▼ 

For several decades, social disparities in access to health care remain a major debate in the U.S. health care system. Despite growing attention to health inequalities, different social classes, especially, minority or ethnic groups, and those without health insurance coverage continue to face challenges to health care. To date, due to the complexities of Socioeconomic Status (SES), it is unclear how SES impacts health and income inequality. The purpose of this dissertation was to examine the association of SES and median household income groups with hospitalization outcomes in the United States from 2008 to 2010. To examine the generalizability of this phenomenon, a retrospective study was used to analyze the pattern of care for hospitalized patients between the ages of 18 and 89, using the National Inpatient Sample (NIS) data of the Healthcare Cost and Utilization Project (HCUP). The study sample consisted of 500,000 admission records and stratified and regression analysis were computed to determine the differences by age, sex, race or ethnicity, income, location, diagnoses, procedures, length of stay, payer, and costs affecting each of the defined income categories. Total hospital costs were examined within the categorical income groups by residential zip code and top 10 diagnoses and procedures showed that high medical costs is an issue across SES groups. Descriptive and inferential statistical analyses were performed. Mean, median, standard deviation, and range were used to calculate continuous variables while frequency counts and chi-square tests of association were conducted to evaluate differences in proportion for categorical variables. Linear regression modeling and multivariable modeling techniques were undertaken to test the hypotheses. Measurement and structural models were tested through structural equation modeling statistical techniques using SPSS version 22.0. When compared the SES differences among the four categorical income groups, the results show that people at the lower quintile were more likely to face higher hospitalization due to their income. Each year, many programs are designed to reduce hospital admissions, but regardless of these efforts the rates of hospitalization continue to increase in U.S population. This study recommended scientific approach in understanding of the role SES and income as they impact health disparities, which will potentially help health providers, researchers, policy makers, and public health planners to design individualized and community-wide programs and policies related to income inequality and hospitalization for high risk populations.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ A SNOMED CT-encoded problem list is required to satisfy the Certification Criteria for Stage 2 “Meaningful Use”. ICD-10-CM has replaced ICD-9-CM as the reimbursement code set in 2015. Having a cross-map from SNOMED CT to ICD-10-CM would promote the use of SNOMED CT as the primary problem list terminology, while easing the transition to ICD-10-CM. There is no established principle and methodology on systematically and semantically linking SNOMED CT to ICD-10-CM. This research project describes the development of mapping principle, mapping guidelines, mapping tools and mapping methodology for a rule-based crosswalk to support semi-automatic generation of ICD-10-CM codes from SNOMED CT-encoded data. A series of mapping guidelines were developed based on the clinical use case, SNOMED CT modeling convention, and ICD-10-CM classification guidelines. One of the important methodology in developing the map set is using triangulation in generating legacy maps. Using the SNOMED CT to ICD-9-CM map and General Equivalence Mappings sequentially, Indirect Map was generated from SNOMED CT to ICD-10-CM for 96.2% of the SNOMED CT concepts within the scope of the study. Another innovation in this crossmapping research is implementation of a principle to handle age specification. The age rule was one type of rule to handle cases in which one SNOMED CT concept can map to different ICD-10-CM codes depending on the age of the patient. The age rule quality assurance (QA) was a mechanism to capture the age specification that can be easily missed by manual mapping. The results showed that the mapping guidelines ensured the mapping consistency, which potentially would reduce the mapping discrepancy between the two independent parallel mapping efforts. It also made it possible that the map set can be used in a meaningful way when data is exchanged. On this triangulation method in generating legacy map, an Indirect Map generated from SNOMED CT to ICD-10-CM covered a very high percentage of SNOMED CT concepts. Overall, this Indirect Map had a moderate degree of agreement with the Direct SNOMED CT to ICD-10-CM map. However, the indirect synonymy maps have much higher precision and can be used for quality assurance (QA) of the three maps. The age rule QA identified 342 out of 7,277 concepts which potentially required age rules, among these 50.3% turned out to be true positives. Without this QA, a large proportion of age rules in the published Map would have been missed. The outcomes of this research project include a set of mapping principle, mapping guidelines, mapping tools and mapping methodology for a rule-based crosswalk from SNOMED CT to ICD-10-CM. All these could be used as a prototype in other cross standard mappings. For example, in the US, ICD-10-PCS officially replaced ICD-9-CM from October 2015 onwards. A project was formulating earlier this year (2015) for the purpose of creating the map from SNOMED CT procedure to ICD-10-PCS. It is a pleasant finding that tooling, principles and guidelines established in… Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

The projection of the U.S. national healthcare expenditure in year 2019 is $4.7 trillion. Medical errors are part of this increasing healthcare costs because they cause tens of thousands of deaths in the U.S. hospitals each year, more than major diseases such as AIDS, breast cancer combined to highway accidents (Chiang S. Jao, Daniel B. Hirer) [10]. Based on a research published in the British Medical Journal (BMJ) in 2016 and conducted by (Michael Daniel & Martin A Makary) [98], Medical error is ranked the third cause of death in the US. With the advancement in technology, we have seen during the last years, important improvements in the design as well as the use of electronic health records (EHRs), Computerized Physician Order Entry (CPOE), and Clinical Decision-Support Systems (CDSS), and Diagnosis Decision-Support Systems (DDSS) to improve the quality of health care delivery; progress have been made but challenges remain. Medication errors can be: •Wrong drug, •Wrong dose, •Wrong route, •Wrong patient, •Bad combination, •Bad reaction to list a few, and are found at every stage from prescription and administration of drugs to monitoring. They hurt about 1.5 million people, and cost billions of dollars each year according to the Institute of Medicine of the National Academies. Medication errors can happen anywhere, from Doctors offices to hospitals, and pharmacies and your home. Sound-Alike / Look-Alike also known as drug name errors, are the most common causes of medication errors, they originate from poor communication between health care providers, poor communication between patients and their providers. To reduce the likelihood of ham related to medications and Adverse Drug Events (ADEs), many interventions have been attempted including notably: The US Food and Drug Administration (FDA), government legislation, policy makers, drug utilization reviews, health professionals, and patients education, all of this with limited success. The aim of this dissertation is to evaluate medication errors related to Sound-Alike drug names, and to propose a new approach of preventing them by "Embedding the Novel Decision Algorithm" coupled with "Confused Drug Names, Generic and Brand drug names" and "Doses" within a computerized provider order entry (CPOE) during the drug prescribing process.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

Background: Ontologies characterize complex and detailed information and are extensively used in healthcare research. Medical information (textbooks, expert opinions, clinical evidence) has information on conditions and its corresponding procedures (treatments), but this information is not captured or structured in any ontology. The objective of the research is to create a condition-procedure ontology from real world data to be utilized in observational research or electronic health record (EHR) system. Methods: Predictive models are developed to learn from five datasets (administrative claims, hospital charge data) to generate two algorithms (diagnostic and therapeutic) to predict condition-procedure relationships in the SNOMED-CT vocabulary. A reference set with 100 positive pairs per algorithm, and 32,132 negative pairs were developed. Predictive models were constructed by designing 51 possible covariates that describe condition-procedure pairs from Optum© De-Identified Clinformatics® Data Mart Database – Socio-Economic Status (Optum) dataset and determining which covariates discriminated between the positive and negative controls, as measured by Area Under Receiver Operator Characteristic Curve (AUC). External validation of the final algorithms was performed on 4 other databases. The final algorithms were applied across the universe of condition and procedure pairs in all five databases to construct the full condition-procedure ontology, and the ontology was evaluated for validity and coverage of condition and procedure concepts from the set of identified condition-procedure pairs. An additional analysis was trained to classify diagnostic vs. therapeutic intervention based on the overlap of pairs within the two algorithms. Results: Algorithms include the following covariates: condition-procedure occurring together, relative risk, support and sensitivity. Both algorithms had AUCs greater than .90, and external validation also showed similar results. In Optum, 98% of conditions and 63% of procedure codes had at least one relationship identified in the ontology. The intervention type analysis resulted in an AUC of 0.79. Conclusions: Real-world data can be utilized to construct a medical ontology of condition-procedure relationships with strong performance and good coverage. These results can be utilized to fuel research efforts in healthcare such as cohort generation and computer provider order entry systems by understanding conditions and procedures and their application to diagnose or treat a patient.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

