Enhancing Pulmonary Embolism Mortality Risk Stratification Using Machine Learning: The Role of the Neutrophil-to-Lymphocyte Ratio

(1) Background: Acute pulmonary embolism (PE) is a significant public health concern that requires efficient risk estimation to optimize patient care and resource allocation. The purpose of this retrospective study was to show the correlation of NLR (neutrophil-to-lymphocyte ratio) and PESI (pulmonary embolism severity index)/sPESI (simplified PESI) in determining the risk of in-hospital mortality in patients with pulmonary thromboembolism. (2) Methods: A total of 160 patients admitted at the County Clinical Emergency Hospital of Sibiu from 2019 to 2022 were included and their hospital records were analyzed. (3) Results: Elevated NLR values were significantly correlated with increased in-hospital mortality. Furthermore, elevated NLR was associated with PESI and sPESI scores and their categories, as well as the individual components of these parameters, namely increasing age, hypotension, hypoxemia, and altered mental status. We leveraged the advantages of machine learning algorithms to integrate elevated NLR into PE risk stratification. Utilizing two-step cluster analysis and CART (classification and regression trees), several distinct patient subgroups emerged with varying in-hospital mortality rates based on combinations of previously validated score categories or their defining elements and elevated NLR, WBC (white blood cell) count, or the presence COVID-19 infection. (4) Conclusion: The findings suggest that integrating these parameters in risk stratification can aid in improving predictive accuracy of estimating the in-hospital mortality of PE patients.

Cross-validated tree-based models for multi-target learning

Multi-target learning (MTL) is a popular machine learning technique which considers simultaneous prediction of multiple targets. MTL schemes utilize a variety of methods, from traditional linear models to more contemporary deep neural networks. In this work we introduce a novel, highly interpretable, tree-based MTL scheme which exploits the correlation between the targets to obtain improved prediction accuracy. Our suggested scheme applies cross-validated splitting criterion to identify correlated targets at every node of the tree. This allows us to benefit from the correlation among the targets while avoiding overfitting. We demonstrate the performance of our proposed scheme in a variety of synthetic and real-world experiments, showing a significant improvement over alternative methods. An implementation of the proposed method is publicly available at the first author's webpage.

Which Intervention Synergies Maximize AGYW's HIV Outcomes? A Classification and Regression Tree Analysis of Layered HIV Prevention Programming

INTRODUCTION

Intersecting behavioral, social, and structural factors increase adolescent girls' (AG) and young women's (YW) HIV vulnerability. Yet, understanding of optimal intervention synergies remains limited. We identified intervention combinations that statistically maximized reductions in AGYW's HIV-related risk.

METHODS

Using data collected in 2018 with Zambian AG (n = 487, aged 15-19 years) and YW (n = 505, aged 20-25 years) after 12-14 months exposure to Determined, Resilient, Empowered, AIDS-free, Mentored, and Safe (multisectoral HIV program), we used classification and regression trees to explore relationships between interventions (safe space/social asset building [SAB] and provision of/linkage to youth-friendly health services [YFHS], education social protection [Educ], economic social protection [Econ]) and HIV-related outcomes (HIV testing, consistent condom use, transactional sex, and sexual violence experience from partners and nonpartners).

RESULTS

Overall, 59.9% completed SAB and 81.5%, 35.4%, and 29.6% received YHFS, Educ, and Econ, respectively. For AG, HIV testing improved (from 73% to 83%) with exposure to all interventions, condom use improved with Econ (from 33% to 46%), transactional sex reduced with SAB + Educ, and sexual violence from partners and nonpartners reduced with Educ and SAB, respectively. For YW, HIV testing increased with Educ (from 77% to 91%), condom use increased with SAB + YFHS (from 36% to 52%), transactional sex reduced with combinations of all interventions, and sexual violence from partners reduced with YFHS and from nonpartners with SAB + Econ.

CONCLUSIONS

Tailored interventions might be more effective than uniform combination intervention packages in reducing AGYW's HIV risk. AG benefitted most from SAB and/or Educ while YFHS, Educ, and/or SAB reduced YW's HIV-related risk. Educational and asset-building interventions could have the greatest impact on AGYW's HIV risk.

A Bayesian Approach Towards Missing Covariate Data in Multilevel Latent Regression Models

The measurement of latent traits and investigation of relations between these and a potentially large set of explaining variables is typical in psychology, economics, and the social sciences. Corresponding analysis often relies on surveyed data from large-scale studies involving hierarchical structures and missing values in the set of considered covariates. This paper proposes a Bayesian estimation approach based on the device of data augmentation that addresses the handling of missing values in multilevel latent regression models. Population heterogeneity is modeled via multiple groups enriched with random intercepts. Bayesian estimation is implemented in terms of a Markov chain Monte Carlo sampling approach. To handle missing values, the sampling scheme is augmented to incorporate sampling from the full conditional distributions of missing values. We suggest to model the full conditional distributions of missing values in terms of non-parametric classification and regression trees. This offers the possibility to consider information from latent quantities functioning as sufficient statistics. A simulation study reveals that this Bayesian approach provides valid inference and outperforms complete cases analysis and multiple imputation in terms of statistical efficiency and computation time involved. An empirical illustration using data on mathematical competencies demonstrates the usefulness of the suggested approach.

Two Clinical Prediction Tools to Inform Rapid Tuberculosis Treatment Decision-making in Children

Background

In the absence of bacteriologic confirmation to diagnose tuberculosis (TB) in children, it is suggested that treatment should be initiated when sufficient clinical evidence of disease is available. However, it is unclear what clinical evidence is sufficient to make this decision. To identify children who would benefit from rapid initiation of TB treatment, we developed 2 clinical prediction tools.

Methods

We conducted a secondary analysis of a prospective intensified TB patient-finding intervention conducted in Pakistan in 2014-2016. TB disease was determined through either bacteriologic confirmation or a clinical diagnosis. We derived 2 tools: 1 uses classification and regression tree (CART) analysis to develop decision trees, while the second uses multivariable logistic regression to calculate a risk score.

Results

Of the 5162 and 5074 children included in the CART and prediction score, respectively, 1417 (27.5%) and 1365 (26.9%) were eligible for TB treatment. CART identified abnormal chest radiographs and family history of TB as the most important predictors (area under the receiver operating characteristic curve [AUC], 0.949). The final prediction score model included age group (0-4, 5-9, 10-14), weight <5th percentile, cough, fever, weight loss, chest radiograph suggestive of TB disease, and family history of TB; the identified best cutoff score was 9 (AUC, 0.985%).

Conclusions

Use of clinical evidence was sufficient to accurately identify children who would benefit from treatment initiation. Our tools performed well compared with existing algorithms, though these results need to be externally validated before operationalization.

