Research on the Correlation Between Multiple Risk Factors and

Diabetes Mellitus

Yuzhen Li

1, *

and Haoze Song

College of Arts and Sciences, The Ohio State University, Columbus, 43210, U.S.A.

School of Science, China University of Mining and Technology-Beijing, Beijing, 100083, China

Keywords: Diabetes Mellitus, Risk Factors, Binary Probit Model.

Abstract: Preventing diabetes is one of the main areas of current study. Numerous studies have revealed a strong

correlation between smoking, age, obesity, and the prevalence of diabetes; nonetheless, a great deal more

study has to be done in this area. Therefore, in this paper, the correlation between multiple risk factors and

diabetes was investigated by collecting data on the physical status of patients at Sylhet Diabetes Hospital and

performing a binary Probit analysis on the data. Polyuria, polydipsia, polyphagia, genital thrush, irritability,

and partial paresis were found to be strong risk factors for developing diabetes, while sudden weight loss,

weakness, visual blurring, delayed healing, muscle stiffness, alopecia, and obesity were found to be less

influential in causing diabetes. The study enables greater attention to be paid to the risk factors associated

with diabetes, facilitates early detection of diabetes and early initiation of treatment to improve patient

prognosis and survival, and points to a clear direction for future related research.

1 INTRODUCTION

The history of diabetes can be dated back to several

millennia ago, especially in ancient Egypt, Greece,

and Asia. Under biological settings, the kidney

maintains fluid and electrolyte homeostasis by

modifying the amount and chemical makeup of urine

in accordance with the body's requirements. A

complicated and varied clinical disease known as

diabetes affects the distribution of water and is

typified by persistent diuresis, which produces huge

amounts of diluted urine (Valenti and Tamma, 2016).

In recent years, the global incidence of diabetes

has climbed dramatically. Diabetes affects 529

million people globally, a figure that is anticipated to

rise to 1.3 billion by 2050 (Liu, 2023). Drawing on

information from the "Global Burden of Disease

2021" research, the freshest and most extensive

assessments to date indicate that diabetes now affects

6.1% of the global population. This positions diabetes

as a leading contributor to mortality and disability

worldwide, ranking it among the top ten causes. The

findings underscore the significant and growing

impact of diabetes on global health, necessitating

urgent and sustained efforts to address this escalating

public health challenge (Liu, 2023). By 2050, it is

anticipated that the prevalence of obesity in North

Africa and the Middle East will skyrocket from 9.3

percent to a staggering 16.8 percent. In a similar vein,

projections indicate that the occurrence rate in Latin

America and the Caribbean is expected to rise

significantly, reaching 11.3 percent (Liu, 2023).

Research conducted in the United States and Europe

has demonstrated a 3-5% annual increase in the

incidence of type 1 diabetes (T1D) throughout time

(Maffi and Secchi, 2017). Diabetes is notably

common among individuals aged 65 and above,

affecting over 20% of this demographic worldwide.

The occurrence is particularly pronounced in those

between 75 and 79 years old (Liu, 2023). To prevent

a rise in the incidence of diabetes, it is essential to

implement measures aimed at assessing the degree of

correlation between diabetes and various risk factors.

Comprehensive studies should be conducted to

thoroughly investigate how these risk variables

interact with the development and progression of

diabetes, thereby enabling the formulation of

effective prevention strategies.

The development of diabetes mellitus involves

multiple pathogenic processes, and the etiologic

classification is very complicated, making it difficult

to diagnose individuals with risk factors (Li et al.,

2024). National and international scholars have found

smoking, age, and obesity to be factored in the onset

Li, Y. and Song, H.

Research on the Correlation Between Multiple Risk Factors and Diabetes Mellitus.

DOI: 10.5220/0012994100004601

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Innovations in Applied Mathematics, Physics and Astronomy (IAMPA 2024), pages 78-85

ISBN: 978-989-758-722-1

and development of diabetes (Li et al., 2019; Gong et

al., 2017 & Qiu et al., 2016). In addition, God et al.

found that other lifestyle factors also influence the

incidence of diabetes (Gode et al., 2024). However,

the risk factors of diabetes mellitus in this literature

failed to be adequately studied. In addition, the

statistically obtained data are not systematic and

complete. As a result, this study concentrates on

examining a set of 16 distinct variables, which

include Gender, Age, Polyuria, Polydipsia, Sudden

weight loss, Weakness, Polyphagia, Genital thrush,

Visual blurring, Itching, Irritability, Delayed healing,

Partial paresis, Muscle stiffness, Alopecia, and

Obesity. The main goal of this research is to discern

and utilize the most appropriate statistical model to

evaluate the degree of association between the

identified risk factors and the prevalence of diabetes.

