# Survey on clinical prediction models for diabetes prediction

- N. Jayanthi
^{1, 3}Email authorView ORCID ID profile, - B. Vijaya Babu
^{1}and - N. Sambasiva Rao
^{2}

**Received: **25 May 2017

**Accepted: **22 June 2017

**Published: **23 August 2017

## Abstract

Predictive analytics has gained a lot of reputation in the emerging technology Big data. Predictive analytics is an advanced form of analytics. Predictive analytics goes beyond data mining. A huge amount of medical data is available today regarding the disease, their symptoms, reasons for illness, and their effects on health. But this data is not analysed properly to predict or to study a disease. The aim of this paper is to give a detailed version of predictive models from base to state-of-art, describing various types of predictive models, steps to develop a predictive model, their applications in health care in a broader way and particularly in diabetes.

## Keywords

## Introduction

Predictive analytics use statistical or machine learning method to make a prediction about future or unknown outcomes [1]. It uses text mining for unstructured data, answers the question “what is next step?” It uses historical and present data to predict future regarding activity, behaviour and trends. To do this it makes use of statistical analysis techniques, analytical queries and automated machine learning algorithms. Predictive analytics need experts to build predictive models. These models are used for prediction.

There are many applications of predictive analytics, out of which one is health care. A most common disease now a day’s is diabetes. People are suffering with it and the patient number increases day by day. The World Health Organization (WHO) predicts that by 2030 there will be approximately 350 million people worldwide affected by diabetes [2, 3]. Mostly whatever food we eat is converted into glucose or sugar. Now, this glucose or sugar is used for energy. Glucose is transported to body cells through insulin. If the body does not produce sufficient insulin or does not make proper use of insulin then it leads to diabetes.

There are four types of diabetes which are TYPE 1, TYPE 2, GESTATIONAL, PRE DIABETES. TYPE 1 diabetes is also known as insulin dependent diabetes [4] where the pancreas does not produce the hormone insulin. TYPE 2 diabetes is also known as non-insulin dependent diabetes [4] where adequate insulin is produced but the body cannot make use of insulin. Gestation diabetes is a type of diabetes which occurs during pregnancy [5]. Pre diabetes refers to a situation where blood glucose levels are higher than normal but not so high to diagnosis as diabetes [6]. Diabetes is a disease in which blindness, nerve damage, blood vessel damage, kidney disease and heart disease can be developed [7]. By the use of predictive analytics in the field of diabetes, diabetes diagnosis, diabetes prediction, diabetes self-management and diabetes prevention can be achieved as per the literature survey.

Future the paper is organized into three sections. “Related work” gives the background of predictive analytics. “Clinical prediction model” describes different predictive models used in health care particularly for diabetes, followed by the summary of predictive models and ends with a conclusion and future work.

## Related work

### Predictive analytics

### Taxonomy of predictive analytics

Difference between linear model and generalized model

S. no | Simple linear model | Generalized linear models |
---|---|---|

1 | μ = E(Y) = β0 + β1 * X1 + β2 * X2+ ··· + βn*Xn | g(μ) = β0 + β1 * X1 + β2 * X2+ ··· + βn*Xn |

2 | Target variable Y does not depend on the value of Y for any other record, only the predictors | Target variable Y does not depend on the value of Y for any other record, only the predictors |

3 | Y is normally distributed | Distribution of Y is a member of the exponential family of distributions(normal, Poisson, gamma, binomial, negative binomial, inverse Gaussian) |

4 | Mean of Y depends on the predictors, but all records have the same variance | Variance of Y is a function of the mean of Y |

5 | Y is related to predictors through simple linear function | g(μ) is linearly related to the predictors. The function g is called the link function |

### Steps to develop predictive model

It was listed that there are six steps for developing predictive models. Which are listed as follows project definition, exploration, data preparation, model building, deployment, model management [9].

