1
Impacts of Data Analysis on the Financial Viability of Health Care Firms
Student’s Name
Institutional Affiliation
Course Number and Course Name
Instructor’s Name
Assignment Due Date
2
Introduction
Background Information
Information has enabled improved planning and new initiative creation. We can be more
structured with more data. This data may be used to predict current and future occurrences,
defined as a process of gathering data. As we became aware of this, we started to collect data on
nearly everything. We are now awash with data from all walks of life, including social,
scientific, occupational, and health. It's like a data flood now. Technological advancements have
enabled us to generate a rising volume of data, which is now unmanageable. Thus, "big data"
refers to vast volumes of unmanageable data. We need new ways to organize and analyze data to
fulfill current and future societal expectations, one such requirement in healthcare. A lot of data
is generated in the healthcare business, which has both benefits and drawbacks. This study of big
data fundamentals will discuss an emphasis on healthcare.
Impacts of data analysis on the financial viability of health care firms
In many sections of the nation, the healthcare business is expanding. Uncertainty about
the future development path is a primary cause of growth discomfort. Between 2010 and 2015,
the United States spent $3.2 trillion on healthcare, the most significant amount ever. This
substantial financial spend is the result of several factors. Every year, around 251,454 Americans
are killed due to medical mistakes. Better data-driven decision-making may relieve these
concerns while also supporting the organization in its move to value-based care. Information
technology is used to run businesses in the healthcare sector. This technology captures massive
3
amounts of data daily. Analytics is a set of tools and methods that help people interpret large data
sets and make better health-related decisions.
Analytics is the use of quantitative and qualitative analysis to maximize the value of your
information. It might be a great concept if well-planned, regulated, and monitored, the lessons
learned are used. For example, analytics was used by the Centers for Medicare and Medicaid
Services (CMS) to minimize hospital readmission rates and avoid $115 million in fraudulent
payments. Data mining and text mining help clinicians anticipate, diagnose, and treat illnesses
while increasing service quality and cutting costs. Each year, data mining has the potential to
save the US healthcare system $450 billion. Over the last decade, a flurry of research on the
subject's practical and theoretical elements was conducted (e.g., methodological or philosophical
issues surrounding data mining)(Zhang, 2020). This project aims to gather and synthesize peer-
reviewed literature on data mining from both a practical and theoretical perspective. Analytics
types (descriptive, predictive, and prescriptive), healthcare applications (clinical decision
support, mental health), and data mining technologies used are described (e.g., classification,
sequential pattern mining).
1. Motivation and Scope
A large body of literature on healthcare and data mining has accumulated due to recent
reviews/conceptual studies. These characteristics are the scope of health care, the date of
publication, and the total number of papers examined. For example, one research examined
publications published between 2001 and 2005 and found that 18 fit the requirements. There is
currently no comprehensive assessment of data mining applications in the health care business
that provides a comprehensive picture of how they operate. There is one review for each of the
4
twenty-one main categories of health care. Three evaluations deal with clinical medicine, two
with detecting adverse drug response signals, two with big data analytics, and four with the use
and performance of data mining techniques. Two trials were conducted with a specific sickness
in mind. None of these publications tell the whole narrative on this issue to the best of our
knowledge. Except for four publications, these investigations lack rigor in terms of technique.
These articles include crucial details regarding the study's timeframe, database search,
and criteria for adding or omitting pertinent content. The breadth of these investigations,
however, is likewise relatively restricted. Our work contributes to the rising number of
theoretical studies on the issue of analytics by condensing the empirical literature. There is
currently a wealth of knowledge on coping with methodological issues and using big data
analytics in health care. According (to Noorein Inamdar, 2007), it covers a wide variety of
application domains, a wide range and depth of data mining methods, and an examination of the
literature's quality. This examination aims to fill the gaps that they discovered before. People
who write about data mining and big data analytics in healthcare provide a more systematic and
complete examination of the practical and theoretical elements.
