Predictive Logistic Regression Model Research Articles

Mortality trends and risk factors in advanced stage-2 Human African Trypanosomiasis: A critical appraisal of 23 years of experience in the Democratic Republic of Congo.

We conducted a retrospective study on mortality trends and risk factors in 781 naïve cases of advanced stage-2 sleeping sickness admitted between 1989 and 2012 at the National Reference Center for Human African Trypanosomiasis (HAT), Department of Neurology, Kinshasa University, Democratic Republic of Congo (DRC). Death was the outcome variable whereas age, gender, duration of disease, location of trypanosomes in body fluids, cytorachy, protidorachy, clinical status (assessed on a syndromic and functional basis) on admission, and treatment regimen were predictors in logistic regression models run at the 0.05 significance level. Death proportions were 17.2% in the standard melarsoprol schedule (3-series of intravenous melarsoprol on 3 successive days at 3.6 mg/kg/d, with a one-week interval between the series, ARS 9); 12.1% in the short schedule melarsoprol (10 consecutive days of intravenous melarsoprol at 2.2 mg/kg/d, ARS 10), 5.4% in the first-line eflornithine (14 days of eflornithine at 400 mg/kg/d in 4 infusions a day DFMO B), 9.1% in the NECT treatment regimen (eflornithine for 7 days at 400, mg/kg/d in 2 infusions a day combined with oral nifurtimox for 10 days at 15 mg/kg/d in 3 doses a day); and high (36%) in the group with select severely affected patients given eflornithine because of their clinical status on admission, at the time when this expensive drug was kept for treatment of relapses (14 days at 400 mg/kg/d in 4 infusions a day, DFMO A). After adjusting for treatment, death odds ratios were as follows: 10.40 [(95% CI: 6.55–16.51); p = .000] for clinical dysfunction (severely impaired clinical status) on admission, 2.14 [(95% CI: 1.35–3.39); p = .001] for high protidorachy, 1.99 [(95% CI: 1.18–3.37); p = .010] for the presence of parasites in the CSF and 1.70 [(95% CI: 1.03–2.81); p = .038] for high cytorachy. A multivariable analysis within treatment groups retained clinical status on admission (in ARS 9, ARS 10 and DFMO B groups) and high protidorachy (in ARS 10 and DFMO B groups) as significant predictors of death. The algorithm for initial clinical status assessment used in the present study may serve as the basis for further development of standardized assessment tools relevant to the clinical management of HAT and information exchange in epidemiological reports.

Modeling Deforestation and Green Houses Gas Emissions in Morowali Utara District, Central Sulawesi Province, Indonesia

Land use cover conversion from forest to another land was constributing for total green houses gas (GHS) emission, especially from forestry sector. The Central Sulawesi Province has forest area about 4.258 million ha. The deforestation in Central Sulawesi Province for period 2000-2009 reached 432,111.50 Ha (10.15 %). Morowali Utara District is one of district in Central Sulawesi Province, which contribution 15,20 % total deforestation in Central Sulawesi province. This study aims to calculate the rate of deforestation and simulation carbon emission in Morowali Utara District by using REDD+ model. The main data used in this study are land use or cover map (2003, 2009 and 2015), driver variable (distance from : forest edge, roads, streams, and slope) as the explanatory variables of deforestation. The results show that the deforestation on the period 2003-2015 reached 62,017.47 (9,19 %) or the rate of deforestation 5,168.12 ha year−1. Logistic regression model for prediction of deforestation is Y1 = −2.2649 + 0.001126 X1 − 0.000152 X2 − 0.000225 X3 + 0.010704 X4 (ROC = 0,7408), meanwhile the amount of carbon that the REDD project will protect given a departure from the business as usual deforestation scenario up to 18.868 million tCO2e.

Can Predictive Modeling Identify Head and Neck Oncology Patients at Risk for Readmission?

Objective Unplanned readmission within 30 days is a contributor to health care costs in the United States. The use of predictive modeling during hospitalization to identify patients at risk for readmission offers a novel approach to quality improvement and cost reduction. Study Design Two-phase study including retrospective analysis of prospectively collected data followed by prospective longitudinal study. Setting Tertiary academic medical center. Subjects and Methods Prospectively collected data for patients undergoing surgical treatment for head and neck cancer from January 2013 to January 2015 were used to build predictive models for readmission within 30 days of discharge using logistic regression, classification and regression tree (CART) analysis, and random forests. One model (logistic regression) was then placed prospectively into the discharge workflow from March 2016 to May 2016 to determine the model's ability to predict which patients would be readmitted within 30 days. Results In total, 174 admissions had descriptive data. Thirty-two were excluded due to incomplete data. Logistic regression, CART, and random forest predictive models were constructed using the remaining 142 admissions. When applied to 106 consecutive prospective head and neck oncology patients at the time of discharge, the logistic regression model predicted readmissions with a specificity of 94%, a sensitivity of 47%, a negative predictive value of 90%, and a positive predictive value of 62% (odds ratio, 14.9; 95% confidence interval, 4.02-55.45). Conclusion Prospectively collected head and neck cancer databases can be used to develop predictive models that can accurately predict which patients will be readmitted. This offers valuable support for quality improvement initiatives and readmission-related cost reduction in head and neck cancer care.