Percentage of usage of prediction model in organizations

Percent of model (%) | Model purpose |
---|---|

65 | Use to guide decision and plans |

52 | To score records |

41 | Import models into BI tools or reports |

36 | Scores to create or augment rules |

33 | Embed rules or models in applications to automate or optimize processes |

### Selecting model according to situation

Depending on situation what model has to be selected is described as for segmentation use clustering algorithm, for developing recommender system use classification algorithm, Use decision tree when linear decision boundary is used, for predicting next outcome of time driven events use regression algorithms, to predict continuous values use regression, use naïve Bayes when features are conditionally independent, Machine learning is used for classifying text problems with ensemble model sometimes [11].

### Deploying the predictive model

Deployment means using a model for the intended purpose. Most predictive models are built depending on the decisions made by the organization [17]. Various ways for deploying the predictive model are sharing the model, score the model, incorporate the model in a BI report, embed the model in application [8].

### Assessment of predictive model

Acceptance of predictive model based on c-statistics

Range | Grade |
---|---|

0.9–1.0 | Excellent |

0.8–0.9 | Good |

0.7–0.8 | Acceptable |

0.6–0.7 | Poor |

0.5–0.6 | Fail |

Measures used model assessment when data is categorical

Measure | Significance |
---|---|

R square | Project the efficiency of the model in terms of independent variables |

Significance F | To check whether results are reliable or not if F less than 0.5 model is OK else stop using those independent variables |

Coefficients | Regression line Y = intercept + A * X1 + −B * X2 A and B are coefficients these are useful for forecasting |

Residuals | These are used to show how far the actual data points are from predicted data points |

- 1.Uplift from model
- a.
Compare the performance of predictive model against random results with lift charts and decline tables.

- b.
Evaluate the validity of the discovery with target shuffling.

- c.
Test predictive model consistency using bootstrap sampling.

- a.
- 2.
Use empirical measures of accuracy such as confidence levels or other statistical quantities if the aim of models is to provide highly accurate predictions or decisions.

### Applications of predictive analytics

Predictive analytics have huge applications in various fields like Homeland security, Crime prevention, Infrastructure management, Cyber security, Intelligent Transportation, Health care and bioinformatics, Text mining, Fraud detection, Social media and decision support [1], Credit scores, Credit card, fraud detection, Mail Sorting, Weather prediction, Hot dogs and hamburgers [16], not only about resource allocation but also about where or how should to allocate resource, what to expect from outcomes of model, how to manage the key drivers of the economic model for better outcome? [19]. There are four ways to monetize predictive models. The first way is for saving cost. The second way is using a predictive model for improving revenue. The third way is for returning investment and lastly is for risk management [20].

### Predictive modeling tools

Predictive modeling tools

Predictive modeling tools | ||
---|---|---|

Risk groupers | These tools are best used for | Acturial |

Underwriting | ||

Profiling perspectives | ||

Statistical models | These tools require lots of historical data | Linear regression |

Logistic regression | ||

Anova | ||

Time series | ||

Trees | ||

Non-linear regression | ||

Survival analysis | ||

Artificial intelligence models | These are new methods | Fuzzy logic |

Neural networks | ||

Genetic algorithm | ||

Nearest neighbor pairing | ||

Conjugate gradient | ||

Rule induction | ||

Principal component analysis | ||

Simulated annealing | ||

Kohonen network |

### Existing predictive models

- 1.
Optum predictive model

This model is used to predict employees of the company, who are interested to join optum health management programs.

- 2.
Netflix

This algorithm determines which movies a customer is likely to enjoy.

- 3.
Match.com

This is a behaviour modeling algorithm as it learns from the behaviour of similar users and factors and recommends those sites to new users who for the same topic on the web.

- 4.
Santa Cruz’s predictive policing program

This algorithm will analyse and detect patterns of past years crime data and predicts areas and windows of time that are high at risk.

- 5.
Harrahs casino

It predicts the pattern of gambling players based on their level of dissatisfaction and intervene.

- 6.
Fighting medicare fraud

Centers for Medicare and Medicaid Services (CMS) is poised to begin using predictive modeling technology to fight Medicare fraud. This algorithm automatic alerts and risk scores for claims.

- 7.
VISA

This algorithm predicts people behaviour.

- 8.
TCS diabetes Readmission predictive analytics model

This predictive model predicts patient readmission rates using a statistical model [21].

## Clinical prediction model

### Clinical model

- i.
Preparation for establishing clinical prediction models.

