Clustering of the Values of a Response Variable and Simultaneous Covariate Selection Using a Stepwise Algorithm

In supervised learning the number of values of a response variable can be very high. Grouping these values in a few clusters can be useful to perform accurate supervised classification analyses. On the other hand selecting relevant covariates is a crucial step to build robust and efficient prediction models. We propose in this paper an algorithm that simultaneously groups the values of a response variable into a limited number of clusters and selects stepwise the best covariates that discriminate this clustering. These objectives are achieved by alternate optimization of a user-defined model selection criterion. This process extends a former version of the algorithm to a more general framework. Moreover possible further developments are discussed in detail.

Identifying High-Risk Medication Prescriptions to Prevent Potentially Severe Adverse Drug Events in Primary-Care Patients with Chronic Multimorbidities:The Polychrome Study

Background:The association between multimorbidities and polypharmacy among elderly individuals is well documented,and polypharmacy has been shown to increase the risk of adverse drug events(ADEs).However,little information is available about the risks associated with the lifelong use of medications to treat chronic multimorbidities.Objective:To determine the prevalence and nature of high-risk prescriptions among primary-care patients with chronic multimorbidities.Methods:We studied a weighted stratified random sample of 105 prescriptions for different patients with chronic multimorbidities taken from the Polychrome database established using information from the French primary-care record database(Observatoire de la Médecine Générale).A medication review was conducted to identify contra-indications and potential drug-drug interactions for each prescription.Results:Contra-indications were identified for 60(57.1%)prescriptions,potential drug-drug interactions for 70(66.7%),absolute contra-indications for 9(8.6%),and inadvisable drug combinations for 11(10.5%).In all,19(18.1%)different patients were at risk for major ADEs.Cardiovascular and nervous-system drugs contributed 66.2% of contra-indications and 69.3% of potential drug-drug interactions.Conclusions:This exploratory study confirms the high prevalence and potential seriousness of prescriptions at risk for ADEs in a population of primary-care patients with chronic multimorbidities.The high prevalence of interactions involving the cardiovascular and nervous systems indicates that efforts to improve prescription practices should target these two categories of conditions and drugs in patients with chronic multimorbidities.

Clinical imaging factors of excellent outcome after thrombolysis in large- vessel stroke: a THRACE subgroup analysis

Background For patients with stroke with large-vessel occlusion(LVO),study of factors predicting response to intravenous thrombolysis(IVT)would allow identifying subgroups with high expected gain,and those for whom it could be considered as futile,and even detrimental.From patients included in the Mechanical Thrombectomy After Intravenous Alteplase vs Alteplase Alone After Stroke trial,we investigated clinical-imaging factors associated with optimal response to IVT.Methods We included patients receiving IVT alone.Excellent outcome was defined by a 3-month modified Rankin Scale(mRS)score≤1.Clinical-imaging predictors were assessed on multivariate analysis after multiple imputations.The predictive performance of the model was assessed with the C-statistic.Results Among 247 patients with LVO treated with IVT alone,77(31%)showed 3-month mRS≤1.Predictors of 3-month mRS≤1 were no medical history of hypertension(OR 2.43;95%CI 1.74 to 3.38;p=0.007);no current smoking(OR 2.76;95%CI 1.79 to 4.26;p=0.02);onset-to IVT time(OR 0.47 per hour increase;95%CI 0.23 to 0.78;p=0.003);diffusion-weighted imaging(DWI)volume(OR 0.78 per 10 mL increase;95%CI 0.68 to 0.89;p=0.0004);presence of susceptibility vessel sign(SVS)(OR 7.89;95%CI 1.65 to 37.78;p=0.01)and SVS length(OR 0.87 per mm increase;95%CI 0.80 to 0.94;p=0.001).The prediction models showed a C-statistic=0.79(95%CI 0.79 to 0.80).Conclusions In patients with stroke with anterior-circulation LVO treated with IVT alone,predictors of excellent outcome at 3 months were no medical history of hypertension or current smoking,reduced onset-to IVT time,small DWI volume,presence of SVS and short SVS length.These predictive factors could help practitioners in decision-making for IVT implementation in reperfusion strategies,all the more for the drip and ship paradigm.Trial registration number NCT01062698.