Chi-square Automatic Interaction Detector Method Research Articles

Prognostic factors in diabetes: Comparison of Chi-square automatic interaction detector (CHAID) decision tree technology and logistic regression.

This study aimed to develop a diabetes prediction model. The model performance was compared with logistic regression, and the decision tree Chi-square automatic interaction detection (CHAID) was used to predict diabetes.In total, 3233 patients were included in the analysis. Of these, 589 patients with diabetes and 2644 patients without diabetes were included after analyzing the study sample from the Korean Genome and Epidemiology Study (KoGES)-8 data. In this study, Diabetes Mellitus (DM) diagnosis prediction was compared with logistic regression and prediction through machine learning (ML) using the CHAID decision classification tree. We performed statistical analysis using the CHAID method with International Business Machine (IBM) statistical program SPSS®.We performed logistic regression analysis to predict the classification of diabetes accurately, and the total classification accuracy of the analysis was 81.7%, and the CHAID decision tree classification accuracy was 81.8%. A diabetes diagnosis decision tree was created, which included seven terminal nodes and three depth levels. This analysis showed that a blood pressure problem and hospital visits were the most decisive variables at the time of classification, and two risk levels were created for diabetes diagnosis.The suggested method is a valuable tool for predicting diabetes. Patients who visit the hospital because of blood pressure problems are more likely to develop diabetes than under-treating hyperlipidemia. The diabetes prediction model can help doctors make decisions by detecting the possibility of diabetes early; however, it is impossible to diagnose diabetes using only the model without the doctor’s opinion.

Effects of robotic priming of bilateral arm training, mirror therapy, and impairment-oriented training on sensorimotor and daily functions in patients with chronic stroke: study protocol of a single-blind, randomized controlled trial

BackgroundCombining robotic therapy (RT) with task-oriented therapy is an emerging strategy to facilitate motor relearning in stroke rehabilitation. This study protocol will compare novel rehabilitation regimens that use bilateral RT as a priming technique to augment two task-oriented therapies: mirror therapy (MT) and bilateral arm training (BAT) with a control intervention: RT combined with impairment-oriented training (IOT).MethodsThis single-blind, randomized, comparative efficacy study will involve 96 participants with chronic stroke. Participants will be randomized into bilateral RT+MT, bilateral RT+BAT, and bilateral RT+IOT groups and receive 18 intervention sessions (90 min/day, 3 d/week for 6 weeks). The outcomes will include the Fugl-Meyer Assessment, Stroke Impact Scale version 3.0, Medical Research Council scale, Revised Nottingham Sensory Assessment, ABILHAND Questionnaire, and accelerometer and will be assessed at baseline, after treatment, and at the 3-month follow-up. Analysis of covariance and the chi-square automatic interaction detector method will be used to examine the comparative efficacy and predictors of outcome, respectively, after bilateral RT+MT, bilateral RT+BAT, and bilateral RT+IOT.DiscussionThe findings are expected to contribute to the research and development of robotic devices, to update the evidence-based protocols in postacute stroke care programs, and to investigate the use of accelerometers for monitoring activity level in real-life situations, which may in turn promote home-based practice by the patients and their caregivers. Directions for further studies and empirical implications for clinical practice will be further discussed in upper-extremity rehabilitation after stroke.Trial registrationThis trial was registered December 12, 2018, at www.clinicaltrials.gov (NCT03773653).

Effects of proximal priority and distal priority robotic priming techniques with impairment-oriented training of upper limb functions in patients with chronic stroke: study protocol for a single-blind, randomized controlled trial

BackgroundThe sequence of establishing a proximal stability or function before facilitation of the distal body part has long been recognized in stroke rehabilitation practice but lacks scientific evidence. This study plans to examine the effects of proximal priority robotic priming and impairment-oriented training (PRI) and distal priority robotic priming and impairment-oriented training (DRI).MethodsThis single-blind, randomized, comparative efficacy study will involve 40 participants with chronic stroke. Participants will be randomized into the PRI or DRI groups and receive 18 intervention sessions (90 min/day, 3 days/weeks for 6 weeks). The Fugl-Meyer Assessment Upper Extremity subscale, Medical Research Council Scale, Revised Nottingham Sensory Assessment, and Wolf Motor Function Test will be administered at baseline, after treatment, and at the 3-month follow-up. Two-way repeated-measures analysis of variance and the chi-square automatic interaction detector method will be used to examine the comparative efficacy and predictors of outcome, respectively, after PRI and DRI.DiscussionThrough manipulating the sequence of applying wrist and forearm robots in therapy, this study will attempt to examine empirically the priming effect of proximal or distal priority robotic therapy in upper extremity impairment-oriented training for people with stroke. The findings will provide directions for further studies and empirical implications for clinical practice in upper extremity rehabilitation after stroke.Trial RegistrationClinicalTrials.gov NCT04446273. Registered on June 23, 2020.