Through this analysis, the study aspires to offer a

thorough comprehension of the individual

contribution of each factor to the probability of

developing diabetes. This understanding is critical for

the formulation of more precise and impactful

prevention and intervention strategies. By elucidating

the specific roles these risk factors play, the research

aims to enhance the effectiveness of public health

initiatives and clinical practices aimed at mitigating

the incidence of diabetes.

In a similar direction, Xue et al. used a variant of

the Cox proportional risk model (Xue et al., 2022),

but due to data limitations, they did not include

individuals aged 35-45 years, whereas diabetes is also

highly prevalent in the 35-45 year age group, so this

may have affected the generalizability of the results.

In addition, the study did not use biomarkers (e.g.,

blood glucose levels or HbA1c) to confirm the

diagnosis of diabetes but relied on self-reporting and

medication use, which may have affected the

accuracy of the results. Ampeir et al. used a logistic

regression model (Ampeire, Kawugeze and Mulogo,

2023), which provided either an Odds Ratio (OR) or

an Adjusted Odds Ratio (AOR), which allows

researchers to quantify the extent to which each

predictor variable affects prediabetes risk.

Nevertheless, it should be noted that the model

presupposes a linear association between the

dependent variable, which in this case is prediabetes,

and the set of independent variables. This assumption

implies that changes in the independent variables are

directly proportional to changes in the likelihood of

prediabetes, simplifying the complexity of potential

nonlinear interactions that might exist in reality. If the

actual relationship is nonlinear, the model may not

accurately capture this relationship, leading to

inaccurate predictions and interpretations. Narjes

Hazar et al. constructed a Dersimonian and Laird

random effects model (Hazar et al., 2024). The model

can handle the heterogeneity that exists between

studies. However, the estimation of heterogeneity is

sensitive to the distribution of the data and the size of

the study. If the sample size of some studies is

extremely large or small, it may have an unbalanced

effect on the overall heterogeneity estimate.

By emphasizing the most important risk variables

that ought to be the focus of intervention, the results

of this study are anticipated to guide the practice of

medicine, influence public health efforts, and offer

novel perspectives on early detection of diabetes

tactics. Furthermore, the discovery of neglected

variables can provide avenues for further

investigation, which would ultimately lead to a more

thorough comprehension of the genesis of diabetes.

2 METHODOLOGY

2.1 Data Sources

The dataset utilized for this research has been sourced

from the repository available on Kaggle's platform.

The data for this study was gathered through the

administration of direct questionnaires to patients

receiving care at Sylhet Diabetes Hospital, located in

Sylhet, Bangladesh. The collected information was

subsequently reviewed and authorized by a medical

professional to ensure its accuracy and reliability.

2.2 Variable Selection

This study utilizes data from a cohort comprising 520

individuals, including both diabetic and non-diabetic

patients. Within this population, there are 328 males

and 192 females. The age distribution of these

subjects spans from 16 to 90 years. The data contains

16 variables (Gender, Age, Polyuria, Polydipsia,

Sudden weight loss, Weakness, Polyphagia, Genital

thrush, Visual blurring, Itching, Irritability, Delayed

healing, Partial paresis, Muscle stiffness, Alopecia,

Obesity).

Research on the Correlation Between Multiple Risk Factors and Diabetes Mellitus

Table 1: Symbols and numerical representations of the

sixteen variables.

Elements Ideogram Number1 Diabetes1

Gender



520 320

Age



520 320

Polyuria



258 243

Polydipsia



233 225

Sudden weight

loss



217 188

Weakness



305 218

Polyphagia



237 189

Genital thrush



116 83

Visual blurring



233 175

Itching



253 154

Irritability



126 110

Delayed healing



239 153

Partial paresis



224 192

Muscle stiffness



195 135

Alopecia



179 78

Obesity



88 61

*Number 1: The prevalence of individuals afflicted by the

disease.

**Diabetes1: The prevalence of individuals diagnosed with

diabetes mellitus who are experiencing the condition.

Table 1 presents the statistics of the general

population alongside the count of individuals

diagnosed with diabetes mellitus. The data sample is

520 individuals, 320 of whom have diabetes.

Table 2 provides a detailed breakdown of the

population, categorizing individuals by gender-

females and males-as well as by diabetic status across

various age brackets. The majority of the dataset

consists of individuals aged between 31 and 60 years.

