Classification And Regression Tree Analysis Research Articles

Use of Atrial Strain to Predict Atrial Fibrillation After Cerebral Ischemia

ObjectivesThis study sought to identify whether atrial strain could be used as an imaging biomarker to predict atrial fibrillation (AF). BackgroundAF is found in up to 30% of cryptogenic cerebrovascular accidents (CVAs), which themselves account for 30% to 40% of ischemic CVA. MethodsThis observational study evaluated all patients who had an echocardiogram (transthoracic echocardiogram [TTE]) following presentation with cryptogenic CVA from 2010 to 2014. The TTEs were evaluated for reservoir strain (ƐR), contractile strain (ƐCt), and conduit atrial strain (ƐCd) using speckle tracking. Baseline clinical and TTE characteristics of patients who developed AF over 5 years of follow-up and those who did not were compared. The independent and incremental predictive value of atrial strain over established clinical models was assessed. Discriminatory cutpoints were defined using a Classification and Regression Tree (CART) analysis to identify patients at risk of developing AF. ResultsOf 538 patients, 61 (11%) developed AF, and this occurred within 2 years in 85% of patients. Patients who developed AF were older, had higher clinical risk scores, had higher LA volume, and had lower atrial strain than did those who did not develop AF. The area under the receiver-operating characteristic curve was 0.85 for ƐR, 0.83 for ƐCt, and 0.76 for ƐCd (all p < 0.001). The nested Cox regression model showed that ƐR (p = 0.03) and ƐCt (p < 0.001) demonstrated independent and incremental predictive value over the clinical risk. CART analysis identified ƐR ≤21.4%, ƐCd >10.4%, and CHARGE-AF (Cohorts for Heart and Aging Research in Genomic Epidemiology Atrial Fibrillation) score >7.8% as discriminatory for AF, with a 13-fold greater hazard of AF (p < 0.001) in patients with increased clinical risk and reduced ƐR. However, validation is needed for these strain cutoffs for detection of AF. ConclusionsLeft atrial strain adds independent and incremental predictive value to current risk-prediction models for AF following cryptogenic CVA. Further studies should examine the implications of these findings for AF monitoring or empiric anticoagulation.

Quality and Clinical Outcomes Associated with a Gentamicin Use System Change for Managing Chorioamnionitis.

Chorioamnionitis is an intra-amniotic infection with serious maternal and neonatal complications. Clinical studies suggest antibiotic administration before delivery reduces the risk of complications compared to after delivery. Our center implemented a standardized intrapartum gentamicin computerized provider order entry and dosage form dispensing system intended to improve treatment initiation efficiency in hospitalized obstetric patients. The primary objective of this retrospective study was to determine if these system changes were associated with decreased time from gentamicin ordering to administration in patients with chorioamnionitis. A secondary objective was to compare clinical outcomes before and after system changes. Classification and regression tree (CART) analyses was applied to identify key predictors. Results demonstrated a trend towards reduced time to administration in the post-implementation group. Clinical outcomes were not altered. CART analysis revealed that post-implementation assignment and length of membrane rupture predicted shorter time to gentamicin initiation. This study suggests that the specific system changes we implemented were safe and improved efficiency, but additional changes are needed to have a clinically significant impact.

A survey of quality of life indicators in the Romanian Roma population following the ‘Decade of Roma Inclusion’

Background: This study explores how the Roma in Romania, the EU’s most concentrated population, are faring in terms of a number of quality of life indicators, including poverty levels, healthcare, education, water, sanitation, and hygiene. It further explores the role of synthetic populations and modelling in identifying at-risk populations and delivering targeted aid. Methods: 135 surveys were conducted across five geographically diverse Romanian communities. Household participants were selected through a comprehensive random walk method. Analyses were conducted on all data using Pandas for Python. Combining land scan data, time-use survey analyses, interview data, and ArcGIS, the resulting synthetic population was analysed via classification and regression tree (CART) analysis to identify hot-spots of need, both ethnically and geographically. Results: These data indicate that the Roma in Romania face significant disparities in education, with Roma students less likely to progress beyond 8 th grade. In addition, the Roma population remains significantly disadvantaged with regard to safe and secure housing, poverty, and healthcare status, particularly in connection to diarrheal disease. In contrast, however, both Roma and non-Roma in rural areas face difficulties regarding full-time employment, sanitation, and water, sanitation, and hygiene infrastructure. In addition, the use of a synthetic population can generate information about ‘hot spots’ of need, based on geography, ethnicity, and type of aid required. Conclusions: These data demonstrate the challenges that remain to the Roma population in Romania, and also point to the myriad of ways in which all rural Romanians, regardless of ethnicity, are encountering hardship. This study highlights an approach that combines traditional survey data with more wide-reaching geographically based data and CART analysis to determine ‘hot spot’ areas of need in a given population. With the appropriate inputs, this tool can be extrapolated to any population in any country.