Centralization and Oncologic Training Reduce Postoperative Morbidity and Failure-to-rescue Rates After Cytoreductive Surgery and Hyperthermic Intraperitoneal Chemotherapy for Peritoneal Surface Malignancies: Study on a 10-year National French Practice.

Evaluate at a national level the postoperative mortality (POM), major morbidity (MM) and failure-to-rescue (FTR) after cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) across time and according to hospital-volume. CRS/HIPEC is an effective therapeutic strategy commonly used to treat peritoneal surface malignancies. However, this aggressive approach has the reputation to be associated with a high POM and MM. All patients treated with CRS/HIPEC between 2009 and 2018 in France were identified through a national medical database. Patients and perioperative outcomes were analyzed. A cut-off value of the annual CRS/HIPEC caseload affecting the 90-day POM was calculated using the Chi-squared Automatic Interaction Detector method. A multivariable logistic model was used to identify factors mediating 90-day POM. A total of 7476 CRS/HIPEC were analyzed. Median age was 59 years with a mean Elixhauser comorbidity index of 3.1, both increasing over time (P < 0.001). Ninety-day POM was 2.6%. MM occurred in 44.2% with a FTR rate of 5.1%. The threshold of CRS/HIPEC number per center per year above which the 90-day POM was significantly reduced was 45 (3.2% vs 1.9%, P = 0.01). High-volume centers had more extended surgery (P < 0.001) with increased MM (55.8% vs 40.4%, P < 0.001) but lower FTR (3.1% vs 6.3%, P = 0.001). After multivariate analysis, independent factors associated with 90-day POM were: age >70 years (P = 0.002), Elixhauser comorbidity index ≥8 (P = 0.006), lower gastro-intestinal origin, (P < 0.010), MM (P < 0.001), and <45 procedures/yr (P = 0.002). In France, CRS/HIPEC is a safe procedure with an acceptable 90-day POM that could even be improved through centralization in high-volume centers.

Establishment of a predictive model for outcomes in patients with severe acute pancreatitis by nucleated red blood cells combined with Ranson score and APACHE II score

Objective To explore a predictive model for outcomes of severe acute pancreatitis (SAP) patients. Methods A retrospective study was conducted of 102 SAP patients from January 1,2016 to April 30,2018 from Taizhou Hospital in this study. The participants were divided into survival group and death group according to the outcome of 90 days after admission,88 cases were in survival group including 57 males and 31 females, aged 55.5 (40.3-69.8) years; 14 cases were in death group including 5 males and 9 females, aged 63.0 (50.8-80.8) years. Clinical data and laboratory indicators were compared between the two groups. Statistical analyses were performed to compare categorical variables. Chi-square automatic interaction detector (CHAID) was used to construct the prediction model of SAP patients′ outcomes. The study cohort consisted of SAP patients from August 1st 2018 to July 1st 2019 were collected to validate the prediction model. Results (1)Statistical analyses were performed by chi square test and Mann-Whitney U test. There were statistically significant differences in the proportion of cardiovascular and cerebrovascular diseases, mechanical ventilation and septic shock, and also the Charlson complication index (CCI), the Ranson score and APACHEⅡscore(χ2=5.554, P=0.018; χ2=5.585,P=0.018;P=0.008;Z=-3.007,P=0.003;Z=-2.982, P=0.003; Z=-3.257, P=0.001), death group were higher than survival group. (2) The MPV, CRP, MCHC, pH, pCO2 and positive rates of NRBC were statistically different between survival group and death group(Z=-2.466,P=0.014;Z=-2.689,P=0.007;Z=-2.238,P=0.025;Z=-1.977,P=0.048;Z=-2.239, P=0.025;P=0.000).The NRBC-positive rate in the death group was higher than that in the survival group, while the other indexes were lower in the survival group.(3)The decision tree CHAID method obtains the prediction scheme: when it meets the Ranson score≤3, the SAP patients were judged to be alive; when it meets the Ranson score>3 and the NRBC in peripheral blood was negative, the SAP patients were judged to be alive; when it meets the Ranson score>3, the NRBC was positive and the APACHEⅡscore≤21, the SAP patients were judged to be alive; when it meets the Ranson score>3, NRBC was positive and APACHE Ⅱ>21, the SAP patients were adverse prognosis. (4) 50 SAP patients were collected in the validation group, with 43 actual survivors and 7 deaths. The accuracy rate of predicting the outcomes of SAP patients in validation group with the scheme was 94.0% (47/50). Conclusion The NRBC combined with Ranson score system and APACHE II score system can predict the outcomes of SAP patients. Key words: Acute disease; Pancreatitis; Erythroblasts; APACHE

