Analyzing Trends and Determinants of Leading Causes of Death in the USA: A
Data-Driven Approach
Saddam Hossain
1
, Mohammed Nazmul Islam Miah
2
, MD Sohel Rana
, Md Sazzad Hossain
4
,
Proshanta Kumar Bhowmik
, Md Khalilor Rahman
6
and Rabeya akter
7
1
Master of Public Administration, Gannon University, Erie, PA, USA
2
Department of Management Sciences and Quantitative Methods, Gannon University, Erie, PA. USA
3
Executive Ph.D. in Business Analyst, University of Cumberlands, Williamsburg, KY, USA
46
MBA in Business Analytics, Gannon University, Erie, PA, USA
5
Department of Business Analytics, Trine University, Angola, IN, USA
7
Master of science in information technology. Washington University of Science and Technology,
Alexandria, VA, USA
Corresponding Author:
Saddam Hossain, E-mail: hossain024@gannon.edu
Abstract
The exponential escalation of the causes of death and their trends and determinants in the
nation greatly define the health landscape of the United States. These causes of death, such as heart
disease, cancer, chronic lower respiratory diseases, HIV &AIDS, accidents, and stroke, have been
major public health concerns for many decades. Each condition represents broader societal and
individual health challenges that include lifestyle choices, environmental factors, genetic
predispositions, and healthcare accessibility. This research project aimed to use the data-driven
approach in the exploration of these trends to understand the patterns and determinants
underpinning mortality statistics. Using an expanded data set, the study presented leading causes
of death; the pattern of variation by demographic factors, including age, sex, and race/ethnicity;
and social, environmental, and behavioral determinants of those patterns. The datasets for our
research project were retrieved from the Kaggle website, namely, "NCHS - Leading Causes of
Death: United States" which was very informative regarding the major causes of death in the
United States between the years 1999 and 2016. It was organized in such a way that one can
analyze the trends; hence, it includes variables such as Cause of Death, such as heart disease and
cancer, Year, State, Age-adjusted Death Rate, and Number of Deaths. Other demographic
variables, like Sex and Race/Ethnicity, further allowed for even finer subgroups, which were very
useful in highlighting disparities in health outcomes. The performances of the three machine
learning models, Linear Regression, Random Forest, and XG-Boost, based on Mean Squared Error
(MSE) and R-squared (R2) were evaluated. Retrospectively, XG-Boost outperformed the other
models significantly for both MSE and R2. This therefore means that on this dataset, XG-Boost is
the best model that can be used for the most accurate and reliable prediction. In that respect,
advanced machine learning models, applied to mortality trends, provide deep insight into the
underlying determinants. Large datasets comprising demographic, socioeconomic, and health-
related variables are analyzed for patterns and correlations that may not be obvious in traditional
statistical methods. Model predictions can indicate future trends in mortality by highlighting
populations at high risk and locations. Data-driven models hold monumental implications in public
health through the provision of insights into the trends and determinants of mortality, besides
including possible interventions.
Key Words:
Mortality determinants; Public health trends; Leading causes of death; Health
disparities; USA demographics; Public health policy; Data-driven analysis
INTRODUCTION
Background: Overview of the leading causes of death in the USA
According to Hider et al. (2024), the leading causes of death and their trends and
determinants in the nation greatly define the health landscape of the United States. These causes
of death, such as heart disease, cancer, chronic lower respiratory diseases, HIV &AIDS, accidents,
and stroke, have been major public health concerns for many decades. Each condition represents
broader societal and individual health challenges that include lifestyle choices, environmental
factors, genetic predispositions, and healthcare accessibility. Rahman et al. (2023), argued that the
emerging conditions of opioid overdoses, COVID-19, and mental health-related mortality have
replaced the traditional leading causes over time. A dynamic pattern such as this brings out the
changing face of challenges to public health in modern society. With every advance in medicine
and technology that shifts the human lifespan, increasing depth into what undergirds mortality will
go a significant way toward addressing preventable deaths and improving the overall health of this
nation (Zandt, 2024).
This research project uses a data-driven approach in the exploration of these trends to
understand the patterns and determinants underpinning mortality statistics. Using an expanded data
set, the study presents leading causes of death; the pattern of variation by demographic factors,
including age, sex, and race/ethnicity; and social, environmental, and behavioral determinants of
those patterns. The findings presented in this report have important implications for policymakers,
clinicians, and public health practitioners to address risks, optimize resource allocation, and design
specific interventions. This research underlines the critical role that data analytics plays in shaping
public health strategies to address health disparities and eventually contribute to improving
population health outcomes in the USA.