Introduction: Abbreviations are often used in clinical documentation to reduce time spent documenting in electronic health records and to save space during documentation. Abbreviations represent a specific challenge in healthcare as they can often contain multiple means. This ambiguous use of abbreviations is a patient safety issue for clinicians who do not properly understand the intended use of the abbreviation and presents a health literacy issue to patients as they try and understand what a provider’s note says about the care provided. Plenty of research has been done on a clinician’s ability to disambiguate abbreviations, but little work has been done to assess how clinicians are using abbreviations or creating tools to assist administrators and clinicians to explore the documentation of their providers. Methods: A semi-supervised approach was taken to identify potential abbreviations within the MIMIC-III database. Over 400 million-word tokens were compared to a list approved abbreviation for Beth Israel Deaconess Hospital. The results of this semi-supervised identification were used to analyze the use of abbreviations and prevalence of abbreviations within the dataset. Results: 463,175,566 raw word tokens were compared to a list of 1,742 approved abbreviations. On average, every document within MIMIC contained almost 14 abbreviation tokens, or roughly 9% of an average note is comprised of potential abbreviations. Some notes contained almost 26% of potential abbreviation tokens. The average count of potential abbreviations for a note created by an RN is 21.87, and the average count of potential abbreviations in a note created by an MD is 11.39. There is a substantial difference in the number of abbreviations used in a note by an RN and MD. MIMIC note events contain a substantial amount of abbreviations >= 5%. Conclusion: Using the MIMIC data set we have shown that clinical abbreviations and complex clinical jargon make up a specific amount of provider documentation. 8.22% of total words within the MIMIC note events table is a term found within the Beth Israel Deaconess approved abbreviation list. We have also shown that there is the capability to replace abbreviations in medical text to provide additional context to patients.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

This study analyzed 10 different types injuries and illnesses. Specifically, amputation of finger, burns 3rd degree of the hand, falls on same level, fracture of upper limbs, heat stress, laceration of the upper limbs, machine accidents, overexertion, sprains/strains, and struck by or against an object. The objective of the study is to discover if age, race, or gender factor in the hospitalization outcomes of the aforementioned injuries and illnesses. An additional objective is to explore whether region, day of incident (weekday or weekend), length of stay, social economics, and total medical charges in the presence of these specific work-related injuries and illnesses are impacting factors. Data was available by the Healthcare Cost and Utilization Project (HCUP) sponsored by the Agency for Health Care Policy and Research1. The National Inpatient Sample (NIS) data in years 2007 through 2011 was assessed and downloaded from HCUP. A total of over 15 million patients aged 18-64 and who did not die when admitted to the hospital in the United States between years 2007 through 2011 (5 years). The data provided patient demographics such as: age, gender, race, insurance type, and income. The Statistical Package for the Social Sciences (SPSS) version 26 was serviced to analyze the data of the study, and all outcomes with a p-value less than 0.05 were found to be significant. Frequencies and multiple linear regression were the appropriate statistical tests to determine the predictors of the study outcomes. White older aged males (31 to 64 years) have the highest frequency of injury and illness. The 76th to 100th percentile income level had the highest frequency of injury and illness. Majority of injuries and illnesses occurred in the South region. The regression model discovered that indicator of sex is they key variable in the amount of time spent in the hospital and the total amount of hospital charges. Falls from same level injury, had 70,226 patients, which is 49% of the total population of all 10 injuries and illnesses investigated. In addition, older aged White females (31 to 64 years) were the highest frequency of patients for falls from same level. Older White males in the 76th to 100th percentile income have the highest risk of injury and illness in the workplace. Preventative measures should improve work-related injuries and illnesses; especially for older ages, provide knowledge through specific training to prevent complacency and help workers to be more aware of risks associated with their age, gender, income, and job duties.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

Breast cancer is the most commonly diagnosed female cancer in the United States. According to the American Cancer Society in 2020 there will be an estimated 276,480 new cases of invasive breast cancer diagnosed in the United States. Surgery continues to be the gold standard for treatment. While breast conserving surgery has been widely accepted, many patients elect mastectomy. The objective of this study was to examine hospital characteristics of mastectomy patients to determine what factors impact length of stay, total charges and in-patient mortality. The objective of this study was to analyze the frequency of mastectomy, whether for risk reduction or the presence of disease. This study utilized the National Inpatient Sample (NIS) database for years 2008 to 2011 to examine patient demographic characteristics such as age, race, insurance type and income. SPSS statistical analysis version22 was utilized for analysis, with p values less than .05 considered significant. Linear regression, Logistic Regression, A NOVA and Chi-Square were used to determine significant predictors of study outcomes Between 2008 and 2011, 55,781 female patients underwent mastectomy with 76.7% electing unilateral mastectomy. Immediate breast reconstruction occurred in 38.6% of patients. The highest incidence of mastectomy occurred in White and Asian women. The total number of discharges revealed privately insured White and Asian women had the highest discharge each year. A further analysis regarding income revealed Whites and Asians also had the highest income. Length of stay remained consistent from 2008 through 2011. Mortality was insignificant among the mastectomy population with Blacks seen to have the highest percentage of in-hospital mortality. Fluid and electrolyte disorders were found to be the highest predictors of mortality. Overall totals costs were on a rising trend and consistent with type of mastectomy. Among hospital inpatients electing mastectomy, there were racial differences that occurred in treatment and outcomes. This study highlighted hospitalization characteristics related to female mastectomy admissions in the United States between 2008 and 2011. Racial disparities in treatment and outcomes highlight areas where efforts may be focused to improve survival among specific population groups. Future research should be targeted on identifying specific causes of racial differences in patients electing mastectomy and outcomes related thereto. This study validates the importance of health education to improve health awareness. Policies promoting early childhood health education may provide the necessary knowledge and life skills to aid in navigating access to resources to promote better health.  

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

Surgical site infection (SSI) is a significant patient safety issue in hospitals that is related to more extended hospital stays, and increased cost burden. Since the infection prevention strategies been inconsistently implemented in hospitals, hospital size must be an influential factor to cause an impact on SSI rates. We retrospectively analyzed 222,845 cases with SSI from National Inpatient Sample (NIS) data developed for Healthcare Cost and Utilization Project (HCUP) database, between the year 2008 and 2012 at small (<250 beds), medium (25-450 beds), and large (100-450+ beds) size hospitals. Risk factors, including demographics, socioeconomic, location, and functioning features then introduced to compare interhospital SSI prevalence and county-specific SSI rates. Finally, risk factors were regressed to assess the association between risk factors and SSI measures. With an overall prevalence of 2.67 per 100 procedures, unadjusted prevalence rates were 2.9% in small, 2.62% in medium, and 2.65% in large hospitals (p<0.0001). Patients with transfers, high severity of the disease, comorbidities, catheterization, and academic hospitals were the vital distinguishing factors for SSI rates amongst the hospital varying in capacities. The elderly patients at the small rural and minorities (Black and Hispanic patients) at the large urban teaching hospitals were at higher SSI risks. Fluid & electrolyte imbalances and weight loss were most recorded comorbidities. Every year hospital administration aims to reduce SSIs without losing gained revenue. Besides infrequent adherence to infection prevention (IP) strategies, it identifies hospital and patient-related conditions that influence SSI rates. According to this study, interhospital SSI rates disparity associations were multifactorial with partial elucidations allied to variances in demographics, transfers, level of severity of the disease, comorbidities, and socioeconomic factors in small and large hospitals. The findings also led to more infection reduction exertions towards hospitals from rural and teaching hospitals from urban counties. Therefore, the assessments of infection prevention deficiencies with the refined overtime data provide more information on modifiable indicators and that if explored in more detail at hospital settings, it can help infection preventionist for benchmarking.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Professions.