2.3 Methodology Introduction

This study employs a binary Probit regression

analysis model to investigate the incidence of

diabetes, designated as the dependent variable (Y), in

relation to 16 independent variables (X). For the

dependent variable, a value of 0 represents the

absence of diabetes, while a value of 1 signifies its

presence. Subsequently, the influence of each of the

16 independent variables on the likelihood of having

diabetes is examined. This analysis is conducted

using SPSSAU software to facilitate the prediction of

early-stage diabetes risk. By comprehensively

assessing the relationship between these factors and

diabetes, this research aims to provide valuable

insights for early diagnosis and intervention

strategies.

2.4 Model Testing

The accuracy of the model's predictions serves as an

indicator of how well the model fits the data.

According to the data presented in Table 3, it is clear

that the research model has achieved an overall

prediction accuracy of 93.27%. This high level of

accuracy indicates that the model's performance is

within acceptable parameters. When we delve deeper

into the specifics, we find that the model exhibits a

prediction accuracy of 91.50% when the actual value

is 0. On the other hand, the accuracy improves to

94.38% when the actual value is 1. These figures

collectively highlight the model's strong performance

across various conditions, thereby validating its

reliability and effectiveness for predictive purposes.

The consistency in accuracy across different actual

values underscores the robustness of the model,

making it a dependable tool for predictive analytics.

Such performance metrics not only demonstrate the

model's capability to generalize well across different

scenarios but also affirm its potential application in

real-world predictive tasks.

Table 2: Gender distribution by age group.

Age [11-20] [21-30] [31-40] [41-50] [51-60] [61-70] [71-80]

[81-90]

Number2 1 44 120 145 127 66 10

Female 0 13 59 55 39 22 0

Male 1 31 61 90 88 44 10

Diabetes2 0 34 69 94 80 36 7

*Number 2: The population distribution across different age brackets.

**Diabetes 2: The number of individuals across various age categories diagnosed with diabetes mellitus.

IAMPA 2024 - International Conference on Innovations in Applied Mathematics, Physics and Astronomy

Table 3: Overview of Prediction Accuracy in Binary Probit Regression.

Forecast value

Prediction accuracy Prediction error rate

0 1

Actual value

0 183 17 91.50% 8.50%

1 18 302 94.38% 5.63%

Summary 93.27% 6.73%

Figure 1: Factors of the prevalence of diabetes in the population.

3 RESULTS AND DISCUSSION

3.1 Descriptive Analysis

Figure 1 presents the percentage of individuals

diagnosed with diabetes and enumerates several

contributing factors associated with the disease.

These contributing factors include Polyuria,

Polydipsia, Sudden weight loss, Weakness,

Polyphagia, Genital thrush, Visual blurring, Itching,

Irritability, Delayed healing, Partial paresis, Muscle

stiffness, Alopecia, and Obesity. The dataset

comprises a total of 520 samples. To determine the

probability of each potential risk factor leading to

diabetes, the percentage of diabetic samples identified

and categorized based on various influencing

variables was meticulously calculated. This analysis

provides a comprehensive understanding of the

correlation between each risk factor and the

likelihood of developing diabetes, thereby offering

valuable insights into the etiology of the disease.

It is evident from Figure 1 above that Polyuria

makes up the largest percentage, reaching up to 76%.

With 19 percent, obesity had the lowest share. The

data in the figure indicates that polyuria is the most

prevalent trigger, which raises the biggest concealed

risk of diabetes.

The analysis presented in the comparison chart

reveals a pronounced disparity between the

proportion of individuals diagnosed with diabetes and

those without the condition. This discrepancy is

particularly striking, with the most substantial

difference being as high as 68.5%. The data

underscore the significant variation in diabetes

prevalence, highlighting the need for further

investigation into the factors contributing to this

considerable gap (Figure 2).

Research on the Correlation Between Multiple Risk Factors and Diabetes Mellitus

Figure 2: Comparative analysis of diabetic and non-diabetic individuals across various factors.

Figure 3: Variable-related schematic.

As illustrated in Figure 3, the study incorporated a

range of variables potentially associated with diabetes

into the analytical model. These variables encompass

a comprehensive list of factors such as Gender, Age,

Polyuria, Polydipsia, Sudden weight loss, Weakness,

Polyphagia, Genital thrush, Visual blurring, Itching,

Irritability, Delayed healing, Partial paresis, Muscle

stiffness, Alopecia, and Obesity. By conducting a

series of computations, this research ultimately

derives the concluding linear regression formula:

𝑃robit





 4.079  0.028x



 2.349x





... 0.119x



(1)

In this context, p denotes the likelihood that the

variable for diabetes equals 1.