2. Methodology
In our review, we followed the PRISMA recommendations. For cross-sectional
investigations, the JBI Critical Appraisal Checklist was utilized, while for qualitative research,
the Critical Appraisal Skills Programmed (CASP) qualitative research checklist was employed.
5
2.1. Sources
This section discussed the literature utilized in the review and selected it. This review
was followed by a single phase (Phase 2) of advanced keyword searching in PubMed and Google
Scholar using the period filter "All Fields" from January 1, 2005, to December 31, 2016. They
used the "All Fields" filter for this search. The site now includes filters for English-language
journal articles. The whole job-searching procedure went as follows: The initial search area
included a lot of articles, which were subsequently weeded away by phase 2 keywords. In the
second part of the experiment, there was 129 Web of Science entries and 5255 PubMed records.
A Google Scholar search for phase 2 keywords yielded 700 items. They preserved the entire
papers on data mining and big data in healthcare decision-making. Getting rid of duplicate
articles early in the PRISMA review process, rather than towards the end when they were
examined, saved a lot of time.
2.2. Quality Assessment and Processing Steps
Two researchers examined each of the 117 papers one by one to ensure no bias. To assess
the quality of the research, we utilized the JBI Critical Appraisal Checklist for Analytical Cross-
Sectional Studies. When doing theoretical work, we employed the CASP qualitative research
checklist. A clear goal and inclusion criteria; a clear description of the sample population and
factors; valid and reliable data collection; ethical concerns; a thorough explanation of the
findings; and a clear description. We adjusted the checklist sections to account for this. The
project's data collection at the conclusion included any research that matched these criteria.
We employed three-step processing in Levy and Ellis and Webster and Watson. First,
extracting information included determining the nature of the issue, addressing it, and
6
emphasizing significant outcomes ("Know the literature"). To assist us in grasping the literature,
we summarized each piece and compared it to related works. It ensured that the study included
no unnecessary data. Second, we kept track of our idea matrix using a spreadsheet, as advised by
Webster and Watson. We modified the idea matrix around 20% to keep up with new results, or
every 23 articles ("Apply"). Third, we devised a classification system based on the notion matrix
that depicts operationally defined each class. It eliminated no duplicate items in the pool
(Analyze and Synthesis). They discussed six publications, and their classifications were
compared. The final categorization method classified articles according to their summary, facts,
and reviewer comments ("Evaluate").
3. Analytics in the healthcare industry
92 of the 117 articles discussed how they might utilize to make choices in the health care
business. These articles address various forms of analytics, the types of data utilized, and the data
mining techniques employed.
3.1 Analyses of Various Kinds
According to the literature, there are three types of analytics: predictive analytics
(predicting what will happen in the future based on past data), descriptive analytics (exploring
and discovering information in a dataset), and prescriptive analytics (making recommendations
based on historical data) (i.e., utilization of scenarios to provide decision support). Only five of
the 92 research employed both descriptive and predictive analytics simultaneously. The most
prevalent analytics utilized in healthcare is descriptive analytics (48 percent). Except for clinical
7
decision support, descriptive analytics was the most prevalent approach to perform analytics. The
only application field that employed descriptive analytics was pharmacovigilance research. It is
due to the application's goal of establishing a relationship between unfavorable pharmacological
effects and medicine(Büsing et al., 2021). In 43 percent of the experiments, they applied
predictive analytics. Because there were so many studies looking at the risk and morbidity of
chest discomfort, heart attack, and other ailments, it was the most prevalent application of
predictive analytics. On the other hand, applied Prescriptive analytics in just 9% of the studies. It
might be because most of the research focused on a specific demographic group or illness
situation.
3.2. Types of Data
We used the approach Raghupathi and Raghupathi to classify different data types. In this
approach, all aspects are considered simultaneously: the kind of data that originated and was
obtained. The most often utilized kind of data in the healthcare business is HG data, including
EMR, EHR, and EPR. Because health care is becoming increasingly digital, online or social
media data (WS) is the second most common data (11 percent). It is only one of many
developments in Natural Language Processing (NLP). It makes it much easier to use WS data
than it was before. Wearable personal health monitoring devices are becoming increasingly
popular, and as a consequence, some people may begin to use more SD and BM data. As a
consequence, data utilization would increase by over 13%.