▼ 

The purpose of this study is to determine the hoop stress around dental implant in a circumferential and vertical pattern and to compare stress values to different cantilever length on a cross arch implant supported framework. Material and Methods: A milled cross-arch metal framework supported by 4 implants imbedded in acrylic in addition to one separated implant in the middle. Five T-strain gauge rosette distributed on five dental implants recorded data as they were loaded by 50 and 100 N forces using MTS 810 loading machine. The loading sites were anterior implant, posterior implant, 0.5 Anterior Posterior ,(AP spread) ratio cantilever, 1.0, 1.5 and 2.0 ratio cantilever. Each one of those 7 sites were loaded 10 times for each of the 50 and 100 N generating 360 different reading for each of those groups. Three-way ANOVA was conducted (twice, one for vertical and one for hoop strain) for analysis of strain with factors of: Magnitude of force, extension of cantilever and position of the implants followed by Post hoc Tukey comparison between the groups. Results: The anterior implant is under tension in vertical direction when forces are applied to the cantilever on the contralateral side. The most posterior implant is under less tension in the vertical direction but shows as much strains in the circumferential direction. Post hoc analysis shows cantilever over 1.0 ratio causes same high amount of stress to the system. Conclusion: A ratio up of 0.5 AP spread for the cantilever is statistically acceptable. However, A ratio up to 1.0 AP spread for the cantilever exerts no more tension to the system than the load on the implants.

Advisors/Committee Members: Srinivasan, Shankar (chair), School of Health Related Professions.

▼ 

Abstract Background: Alzheimer's disease (AD) is the commonest dementia, which has no recognized cure. It causes deterioration as it advances, and ultimately results in death. AD was primarily defined by German psychiatric specialist and neuropathologist, Alois Alzheimer. Frequently, AD Alzheimer’s disease is diagnosed in persons above the age of 65 years, even though the less common early-onset of the disease can happen. It is anticipated that more than 3 million individuals aged 85 and more will have Alzheimer's. 33% of Americans over age 85 are burdened with the disease while 5.3 million Americans are living with Alzheimer's disease. Unless a cure is found, close to 16 million Americans will have the disease by 2050. Alzheimer’s is one of the most expensive diseases. The impact of the disease in the U.S. is the main objective of this study. To study this impact, the length of stay, mortality, and cost will be studied in terms of different patient characteristics and hospital contexts. Method: This study’s main objective was to find the influence of patient characteristics and hospital contexts on three outcomes, namely; length of stay, mortality, and costs. To achieve this objective, The Nationwide Inpatient Sample (NIS) was analyzed after using a filtering method to get a net sample size of 698,170 entries. The sample was obtained for statistical analysis for the six-year period covering 2007-2012. Descriptive and inferential statistic analysis were conducted in order to answer the research questions. Descriptive analysis includes frequencies, mean, and median. Inferential analysis includes multiple and logistic regression and qui-square models were utilized to test the significance of the relationships between independent and dependent variables of the study. Results: Some of the important results found in this study were: 1. The patient characteristics including the age and gender are a highly risk factor of the incidence of Alzheimer’s disease while the race is not a significant risk factor. 2. Alzheimer’s patients who were admitted to the hospital with psychosis on average stayed 2.20 days longer than those without psychosis (p < .001). 3. Alzheimer’s patients who were admitted to the hospital with normal pressure hydrocephalus on average were charged 4569.03 more than those without with normal pressure hydrocephalus (p < .001). 4. Alzheimer’s patients on average were billed 11,895.48 more per procedure performed (p <.001). 5. Alzheimer’s patients who were admitted to the hospital with diabetes were .92 times as likely to die as those without diabetes (not statistically significant). 6. The age group 65 and less has a length of stay of 1.6 more than other patients on other age groups.

Advisors/Committee Members: Haque, Syed (chair), Srinivasan, Shankar (internal member), Coffman, Frederick (internal member), School of Health Professions.

▼ 

Type II diabetes mellitus (T2DM) is a growing health concern in the United States, affecting almost 30 million individuals, and currently ranking as the 7th leading cause of mortality. In addition, T2DM is associated with multi-systemic complications that contribute to both early mortality and decreased quality of life. Individuals with T2DM can be diagnosed by blood glucose tests, and previous studies have demonstrated increased risk factors for T2DM development, including obesity, particular ethnicities, personal history of polycystic ovary disease, or a family history of T2DM. The present study aimed to find the connection between T2DM complications including ketoacidosis, hyperosmolarity, renal manifestations, ophthalmic manifestations, neurological manifestations, and peripheral circulatory diseases with the most widespread risk factors, including gender, race, family history of diabetes, obesity, smoking, alcohol-related disorders, hyperlipidemia, hypertension, hypercholesterolemia, asthma, Vitamin D deficiency, and age. The strongest association was found between increasing age and peripheral circulatory disorders, with those over 65 years showing the highest correlation (OR=22.081). Strong connections were also found between Asian/Pacific Islanders and age >65 with renal complications, as well as between alcohol abuse and hyperglyceridemia with ketoacidosis (OR=3.303 and 2.992 respectively). This study also tested two predictive models, Logistic Regression and Neural Network (ANN), in modeling T2DM with complications. Classification methods tests showed that three complications – renal manifestations, neurological manifestations, and ketoacidosis – were better predicted by these models than the other complications, and that both models performed very similarly in both sensitivity and specificity. This study demonstrates that specific combinations of risk factors can predict increased probabilities of specific complications in T2DM patients, and that a neural network analysis model can predict these relationships as accurately and with the same sensitivity as a standard linear regression model.

Advisors/Committee Members: Coffman, Frederick (chair), Haque, Syed (internal member), Srinivasan, Shankar (internal member), School of Health Professions.

▼ 

Lung cancer (LC) is a life threatening disease associated with significant cost and high mortality. LC is diagnosed in either early stage or more frequently in late stage, the face of lung cancer. Objective: To make a comparison between early and late stage lung cancer (SLC) in relation to cost and mortality Methods: The study is a random effects data analysis of a historical dataset the Nationwide Inpatient Sample (NIS). The study is based on the time period 2002, 2006 and 2011. The primary outcomes of interest is cost (total cost per day) and mortality (died/did not die). Two replicates samples for the years 2002, 2006 and 2011 were taken. Demographic factors that influence cost and mortality were co-varied out of the analysis. Descriptive Statistical analysis and bivariate analysis were done for cost includes ANOVA and ANCOVA. A statistical analysis for mortality includes logistic regression. Cost and mortality for early versus late (SLC) were measured in isolation and after accounting for age, gender, race, socio-economic status, number of diagnoses, length of stay, and number of procedures.Results: In the three years, 3 samples of 2173, 13,032, and 15,771 including 3 replicate samples of 2060, 13,032 and 15,772 participated in the study. All significant relationships tested at an alpha level of (P<0.05). The cost for early (SLC) was higher compared to late (SLC) and is statistically significant. The number of procedures in part accounted for the difference. Late (SLC) had higher mortality compared to early (SLC) and is statistically significant. The number of diagnoses in part accounted for the difference. The study showed early (SLC) costs 14% more than late (SLC). Late stage is more deadly, however, the gap is surprisingly small at 30% or an odds ratio of 1.3 to 1.5 after adjusting for covariates. Conclusion: This study of HCUP data revealed that early (SLC) is more expensive than late (SLC). Additionally, the data revealed that mortality is higher in late (SLC) compared to early (SLC). Overall, these finding highlight the important role of Health Informatics in understanding the cost and mortality of early and late (SLC).

