Exhaustive Chi-squared Automatic Interaction Detector Research Articles

Reactions to Resistance: The Role of Contextual Factors in Sex Offending

Upon encountering victim resistance during the commission of a sexual assault, an offender makes a decision as to how to react to the resistance via the use of varying degrees of coercive or noncoercive tactics. Exhaustive chi-squared automatic interaction detection (CHAID) analyses were performed on a sample of 369 criminal events to detect the significant situational and crime-specific factors, the pertinent interactions between these factors, and how these affect an offender's reaction to victim resistance depending on whether the victim is a child or an adult. Findings show that multiple interactions do occur and, moreover, factors affecting the offender's reaction to victim resistance are very different depending on whether the victim is a child or an adult. Planning and preparation variables, such as the offender's attack strategy, were found to be significant in the child victim model, whereas reactive variables, such as the type of victim resistance, were found to be significant in the adult victim model. The findings lend support to the social interac-tionist perspective and the importance of victim-offender dynamics.

Intersectionality and Disability Harassment

A possible interaction among the characteristics of disability, race, gender, and age was examined with respect to formal allegations of disability harassment. Using data from the National Equal Employment Opportunity Commission (EEOC) Americans With Disabilities Act (ADA) Research Project, the authors examined whether there was an interaction among disability, gender, age, race, and employer characteristics when considering the proportion of harassment versus other forms of discrimination allegations. Using Exhaustive Chi-Squared Automatic Interaction Detector (Exhaustive CHAID) analysis, the authors detected several interaction effects. They discovered unique clusters of characteristics that place certain groups at a very high and very low risk for experiencing disability harassment. The findings are discussed in the context of past and future research on intersectionality and workplace implications.

Gingival changes during pregnancy: III. Impact of clinical, microbiological, immunological and socio‐demographic factors on gingival inflammation

To identify predictor variables involved in exacerbated gingival inflammation associated with pregnancy. In this cohort study, 48 pregnant and 28 non-pregnant women without periodontitis were included. The pregnant women were evaluated in the first, second and third trimester and at 3 months postpartum, whilst the non-pregnant women were evaluated twice, with a 6-month interval. At each visit, clinical [plaque index (PlI) and gingival index (GI)], hormonal (salivary progesterone and estradiol), immunological [gingival crevicular fluid interleukin-1β, interleukin-6, tumour necrosis factor-α (TNF-α) and prostaglandin-E(2) ] and microbiological (periodontal pathogens culture) evaluations were performed. Statistical analysis was undertaken using exhaustive chi-square automatic interaction detection (exhaustive CHAID) to analyse the predictive value of the independent outcomes to develop pregnancy GI. PlI was the strongest predictor implicated in the GI throughout pregnancy and after delivery. During the second and third trimesters the presence of Porphyromonas gingivalis significantly contributed to the worsening of gingival inflammation. When compared with the non-pregnant group, significant differences were found in TNF-α amounts and concentrations and in the third trimester site-specific GI. Bacterial challenge to the gingival tissues, both quantitatively (PlI) and qualitatively (harbouring P. gingivalis) appears to affect the level of gingival inflammation observed during pregnancy.

Seroprevalence and risk factors associated with Visna/Maedi virus in semi-intensive lamb-producing flocks in northwestern Spain

A cross-sectional study was carried out to determine Visna/Maedi virus (VMV) seroprevalence and risk factors in semi-intensive lamb-producing flocks as a prelude to establishing a monitoring program in northwestern (NW) Spain. A total of 15,155 serum samples were taken from 78 commercial flocks and were submitted to an indirect VMV ELISA. Association between potential risk factors and seroprevalence at the flock level was assessed using a multivariable logistic regression model. A Generalized Estimating Equations (GEE) model and Exhaustive Chi-squared Automatic Interaction Detector (CHAID) were used to determine the seropositivity against VMV at the individual animal level. Individual seropositivity was 24.8% while 52.6% of the flocks examined had a true seroprevalence ≥1%. Flock size and introduction of new animals in the flock were significantly associated with seropositivity at the flock level. Flock size, sheep–goat contact, type of housing of lambs prior to weaning and age were significantly associated with individual VMV seropositivity. Confinement of lambs in preweaning lamb groups and high sheep–goat contact, regardless of the low number of goats per flock, were risk factors associated with individual VMV seropositivity, suggesting that these two factors are crucial for VMV control in semi-intensive lamb-producing flocks. These factors should be considered for developing more efficient strategies that will reduce the rate of VMV transmission.