3.3 Information extraction
The articles under examination use a variety of data mining techniques. It doesn't say
much about each method, but it does say how often and where each technique is used. In 57 of
8
the 57 studies included in the analysis, categorization was used to assess data. Twenty-one
papers used association and 18 articles that employed clustering simultaneously, but not both at
the same time. Utilized other approaches on a less frequent basis. Used it in eight out of nine
pharmacovigilance reports. When it came to categorizing, everything else was secondary. In
health care, data warehousing is frequently used.
We focused on categorization since it was common in most research (57 out of 92). The
most often used algorithms are known as "ANN," "LR," "DT," and "DT-based" algorithms.
Radiofrequency, Bayesian networks, and fuzzy logic algorithms were often, but not always,
used. Three articles addressed innovative algorithms that devised a technique for categorizing
breast cancer patients based on discrete particle swarm optimization. Clustering algorithms in
healthcare often used self-organizing maps and K-means to help them figure out what groups of
people were associated(Büsing et al., 2021). The accuracy, sensitivity, specificity, AUC, positive
predictive value, negative predictive value, and so on of the algorithms varied depending on what
they used them for and the kind of data they evaluated. We propose that you test several
algorithms before choosing the best one.
4. The Use of Analytics in Healthcare
4.1 Cardiovascular Illness (CVD) Cardiovascular Illness (CVD)
Cardiovascular disease (CVD) is the leading cause of mortality globally. Seven research
emphasized its importance for public health (18 percent of articles in clinical decision support).
A decision tree is utilized to classify CHD risk variables. First, the ACGs were all checked (MI).
CABG, PCI, and MI occur. People may be in danger of four things segregated unchangeable
variables from the changeable variables (such as age or gender). Then followed the key danger
9
signs. The Framingham equation determined the risk. Aside from family history, I discovered an
age-related risk factor. Demographics, medical history, essential physical examinations, and
blood tests, as well as noninvasive testing like heart rate, glucose levels, and BMI, are not
required. It is quicker, cheaper, and more accurate. Create a decision tree to organize your
research processes. The fuzzy model outperformed an ANN (ANN).SVM and PNN may not
work well with large ACVD datasets. They built a Fuzzy SAM to aid sort. They employed a GA
to reduce unsupervised fuzzy rules (GA)(Batko & ŚLęzak, 2022). This game's rules changed.
Because the fuzzy system was slow, it reduced the qualities. ARTMAP (63.46%) and ANFI
outperformed SVM. Data filtering is critical in cardiovascular prediction (i.e., patients are not
followed up after discharge until a new incident occurs; the available data gets right-censored).
Encryption changes predict outcomes. Censorship is Inverse Probability Two research solved the
data filtering issue. IPCW computes weights on data before using a Bayesian Network to classify
it. One of these researches employed the IPCW approach.
Electrocardiography is frequently the best approach to diagnose heart illness (ECG).
Machine learning may aid decompress ECG analysis. It will take up computer time and space,
but it's worth it; I did it in one research. They created a mobile patient monitoring gadget to
speed up diagnosis and treatment. The hospital server could differentiate a normal ECG from an
irregular one. The approach discovered irregular heartbeats 97% of the time, must be switched
on the patient's phone to notify the hospital or emergency services. Insight into unusual
CVDs.Interruption of the hypoplastic left heart(HLHS). If our baby's heart doesn't operate
properly, we'll require surgery. Pulse, heart rate, systemic and mixed venous oxygen saturation,
and other measures may assess health. The body may perish if you don't detect a minor
alteration. The researchers created a model by comparing physiological data to treatments. There
10
were 17 pediatric ICU records (PICU). It also has nurses' and machines' feelings. Each paper got
a health rating. With each well-being score, there are guidelines for selecting interventions.