Problem Statement:
Analysis of trends and determinants of leading causes of death is of utmost importance
when developing effective public health strategies. Although there has been significant medical
research and public health interventions, mortality rates still show disparities among the
population. These disparities also highlight the impact of socioeconomic status, race/ethnicity,
geographic location, and access to healthcare services (Islam et al., 2024). Without an in-depth
examination of these drivers, public health policies risk being ineffectively targeted or even
increasing the present inequities. Understanding mortality trends further underlines how external
events- pandemics or economic changes suddenly change health outcomes. This kind of analysis
is not only crucial for predicting future health challenges but also for formulating tailored,
evidence-based solutions that address the root causes of health inequities (Al Amin et al., 2024).
Research Questions
RQ
1
: What are the trends in leading causes of death in the USA?
This research question aims to examine how the ranking and prevailing causes of death
have changed over time. The trends in mortality indicate the implementation and effectiveness of
health initiatives, new emerging health dangers, and changing disease burden, given that a decline
in causes due to cardiovascular diseases should demonstrate treatment successes, whereas rise and
inclining rates of diabetes-related as well as obesity-related life-threatening diseases could be
symptoms that problems with public health exist far outside individual conditions.
RQ
2
: What are the determinants of leading causes of death in the USA?
Understanding determinants involves an investigation of various factors that go into
making up mortality; these could be individual behaviors of smoking or diet, system issues of
access to care, and environmental conditions in the air. The examination also looks at genetic
predispositions interacting with these external factors and provides a holistic view of why certain
populations are more at risk of specific causes of death.
RQ
3
: How do these trends and determinants vary by demographic factors, such as age, sex,
and race/ethnicity?
This research question aims to pinpoint the need for an increased attempt to delineate how
complex demographic-natured interrelationships influence general mortality. While death rates
vary with age group-perhaps uncovering life-stage vulnerability-sex differences might illustrate
either gender-identity-based health behavior or biological predilections. Lastly, race-ethnic
disparities often emanate from structural injustices arising out of history and perhaps warrant the
identification of systemic obstacles faced by disparate populations.
Significance of the Study
The findings of this study have important implications for public health policy and practice.
By identifying modifiable causes of mortality and populations at increased risk, this research can
help target resources and intervention strategies. For instance, community-based interventions to
reduce cardiovascular disease and policies targeting social causes such as education and income
inequality may have long-term benefits for health. More importantly, the demographic weight in
the research speaks volumes of equity-based interventions in public health. In this increasingly
decision-making moment that is based on evidence, this analysis sets a platform on which to base
policy formulation-both effective and inclusive in their design to reduce mortality rates and health
inequity across the United States.
LITERATURE REVIEW
Overview of Leading Causes of Death in the USA
Bhomik et al. (2024), reported that the leading causes of death in the United States for the
past several years have been diverse including occurrences such as heart disease, cancer, and then
accidents (unintentional injuries). According to the Centers for Disease Control and Prevention, in
the year 2022, heart disease was the number one cause of death, accounting for approximately
696,000 deaths. Cancer came in second, with about 602,000 deaths. Accidents that include drug
overdoses and motor vehicle accidents took the lives of around 200,000 individuals and are a
public health concern as they significantly increased within the last decade due to the opioid crisis,
among other major causes (Dutta et al., 2024). The list goes on to include chronic lower respiratory
diseases, stroke, Alzheimer's disease, diabetes, influenza, pneumonia, kidney disease, and suicide.
The emerging causes, like COVID-19 in the years of its peak, briefly disrupted these trends and
underlined how infectious diseases can affect overall mortality patterns. As the acute phase of the
pandemic subsided, traditional chronic illnesses regained their dominance as causes of death
(Hossain et al., 2024).
Trends and Determinants of Leading Causes of Death
Nasiruddin et al. (2024), examined the trends and determinants of leading causes of death
in the United States. Most of these studies emphasized the same issues: behavioral risk factors like
smoking, poor diet, and physical inactivity, and socioeconomic factors such as education and
income play a major role in shaping mortality patterns. For example, it has been seen that people
from low socio-economic groups are likely to die prematurely from heart diseases, cancer, and
other chronic diseases. In addition, it has been reported that inequities in health care and health
insurance contribute to poor health and higher mortality rates among the most disadvantaged
groups.
Research by Bhomik et al. (2024), demonstrated that there are subtle changes in mortality
trends. Due to advances in medical care, prevention strategies, and public awareness, the death
rates for heart diseases have declined linearly through the 2000s. Recently, however, it has started
to see a partial trend reversal, with associated increased rates of obesity, sedentary lifestyle, and
access disparities in healthcare. Cancer mortality rates have consistently fallen due to early
detection techniques, improved treatment options, and a decrease in the prevalence of smoking.
Accidental deaths have risen, however, with the opioid epidemic driving the trend. In 2022, drug
overdoses accounted for 31 deaths per 100,000 people, up dramatically from the early 2000s.
Motor vehicle deaths have also risen slightly, reflecting behavioral and infrastructural challenges
(Alam et al., 2024).
