In this research, we apply data mining techniques to a nationally-representative expenditure data from the US to predict very high-cost patients in the top 5 cost percentiles, among the general population. Samples are derived from the Medical Expenditure Panel Survey’s Household Component data for 2006-2008 including 98,175 records. After pre-processing, partitioning and balancing the data, the final MEPS dataset with 31,704 records is modeled by Decision Trees (including C5.0 and CHAID), Neural Networks. Multiple predictive models are built and their performances are analyzed using various measures including correctness accuracy, G-mean, and Area under ROC Curve. We conclude that the CHAID tree returns the best G-mean and AUC measures for top performing predictive models ranging from 76% to 85%, and 0.812 to 0.942 units, respectively. Among a primary set of 66 attributes, the best predictors to estimate the top 5% high-cost population include individual’s overall health perception, history of blood cholesterol check, history of physical/sensory/mental limitations, age, and history of colonic prevention measures. It is worthy to note that we do not consider number of visits to care providers as a predictor since it has a high correlation with the expenditure, and does not offer a new insight to the data (i.e. it is a trivial predictor). We predict high-cost patients without knowing how many times the patient was visited by doctors or hospitalized. Consequently, the results from this study can be used by policy makers, health planners, and insurers to plan and improve delivery of health services.

This study explores healthcare coverage disparity using a quantitative analysis on a large dataset from the United States. One of the objectives is to build supervised models including decision tree and neural network to study the efficient factors in healthcare coverage. We also discover groups of people with health coverage problems and inconsistencies by employing unsupervised modeling including K-Means clustering algorithm. Our modeling is based on the dataset retrieved from Medical Expenditure Panel Survey with 98,175 records in the original dataset. After pre-processing the data, including binning, cleaning, dealing with missing values, and balancing, it contains 26,932 records and 23 variables. We build 50 classification models in IBM SPSS Modeler employing decision tree and neural networks. The accuracy of the models varies between 76% and 81%. The models can predict the healthcare coverage for a new sample based on its significant attributes. We demonstrate that the decision tree models provide higher accuracy that the models based on neural networks. Also, having extensively analyzed the results, we discover the most efficient factors in healthcare coverage to be: access to care, age, poverty level of family, and race/ethnicity.

Back pain is the 6th most costly condition in the United States and is responsible for the most workdays lost. Approximately 33 million American adults suffered from back and neck problems in 2005. The societal cost of chronic back pain (CBP) has not been calculated from a US perspective. Longitudinal data files from Panels 12, 13, and 14 of the Medical Expenditure Panel Survey (MEPS) were used to estimate excess direct (ambulatory visits, inpatient admissions, emergency room visits, and prescription medication) costs and indirect (lost productivity) costs for persons 18 years and older reporting CBP compared to those not reporting back pain. Persons were included in the CBP group if they reported back pain (ICD-9-CM codes 720, 721, 722, 723, 724, 737, 805, 806, 839, 846, 847) in at least 3 consecutive interview rounds. The complex sampling design of MEPS was taken into account to get accurate national estimates. All costs were adjusted to 2011 using Consumer Price Indices. All mean costs were computed using age-stratified regression models, after adjusting for demographic and clinical covariates. Utilization of provider-based complementary and alternative medicine (CAM) among CBP patients was studied, and differences in costs between CAM users and non-users examined. Based on this analysis, the prevalence of CBP in the adult US population was estimated to be 3.76%. Total all-cause costs for CBP patients were estimated to be $187 billion over 2 years (direct costs = $176 billion, indirect cost = $11 billion). Overall estimates of excess costs of CBP over 2 years per person for direct medical costs were $37,129 ($25,273 vs. $48,984; p<0.001). This breaks down to $11,711 ($14,929 vs. $3,219; p<0.001) for ambulatory visits; $3,560 ($6,514 vs. $2,914; p<0.001) for inpatient admissions; $300 ($690 vs. $390; p<0.001) for emergency department visits; and $19,849 ($23,873 vs. $4,024; p<0.001) for prescription medications. Excess indirect costs for CBP patients were $1,668 ($2,329 vs. $661; p<0.001). Thirty-seven percent of CBP patients reported at least one CAM visit. There was no significant difference in overall costs between CAM users and non-users. The high cost of chronic back pain in the US population has potential implications for prioritizing policy, and in attempting to improve care and outcomes for these patients. / text