Comparison of regression tree data mining methods for prediction of mortality in head injury

With this research, we sought to examine the performance of six different regression tree data mining methods to predict mortality in head injury. Using a data set consisting of 1603 head injury cases, we assessed the performance of: the Classification and Regression Trees (CART) method; the Chi-squared Automatic Interaction Detector (CHAID) method; the Exhaustive CHAID (E-CHAID) method; the Quick, Unbiased, Efficient Statistical Tree (QUEST) method; the Random Forest Regression and Classification (RFRC) method; and the Boosted Tree Classifiers and Regression (BTCR) method, in each case based on sensitivity, specificity, positive/negative predictive, and accuracy rates. Next, we compared their areas under the (Receiver Operating Characteristic) ROC curves. Finally, we examined whether they could be grouped in meaningful clusters with hierarchical cluster analysis. Areas under the ROC curves of regression tree data mining methods ranged from 0.801 to 0.954 ( p < 0.001 for all). In predicting mortality in head injury under the ROC curve, the BTCR method achieved both the highest area (0.954) and accuracy rate (93.0%), while the CART method achieved both the lowest area (0.801) and accuracy rate (91.1%). All of the regression tree data mining methods were clustered in the same grouping, but the BTCR method was at the origin of the cluster while the CART and QUEST methods produced results that were least like the others. The BTCR, demonstrating a 93.0% accuracy rate and showing statistically significantly differences from the others, may be a helpful tool in medical decision-making for predicting mortality in head injury.

Predicting the incidence risk of ischemic stroke in a hospital population of southern China: A classification tree analysis

Objective To determine the major risk factors and their interactions of ischemic stroke (IS) and to develop a classification tree model to predict the incidence risk of IS for a Chinese population. Methods Exhaustive Chi-squared Automatic Interaction Detection (Exhaustive CHAID) algorithm of classification tree method was applied to build a prediction model for the incidence risk of IS under the design of 1:1 matched case-control study. The statistics of misclassification risk was used to evaluate the fitness of the model. Results In the prediction model, six variables of physical exercise, history of hypertension, tea drinking, HDL-c level, smoking status and educational level were in turn selected as the predictors of IS incidence risk. In the subgroup of lacking of physical exercise, individuals who had history of hypertension would have a significantly higher IS risk (92%) than that of the ones who had no history of hypertension (64%). The misclassification risk estimate of the prediction model was 0.21 with the standard error of 0.02, indicating that 79% of the cases could be classified correctly based on current prediction model. Conclusions Lacking of physical exercise and history of hypertension are identified to be the prominent predicting variables of IS risk for a hospital population of southern China. Although CHAID analysis could provide detailed information and insight about interactions among risk factors of IS, we still need to validate our model and improve the vascular risk prediction for Chinese subjects in further studies.

Merit Determinants of ADA Title I Allegations Involving Discharge: Implications for Human Resources Management and Development

Nonparametric tests of proportion were performed comparing discharge resolutions (the most prevalent discrimination issue filed with the EEOC between 1992 and 2008) with resolutions of a composite group of the next most prevalent discrimination issues (reasonable accommodation, terms and condition of employment, harassment and intimidation, and hiring). Results indicated that when an employee files an allegation of discharge discrimination, the resolution status of that allegation is more likely to be without Merit. Two exhaustive chi-square automatic interaction detector (CHAID) analyses were performed, with the first including 44 impairment types (e.g., back, traumatic brain injury, learning disability) and the second including those 44 impairment types combined into six categories (e.g., physical, neurological, sensory, and behavioral impairments). Results of the first exhaustive CHAID proved quite unwieldy but indicated that behavioral disabilities have relatively low merit resolution rates when it comes to discharge allegations (i.e., there is less actual discharge discrimination occurring for persons with behavioral disabilities). One additional finding was that filing discharge discrimination under one of the alternate prongs of the ADA’s definition of disability (i.e., “regarded as a person with a disability,” “record of a disability,” or “association with a person with a disability”) was the main driver of merit resolution activity. Results from a second exhaustive CHAID indicated that in the case of discharge merit resolution activity, employer industry drives the alternate prongs, employee race drives physical impairments, and employer industry drives behavioral impairments. Implications for human resources management and development research and practice are addressed.