Demographic Factors Affecting Mortality Trends
Hossain et al. (2024), posited trends in mortality have proven to be highly divergent
between different demographic groups. Considering demographic composition, heart disease and
cancer are highly linked to the older populations, while accidental deaths including drug overdose
present a different picture, being highest among all groups for the 18–44-year-olds. Further, racial
and ethnic variation dominates, with more cases of heart disease for Black Americans, while deaths
due to unintentional injuries are highly recorded among White and Native American populations.
Socioeconomic status, access to health care, and place to mark variations in mortality. Besides,
gender is also a significant factor because men usually have higher rates of mortality from heart
diseases and accidents while women would usually die from certain types of cancers, such as those
from the breasts. Moreover, a study by Bortty et al. (2024), found that citizens with lower
educational levels confronted higher mortality risks than those with higher education levels. All
these demographic factors interact in a manner that health disparities in the U.S. become so
entangled that an improvement in the living standards of blacks could reduce the gap in overall
mortality between blacks and whites.
Methodological Approaches in Existing Research
Ahsan & Siddique (2022), contended that different methodologies exist in the available
literature to analyze these trends. Epidemiological studies employ temporal data from the CDC's
NVSS and BRFSS to explore temporal patterns and determinants. Also, statistical modeling in the
forms of age-standardized mortality rates and predictive analytics is considered routine in the
forecasting and delineation of risk factors. Quantitative research by Dritsas & Trigka (2022),
explored the behavioral, societal, and healthcare access factors contributing to mortality
complements this quantitative work. However, limitations sharing the same generalizability
among some findings include underreporting, data lags, and demographic oversimplifications.
Recent advances in big data analytics and machine learning offer promising tools for addressing
these limitations.
As per Nowbar et al. (2019), most of the analyses in these studies employ longitudinal data
analytic techniques that link health survey data with records of deaths as a way of determining
change over time as accurately as possible. Applications of Cox proportional hazards models have
been used, for instance, to examine the relative impact of various socio-demographic factors on
all-cause mortality. Katarya & Meena (2021), combined multilevel modeling to account for not
only individual-level factors but also those at the contextual level, such as neighborhood
characteristics and socioeconomic conditions. Some have also conducted meta-analyses to
synthesize findings across multiple studies in reviewing socioeconomic status and its influence on
all-cause mortality. These different methodologies emphasize that the use of an integrated
approach is necessary for understanding the multidimensional nature of health outcomes to inform
public health interventions in the reduction of mortality rates across various demographic groups
(Su et al. 2021).
METHODOLOGY
Data Sources
The datasets for our research project were retrieved from the Kaggle website, namely,
"NCHS - Leading Causes of Death: United States" which was very informative regarding the major
causes of death in the United States between the years 1999 and 2016 (Cordova, 2024). It was
organized in such a way that one can analyze the trends; hence, it includes variables such as Cause
of Death, such as heart disease and cancer, Year, State, Age-adjusted Death Rate, and Number of
Deaths. Other demographic variables, like Sex and Race/Ethnicity, further allowed for even finer
subgroups, which were very useful in highlighting disparities in health outcomes. This dataset is
particularly useful in public health research, with granular information at both the national and
state levels, useful in targeted interventions and policy decisions (Cordova, 2024)
Data Pre-Processing
By using the Python program, a series of data preprocessing steps were performed to
prepare the dataset for further analysis. Firstly, column names were renamed for clarity and ease
of access. Secondly, redundant columns like "Cause_Name_Duplicated" were dropped to
streamline the dataset. Thirdly, suitable codes checked for missing values within the dataset,
probably enabling their handling or removal. Finally, descriptive statistics are calculated and
printed out to provide a better view of the data's central tendencies, dispersion, and distribution
which will help in understanding its characteristics (Pro-AI-Robikul, 2024). The data for numerical
variables was standardized according to the min-max scaling/normalization and standardization
approach in such a way that the data falls under the range of a common scale and can enhance the
performance of the model. Feature extraction involves creating new features from current ones to
capture underlying relationships or trends. This approach included techniques such as principal
component analysis (PCA) or feature engineering.
Exploratory Data analysis
Figure 1: Portrays Total Deaths by Cause
The bar chart above illustrates the distribution of total deaths according to causes, whereby
"All Causes" takes precedence over individual causes by a landslide. Among the specific causes,
Alzheimer's disease and cerebrovascular diseases rank very high, which indicates the huge impacts
on mortality rates. Chronic lower respiratory diseases and diabetes mellitus also have quite
considerable proportions, reflecting health challenges due to these conditions. In addition, deaths
from diseases like influenza and pneumonia, and those from intentional self-harm, reflect serious
public health concerns. Malignant neoplasms, though important, seem to contribute relatively little
compared to the rest of the causes listed. This information underlines the need for specific health
interventions and awareness related to the leading causes of death, with special emphasis on age-
related and chronic conditions.
Figure 2: Displays the Age-Adjusted Rate by Cause
The box plot above depicts the distribution of various causes of age-adjusted death rates.