Mild Elevation of Pulmonary Arterial Pressure as a Predictor of Mortality.

Normal mean pulmonary arterial pressure (mPAP) is 14.0 ± 3.3 mm Hg (mean ± SD). The prognostic relevance of mildly elevated mPAP not fulfilling the definition of pulmonary hypertension (PH; mPAP ≥ 25 mm Hg) has not been prospectively evaluated in a real-world setting. To assess the association of resting mPAP with all-cause mortality in a retrospective and a prospective cohort of patients with unexplained dyspnea and/or at risk of PH. Prognostic cutoffs were calculated by means of 1) classification and regression tree (CART) analysis without any preset thresholds, and 2) preset thresholds on the basis of literature data defining mPAP as lower-normal (≤mean + 1 SD), upper-normal (between mean + 1 SD and mean + 2 SD), borderline (between mean + 2 SD and 25 mm Hg), and manifest PH (≥25 mm Hg). We performed univariate and multivariate survival analysis adjusted for age and comorbidities. We enrolled 547 patients, of whom 137, 56, 64, and 290 presented with lower-normal, upper-normal, or borderline mPAP, and manifest PH, respectively. The CART analysis on mPAP discriminated three prognostic groups: mPAP less than 17 mm Hg, 17 to 26 mm Hg, and greater than 26 mm Hg, with significantly decreasing survival. The univariate analysis on the basis of preset thresholds showed that upper-normal mPAP, borderline mPAP, and manifest PH were significantly associated with poor survival compared with lower-normal mPAP. In the multivariate model, considering age and comorbidities, only borderline mPAP (hazard ratio, 2.37; 95% confidence interval, 1.14-4.97; P = 0.022) and manifest PH (hazard ratio, 5.05; 95% confidence interval, 2.79-9.12; P < 0.001) were significantly associated with poor survival. In patients at risk for PH and/or with unexplained dyspnea, CART analysis detects prognostic thresholds at a resting mPAP of 17 mm Hg and 26 mm Hg, and values between 20 mm Hg and 25 mm Hg represent an independent predictor of poor survival. Clinical trial registered with www.clinicaltrials.gov (NCT 01607502).

Tip-Apex Distance Is Most Important of Six Predictors of Screw Cutout After Internal Fixation of Intertrochanteric Fractures in Women.

Background:Six risk factors for screw cutout after internal fixation of intertrochanteric fractures have been reported. The purpose of the present study was to evaluate and compare the impact of the 6 risk factors of screw cutout to clarify the most important one.Methods:We enrolled 8 consecutive patients who had screw cutout and 48 random control subjects after internal fixation of intertrochanteric fractures treated with proximal femoral nail antirotation systems at our institution. All of the patients were female. The group that had screw cutout and the control group were retrospectively evaluated and compared with respect to the OTA/AO classification, presence of a posterolateral fragment, types of reduction pattern on anteroposterior and lateral radiographic images, position of the screw, and the presence of a tip-apex distance (TAD) of ≥20 mm. The impact of each factor on screw cutout was assessed using backward stepwise multivariable logistic regression analysis with the Akaike information criterion. Risk stratification was assessed using classification and regression tree (CART) analysis.Results:Among 6 risk factors, only a TAD of ≥20 mm had a significant impact on screw cutout, with an adjusted odds ratio of 12.4 (95% confidence interval, 1.6 to 129.0; p = 0.019). CART analysis also demonstrated that a TAD of ≥20 mm was the most important risk stratification factor (p < 0.001).Conclusions:Among the 6 previously reported screw cutout-related factors, only a TAD of ≥20 mm was associated with screw cutout after internal fixation of intertrochanteric fractures with proximal femoral nail antirotation systems.Level of Evidence:Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

Modeling flash floods in ungauged mountain catchments of China: A decision tree learning approach for parameter regionalization