Derivation of a clinical decision rule for a bone marrow aspirate concentrate injection in knee osteoarthritis

Aim: To develop a simple clinical decision rule (CDR) to identify people with knee osteoarthritis who are likely or unlikely to benefit from bone marrow aspirate concentrate (BMAC) injection. Materials & methods: A total of 92 people with clinical and radiographic evidence of refractory knee osteoarthritis received a single intra-articular (IA) BMAC injection. Multiple logistic regression analysis was used to determine which combination of risk factors predicted BMAC responsiveness. A responder was defined as a person whose knee pain improved more than 15% from baseline 6 months post procedure. Results: The CDR demonstrated that those with lower pain levels, or high pain levels with previous surgery, could be predicted to benefit from a single IA BMAC injection. Conclusion: A simple CDR containing three variables predicted responsiveness to a single IA knee BMAC injection with high accuracy. Further validation of the CDR is required prior to routine use in clinical practice.

Preoperative Serum Markers and Risk Classification in Intrahepatic Cholangiocarcinoma: A Multicenter Retrospective Study

Accurate risk stratification selects patients who are expected to benefit most from surgery. This retrospective study enrolled 225 Japanese patients with intrahepatic cholangiocellular carcinoma (ICC) who underwent hepatectomy between January 2009 and December 2020 and identified preoperative blood test biomarkers to formulate a classification system that predicted prognosis. The optimal cut-off values of blood test parameters were determined by ROC curve analysis, with Cox univariate and multivariate analyses identifying prognostic factors. Risk classifications were established using classification and regression tree (CART) analysis. CART analysis revealed decision trees for recurrence-free survival (RFS) and overall survival (OS) and created three risk classifications based on machine learning of preoperative serum markers. Five-year rates differed significantly (p < 0.001) between groups: 60.4% (low-risk), 22.8% (moderate-risk), and 4.1% (high-risk) for RFS and 69.2% (low-risk), 32.3% (moderate-risk), and 9.2% (high-risk) for OS. No difference in OS was observed between patients in the low-risk group with or without postoperative adjuvant chemotherapy, although OS improved in the moderate group and was prolonged significantly in the high-risk group receiving chemotherapy. Stratification of patients with ICC who underwent hepatectomy into three risk groups for RFS and OS identified preoperative prognostic factors that predicted prognosis and were easy to understand and apply clinically.

Motivation of Teleworkers and Non-teleworkers in Times of COVID-19 in Spain: An Exploratory Study Using Non-parametric Analysis and Classification and Regression Trees

With the outbreak of COVID-19 in spring 2020, small, medium, and large companies were forced to cope with the unexpected circumstances. Faced by this health emergency, it was necessary to ensure that staff remained motivated and that they could continue to carry out their duties despite the obstacles. The main goal of this exploratory research was to characterize employees who teleworked and who did not, and their motivation during the lockdown. A total of 11,779 workers from different-sized companies in various sectors answered an ad hoc questionnaire. By using non-parametric comparisons and Classification and Regression Trees (CRTs), the results show differences in both the assessment of strategies put into practice by the companies and the level of motivation of teleworkers and non-teleworkers, with the latter being more highly motivated. Nonetheless, teleworkers assessed their companies’ strategies and the role of their managers and colleagues more positively. This research helps to understand how different sectors have dealt with the crisis, according to the degree of teleworking implemented in each sector, and to what extent the motivation of the employees has been affected. The analysis of the large amount of data obtained confirms the importance of the role of managers in sustaining the motivation of their subordinates in times of crisis. In this sense, it is necessary to develop managers’ competencies in order to develop and maintain relations of trust and support with their coworkers. On the other hand, it is necessary to foster employees’ sense of meaningfulness and responsibility at work in order to keep them motivated.

Characterization of Drug Resistance in Chronic Myeloid Leukemia Cells Based on Laser Tweezers Raman Spectroscopy

Multidrug resistance is highly associated with poor prognosis of chronic myeloid leukemia. This work aims to explore whether the laser tweezers Raman spectroscopy (LTRS) could be practical in separating adriamycin-resistant chronic myeloid leukemia cells K562/adriamycin from its parental cells K562, and to explore the potential mechanisms. Detection of LTRS initially reflected the spectral differences caused by chemoresistance including bands assigned to carbohydrates, amino acid, protein, lipids, and nucleic acid. In addition, principal components analysis as well as the classification and regression trees algorithms showed that the specificity and sensitivity were above 90%. Moreover, the band data-based classification and regression tree model and receiver operating characteristic curve further determined some important bands and band intensity ratios to be reliable indexes in discriminating K562 chemoresistance status. Finally, we highlighted three metabolism pathways correlated with chemoresistance. This work demonstrates that the label-free LTRS analysis combined with multivariate statistical analyses have great potential to be a novel analytical strategy at the single-cell level for rapid evaluation of the chemoresistance status of K562 cells.

Impact of the COVID-19 Pandemic on Higher Education: Characterizing the Psychosocial Context of the Positive and Negative Affective States Using Classification and Regression Trees

Our aim is to analyze the extent to which the psychosocial aspects can characterize the affective states of the teachers, administrative staff, and undergraduate and postgraduate students during the quarantine. A questionnaire was answered by 1,328 people from the community of the Universitat de Barcelona (UB), Spain. The survey was partially designed ad hoc, collecting indicators related to sociodemographic variables, the impact of COVID on the subjects or in their personal context, the psychosocial context of coexistence and perceived social support, characteristics related to the physical context during the quarantine, and labor conditions. Additionally, it included two validated instruments: the Survey Work-Home Interaction-Nijmegen for Spanish Speaking Countries (SWING-SSC) validated in Spanish and PANAS, the Positive and Negative Affect Schedule. Classification and Regression Trees (CART) were performed to identify which variables better characterize the participants' level of positive and negative affective states. Results according to groups showed that students are the ones who have suffered the most as a result of this situation (temporary employment regulation, higher scores in negative work-home and home-work interaction, lower scores in positive home-work interaction, and negative effects of teleworking). Additionally, they reported a higher mean score in interpersonal conflict and worse scores with regard to negative affective states. Based on sex, women were the ones whose environment was shown to be more frequently affected by the pandemic and who exhibited more negative effects of teleworking. In general terms, participants with the highest scores in negative affective states were those who perceived an increase in conflict and a high negative effect from work spilling over into their personal lives. On the contrary, participants with the highest levels of positive affective states were those with medium to low levels of negative home-work interaction, over 42.5 years old, and with medium to high levels of positive work-home interaction. Our results aim to help higher education to reflect on the need to adapt to this new reality, since the institutions that keep pace with evolving trends will be able to better attract, retain, and engage all the members of the university community in the years ahead.