Advisors/Committee Members: Haque, Syed (chair), Coffman, Frederick (internal member), Srinivasan, Shankar (internal member), School of Health Professions.

▼ 

Colorectal cancer is the 3rd most prevalent type of non-cancer, and the 2nd leading cause of cancer-related death in both men and women in the United States. This is despite being one of the most preventable and curable cancer types, when detected early. While the incidence and mortality from the disease has been declining over the past decade, its decline can be further accelerated by improving screening rates in order to identify the disease at the earliest stage, when the cure rate is at its highest. The goal of this study was to identify the socioeconomic attributes of the colorectal cancer screening population, and to assess if any socioeconomic attribute statistically increases the likelihood that a patient tests positive at screening. This was achieved by stratifying socioeconomic attributes, and the colorectal cancer screening population data from a national clinical diagnostics lab for patients screened in a 6 year period, beginning in 2012 through 2015. Some of the key findings from the study are outlined below: - There was a significant increase in CRC screening volume over the 6 years studied. This was an encouraging trend that shows a possible increase in CRC screening benefits awareness, which is pivotal in the ongoing effort to reduce the incidence and mortality rates from the disease - The population median income showed a decline over the 6 year studied, even as the CRC screening volume grew over the same period. This was an encouraging finding because it was indicative of the less affluent becoming more aware of CRC screening, as well as having better access to screening - There was an inverse relationship between the population’s median income and the positivity rate; as the population’s median income increased, the positivity rate declined - Poverty rate in the CRC screening population grew over the study period, in support of the trend observed with the population median income attribute. Another indication that CRC screening was becoming more accessible to the poorer population - The screening population’s bachelor’s degree attainment rate remained relatively stable of the 6 year study period, even as the CRC screening volume grew. Bachelor’s degree attainment had an inverse relationship with positivity rate

Advisors/Committee Members: Haque, Syed (chair), Srinivasan, Shankar (internal member), Hoffman, Frederick (internal member), School of Health Professions.

▼ 

Background: Crohn's Disease is a chronic disease of the gastrointestinal tract. Crohn's disease most commonly, affects the small intestine and colon, but any portion of the bowel from the mouth to the anus may be involved. The disease is characterized by intermittent episodes of relapse, remission, and recurrence, often requiring surgical intervention and/or therapeutic agents such as steroids, and novel immunosuppressive drugs, as part of the medical management plan. This investigation offers an understanding on how age, race gender, medical insurance and medical comorbidities play a role in the hospitalization of patients with Crohn's disease across the four regions (Northeast, Midwest, South and West) within the United States. Descriptive statistical analysis was conducted to detect observations that were statistically significant to further conduction other statistical applications. Analysis of variance was performed using a statistical model ANOVA in an effort to expose and uncover differences that are statistically significant between patient length of stay and four numerical variables ( LOS = Pay1, Race, Female and Agecat). Further, another ANOVA analysis was done to incorporate categorical variables (TOTCHG = Pay1, Race, Female, and Agecat). Logistic regression analysis was done to better understand the relation between patient demographic characteristics and outcomes of patients with Crohn's disease. Methods: This is a retrospective, observational, cohort study of Crohn's disease patients in the existing HCUP database for the years beginning in 2008-2012. Patients being admitted to the hospital based on the following ICD-9-CM code for Crohn's disease (555.0, 555.1, 555.2, 555.9), inflammatory bowel disease (569.89) and ulcerative colitis (556.1, 556.2, 556.3, 556.4, 556.5, 556.6, 556.8, 556.9). Results: Crohn's disease, inflammatory bowel disease and ulcerative colitis are within the age group of 18-40. The disease is more prevalence in whites. Logistic regression suggests that there are statistically significant predictive relations between patients demographics and outcomes of Crohn's disease in 65.5% of what is reported in the database. Conclusion: Young patients 18-40 exhibit higher hospitalization rates for Crohn's disease. Various medical comorbidities did not play a role in patient outcomes.

Advisors/Committee Members: Srinivasan, Shankar (chair), Coffman, Frederick (internal member), Casimir, Patrick (internal member), School of Health Professions.

▼ 

Hypertension is one of the most common worldwide diseases in the adult population and is a major risk factor for stroke, myocardial infarction, vascular disease, and chronic kidney disease. Numerous genetic, environmental, and lifestyle factors influence the development of Hypertension. The key objective of this study is to evaluate the association of metabolic variables with Hypertension individually and with the combination of co-factors (Age, BMI). The study utilizes a series of statistical procedures to achieve its objectives. Statistical Analysis was conducted using 10 Years of NHANES data from 2005 - 2014 datasets. The analysis only included an adult population of 25 years and older. Our study is in-line with studies which support that Hypertension is associated with the characteristic variables Age and BMI. Age and BMI are common threads in many organ abnormalities. The study further continued to analyze the association of Hypertension and characteristic variables with metabolic abnormalities. Based on our statistical analysis, we determined the association between our study variables and concluded that Hypertension is interrelated with most of the metabolic abnormities. Our study results showed that Hypertensive adults are more likely to have abnormal levels of Glycohemoglobin, Total Cholesterol, Albumin, ALP, AST, ALT and Creatinine irrespective of its underlying factors. However, Hypertension has no association with Total Bilirubin. Our study and evaluation were successful in achieving its objectives. We are 95% confident that Hypertension is either the leading indicator or a cause of metabolic abnormalities in target organs.

Advisors/Committee Members: Haque, Syed (chair), Srinivasan, Shankar (internal member), Coffman, Frederick (internal member), School of Health Professions.

▼ 

Background: Complete denture is considered one of the most challenging treatment modalities in dentistry. Any error occurs during any step of the complete denture fabrication can result in patient’s complaint. Troubleshooting complete denture complaints is very challenging for several reasons. First, all the teeth are connected into one denture base that is supported by a movable resilient mucosa make it difficult to isolate the errors. Second, one patient’s complaint could be due to several causes and at the same time one cause can lead to many patient’s complaints. Third, the diagnosis of each patient’s complaint must be approached systematically because adjusting the denture in a wrong spot may add another problem. Most of the complete denture treatment is provided by general dentists and depend only on their undergraduate training and clinical experience. There is a concern about the competency and level of skills of the undergraduate students in mastering the complete denture skills due to limited exposure to enough cases. The purpose of this research is to develop a new clinical decision support and training system to aid the general dentists and undergraduate students in diagnosis and management of complete denture complaints. Method: The new clinical decision support and training system was developed using Exsys Corvid Core software. The knowledge base of the system for the complete denture complaints, the complaints causes and the management protocol for each complaint were retrieved from the literature. The software was successfully loaded with 123 rules representing 48 patients’ complaints along with its detailed diagnostic methods and management protocol. After system development, the system was validated by ten expert prosthodontists using a survey questionnaire. The questionnaire results were statistically evaluated using Cronbach’s Alpha test. Results: The Cronbach’s Alpha reliability coefficient was 0.847, which represent a good internal consistency. The validation questionnaire results showed that all ten prosthodontists agreed on the need of such system and its user friendliness. Also, all prosthodontists agreed that the system is a good tool to assist the general dentists and undergraduate students in diagnosis and management of complete denture complaints. 90% of the prosthodontist agreed with the knowledge base of the system for the complete denture complaints, the causes of each complaints and the management protocol for each complaint. Conclusion: The clinical decision support and training system to aid the general dentists and undergraduate students in diagnosis and management of complete denture complaints was developed. The overall agreement of the ten evaluating prosthodontists with the system indicates that the system was successfully developed.