POTENSI MODIFIKASI MODEL PIERS UNTUK PREDIKSI LUARAN IBU DENGAN PREEKLAMSIA PADA RUMAH SAKIT TIPE B DI INDONESIA: PENELITIAN RETROSPEKTIF DI RSUD ANSARI SALEH BANJARMASIN, KALIMANTAN SELATAN

ABSTRACTBackground: Preeclampsia is the most common cause (32%) of maternal deaths in Indonesia. Model Piers(Pre-eclampsia Integrated Estimate Risk) can predict the outcomes / complications of pre-eclampsia andassociated with a reduction in maternal morbidity. This study aims to assess the potential modification of themodel to predict the outcome Piers mothers with preeclampsia in Type B hospitals in Indonesia. Methods:This research is a prognostic with the analytical method and quantitative approach retrospectively. This studyused the capture technique stratified random sampling to obtain the medical records of 160 people withpreeclampsia and 160 controls in hospitals Banjarmasin Ansari Saleh 2014. Identified availability ofvariables in the model along with the outer piers preeclampsia. Performed univariate, bivariate andmultivariate corresponding availability of variables the model Piers. Result: eight of the 48 variables in themodel Piers consistently found in all samples, while output available were complication preeclampsiaeclampsia and/or HELLP syndrome. The gestational age was significantly influent eclampsia for the subjectin this study (P <0.05), HELLP syndrome (P <0.05), and a combination of both (P <0.01). Predictive value ofa logistic regression model with the variables that the outer form of the syndrome HELLP or in combinationwith eclampsia is 0.811 (AUC ROC; CI 95% from 0.636 to 0.986) and 0.767 (AUC ROC; CI 95% 0.666 to0.868), while the prediction of eclampsia was not significant (P> 0.05). Conclusion: Modifications Piersmodels to predict potential outcomes of women with preeclampsia in RS type B in Indonesia if it is supportedby policies and standards that consider this model and supported a larger number of samples.

Understanding Prediabetes in a Medicare Advantage Population Using Data Adaptive Techniques.

The objective was to identify individuals with undiagnosed prediabetes from administrative data using adaptive techniques. The data source was a national Medicare Advantage Prescription Drug (MAPD) plan administrative data set. A retrospective, cross-sectional study developed and evaluated data adaptive logistic regression, decision tree, neural network, and ensemble predictive models for metabolic syndrome and prediabetes using 3 mutually exclusive cohorts (N = 279,903). The misclassification rate (MCR), average squared error (ASE), c-statistics, sensitivity (SN), and false positive (FP) rates were compared to select the final predictive models. MAPD individuals with continuous enrollment from 2013 to 2014 were included. Metabolic syndrome and prediabetes were defined using clinical guidelines, diagnosis, and laboratory data. A total of 512 variables identified through subject matter expertise in addition to utilizing all data available were evaluated for the modeling. The ensemble model demonstrated better discrimination (c-statistics, MCR, and ASE of 0.83, 0.24, and 0.16, respectively), high SN, and low FP rate in predicting metabolic syndrome than the individual data adaptive modeling techniques. Logistic regression demonstrated better discrimination (c-statistics, MCR, and ASE of 0.67, 0.13, and 0.11 respectively), high SN, and low FP rate in predicting prediabetes than the other adaptive modeling techniques or ensemble methods. The scored data predicted prediabetes in 44% of the MAPD population, which is comparable to 2005-2006 National Health and Nutrition Examination Survey prediabetes rates of 41%. The logistic regression model demonstrated good performance in predicting undiagnosed prediabetes in MAPD individuals.