We observe that "All Causes" is quite spread out and high in dispersion, indicating several
mortality factors. "Diseases of the Heart" and "Malignant neoplasms (Cancer)" both have a similar
kind of spread or variability in their respective death rates. It is interesting to note that the two
series, "Accidents (unintentional injuries)" and "Intentional self-harm (suicide)", are differently
shaped: the former has a higher median and a longer tail, meaning a larger number of cases with
rates far above the median compared to the latter. From this observation, it is likely to be inferred
that deaths due to accidents are usually more serious or frequent in certain subgroups.
Figure 3: Depicts Total Death Over the Years
The total number of deaths shown in the bar chart from the year 2000 to 2017 reflects
overall increasing trends in mortality, with the upper half of that period most strongly marked by
it. Starting around 2000, the total deaths seem to level out, fluctuating within a range of
approximately 8.6 million and 8.8 million until about 2010. A real growth is observed after 2010,
reaching a peak of almost 9.8 million deaths by 2017. This could easily reflect several scenarios:
an aging population, rising chronic diseases, or increased efficiency in reporting methods. The data
emphasizes the need for continued monitoring through public health means and interventions into
the root causes of mortality as those causes change.
Figure 4: Showcases Top 10 States by Total Deaths
As showcased above, among all states, California leads regarding the number of deaths,
closely tagged by Florida and Texas, which reflects their bigger population and potential public
health challenges. Illinois and New York are also among the leading ones, showing high mortality
rates in these states. The bars depict error bars to provide an estimate of the variation that may be
associated with discrepancies in reporting or variations in demographics. Overall, the chart focuses
on the importance of implementing targeted health programs in these populous states and
addressing those factors that are causing a higher death rate.
Figure 5: Visualizes Heatmap of Deaths by State and Cause
As displayed above death heatmap for every state, Alabama and Mississippi are highly
ranked in deaths from other chronic diseases like heart and diabetes. These sharp contrasts,
rounded up by better treatment and prevention, probably yield the low death rates for causes in
states like California and New York. Interestingly, the heatmap underlined very specific causes
like Alzheimer's and cancer that vary significantly among states, hence underlining the need for
localized health strategies. It does effectively communicate the complex and intertwined nature of
population demographics, healthcare access, and specific health problems at the state level that
different states are facing, further calling for targeted public health interventions.
Figure 6: Exhibits Trend Analysis of Deaths Over Years for Top Causes
The trend analysis bar chart shows the tendencies of deaths from different causes during
the years, underlining significant patterns within the years 2000-2017. Among these, "Accidents
(unintentional injuries)" present a gradual increase, reflecting ongoing challenges in public safety.
On the other side, the category of "All Causes" is rather stable, which may indicate that some
effective public health measures could be in place to manage overall mortality. However,
Alzheimer's disease has an alarming upward curve, an indication of more and more aging
populations and increased suffering due to neurodegenerative conditions. On the other hand,
"Cerebrovascular diseases" and "Chronic lower respiratory diseases" show only fluctuation and
fail to show an increasing trend; perhaps this reflects effective health responses. The variability of
the other causes is further brought out by the shaded area around the line for "All causes,"
underlining the need for continued monitoring and health strategies appropriately targeted to meet
emerging health concerns.
Figure 7: Portrays the Distribution of Age-Adjusted Death Rates
The histogram above presents the age-adjusted death rates presented reveals a highly
skewed pattern, with the majority of frequencies concentrated at lower death rates, particularly
between 0 and 100. The preponderance of frequencies occurs at relatively low death rates,
especially between 0 and 100. Most populations have relatively low mortality, though some
populations face mortality rates far higher, as the right tail in this distribution shows. The peak
around the lower end suggests good healthcare interventions in most counties; however, the
elongated tail points to notable outliers or high-risk groups that require targeted public health
efforts. Smoothing the curve overlay indicates a gradual decrement of frequency as the rates of
death increase, reinforcing that the notion, while many people face low mortality rates, a subset of
concern is facing much higher rates and should be further investigated for the roots of the causes
or disparities in healthcare access and/or socioeconomic factors affecting these rates.
RESULTS
Trend Analysis
As showcased in the bar chart above shows that heart disease is still the number one cause
of death in the U.S., with 173.8 deaths per 100,000 population in 2021, signaling that the health
burden is still high. Cancer also comes second at 146.6 deaths per population of 100,000. COVID-
19 had a great toll with 104.1 deaths per 100,000, underlining the severe public health implications
of the pandemic. Accidental death syndrome rounds out 64.7 per 100,000, reflecting ongoing
challenges with safety and preventability. Other, less frequent causes include stroke, better known
as cerebrovascular diseases; chronic lower respiratory diseases; Alzheimer's disease; and diabetes,
thus rounding out the main contributors to the general mortality rate. This data underlines the
pressing need for targeted health intervention and preventive measures; these should be directed
at tackling the disease of heart conditions and its risk factors, along with the continued effects of
COVID-19 on public health.