Advisors/Committee Members: Srinivasan, Shankar (chair), Coffman, Frederick (internal member), Kamel, Mohamed (outside member), School of Health Professions.

▼ 

Although the advancement of surgical techniques and immunosuppressive therapy has significantly improved the outcomes of kidney transplantation in patients with end stage renal disease (ESRD), post-transplantation outcomes remain a big challenge. Accordingly, the goal of this dissertation was to determine which variables might have the most critical impact on the graft and patient survival time. One such variable which seemed significant but not well studied was the Blood Urea Nitrogen (BUN) level of the donor. Therefore, using the United Network for Organ Sharing (UNOS) registry database (October 1987 to March 2016), a retrospective (longitudinal) cohort study was setup to examine the relationship between the BUN level of the deceased donor and the survival of the graft and the patient while controlling for certain other variables. The final sample consisted of 168,081 patients in the United States. Multivariate cox regression analysis revealed that high log BUN level of deceased donor remained an independent predictor of graft loss (hazard ratio [HR], 1.080; 95% hazard ratio confidence limits [CI], 1.032 - 1.131; P = 0.0009) and patient death (hazard ratio [HR], 1.063; 95% hazard ratio confidence limits [CI], 1.007 - 1.121; P = 0.0262) compared to low log BUN level of deceased donor. Significant findings from this study indicate that high log BUN level (> 2.79 mg/dl) of deceased donor is independently associated with decreased graft and patient survival time compared to low log BUN level of deceased donor. It is to be noted that White, Black and Hispanic donors races have significant differences at 5 year graft and patient survival time while donors of other races (Asian, American Indian/Alaska native, Native Hawaiian/other Pacific Islander, and multiracial) did not show a statistically significant influence on graft and patient survival due to genetic influences. These results can potentially contribute to a more efficient allocation of resources to donor sources with better outcome prospect.

Advisors/Committee Members: Coffman, Frederick (chair), Srinivasan, Shankar (co-chair), Gohel, Suril (co-chair), School of Health Professions.

▼ Treatment planning with regard to retention or extraction decision-making is often challenging, especially to dental students and inexperienced dentists. Such a decision requires consideration of many factors and clinical parameters to achieve an accurate definitive plan and therefore proper dental care. Contemporary dentistry has adopted the concept of evidence-based practice guidelines. Our project has followed this concept, thus literature evidence, clinician’s expertise, patient’s desires and preferences were factored-in to develop a contemporary approach for determining the overall tooth prognosis accurately. Although, several studies have been published to report the affect of specific clinical parameters on individual tooth prognosis, these studies were isolated and focused on particular aspects for developing tooth prognosis scale rather than evaluation of teeth in comprehensive manner covering all factors relevant to retention or extraction decision-making. Retention of non-salvageable teeth as well as extraction of salvageable teeth can be drastically devastating to patients since they are time consuming, cost more time and money and may lead to loss of trust and confidence in care givers. Thus, a multi-factorial approach for development of accurate prognosis is required. This approach requires sound knowledge of principles that span across multiple dental specialties of restorative/prosthodontics, periodontics and endodontics. It is extremely difficult for clinicians to recall an extensive list of factors that determine the overall tooth prognosis, especially for dental students and less experienced clinicians. On the other hand, expert clinicians with their knowledge and years of experience are accustomed to critical thinking utilizing comprehensive list of factors on day-to-day clinical practice. However, experts may not always be accessible for consultation at the point of care. To fulfill this ever-outstanding need, we propose the development of a clinical decision support system that satisfies the concept of evidence-based dentistry considering all factors relative to research evidence, clinician’s expertise, patient’s desires and preferences. Utilizing Exsys Corvid expert system development platform, we have developed an efficient, interactive and user-friendly tool that can easily be hosted on the web, implemented in clinical settings, and integrated with treatment planning procedures into the daily clinical practice. In addition to assistance in clinical setting, our system can also be considered as an educational tool that helps dental students and inexperienced dentists to process challenging clinical scenarios like expert clinicians since it includes comprehensive list of factors that acquires evaluation and response prior to reaching final recommendations. Based on information entered by the user, our system provides clinical recommendations, options and alerts labeled with patients and providers identification numbers, time and date for documentation purposes. Since our system was designed to… Advisors/Committee Members: Srinivasan, Shankar (chair), Coffman, Frederick (internal member), DiPede, Louis (internal member), School of Health Professions.

▼ Obesity has reached near epidemic proportions in the United States. Rising obesity and its associated comorbidities result in deleterious effects on health status[4-6] and a significant increase in health burdens [7, 8]. Excess cost attributable to overweight and obesity was reported to be approximately $92.6 billion dollars, comprising between 6 – 10% of the total health care expenditure of the US [9, 10]. Obese individuals had 36% higher annual health care costs than non-obese individuals [11]. Type 2 Diabetes (T2DM) has the common characteristic of obesity or being overweight. In addition, the researchers found that while two out of every 1,000 normal weight people had been diagnosed with diabetes, some 18 out of 1,000 obese people had the disease and there was a 41% increase in the incidence of diagnosed diabetes during that time. Researchers confirmed that the more fat tissue a person has the less sensitive that person becomes to insulin. Therefore a greater amount of insulin is required to maintain the body's regulation of blood glucose levels. Fat cells release a protein that leads to the development of T2DM [17]. Obesity prevalence of the pre-diabetic and diabetic conditions is more common in certain subgroups of the population. For African Americans, the prevalence of obesity is high, particularly African American women. The risks of morbidity and mortality associated with diabetes poses serious problems African American women as they affected by obesity related comorbidities disproportionately [14]. Although prevalence rates of obesity and diabetes have reached epidemic proportions in the African American population, the relationship between obesity and hospital health care use, cost and length of stay has received limited attention and failed to provide consistent results. Even though obesity is one of the biggest drivers of preventable chronic diseases and healthcare cost in the United States, obesity rates continue to grow. Taking account of culture and social economic factors, this study serves as a model for future studies on hospital length of stay and health care cost in high risk populations of primary diseases with comorbidities. The study provides a baseline for obese African American women with T2DM. The study design is a retrospective, correlation, quantitative analysis on lengths of hospital stay and cost among adult African American women categorized according to their weight status with T2DM. This study will be driven by the following four research questions and associated statistical hypotheses: Research Question 1 (RQ1). Is there a relationship between individual health factors of interest (obesity and diabetes) and hospital length of stay (LOS) among African American females? Research Question 2 (RQ2). Is there a relationship between individual health factors of interest (obesity and diabetes) and hospital costs among African American females? Research Question 3 (RQ3). How co-morbidities and life factors are related to individual health factors of interest (obesity and diabetes)… Advisors/Committee Members: Haque, Syed (chair), Srinivasan, Shankar (co-chair), Shabata, Masayuki (internal member), School of Health Professions.