Establishing Risk and Targeting Profiles Using Dara Mining: Decision Trees

The application of technology and the computerisation of management processes in customs administrations have undoubtedly accelerated the processes related to data storage. In this context, customs administrations possess vast amounts of data on trade and financial flows. Data mining tools can be effective in analysing huge reams of data. Data mining consists of understanding, preparing, modelling and analysing data using different techniques, such as machine learning. Many of these techniques have advanced predictive analytical capacities that can ultimately lead to improved analytical capabilities in risk management. The Chi-square Automatic Interaction Detector (CHAID) decision tree method was selected for the purposes of this paper to determine the customs risk factors associated with import declarations recorded in the customs clearance system. The CHAID method is also used to create risk profiles and predict non-compliant customs declarations based on established rules.

Prevalence, geographic risk factor, and development of a standardized protocol for fungal isolation in cystic fibrosis: Results from the international prospective study “MFIP”

BackgroundFungi are increasingly recognized for their potential role in contributing to pulmonary damage in Cystic Fibrosis (CF). We therefore designed a prospective international study aimed at (i) determining the prevalence of fungi isolated from sputum samples collected from a large CF population, (ii) comparing the performance of different media used for fungal culture, and (iii) proposing a standardized protocol suitable for CF routine microbiology. MethodsAn international, consensually designed prospective study was set up (https://www.ecfs.eu/special-projects/mucofong-international-project). All centers worked according to the same protocol approved by Lille Ethical Committee. CF sputa were inoculated onto eight semi-selective media incubated at 37 °C or 25 °C–30 °C for 15 days, and inspected twice weekly for fungal growth. ResultsA total of 469 sputa were collected from patients at 18 European and one Australian CF centers. Positive cultures for fungal growth were significantly associated with patient ages. Aspergillus fumigatus was the most frequently isolated mold. We identified a growing European North-to-South gradient of Scedosporium prevalence, while yeasts, Aspergillus section Fumigati, Cladosporium and Penicillium were significantly more prevalent in the Northern regions. ConclusionsAccording to the CHi-squared Automatic Interaction Detector method, we propose a consensual protocol based on two media (YPDA or Sabouraud medium, and B(+) medium) to detect the main opportunistic molds in CF context; the use of an additional medium being recommended according to the patient's clinical status. This standardized protocol allows us to have an accurate overview of the respiratory mycobiome on the culturomic side in CF.

A screening focusing on aftereffects of alcohol consumption in a student population. A National cross-sectional survey.

Students overestimate alcohol consumption of those around them and underestimate their own, so that quantitative approach may not be the most relevant to assess students' drinking. The main objective was to provide an appropriate tool for screening for students with potential drinking problems. A multicentre cross-sectional survey was conducted by internet between February and June, 2013 in France. Thirteen questions explored alcohol consumption, including 8 concerning after-effects of drinking episodes (4 items of the AUDIT) and alcohol behaviour (CAGE test). A multiple correspondence analysis (MCA) was conducted to identify profiles of student's alcohol consumption. Partitioning methods were used to group students by mode of alcohol use. The most relevant items included in the MCA were identified. Three questions were identified as most pertinent among the students with potential drinking problems and ranked by a decision tree with the Chi-square Automatic Interaction Detector method. Finally, we assessed the generalisation of the model. A total of 36,427 students participated in the survey: 25,679 were women (70.5% of respondents), sex ratio 0.42 and mean aged 21.2 (sd 3.7 years). Among those who had experimented with alcohol (N = 33,113), three consumption profiles were identified: "simple/non-use" (66.9%), "intermediate consumption" (25.9%) and "problem drinking" (7.2%). For the latter group, the three most relevant items were (Q20) "not able to stop drinking after starting", (Q21) "failed to do what was normally expected", and (Q23) "unable to remember what happened the night before". These results provide healthcare professionals with a 3-item screening tool for students "problem drinking".