Study on learning effect prediction models based on principal component analysis in MOOCs

In recent years, the rise and development of massive open online courses (MOOCs) have promoted the boom of online education and also promoted the research on learning analysis and mining based on big data of education. However, while offering a large number of high quality courses, there is also a phenomenon that the overall learning effect is not ideal. How to make effective use of MOOCs for teaching activities poses urgent practical requirements for educators and researchers. MOOCs store massive learners’ learning behavior data, and mining these data is of great significance to learners’ learning effects. Due to the deviation caused by the correlation between many behavior indexes, the paper analyze the nine measurable performance indexes by principal component analysis (PCA), and get three principal components factors in order to reduce the computational dimension and increase the comprehensibility, and then obtain a model of learning effect prediction, namely the PCA prediction model (PCA_PM). Logistic regression algorithm has a good fitting effect on behavioral indexes. Therefore, the paper use the logistic regression method to construct another model of learning effect prediction, namely the principal component logistic regression prediction model (PCL_PM), and predictive effects of two models were compared and analyzed. The results show that it is an effective way to help learners to improve the passing rate of the course by establishing the predictive models. The prediction accuracy of both PCA_PM and PCL_PM is high, but overall, the prediction effect of PCA_PM is better.

Relationship between blood parameters and Clonorchis sinensis infection: A retrospective single center study

Our study aims to retrospectively investigate neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) and eosinophil-to-lymphocyte ratio (ELR) in patients infected with Clonorchis sinensis. This study analyzes a total of 151 patients with C. sinensis infections and 53 healthy control patients from our hospital. We found close relationships between the three candidate markers and the stages of C. sinensis infection-related biliary obstruction. The NLRs, PLRs and ELRs of patients with C. sinensis infections were significantly higher than those of healthy individuals; of those, ELRs showed the most superior diagnostic accuracy (sensitivity = 62.9%, specificity = 92.5%). Further, we constituted a logistic regression prediction model; applying two variables (age and NLR) with a sensitivity of 88.89% and a specificity of 83.78% in differentiating C. sinensis-related cholelithiasis from C. sinensis-untreated patients. Cancer antigen 19-9 (CA19-9) is a commonly used marker in the diagnosis of cholangiocarcinoma. Significant correlation was observed between NLR and CA19-9 in patients with C. sinensis–related cholangiocarcinoma (r = 0.590, P = 0.000). In the receiver operating characteristic analysis for separating C. sinensis–related cholelithiasis and cholangiocarcinoma, the cutoff value of PLR was 145.14 with a sensitivity of 65.62% and a specificity of 68.89%; the sensitivity of CA19-9 was 75.00% with a specificity of 77.78%. PLR showed acceptable efficiency to separate C. sinensis-related cholelithiasis from cholangiocarcinoma. In conclusion, all of the candidate markers (PLRs, NLRs and ELRs) may act as the valuable supplement in detecting C. sinensis infections and diseases.

Development of prognostic model for fistula maturation in patients with advanced renal failure

BackgroundThis study aimed to explore the role of patient’s characteristic and haematological factors as predictive on the maturation of arteriovenous fistulae in patients who underwent vascular access surgery at the Royal Infirmary of Edinburgh.MethodsRetrospective data from 300 patients who had undergone fistula creation between February 2007 and October 2010 was examined. A predictive logistic regression model was developed using the backward stepwise procedure. Model performance, discrimination and calibration, was assessed using the receiver operating characteristics (ROC) curve and Hosmer and Lemeshow goodness of fit test.ResultsThree variables were identified which independently influenced fistula maturation. Males were twice as likely to undergo fistula maturation, compared to that of females (odds ratio (OR) 0.514; 95% confidence interval (CI) 0.308–0.857), patients with no evidence of peripheral vascular disease (PVD) were three times more likely to mature their fistula (OR 3.140; 95% CI 1.596–6.177) and a pre-operative vein diameter > 2.5 mm resulted in a fivefold increase in fistula maturation compared to a vein size less than 2.5 mm (OR 4.532; 95% CI 2.063–9.958). The model for fistula maturation had fair discrimination as indicated by the area under the ROC curve (0.68; 95% CI 0.615–0.738) but good calibration indicated by Hosmer and Lemeshow test (p = 0.79).ConclusionGender, PVD and vein size are independent predictors of arteriovenous fistula maturation. The clinical utility of these risk equation in the maturation of arteriovenous fistulae requires further validation in the newly treated patients.