- ii.
Dataset selection.

- iii.
Handling variables.

- iv.
Model generation.

- v.
Model evaluation and validation.

Predictive analytics in health now a day

Model | Data | ||
---|---|---|---|

Size | Sources | Quality | |

Old | Limited data | Claims data Inpatient data only | Poor Unstructured Data can not be accessed |

Morden | Large data | Emr + claims + socioeconomic + care management Inpatient + outpatient + ED | Excellent Unstructured data can be accessed |

### Different prediction models used for diabetes

A multi stage adjustment model with low misclassification rate which predicts which persons are most likely to develop diabetes is built by using KoGES dataset [23]. A physiological model which can predict the blood glucose level 30 min in advance was developed using five patients data by training SVR with physiological features. This helped in producing best results than doctors [24]. Another type of predictive model is sparse factor graph model. By using which diabetes complications are not the only forecast but also can discover the underlying associations between diabetes complications and lab test types. All algorithms were implemented in C++, and all experiments were performed on a Mac running Mac OS X with Intel Core i7 2.66 GHz and 4 GB of memory. The data set used for the experiment is collected from a geriatric hospital. The data set contain 1-year span data with 181,933 medical records, 35,525 patients data and 1945 types of lab tests. 60% of data was chosen for training the model and the rest for testing. The proposed model addresses two challenges feature sparseness and knowledge skewness [25].

In paper [28] the authors used two different types of neural networks to express which will output the accurate classifier in predicting diabetes. The two neural network models are multilayer neural network and probabilistic neural network. The dataset contains Pima Indian diabetes, having two classes and 768 samples. 576 samples were used for training and 192 were used for testing. The proposed methods were proved to better when compared with other previous methods.

In paper [29] the author developed a prediction model based on Hybrid-Twin Support Vector Machine (H-TSVM), which predicts whether a new patient is suffering from diabetes or not. They used Pima dataset for conducting an experiment. The factor that keeps this proposed method different from others is kernel function. The classifier produces an accuracy of 87.46%.

In paper [30] the author proposed a predicting model that classify type 2 diabetic treatment plans into three groups such as insulin, diet and medication. The dataset used for developing the model was JABER ABN ABU ALIZ clinic centre which contains 318 medical records. The model was developed using WEKA tool by applying J48 classifier and it has produced an accuracy of 70.8%.

In paper [31] the author developed a prediction model which predicts what are different types of disease a diabetic patient can develop. To develop the model a data set of 3 years span is collected from AR hospital with 739 patient details and 31 attributes. The pre processed data after deleting outliers by using distance based outlier detection (DBOD), is given as input to logistic regression model which was built by Bipolar Sigmoid Function that is calculated using Neuro based Weight Activation function. The model produced prediction accuracy of 90.4%.

In paper [33] author developed a hybrid model KSVM. The important criteria that make this model different from other methods are feature selection algorithm. PIMA data set was utilized to do experiments and results were produced. It was shown that diagnosis results using K-SVM are 99.74, 99.78, and 99.81 for learning experiments with amount 50, 60, and 70% data respectively, and 99.82, 99.85, and 99.90 for testing experiments with amount 50, 60, and 70% data respectively.

In paper [34] authors developed a prediction model that would predict whether a person would develop diabetes by considering daily lifestyle activities. To build prediction model PIMA diabetes data set was used and CART (Classification and Regression Trees) machine learning classifier was applied. The proposed model could provide an accuracy of 75%.

In paper [36] authors developed a decision tree model for the diagnosis of type 2 diabetes. They used Pima Indian diabetes dataset. Pre-processing techniques like attributes identification and selection, handling missing values, and numerical discretization was used to improve the quality of data. Weka tool was used, J48 decision tree classifier was applied to construct the decision tree model. The model produced an accuracy of 78.17%.

In paper [38] the authors have developed an expert healthcare predictive decision support system that predicts diabetes. This model is trained on Pima diabetes dataset. Decision tree and K-nearest neighbor algorithms are used to develop the model and found that C4.5 algorithm has achieved 90.43% accuracy.