Changing feature values influences the classification process as assessed by Combined
Classification Quality (CQC). Unlike other quality measures, CQC enhances categorization
accuracy (CQ) can be accurately classified Two CQ = 1 characteristics 35.5-75% of the time.
4.2 Diabetes
Global diabetes rates rise. Oman will be sixth by 2030. Utilized data mining to discover
people's diabetes histories. We looked at seven diabetes analytics studies. The history of diabetes
indicates that the illness may manifest in numerous ways and have many implications. The first-
level patients received weekly blood glucose and urine testing (e.g., eye tests for diabetic
retinopathy). The second level is ideal for diabetes with particular requirements (e.g.,
cardiovascular, eye, liver, and kidney-related complications). This group taught the authors about
unusual diabetes complications. Density-based clustering algorithms like DBSCAN can handle
noisy data. It is why I picked it. Use this patient grouping to compare the prices of similar
treatments. Also examined are the challenges of managing T2D. The FSSMC rated vital sections.
FSSMC employed naive Bayes, IB1, and C4.5 algorithms to compare each distinct group. Tested
on 3857 people's physiological and lab data, classifier algorithms. Age, diagnosis, and insulin
therapy are vital to one's health.
We developed an algorithm based on data from two hospitals and one research. Patients'
data are lost when they change hospitals. Using the HTPC approach may result in
misclassification. Rural residents experienced a 32.9 percent false-negative rate if they utilized
two health care institutions instead of one. The findings from another study predicted T2D. So it
11
was worth more. The dataset contains medical illnesses and disorders. It was calculated using a
hybrid technique. Used SVM to predict diabetes risk based on food and body measures. The
diabetes survey data had more excellent organization than the insurance claims data. Statistics
and AI classify things. This research employed two statistical approaches to compare SVM to
ANN and Random Forest: (LR and Fisher linear discriminant analysis). The system is tested on
age, gender, BMI, waist circumference, smoking, job, hypertension, and diabetes family history
(diabetes or no diabetes). Every test, an ANN using SVM parameters won. SVM outperformed
other algorithms by 5%. Also, statistical approaches agreed. Diabetes was rare in the sample,
making classification challenging. Convolutional nonnegative matrix factorization is a novel
pattern recognition method. A patient was to them a single unit of time. Patterns may help you
organize and manage patients. It's a grid, so it's simple to grasp.
4.3 Cancer
Cancer is a severe health hazard. It has been used to detect and predict cancer. Chin
cancer-analytics studies were viewed it's challenging to forecast how long cancer patients will
survive due to the disease's complexity, treatment choices, and patient kinds. Prostate cancer
survival rates are classified. The model uses a public database, SEER. This study employed
stratified ten-fold sampling. DT, ANN, and SVM were employed. SVM outclassed DT and ANN
in classification, with 92.85% accuracy. It may also indicate how long you will survive lung
cancer. The number of excised lymph nodes and malignant or in situ tumors were the most
significant predictors. A group picked them to live for 6, 9, 1, 2, and 5 years (J48 DT, RF,
LogitBoost, Random Subspace, and Alternating DT).
12
A lung cancer calculator predicts mortality at 6, 9, 1, 2, and 5 years. Machine learning
has been used to diagnose and predict the outcome of cancer cases in the past. Researchers
developed a hybrid algorithm to distinguish between cancer patients and non-cancer patients. We
started by looking at correlation and regression. DPSFO discovered the data (PSO). This
combination method worked well on the Wisconsin Breast Cancer UCI. At least 96% and 93%
of respondents agreed. This research also examines demographics and cancer kinds. The
researchers utilized SAM with GA (CVD). They utilized a UCI machine learning repository to
increase data classification. Download the application to see how age, gender, diseases, and
cancer statistics vary over time (trend, correlation) (e.g., diabetes, kidney infection). Researchers
employed data mining to identify and cure cancer quicker.
4.4Cares in an Emergency
The ER is the main hospital entrance. In 2011, 20% of Americans visited the ER.
