled diabetic

patient readmission prediction.

At the end ,2 sub data sets were extracted from the

original one. The first subset for the prediction of all

readmission cases (within 30 days or after 30 days

counted as yes). The second subset for the prediction

of the early readmission cases (excluding all

readmission after 30 days).

This ended with 35 attributes in the first subset

(long- term and short-term readmission) and 9102

instances. A 6282 instances and 34 attributes in the

second subset (short- term readmission only).

The following charts shows the distribution of the

prediction variable (readmission) in the data set and

the two sub sets. Chart 1 and 2 shows the distribution

of the readmission through the subsets. Chart 3 shows

the distribution of the excluded data set for

uncontrolled diabetics.

Chart 1: Distribution of short-term readmission.

Chart 2: Distribution of uncontrolled diabetic patient.

Chart 3: distribution of the excluded data set for

uncontrolled diabetics.

D. Modelling

WEKA was the tool that has been chosen for this step

because of its variety of classification methods, this

study used a tree-based Algorithm (RF), a Bayesian

learning algorithms (NB), a function algorithm

(SVM, NN), a meta algorithm (Ad boost), and a lazy

algorithm (KNN). After choosing the algorithms,

Sampling been done with

10 folds cross validation while 30 % used for the

test set and 70 % used for the training set. Cross

validation has been used to give the model an

opportunity to be trained on multiple (10) train test

splits as well as it reduces over fitting. Also, all these

Short-Term and Long-Term Readmission Prediction in Uncontrolled Diabetic Patients using Machine Learning Techniques

685

algorithms used through the filtered classifier

algorism in WEKA to apply the over sampling for the

training set only not on both training and test sets.

Finally, two prediction scenarios have been

developed to compare the results. The first scenario

for the prediction of all readmission cases (within 30

days or after 30 days). The second scenario for the

prediction of the early readmission cases (excluding

all readmission after 30 days). the output of this step

were 2 sub data sets one for the early readmission

prediction and the second for the readmission

prediction.

E. Evaluation

The basic performance parameters this study

considers are the model accuracy and AUROC (Area

Under curve for the ROC). While AUROC is the

measure of the ability of a classifier to distinguish

between classes, the accuracy is the fraction of

predictions our model got right.

RESULTS

For the first scenario, SVM achieved the highest

accuracy of 64.2 % and NB achieved the best

AUROC area of 0.65. For the second scenario, RF

achieved the highest accuracy of 86.38 % and the best

AUROC area of 0.63. Table 2 summarizes the results:

Table 2: Results summary.

Also, as noticed from the results, the second scenario

shows a much better accuracy of 86 %, but the first

scenario shows a little better AUROC (.65). This

figure compares the AUROC and accuracy for each

algorithm in both scenarios.

Chart 4: Results graph.

DISCUSSION

Although this research targets uncontrolled diabetics

and all research in literature targets all diabetic

patients. The results comes, as expected, in context

with other research in literature, especially the whole

readmission predictions as example ,Pham et al

ensemble model achieved an accuracy of 63 %

accuracy , our model achieved a better accuracy of

64.5 % with SVM and attribute selection .For the

second scenario for the short term uncontrolled

diabetic readmission prediction, although, Sharma et

al RF model achieved 94 % accuracy and Alars et al

NB model achieved 93.5 % ,our model is still in

context with other research in literature, as example

Neto et al RF model achieved 89.8 % and Farajollahi

et al achieved 86.8 % using deep learning. the

difference in the data sample used in this research

(uncontrolled diabetic patient may explain the

difference in accuracy with other researchers.

CONCLUSION

In this study, several machinebased methods were

proposed to predict short-term and long-term

uncontrolled diabetic readmission. SMOTE-based

data pre-processing is introduced to address the

imbalanced data. In addition, comparisons have been

done between Random forest, Neural network, KNN,

Naïve Bayes, SVM, and Adaboost. The experimental

results indicate that in the first scenario, SVM

outperforms other methods in the prediction of short-

term and long-term readmission with an accuracy of

64 % but NB achieved a better AUROC 0.65 in both

cases with and without attribute selection. Also, In the

second scenario, the prediction of early readmission

with Random forest outperforms other methods with

an accuracy of 86,38 % and an AUROC of 0.63 in

both experiments with and without attribute selection.