In paper [39] the authors have developed a prediction model using Chi squared test to find not only dependencies between factors but also independences. Then CART is applied to build a prediction model which has 75% accuracy. Data was collected through questioners from 200 people and model was built using R tool.

In paper [40] authors developed an elastic net model which improves the accuracy for estimating glucose. The authors have collected 45 experimental sessions data set from diabetic patients. The data was collected from a noninvasive glucose device i.e., a blood sample is not taken. Three models were constructed using regularized methods LASSO, Ridged and Elastic net model. The elastic net model has compared with LASSO, ridged and partial least square regression and found Elastic net model is best.

From all of the techniques and prediction models discussed above, we want a prediction model that predicts diabetes of a diagnosed person. Since this output can be obtained depending on the time we would lie to use regression model. Of all regressions, Elastic Net is most useful as categorical, numerical and image or signal form data can be given as input to the model. The elastic net regression model is a combination of LASSO (Least Absolute Shrinkage And Selection Operator) and Ridged Regressions. Thus elastic net regression support shrinkage of coefficients as well as grouping effect. One more interesting point is numerical, categorical and image form data can be given as input to the model.

## Summary

Summary of different prediction models used for diabetes

Paper no | Dataset | Prediction model | Technique | Tool | Outcome | Accuracy |
---|---|---|---|---|---|---|

11 | Koges | Multi stage adjustment model | Not mentioned | Not mentioned | Which person is most likely to develop diabetes | Not mentioned |

12 | Five patients data | Physiological model | Svr | Not mentioned | Predicts blood glucose level 30 min in advance | Not mentioned |

17 | Geriatric Hospital | Sparse factor graph model | Not mentioned | Not mentioned | Forecast diabetes complications and uncover underlying relationship between diabetes and lab reports | Not mentioned |

18 | Pima | Hybrid model to predict | Clustering + C4.5 | Weka | Predict whether the diagnosed patient may develop diabetes within 5 years or not | 92.38% |

19 | Pima | Hybrid prediction model | Clustering + SVM | Weka | Optimal feature subset which helps in detecting diabetes with high accuracy | 98.9247%. |

20 | Pima | Neural networks | Multilayer neural network and probabilistic neural network | Not mentioned | Output the accurate classifier in predicting diabetes | |

21 | Pima | Hybrid-twin support vector machine | Kernel functions | Not mentioned | Predicts whether a new patient is suffering from diabetes or not | 87.46%. |

22 | Jaber Abn Abu Aliz | Prediction model | J48 classifier | Weka | Classify type 2 diabetic treatment plans | 70.8%. |

23 | Ar Hospital | Logistic regression model | Bipolar sigmoid function that is calculated using neuro based weight activation function | Not mentioned | Predicts what are different types of disease a diabetic patient can develop | 90.4% |

24 | Not mentioned | Fnc model | Fuzzy logic, neural network, case based reasoning, rule based algorithm | Matlab and Mycbr plug-in | Used for diabetes diagnosing | Not mentioned |

25 | Pima | Ksvm | Feature selection algorithm | Not mentioned | Used for diabetes diagnosing | 99.82–50, 99.85–60, and 99.90–70% of data |

26 | Manual collection | Cart | Manual | Used to predict whether a person would develop diabetes or not | 75% | |

27 | Pima | Correlation analysis | Multiple regression | Manual | Predicts whether patient develops diabetes or not | 77.85% |

36 | Pima | CART | J48 | weka | Predicts whether patient develops diabetes or not | 78.17% |

37 | Manual | Neural networks | Memetic algorithm | Not mentioned | Classify and diagnose onset and progression of diabetes | 93.2% |

38 | Pima | Prediction model | C4.5 and KNN | Not mentioned | Predicts diabetes or not | 93.43% |

39 | Questioner | Prediction model | CART | R | Predicts whether a person fall into diabetic in future | 75% |

40 | Manual | Prediction model | LASSO, ridge and elastic net regressions | R | Predicts glucose level accurately | Not mentioned |

Current work | Pima | Prediction model | Elastic net regression | R | Predicts whether a person develops diabetes or not with in 6 months | To be worked out |

## Conclusions

In this paper a detail description of predictive modeling is presented, a combination of tradition and hybrid prediction models Modeling, This paper showed that hybrid models produce more accuracy than traditional models. A researcher who is willing to do research in developing clinical prediction model would be benefited by this paper. There is a wide range of scope for the development of clinical prediction models especially for diabetes as this is a modern disease in developing countries like India.