For employees to optimally perform at work, it is important that they are healthy. The employee is under constant work pressures that affects their health behaviour. The aim of this research is to look at the evident health risks of employees working within a financial institution, to analyze how physical activity influences these health risk factors and their medical costs. In this study a sample of 9860 self selected employees, aged between 18 and 64 (35.3 ± 10.7 years), was used. These employees are spread over all the provinces of South Africa and from all racial groupings. Differentiation was made between employees who were on chronic medication and those who were not. The Health Risk Assessment (HRA) questionnaire developed and provided by the medical aid of the institution was used as the analysis tool. Medical expenditures of the sample group were also provided by the medical aid and investigated. A national network of registered Biokineticists administered the implementation of the HRA, based on a set protocol. ANOVA was used for statistical data analysis - providing descriptive and summarising statistics. One-way analysis of variance was used to determine relationships between variables. It is clear from the descriptive data that the tendencies of selected health risks were high. The results also show that 37.6% (Diastolic) and 47.87% (Systolic) of the sample group comply with the normal borders of blood pressure. With regards to BM3, 32.3% were overweight and 25.3% within the boundaries of obesity. The average cholesterol of the group is 4.4mmol.L-l. The Physical activity levels were determined using an activity algorithm developed by the medical aid of the institution (described in detail in the thesis). Results show a low level of physical activity index (7.18 + 3.05) within the sample group. No statistical significance could be found between physical activity levels and medical expenditure, although those who are highly active seemed to have higher expenditure than those who are inactive. / Thesis (M.A. (Human Movement Science))--North-West University, Potchefstroom Campus, 2009.

For employees to optimally perform at work, it is important that they are healthy. The employee is under constant work pressures that affects their health behaviour. The aim of this research is to look at the evident health risks of employees working within a financial institution, to analyze how physical activity influences these health risk factors and their medical costs. In this study a sample of 9860 self selected employees, aged between 18 and 64 (35.3 ± 10.7 years), was used. These employees are spread over all the provinces of South Africa and from all racial groupings. Differentiation was made between employees who were on chronic medication and those who were not. The Health Risk Assessment (HRA) questionnaire developed and provided by the medical aid of the institution was used as the analysis tool. Medical expenditures of the sample group were also provided by the medical aid and investigated. A national network of registered Biokineticists administered the implementation of the HRA, based on a set protocol. ANOVA was used for statistical data analysis - providing descriptive and summarising statistics. One-way analysis of variance was used to determine relationships between variables. It is clear from the descriptive data that the tendencies of selected health risks were high. The results also show that 37.6% (Diastolic) and 47.87% (Systolic) of the sample group comply with the normal borders of blood pressure. With regards to BM3, 32.3% were overweight and 25.3% within the boundaries of obesity. The average cholesterol of the group is 4.4mmol.L-l. The Physical activity levels were determined using an activity algorithm developed by the medical aid of the institution (described in detail in the thesis). Results show a low level of physical activity index (7.18 + 3.05) within the sample group. No statistical significance could be found between physical activity levels and medical expenditure, although those who are highly active seemed to have higher expenditure than those who are inactive. / Thesis (M.A. (Human Movement Science))--North-West University, Potchefstroom Campus, 2009.

This study explores healthcare coverage disparity using a quantitative analysis on a large dataset from the United States. One of the objectives is to build supervised models including decision tree and neural network to study the efficient factors in healthcare coverage. We also discover groups of people with health coverage problems and inconsistencies by employing unsupervised modeling including K-Means clustering algorithm. Our modeling is based on the dataset retrieved from Medical Expenditure Panel Survey with 98,175 records in the original dataset. After pre-processing the data, including binning, cleaning, dealing with missing values, and balancing, it contains 26,932 records and 23 variables. We build 50 classification models in IBM SPSS Modeler employing decision tree and neural networks. The accuracy of the models varies between 76% and 81%. The models can predict the healthcare coverage for a new sample based on its significant attributes. We demonstrate that the decision tree models provide higher accuracy that the models based on neural networks. Also, having extensively analyzed the results, we discover the most efficient factors in healthcare coverage to be: access to care, age, poverty level of family, and race/ethnicity.