3.2 Binary Probit Model Results

Table 4 clearly illustrates that the model assesses

whether the p-value exceeds the threshold of 0.05.

This evaluation enables the study to delve deeper into

identifying potential risk factors associated with

diabetes mellitus. Through this detailed analysis, we

IAMPA 2024 - International Conference on Innovations in Applied Mathematics, Physics and Astronomy

can infer connections that might otherwise remain

undetected, thereby contributing valuable insights to

the ongoing research on diabetes risk factors. On the

one hand, diabetes is unaffected by the p values of

abrupt weight loss, weakness, blurred vision, delayed

healing, muscular stiffness, baldness, and obesity, all

of which are bigger than 0.05. Conversely, the p

values for partial paresis, age, and polyphagia fall

within the range of greater than 0.01 but less than

0.05. This statistical outcome implies that these

variables exert a notable influence on the

development or progression of diabetes. Specifically,

the p values associated with age, polyphagia, and

partial paresis, while exceeding 0.01, still remain

below 0.05, thereby underscoring their significant

contribution to the condition.

This statistical evidence highlights the meaningful

impact of these factors, suggesting that they play an

important role in the pathophysiology of diabetes.

Such findings are critical for understanding the

multifaceted nature of diabetes and underscore the

importance of considering these variables in both

clinical and research settings. However, the p values

for gender, polyuria, polydipsia, genital thrush,

itching, and irritability are less than 0.01, indicating

that these factors have an especially significant

association with diabetes. In terms of influence

relationships, the regression coefficient and marginal

effect values reveal the specific factors impacting

diabetes. For each unit increase in Polyuria,

Polydipsia, Polyphagia, Genital Thrush, Irritability,

and Partial Paresis, the incidence of diabetes is

projected to rise by 852.30%, 827.08%, 223.65%,

359.85%, 428.05%, and 230.99%, respectively.

Given that each of these statistics is significant when

compared to one another, it is evident that these

elements play a crucial role in influencing the

likelihood of diabetes onset. This suggests that the

data have some research and a value of reference.

Table 4: Results overview from the binary Probit regression analysis.

Ideogram regression coefficient SE z-value p-value 95% CI marginal effect



-0.028 0.013 -2.082 0.037 -0.054 ~ -0.002 -0.003



-2.349 0.301 -7.791 0.000 -2.940 ~ -1.758 -0.217



2.380 0.353 6.748 0.000 1.689 ~ 3.072 0.220



2.792 0.426 6.547 0.000 1.956 ~ 3.628 0.258



0.188 0.289 0.651 0.515 -0.378 ~ 0.754 0.017



0.447 0.282 1.584 0.113 -0.106 ~ 1.001 0.041



0.630 0.287 2.195 0.028 0.067 ~ 1.192 0.058



1.049 0.308 3.410 0.001 0.446 ~ 1.653 0.097



0.479 0.351 1.366 0.172 -0.208 ~ 1.167 0.044



-1.521 0.349 -4.354 0.000 -2.205 ~ -0.836 -0.141



1.270 0.319 3.984 0.000 0.645 ~ 1.895 0.117



-0.226 0.296 -0.766 0.444 -0.806 ~ 0.353 -0.021



0.638 0.283 2.258 0.024 0.084 ~ 1.193 0.059



-0.346 0.301 -1.150 0.250 -0.935 ~ 0.244 -0.032



0.178 0.319 0.559 0.576 -0.447~0.804 0.016



-0.119 0.307 -0.389 0.697 -0.721~0.482 -0.011

Constant -4.079 0.915 -4.456 0.000 -5.874~-2.285 -

Research on the Correlation Between Multiple Risk Factors and Diabetes Mellitus

3.3 Comparison with Existing

Literature

In several earlier studies, researchers frequently

employed a narrow examination of a single

component, such as genetic inheritance, which was

restricted to established risk factors for diabetes

development. Additionally, several scholars suggest

that the increased incidence of diabetes may be linked

to the adoption of a Western lifestyle. This assertion

is supported by migration studies, which highlight the

transition from conventional agricultural practices to

contemporary American living as a contributing

factor (Wu et al., 2022). On the other hand, the factors

included in this study are more thorough, which can

help prevent errors that arise from failing to account

for a single variable. It can also expand the ideas of

forthcoming studies on diabetes, assisting medical

professionals in identifying additional treatment

avenues and early diabetes detection for prompt

treatment.