What cut-point in gait speed best discriminates community dwelling older adults with mobility complaints from those without? A pooled analysis from the Sarcopenia Definitions and Outcomes Consortium

BACKGROUND

Cut-points to define slow walking speed have largely been derived from expert opinion.

METHODS

Study participants (13,589 men and 5,043 women aged ≥65years) had walking speed (m/s) measured over 4-6 meters (mean ± SD: 1.20 ± 0.27 m/s in men and 0.94 ± 0.24 m/s in women.) Mobility limitation was defined as self-reported any difficulty with walking ~1/4 mile (prevalence: 12.6% men, 26.4% women). Sex-stratified classification and regression tree (CART) models with 10-fold cross-validation identified walking speed cut-points that optimally discriminated those who reported mobility limitation from those who did not.

RESULTS

Among 5,043 women, CART analysis identified two cut-points, classifying 4,144 (82.2%) with walking speed ≥0.75 m/s, which we labeled as "fast"; 478 (9.5%) as "intermediate" (walking speed ≥0.62 m/s but <0.75 m/s); and 421 (8.3%) as "slow" (walking speed <0.62 m/s). Among 13,589 men, CART analysis identified three cut-points, classifying 10,001 (73.6%) with walking speed ≥1.00 m/s ("very fast"); 2,901 (21.3%) as "fast" (walking speed ≥0.74 m/s but <1.00 m/s); 497 (3.7%) as "intermediate" (walking speed ≥0.57 m/s but <0.74 m/s); and 190 (1.4%) as "slow" (walking speed <0.57 m/s). Prevalence of self-reported mobility limitation was lowest in the "fast" or "very fast" (11% for men and 19% for women) and highest in the "slow" (60.5% in men and 71.0% in women). Rounding the two slower cut-points to 0.60 m/s and 0.75 m/s reclassified very few participants.

CONCLUSIONS

Cut-points in walking speed of ~0.60 m/s and 0.75 m/s discriminate those with self-reported mobility limitation from those without.

Trends in Catastrophic Occupational Incidents among Electrical Contractors, 2007-2013

This study used methodologies of descriptive and quantitative statistics to identify the contributing factors most affecting occupational accident outcomes among electrical contracting enterprises, given an accident occurred. Accident reports were collected from the Occupational Safety and Health Administration's fatality and catastrophe database. To ensure the reliability of the data, the team manually codified more than 600 incidents through a comprehensive content analysis using injury-classification standards. Inclusive of both fatal and non-fatal injuries, the results showed that most accidents happened in nonresidential buildings, new construction, and small projects (i.e., $50,000 or less). The main source of injuries manifested in parts and materials (46%), followed by tools, instruments, and equipment (19%), and structure and surfaces (16%). The most frequent types of injuries were fractures (31%), electrocutions (27%), and electrical burns (14%); the main injured body parts were upper extremities (25%), head (23%), and body system (18%). Among non-fatal cases, falls (37%), exposure to electricity (36%), and contact with objects (19%) caused most injuries; among fatal cases, exposure to electricity was the leading cause of death (50%), followed by falls (28%) and contact with objects (19%). The analysis also investigated the impact of several accident factors on the degree of injuries and found significant effects from such factors such as project type, source of injury, cause of injury, injured part of body, nature of injury, and eventtype. In other words, the statistical probability of a fatal accident-given an accident occurrence-changes significantly based on the degree of these factors. The results of this study, as depicted in the proposed decision tree model, revealed that the most important factor for predicting the nature of injury (electrical or non-electrical) is: whether the source of injury is parts and materials; followed by whether the source of injury is tools, instruments, and equipment. In other words, in predicting (with a 94.31% accuracy) the nature of injury as electrical or non-electrical, whether the source of injury is parts and materials and whether the source of injury is tools, instruments, and equipment are very important. Seven decision rules were derived from the proposed decision tree model. Beyond these outcomes, the described methodology contributes to the accident-analysis body of knowledge by providing a framework for codifying data from accident reports to facilitate future analysis and modeling attempts to subsequently mitigate more injuries in other fields.

Water stress assessment on grapevines by using classification and regression trees

Multiple factors, such as the vineyard environment and winemaking practices, are known to affect the development of vines as well as the final composition of grapes. Water stress promotes the synthesis of phenols and is associated with grape quality as long as it does not inhibit production. To identify the key parameters for managing water stress and grape quality, multivariate statistical analysis is essential. Classification and regression trees are methods for constructing prediction models from data, especially when data are complex and when constructing a single global model is difficult and models are challenging to interpret. The models were obtained by recursively partitioning the data space and fitting a simple prediction model within each partition. The partitioning can be represented graphically as a decision tree. This approach permitted the most decisive variables for predicting the most vulnerable vineyards and wine quality parameters associated with water stress. In Priorat AOC, Carignan grapevines had the highest water potential and abscisic acid concentration in the early growth plant stages and permitted vineyards to be classified by mesoclimate. This information is useful for identifying which measurements could most easily differentiate between early and late-ripening vineyards. LWP and Ts during an early physiological stage (pea size) permitted warm and cold areas to be differentiated.

A volatolomic approach using cerumen as biofluid to diagnose bovine intoxication by Stryphnodendron rotundifolium

An innovative volatolomic approach employs the detection of biomarkers present in cerumen (earwax) to identify cattle intoxication by Stryphnodendron rotundifolium Mart., Fabaceae (popularly known as barbatimão). S. rotundifolium is a poisonous plant with the toxic compound undefined and widely distributed throughout the Brazilian territory. Cerumen samples from cattle of two local Brazilian breeds ('Curraleiro Pé-Duro' and 'Pantaneiro') were collected during an experimental intoxication protocol and analyzed using headspace (HS)/GC-MS followed by multivariate analysis (genetic algorithm for a partial least squares, cluster analysis, and classification and regression trees). A total of 106 volatile organic metabolites were identified in the cerumen samples of bovines. The intoxication by S. rotundifolium influenced the cerumen volatolomic profile of the bovines throughout the intoxication protocol. In this way, it was possible to detect biomarkers for cattle intoxication. Among the biomarkers, 2-octyldecanol and 9-tetradecen-1-ol were able to discriminate all samples between intoxicated and nonintoxicated bovines. The cattle intoxication diagnosis by S. rotundifolium was accomplished by applying the cerumen analysis using HS/GC-MS, in an easy, accurate, and noninvasive way. Thus, the proposed bioanalytical chromatography protocol is a useful tool in veterinary applications to determine this kind of intoxication.