Flash floods in small mountain catchments are one of the most frequent causes of loss of life and property from natural hazards in China. Hydrological models can be a useful tool for the anticipation of these events and the issuing of timely warnings. One of the main challenges of setting up such a system is finding appropriate model parameter values for ungauged catchments. Previous studies have shown that the transfer of parameter sets from hydrologically similar gauged catchments is one of the best performing regionalization methods. However, a remaining key issue is the identification of suitable descriptors of similarity. In this study, we use decision tree learning to explore parameter set transferability in the full space of catchment descriptors. For this purpose, a semi-distributed rainfall-runoff model is set up for 35 catchments in ten Chinese provinces. Hourly runoff data from in total 858 storm events are used to calibrate the model and to evaluate the performance of parameter set transfers between catchments. We then present a novel technique that uses the splitting rules of classification and regression trees (CART) for finding suitable donor catchments for ungauged target catchments. The ability of the model to detect flood events in assumed ungauged catchments is evaluated in series of leave-one-out tests. We show that CART analysis increases the probability of detection of 10-year flood events in comparison to a conventional measure of physiographic-climatic similarity by up to 20%. Decision tree learning can outperform other regionalization approaches because it generates rules that optimally consider spatial proximity and physical similarity. Spatial proximity can be used as a selection criteria but is skipped in the case where no similar gauged catchments are in the vicinity. We conclude that the CART regionalization concept is particularly suitable for implementation in sparsely gauged and topographically complex environments where a proximity-based regionalization concept is not applicable.

The TWIST Algorithm Predicts Time to Walking Independently After Stroke

Background and Objective. The likelihood of regaining independent walking after stroke is of concern to patients and their families and influences hospital discharge planning. The objective of this study was to explore factors that could be combined in an algorithm for predicting whether and when a patient will walk independently after stroke. Methods. Adults with new lower limb weakness were recruited within 3 days of having a stroke. Clinical assessment, transcranial magnetic stimulation, and magnetic resonance imaging were completed 1 to 2 weeks poststroke. Classification and regression tree (CART) analysis was used to identify factors that predicted whether a patient achieved independent walking by 6 or 12 weeks, or remained dependent at 12 weeks. Results. We recruited 41 patients (24 women; median age 72 years, range 43-96 years). The CART analysis results were used to create the Time to Walking Independently after STroke (TWIST) algorithm, which made accurate predictions for 95% of patients. Patients with a trunk control test score >40 at 1 week walked independently within 6 weeks. Patients with a trunk control test score <40 only achieved independent walking by 12 weeks if they also had hip extension strength of Medical Research Council grade 3 or more. Neurophysiological and neuroimaging measures did not predict independent walking after stroke. Conclusions. In this exploratory study, the TWIST algorithm accurately predicted whether and when an individual patient walked independently after stroke using simple bedside measures 1 week poststroke. Further work is required to develop and validate this algorithm in a larger study.

P20-4 - Comparable Prognostic Impacts of Plasma BNP Levels Among HFpEF, Borderline HFpEF, and HFpEF-A Report From the CHART-2 Study

Background: Although the usefulness of plasma levels of B-type natriuretic peptide (BNP) as a prognostic indicator has been established for patients with heart failure (HF) with reduced ejection fraction (EF) (HFrEF, EF ≤ 40%), it remains unknown whether BNP is also a useful prognostic marker for HF with preserved EF (HFpEF, EF > 50%) or borderline HFpEF (40% < EF ≤ 50%). Methods and Results: We analyzed the data of 4301 consecutive chronic HF patients in stage C/D (mean 69 years, female 32%) in our CHART-2 Study (N = 10,219). During median 6.3-year follow up, 887, 218 and 330 all-cause deaths occurred in HFpEF (N = 2893), borderline HFpEF (N = 666) and HFrEF (N = 742), respectively. Although median BNP levels progressively increased from HFpEF, borderline HFpEF to HFrEF (85.3, 127 and 216 pg/mL, respectively P < .001), the relationship between log2 BNP levels and the mortality risk was comparable among the 3 groups. Moreover, log2 BNP levels were similarly associated with cardiovascular and non-cardiovascular deaths and HF admission in all groups. In addition, when evaluated with the cutoff BNP levels obtained by the survival CART (classification and regression trees) analysis, the mortality risk was effectively stratified by the BNP levels in each group; patients with 30–100, 100–300 and ≥300 pg/mL had comparably increasing mortality risk in all classes. Conclusion: BNP levels had comparable prognostic impacts among HFpEF, borderline HFpEF and HFrEF.

