A comparison of regression trees, logistic regression, generalized additive models, and multivariate adaptive regression splines for predicting AMI mortality

Austin, Peter C.

doi:10.1002/sim.2770

Cited by 133 publications

(100 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We used data on 9,081 patients who were discharged alive with an acute myocardial infarction (AMI or heart attack) from 102 hospitals in Ontario, Canada, between April 1, 1999 and March 31, 2001 and who did not die on the day of discharge. These data are similar to those reported elsewhere (Austin 2007c;) and were collected as part of the Enhanced Feedback for Effective Cardiac Treatment (EFFECT) Study, an initiative focused on improving the quality of care for cardiovascular disease patients in Ontario (Tu et al 2004). Data on patient demographics, presenting signs and symptoms, classic cardiac risk factors, comorbid conditions and vascular history, vital signs on admission, and results of laboratory tests were abstracted directly from patients' medical records.…”

supporting

confidence: 61%

Type I Error Rates, Coverage of Confidence Intervals, and Variance Estimation in Propensity-Score Matched Analyses

Austin¹

2009

The International Journal of Biostatistics

156

153

View full text Add to dashboard Cite

Propensity-score matching is frequently used in the medical literature to reduce or eliminate the effect of treatment selection bias when estimating the effect of treatments or exposures on outcomes using observational data. In propensity-score matching, pairs of treated and untreated subjects with similar propensity scores are formed. Recent systematic reviews of the use of propensity-score matching found that the large majority of researchers ignore the matched nature of the propensity-score matched sample when estimating the statistical significance of the treatment effect. We conducted a series of Monte Carlo simulations to examine the impact of ignoring the matched nature of the propensity-score matched sample on Type I error rates, coverage of confidence intervals, and variance estimation of the treatment effect. We examined estimating differences in means, relative risks, odds ratios, rate ratios from Poisson models, and hazard ratios from Cox regression models. We demonstrated that accounting for the matched nature of the propensity-score matched sample tended to result in type I error rates that were closer to the advertised level compared to when matching was not incorporated into the analyses. Similarly, accounting for the matched nature of the sample tended to result in confidence intervals with coverage rates that were closer to the nominal level, compared to when matching was not taken into account. Finally, accounting for the matched nature of the sample resulted in estimates of standard error that more closely reflected the sampling variability of the treatment effect compared to when matching was not taken into account.

show abstract

supporting

confidence: 61%

Type I Error Rates, Coverage of Confidence Intervals, and Variance Estimation in Propensity-Score Matched Analyses

Austin¹

2009

The International Journal of Biostatistics

156

153

View full text Add to dashboard Cite

show abstract

“…In some cases, it has been shown that the predictive accuracy of C&RT is somewhat lower than for logistic models [18]. In our case, the results obtained using C&RT and the regression model are comparable.…”

Section: Discussionsupporting

confidence: 58%

“…A more recently developed non-parametric tool suitable for this type of problems is called Classification and Regression Tree (C&RT) [13] . This technique is employed in clinical research with the aim to obtain a simple pattern to classify subjects between ill and healthy, and to get information about which groups of individuals could benefit more from targeted interventions [14][15][16][17][18]. One of the main advantages of C&RT is that the result of the analysis is a classification tree, which is easier to interpret in clinical practice [14].…”

Section: Introductionmentioning

confidence: 99%

Factors affecting recall rate and false positive fraction in breast cancer screening with breast tomosynthesis – A statistical approach

et al. 2015

View full text Add to dashboard Cite

Link to publication Citation for published version (APA): Rosso, A., Lång, K., Petersson, I., & Zackrisson, S. (2015). Factors affecting recall rate and false positive fraction in breast cancer screening with breast tomosynthesis -A statistical approach. Breast, 24(5), 680-686. DOI: 10.1016680-686. DOI: 10. /j.breast.2015 General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.• You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal AbstractIn this study, we investigate which factors affect the false positive fraction (FPF) for digital breast tomosynthesis (DBT) compared to digital mammography (DM) in a screening population by using classification and regression trees (C&RT) and binary marginal generalized linear models.The data was obtained from the Malmö Breast Tomosynthesis Screening Trial, which aimed to compare the performance of DBT to DM in breast cancer screening. By using data from the first half of the study population (7 500 women), a tree with the recall probability for different groups was calculated. The effect of age and breast density on the FPF was estimated using a binary marginal generalized linear model.Our results show that breast density and breast cancer were the main factors influencing recall. The FPF is mainly affected by breast density and increases with breast density for DBT and DM.In conclusion, the results obtained with C&RT are easy to interpret and similar to those obtained using binary marginal generalized linear models. The FPF is approximately 40% higher for DBT compared to DM for all breast density categories.

show abstract

“…Our model differs from the study by BONDERMAN et al [35], as we specifically aimed to identify LHF as alternative cause of PH, whereas the model of BONDERMAN et al [35] aimed to exclude PAH and included patients with normal pulmonary pressures. Considering this, and as logistic regression models have better predictive characteristics compared to regression trees, the model of BONDERMAN et al [35] and our own model may have additional value in reducing the need for RHC [36]. The LHF patient cohort in our study displays specific characteristics, with relatively elevated PVR and pulmonary artery pressures.…”

Section: Figurementioning

confidence: 65%

Noninvasive identification of left-sided heart failure in a population suspected of pulmonary arterial hypertension

et al. 2015

View full text Add to dashboard Cite

Exclusion of pulmonary hypertension secondary to left-sided heart disease (left heart failure (LHF)) is pivotal in the diagnosis of pulmonary arterial hypertension (PAH). In case of doubt, invasive measurements are recommended. The aim of the present study was to investigate whether it is possible to diagnose LHF using noninvasive parameters in a population suspected of PAH. 300 PAH and 80 LHF patients attended our pulmonary hypertension clinic before August 2010, and were used to build the predictive model. 79 PAH and 55 LHF patients attended our clinic from August 2010, and were used for prospective validation.A medical history of left heart disease, S deflection in V1 plus R deflection in V6 in millimetres on ECG, and left atrial dilation or left valvular heart disease that is worse than mild on echocardiography were independent predictors of LHF. The derived risk score system showed good predictive characteristics: R 2 =0.66 and area under the curve 0.93. In patients with a risk score ⩾72, there is 100% certainty that the cause of pulmonary hypertension is LHF. Using this risk score system, the number of right heart catheterisations in LHF may be reduced by 20%.In a population referred under suspicion of PAH, a predictive model incorporating medical history, ECG and echocardiography data can diagnose LHF noninvasively in a substantial percentage of cases. @ERSpublications A risk score system can identify left-heart failure as alternative PH cause in patients suspected of PAH clinically http://ow.ly/JAIkR

show abstract

A comparison of regression trees, logistic regression, generalized additive models, and multivariate adaptive regression splines for predicting AMI mortality

Cited by 133 publications

References 49 publications

Type I Error Rates, Coverage of Confidence Intervals, and Variance Estimation in Propensity-Score Matched Analyses

Type I Error Rates, Coverage of Confidence Intervals, and Variance Estimation in Propensity-Score Matched Analyses

Factors affecting recall rate and false positive fraction in breast cancer screening with breast tomosynthesis – A statistical approach

Noninvasive identification of left-sided heart failure in a population suspected of pulmonary arterial hypertension

Contact Info

Product

Resources

About