Affordability improves efficiency and speeds up patient flow. It may aid with specific ED
procedures (DES). PEOPLE, THERAPEUTIC OPTIONS, URGENCY, AND UN To facilitate
therapy, a patient's treatment demands may be categorized. Case-mix isn't appropriate in
telemetry or nephrology. Used data mining (clustering) to characterize ER visitors (e.g., full
ward test, head injury observation, ECG, blood glucose, CT scan, X-ray). ER doctors studied the
clustering model. Each group developed its ED method using discrete event simulation
(primarily duration of stay). It checked for chest pain in ED patients. Determined the severity of
the chest discomfort in three steps. The initial step was to collect samples and identify them.
Experts devised criteria for connecting lab tests to diagnosis to help clinicians diagnose quicker
and reduce unneeded lab testing. Then they created a category tree using lab test data. It was in
an ER. They utilized C5.0, NN, and SVM to classify lab tests. NN and SVM were both 88.89%
13
correct. They treat clients with life-threatening diseases or injuries. ICUs cause 22.4 percent of
hospital mortality. Calculating life expectancy and risk variables may help physicians plan
therapy. They designed ICU deaths. For example, one research employed student t-tests to
uncover 15 of 40 essential features. Use these tools to predict death probabilities. DT, SVM,
ANN, APACHE 3. DT leads the other teams by 0.02 AUC. The study focused on the first 24
hours in the ICU. Another group utilized a MIMIC-II similarity score to predict 30-day
mortality. It grew better at predicting because it utilized data from comparable cases (such as
vital signs and lab findings). When included all patient data, it became 0.81.
4.5 Data Mining
People utilized data mining to predict pressure ulcer risk, create issue lists, and
personalize medical therapy. Their work led to two classification-based pressure ulcer prediction
systems. Two models were created using the 14 characteristics (age and gender): (5 in DT
model, 7 in SVM, LR, and Mahalanobis Taguchi System model). MTS and SVM were used to
choose significant characteristics (in the second model only). MTS outperformed LR and SVM
by 10 to 15%, except for sensitivity and specificity. However, it can't predict pressure ulcers and
the Norton and Braden medical scale. 8 out of 168 research participants experienced pressure
sores. Another study team developed a new scale. They employed data mining to forecast
pressure ulcers based on four years of patient data (i.e., days of stay in the hospital, serum
albumin, and age). Each of DTS (0.63), LR (0.82), and adaptive regression splines Weak data
mining approaches create subsampling by not showing one class is more valuable than others.
They wanted to try subsampling and not subsampling to discover which was superior. The
14
classifications became better for eight illnesses with a 5% prevalence rate in the Healthcare Cost
and Utilization Project's National Inpatient Sample (NIS) (HCUP). Better than the other two
approaches (+0.01 AUC), maybe due to more uniformly distributed data in the datasets.
4.6 Additional Uses
It's vital in clinical practice. QA/PM, But it seldom works. Researchers claim that
diagnosing all issues increases treatment outcomes. Professionals use problem lists to diagnose.
Data mining linked patient complaints, medicines, and lab testing. ARF might save time and
money. Unstructured data includes doctor and nurse observations. They may help physicians
avoid errors and speed up therapy. Both can coexist. For the same disorders, physicians use the
same data. One study showed 90% of pharmaceutical difficulties and 55% lab issues. Outpatients
with respiratory illnesses took 92.79 percent of the time. Practitioners have access to 100,013
medicines. It didn't matter what others said or how old they were. Age, ethnicity, gender, and
patient complaints did not determine which medications were the most popular. Medicine
focuses on a patient's capacity to recover. One study team employed big data to ensure patients
received the most excellent treatment. In this situation, each patient's data is examined. This
study compared people's susceptibility to infection. Another team employed structured and
unstructured data (such as illness codes and demographics) to provide individualized treatment.
Notes or remarks made by a doctor or nurse while treating someone A clinical ontology matched
unstructured clinical terminology. These profiles were created using mapped codes and data
(disease code). Some patients were categorized by appearance. A patient's condition was two or
more ailments. Then a protein network was created using the illness protein or structure. Gene's
proteins share a structure, indicating a connection.