As per the survey of above papers we can find many gaps that are to be filled, which are usage of larger dataset [23, 34], outlier detection [35], improving prediction model [34], integration of optimization techniques to hybrid prediction model [33], implementation of prediction models for other diseases on android mobile [31], development of prediction model that include type 1 treatment plans with more attributes [30], usage of datasets of multiple classes [4].

## Future work

The problems that are studies in the above papers for improving accuracy for prediction and, diagnosis of diabetes would be worked out further using elastic net regression. Elastic net regression is a combination of LASSO and Ridged Regression techniques to which categorical, numeric and image form data can be given to the regression.

## Declarations

### Authors’ contributions

NJ has contributed for acquisition of data, analysis and interpretation of data, Drafting of the manuscript. VB has served as scientific advisors in study conception, design and for critical revision. SR has critically reviewed the study proposal. All authors read and approved the final manuscript.

### Acknowledgements

I would like to show our gratitude to Dr. Vasantha Vedachalam for sharing her pearls of wisdom with me during the course of this research. Dr. Subba Laxmi for providing necessary information. Dr. P.V. Rao for his comments that greatly improved the manuscript. I thank Dr. Madhu Bala for her insight and expertise that greatly assisted the research. I am also immensely grateful to Ms. B. Padmaja for her comments and interpretations on an earlier version of the manuscript, which was helpful for improvement of the manuscript.

### Competing interests

The authors declare that they have no competing interests.

### Availability of data and materials

PIMA Indian data set—UCI Repository.

### Consent for publication

Not applicable.

### Ethics approval and consent to participate

Not applicable.

### Funding

I certify that no funding has been received for the conduct of this study and/or preparation of this manuscript.