CCH 2023 - Special Session on Machine Learning and Deep Learning for Preventive Healthcare and Clinical Decision Support

686

In this study, uncontrolled diabetic patients are

targeted; nevertheless, we expect that this early study

will pave the way for future research that can improve

the accuracy of readmission risk estimates for other

health conditions like heart and kidney diseases. Also,

an improved data set, including other important

features such as age, weight, and laboratory values,

could prove valuable and warrant further study.

REFERENCES

Alamer, A. A., Patanwala, A. E., Aldayyen, A. M., & Fazel,

M. T. (2019). Validation and comparison of two 30-day

re-admission prediction models in patients with

diabetes. Endocrine Practice, 25(11), 1151-1157.

Al-Ars, Z. T., & Aldabbagh, A. M. (2021). Predicting the

Early Re-admission of Diabetic Patients Using

Different Data Mining Techinques. In 2021 Fourth

International Conference on Electrical, Computer and

Communication Technologies (ICECCT) (pp. 1-8).

IEEE.

Ali, F., El-Sappagh, S., Islam, S. R., Kwak, D., Ali, A.,

Imran, M., & Kwak, K. S. (2020). A smart healthcare

monitoring system for heart disease prediction based on

ensemble deep learning and feature fusion. Information

Fusion, 63, 208-222.

Battineni, G., Sagaro, G. G., Chinatalapudi, N., & Amenta,

F. (2020). Applications of machine learning predictive

models in the chronic disease diagnosis. Journal of

personalized medicine, 10(2), 21.

Bhuvan, M. S., Kumar, A., Zafar, A., & Kishore, V. (2016).

Identifying diabetic patients with high risk of

readmission. arXiv preprint arXiv:1602.04257.

Boutayeb, A., Twizell, E. H., Achouayb, K., & Chetouani,

A. (2004). A mathematical model for the burden of

diabetes and its complications. Biomedical engineering

online, 3(1), 1-8.

Dagliati, A., Marini, S., Sacchi, L., Cogni, G., Teliti, M.,

Tibollo, V., & Bellazzi, R. (2018). Machine learning

methods to predict diabetes complications. Journal of

diabetes science and technology, 12(2), 295-302.

Donsa, K., Spat, S., Beck, P., Pieber, T. R., & Holzinger, A.

(2015). Towards personalization of diabetes therapy

using computerized decision support and machine

learning: some open problems and challenges. In Smart

Health (pp. 237-260). Springer, Cham.

Dua, D. and Graff, C. (2019). UCI Machine Learning

Repository [http://archive.ics.uci.edu/ml]. Irvine, CA:

University of California, School of Information and

Computer Science

Eldarrat, A. H. (2011). Diabetic patients: their knowledge

and perception of oral health. Libyan Journal of

Medicine, 6(1), 5691.

Farajollahi, B., Mehmannavaz, M., Mehrjoo, H., Moghbeli,

F., & Sayadi, M. J. (2021). Diabetes diagnosis using

machine learning. Frontiers in Health Informatics,

10(1), 65.

Farajollahi, B., Mehmannavaz, M., Mehrjoo, H., Moghbeli,

F., & Sayadi, M. J. (2021). Predicting hospital

readmission of diabetic patients using machine

learning. Frontiers in Health Informatics, 10(1), 74.

Hu, P., Li, S., Huang, Y. A., & Hu, L. (2019, June).

Predicting hospital readmission of diabetics using deep

forest. In 2019 IEEE International Conference on

Healthcare Informatics (ICHI) (pp. 1-2). IEEE.

Kavakiotis, I., Tsave, O., Salifoglou, A., Maglaveras, N.,

Vlahavas, I., & Chouvarda, I. (2017). Machine learning

and data mining methods in diabetes research.

Computational and structural biotechnology journal,

15, 104-116.