Risk factors for amendment in type, duration and setting of prescribed outpatient parenteral antimicrobial therapy (OPAT) for adult patients with cellulitis: a retrospective cohort study and CART analysis

Purpose of the studyTo measure the percentage rate and risk factors for amendment in the type, duration and setting of outpatient parenteral antimicrobial therapy (OPAT) for the treatment of cellulitis.Study...

Acute Pediatric Monteggia Fractures: A Conservative Approach to Stabilization.

In 2015, a multicenter study group proposed a treatment algorithm for pediatric Monteggia fractures based upon the ulnar fracture pattern. This strategy recommends surgical stabilization for all complete ulna fractures. The purpose of this study was to evaluate whether an initial nonoperative approach to pediatric Monteggia fractures resulted in poorer outcomes and a higher rate of complications. This institutional review board approved retrospective study evaluated all Monteggia fractures presenting to a level 1 pediatric trauma center between 2008 and 2014. Chart and radiographic reviews were performed on 94 patients who met inclusion criteria. The mean age was 5.5 years (range, 1 to 13 y). The mean clinical follow-up was 18 weeks. Major complications were defined as those requiring an unplanned second procedure (other than implant removal) or that may result in long-term disability (residual radial head subluxation/dislocation). Univariate (P<0.05) and Multivariate Classification and Regression Tree (CART) (P<0.05) analyses were used to identify variables associated with the need for surgical stabilization. At final follow-up, there were no cases of residual radiocapitellar joint subluxation or dislocation and all fractures had healed. The majority (83%) of patients were successfully managed with a cast. Univariate analysis found Bado type and maximum ulna angulation as significant predictors (P<0.05), whereas the CART analysis found ulna angulation >36.5 degrees as the only primary predictor of requiring surgical stabilization. Overall, good outcomes were achieved in all patients with few major complications. Although treatment algorithms are intended to minimize complications and maximize good outcomes, we believe that an unintentional consequence of the recently proposed pediatric Monteggia fracture treatment guideline may be the overtreatment of these injuries. In our cohort, the majority of patients were able to avoid the operating room and surgical implants without compromising outcomes or complications. This more conservative approach, however, requires close monitoring of patients in the first 3 weeks during which most reductions were lost. Level IV-therapeutic studies, case series.

Association of genetic variations in the serotonin and dopamine systems with aggressive behavior in the Chinese adolescent population: Single- and multiple-risk genetic variants

BackgroundGenetic predisposition is an important factor leading to aggressive behavior. However, the relationship between genetic polymorphisms and aggressive behavior has not been elucidated. MethodsWe identified candidate genes located in the dopaminergic and serotonin system (DRD3, DRD4, and FEV) that had been previously reported to be associated with aggressive behavior. We investigated 14 tag single-nucleotide polymorphisms (SNPs) using a multi-analytic strategy combining logistic regression (LR) and classification and regression tree (CART) to explore higher-order interactions between these SNPs and aggressive behavior in 318 patients and 558 controls. ResultsBoth LR and CART analyses suggested that the rs16859448 polymorphism is the strongest individual factor associated with aggressive behavior risk. In CART analysis, individuals carrying the combined genotypes of rs16859448TT/GT-rs11246228CT/TT-rs3773679TT had the highest risk, while rs16859448GG-rs2134655CT had the lowest risk (OR = 5.25, 95% CI: 2.53–10.86). ConclusionThis study adds to the growing evidence on the association of single- and multiple-risk variants in DRD3, DRD4, and FEV with aggressive behavior in Chinese adolescents. However, the aggressive behavior scale used to diagnose aggression in this study did not account for comorbid conditions; therefore, further studies are needed to confirm our observations.

Epidemiologic characteristics and outcomes of open globe injury in Shanghai.