Identification of Risk Factors of Multidrug-Resistant Tuberculosis by using Classification Tree Method

Multidrug-resistant tuberculosis (MDR-TB) has become a major public health problem. We tried to apply the classification tree model in building and evaluating a risk prediction model for MDR-TB. In this case-control study, 74 newly diagnosed MDR-TB patients served as the case group, and 95 patients without TB from the same medical institution served as the control group. The classification tree model was built using Chi-square Automatic Interaction Detectormethod and evaluated by income diagram, index map, risk statistic, and the area under receiver operating characteristic (ROC) curve. Four explanatory variables (history of exposure to TB patients, family with financial difficulties, history of other chronic respiratory diseases, and history of smoking) were included in the prediction model. The risk statistic of misclassification probability of the model was 0.160, and the area under ROC curve was 0.838 (P < 0.01). These suggest that the classification tree model works well for predicting MDR-TB. Classification tree model can not only predict the risk of MDR-TB effectively but also can reveal the interactions among variables.

Toward the Standardization of Mycological Examination of Sputum Samples in Cystic Fibrosis: Results from a French Multicenter Prospective Study.

Fungal respiratory colonization of cystic fibrosis (CF) patients emerges as a new concern; however, the heterogeneity of mycological protocols limits investigations. We first aimed at setting up an efficient standardized protocol for mycological analysis of CF sputa that was assessed during a prospective, multicenter study: "MucoFong" program (PHRC-06/1902). Sputa from 243 CF patients from seven centers in France were collected over a 15-month period and submitted to a standardized protocol based on 6 semi-selective media. After mucolytic pretreatment, sputa were plated in parallel on cycloheximide-enriched (ACT37), erythritol-enriched (ERY37), benomyl dichloran-rose bengal (BENO37) and chromogenic (CAN37) media incubated at 37°C and on Sabouraud-chloramphenicol (SAB27) and erythritol-enriched (ERY27) media incubated at 20-27°C. Each plate was checked twice a week during 3weeks. Fungi were conventionally identified; time for detection of fungal growth was noted for each species. Fungal prevalences and media performances were assessed; an optimal combination of media was determined using the Chi-squared automatic interaction detector method. At least one fungal species was isolated from 81% of sputa. Candida albicans was the most prevalent species (58.8%), followed by Aspergillus fumigatus (35.4%). Cultivation on CAN37, SAB27, ACT37 and ERY27 during 16days provided an optimal combination, detecting C. albicans, A. fumigatus, Scedosporium apiospermum complex and Exophiala spp. with sensitivities of 96.5, 98.8, 100 and 100%. Combination of these four culture media is recommended to ensure the growth of key fungal pathogens in CF respiratory specimens. The use of such consensual protocol is of major interest for merging results from future epidemiological studies.

Control measures to trace ≤ 15-year-old contacts of index cases of active pulmonary tuberculosis.

This was descriptive study carried out in a medium-sized Brazilian city. In ≤ 15-year-old contacts of index cases of active pulmonary tuberculosis, we assessed compliance with the Brazilian national guidelines for tuberculosis control. We interviewed 43 contacts and their legal guardians. Approximately 80% of the contacts were not assessed by the municipal public health care system, and only 21% underwent tuberculin skin testing. The results obtained with the Chi-square Automatic Interaction Detector method suggest that health care teams have a biased attitude toward assessing such contacts and underscore the need for training health professionals regarding tuberculosis control programs.

Segmenting visitors based on willingness to pay for recreational benefits

This article uses a criterion-based method to segment visitors in the context of a contingent valuation survey. The National Forest of Leiria, located in central Portugal, is used as case study. Respondents are assigned to segments through a chi-squared automatic interaction detector method according to their willingness to pay for recreational improvements. This method identifies the main socio-economic and behavioural variables that differentiate the segments. The results show that the segment with the largest willingness to pay is formed by visitors who contribute to environmental protection and rate their recreational experience in the forest as very good. Besides these ‘environmentally friendly’ visitors, two other segments show willingness to pay above average: ‘graduates’ and ‘forest neighbours’. The study shows how the rich set of data provided by contingent valuation studies can be used to segment visitors, and therefore to support the planning and management of recreational facilities.