▼ The overall goal of the project was to identify the factors and costs associated with Colon Cancer patients in terms of mortality, length of stay and costs in different types of clinical settings across the United States. Accordingly this research study utitlized the datasets for 2008 to 2010 available from the Nationwide Inpatient Sample (NIS) database with hospitalization characteristics of patients admitted with Colon Cancer as the principal diagnosis. Some of the important results found in this study were: Between 2008 and 2010 the age and population adjusted incidences and the hospital discharges both decreased significantly which is a promising trend speaking well of the state of health care in the United States as also possibly due to the effectiveness of nutritional counseling, patient education, screening for men aged 50 and above. It was found that while the total number of colon cancer patient discharges decreased significantly between 2008 to 2010, the Total Charges however significantly risen up between 2008 and 2010. The mean charges increased by nearly 8 %. The number of discharges across the various hospital types and their locations across the United States as shown above revealed that those large hospitals in metropolitan regions and those that are private not-for-profit have more discharges compared to the other types. Patients who are uninsured and those on Medicaid (low income) are more in number over the years 2008 to 2010 as compared to those on Medicare and Private Insurance which have decreasing trends. It was found that the mean and median length of stay of colon cancer patient discharges remained more or less the same between 2008 and 2010. It was found that the number of in-hospital mortality or deaths significantly reduced between 2008 and 2010. Alongside Home Health Care increased while discharges to another hospital also decreased (with a smaller decrease in discharge to another institution such as rehab facility and nursing home). The number of in-hospital deaths has a decreasing trend in the number of deaths over the years 2008 to 2010. Southern United States has more (nearly 2 times) in-hospital deaths compared to the other regions in all the 3 years. This study seems to indicate that mortality is positively correlated with the total costs and this may be due to a significant admission source is from emergency department. The in-hospital mortality prediction model above revealed significant risk for patients with hypertension (nearly 100%) and with obesity (62 % more) Patients with white ethnicity have lower risk of dying in the hospital compared to the other ethnicities which all have similar odds ratio intervals. This research study was limited to the datasets available from the Nationwide Inpatient Sample (NIS) database with hospitalization characteristics of patients admitted with Colon Cancer as the principal diagnosis. A similar large scale dataset based future study is indeed warranted to analyze demographic and hospital based outcomes… Advisors/Committee Members: Haque, Syed (chair), Srinivasan, Shankar (internal member), Mital, Dinesh (internal member), School of Health Related Professions.

▼ 

Background: Despite the fact that cervical cancer is known to be a preventable cancer, it remains one of the major causes of cancer-related deaths in females. A number of studies have attributed differences in clinical outcomes of cervical cancer to several factors such as stage at presentation, treatment pattern, and socioeconomic status. However, the association of specific genetic alterations with differences in clinical outcomes remains largely unexplored. Objectives: The initial research purpose was to identify the most common oncogene and tumor suppressor gene in cervical cancer with mutations and copy number alterations (CNAs). The focused research purpose was to examine the association of the identified oncogene and tumor suppressor gene with clinical outcomes and racial differences. Methodology: This study made use of the Cancer Genome Atlas (TCGA) database. The TCGA cervical cancer data were submitted between 2011 and 2014. The two genomic profiles used were mutation data and CNA data. The Fisher’s exact and chi-square tests were used to test for associations between the categorical variables. Logistic regression analysis was used to quantify the strength of associations.   Results: There were 309 cervical cancer cases. Phosphatidylinositol3-Kinase Catalytic Subunit Alpha (PIK3CA) and Phosphatase and Tensin Homolog (PTEN) genes were identified as the most common oncogene and tumor suppressor gene respectively. 63 patients had mutations in PIK3CA or PTEN or both, and 70 patients had CNAs. The ORs (Exp(B)) of death and tumor recurrence for patients with mutations were 3.300(1.625– 6.700) and 2.461(1.120–5.407) respectively. The ORs of death and tumor recurrence for patients with CNAs were 2.316(1.282–4.186) and 2.383(1.228–4.624) respectively. The ORs for CNA positive for the Black race compared to White race was 2.378(1.137–5.452). Conclusions: Genetic alterations in PIK3CA or PTEN or both are associated with a higher risk of cervical cancer mortality and tumor recurrence. These genes can be explored as therapeutic targets to improve cervical cancer treatment. High prevalence of CNAs in African American women could be due to the fact that a larger percentage presented at a later stage as stages III and IV are significant predictors of the presence of CNAs in these genes.

Advisors/Committee Members: Coffman, Frederick (chair), Srinivasan, Shankar (internal member), Mitrofanova, Antonina (internal member), School of Health Professions.

▼ 

Background: Celiac disease is a genetic autoimmune disease affecting people of all ages that results in small intestine enteropathy and is caused by the permanent intolerance to gliadin and glutenin, two proteins found in gluten containing grains. Celiac disease is considered to be a clinical chameleon. The disease can also be asymptomatic. Average prevalence of celiac disease in the population is one out of 100 people with data indicating the risk may be as high as 22% for those with first-degree relatives that have the disease. Research suggests 83% of people with celiac disease may be undiagnosed and the average duration for diagnosis is 10 years. Data indicates there is a lack of consensus regarding methodology used to diagnose celiac disease and poor knowledge of associated diseases and symptomatology. A review of the literature determined a celiac disease risk estimation and decision-making expert system including signs, symptomatology, manifestations and associations, with serology and histology based on the Mayo Clinic algorithm, using Exsys Corvid Software, did not currently exist. Method: A new clinical decision support system (CDSS) was developed using Exsys Corvid for expert analysis. The CDSS was divided into symptoms and manifestations with 80 points of navigation, and a serology section, and was validated by 13 experts in the field of celiac disease using a 10 statement, 5-point Likert scale. Results: This scale was analyzed using Cronbach’s alpha reliability coefficient, which was calculated using SPSS. Cronbach’s alpha revealed good internal consistency and reliability with a result of 0.813. One-hundred percent of the experts agreed with the system and that the CDSS is capable of guiding a healthcare professional through the diagnostic process, contains an accurate list of symptoms based on the clinical literature, can foster improved awareness and education about celiac disease, and that there is a need for this system. Over 90% agreed the system is a good tool for training medical students or residents. Conclusion: A celiac disease risk estimation and decision-making expert system was successfully developed and evaluated by medical professionals, with 100% agreeing that this CDSS is medically accurate and can guide healthcare professionals through the diagnostic process.

Advisors/Committee Members: Srinivasan, Shankar (chair), Mitrofanova, Antonina (internal member), Coffman, Fredrick (internal member), School of Health Professions.

▼ 

With the push for national EHR adoption and the subsequent increase in meaningful use of HIT applications, the healthcare industry has sought to realize reduced cost, increased safety, and improved patient outcomes. In an effort to evaluate the goal of improved patient outcomes, this study examines the effect of HIT adoption stage on the length of stay (LOS) for children admitted with an asthma diagnosis. Asthma is a chronic disease affecting millions of children each year, and has significant health, monetary, and emotional costs. As asthma is in the top three of most common conditions requiring hospital admissions for children and that nearly 50% of inpatient pediatric patients are covered by Medicaid, improving quality outcomes for this condition has large implications across the healthcare delivery system. Using comparisons from the KID 2009 and 2012 datasets, the differences between mean LOS for pediatric asthma patients between stages of adoption of Health IT as measured by the EMRAM scale are statistically significant at the p<.05 level, demonstrating that increased use of Health IT has lowered the mean length of stay for this population. Thus, the utilization of a medical best practice, here the adoption of Health IT, resulted in shorter hospital stays and thus cost savings, in this defined pediatric patient population. While further studies examining Health IT implementation in other patient populations are necessary, these results demonstrate that the implementation of Health IT can lead to both better standards of care and lower healthcare costs, which should be of significant interest to those charting the future course of healthcare and healthcare reimbursement in this country.