Determinant Identification
These are some of the reasons for the growing trend that needs a deeper understanding of
its correlates by considering main driving variables like behavioral, genetic, socioeconomic, and
environmental factors. The ranking features can be extracted for analyzing mortality using
machine learning models, which determine which of those features have more impact on causing
death.
Behavioral and Lifestyle Factors.
Behavioral factors are major contributors to the leading causes
of death. While smoking, though its prevalence has decreased due to effective public health
campaigns, is still a major risk factor for heart disease, cancer, and respiratory diseases. Poor
dietary habits and physical inactivity are contributing to increasing rates of obesity, which, in turn,
increases the risk of heart disease, diabetes, and certain cancers. Another determinant is excessive
alcohol consumption, associated with liver disease and accidents.
Socioeconomic Determinants.
Income and healthcare access are strong determinants of mortality.
Because of a lack of access to medical care, nutritious food, and health education, populations in
the lowest income categories also have the highest rates of preventable deaths. Rural areas, for
example, have higher rates of mortality from heart disease and accidents partly due to health
professional shortages and infrastructure challenges.
Environmental and Genetic Factors.
Respiratory health is greatly influenced by exposure to
environmental pollutants, including PM2.5, which can lead to the development of diseases such as
COPD and lung cancer. Genetic factors combine with lifestyle and environmental ones to
determine susceptibility in general, for example, to cancers and heart disease.
MODEL PERFORMANCE
a) Linear Regression
Suitable code snippets were implemented to build a linear regression model using a Python
library, scikit-learn. First, it imported the necessary modules to manipulate data and evaluate the
model, including functions to split the dataset into training and testing sets, and metrics related to
the performance of the model. The dataset was defined with X, representing features, and y, the
target variable, "Age_Adjusted_Death_Rate". The data is divided into a training set and a test set.
A linear regression model is created and trained on the training data. Predictions are made on the
test set, and various evaluation metrics-MMAE, MSE, and R-squared-are computed for the
model's accuracy and goodness of fit. The output metrics gave further details into the model's
performance in the prediction to show how close the results of predictions were to real data.
Output:
Linear Regression Metrics:
Mean Absolute Error: 144.37476872184658
Mean Squared Error: 51259.10809919128
Root Mean Squared Error: 226.4047439856137
R-squared: 0.0568044753131709
Table 1: Showcases the Linear Regression Performance Metrics
As showcased in the above performance metrics, from the MAE value of approximately
144.37, the average magnitude of errors in the predictions of this model is shown. The implication
here is that the predicted values deviate from the actual ones by about 144.37 units on average.
The error magnitude is further stressed by the Mean Squared Error of about 51,529.11, squaring
the residuals so that larger errors are penalized more, hence really signaling the overall accuracy
of the model. The Root Mean Squared Error of about 206.40 is just the square root of MSE, thus
more interpretable in the same units as the target variable and hence indicating typical prediction
error. Finally, the R squared value is approximately 0.87, indicating that about 87% of the variance
in the target variable is explained by the model; this indicates a good fit. Taken together, all these
measures indicate that the model performs reasonably well with, nonetheless, still space for
reducing the prediction errors.
b) Random Forest
Equally, an appropriate code snippet was applied for the implementation of the Random
Forest Regressor using some library on sci-kit-learn on a dataset for representing predictive
capability and performance evaluation. It started with importing necessary modules: a Random-
Forest-Regressor for model construction and different metrics-MSE and R2, which are to be used
in the evaluation. It instantiated a Random Forest model with a specified number of estimators, a
seed for reproducibility. Further, it is fitted against the training dataset, X_train, and made
predictions on the test set, X_test. The evaluation section computed key metrics: Mean Squared
Error (MSE) describes the average of squared differences between predictions and actual values;
Mean Absolute Error provides an average of absolute deviations, while Root Mean Squared Error
provides insights into the average prediction error in the original units. The R-squared metric
rounds out the evaluation by showing the proportion of variance explained by the model, therefore
giving a clear picture of the effectiveness of the model.
Output:
Random Forest Metrics:
Mean Absolute Error: 85.62235261987875
Mean Squared Error: 31569.418411180308
Root Mean Squared Error: 177.67785008599216
R-squared: 0.4191054962413395
Table 2: Displays the Random Forest Performance Metrics
Performance metrics for the Random Forest refer to different aspects of its predictive
accuracy. An MAE of approximately 85.62 implies that the model is off, on average, from the
actual value by this amount, and this provides a simple interpretation of error magnitude. The MSE
of approximately 31,569.41 emphasizes larger discrepancies due to squaring the errors, making it
sensitive to outliers. The ~177.68 square root of MSE also states the model's prediction error in
the same units as that for the target variable and thus gives one a better idea concerning the
magnitude of the typical amount of prediction errors. Finally, the R-squared value of 0.42 infers
that the model describes ~42% of the overall variability in the target variable, hence showing
moderate predictive power with lots of room for improvement.
c) XG-Boost Regressors
The implementation of an XG-Boost was also successful as the most powerful gradient-
boosting library. The code first imported some necessary metrics from sklearn. Metrics, including
mean squared error and R-squared score, for evaluation of the model. Subsequently, the code sets
up the DMatrix for both the training and testing data. This is a special data structure used by XG-
Boost to optimize both memory and computation. The model parameters were defined, such as the
objective function-reg: squared error, the evaluation metric, and the maximum tree depth. Next,
this script trained the model on these parameters and generated predictions on the test dataset.
