Use of classification and regression tree (CART), to identify hemoglobin A1C (HbA 1C ) cut-off thresholds predictive of poor tuberculosis treatment outcomes and associated risk factors

Mburu, Josephine W.; Kingwara, Leonard; Ester, Magiri; Nyerere, Andrew

doi:10.1016/j.jctube.2018.01.002

Cited by 14 publications

(11 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similarly, the above steps were repeated for parameter tuning of the C4.5. Considering α=1.414 the low and high level of MNIL has been considered 8 and 42, respectively to cover its region of interest, which is [1,49]. Also, the low and high levels of confidence factor (CF) were set to 0.074 and 0.426 respectively to cover its range of (0,0.5].…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Detecting factors associated with polypharmacy in general practitioners’ prescriptions: A data mining approach

2020

View full text Add to dashboard Cite

Prescribing and consuming drugs more than necessary is considered as polypharmacy, which is both wasteful and harmful. The purpose of this paper is to establish an innovative data mining framework for analyzing physicians' prescriptions regarding polypharmacy. The approach consists of three main steps: pre-modeling, modeling, and post-modeling. In the first step, after collecting and cleaning the raw data, several novel physicians' features are extracted. In the modeling step, two popular decision trees, i.e., C4.5 and Classification and Regression Tree (CART), are applied to generate a set of If-Then rules in a tree-shaped structure to detect and describe physicians' features associated with polypharmacy. In a novel approach, the response surface method (RSM) as a tool for hyper-parameter tuning is simultaneously applied along with correlation-based feature selection (CFS) to enhance the performance of the algorithms. In the post-modeling step, the discovered knowledge is visualized to make the results more perceptible, then is presented to domain experts to evaluate whether they make sense or not. The framework has been applied to a real-world dataset of prescriptions. The results have been confirmed by the experts, which demonstrates the capabilities of the data mining framework in the detection and analysis of polypharmacy. The derived If-Then rules can be beneficial for healthcare managers and policymakers to recognize physicians' prescribing pattern and take suitable action to support medicines management and develop high quality prescribing guidelines.

show abstract

Section: Resultsmentioning

confidence: 99%

“…It also can handle the missing values. CART is a popular machine learning algorithm in the field of healthcare and medical research for more than two decades and has been utilized in various medical researches [42], [45], [46], [47], [48], [49].…”

Section: Modeling 421 Model Buildingmentioning

confidence: 99%

Detecting factors associated with polypharmacy in general practitioners’ prescriptions: A data mining approach

2020

View full text Add to dashboard Cite

show abstract

“…Typically these studies gather demographic and medical statistics on a cohort, observe their adherence and outcomes throughout the trial, then retrospectively apply survival [18,35] or logistic regression [32] analysis to determine covariates predictive of failure. Newer work has improved classification accuracy via machine learning techniques such as Decision Trees, Neural Networks, Support Vector Machines and more [14,15,22,34]. However, the conclusions connecting predictors to risk are largely the same as in previous medical literature.…”

Section: Related Workmentioning

confidence: 99%

Learning to Prescribe Interventions for Tuberculosis Patients Using Digital Adherence Data

Killian

Wilder

Sharma

et al. 2019

Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery &Amp; Data Mining

View full text Add to dashboard Cite

Digital Adherence Technologies (DATs) are an increasingly popular method for verifying patient adherence to many medications. We analyze data from one city served by 99DOTS, a phone-call-based DAT deployed for Tuberculosis (TB) treatment in India where nearly 3 million people are afflicted with the disease each year. The data contains nearly 17,000 patients and 2.1M dose records. We lay the groundwork for learning from this real-world data, including a method for avoiding the effects of unobserved interventions in training data used for machine learning. We then construct a deep learning model, demonstrate its interpretability, and show how it can be adapted and trained in different clinical scenarios to better target and improve patient care. In the real-time risk prediction setting our model could be used to proactively intervene with 21% more patients and before 76% more missed doses than current heuristic baselines. For outcome prediction, our model performs 40% better than baseline methods, allowing cities to target more resources to clinics with a heavier burden of patients at risk of failure. Finally, we present a case study demonstrating how our model can be trained in an end-to-end decision focused learning setting to achieve 15% better solution quality in an example decision problem faced by health workers.

show abstract

“…Therefore, examining and understanding the risk factors and their interactions may be central to understanding the epidemiology of CCPP and helping in the design and analysis of other related CCPP epidemiological studies. The Classification and Regression Tree (CART) is a machine-learning algorithm that has been utilized in clinical settings as an ideal tool for clinical decision-making and to assess risk factors [ 25 , 26 ]. Despite its potential, the CART has been seldom used in population-based epidemiology and genetic epidemiology studies.…”

Section: Introductionmentioning

confidence: 99%

Determination of Seroprevalence of Contagious Caprine Pleuropneumonia and Associated Risk Factors in Goats and Sheep Using Classification and Regression Tree

Selim

Megahed

Kandeel

et al. 2021

Animals

View full text Add to dashboard Cite

Classification and Regression Tree (CART) analysis is a potentially powerful tool for identifying risk factors associated with contagious caprine pleuropneumonia (CCPP) and the important interactions between them. Our objective was therefore to determine the seroprevalence and identify the risk factors associated with CCPP using CART data mining modeling in the most densely sheep- and goat-populated governorates. A cross-sectional study was conducted on 620 animals (390 sheep, 230 goats) distributed over four governorates in the Nile Delta of Egypt in 2019. The randomly selected sheep and goats from different geographical study areas were serologically tested for CCPP, and the animals’ information was obtained from flock men and farm owners. Six variables (geographic location, species, flock size, age, gender, and communal feeding and watering) were used for risk analysis. Multiple stepwise logistic regression and CART modeling were used for data analysis. A total of 124 (20%) serum samples were serologically positive for CCPP. The highest prevalence of CCPP was between aged animals (>4 y; 48.7%) raised in a flock size ≥200 (100%) having communal feeding and watering (28.2%). Based on logistic regression modeling (area under the curve, AUC = 0.89; 95% CI 0.86 to 0.91), communal feeding and watering showed the highest prevalence odds ratios (POR) of CCPP (POR = 3.7, 95% CI 1.9 to 7.3), followed by age (POR = 2.1, 95% CI 1.6 to 2.8) and flock size (POR = 1.1, 95% CI 1.0 to 1.2). However, higher-accuracy CART modeling (AUC = 0.92, 95% CI 0.90 to 0.95) showed that a flock size >100 animals is the most important risk factor (importance score = 8.9), followed by age >4 y (5.3) followed by communal feeding and watering (3.1). Our results strongly suggest that the CCPP is most likely to be found in animals raised in a flock size >100 animals and with age >4 y having communal feeding and watering. Additionally, sheep seem to have an important role in the CCPP epidemiology. The CART data mining modeling showed better accuracy than the traditional logistic regression.

show abstract

Use of classification and regression tree (CART), to identify hemoglobin A1C (HbA 1C ) cut-off thresholds predictive of poor tuberculosis treatment outcomes and associated risk factors

Cited by 14 publications

References 46 publications

Detecting factors associated with polypharmacy in general practitioners’ prescriptions: A data mining approach

Detecting factors associated with polypharmacy in general practitioners’ prescriptions: A data mining approach

Learning to Prescribe Interventions for Tuberculosis Patients Using Digital Adherence Data

Determination of Seroprevalence of Contagious Caprine Pleuropneumonia and Associated Risk Factors in Goats and Sheep Using Classification and Regression Tree

Contact Info

Product

Resources

About