Advisors/Committee Members: Coffman, Frederick (chair), Srinivasan, Shankar (internal member), King, Leslie (outside member), School of Health Professions.

▼ BACKGROUND: Hypo- and hyperthyroidism are the most common types of autoimmune diseases of the thyroid gland. Although the prevalence of overt hypo- and hyperthyroidism is 0.3% and 0.5%, respectively, the majority of patients with these disorders suffer from cardiovascular complications, which are considered to be a significant risk of mortality. The thyroid disorders and their complications affect patient quality of life and life spans, and elevate the government’s economic burdens regarding health care. The objective of the present study is to highlight the similarities and differences of hypo- and hyperthyroidism in terms of risk factors related to hospitalization outcomes such as mortality, length of stay, and total medical charges when there is a presence of cardiovascular and other complications. METHOD: The study implemented a cross-sectional design to achieve the primary objectives. Data were downloaded and extracted, with permission, from Nationwide Inpatient Sample (NIS). A total of 721,958 patients with hypo- and hyperthyroidism were admitted to hospitals in the United States in 2012. The collected data included patient demographic characteristics, such as age, gender, race, insurance type, and income. Patient medical information included the number of medical procedures, chronic diseases, co-morbidities, and the type of thyroid disorder. Statistical Package for the Social Sciences (SPSS) version 22 was used to analyze the data of the present study, and all outcomes with a p-value less than 0.05 were found to be significant. Multinomial logistic regression and multiple linear regressions (the dummy method) were the appropriate statistical tests to determine the predictors of the study outcomes. RESULTS: A descriptive analysis of the present study revealed the highest incidences of thyroid disorders to be in those who were older than 80 years of age (29.3%), white (76.7%), female (74.9%), on Medicare (68.2%), and who had a household income in the 25th percentile (27.1%). Patient medical information showed the highest comorbidities to be hypertension (63.8%), fluid-electrolyte disorders (29.1%) and uncomplicated diabetes (24.4%). The incidence of mortality for patients with thyroid diseases was 2.4%. The mean (± SD) length of hospital stay and total medical charges were 5.06 (±6.113) days and $41829.47 (±60920.47), respectively. There was a higher prevalence of hypothyroidism than hyperthyroidism (97% vs. 3%). Overall mortality showed a higher incidence of hypothyroidism than of hyperthyroidism (2.4% vs. 1.75%). The incidence of mortality increased with cardiovascular complications, to 5.42% vs. 4.87% for congestive heart failure (CHF) and 2.47% vs. 1.99% for hypertension (HT), for patients with hypo- and hyperthyroidism, respectively. Risk factors for patients with hyperthyroidism related to length of stay were paralysis, weight loss, pulmonary circulation, fluid and electrolyte disorders, age, neurological disorders, coagulopathy, psychosis, and the number of procedures. Risk factors of length of stay… Advisors/Committee Members: Srinivasan, Shankar (chair), Mital, Denish (internal member), Manchiano, Mario (outside member), School of Health Professions.

▼ 

Radiation oncology has made great strides forward specifically in the treatment of lung cancer. However, these advances have themselves delivered new questions that clinicians face when attempting to treat tumors in the lungs. The first of which is how to best deliver an increasing radiation dose to a small moving target. The second is how best to estimate and predict the damage to healthy lung tissue as a consequence of these higher doses. Clinicians and academics from around the country have tabulated data, the purpose of which is to assess the risk of radiation damage to their patients during and after treatment. The consensus among these various groups is that the risk is best assessed by two or three volumetric data points. These dose indices are believed to allow clinicians to better assess toxicity endpoints in the lungs. The literature is rich with this guidance. However, that same literature search will also reveal that there is little to no data that focuses on the changes that occur in the previously mentioned evaluation metrics during respiration. The “V’s” in the V5 and the V20 are incorrectly assumed constant and unchanging. This retrospective analysis of 10 lung cancer patients shows that those clinically used metrics of evaluation that are treated as static numbers are in fact dynamic. It shows the degree to which these volumetric numbers vary from what is currently accepted. And it presents a more stable, mass-based alternative to volumetric metrics that may be more suited to assessing dose to healthy lung tissue during radiation therapy due to its stability throughout the patient’s breathing cycle. These mass-based alternative metrics are derived from each patient’s own lung volume using novel techniques involving the CT Hounsfield units. Yet, through ANOVA and two sample t-tests they show statistical significance in their difference from the volume in a rate of change analysis. The mass metrics also present more stability in their rate of change via one sample t-test and also exhibit lower standard deviations in all 10 patient’s breathing cycle and therefore has the potential to replace the current metrics for assessing radiation toxicity.

Advisors/Committee Members: Gohel, Suril (chair), Srinivasan, Shankar (internal member), Coffman, Frederick (internal member), School of Health Professions.

▼ BACKGROUND: Type 1 diabetes (T1D) is a complex autoimmune disease resulting in the destruction of β-cells encompassed by a combination of genotype and phenotype etiologies. With the etiological differences, high-dimensional multi-omics data, and stochastic components of T1D, a data-driven unsupervised machine learning and topology-based approach may identify new T1D endotypes that might go undetected with classical statistical approaches. Unsupervised machine learning, and topology-based approaches have found subtypes of various other diseases and disorders. Clustering techniques play a pivotal role in various elements of data analysis. They can provide important clues to the structure of data sets, signifying results and hypotheses of the underlying pathogenesis. METHOD: This work builds upon published TEDDY results. TEDDY has shown several single nucleotide polymorphisms (SNPs) can distinguish IAA-only from GADA-only as the first appearing IA and early exposures (infectious episodes) influence both in different ways, depending on the genetic factors. This would strongly suggest there are at least two or more different endotypes. What is unknown is the specific biological pathways that would explain the observable diversity in IA phenotypes, first appearing and progression. Unsupervised cluster and topology based analytical analyses of the T1D cases may distinguish phenotypes and help generate hypotheses regarding the biological pathways. Hierarchical cluster analysis was used for this particular analysis. We performed various analyses on the data: the first clustering with eight agglomerative hierarchical clustering methods; ward.D, ward.D2, average, complete, single, mcquitty, median, and centroid. We additionally performed k-means clustering, model based clustering, and topological data analysis. RESULTS: This was an exploratory study conducted to investigate the classification of T1D patient populations into distinct endotypes through procedures that utilize unsupervised machine learning techniques (hierarchical) and external validation through k-means clustering, model based clustering, and topological data analysis (TDA). The research analyzed data from a case-control cohort of genetically at risk study participants from the TEDDY study to explore the possibility of T1D endotypes. These study participants, enrolled from birth, carry HLA-susceptibility genotypes for development of islet autoantibodies (IA) and T1D. A novel exploratory approach to classify disease endotypes is presented in this study. By means of hierarchical clustering methods and exploration, the results of this study suggest that classification of T1D patient populations is plausible. This study identified three distinct clusters of T1D diagnosed patients among genetically at risk individuals who carry HLA-susceptibility genotypes for development of islet autoantibodies (IA) and T1D. CONCLUSION: While this exploratory study has limitations, the novel methodical approach taken to identify possible endotypes through clustering can be used… Advisors/Committee Members: Srinivasan, Shankar (chair), Gohel, Suril (internal member), Parikh, Hemang (outside member), School of Health Professions.