Validation of an Optimized Algorithm for Identifying Persons Living With Diagnosed HIV From New York State Medicaid Data, 2006-2014

Algorithms are regularly used to identify persons living with diagnosed human immunodeficiency virus (HIV) (PLWDH) in Medicaid data. To our knowledge, there are no published reports of an HIV algorithm from Medicaid claims codes that have been compared with an HIV surveillance system to assess its sensitivity, specificity, positive predictive value, and negative predictive value in identifying PLWDH. Therefore, our aims in this study were to 1) develop an algorithm that could identify PLWDH in New York State Medicaid data from 2006-2014 and 2) validate this algorithm using the New York State HIV surveillance system. Classification and regression tree analysis identified 16 nodes that we combined to create a case-finding algorithm with 5 criteria. This algorithm identified 86,930 presumed PLWDH, 88.0% of which were verified by matching to the surveillance system. The algorithm yielded a sensitivity of 94.5%, a specificity of 94.4%, a positive predictive value of 88.0%, and a negative predictive value of 97.6%. This validated algorithm has the potential to improve the utility of Medicaid data for assessing health outcomes and programmatic interventions.

Tree-based classification system incorporating the HVTT-PVTT score for personalized management of hepatocellular carcinoma patients with macroscopic vascular invasion

Purpose: To develop a decision tree algorithm-based classification system for personalized management of hepatocellular carcinoma (HCC) patients with macroscopic vascular invasion.

Results: The HVTT-PVTT score could differentiate two groups of patients (< 3 and ≥ 3 points) with different survival outcomes (7.4 vs 4.6 months, P < 0.001) and surgical proportion (24.4% vs 3.6%, P < 0.001). Using the Cox regression model and classification and regression tree (CART) algorithm, patients in the training set were automatically separated into three subgroups with different prognosis (10.3 vs 6.1 vs 3.3 months). The predictive accuracy was verified in the validation group (12.3 vs 6.9 vs 5.6 months) and was better than other commonly used staging systems.

Conclusions: Our study proposed a new classification system for HCC patients with macroscopic vascular invasion that could be meaningful for personalized management of these patients.

Methods: A total of 869 HCC patients initially diagnosed with macroscopic vascular invasion were randomly divided into training and validation sets. A comprehensive and simplified HVTT-PVTT score was set up for subdivision of vascular invasion according to the patients’ survival outcome. Then, a decision tree algorithm-based classification system was used to establish the refined subdivision system incorporating all independent prognostic factors.

Quantitative assessment of the activity of antituberculosis drugs and regimens

Introduction: Identification of optimal drug doses and drug combinations is crucial for optimized treatment of tuberculosis. Areas covered: An unprecedented level of research activity involving multiple approaches is seeking to improve tuberculosis treatment. This report is a review of the quantitative methods currently used on clinical data sets to identify drug exposure targets and optimal drug combinations for tuberculosis treatment. A high-level summary of the methods, including the strengths and weaknesses of each method and potential methodological improvements is presented. Methods incorporating data generated from multiple sources such as in vitro and clinical studies, and their potential to provide better estimates of pharmacokinetic/pharmacodynamic (PK/PD) targets, are discussed. PK/PD relationships identified are compared between different studies and data analysis methods. Expert opinion: The relationships between drug exposures and tuberculosis treatment outcomes are complex and require analytical methods capable of handling the multidimensional nature of the relationships. The choice of a method is guided by its complexity, interpretability of results, and type of data available.

Leveraging Machine Learning Approaches for Predicting Antidepressant Treatment Response Using Electroencephalography (EEG) and Clinical Data

Background: Individuals with major depressive disorder (MDD) vary in their response to antidepressants. However, identifying objective biomarkers, prior to or early in the course of treatment that can predict antidepressant efficacy, remains a challenge. Methods: Individuals with MDD participated in a 12-week antidepressant pharmacotherapy trial. Electroencephalographic (EEG) data was collected before and 1 week post-treatment initiation in 51 patients. Response status at week 12 was established with the Montgomery-Asberg Depression Scale (MADRS), with a ≥50% decrease characterizing responders (N = 27/24 responders/non-responders). We used a machine learning (ML)-approach for predicting response status. We focused on Random Forests, though other ML methods were compared. First, we used a tree-based estimator to select a relatively small number of significant features from: (a) demographic/clinical data (age, sex, individual item/total MADRS scores at baseline, week 1, change scores); (b) scalp-level EEG power; (c) source-localized current density (via exact low-resolution electromagnetic tomography [eLORETA] software). Second, we applied kernel principal component analysis to reduce and map important features. Third, a set of ML models were constructed to classify response outcome based on mapped features. For each dataset, predictive features were extracted, followed by a model of all predictive features, and finally by a model of the most predictive features. Results: Fifty eLORETA features were predictive of response (across bands, both time-points); alpha₁/theta eLORETA features showed the highest predictive value. Eighty-eight scalp EEG features were predictive of response (across bands, both time-points), with theta/alpha₂ being most predictive. Clinical/demographic data consisted of 31 features, with the most important being week 1 "concentration difficulty" scores. When all features were included into one model, its predictive utility was high (88% accuracy). When the most important features were extracted in the final model, 12 predictive features emerged (78% accuracy), including baseline scalp-EEG frontopolar theta, parietal alpha₂ and frontopolar alpha₁. Conclusions: These findings suggest that ML models of pre- and early treatment-emergent EEG profiles and clinical features can serve as tools for predicting antidepressant response. While this must be replicated using large independent samples, it lays the groundwork for research on personalized, "biomarker"-based treatment approaches.

Computerized adaptive testing with decision regression trees: an alternative to item response theory for quality of life measurement in multiple sclerosis

BACKGROUND

The aim of this study was to propose an alternative approach to item response theory (IRT) in the development of computerized adaptive testing (CAT) in quality of life (QoL) for patients with multiple sclerosis (MS). This approach relied on decision regression trees (DRTs). A comparison with IRT was undertaken based on precision and validity properties.

MATERIALS AND METHODS

DRT- and IRT-based CATs were applied on items from a unidi-mensional item bank measuring QoL related to mental health in MS. The DRT-based approach consisted of CAT simulations based on a minsplit parameter that defines the minimal size of nodes in a tree. The IRT-based approach consisted of CAT simulations based on a specified level of measurement precision. The best CAT simulation showed the lowest number of items and the best levels of precision. Validity of the CAT was examined using sociodemographic, clinical and QoL data.

RESULTS

CAT simulations were performed using the responses of 1,992 MS patients. The DRT-based CAT algorithm with minsplit = 10 was the most satisfactory model, superior to the best IRT-based CAT algorithm. This CAT administered an average of nine items and showed satisfactory precision indicators (R = 0.98, root mean square error [RMSE] = 0.18). The DRT-based CAT showed convergent validity as its score correlated significantly with other QoL scores and showed satisfactory discriminant validity.