Assemblage-level analysis of sex-ratios in Coloradan bats in relation to climate variables: A model for future expectations

Bats act as important bioindicators of environmental changes, including climate disruptions. In the Rocky Mountain West, that includes western Colorado, the pace of climate warming outpaces other areas of North America. Using data on bats captured between 1996 and 2015 we found significant trends towards male-bias (R2 = 0.38) over the 20-year period for both adults and juveniles. Capture data for juveniles showed markedly less females captured than males after 2000, hitting all-time lows in 2000, 2002, and 2007 (R2 = 0.12). When juvenile captures were plotted against an index of annual monthly temperatures for the June, July, and August, divided by annual precipitation for the same months (T/P index) we found that sex ratios favored males at the extreme ends of the spectrum (cooler/wetter and hotter/drier), however, male-biased outcomes were much more prevalent in hotter/drier years. We found that nearly 1:1 sex ratios occurred within an optimal range of temperature and precipitation that peaked at 28.7° C with 7.5 cm precipitation. To explore the possible impacts of future climate conditions on sex ratios, we used juvenile capture data to build a predictive logistic regression model and projected it through year 2090 using four carbon emission scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5). All scenarios showed overall declining proportions of females in future years. Thus, mitigation of climate change on bat populations must take into account not only reproductive declines in some species, but also shifting sex ratio outcomes towards increasing male abundances in populations.

The Use of a Nomogram to Visually Interpret a Logistic Regression Prediction Model for Giant Cell Arteritis

ABSTRACTObjective: To illustrate the utility of a nomogram for the prediction of giant cell arteritis (GCA).Method: A nomogram was constructed from a multivariable logistic regression prediction model with 10 covariates: age, sex, clinical temporal artery abnormality, new-onset headache, jaw claudication, vision loss, diplopia, erythrocyte sedimentation rate, C-reactive protein, and platelet level.Results: The magnitude and location of the nomogram scale for each predictor variable graphically illustrates the net effect of each covariate and is especially useful for continuous variables such as age and bloodwork values.Conclusions: Nomograms allow integration and synthesis of the relative importance of clinical variables and provide a graphic representation of the odds ratios, p values, and confidence intervals of logistic regression prediction models. Although nomograms and prediction rules cannot substitute for clinical judgment, they help objectify and optimize the individualized risk assessments for patients with suspected GCA.

712: Beyond preterm birth history and cervical length: analysis of routinely collected clinical data improves prediction of preterm delivery

A history of preterm delivery and shortened cervical length are widely recognized risk factors for preterm birth (PTB) and have been used as inclusion criteria for trials of therapies to prevent PTB. We sought to determine if other readily accessible clinical or demographic variables could improve risk stratification or identification of pregnancies at increased risk for PTB. This is a secondary analysis of the MFMU-sponsored Singleton Preterm Prediction Study, a prospective observational study to identify risk factors for PTB conducted from 1992-1994. Participants were not selected based on risk factors for PTB, had extensive demographic and clinical data collected and, amongst other tests, underwent sonographic cervical length (CL) measurement around 24 weeks. We used cervical length measured between 23w0d and 25w0d, PTB history and other data readily accessible in a clinical setting (final predictors: age, race, reported family income, chronic hypertension history, prior full term delivery, prior miscarriage, vaginal bleeding prior to 25 weeks and pre-pregnancy BMI) to predict spontaneous PTB (<37 weeks). We compared 4 logistic regression models for prediction of PTB using the following predictors: 1) PTB history (PTB only); 2) Cervical length (CL only); 3) Listed clinical variables and PTB history (Clinical only); and 4) PTB history, CL and clinical variables (Clinical+CL). The area under the curve (AUC) for the receiver operating characteristic (ROC) curve was compared across the 4 models. In the study population of 2,796, there were 402 spontaneous preterm births (14.4%). The Clinical+CL model predicted PTB best, followed by the Clinical only model (Figure). The CL only and PTB only models performed worst, with no significant difference between their AUC. Amongst the 280 participants predicted to be in the top decile of risk for PTB by the Clinical+CL model (Probability of PTB >0.283), 63 (22.5%) had no history of PTB and 232 (82.9%) had a cervical length >20mm. Beyond PTB history and CL, clinical variables that are available at essentially no cost significantly improve risk-stratification for PTB. Clinical trials of PTB prevention strategies in this easily identified and potentially overlooked population are warranted.

An Integrated Framework With Feature Selection for Dropout Prediction in Massive Open Online Courses

Massive open online courses (MOOCs) have flourished in recent years, which is conducive to the redistribution of high-quality educational resources globally. However, the high dropout rate in the course of operation has seriously affected its development. Therefore, in order to improve the degree of completion, it is an effective way to study how to effectively predict the dropout in MOOCs and intervene in advance. Traditional methods rely on manually extracted features, which is difficult to guarantee the final prediction effect. In order to solve this problem, this paper proposes an integrated framework with feature selection (FSPred) to predict the dropout in MOOCs, which includes feature generation, feature selection, and dropout prediction. Specifically, FSPred applies a fine-grained feature generation method in days to generate features and then uses an ensemble feature selection method to select valid features and feed them into a logistic regression model for prediction. Extensive experiments on a public data set have shown that FSPred can achieve the comparable results with other dropout prediction methods in terms of precision, recall, F1 score, and AUC score. Finally, through the analysis of the features of the final selection, the suggestions for the construction of the MOOCs are put forward.