Kim, H., Ross, J. S., Melkus, G. D., Zhao, Z., & Boockvar,

K. (2010). Scheduled and unscheduled hospital

readmissions among diabetes patients. The American

journal of managed care, 16(10), 760.

Kim, S. (2007). Burden of hospitalizations primarily due to

uncontrolled diabetes: implications of inadequate

primary health care in the United States. Diabetes care,

30(5), 1281-1282.

Kohli, P. S., & Arora, S. (2018, December). Application of

machine learning in disease prediction. In 2018 4th

International conference on computing communication

and automation (ICCCA) (pp. 1-4). IEEE.

"Kumar, P. M1. Battineni, G., Sagaro, G. G.,

Chinatalapudi, N., & Amenta, F. (2020). Applications

of machine learning predictive models in the chronic

disease diagnosis. Journal of personalized medicine,

10(2), 21.., Lokesh, S., Varatharajan, R., Babu, G. C.,

& Parthasarathy, P. (2018). Cloud and IoT based

disease prediction and diagnosis system for healthcare

using Fuzzy neural classifier. Future Generation

Computer Systems, 86, 527-534."

Navarro-Pérez, J., Orozco-Beltran, D., Gil-Guillen, V.,

Pallares, V., Valls, F., Fernandez, A., & Tellez-Plaza,

M. (2018). Mortality and cardiovascular disease burden

of uncontrolled diabetes in a registry-based cohort: the

ESCARVAL-risk study. BMC cardiovascular

disorders, 18(1), 1-9.

Neto, C., Senra, F., Leite, J., Rei, N., Rodrigues, R.,

Ferreira, D., & Machado, J. (2021). Different scenarios

for the prediction of hospital readmission of diabetic

patients. Journal of Medical Systems, 45(1), 1-9.

Pham, H. N., Chatterjee, A., Narasimhan, B., Lee, C. W.,

Jha, D. K., Wong, E. Y. F., & Chua, M. C. (2019, July).

Predicting hospital readmission patterns of diabetic

patients using ensemble model and cluster analysis. In

2019 International Conference on System Science and

Engineering (ICSSE) (pp. 273-278). IEEE.

Pujianto, U., Setiawan, A. L., Rosyid, H. A., & Salah, A.

M. M. (2019). Comparison of Naïve Bayes Algorithm

and Decision Tree C4. 5 for Hospital Readmission

Diabetes Patients using HbA1c Measurement. Knowl.

Eng. Data Sci., 2(2), 58-71.

Sharma, A., Agrawal, P., Madaan, V., & Goyal, S. (2019,

June). Prediction on diabetes patient's hospital

readmission rates. In Proceedings of the Third

International Conference on Advanced Informatics for

Computing Research (pp. 1-5).

Short-Term and Long-Term Readmission Prediction in Uncontrolled Diabetic Patients using Machine Learning Techniques

687

Sharma, A., Agrawal, P., Madaan, V., & Goyal, S. (2019,

June). Prediction on diabetes patient's hospital

readmission rates. In Proceedings of the Third

International Conference on Advanced Informatics for

Computing Research (pp. 1-5).

Strack, B., DeShazo, J. P., Gennings, C., Olmo, J. L.,

Ventura, S., Cios, K. J., & Clore, J. N. (2014). Impact

of HbA1c measurement on hospital readmission rates:

analysis of 70,000 clinical database patient records.

BioMed research international, 2014.

Threatt, J., Williamson, J. F., Huynh, K., Davis, R. M., &

Hermayer, K. (2013). Ocular disease, knowledge and

technology applications in patients with diabetes. The

American journal of the medical sciences, 345(4), 266-

270.

Urban, G., Tripathi, P., Alkayali, T., Mittal, M., Jalali, F.,

Karnes, W., & Baldi, P. (2018). Deep learning localizes

and identifies polyps in real time with 96% accuracy in

screening colonoscopy. Gastroenterology, 155(4),

1069-1078.

CCH 2023 - Special Session on Machine Learning and Deep Learning for Preventive Healthcare and Clinical Decision Support

688