CONCLUSION

We presented a new adaptive testing algorithm based on DRT, which has equivalent level of performance to IRT-based approach. The use of DRT is a natural and intuitive way to develop CAT, and this approach may be an alternative to IRT.

Which melanoma patient carries a BRAF-mutation? A comparison of predictive models

Background

In patients with advanced melanoma the detection of BRAF mutations is considered mandatory before the initiation of an expensive treatment with BRAF/MEK inhibitors. Sometimes it is difficult to perform such an analysis if archival tumor tissue is not available and fresh tissue has to be collected.

Results

514 of 1170 patients (44%) carried a BRAF mutation. All models revealed age and histological subtype of melanoma as the two major predictive variables. Accuracy ranged from 0.65–0.71, being best in the random forest model. Sensitivity ranged 0.76–0.84, again best in the random forest model. Specificity was low in all models ranging 0.51–0.55.

Methods

We collected the clinical data and mutational status of 1170 patients with advanced melanoma and established three different predictive models (binary logistic regression, classification and regression trees, and random forest) to forecast the BRAF status.

Conclusions

Up to date statistical models are not able to predict BRAF mutations in an acceptable accuracy. The analysis of the mutational status by sequencing or immunohistochemistry must still be considered as standard of care.

Subregional Nowcasts of Seasonal Influenza Using Search Trends

Background

Limiting the adverse effects of seasonal influenza outbreaks at state or city level requires close monitoring of localized outbreaks and reliable forecasts of their progression. Whereas forecasting models for influenza or influenza-like illness (ILI) are becoming increasingly available, their applicability to localized outbreaks is limited by the nonavailability of real-time observations of the current outbreak state at local scales. Surveillance data collected by various health departments are widely accepted as the reference standard for estimating the state of outbreaks, and in the absence of surveillance data, nowcast proxies built using Web-based activities such as search engine queries, tweets, and access of health-related webpages can be useful. Nowcast estimates of state and municipal ILI were previously published by Google Flu Trends (GFT); however, validations of these estimates were seldom reported.

Objective

The aim of this study was to develop and validate models to nowcast ILI at subregional geographic scales.

Methods

We built nowcast models based on autoregressive (autoregressive integrated moving average; ARIMA) and supervised regression methods (Random forests) at the US state level using regional weighted ILI and Web-based search activity derived from Google's Extended Trends application programming interface. We validated the performance of these methods using actual surveillance data for the 50 states across six seasons. We also built state-level nowcast models using state-level estimates of ILI and compared the accuracy of these estimates with the estimates of the regional models extrapolated to the state level and with the nowcast estimates published by GFT.

Results

Models built using regional ILI extrapolated to state level had a median correlation of 0.84 (interquartile range: 0.74-0.91) and a median root mean square error (RMSE) of 1.01 (IQR: 0.74-1.50), with noticeable variability across seasons and by state population size. Model forms that hypothesize the availability of timely state-level surveillance data show significantly lower errors of 0.83 (0.55-0.23). Compared with GFT, the latter model forms have lower errors but also lower correlation.

Conclusions

These results suggest that the proposed methods may be an alternative to the discontinued GFT and that further improvements in the quality of subregional nowcasts may require increased access to more finely resolved surveillance data.

Analyzing injury severity of bus passengers with different movements

OBJECTIVE

Though public transport vehicles are rarely involved in mass casualty accidents, when they are, the number of injuries and fatalities is usually high due to the high passenger capacity. Of the few studies that have been conducted on bus safety, the majority focused on vehicle safety features, road environmental factors, as well as driver characteristics. Nevertheless, few studies have attempted to investigate the underlying risk factors related to bus occupants. This article presents an investigation aimed at identifying the risk factors affecting injury severity of bus passengers with different movements.

METHOD

Three different passenger movement types including standing, seated, and boarding/alighting were analyzed individually using classification and regression tree (CART) method based on publicly available accident database of Great Britain.

RESULTS

According to the results of exploratory analyses, passenger age and vehicle maneuver are associated with passenger injury severity in all 3 types of accidents. Moreover, the variable "skidding and overturning" is associated with injury severity of seated passengers and driver age is correlated with injury severity of standing and boarding/alighting passengers.

CONCLUSIONS

The CART method shows its ability to identify and easily explain the complicated patterns affecting passenger injury severity. Several countermeasures to reduce bus passenger injury severity are recommended.

Protecting Confidentiality in Cancer Registry Data With Geographic Identifiers

The National Cancer Institute's Surveillance, Epidemiology, and End Results Program releases research files of cancer registry data. These files include geographic information at the county level, but no finer. Access to finer geography, such as census tract identifiers, would enable richer analyses-for example, examination of health disparities across neighborhoods. To date, tract identifiers have been left off the research files because they could compromise the confidentiality of patients' identities. We present an approach to inclusion of tract identifiers based on multiply imputed, synthetic data. The idea is to build a predictive model of tract locations, given patient and tumor characteristics, and randomly simulate the tract of each patient by sampling from this model. For the predictive model, we use multivariate regression trees fitted to the latitude and longitude of the population centroid of each tract. We implement the approach in the registry data from California. The method results in synthetic data that reproduce a wide range (but not all) of analyses of census tract socioeconomic cancer disparities and have relatively low disclosure risks, which we assess by comparing individual patients' actual and synthetic tract locations. We conclude with a discussion of how synthetic data sets can be used by researchers with cancer registry data.

Data mining in road crash analysis: the context of developing countries

The recent advancements in the field of data mining have made vast progress in extracting new information and hidden patterns from large datasets which are often overlooked by the traditional statistical approaches. These methods focus on searching for new and interesting hypothesis which were previously unobserved. Road safety researchers working with the crash data from developed world have seen encouraging success in obtaining new insight into crash mechanism through data mining. An attempt was made in this study to apply these advance methods and evaluate their performance in manifesting crash causes for Bangladesh. The study applies hierarchical clustering to identify hazardous clusters, random forest to find important variables explaining each of these clusters, and classification and regression trees to unveil their respective crash mechanisms for the road crash data of Bangladesh. The results identified several new interesting relationships and acknowledged issues related to quality of data.

The impact of post-stroke complications on in-hospital mortality depends on stroke severity

INTRODUCTION

Controversies remain on whether post-stroke complications represent an independent predictor of poor outcome or just a reflection of stroke severity. We aimed to identify which post-stroke complications have the highest impact on in-hospital mortality by using machine learning techniques. Secondary aim was identification of patient's subgroups in which complications have the highest impact.