Prediction models to identify individuals at risk of metabolic syndrome who are unlikely to participate in a health intervention program

ObjectivesSince the launch of a nationwide general health check-up and instruction program in Japan in 2008, interest in strategies to improve implementation of the program based on predictive analytics has grown. We investigated the performance of prediction models developed to identify individuals classified as “requiring instruction” (high-risk) who were unlikely to participate in a health intervention program. MethodsData were obtained from one large health insurance union in Japan. The study population included individuals who underwent at least one general health check-up between 2008 and 2013 and were identified as “requiring instruction” in 2013. We developed three prediction models based on the gradient boosted trees (GBT), random forest (RF), and logistic regression (LR) algorithms using machine-learning techniques and compared the areas under the curve (AUC) of the developed models with those of two conventional methods The aim of the models was to identify at-risk individuals who were unlikely to participate in the instruction program in 2013 after being classified as requiring instruction at their general health check-up that year. ResultsAt first we performed the analysis using data without multiple imputation. The AUC values for the GBT, RF, and LR prediction models and conventional methods: 1, and 2 were 0.893 (95%CI: 0.882–0.905), 0.889 (95%CI: 0.877–0.901), 0.885 (95%CI: 0.872–0.897), 0.784 (95%CI: 0.767–0.800), and 0.757 (95%CI: 0.741–0.773), respectively. Subsequently, we performed the analysis using data after multiple imputation. The AUC values for the GBT, RF, and LR prediction models and conventional methods: 1, and 2 were 0.894 (95%CI: 0.882–0.906), 0.889 (95%CI: 0.887–0.901), 0.885 (95%CI: 0.872–0.898), 0.784 (95%CI: 0.767–0.800), and 0.757 (95%CI: 0.741–0.773), respectively. In both analyses, the GBT model showed the highest AUC among that of other models, and statistically significant difference were found in comparison with the LR model, conventional method 1, and conventional method 2. ConclusionThe prediction models using machine-learning techniques outperformed existing conventional methods: for predicting participation in the instruction program among participants identified as “requiring instruction” (high-risk).

Risk Modeling to Optimize Patient Selection for Management of the Descending Thoracic Aortic Aneurysm

A single-institutional study comparing early and long-term outcomes of thoracic endovascular aortic repair (TEVAR) and open surgical repair (OSR) was performed to determine the appropriate treatment option for descending thoracic aortic aneurysm (DTAA). Between 2005 and 2014, 438 DTAA patients were treated (TEVAR, 88; OSR, 350). Acute dissection and traumatic injury were excluded. Perioperative and follow-up data were reviewed. Stratified analyses were conducted to identify patients most likely to benefit from TEVAR. A propensity score for TEVAR was developed by logistic regression, and predictive logistic and Cox regression models for death were adjusted for propensity score. TEVAR patients were frequently older women with emergent status, chronic obstructive pulmonary disease, or coronary artery disease. TEVAR had similar immediate (0% vs 1%; p= 0.588) and delayed (5% vs 6%, p= 1.000) motor deficits and early mortality (6% vs 12%, p= 0.121) but lower dialysis (3% vs 18%, p < 0.001), respiratory failure (10% vs 34%, p < 0.001), and intensive care unit stay (2.0 vs 5.0 days, p < 0.001). Early mortality after TEVAR was lower in septuagenarians (3% vs 16%, p < 0.02), glomerular filtration rate of less than 60 mL/min (8% vs 32%, p < 0.049), chronic obstructive pulmonary disease (6% vs 21%, p < 0.02), defined as target population that had fourfold mortality reduction (p < 0.006) attributable to TEVAR. Propensity-adjusted predictors of early mortality predictors included OSR (odds ratio [OR], 4.3; p < 0.024), target population (OR, 7.7; p < 0.001), diabetes (OR, 3; p < 0.009), peripheral vascular disease (OR, 4.7; p < 0.001), and emergent status (OR, 4.6; p < 0.001). Propensity-adjusted determinants of survival were age, glomerular filtration rate of less than 60 mL/min, peripheral vascular disease, chronic obstructive pulmonary disease, and emergent status. In older patients with significant comorbidities, TEVAR demonstrated superior results compared with OSR and may be preferable in this target population.