This study explores healthcare coverage disparity using a quantitative analysis on a large dataset from the United States. One of the objectives is to build supervised models including decision tree and neural network to study the efficient factors in healthcare coverage. We also discover groups of people with health coverage problems and inconsistencies by employing unsupervised modeling including K-Means clustering algorithm. Our modeling is based on the dataset retrieved from Medical Expenditure Panel Survey with 98,175 records in the original dataset. After pre-processing the data, including binning, cleaning, dealing with missing values, and balancing, it contains 26,932 records and 23 variables. We build 50 classification models in IBM SPSS Modeler employing decision tree and neural networks. The accuracy of the models varies between 76% and 81%. The models can predict the healthcare coverage for a new sample based on its significant attributes. We demonstrate that the decision tree models provide higher accuracy that the models based on neural networks. Also, having extensively analyzed the results, we discover the most efficient factors in healthcare coverage to be: access to care, age, poverty level of family, and race/ethnicity.

In this research, we apply data mining techniques to a nationally-representative expenditure data from the US to predict very high-cost patients in the top 5 cost percentiles, among the general population. Samples are derived from the Medical Expenditure Panel Survey’s Household Component data for 2006-2008 including 98,175 records. After pre-processing, partitioning and balancing the data, the final MEPS dataset with 31,704 records is modeled by Decision Trees (including C5.0 and CHAID), Neural Networks. Multiple predictive models are built and their performances are analyzed using various measures including correctness accuracy, G-mean, and Area under ROC Curve. We conclude that the CHAID tree returns the best G-mean and AUC measures for top performing predictive models ranging from 76% to 85%, and 0.812 to 0.942 units, respectively. Among a primary set of 66 attributes, the best predictors to estimate the top 5% high-cost population include individual’s overall health perception, history of blood cholesterol check, history of physical/sensory/mental limitations, age, and history of colonic prevention measures. It is worthy to note that we do not consider number of visits to care providers as a predictor since it has a high correlation with the expenditure, and does not offer a new insight to the data (i.e. it is a trivial predictor). We predict high-cost patients without knowing how many times the patient was visited by doctors or hospitalized. Consequently, the results from this study can be used by policy makers, health planners, and insurers to plan and improve delivery of health services.

No description available.

In Healthy People 2010, mental health is listed as a major public health concern as evidenced by an alarming increase in the number of individuals who suffer from mental disorders. Mental disorders are a treatable public health condition. However, health disparities in the treatment of mental disorders are evident. The purpose of this study was to examine factors that affected health outcomes of persons with mental disorders. Two specific aims were addressed: Aim 1: to examine the relationships of population characteristics (predisposing factors and enabling resources), health behaviors (health services use and health practice); and health outcomes (physical health status and mental health status); Aim 2: to determine the differences in the usual source of care and health outcomes between individuals with self-reported mental disorders and individuals without mental disorders. This study was a secondary analysis of existing data collected from 2006 Medical Expenditure Panel Survey Household Component Consolidated file. A sample of U.S. civilian non-institutionalized adults (N=622) was grouped according to three self-reported health conditions: mental disorders (MD), physical illnesses (PI) and co-morbid mental disorders and physical illnesses (CM). This sample was predominantly male, White non-Hispanic and married; had a high school diploma, middle to high income, and private insurance; and preferred office-based clinics as the usual source of care, F(2,29)=5.94, p = .007. No statistically significant differences between groups in usual source of care (p=.069) and physical health status (p=.490) but there was a significant difference in mental health status (p=.001). Participants with CM had a poorer mental health status than those with PI and MD, F (2,619) =21.8, p= .000. The mental health status of individuals with PI was significantly better than that of participants with MD. Awareness of disparities in the usual source of care, health services use, and health outcomes among individuals with mental health conditions is imperative if barriers to care are to be eliminated. Innovative interventions pertinent to decreasing barriers to accessing health care and improving the health outcomes among individuals with MD must be tested. Advocating for mental health care policies that reduce health care services disparities among individuals with self-reported MD must be encouraged.