DATA MINING FOR OCCUPATIONAL INJURY RISK: A CASE STUDY

The management of occupational injury is of strategic importance in a company from the organizational, engineering and economic point of view. This work is an attempt to apply data mining techniques to data regarding accidents in a medium-sized refinery. Several techniques were adopted in order to identify the important relationships between risk level and immediate/root causes and corrective actions. As a Data Mining technique were tested: Negative Binomial Regression (NBR), Chi-Squared Automatic Interaction Detection (CHAID); Exhaustive CHAID; Classification And Regression Trees (CART); Quick, Unbiased, Efficient Statistical Tree (QUEST), Artificial Neural Network (ANN) and Neuro-Fuzzy Systems (FIS). The comparison carried out in this study shows through a real application the flexibility and advantages of using the neuro-fuzzy network, a typical soft computing tool. Using these innovative techniques to analyse injury data this study aims to: • obtain a classification of input data according to their importance and/or influence on the risk level in injuries; • assess how a variation in one or more pieces of input data can effect occupational injury and subsequently carry out a sensitivity analysis concerning the probability, the consequences and risks of the injurie events; The analyses carried out indicated important relationships between the variables, providing useful decision-making rules which can be followed when adopting measures for improvement.

A Decision Support Tool for Identifying Abuse of Controlled Substances by ForwardHealth Medicaid Members

Our objective was to use Wisconsin's Medicaid Evaluation and Decision Support (MEDS) data warehouse to develop and validate a decision support tool (DST) that (1) identifies Wisconsin Medicaid fee-for-service recipients who are abusing controlled substances, (2) effectively replicates clinical pharmacist recommendations for interventions intended to curb abuse of physician and pharmacy services, and (3) automates data extraction, profile generation and tracking of recommendations and interventions. From pharmacist manual reviews of medication profiles, seven measures of overutilization of controlled substances were developed, including (1-2) 6-month and 2-month "shopping" scores, (3-4) 6-month and 2-month forgery scores, (5) duplicate/same day prescriptions, (6) count of controlled substance claims, and the (7) shopping 6-month score for the individual therapeutic class with the highest score. The pattern analysis logic for the measures was encoded into SQL and applied to the medication profiles of 190 recipients who had already undergone manual review. The scores for each measure and numbers of providers were analyzed by exhaustive chi-squared automatic interaction detection (CHAID) to determine significant thresholds and combinations of predictors of pharmacist recommendations, resulting in a decision tree to classify recipients by pharmacist recommendations. The overall correct classification rate of the decision tree was 95.3%, with a 2.4% false positive rate and 4.0% false negative rate for lock-in versus prescriber-alert letter recommendations. Measures used by the decision tree include the 2-month and 6-month shopping scores, and the number of pharmacies and prescribers. The number of pharmacies was the best predictor of abuse of controlled substances. When a Medicaid recipient receives prescriptions for controlled substances at 8 or more pharmacies, the likelihood of a lock-in recommendation is 90%. The availability of the Wisconsin MEDS data warehouse has enabled development and application of a decision tree for detecting recipient fraud and abuse of controlled substance medications. Using standard pharmacy claims data, the decision tree effectively replicates pharmacist manual review recommendations. The DST has automated extraction and evaluation of pharmacy claims data for creating recommendations for guiding pharmacists in the selection of profiles for manual review. The DST is now the primary method used by the Wisconsin Medicaid program to detect fraud and abuse of physician and pharmacy services committed by recipients.

P aco2 and Neurodevelopment in Extremely Low Birth Weight Infants

To determine the relationship between PaCO2 in the first 4 days of life and neurodevelopment at 18 to 22 months. Stepwise regression and exhaustive CHAID (Chi-squared Automatic Interaction Detector) analyses were done for neurodevelopmental impairment (NDI), mental developmental index (MDI), and psychomotor developmental index (PDI) using clinical variables in combination with the maximum (max), time-weighted average (avg), and max-minimum (max-min) PaCO2 in 400 infants of 401 to 1000 g birth weight (BW). By regression, NDI predictors were male sex, non-Caucasian race, premature prolonged rupture of membranes (PPROM), lower BW, IVH 3 to 4, and lower 1-minute Apgar score. For lower MDI, predictors were male sex, non-Caucasian race, PPROM, IVH 3 to 4, sepsis, and higher max-min PaCO2. For lower PDI, predictors were male sex, PPROM, lower BW, IVH 3 to 4, sepsis, and higher avg PaCO2. By CHAID, the most important predictor of NDI was sex. For MDI, sex was most important followed by max-min PaCO2 >42 mm Hg in boys. For PDI, IVH was most important, followed by max-min PaCO2 >42 mm Hg for grade < or = 2 IVH. Extreme fluctuations in PaCO2 and higher max PaCO2 are associated with worse neurodevelopmental outcomes and may indicate either a greater severity of illness or contribution of PaCO2 to pathophysiology of adverse neurodevelopment.