Prediction of fetal growth restriction using estimated fetal weight vs a combined screening model in the third trimester

To compare the performance of third-trimester screening, based on estimated fetal weight centile (EFWc) vs a combined model including maternal baseline characteristics, fetoplacental ultrasound and maternal biochemical markers, for the prediction of small-for-gestational-age (SGA) neonates and late-onset fetal growth restriction (FGR). This was a nested case-control study within a prospective cohort of 1590 singleton gestations undergoing third-trimester (32 + 0 to 36 + 6 weeks' gestation) evaluation. Maternal baseline characteristics, mean arterial pressure, fetoplacental ultrasound and circulating biochemical markers (placental growth factor (PlGF), lipocalin-2, unconjugated estriol and inhibin A) were assessed in all women who subsequently delivered a SGA neonate (n = 175), defined as birth weight < 10th centile according to customized standards, and in a control group (n = 875). Among SGA cases, those with birth weight < 3rd centile and/or abnormal uterine artery pulsatility index (UtA-PI) and/or abnormal cerebroplacental ratio (CPR) were classified as FGR. Logistic regression predictive models were developed for SGA and FGR, and their performance was compared with that obtained using EFWc alone. In SGA cases, EFWc, CPR Z-score and maternal serum concentrations of unconjugated estriol and PlGF were significantly lower, while mean UtA-PI Z-score and lipocalin-2 and inhibin A concentrations were significantly higher, compared with controls. Using EFWc alone, 52% (area under receiver-operating characteristics curve (AUC), 0.82 (95% CI, 0.77-0.85)) of SGA and 64% (AUC, 0.86 (95% CI, 0.81-0.91)) of FGR cases were predicted at a 10% false-positive rate. A combined screening model including a-priori risk (maternal characteristics), EFWc, UtA-PI, PlGF and estriol (with lipocalin-2 for SGA) achieved a detection rate of 61% (AUC, 0.86 (95% CI, 0.83-0.89)) for SGA cases and 77% (AUC, 0.92 (95% CI, 0.88-0.95)) for FGR. The combined model for the prediction of SGA and FGR performed significantly better than did using EFWc alone (P < 0.001 and P = 0.002, respectively). A multivariable integrative model of maternal characteristics, fetoplacental ultrasound and maternal biochemical markers modestly improved the detection of SGA and FGR cases at 32-36 weeks' gestation when compared with screening based on EFWc alone. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.

Early pregnancy immune biomarkers in peripheral blood may predict preeclampsia

We performed a prospective cohort study in 197 pregnant women. Peripheral blood was collected between 5 and 16 weeks of gestation. Intracellular cytokine analysis and immunophenotype were performed by flow-cytometry. Serum levels of cytokines and chemokines were analyzed by multiplex assay. 86 patients were eligible for the analysis and 10.5% (n=9) developed preeclampsia. Patients with preeclampsia had significantly higher percentage of CD3+CD4+TNFα+ T helper (Th) 1 cells (45.4±10.3 vs 37.1±8.5, P=0.032) and CD3+CD4+IL17+ Th 17 cells (2.4±1.3 vs 1.6±1.1, P=0.029) when compared to those of patients without preeclampsia. CD3+CD4+CD25+CD127dim/− T regulatory cells (Treg) cells (5.7±1.2% vs 7.0±1.6%, P=0.015) were significantly lower in patients with preeclampsia when compared to those without preeclampsia. Patients with preeclampsia had significantly higher TNFα/IL-10 cell ratio (43.8±10.3 vs 34.3±7.9, P=0.005) and Th17/Treg cell ratio (0.5±0.3 vs 0.2±0.2, P=0.011) when compared to those of patients without preeclampsia. IL-8 and Macrophage inflammatory protein (MIP)-1α serum levels were significantly higher in patients with preeclampsia when compared with patients without preeclampsia (Median=341.0 vs 87.6, U=152, P=0.020 and Median=35.7 vs 17.7, U=120, P=0.029 respectively). Serum MCP-1 levels were significantly lower in patients with preeclampsia when compared with patients without preeclampsia (Median=233.8 vs 390.9, U=183, P=0.021). The logistic regression predictive model combining TNFα/IL-10 ratios, IL-8 and MCP-1 serum levels had the best performance (AUC=0.886, 95%CI 0.8–0.9). We concluded that elevated Th1 and Th17 cell percentages, elevated TNFα/IL-10 and Th17/Treg cell ratios and decreased Treg cell percentages in early pregnancy are associated with preeclampsia.