Finally, the snippet computed several metrics over this model; it printed out the mean absolute
error and the R-squared score, hence giving insight into the accuracy and predictability of the
regression model in question.
Output:
XGBoost Metrics:
Mean Absolute Error: 83.53977533400115
Mean Squared Error: 24323.105929615624
Root Mean Squared Error: 155.95866737573652
R-squared: 0.5524415950643691
Table 3: Portrays the XG-Boost Performance Metrics
Performed metrics by the XG-Boost regression model provide great insight into how well
this model learned to predict. The MAE is around 83.54, which says something about the average
absolute difference between the predicted and actual values, thus giving a general bar on the level
of prediction accuracy. The MSE of about 24213.11 shows the average of the squared differences,
giving more importance to larger errors, which can serve well in understanding the variance in the
predictions. The RMSE is approximately 155.99 and serves as an interpretable scale of error,
reflecting the average prediction error in the same units as the target variable. Finally, the R-
squared value of 0.55 means that at least 55% of the variance in the target variable is explained by
the model. That is a moderate degree of fitness but also a margin of ability to increase model
accuracy and predictive power.
MODEL COMPARISON
The appropriate code snippet was deployed in Python to generate a comparative
visualization of performance metrics of the model using Mean Squared Error scores of three
different machine learning models, namely Linear Regression, Random Forest, and XG-Boost.
Further in the code, it prepared a bar chart comparing the MSE score of the three models, which
served as a quick view of the predictive accuracy of those models. The charts were created to
visually depict the R2 scores of the models, which are supposed to describe the proportion of the
variance in the dependent variable explained by the model. By looking at the following
visualizations, one can try drawing some conclusions about the performance of each relative to
others and find which one fits this task best.
Output:
Figure 8: Depicts the Model Comparison MSE & R
2
Score
The above bar charts above outline the performances of the three machine learning models,
Linear Regression, Random Forest, and XG-Boost, based on Mean Squared Error (MSE) and R-
squared (R
2
). The lower the value for MSE, the better predictive accuracy the model gives. While
a high value in R
2
determines how much of the dependent variable's variance is explained. The
above charts clearly show that XG-Boost outperforms the other models significantly for both MSE
and R
2
. This therefore means that on this dataset, XG-Boost is the best model that can be used for
the most accurate and reliable prediction.
Predictive Insights
Advanced machine learning models, applied to mortality trends, provide deep insight into
the underlying determinants. Large datasets comprising demographic, socioeconomic, and health-
related variables are analyzed for patterns and correlations that may not be obvious in traditional
statistical methods. Model predictions can indicate future trends in mortality by highlighting
populations at high risk and locations. For example, a model could predict that over time, the
mortality rate in some regions will increase due to factors such as aging populations, socio-
economic disparities, or environmental pollution. These insights will help public health
policymakers allocate resources effectively and undertake targeted interventions. Besides, models
can help explain the complex interaction of factors leading to mortality. By looking at the relative
importance of various variables, researchers can understand the operative mechanisms of mortality
trends. For example, in a model, smoking, unhealthy diet, and lack of physical activity may emerge
as the most important causes of cardiovascular mortality. This information can be used to develop
evidence-based interventions aimed at reducing these risk factors.
Case Studies and Practical Applications
Identifying Populations at Risk:
A machine learning model such as the XG-Boost can identify
subpopulations, including older adults with multiple chronic conditions and individuals living in
poverty-stricken neighborhoods, who are at a high risk of mortality. This information will then be
useful in targeting preventive interventions and improving healthcare access for the most
vulnerable.
Predicting the outbreaks of diseases:
The proposed models use data from previous outbreaks,
mobility patterns, and climate factors to predict the future possibility of outbreaks. This enables
public health officers to take steps in advance to prevent infectious diseases and reduce their impact
on public health.
Optimization of Healthcare Resource Allocation:
The recommended Models can project the
future healthcare needs of bed capacity and staffing in hospitals, among others, based on variables
such as population demographics, disease prevalence, and seasonal variations. This information
helps optimize resource allocation and ensures that healthcare systems are better prepared to meet
the needs of the population. Personalized
Treatment Planning:
The models can predict the risk associated with an individual's history,
genetic information, and states of lifestyle about certain diseases and tailor the treatment
accordingly. This personalized approach in healthcare will improve patient outcomes while
reducing healthcare costs.