4 CONCLUSION

Based on these experimental results, the main

conclusion can be summarized as follows. 16

variables may lead to diabetes, in this study, a binary

Probit regression analysis model is used and indicates

that the most significant risk factors identified by

these analyses and the data in the table are polyuria,

polydipsia, polyphagia, genital thrush, irritability,

and partial paresis. Sudden weight loss, weakness,

visual blurring, delayed healing, muscle stiffness,

alopecia, and obesity have a weak impact on causing

diabetes.

However, it should be noted that this study has

examined only some people from 11 to 90, which

means that this conclusion may not be used to study

the risk factors for children who are less than 11 years

old and infants. And it is undeniable that this model

may have mistakes in addition to the components

because of the small amount of data, and that the

sample did not include all ethnicities, which might

have resulted in differences that could have also

affected the accuracy of the findings.

Notwithstanding its limitations, this research reported

here would seem to indicate some valuable aspects.

Firstly, the analysis indicates that Polyuria is the most

common trigger and poses the highest hidden risk for

diabetes, as demonstrated through a graphical method

used to visually compare the proportions of each

factor among individuals. Further research is required

to prevent Polyuria from developing into diabetes,

and more drugs are being developed to slow the

disease's progression to provide a complete picture of

the diabetes treatment. Secondly, researchers create a

representation of the experiment and make the results

easier to comprehend and intuitive by comparing

relative data, which includes those with diabetes and

those without the disease. These tables enable people

to pay more attention and emphasize the factors

related to diabetes. The results of this study will guide

future relevant research since it is necessary to

determine whether these characteristics are linked to

the development of diabetes. By identifying new

causal components in addition to previously known

causes, the prognosis and survival of patients can be

improved through early detection of diabetes and

prompt initiation of therapy.

AUTHORS CONTRIBUTION

All the authors contributed equally and their names

were listed in alphabetical order.

REFERENCES

Valenti G and Tamma G 2016 History of Diabetes

Insipidus. Giornale italiano di nefrologia (Aldo Moro:

Biotechnologies and Biopharmaceutics University of

Bari).

Xia Liu 2023 By 2050 there will be 1.3 billion people with

diabetes worldwide. Science and Technology Daily.

3465.

Maffi P and Secchi A 2017 The Burden of Diabetes:

Emerging Data. Management of Diabetic Retinopathy.

60 1-5.

Li T, Sun Y Y, Li X L and Dong H 2024 Auxiliary

diagnosis of diabetes mellitus based on machine

learning classification algorithm. Computer Knowledge

and Technology 20 27-29.

Li H Q, Liu L J and Xu Y C 2019 Effects of onset age on

islet function and related indicators of newly diagnosed

type 2 diabetes mellitus. Public Health and Preventive

Medicine 30 134-137.

Gong H Y, Liu X F, He Y, Wang J Y and Deng Y 2017 The

interaction of obesity and smoking on diabetes mellitus.

Chronic Disease Prevention and Control in China. 25

592-594.

Qiu J Y Z, Hou X H and Jia W P 2016 Research progress

on the correlation between smoking and diabetes

mellitus. Journal of Shanghai Jiao Tong University

(Medical Science Edition) 36 110-114.

Gode Y, Patond S, Wankhade V, Ghodki S, Jadhav D,

Dhawade M R and Wankhade Y 2024 Study of Impact

of Lifestyle Modification on Diabetes and Prediabetes

IAMPA 2024 - International Conference on Innovations in Applied Mathematics, Physics and Astronomy

in an Urban Population. E3S Web of Conferences 491

3002.

Xue L, Wang H, He Y, Sui M, Li H, Mei L and Ying X

2022 Incidence and risk factors of diabetes mellitus in

the Chinese population: a dynamic cohort study. BMJ

Open 12.

Ampeire I P, Kawugezi P C and Mulogo E M 2023

Prevalence of prediabetes and associated factors among

community members in rural Isingiro district. BMC

Public Health 23 958.

Hazar N, Jokar M, Namavari N, Hosseini S and Rahmanian

V 2024 An updated systematic review and Meta-

analysis of the prevalence of type 2 diabetes in Iran,

1996-2023. Frontiers in Public Health. 12.

Wu Y L, Yu Y W, Zhou J, Wang Y Y, Yu L S, Zhang J and

Liu T 2022 Prospective study on the relationship

between healthy lifestyle and diabetes in prediabetic

population. Modern Preventive Medicine 49 1350-

1355.

Research on the Correlation Between Multiple Risk Factors and Diabetes Mellitus