Multifactorial Predictive Model Research Articles

Development of a multifactorial prediction model for commute mode choice in 10 983 Finnish public sector employees: a cross-sectional study

ObjectiveThe objective of this study is to examine the feasibility of using survey data to identify factors that predict commute mode choice.DesignThe study design is cross-sectional.SettingSurvey data from the Finnish...

Combining 2.5D deep learning and conventional features in a joint model for the early detection of sICH expansion.

The study aims to investigate the potential of training efficient deep learning models by using 2.5D (2.5-Dimension) masks of sICH. Furthermore, it intends to evaluate and compare the predictive performance of a joint model incorporating four types of features with standalone 2.5D deep learning, radiomics, radiology, and clinical models for early expansion in sICH. A total of 254 sICH patients were enrolled retrospectively and divided into two groups according to whether the hematoma was enlarged or not. The 2.5D mask of sICH is constructed with the maximum axial, coronal and sagittal planes of the hematoma, which is used to train the deep learning model and extract deep learning features. Predictive models were built on clinic, radiology, radiomics and deep learning features separately and four type features jointly. The diagnostic performance of each model was measured using the receiver operating characteristic curve (AUC), Accuracy, Recall, F1 and decision curve analysis (DCA). The AUCs of the clinic model, radiology model, radiomics model, deep learning model, joint model, and nomogram model on the train set (training and Cross-validation) were 0.639, 0.682, 0.859, 0.807, 0.939, and 0.942, respectively, while the AUCs on the test set (external validation) were 0.680, 0.758, 0.802, 0.857, 0.929, and 0.926. Decision curve analysis showed that the joint model was superior to the other models and demonstrated good consistency between the predicted probability of early hematoma expansion and the actual occurrence probability. Our study demonstrates that the joint model is a more efficient and robust prediction model, as verified by multicenter data. This finding highlights the potential clinical utility of a multifactorial prediction model that integrates various data sources for prognostication in patients with intracerebral hemorrhage. The Critical Relevance Statement: Combining 2.5D deep learning features with clinic features, radiology markers, and radiomics signatures to establish a joint model enabling physicians to conduct better-individualized assessments the risk of early expansion of sICH.

Impact of Consumption of Specific Food Groups on Metabolic and Cardiovascular Disorders among Nurses: Framingham's Multifactorial Predictive Model.

Objective: This study aimed to analyze the relationship between the consumption of selected food products and the risk of prevalence of selected metabolic and cardiovascular disorders among nurses. Methodology: This cross-sectional study was conducted among 405 nurses. To achieve the study objective, body composition analysis (Tanita MC-980), blood pressure measurement (Welch Allyn 4200B), anthropometric measurements, lipid profile, fasting blood glucose (CardioChek PA), and surveys regarding the consumption of specific food groups were conducted. Results: More than half of the respondents were overweight or/and obese, and almost 40% had elevated blood pressure levels. The results obtained from logistic regression models indicated that the consumption of specific food product groups may predispose to/increase the risk of hypertension, abdominal obesity, overweight, obesity, body fat accumulation, and the risk of cardiovascular events. Conclusions: These findings highlight the importance of targeted nutritional strategies to enhance the health and professional efficacy of nursing staff, paving the way for improved healthcare practices.

Risk prediction for severe COVID-19 progressing to critical illness and death in the ICU and efficacy analysis of using traditional Chinese medicine.

To reveal the key factors influencing the progression of severe COVID-19 to critical illness and death in the intensive care unit (ICU) and to accurately predict the risk, as well as to validate the efficacy of treatment using traditional Chinese medicine (TCM), thus providing valuable recommendations for the clinical management of patients. A total of 189 patients with COVID-19 in 25 ICUs in Chongqing, China, were enrolled, and 16 eventually died. Statistical models shown that factors influencing the progression of COVID-19 to critical illness include the severity of illness at diagnosis, the mode of respiratory support, and the use of TCM. Risk factors for death include a history of metabolic disease, the use of antiviral drugs and TCM, and invasive endotracheal intubation. The area under curve of the noncollinearity model predicted the risk of progression to critical illness and the risk of death reached 0.847 and 0.876, respectively. The use of TCM is an independent protective factor for the prevention of the progression of severe COVID-19, while uncorrectable hypoxemia and invasive respiratory support are independent risk factors, and antiviral drugs can help reduce mortality. The multifactorial prediction model can assess the risk of critical illness and death in ICU COVID-19 patients, and inform clinicians in choosing the treatment options and medications.

Risk Factors Analysis and Prediction of Rotator Cuff Tears: A Retrospective Study.