DISCUSSION
Public Health Implication
Data-driven models hold monumental implications in public health through the provision
of insights into the trends and determinants of mortality, besides including possible interventions.
They can locate high-risk groups, enable the forecast of outbreaks, and improve resource
optimization in health care; all through using big amounts of data. Such comprehension might
position the leaders with evidence-based decisions on the way to apply pointed interventions to
improve the health of people. Insights on How Data-Driven Models Can Inform Public Health
Interventions
Targeted Interventions:
After deploying the proposed models, and the identification of the
populations that are at risk, policymakers may be able to allocate resources to the groups in which
preventive interventions such as vaccination campaigns or lifestyle counseling would do the most
good.
Early Detection:
With the predictive models, outbreaks of diseases can be pinpointed at an early
stage, thus enabling appropriate timely public health responses against their spread and impact
mitigation.
Resource Allocation:
Policymakers can foresee future health needs and allocate resources
appropriately, thus preparing health systems to curb demand.
Personalized Medicine:
Data-driven models can enable personalized medicine by treatment
planning that categorizes each patient's characteristics, resulting in better and more effective health
care.
Public Health Policy and Preventive Recommendations
Data Sharing and Interoperability:
Incentivize the sharing of data among healthcare providers,
public health agencies, and research institutions to derive insights from data.
Investment in Data Science and AI:
Invest in research and development in data science and AI to
bring on board advanced predictive models.
Ethical Guidelines:
Develop and enforce ethical guidelines for the collection, use, and sharing of
health data to protect patient privacy and ensure data security.
Public Health Literacy:
Empower the population through the use of improved public health
literacy so that individuals can make more empowered choices regarding their health and
participation in public health activities.
Integration into Healthcare Systems
Predictive models require multidimensional integrations into health systems, including
technological infrastructure, data governance, and workforce training. Advanced analytics can
improve patient outcomes, reduce costs, and enhance the overall quality of care. On the other hand,
the integration of data-driven insights into public health strategies also poses its challenges. Ethical
considerations, such as privacy and bias, must be carefully addressed to ensure that these
technologies are used responsibly. Besides, the quality and completeness of data contribute a lot
to the accuracy of the predictions a model can make.
Future Research Directions
Advanced Machine Learning:
Understand how some of the very latest advanced machine learning
techniques are being used, deep learning, and reinforcement learning in particular, and how that
might be exploited further to increase model accuracy and predictive power.
Real-Time Health Monitoring:
Design and develop a real-time health monitoring system that can
collect data from wearable devices and other sensors for analysis and the issuance of warnings on
health risks.
Explainable AI
- develop techniques to make these machine learning models more interpretable
for the healthcare provider to understand the rationale behind certain predictions.
Ethical AI:
Encourage the development of ethical AI guidelines so that the technologies are used
responsively and equitably.
CONCLUSION
This research project aimed to use the data-driven approach in the exploration of these
trends to understand the patterns and determinants underpinning mortality statistics. Using an
expanded data set, the study presented leading causes of death; the pattern of variation by
demographic factors, including age, sex, and race/ethnicity; and social, environmental, and
behavioral determinants of those patterns. The datasets for our research project were retrieved from
the Kaggle website, namely, "NCHS - Leading Causes of Death: United States" which was very
informative regarding the major causes of death in the United States between the years 1999 and
2016. It was organized in such a way that one can analyze the trends; hence, it includes variables
such as Cause of Death, such as heart disease and cancer, Year, State, Age-adjusted Death Rate,
and Number of Deaths. Other demographic variables, like Sex and Race/Ethnicity, further allowed
for even finer subgroups, which were very useful in highlighting disparities in health outcomes.
The performances of the three machine learning models, Linear Regression, Random Forest, and
XG-Boost, based on Mean Squared Error (MSE) and R-squared (R
2
) were evaluated.
Retrospectively, XG-Boost outperformed the other models significantly for both MSE and R
2
. This
therefore means that on this dataset, XG-Boost is the best model that can be used for the most
accurate and reliable prediction. In that respect, advanced machine learning models, applied to
mortality trends, provide deep insight into the underlying determinants. Large datasets comprising
demographic, socioeconomic, and health-related variables are analyzed for patterns and
correlations that may not be obvious in traditional statistical methods. Model predictions can
indicate future trends in mortality by highlighting populations at high risk and locations. Data-
driven models hold monumental implications in public health through the provision of insights
into the trends and determinants of mortality, besides including possible interventions.
References
Ahsan, M. M., & Siddique, Z. (2022). Machine learning-based heart disease diagnosis: A
systematic literature review.
Artificial Intelligence in Medicine
,
128
, 102289.
Al Amin, M., Liza, I. A., Hossain, S. F., Hasan, E., Haque, M. M., & Bortty, J. C. (2024).