▼ 

Depression has been cited by multiple sources to worsen outcomes of patients with melanoma. There have been exhaustive studies that have provided correlative and implicative evidence that comorbid depression leads to worse outcomes for melanoma patients. These reasons include poor adherence to follow-up care, initial diagnosis occurring at a later stage, among others. However, there have been few studies that have been able to quantify these relationships. This study quantified these observations via retrospective cohort data and found that melanoma patients with depression and higher PHQ4 scores presented to their healthcare providers with higher frequency and that melanoma patients with depression were more likely to be unmarried and white. However, no difference was found in regards to stage of malignancy at time of initial diagnosis among melanoma patients with and without depression. Therefore, this study identifies an inefficiency in healthcare provided to this subset of melanoma patients and advises healthcare providers to consider providing screening or referrals to specialists.

Advisors/Committee Members: Coffman, Frederick (chair), Srinivasan, Shankar (internal member), Gohel, Suril (internal member), School of Health Professions.

▼ BACKGROUND: Drug abuse has been on the increase over the last few years, contributing to the healthcare cost. An understanding of the overall impact of drug abuse hospitalizations is essential in combatting the drug abuse epidemic. OBJECTIVE: The objective of this study is to examine hospitalization outcomes of total charges, and length of stay, among other elements associated with drug abuse comorbidity patients. The study will compare drug abuse comorbidity patients with non-drug abuse admission. The focus is on patients that were discharged in the United States between 2010 and 2014. Drug abuse comorbidity increases the intricacy of hospitalized patients; it is necessary to analyze the outcomes. The Center for Medicare and Medicaid (CMS) also implemented a value-based care model which allows healthcare providers, including hospitals and physicians, to be paid based on patient hospitalization outcomes. Therefore an understanding of this outcome is necessary for payment and resource allocation. METHOD: This study utilized the National (Nationwide) Inpatient Sample (NIS) for the years 2010 to 2014. The data source is an inpatient hospitalization dataset produced every year. The NIS is a publicly available all-payer inpatient health care dataset with national estimates of hospital inpatient stays. NIS collects data from more than 7 million hospital stays each year. It is estimated to be collecting more than 35 million hospitalizations nationally. NIS is a Federal-State-Industry partnership sponsored by the Agency for Healthcare Research and Quality (AHRQ). In this retrospective study, we demonstrated the estimation of inpatient outcomes for total charges and length of stay. The SPSS statistical analysis software was used to analyze the data. Various descriptive and inferential analysis was performed on the filtered data sets above for the years 2010 to 2014. Results of the outcome analysis that had a p-value less than 0.05 were noted to be significant. RESULTS: Drug abuse comorbidity cases within the five years were 2,258,235. Descriptive analysis showed that the prevalence of drug abuse comorbidity to be among males (58%), and they were more likely to be admitted compared to females (42%). This population, the median age at admissions, was 42 for males, and 40 for female. The average hospitalization length of stay was 4.5 days for non-drug abuse and 5.5 days for drug abuse comorbidity (P<0.001). Most drug abuse comorbidity hospitalization cases were charged to government-related insurance Medicaid (36.7%), Medicare (22.6%), and Private (18.2%), Self-pay (15.1% and other or unknown insurance (5.4%) P<0.001. Mean charges for drug abuse comorbidity (3.6% of population) was 36,735.98 while non-drug abuse cases (96.4% of study population) was 35,200.85 P< 0.001. The mean charges were highest in the Midwest 13,500.00 for non-drug abuse and 14,000.00 for those with drug abuse comorbidity on record. The lowest charges of 12,900 for drug abuse comorbidity and 13,300 for non-drug abuse were recorded in the Northeast.… Advisors/Committee Members: Srinivasan, Shankar (chair), Suchismita, Ray (internal member), Hunt, Thomas J (internal member), School of Health Professions.

▼ 

Methicillin-resistant Staphylococcus aureus (MRSA) is the leading cause of antibiotic resistance related mortality in surgical patients. Effective prediction of MRSA and MRSA-related SSI would facilitate the prophylactic use of appropriate antibiotics or application of other prevention techniques, which have been shown to improve clinical outcomes. While there is a range of factors that have been shown to increase the risk of MRSA-related infections, research is less clear on the best approaches to developing predictive models for incorporation into a clinical decision support system. This study compared two common modeling approaches — logistic regression (LR) and artificial neural networks (ANN) — for the prediction of MRSA infection and MRSA-related SSI in patients undergoing major surgical procedures (MSPs) in the United States. The data source for analysis is the National Inpatient Sample, which contains approximately 7 million discharges each year. A descriptive analysis was performed to identify potential predictors for each of three research hypotheses and ANN and LR models were developed and evaluated for the prediction of: (1) MRSA infection in patients undergoing MSPs; (2) MRSA-related SSI in patients undergoing MSPs; and (3) MRSA-related SSI in patients with S. aureus infection. The ANN model performed best for Hypothesis 1, with an AUC of 0.87, sensitivity of 0.86 and specificity of 0.74; the LR model achieved an AUC of 0.85, sensitivity of 0.79 and specificity of 0.75. For Hypothesis 2, the ANN model achieved an AUC of 0.86, sensitivity of 0.73 and specificity of 0.87; the LR model achieved an AUC of 0.85, sensitivity of 0.77 and specificity of 0.76. For Hypothesis 3, the ANN model achieved an AUC of 0.67, sensitivity of 0.57 and specificity of 0.67; the LR model achieved an AUC of 0.68, sensitivity of 0.61 and specificity of 0.64. This study assessed the feasibility of LR and ANN for the prediction of MRSA-related infections in surgical patients using a range of demographic, clinical, procedural, and hospital-related factors. The results showed that both algorithms are effective modeling approaches with reasonable sensitivity and specificity and suggest that a clinical decision support tool based on either model could be informative in clinical practice. ?

Advisors/Committee Members: Srinivasan, Shankar (chair), Coffman, Frederick (internal member), Gohel, Suril (internal member), School of Health Professions.

▼ 

Background: Chronic Kidney disease is recognized as a significant health problem affecting about 14.8% of the US population. CKD is the ninth leading cause of death in the united states . CDC projected the prevalence of CKD for 65 and older to slightly increase to 37.8 %.4 CKD not only accounts for more death than prostate and breast cancer put together but also contribute to other diseases which increase the probability of higher prevalence in heart disease and . In a study that focused on the future burden of CKD, they found that by 2030 adults of 30 years or older may increase from 14.8% to 16.7%.Methods: The data were obtained from the Nationwide Inpatient Sample (NIS) and were used to identify the relationship between length of stay, total charges, mortality and CKD. After merging the data, a total sample of 660,663 out of 30,931,761 discharge records of patients were diagnosed with chronic kidney disease. SAS Enterprise was used to perform descriptive and inferential analysis. RESULTS: The results showed that total charges, length of stay and mortality for CKD patients increases with the presence of comorbidities. Also, patients between the age of 0-19 and have hypertension their risk of developing CKD increases by 18x. Patients in their 20s and has hypertension are at a risk of developing CKD by almost 3 times when they reach 50 if they do not control their hypertension. Also, patients with complicated diabetes their risk increases by 8x where anemia at 4x. Conclusion: In this study comorbidities across all age groups such as diabetes, hypertension, obesity, anemia, and congestive heart disease increases the likelihood of developing CKD. Patients with these risk factors should follow guidelines to control their condition to avoid developing CKD.

Advisors/Committee Members: Srinivasan, Shankar (chair), Mital, Dinesh (internal member), Coffman, Frederick (internal member), School of Health Professions.