15
Researchers developed an algorithm that considers both the surroundings and the users'
health to prevent asthma episodes. They looked at-home care and the surroundings (air pollutants
and weather data). Used DT and association rule mining to choose attributes and predict hazards.
It correctly predicts asthma attacks by 86%. People liked knowing their risk of severe asthma.
Researchers created a mechanism for Parkinson's patients to analyze and evaluate data. Minority
Oversampling was utilized to level out the data (SMOTE). SMOTE rebalancing improves
AdaBoost classification accuracy by 96%. Machine learning-based classification algorithms
surpassed statistical classification strategies regarding accuracy and reliability. A model for
predicting renal disease survival was developed. The report contains 188 patients, 707 visits,
demographics, and dialysis solution content. A patient group has two comparable features. The
data were divided into eight groups at random and assigned a number. The RS and DT
algorithms classified the eight groupings. There are 3-year, 3-month, and unknown survival
courses. Utilized Most of the 16 criteria to choose the final class. It implies DT outperformed RS
by 11%. At least 67%. (The accuracy is 56%.) The clinical trial design might benefit from
predictive analysis and tailored therapy selection. Others utilized biochemical data to predict
dialysis demand. High albumin levels in the blood increase the risk of hospitalization.
Researchers used EHRs from 7463 patients to predict 5-year mortality. Enlisting patients
with a life expectancy of five years or fewer may be done using the Ensemble Rotating Forest.
Aside from age, BUN and comorbidity were important. AUC 0.81 for the adjusted Charlson and
Walter indices (AUC 0.86).AUC 0.78 Sickness was more prevalent among older adults who had
been to the hospital previously. This approach may help individuals get the most out of services
like cancer screenings. Current disease prediction and detection software cannot handle
sophisticated and large systems (like a Bayesian Network) and needs extensive training. (INN).
16
So they created the "VFDT." VFDT can categorize quicker and use less energy than IBM. In
glaucoma, data mining is utilized to diagnose. All three scanning methods are costly. The SVM
classifier discovered glaucoma. A discrete wavelet transform was employed to classify the data
(DWT). Used the same SVM kernel functions. It was 95% correct regarding glaucoma. Follow-
up testing indicated delays in chest imaging for malignancy. The patient's exit was anticipated.
The algorithm was 98% accurate in standard imaging and 38% accurate in pathological imaging.
Data mining looked at three things to check for health issues (CABI, CAUTI, and VAP).
Care Fusion is a San Diego-based corporation (ICD-9-CM)(Batko & ŚLęzak, 2022). Traditional
approaches outperformed data mining in rate estimation. Utilized it in 38 publications on CVD,
diabetes, cancer, emergency, and critical care (16 articles). Innovated forecasting methods and
tools SVM frequently outperforms other algorithms. However, a doctor or other healthcare
professional was not used to constructing models or analyzing data (see the study characteristics
in Supplementary Materials Table S3). We don't know how physicians and other health care
providers feel about these algorithms or whether they utilize them. Not looked into.
5. Future Research
Data mining is used in marketing and production. Its application in medicine is growing.
Data mining challenges include noisy data, heterogeneity, massive data, dynamic nature, and
computing time. Individualized care, data mining in healthcare, interdisciplinary study, domain
expert knowledge, and automation for non-experts. The EMR is increasingly utilized to record
patient demographics [1]. Personalized treatment plans based on EMR data may increase patient
satisfaction [162]. MISSING DATA WERE OFTEN ERASED OR DELETED IN; 46.5 percent
of the data and 363 out of 410 variables were missing. Another used just 2064 of 4948
17
observations (42%). Enlarging the range of data, we lose information. Instead of deleting missing
data, future studies should estimate them. It's also needed to develop new data gathering
methods. It is common to remove outliers. However, one of our investigations showed that
outliers might reveal unique conditions. Therefore, dismissing outliers is a mistake.