The Combination of Frailty and ISS Scores Identifies a Simple Prognostic Index for Overall Survival in Elderly Patients Treated with Novel Agents-Based Induction Therapy

▪Background: Several biological parameters define patients with multiple myeloma (MM) at high-risk of progression or death. The well-known International Staging System (ISS), as well as age per se, are insufficient to explain differences of overall survival (OS) in patients over 65 years, who are 2/3 of newly diagnosed (ND) MM patients. We have recently showed that a frailty score combining age, functional status (Activity of Daily Living and Instrumental Activity of Daily living scores) and comorbidities (Charlson index) defines 3 categories of patients - fit, intermediate-fitness, frail - with significantly differences in OS and progression-free survival (Larocca A, et al. Blood 2013 122:687). Here we assess the causes of the different mortality in intermediate-fitness and frail groups compared to fit ones and present a final prognostic score based on the combination of ISS and frailty scores.Methods: NDMM patients over 65 years enrolled in 3 clinical trials, receiving either lenalidomide, bortezomib or carfilzomib were included in the analysis. Details on treatment regimens and results of these studies have previously been reported (Palumbo A, et al. Blood 2013 122:536; Larocca A, et al. Blood 2013 122:539, Bringhen S et al. Blood 2014 Jul 3;124(1):63-9). The cumulative incidences of discontinuation and toxicities were calculated using the Fine & Gray model. The frailty score was combined with ISS with the CHi-squared Automatic Interaction Detector method used as an iterative decision tree.Results: 869 patients (median age 74 years) were included in the analysis; 260 (30%) were frail, 269 (31%) intermediated-fitness and 340 (39%) fit. The 3-year OS was 57% in frail, 76% in intermediated-fitness and 84% in fit patients. Overall, 143 patients (16%) died, 70 (27%) frail, 39 (14%) intermediate-fitness and 34 (10%) fit. The causes of death were: disease progression [35 (13%) in frail, 22 (8%) in intermediate-fitness and 18 (5%) in fit patients] and toxicity [21 (8%), 10 (4%) and 11 (3%), respectively]. The higher risk of death for progression was related with the lower dose-intensity due to the higher rate of drug discontinuation and/or dose reduction. The average dose intensity was lower in frail (74%, p=0.0006) and intermediate-fitness patients (80%, p=0.07) compared with fit patients (85%). The cumulative incidence of drug discontinuation for any cause, excluding progression and death, was higher in frail (25%; HR 2.21, p<0.001) and intermediate-fitness (22%; HR: 1.41, p=0.052) patients compared with fit ones (17%). The most frequent reasons for toxicity-related death were cardiac events [11 (4%) in frail patients, 2 (1%) in intermediate-fitness, 3 (1%) in fit] and infections [8 (3%), 2 (1%) and 2 (1%), respectively]. When we combined the frailty score with the ISS, 6 groups of patients and 4 risk categories were identified: fit patients with ISS I at low risk (15%; 3-year OS: 94%), fit patients with ISS stage II or III and intermediate-fitness patients with ISS I, II or III at intermediate risk (55%; 3-year OS: 75-77%.), frail patients with ISS stage I or II at high risk (19%; 3-year OS: 61%) and frail patients with ISS stage III at very-high risk (11%, 3-year OS: 55%) (Figure 1).Conclusion: The inferior survival observed among intermediate-fitness and in frail patients as compared to fit ones, is related to a higher rate of toxic deaths and disease progression, due to a lower dose intensity. The combination of the frailty score, evaluating the patient’s status, and the standard ISS, taking into account the biological characteristics of the disease, can predict survival and enhances the single predictive values of the scores, thus representing a valuable tool for treatment-decision in the clinical practice. [Display omitted] DisclosuresLarocca:Janssen Cilag: Honoraria; Celgene: Honoraria. Off Label Use: Use off-label of lenalidomide (immunomodulatory drug), carfilzomib (proteasome inhibitor), subcutaneous bortezomib (proteasome inhibitor) in terms of schedule used and combination.. Bringhen:Onyx: Consultancy; Merck Sharp & Dohme: Membership on an entity’s Board of Directors or advisory committees; Novartis: Honoraria; Janssen and Cilag: Honoraria; Celgene: Honoraria. Hajek:Janssen: Honoraria; Celgene: Consultancy, Honoraria; Merck: Consultancy, Honoraria. Offidani:Celgene: Honoraria; Janssen: Honoraria. Maracci:Mundipharma: Honoraria. Gay:Sanofi: Membership on an entity’s Board of Directors or advisory committees; Janssen: Honoraria; Celgene: Honoraria, Membership on an entity’s Board of Directors or advisory committees. Marasca:Janssen: Honoraria; Celgene: Honoraria. Giuliani:Celgene: Research Funding. Musto:Janssen: Honoraria; Celgene: Honoraria. Boccadoro:Sanofi: Consultancy, Membership on an entity’s Board of Directors or advisory committees; Onyx: Consultancy, Membership on an entity’s Board of Directors or advisory committees; Janssen-Cilag: Consultancy, Membership on an entity’s Board of Directors or advisory committees; Celgene: Consultancy, Membership on an entity’s Board of Directors or advisory committees. Sonneveld:Millenium: Honoraria, Research Funding; Onyx: Honoraria, Research Funding; Janssen: Honoraria, Research Funding; Celgene: Honoraria, Research Funding. Palumbo:Celgene: Consultancy, Honoraria; Janssen-Cilag: Consultancy, Honoraria; Millennium Pharmaceuticals, Inc.: Consultancy, Honoraria; Onyx Pharmaceuticals: Consultancy, Honoraria; Array BioPharma: Honoraria; Amgen: Consultancy, Honoraria; Sanofi: Honoraria; Genmab A/S: Consultancy, Honoraria; Bristol-Myers Squibb: Consultancy, Honoraria.