PATIENTS AND METHODS

Registro Nacional de Ictus de la Sociedad Española de Neurología is a stroke registry from 42 centers from the Spanish Neurological Society. Data from ischemic stroke patients were used to build a random forest by combining 500 classification and regression trees, to weight up the impact of baseline characteristics and post-stroke complications on in-hospital mortality. With the selected variables, a logistic regression analysis was performed to test for interactions.

RESULTS

12,227 ischemic stroke patients were included. In-hospital mortality was 5.9% and median hospital stay was 7(4-10) days. Stroke severity [National Institutes of Health Stroke Scale > 10, OR = 5.54(4.55-6.99)], brain edema [OR = 18.93(14.65-24.46)], respiratory infections [OR = 3.67(3.02-4.45)] and age [OR = 2.50(2.07-3.03) for >77 years] had the highest impact on in-hospital mortality in random forest, being independently associated with in-hospital mortality. Complications have higher odds ratios in patients with baseline National Institutes of Health Stroke Scale <10.

DISCUSSION

Our study identified brain edema and respiratory infections as independent predictors of in-hospital mortality, rather than just markers of more severe strokes. Moreover, its impact was higher in less severe strokes, despite lower frequency.

CONCLUSION

Brain edema and respiratory infections were the complications with a greater impact on in-hospital mortality, with the highest impact in patients with mild strokes. Further efforts on the prediction of these complications could improve stroke outcome.

Meta-CART: A tool to identify interactions between moderators in meta-analysis

In the framework of meta-analysis, moderator analysis is usually performed only univariately. When several study characteristics are available that may account for treatment effect, standard meta-regression has difficulties in identifying interactions between them. To overcome this problem, meta-CART has been proposed: an approach that applies classification and regression trees (CART) to identify interactions, and then subgroup meta-analysis to test the significance of moderator effects. The previous version of meta-CART has its shortcomings: when applying CART, the sample sizes of studies are not taken into account, and the effect sizes are dichotomized around the median value. Therefore, this article proposes new meta-CART extensions, weighting study effect sizes by their accuracy, and using a regression tree to avoid dichotomization. In addition, new pruning rules are proposed. The performance of all versions of meta-CART was evaluated via a Monte Carlo simulation study. The simulation results revealed that meta-regression trees with random-effects weights and a 0.5-standard-error pruning rule perform best. The required sample size for meta-CART to achieve satisfactory performance depends on the number of study characteristics, the magnitude of the interactions, and the residual heterogeneity.

Development of a CART Model to Predict the Synthesis of Cardiotoxic Daunorubicinol in Heart Tissue Samples From Donors With and Without Down Syndrome

Daunorubicin (DAUN) and doxorubicin (DOX) are used to treat a variety of cancers. The use of DAUN and DOX is hampered by the development of cardiotoxicity. Clinical evidence suggests that patients with leukemia and Down syndrome are at increased risk for anthracycline-related cardiotoxicity. Carbonyl reductases and aldo-keto reductases (AKRs) catalyze the reduction of DAUN and DOX into cardiotoxic C-13 alcohol metabolites. Anthracyclines also exert cardiotoxicity by triggering mitochondrial dysfunction. In recent studies, a collection of heart samples from donors with and without Down syndrome was used to investigate determinants for anthracycline-related cardiotoxicity including cardiac daunorubicin reductase activity (DA), carbonyl reductase/AKRs protein expression, mitochondrial DNA content (mtDNA), and AKR7A2 DNA methylation status. In this study, the available demographic, biochemical, genetic, and epigenetic data were integrated through classification and regression trees analysis with the aim of pinpointing the most relevant variables for the synthesis of cardiotoxic daunorubicinol (i.e., DA). Seventeen variables were considered as potential predictors. Leave-one-out-cross-validation was performed for model selection and to estimate the generalization error. The classification and regression trees analysis model and variable importance measures suggest that cardiac mtDNA content, mtDNA(4977) deletion frequency, and AKR7A2 protein content are the most important variables in determining DA.

Associations between ambient air pollutant mixtures and pediatric asthma emergency department visits in three cities: a classification and regression tree approach

BACKGROUND

Characterizing multipollutant health effects is challenging. We use classification and regression trees to identify multipollutant joint effects associated with pediatric asthma exacerbations and compare these results with those from a multipollutant regression model with continuous joint effects.

METHODS

We investigate the joint effects of ozone, NO2 and PM2.5 on emergency department visits for pediatric asthma in Atlanta (1999-2009), Dallas (2006-2009) and St. Louis (2001-2007). Daily concentrations of each pollutant were categorized into four levels, resulting in 64 different combinations or "Day-Types" that can occur. Days when all pollutants were in the lowest level were withheld as the reference group. Separate regression trees were grown for each city, with partitioning based on Day-Type in a model with control for confounding. Day-Types that appeared together in the same terminal node in all three trees were considered to be mixtures of potential interest and were included as indicator variables in a three-city Poisson generalized linear model with confounding control and rate ratios calculated relative to the reference group. For comparison, we estimated analogous joint effects from a multipollutant Poisson model that included terms for each pollutant, with concentrations modeled continuously.

RESULTS AND DISCUSSION

No single mixture emerged as the most harmful. Instead, the rate ratios for the mixtures suggest that all three pollutants drive the health association, and that the rate plateaus in the mixtures with the highest concentrations. In contrast, the results from the comparison model are dominated by an association with ozone and suggest that the rate increases with concentration.

CONCLUSION

The use of classification and regression trees to identify joint effects may lead to different conclusions than multipollutant models with continuous joint effects and may serve as a complementary approach for understanding health effects of multipollutant mixtures.

Comparative effectiveness research of chronic hepatitis B and C cohort study (CHeCS): improving data collection and cohort identification

BACKGROUND AND AIMS

The Chronic Hepatitis Cohort Study (CHeCS) is a longitudinal observational study of risks and benefits of treatments and care in patients with chronic hepatitis B (HBV) and C (HCV) infection from four US health systems. We hypothesized that comparative effectiveness methods-including a centralized data management system and an adaptive approach for cohort selection-would improve cohort selection while controlling data quality and reducing the cost.

METHODS

Cohort selection and data collection were performed primarily via the electronic health record (EHR); cases were confirmed via chart abstraction. Two parallel sources fed data to a centralized data management system: direct EHR data collection with common data elements, and chart abstraction via electronic data capture. An adaptive Classification and Regression Tree (CART) identified a set of electronic variables to improve case ascertainment accuracy.

RESULTS

Over 16 million patient records were collected on 23 case report forms in 2006-2008. The vast majority of data (99.2%) were collected electronically from EHR; only 0.8% was collected via chart abstraction. Initial electronic criteria identified 12,144 chronic hepatitis patients; 10,098 were confirmed via chart abstraction with positive predictive values (PPV) 79 and 83% for HBV and HCV, respectively. CART-optimized models significantly increased PPV to 88 for HBV and 95% for HCV.