5.1 Individualized care
Data collecting in healthcare is used to track patient progress and plan treatment. Include
research goals in data collection to improve data quality and reduce errors. The Pomeranian
Health Study (SHIP) collected valuable data. Northern Germany SHIP's purpose was to identify
prevalent illnesses, risk factors, and overall health. Consequently, the research found a
systematic data collection with high dependability, minimum noise, and few missing data points.
Use existing documentation strategies to acquire better-organized data (EMR or EHR). Adding a
research goal to healthcare data gathering takes advanced preparation. Data mining for non-
experts, Doctors, nurses, and other healthcare workers lack analytical abilities. An automated
(human-free) system for end-users is one option. A cloud-based automated method to prevent
medical errors is possible but complicated. Interdisciplinary and domain knowledge is a cross-
disciplinary field. Oncologists, for cancer research, and cardiologists, for CVD research. About
32% of analytics papers had no professional input. Future studies should involve healthcare staff.
5.2 Inclusion in the healthcare system
System integration, few studies assessed aimed to incorporate data mining into decision-
making. The effect of data mining on healthcare providers' time and effort is unknown. Future
research should examine the system's impact on workplaces. The Black Swan and error
prediction is risky. Even though half of the analytics literature we reviewed is about prediction, a
18
wrong forecast has implications. Prediction of cancer or other illness does not imply decision-
making. Also, the models may predict more frequent occurrences than unusual ones. Like in
"The Black Swan," a paradigm for the unexpected is required. One research looked at evidence-
based prescriptions. How much proof is needed before making a recommendation? Many of the
research in this review ignore these issues. Predictive models are challenging to use, mainly
when mistakes and unexpected events occur.
Conclusion
Created this framework to provide high-quality patient-centered healthcare. Big data has
opened up new research opportunities for academics and practitioners. A rising body of research
shows the benefits of health care analytics in improving global health systems. An integrated and
informed health care system is the ultimate aim. This paper is the first thorough assessment of
health care analytics and data mining. It addresses both practical and theoretical elements of
healthcare analytics and data mining. The review procedure lets readers evaluate more
thoroughly by classifying the outcome as analytical or theoretical. Future research should include
domain-expert knowledge, reduce prediction error, and integrate predictive models in real-world
work situations. In addition, the analysis should respond to mistakes and unexpected events in
the future. Despite these valuable findings, our suggested review process has flaws. As a result,
our study is limited by the absence of relevant conference materials. A 12-year search span may
have missed prior research on this issue, while the growing trend since 2005 and a smaller
quantity of publications before 2008 may have compensated for this constraint. It may also
exclude works published in languages other than English. Like Levy and Ellis, we did not
undertake a reverse search (reviewing literature that mentioned the chosen article). Despite these
flaws, may use the review's analytical approach in any healthcare setting.
19
References
20
Batko, K., & ŚLęzak, A. (2022). The use of Big Data Analytics in healthcare. Journal of Big
Data, 9(1). https://doi.org/10.1186/s40537-021-00553-4
Büsing, C., Rachuba, S., &Thielen, C. (2021). Special Issue on Healthcare Analytics. Operations
Research for Health Care, 30, 100309. https://doi.org/10.1016/j.orhc.2021.100309
Liu, C., & Chen, A. S. (2015). Do firms use dividend changes to signal future profitability? A
simultaneous equation analysis. International Review of Financial Analysis, 37, 194–207.
https://doi.org/10.1016/j.irfa.2014.12.001
Noorein Inamdar, S. (2007). Examining the Scope of Multibusiness Health Care Firms:
Implications for Strategy and Financial Performance. Health Services Research, 42(4),
1691–1717. https://doi.org/10.1111/j.1475-6773.2006.00686.x
Zhang, D. (2020). How do firms overcome financial constraint anxiety to survive in the market?
Evidence from extensive manufacturing data. International Review of Financial Analysis,
70, 101503. https://doi.org/10.1016/j.irfa.2020.101503