Behaviour analysis of internet survey completion using decision trees

PurposeThe purpose of this paper is to explore teachers' behaviours in completing an internet survey using decision trees. Furthermore, to reduce the complexity of the decision trees, a statistical technique was used to decrease the number of input variables in the decision trees.Design/methodology/approachA dataset of 47,647 samples was used to build the decision trees. These samples were collected from an internet survey of teachers in Taiwan. The output of the decision trees was the answering time (the time taken to complete the internet questionnaire). Eight variables were selected as the inputs for the decision trees. Two techniques were employed to build the decision trees – the exhaustive chi‐squared automatic interaction detector (ECHAID) and classification and regression tree (CRT) analysis. To reduce the complexity of the decision models, factor analysis technique was used to decrease the data dimensions (number of input variables) and to obtain a simplified decision model. One‐way ANOVA was used to validate the effects of the dimension reduction.FindingsFrom the results of the factor analysis, a simplified decision tree is recommended using four input variables – teaching years, school level, sex and area. The classification accuracy of the simplified model is statistically equivalent to that of the original one, which used eight input variables.Originality/valueThe complexity of decision trees theoretically depends on the number of input variables. This study used a statistical technique to decrease the number of input variables and thereby reduce the complexity of the decision trees. A statistical technique was employed to validate that the classification accuracy is not statistically different between the original decision model and the simplified one. The decision models proposed in this paper can be applied in estimating the answering time for completing a questionnaire during an internet survey.

Which physicians have access to electronic prescribing and which ones end up using it?

This study examines the availability of electronic prescription and the utilization of e-prescribing by physicians in the US. Nationally representative data from the 2004-2005 Community Tracking Study Physician Survey were used to identify which subgroups of physicians have access to e-prescribing technology and which subgroups are using this technology more or less intensively. Exhaustive Chi-squared Automatic Interaction Detection (CHAID) was employed for statistical data segmentation. Results indicate that the rapidly increasing adoption of electronic prescription is diminished by relatively low physician utilization. E-prescription utilization was segmented among practice size and type. There were also differences in e-prescription use by age, gender, and ethnicity/race in some subgroups. Actual use of e-prescription was very low for female physicians in surgical specialties, psychiatry, and obstetrics/gynecology, and for Hispanic physicians in pediatrics, internal medicine, and family/general practice in solo/two physician practices, medical schools, and hospitals. Insights from segmentation analyses could be used to identify adoption barriers and to develop targeted interventions to accelerate the implementation of e-prescription systems in physician practices.

Using Decision Trees to Identify Tourism Stakeholders: The Case of Two Eastern North Carolina Counties

This paper explores stakeholder involvement in tourism planning, development, and management. For tourism planners to include stakeholders in the tourism planning process those stakeholders and their interests need to be identified. The research reported in this paper describes and applies an analytical technique that is not traditionally used to identify stakeholders. A questionnaire was developed and mailed to stakeholders in two rural communities in North Carolina. The data were analysed with an Exhaustive Chi-square Automatic Interaction Detection decision tree. From the results of the decision tree, stakeholder groups were identified in relation to their support for sustainable tourism development in their community.

The semi-automated classification of sedimentary organic matter in palynological preparations

The capture, analysis and classification of sedimentary organic matter in palynological preparations have been semi-automated. First, the morphological and textural discriminatory features used in classification schemes are measured using a computer-controlled stage and a digital camera mounted on a microscope in combination with Halcon image analysis algorithms. Second, the Exhaustive CHi-square Automatic Interaction Detector classification tree algorithm is applied to all feature measurements to establish their saliency as classification discriminators. Thirdly, the results of the classification tree algorithm are used to determine the inputs used by the actual classifier, which consists of a series of artificial neural networks (ANNs). The Gamma test (GT) is introduced as a tool to help facilitate the best use of limited data and to ensure that the ANNs are not over trained. The results show that the system developed is able to achieve an average correct classification rate of 87%. This is encouraging enough to prompt further research that could result in a commercially viable system. In the future, work will concentrate on refining the image capture component of the system and increasing the size of those databases that have been shown both empirically and by the GT to be too small to facilitate the construction of accurate classifiers.