Reconsidering lactate as a sepsis risk biomarker.

ObjectivesThere has been renewed interest in lactate as a risk biomarker in sepsis and septic shock. However, the ability of the odds ratio (OR) and change in the area under the receiver operator characteristic curve (AUC-ROC) to assess biomarker added-value has been questioned.Design, setting and participantsA sepsis cohort was identified from the ICU database of an Australian tertiary referral hospital using APACHE III diagnostic codes. Demographic information, APACHE III scores, 24-hour post-admission patient lactate levels, and hospital mortality were accessed.Measurements and main resultsHospital mortality was modelled using a base predictive logistic regression model and sequential addition of admission lactate, lactate clearance ([lactateadmission—lactatefinal]/lactateadmission), and area under the lactate-time curve (LTC). Added-value was assessed using lactate index OR; AUC-ROC difference (base-model versus lactate index addition); net (mortality) reclassification index (NRI; range -2 to +2); and net benefit (NB), the number of true positives per patient adjusted for the number of false positives. The data set comprised 717 patients with mean(SD) age and APACHE III score 61.1(16.5) years and 68.3(28.2) respectively; 59.2% were male. Admission lactate was 2.3(2.5) mmol/l; with lactate of ≥ 4 mmol/L (37% hospital mortality) in 17% and patients with lactate < 4 mmol/L having 18% hospital mortality. The admission base-model had an AUC-ROC = 0.81 with admission lactate OR = 1.127 (95%CI: 1.038, 1.224), AUC-ROC difference of 0.0032 (-0.0037, 0.01615; P = 0.61), and NRI 0.240(0.030, 0.464). The over-time model had an AUC-ROC = 0.86 with (i) clearance OR = 0.771, 95%CI: 0.578, 1.030; P = 0.08; AUC-ROC difference 0.001 (-0.003, 0.014; P = 0.78), and NRI 0.109(-0.193, 0.425) and (ii) LTC OR = 0.997, 95%CI: 0.989, 1.005, P = 0.49; AUC-ROC difference 0.004 (-0.002, 0.004; P = 0.34), and NRI 0.111(-0.222, 0.403). NB was not incremented by any lactate index.ConclusionsLactate added-value assessment is dependent upon the performance of the underlying predictive model and should incorporate risk reclassification and net benefit measures.

Impact of Obesity on Acute Kidney Injury Incidence Among Patients Treated with Piperacillin–Tazobactam and Vancomycin

BackgroundObesity is associated with worse patient outcomes in a variety of clinical scenarios. It is unclear from previous research if obesity is associated with increased acute kidney injury (AKI) among patients receiving concomitant piperacillin-tazobactam (TZP) and vancomycin (VAN).MethodsClinical and demographic data were collected from the University of Kentucky Center for Clinical and Translational Science Enterprise Data Trust. Patients who received TZP+VAN for at least 48 hours in combination were included. Patients with CKD, a baseline creatinine clearance (CrCl) < 30 mL/minute, CF, or missing height and weight information were excluded from analysis. AKI was defined using the Risk, Injury, Failure, Loss, End-stage (RIFLE) criteria. A weight cutoff point of 91 kg was determined by finding the most predictive bivariable logistic regression model with weights varying from 70 kg through 120 kg via minimization of the Akaike information criterion. Basic descriptive statistics were performed and bivariable and multivariable logistic regressions were performed.ResultsIn total, 8125 patients were included in the final analysis. A total of 2452 (30.2%) of patients weighed ≥91 kg. Patients weighing less 91kg were less likely to receive concomitant nephrotoxins and had higher baseline CrCl (97.3 [70.1–128.1] mL/minute vs. 91.8 [68.1–116.5] mL/minute, <0.00001). Baseline severity of illness was similar between groups; however, diabetes (38.9% vs. 20.8%, P < 0.00001), hypertension (63.5% vs. 46.7%, P < 0.00001), and heart failure (14.8% vs. 12.5%, P = 0.007) were more common among the 91kg+ patients. Median daily VAN doses were less in the sub-91kg patients (2000 [1250–2500] mg vs. 3000 [2000–3500] mg, P < 0.00001); however, weight-based doses were lower in the ≥91kg group (25.5 [16.3–31.5] mg/kg/day vs 27.9 [18.7–34.2] mg/kg/day, P < 0.00001). AKI was more common in the patients weighing ≥91kg (23.8% vs. 17.8%, P < 0.00001; adjusted odds ratio 1.46 [95% CI 1.28–1.66]).ConclusionObesity appears to increase the incidence of AKI among patients treated with TZP+VAN, independent of clinically important confounders, with an important breakpoint occurring at 91 kg.Disclosures R. Hall, Genentech: Scientific Advisor, Consulting fee Merck: Grant Investigator, Grant recipient