To investigate the epidemiologic characteristics and outcomes of open globe injury in Shanghai. A retrospective study was conducted for 148 unilateral open globe injury cases presenting to a tertiary referral hospital of Shanghai. Electronic medical records were reviewed and phone surveys were conducted to collect and analyze 1) background of patient; 2) setting of injury and clinical signs at presentation; 3) treatment procedure and outcome; 4) quality of life after injury. There were more male patients (77.03%) than females (22.97%), more temporary habitants (79.05%) than residents (20.95%). The subjects in this study presented a significantly lower constitutional status of education than that of the whole Shanghai population (P<0.001). Occupational injury was the first cause of injuries (39.86%), followed by home accident (20.27%), road accident (16.89%), violent behavior (16.89%) and outdoor injury (6.08%). The 143 subjects (96.62%) were not wearing spectacles at the time of injury. Of all patients, 77 subjects (52.03%) had the outcome of no vision (including enucleation). The classification and regression tree (CART) prognosis presents 59.58% sensitivity to predict visual survival correctly and 80.19% specificity to predict no vision correctly. The patients whose injured eye had no vision reported more reduction of life quality. We found that male subject, temporary habitants, low educational status and no eyewear are risk factors of open globe injury in Shanghai. Occupational injury is the leading cause. CART analysis presents a certain agreement to the actual visual outcome. The injury imposes negative impact on quality of life especially in no vision cases. The education of eye protection may help to avoid the injury.

Development of an algorithm using clinical tests to avoid post-operative residual neuromuscular block

BackgroundQuantitative neuromuscular monitoring is the gold standard to detect postoperative residual curarization (PORC). Many anesthesiologists, however, use insensitive, qualitative neuromuscular monitoring or unreliable, clinical tests. Goal of this multicentre, prospective, double-blinded, assessor controlled study was to develop an algorithm of muscle function tests to identify PORC.MethodsAfter extubation a blinded anesthetist performed eight clinical tests in 165 patients. Test results were correlated to calibrated electromyography train-of-four (TOF) ratio and to a postoperatively applied uncalibrated acceleromyography. A classification and regression tree (CART) was calculated developing the algorithm to identify PORC. This was validated against uncalibrated acceleromyography and tactile judgement of TOF fading in separate 100 patients.ResultsAfter eliminating three tests with poor correlation, a model with four tests (r = 0.844) and uncalibrated acceleromyography (r = 0.873) were correlated to electromyographical TOF-values without losing quality of prediction. CART analysis showed that three consecutively performed tests (arm lift, head lift and swallowing or eye opening) can predict electromyographical TOF. Prediction coefficients reveal an advantage of the uncalibrated acceleromyography in terms of specificity to identify the EMG measured train-of-four ratio < 0.7 (100% vs. 42.9%) and <0.9 (89.7% vs. 34.5%) compared to the algorithm. However, due to the high sensitivity of the algorithm (100% vs. 94.4%), the risk to overlook an awake patient with a train-of-four ratio < 0.7 was minimal. Tactile judgement of TOF fading showed poorest sensitivity and specifity at train of four ratio < 0.9 (33.7%, 0%) and <0.7 (18.8%, 16.7%).ConclusionsResidual neuromuscular blockade can be detected by uncalibrated acceleromyography and if not available by a pathway of four clinical muscle function tests in awake patients. The algorithm has a discriminative power comparable to uncalibrated AMG within TOF-values >0.7 and <0.3.Trial registrationClinical Trials.gov (principal investigator’s name: CU, and identifier: NCT03219138) on July 8, 2017.

Development of Korean Liver Allocation System Based on Model for End Stage Liver Disease Score

Background Current Korean liver allocation system is based on Child-Turcotte-Pugh score. As the number of deceased donor has been increased, there is a pressing need to develop equitable and objective allocation system based on model for end-stage liver disease score (MELD score). Methods We conducted a nationwide, multicenter, retrospective cohort study of candidates who were listed for liver transplantation from January 2009 to December 2011 at 11 transplant centers. Classification and regression tree (CART) analysis was used to stratify MELD score ranges according to the waitlist removal. Results Among 2702 patients listed for liver transplantation, 2453 chronic liver disease patients were eligible. Of the patients, 1068 patients (47.6%) were with hepatocellular carcinoma (HCC). The CART analysis indicated several MELD scores for the significant determinant points for the waitlist removal. The 90-day waitlist survival rates were 16.2% for MELD 31-40, 64.1% for MELD 21-30, and 95.9% for MELD ≤20, respectively (P<0.001). We further partitioned severely ill patients (MELD 31-40, n=240) into subgroups according to 14-day waitlist removal. The 14-day waitlist survival rates were 61.5% for patients with MELD 31-37 and 41.6% for those with MELD 38-40 (P=0.003). In patients with MELD >20, the presence of HCC did not affect the waitlist removal. We determined MELD ranges for HCC patients with waitlist removal rates equivalent to those of non-HCC patients (HCC MELD 14-20 vs. non-HCC MELD 21-25; HCC MELD <14 vs. non-HCC MELD 14-17). Conclusion MELD score successfully discriminates Korean waitlist patients according to short-term mortality. Considering lack of donor organs and geographic disparities in Korea, we suggest staged broader sharing model based on survival CART analysis. In HCC prevalent area, prioritization within HCC candidates on the basis of native MELD score is required.