Anesthetic practice for Caesarean section and factors influencing anesthesiologists' choice of anesthesia: a population-based study.

We examined the change in anesthetic practice for Caesarean section (CS) during the past decade and determined factors influencing anethesiologists' decisions. The cases were identified from data retrieved from Longitudinal Health Insurance Database released by the Taiwan National Health Research Institute in 2008. Trend analysis was performed using logistic regression models. The decision tree analysis was performed using the chi-squared automatic interaction detector method and multivariable logistic regression analysis was performed to identify predictors of general anesthesia. A total of 25,606 women undergoing CS were studied. Logistic regression analyses revealed an upward trend of spinal anesthesia from 2000 to 2008 [57.8-67.5%, adjusted odds ratio (OR) = 1.06, 95% confidence interval (CI) = 1.05-1.07, P < 0.001] and a decreasing trend across time for both general and epidural anesthesia (5.5-3.9% and 36.7-28.6%; both OR < 1, both P < 0.001). Patterns of change in anesthetic practice across time for emergency and non-emergency CS were similar (all P < 0.05). Our data further demonstrated that early or threatened labor, a history of preeclampsia, antepartum hemorrhage, emergency CS, and previous CS were important predictors that influenced the anesthesiologists' choice of general anesthesia versus neuraxial anesthesia for women undergoing CS. Spinal anesthesia was the most common mode of anesthesia for CS deliveries in Taiwan during the past decade. Early or threatened labor, antepartum hemorrhage, emergency CS, previous CS, and preeclampsia are significant determinants of general anesthesia in CS deliveries.

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.

Adulteration detection of argan oil by inductively coupled plasma optical emission spectrometry

Abstract The singularity of the trace element profile of argan oil has been demonstrated by means of inductively coupled plasma optical emission measurement in combination with different chemometric approaches. The ability of multivariate analysis methods; such as hierarchical cluster analysis (HCA), principal component analysis (PCA), classification trees using Chi -squared Automatic Interaction Detector (CHAID) and discriminant analysis (DA) to achieve edible oils classification based on its type or variety from their elemental content have been investigated. The calculations were performed using 16 variables (contents of Na, Mg, Al, K, Ca, Ti, Fe, Co, Ni, Cu, Zn, Cd, Pr, Sm, Er and Bi at μg g −1 level determined by ICP-OES). HCA is able to differentiate sunflower oil samples from the rest, however the discrimination of argan oil from olive, seeds and soya oils based on their different trace element composition is hard to achieve. The PCA analysis shows three different classes in the multidimensional space (PC1-3) representing sunflower, argan and a third group comprising olive, seeds and soya oils. CHAID method allows separating the entire vegetable oil dataset, providing a correct re-substitution rate of 94.12% for argan oil using only the concentration of K. DA performed using the same variables, provides also an acceptable average accuracy results of 93.65%, by the re-substitution method. DA has been successfully applied to the analysis adulterated argan oil by addition of cheaper vegetable oils.