CONCLUSIONS

CHeCS is a comparative effectiveness research project that leverages electronic centralized data collection and adaptive cohort identification approaches to enhance study efficiency. The adaptive CART model significantly improved the positive predictive value of cohort identification methods.

Supervised classification in the presence of misclassified training data: a Monte Carlo simulation study in the three group case

Statistical classification of phenomena into observed groups is very common in the social and behavioral sciences. Statistical classification methods, however, are affected by the characteristics of the data under study. Statistical classification can be further complicated by initial misclassification of the observed groups. The purpose of this study is to investigate the impact of initial training data misclassification on several statistical classification and data mining techniques. Misclassification conditions in the three group case will be simulated and results will be presented in terms of overall as well as subgroup classification accuracy. Results show decreased classification accuracy as sample size, group separation and group size ratio decrease and as misclassification percentage increases with random forests demonstrating the highest accuracy across conditions.

Classification and regression trees for epidemiologic research: an air pollution example

BACKGROUND

Identifying and characterizing how mixtures of exposures are associated with health endpoints is challenging. We demonstrate how classification and regression trees can be used to generate hypotheses regarding joint effects from exposure mixtures.

METHODS

We illustrate the approach by investigating the joint effects of CO, NO2, O3, and PM2.5 on emergency department visits for pediatric asthma in Atlanta, Georgia. Pollutant concentrations were categorized as quartiles. Days when all pollutants were in the lowest quartile were held out as the referent group (n = 131) and the remaining 3,879 days were used to estimate the regression tree. Pollutants were parameterized as dichotomous variables representing each ordinal split of the quartiles (e.g. comparing CO quartile 1 vs. CO quartiles 2-4) and considered one at a time in a Poisson case-crossover model with control for confounding. The pollutant-split resulting in the smallest P-value was selected as the first split and the dataset was partitioned accordingly. This process repeated for each subset of the data until the P-values for the remaining splits were not below a given alpha, resulting in the formation of a "terminal node". We used the case-crossover model to estimate the adjusted risk ratio for each terminal node compared to the referent group, as well as the likelihood ratio test for the inclusion of the terminal nodes in the final model.

RESULTS

The largest risk ratio corresponded to days when PM2.5 was in the highest quartile and NO2 was in the lowest two quartiles (RR: 1.10, 95% CI: 1.05, 1.16). A simultaneous Wald test for the inclusion of all terminal nodes in the model was significant, with a chi-square statistic of 34.3 (p = 0.001, with 13 degrees of freedom).

CONCLUSIONS

Regression trees can be used to hypothesize about joint effects of exposure mixtures and may be particularly useful in the field of air pollution epidemiology for gaining a better understanding of complex multipollutant exposures.

Summarizing techniques that combine three non-parametric scores to detect disease-associated 2-way SNP-SNP interactions

Identifying susceptibility genes that influence complex diseases is extremely difficult because loci often influence the disease state through genetic interactions. Numerous approaches to detect disease-associated SNP-SNP interactions have been developed, but none consistently generates high-quality results under different disease scenarios. Using summarizing techniques to combine a number of existing methods may provide a solution to this problem. Here we used three popular non-parametric methods-Gini, absolute probability difference (APD), and entropy-to develop two novel summary scores, namely principle component score (PCS) and Z-sum score (ZSS), with which to predict disease-associated genetic interactions. We used a simulation study to compare performance of the non-parametric scores, the summary scores, the scaled-sum score (SSS; used in polymorphism interaction analysis (PIA)), and the multifactor dimensionality reduction (MDR). The non-parametric methods achieved high power, but no non-parametric method outperformed all others under a variety of epistatic scenarios. PCS and ZSS, however, outperformed MDR. PCS, ZSS and SSS displayed controlled type-I-errors (<0.05) compared to GS, APDS, ES (>0.05). A real data study using the genetic-analysis-workshop 16 (GAW 16) rheumatoid arthritis dataset identified a number of interesting SNP-SNP interactions.

Screening of synthetic PDE-5 inhibitors and their analogues as adulterants: analytical techniques and challenges

The popularity of phosphodiesterase type 5 (PDE-5) enzyme inhibitors for the treatment of erectile dysfunction has led to the increase in prevalence of illicit sexual performance enhancement products. PDE-5 inhibitors, namely sildenafil, tadalafil and vardenafil, and their unapproved designer analogues are being increasingly used as adulterants in the herbal products and health supplements marketed for sexual performance enhancement. To date, more than 50 unapproved analogues of prescription PDE-5 inhibitors were found as adulterants in the literature. To avoid detection of such adulteration by standard screening protocols, the perpetrators of such illegal products are investing time and resources to synthesize exotic analogues and devise novel means for adulteration. A comprehensive review of conventional and advance analytical techniques to detect and characterize the adulterants is presented. The rapid identification and structural elucidation of unknown analogues as adulterants is greatly enhanced by the wide myriad of analytical techniques employed, including high performance liquid chromatography (HPLC), gas chromatography-mass spectrometry (GC-MS), liquid chromatography mass-spectrometry (LC-MS), nuclear magnetic resonance (NMR) spectroscopy, vibrational spectroscopy, liquid chromatography-Fourier transform ion cyclotron resonance-mass spectrometry (LC-FT-ICR-MS), liquid chromatograph-hybrid triple quadrupole linear ion trap mass spectrometer with information dependent acquisition, ultra high performance liquid chromatography-time of flight-mass spectrometry (UHPLC-TOF-MS), ion mobility spectroscopy (IMS) and immunoassay methods. The many challenges in detecting and characterizing such adulterants, and the need for concerted effort to curb adulteration in order to safe guard public safety and interest are discussed.

Improved biomarker performance for the detection of hepatocellular carcinoma by inclusion of clinical parameters

We have previously identified several biomarkers of hepatocellular carcinoma (HCC). The levels of three of these biomarkers were analyzed individually and in combination with the currently used marker, alpha fetoprotein (AFP), for the ability to distinguish between a diagnosis of cirrhosis (n=113) and HCC (n=164). We have utilized several novel biostatistical tools, along with the inclusion of clinical factors such as age and gender, to determine if improved algorithms could be used to increase the probability of cancer detection. Using several of these methods, we are able to detect HCC in the background of cirrhosis with an AUC of at least 0.95. The use of clinical factors in combination with biomarker values to detect HCC is discussed.

Poor response to sertraline in methamphetamine dependence is associated with sustained craving for methamphetamine

BACKGROUND

Depression is common among individuals with methamphetamine (MA) use disorders. As agents that enhance serotonergic function are frequently used to treat depression, one might predict that they would be useful medications for MA dependence. However, clinical trials of serotonergic agents for MA addiction have been unsuccessful.