Predictors of Prolonged Hospitalizations in Pediatric Complicated Pneumonia

Pediatric community-acquired complicated pneumonia (PCACP) is characterized by a prolonged clinical course, but this may be highly variable. A multicenter observational study was conducted to develop and validate a clinical prediction tool for prolonged hospitalizations in PCACP. The derivation and validation cohorts consisted of 144 and 169 patients with PCACP, respectively, hospitalized between the years 1997 and 2017 in three tertiary care hospitals. Logistic regression analyses were used to identify parameters associated with a prolonged hospitalization and to develop and validate a prediction model for constructing a useful clinical tool. Higher levels of lactate dehydrogenase (LDH) (P< .026) and lower levels of glucose (P= .018) in pleural fluid were significantly associated with prolonged hospitalization. A predictive stepwise logistic regression model was developed and applied to the validation cohort. The area under the receiver operating characteristic curve (AUROC) constructed indicated that the model retained good predictive value (AUROC for the derivation vsvalidation data, [0.77 (95%CI, 0.66-0.87) vs0.82 (95%CI, 0.72-0.91)], respectively). From these data, a clinical tool was derived; the combination of pleural LDH >1,000 units/L and pleural glucose levels< 1mmol/L or pleural LDH levels > 2,000 units/L and pleural glucose levels< 2mmol/L or pleural LDH levels > 3,000 units/L and pleural glucose< 3mmol/L predict prolonged hospitalization with positive and negative predictive values of 78%(95%CI, 0.71-0.85) and 73%(95%CI, 0.59-0.85), respectively. In children, pleural fluid LDH and glucose levels are useful parameters for assessing the severity of PCACP. The model developed in this study accurately predicts patients who will have prolonged hospitalization.

372 Derivation of a Clinical Decision Instrument to Identify Adults in a Community Setting With Mild Traumatic Intracranial Hemorrhage at Low Risk for Requiring Critical Care Intervention

Adults presenting to the ED with mild traumatic intracranial hemorrhage (tICH) are commonly admitted to the intensive care unit (ICU), although critical care interventions are often unnecessary. In prior research, we found that a clinical decision instrument (CDI) derived at an academic Level I trauma center to identify adults at low risk for critical care interventions performed poorly in a community setting, likely because of substantial differences in patient case-mix. We sought to derive a more generalizable CDI among patients of a large community-based integrated health care delivery system. This retrospective cohort study included non-anticoagulated adults (≥18 years of age) with mild tICH, defined as GCS score ≥13, across 21 community EDs from 01/2012 to 12/2013. No study facility carried a Level I trauma center designation, and only one carried a Level II designation. The primary outcome of at least one critical care intervention within 48 hours of ED arrival included intubation, neurosurgical intervention, vasopressor or inotrope use, invasive monitoring, or cardiopulmonary resuscitation. Using logistic regression with single independent variables, we identified potential predictors of the outcome, developed a 6-variable predictive logistic regression model, and created a simplified CDI to identify low-risk patients based on predicted probabilities from the model. We examined the prevalence of patients designated low-risk by both CDIs and compared performance metrics. We calculated sensitivity and specificity with Clopper-Pearson, positive and negative predictive values with standard logit, and c-statistics with Wald confidence interval (CI) estimates. Our cohort included 929 patients with mean age of 73.3 (SD 17.0) years; 50% were female, and 82% were injured by a ground-level fall. Of these, 110 (11.8%) received at least one critical care intervention, 100 of whom received a neurosurgical operation, mannitol or hypertonic saline. Patients identified as low risk had none of the following: time from trauma to ED >7 days, ED admitting GCS <15, lowest ED systolic blood pressure >140 mmHg, CT evidence of skull fracture, mass effect, or midline shift >5 mm. The new CDI designated nearly three times as many patients as low-risk than the academic-based CDI with similar sensitivity and significantly higher specificity (see Table). We derived a CDI to identify patients with mild tICH at low risk for requiring ICU-level care in this community setting. The community-based instrument was equally sensitive as its academic counterpart, but outperformed it with a greater specificity and positive predictive value. This lower rate of false positives may increase its ultimate clinical utility. Continued model development may further improve specificity, and a prospective validation study will be needed prior to clinical implementation.