The Role of Occupational Voice Demand and Patient-Rated Impairment in Predicting Voice Therapy Adherence

Examine the relationship among the severity of patient-perceived voice impairment, perceptual dysphonia severity, occupational voice demand, and voice therapy adherence. Identify clinical predictors of increased risk for therapy nonadherence. A retrospective cohort study of patients presenting with a chief complaint of persistent dysphonia at an interdisciplinary voice center was done. The Voice Handicap Index-10 (VHI-10) and the Voice-Related Quality of Life (V-RQOL) survey scores, clinician rating of dysphonia severity using the Grade score from the Grade, Roughness Breathiness, Asthenia, and Strain scale, occupational voice demand, and patient demographics were tested for associations with therapy adherence, defined as completion of the treatment plan. Classification and Regression Tree (CART) analysis was performed to establish thresholds for nonadherence risk. Of 166 patients evaluated, 111 were recommended for voice therapy. The therapy nonadherence rate was 56%. Occupational voice demand category, VHI-10, and V-RQOL scores were the only factors significantly correlated with therapy adherence (P < 0.0001, P = 0.018, and P = 0.008, respectively). CART analysis found that patients with low or no occupational voice demand are significantly more likely to be nonadherent with therapy than those with high occupational voice demand (P < 0.001). Furthermore, a VHI-10 score of ≤29 or a V-RQOL score of >40 is a significant cutoff point for predicting therapy nonadherence (P < 0.011 and P < 0.004, respectively). Occupational voice demand and patient perception of impairment are significantly and independently correlated with therapy adherence. A VHI-10 score of ≤9 or a V-RQOL score of >40 is a significant cutoff point for predicting nonadherence risk.

Structuring subjectivity in Asian Englishes

This study investigates the usage patterns of four near-synonymous mental predicates (believe, guess, suppose and think) across three Asian ESL (English as a Second Language) varieties as well as British and American Englishes. Using two multivariate techniques, multiple correspondence analysis and classification and regression tree analysis, the study shows the benefits of exploring cross-varietal variation through the lens of lexicalization patterns. The study also demonstrates that to make sense of semantic patterns it is crucial to account for extra-linguistic factors such as genre, as different ESL writers structure the meaning of believe, guess, suppose and think differently depending on their type of writing. Ultimately, in the broader context of the emancipation of ESL varieties, the results raise important questions about the developmental process of Asian Englishes and the place that semantic structure holds in this endeavor.

Understanding hydrologic variability across Europe through catchment classification

Abstract. This study contributes to better understanding the physical controls on spatial patterns of pan-European flow signatures – taking advantage of large open datasets for catchment classification and comparative hydrology. Similarities in 16 flow signatures and 35 catchment descriptors were explored for 35 215 catchments and 1366 river gauges across Europe. Correlation analyses and stepwise regressions were used to identify the best explanatory variables for each signature. Catchments were clustered and analyzed for similarities in flow signature values, physiography and the combination of the two. We found the following. (i) A 15 to 33 % (depending on the classification used) improvement in regression model skills when combined with catchment classification versus simply using all catchments at once. (ii) Twelve out of 16 flow signatures were mainly controlled by climatic characteristics, especially those related to average and high flows. For the baseflow index, geology was more important and topography was the main control for the flashiness of flow. For most of the flow signatures, the second most important descriptor is generally land cover (mean flow, high flows, runoff coefficient, ET, variability of reversals). (iii) Using a classification and regression tree (CART), we further show that Europe can be divided into 10 classes with both similar flow signatures and physiography. The most dominant separation found was between energy-limited and moisture-limited catchments. The CART analyses also separated different explanatory variables for the same class of catchments. For example, the damped peak response for one class was explained by the presence of large water bodies for some catchments, while large flatland areas explained it for other catchments in the same class. In conclusion, we find that this type of comparative hydrology is a helpful tool for understanding hydrological variability, but is constrained by unknown human impacts on the water cycle and by relatively crude explanatory variables.