OBJECTIVE

To identify factors that distinguish MA-dependent research participants who increased MA self-administration while receiving treatment with the selective serotonin reuptake inhibitor (SSRI) sertraline from other groups of participants.

METHOD

Using a dataset from a 12-week randomized, placebo-controlled trial of sertraline (100mg daily) for MA addiction, we identified participants who had completed at least 8 weeks of the trial (n=61 sertraline, n=68 placebo). We compared the proportions of MA-positive urine tests for weeks 8-12 of the trial for these subjects to their pre-randomization baseline, and identified those subjects who increased MA use during treatment. Using classification trees, we then assessed all data collected during the study to identify factors associated with increasing MA use during treatment with sertraline, compared to placebo.

RESULTS

More subjects in the sertraline condition increased MA use during treatment (n=13) than in the placebo condition (n=5; p=0.03). Classification trees identified multiple factors from both pre-treatment and in-treatment data that were associated with increased MA use during treatment. Only elevated in-treatment craving for MA specifically characterized subjects in the sertraline group who increased their MA use.

CONCLUSIONS

Some MA-abusing individuals treated with SSRIs have sustained craving with an increased propensity to relapse during treatment despite psychosocial treatment interventions.

Altered gut flora are associated with septic complications and death in critically ill patients with systemic inflammatory response syndrome

BACKGROUND

Gut under severe insult is considered to have an important role in promoting infection and multiple organ dysfunction syndrome from the viewpoint of altered intestinal epithelium, immune system and commensal bacteria. There are few reports, however, about the relationship between gut flora and septic complications.

METHODS

We analyzed gut flora in patients with systemic inflammatory response syndrome (SIRS) and evaluated key bacteria and their cutoff values for infectious complications and mortality by using classification and regression trees (CART). Eighty-one SIRS patients with a serum C-reactive protein level higher than 10 mg/dL treated in the intensive care unit (ICU) for more than 2 days were included for the study. We quantitatively evaluated nine types of bacteria in fecal samples by plate or tube technique. Two hundred seventy-one samples were analyzed using CART and logistic regression.

RESULTS

The dominant factors for complication of enteritis were the minimum number of total obligate anaerobes and the maximum number of Staphylococcus and Enterococcus. The dominant factors for complication of bacteremia were the minimum numbers of total obligate anaerobes and total facultative anaerobes. The dominant factors for mortality were the numbers of total obligate anaerobes and total facultative anaerobes and age.

CONCLUSIONS

A decrease in total obligate anaerobes and an increase in pathogenic bacteria in the gut are associated with septic complications and mortality in patients with SIRS. The altered gut flora may be a potential prognostic marker in SIRS patients.

Availability and capacity of substance abuse programs in correctional settings: A classification and regression tree analysis

OBJECTIVE TO BE ADDRESSED: The purpose of this study was to investigate the structural and organizational factors that contribute to the availability and increased capacity for substance abuse treatment programs in correctional settings. We used classification and regression tree statistical procedures to identify how multi-level data can explain the variability in availability and capacity of substance abuse treatment programs in jails and probation/parole offices.

METHODS

The data for this study combined the National Criminal Justice Treatment Practices (NCJTP) Survey and the 2000 Census. The NCJTP survey was a nationally representative sample of correctional administrators for jails and probation/parole agencies. The sample size included 295 substance abuse treatment programs that were classified according to the intensity of their services: high, medium, and low. The independent variables included jurisdictional-level structural variables, attributes of the correctional administrators, and program and service delivery characteristics of the correctional agency.

RESULTS

The two most important variables in predicting the availability of all three types of services were stronger working relationships with other organizations and the adoption of a standardized substance abuse screening tool by correctional agencies. For high and medium intensive programs, the capacity increased when an organizational learning strategy was used by administrators and the organization used a substance abuse screening tool. Implications on advancing treatment practices in correctional settings are discussed, including further work to test theories on how to better understand access to intensive treatment services. This study presents the first phase of understanding capacity-related issues regarding treatment programs offered in correctional settings.

Using administrative data to identify indications for elective primary cesarean delivery

OBJECTIVE

To develop a methodology to identify indications and normative rates for elective primary cesarean delivery using administrative data.

DATA SOURCES/STUDY SETTING

All delivery discharges in 1995, as reported to the California Office of Statewide Health Planning and Development (secondary data).

STUDY DESIGN

Retrospective population based study.

DATA COLLECTION/EXTRACTION

Data were entered into a recursive partitioning algorithm to develop a hierarchy of conditions by which patients with multiple conditions could be sorted with respect to the binary outcome of labor or elective primary cesarean without labor. This hierarchy was examined for its clinical consistency, validated on a second sample, and compared with results obtained from logistic regression.

PRINCIPAL FINDINGS

Four percent (19,664) of delivery discharges in 1995 underwent elective primary cesarean. Twelve clinical conditions contributed to the hierarchy, and accounted for 92.9 percent of all women experiencing elective primary cesarean delivery. The remaining 7.1 percent of the elective primary cesarean cases were classified as "unspecified."

CONCLUSIONS

A standardized methodology (utilizing recursive partitioning algorithms) for assigning indications for elective primary cesarean is presented. This methodology relies on administrative data, classifies women with complex comorbidity patterns into clinically relevant subpopulations, and can be used to establish normative rates for benchmarking specific indications for cesarean delivery.

Novel statistical classification model of type 2 diabetes mellitus patients for tailor-made prevention using data mining algorithm

To estimate the usefulness of data mining algorithms for extracting risk predictors of diabetic vascular complications in proper order in the future, we tried applying the Classification and Regression Trees (CART) method to the prevalence data of 165 type 2 diabetic outpatients and already known risk factors. Among the 6 categorical and 15 continuous risk factors, age (cutoff: 65.4) was the best predictor for classifying patients into groups with and without macroangiopathy (p=0.000). Body weight (cutoff: 53.9) was the best predictor (p=0.006) in the older group (age >65.4), whereas systolic blood pressure (cutoff: 144.5) was the best predictor in the remaining group (p=0.002). Age (cutoff: 64.8) was also the best predictor for categorizing them into groups with and without microangiopathy (p=0.000). In the older group (age >64.8), BMI (cutoff: 21.5) was the best predictor (p=0.001), whereas morbidity term (cutoff: 15.5) was the best predictor in the other group (p=0.01 0). Because the orders and values of all risk factors and cutoff points mined were reasonable clinically, this method may have the potential to highlight predictors in order of importance to apply tailor-made prevention of diabetic vascular complications.