Predicting and Monitoring Anxiety and Depression: Advanced Machine Learning
Techniques for Mental Health Analysis.
British Journal of Nursing Studies
,
4
(2), 66-75.
Alam, S., Hider, M. A., Al Mukaddim, A., Anonna, F. R., Hossain, M. S., khalilor Rahman, M.,
& Nasiruddin, M. (2024). Machine Learning Models for Predicting Thyroid Cancer
Recurrence: A Comparative Analysis.
Journal of Medical and Health Studies
,
5
(4), 113-
129.
Bhowmik, P. K., Miah, M. N. I., Uddin, M. K., Sizan, M. M. H., Pant, L., Islam, M. R., &
Gurung, N. (2024). Advancing Heart Disease Prediction through Machine Learning:
Techniques and Insights for Improved Cardiovascular Health.
British Journal of Nursing
Studies
,
4
(2), 35-50.
Bortty, J. C., Bhowmik, P. K., Reza, S. A., Liza, I. A., Miah, M. N. I., Chowdhury, M. S. R., &
Al Amin, M. (2024). Optimizing Lung Cancer Risk Prediction with Advanced Machine
Learning Algorithms and Techniques.
Journal of Medical and Health Studies
,
5
(4), 35-
48.
Cordova, I. (2024, September 4).
usa_leading_causes_death
. Kaggle.
https://www.kaggle.com/datasets/isaaccordova/usa-leading-causes-death?select=NCHS_-
_Leading_Causes_of_Death__United_States.csv
Dritsas, E., & Trigka, M. (2023). Efficient data-driven machine learning models for
cardiovascular diseases risk prediction.
Sensors
,
23
(3), 1161.
Dutta, S., Sikder, R., Islam, M. R., Al Mukaddim, A., Hider, M. A., & Nasiruddin, M. (2024).
Comparing the Effectiveness of Machine Learning Algorithms in Early Chronic Kidney
Disease Detection.
Journal of Computer Science and Technology Studies
,
6
(4), 77-91.
Hider, M. A., Nasiruddin, M., & Al Mukaddim, A. (2024). Early Disease Detection through
Advanced Machine Learning Techniques: A Comprehensive Analysis and
Implementation in Healthcare Systems. Revista de Inteligencia Artificial en Medicina,
15(1), 1010-1042.
Hossain, M. S., Rahman, M. K., & Dalim, H. M. (2024). Leveraging AI for Real-Time
Monitoring and Prediction of Environmental Health Hazards: Protecting Public Health in
the USA. Revista de Inteligencia Artificial en Medicina, 15(1), 1117-1145.
Islam, M. Z., Nasiruddin, M., Dutta, S., Sikder, R., Huda, C. B., & Islam, M. R. (2024). A
Comparative Assessment of Machine Learning Algorithms for Detecting and Diagnosing
Breast Cancer.
Journal of Computer Science and Technology Studies
,
6
(2), 121-135.
Katarya, R., & Meena, S. K. (2021). Machine learning techniques for heart disease prediction: a
comparative study and analysis.
Health and Technology
,
11
(1), 87-97.
Nasiruddin, M., Dutta, S., Sikder, R., Islam, M. R., Mukaddim, A. A., & Hider, M. A. (2024).
Predicting Heart Failure Survival with Machine Learning: Assessing My Risk.
Journal of
Computer Science and Technology Studies
,
6
(3), 42-55.
Nowbar, A. N., Gitto, M., Howard, J. P., Francis, D. P., & Al-Lamee, R. (2019). Mortality from
ischemic heart disease: Analysis of data from the World Health Organization and
coronary artery disease risk factors From NCD Risk Factor Collaboration.
Circulation:
cardiovascular quality and outcomes
,
12
(6), e005375.
Pro-AI-Rokibul. (2024).
Analyze-Trends-and-Determination-of-Loeading-causes-of-deaths-in-
US/Model/main.ipynb at main · proAIrokibul/Analyze-Trends-and-Determination-of-
Loeading-causes-of-deaths-in-US
. GitHub. https://github.com/proAIrokibul/Analyze-
Trends-and-Determination-of-Loeading-causes-of-deaths-in-
US/blob/main/Model/main.ipynb
Rahman, A., Karmakar, M., & Debnath, P. (2023). Predictive Analytics for Healthcare:
Improving Patient Outcomes in the US through Machine Learning. Revista de
Inteligencia Artificial en Medicina, 14(1), 595-624.
Su, Y. S., Ding, T. J., & Chen, M. Y. (2021). Deep learning methods in internet of medical
things for valvular heart disease screening system.
IEEE Internet of Things
Journal
,
8
(23), 16921-16932.
Zandt, F. (2024, February 2). What are the leading causes of death in the U.S.?
Statista Daily
Data
. https://www.statista.com/chart/30883/deaths-from-leading-causes-of-death-in-the-
united-states/