Unmet Needs in Groups of Traditionally Underserved Individuals with HIV/AIDS: Empirical Models

SUMMARY Over the course of the HIV epidemic, the demographics of the populations of affected individuals have changed. Groups that traditionally have been underserved in systems of care have a number of unmet service needs. This article presents results based on data from 478 patients in five national demonstration projects which were funded to enroll individuals from traditionally underserved groups and to help them access services using different strategies. The participants in these programs had a high level of unmet need prior to enrolling in care. Data on client service needs were related to 17 indicators of traditionally underserved status including demographic characteristics and risk behaviors, using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector). Crack cocaine users with HIV/AIDS were more likely than other patient groups to have unmet service needs. Patients who were homeless or in precarious housing also were vulnerable. Results are discussed in terms of designing and evaluating innovative service models to close these service gaps.

Satisfaction with Services in Innovative Managed Care Programs for Groups of Traditionally Underserved Individuals with HIV/AIDS: Empirical Models

SUMMARY As the number of people with HIV/AIDS receiving services in managed care models increases, concerns over quality of care and satisfaction with services have grown. This article examined data from three national demonstration projects that were funded to enroll traditionally underserved individuals and provide innovative medical services in programs developing models appropriate for managed care funding. Assessments of patient satisfaction were related to indicators of traditionally underserved status including demographic characteristics, behaviors, and other risk factors using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector).Overall patient satisfaction levels with these programs were very high. Through the modeling methods, the groups most likely to experience the greatest program satisfaction are identified. In general, all groups were highly satisfied with the programs.

Satisfaction with Innovative Community and University Health Clinic Programs for Groups of Traditionally Underserved Individuals with HIV/AIDS: Empirical Models

SUMMARY As the demographics of the populations of affected individuals have changed, systems of care have needed to adapt to be responsive to client needs. This article examines client satisfaction data from seven national demonstration projects funded to enroll individuals from traditionally underserved groups and help them access services using different strategies. Data on client satisfaction ratings were related to indicators of traditionally underserved status, including demographic characteristics, behaviors, and other risk factors using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector). Client groups that were most likely to experience relatively higher and lower levels of satisfaction with services are identified. Overall, all client groups were highly satisfied with the innovative HIV/AIDS services received. The findings illustrate the success of these innovative HIV care models in being responsive and sensitive to the needs of their target populations.

Perceived Barriers to Receiving HIV Services in Groups of Traditionally Underserved Individuals: Empirical Models

SUMMARY Persons living with HIV/AIDS face many issues that make them highly vulnerable to a number of health and social problems. As the demographics of the epidemic have shifted in recent years, many members of traditionally underserved groups have encountered barriers to entering the services system. This article uses data from seven national demonstration projects funded to enroll persons with HIV/AIDS who tend to “fall through the cracks” and help them access needed services. Data on the initial perceptions of the participants about barriers to accessing services were related to 17 indicators of traditionally underserved status including demographic characteristics and behavioral variables using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector). Through the modeling methods, the groups most likely to experience a large number of barriers to service participation are identified. Having children needing care is particularly predictive of the level of barriers to care.

Service Needs and Factors Related to Quality of Life at Time of Service Enrollment Among Persons Living with HIV

ABSTRACT This article explores the relationships of HIV risk factors, service needs, and vulnerabilities to health-related quality of life in a sample of 1,371 participants newly enrolled into 13 innovative HIV/AIDS treatment model service demonstration projects. These projects targeted services to traditionally underserved populations. Eight distinct quality of life clusters of HIV patients were used in this analysis along with patient self-identified risk factors. The quality of life clusters were based on patient self-reported quality of life dimensions. The eight clusters were differentiated based on relative strengths and weaknesses in physical functioning, energy levels, and social functioning. Data on patient need-vulnerability factors and demographic characteristics were related to these eight clusters using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector). Through this method, the characteristics most likely to be associated with higher and lower levels of quality of life at the time of enrollment into services were identified. The results provide further support that quality of life assessment is a useful clinical tool for monitoring patient progress.

Key elements for implementing comprehensive health care models for persons with HIV: a stakeholder analysis.

A semistructured interview was conducted with 69 stakeholders in three university-based health care projects that were funded to provide an integrated continuum of care for persons living with HIV/AIDS. Data from the key informant interviews yielded composite indicators of familiarity with the service model, the importance of the elements in the service model, and the perceived quality of services provided by these innovative HIV service demonstration projects. Ratings of service quality were related to ratings of the respondent's knowledge of the service demonstration project, the importance of the various elements in the service continuum, and several indicators of stakeholder characteristics using the data modeling method of Exhaustive CHAID (Chi-squared Automatic Interaction Detector). The groups of stakeholders most likely to give the highest quality or success ratings for these projects are identified. The implications of these findings for developing collaborative and comprehensive service models for persons with HIV/AIDS are discussed.