Use of Wind Fetch and Shoreline Relief to Predict Nearshore Substrate Composition in a North Temperate Lake

AbstractSpawning habitat assessments often focus on substrate composition, but few studies have predicted shoal substrates by using environmental factors. We developed a model for predicting shoal substrates in Belle Lake, Minnesota, using wind fetch and shoreline relief characteristics. Percent composition of four substrate classes (silt, sand, gravel, and rock), water depth estimated at 1 m from shore (shoal slope), effective wind fetch measured using a GIS model, and riparian bank height derived from LIDAR imaging were determined at 50 transects. Classification and regression tree (CART) analysis grouped substrates into categories, and general additive modeling described the effects of three predictor variables on the percent composition of substrate classes. The CART analysis correctly grouped 39 of 50 transects into four categories, and misclassifications primarily resulted from the movement of sand. Effective fetch most influenced silt (low fetch) and rock (high fetch) substrate classes, shoal slope was predictive of rock, and riparian height was useful in distinguishing sand from gravel. These results demonstrate the utility of a single empirical model for determining shoal substrate composition. Fisheries managers can use this technique to determine potential fish spawning locations and identify potential areas for habitat restoration or protection projects.Received December 5, 2016; accepted May 5, 2017 Published online July 19, 2017

Predictors of success of external cephalic version and cephalic presentation at birth among 1253 women with non-cephalic presentation using logistic regression and classification tree analyses.

Among women with a fetus with a non-cephalic presentation, external cephalic version (ECV) has been shown to reduce the rate of breech presentation at birth and cesarean birth. Compared with ECV at term, beginning ECV prior to 37weeks' gestation decreases the number of infants in a non-cephalic presentation at birth. The purpose of this secondary analysis was to investigate factors associated with a successful ECV procedure and to present this in a clinically useful format. Data were collected as part of the Early ECV Pilot and Early ECV2 Trials, which randomized 1776 women with a fetus in breech presentation to either early ECV (34-36weeks' gestation) or delayed ECV (at or after 37weeks). The outcome of interest was successful ECV, defined as the fetus being in a cephalic presentation immediately following the procedure, as well as at the time of birth. The importance of several factors in predicting successful ECV was investigated using two statistical methods: logistic regression and classification and regression tree (CART) analyses. Among nulliparas, non-engagement of the presenting part and an easily palpable fetal head were independently associated with success. Among multiparas, non-engagement of the presenting part, gestation less than 37weeks and an easily palpable fetal head were found to be independent predictors of success. These findings were consistent with results of the CART analyses. Regardless of parity, descent of the presenting part was the most discriminating factor in predicting successful ECV and cephalic presentation at birth.

Predicting Residual Disease in Incidental Gallbladder Cancer: Risk Stratification for Modified Treatment Strategies

IntroductionRe-operation is advised for patients with T1b or greater incidental gallbladder cancer (GBCA). The presence of residual disease (RD) impacts resectability, chemotherapy, and survival. This study created a preoperative model to predict RD at re-operation. MethodsPatients with re-operation for incidental GBCA from 1992–2015 were included. The relationship between pathology data from initial cholecystectomy and RD at re-operation was assessed with logistic regression and classification and regression tree (CART) analysis. ResultsTwo hundred fifty-four patients were included and 188 underwent definitive re-resection (74.0%). Distant RD was identified in 69 (27.2%) patients and locoregional only RD in 82 (32.3%). On multivariate analysis, T3 (OR 22.7, 95% CI 5.5–94.4) and poorly differentiated tumors (OR 4.3, 95% CI 1.4–13.3) were associated with RD (p < 0.001–0.012). AUC of multivariate model was 0.78 (95% CI 0.72–0.83). CART analysis split patients into groups based on percentage with RD: 87% RD with T3, 67% RD with T1b/T2 and poorly differentiated, and 35% RD with T1b/T2 and well/moderate differentiated tumors. ConclusionBased on T stage and grade from cholecystectomy, this study developed a model for predicting RD at re-operation in incidental GBCA. This model delineates patient groups with variable percentages of RD and could be used to stratify high-risk patients for prospective trials.