### Publisher’s Note

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

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

## Authors’ Affiliations

## References

- Brown DE, et al. Predictive analytics. Washington: IEEE Computer Society; 2015.Google Scholar
- http://www.predictiveanalyticsworld.com/patimes/intro-to-machine-learning-algorithms-for-it-professionals-0620152/5580/. Accessed 2 July 2017.
- http://www.who.int/diabetes/publications/en/screening_mnc03.pdf. Accessed 29 Mar 2017.
- Sanakal R, Jayakumari T. Prognosis of diabetes using data mining approach-fuzzy C means clustering and support vector machine. Int J Comput Trends Technol. 2014;11(2):94–8.View ArticleGoogle Scholar
- Lakshmi KR, Kumar SP. Utilization of data mining techniques for prediction of diabetes disease survivability. Int J Sci Eng Res. 2013;4(6):933–40.Google Scholar
- Repalli P. Prediction on diabetes using data mining approach. Stillwater: Oklahoma State University; 2011.Google Scholar
- Motka R, et al. Diabetes mellitus forecast using different data mining techniques. In: Computer and communication technology (ICCCT), IEEE, 4th international conference. New York: IEEE; 2013.Google Scholar
- Eckerson WW. Predictive analytics. Tdwi Research. 2006.Google Scholar
- http://data-magnum.com/types-and-uses-of-predictive-analytics-what-they-are-and-where-you-can-put-them-to-work/. Accessed 15 Apr 2017.
- https://link.springer.com/chapter/10.1057%2F9781137379283_6#page-1. Accessed 5 July 2017.
- Kalechofsky H. A simple framework for building predictive models. 2016.Google Scholar
- Tevet D, et al. Introduction to predictive modeling using GLMs a practitioner’s viewpoint.Google Scholar
- https://www.analyticsvidhya.com/blog/2015/08/comprehensive-guide-regression/. Accessed 20 Apr 2017.
- Gemson Andrew Ebenezer J. Big data analytics in healthcare: a survey. ARPN J Eng Appl Sci. 2015;10(8).Google Scholar
- http://www.dummies.com/programming/big-data/data-science/data-science-for-dummies-cheat-sheet/. Accessed 30 Mar 2017.
- Predictive modeling, Julie Chambers, the 56th annual Canadian reinsurance conference.Google Scholar
- Abbott Analytics. Strategies for building predictive models. 2014.Google Scholar
- Predictive analytics: poised to drive population health White Paper, Optum.Google Scholar
- Duncan I. Introduction to predictive modeling. 2015.Google Scholar
- https://www.linkedin.com/pulse/4-types-predictive-analytics-models-mark-rabkin. Accessed 3 July 2017.
- ]http://234w.tc.tracom.net/healthcare/Pages/Diabetes-Readmission-Predictive-Analytics.aspx. Accessed 25 Mar 2017.
- Lee YH, et al. How to establish clinical prediction models. Seoul: Korean Endocrine Society; 2016.Google Scholar
- Lee J, et al. Development of a predictive model for type 2 diabetes mellitus using genetic and clinical data. Osong Public Health Res Perspect. 2011;2(2):75–82.View ArticleGoogle Scholar
- Plis K, et al. A machine learning approach to predicting blood glucose levels for diabetes management. Association for the Advancement of Artificial Intelligence. 2014.Google Scholar
- Yang Y, et.al. Forecasting potential diabetes complications. In: Proceedings of the twenty-eighth AAAI Conference on artificial intelligence, Copyright c. Association for the Advancement of Artificial. 2014.Google Scholar
- Patil BM, et al. Hybrid prediction model for type-2 diabetic patients. Expert Syst Appl. 2010;37(12):8102–8.View ArticleGoogle Scholar
- Sarojini Ilango, B. et al. A hybrid prediction model with F-score feature selection for type ii diabetes databases. In: A2CWiC. 2010.Google Scholar
- Temurtas H., et al. A comparative study on diabetes disease diagnosis using neural networks. Expert Syst Appl. 2009;36(4):8610–5.View ArticleGoogle Scholar
- Divya et al. Predictive model for diabetic patients using hybrid twin support vector machine. In: Proc. of int. conf. on advances in communication, network, and computing, CNC. Amsterdam: Elsevier; 2014.Google Scholar
- Ahmed TM. Developing a predicted model for diabetes type 2 treatment plans by using data mining. J Theor Appl Inf Technol. 2016;90(2):181–7.Google Scholar
- Devi MN, et al. Developing a modified logistic regression model for diabetes mellitus and identifying the important factors of type II DM. Indian J Sci Technol. 2016; 9(4).Google Scholar
- Thirugnanam M, et al. Hybrid tool for diagnosis of diabetes. IIOAB J. 2016;7(5).Google Scholar
- Osman AH, et al. Diabetes disease diagnosis method based on feature extraction using K-SVM. Int J Adv Comput Sci Appl. 2017;8(1).Google Scholar
- Anand A. Prediction of diabetes based on personal lifestyle indicators. In: 2015 1st international conference on next generation computing technologies (Ngct-2015) Dehradun, India, 4–5 September 2015.Google Scholar
- Jakhmola S. A computational approach of data smoothening and prediction of diabetes dataset. New York City: ACM; 2015.View ArticleGoogle Scholar
- AlJarullah AA. Decision tree discovery for the diagnosis of type II diabetes. In: International conference on innovations in information technology. New York: IEEE; 2011.Google Scholar
- Jahani Meysam, Mahdavi Mahdi. Comparison of predictive models for the early diagnosis of diabetes. Healthc Inform Res. 2016;22(2):95–100.View ArticleGoogle Scholar
- Hashi EK, et al. An expert clinical decision support system to predict disease using classification techniques. In: International conference on electrical, computer and communication engineering (ECCE), ©2017 IEEE, February 16–18, 2017, Cox’s Bazar, Bangladesh.Google Scholar
- Anand A, Shakti D. Prediction of diabetes based on personal lifestyle indicators. In: Next generation computing technologies (NGCT), 2015 1st international conference on 4–5 Sept. New York: IEEE; 2015.Google Scholar
- Zanon M, et al. Regularised model identification improves accuracy of multisensor systems for noninvasive continuous glucose monitoring in diabetes management. J Appl Math. 2013;2013.Google Scholar