Rotator cuff tears (RCTs) are a major cause of shoulder pain and disability, affecting millions worldwide. Understanding the risk factors and developing reliable predictive measures for RCTs is essential for early diagnosis, targeted prevention, and effective treatment of this patient population. This study seeks to enhance our understanding by analyzing the acromiohumeral distance (AHD) and Constant-Murley Score (CMS) in patients with and without RCTs, thereby aiding the development of a predictive model aimed at improving clinical outcomes and prevention strategies in rotator cuff pathology. This retrospective analysis involved 201 patients with shoulder pain, categorized into RCT (n = 72) and no RCTs (N-RCTs, n = 129) groups based on Magnetic Resonance Imaging (MRI) findings. We compared demographics, AHD, CMS, and rotator cuff status between groups and utilized logistic regression for identifying RCT predictors, leading to the development of a multifactorial predictive model. The mean AHD was 6.60 ± 1.12 mm. The RCT group showed a marginally higher AHD than the N-RCT group (p = 0.669). CMS scores were significantly lower in the RCT group (p < 0.001). Dominant side involvement (Odds Ratio (OR) 2.244), type III acromion (OR 6.106), and lower CMS (OR 0.938) significantly correlated with RCTs. The predictive model demonstrated an area under the curve (AUC) of 0.701 for RCT diagnosis. Reduced CMS, dominance of the affected side, and type III acromion emerged as key risk factors for RCTs. Our predictive model, incorporating these factors, holds promise for RCT diagnosis, with future studies needed for further validation.

The Bengaluru Score - a Prognostic Predictive Score for Clinical Efficacy Outcomes Based on Discovery Cohort of Patients Treated with Varnimcabtagene Autoleucel (IMN-003A) in the Phase-2 (IMAGINE) Study

Background: Recurrence of disease is seen in some patients post CD19 targeted CAR-T cell therapy. There is a need to develop CAR-specific efficacy risk-classification systems as established parameters of disease relapse are not predictive. In the Phase-2 IMAGINE study, multiple variables that may predict clinical outcomes were studied and the relevant clinical, laboratory and translational information were integrated. This abstract reviews the development of the Bengaluru Score as a predictive prognostic score for clinical efficacy outcomes based on discovery cohort of patients treated with varnimcabtagene autoleucel (IMN-003A). Methods: Multiple characteristics for disease progression were reviewed (n=36) and risk stratified as: 1) Baseline parameters (age, lines of treatment, bridging therapy, cytogenetics, blasts, total metabolic tumour volume (TMTV), sum of perpendicular diameter (SPD), refractory, relapse within 6m, laboratory parameters - haemoglobin, neutrophil count, lymphocyte count, platelet count, CRP, LDH, ferritin, baseline interleukin-6 (IL-6), baseline total T cells, CD4, and CD8 subsets in peripheral blood); 2) Apheresis (total T cells, naïve / central memory in CD4 and CD8 subsets); 3) Final Product characteristics [total CAR+ cells, naïve / central memory in CD4 and CD8 subsets, CD8 (naïve &amp; central memory subsets / tumour burden (TMTV and SPD), product doubling time] and 4) Post Infusion (CAR peak &amp; persistence, B cell aplasia, hypogammaglobulinemia, infections and intravenous immunoglobulin (IVIg) replacement). These parameters were analyzed using chi-square and appropriate multivariant odds ratio (OR) to examine the association with disease progression. Independent risk factors were identified and weighted score allocated for OR ≥10 to obtain a multifactorial prediction model - The Bengaluru Score. The specificity and sensitivity of this model was then studied in the discovery cohort. Results: In the IMAGINE study, with ORR 80.9% at primary endpoint (n=17/21) and median PFS not reached (range 12 to NR), relapse was seen in 37.5% (n=9/24) pts with median PFS of 178 days (range 28 to 319). Multivariant analysis of these parameters showed some statistical significance for following variables: Baseline Platelet count &amp;lt;200 x 10 9/L (overall); Final Product doubling time ≥1 day (B-ALL); Baseline SPD ≥2000 mm 2, Hb &amp;lt;12 g/dL, Platelet count &amp;lt;200 x 10 9/L and post infusion IVIg replacement (B-NHL), Of these variables, statistically significant association was seen with Final Product doubling time ≥1 day (B-ALL) and Baseline SPD ≥2000 mm 2 with Hb &amp;lt;12 g/dL and Platelet count &amp;lt;200 x 10 9/L (B-NHL). The impact of IVIg replacement on efficacy needs further study. A multiple factor-based prognostic score system - The Bengaluru Score was developed classifying patients into B-ALL low-risk (1 point) and high-risk (3 points) and for B-NHL low risk (&amp;lt;5 points) and high-risk (≥5 points), Table 1. The Bengaluru Score when applied to discovery cohort in the IMAGINE study was able to risk stratify patients for disease progression with Sensitivity, Specificity, Positive Predictive Value and Accuracy of 100.00%, 66.67%, 60.00% and 77.78% respectively (for B-ALL) and 100.00%, 100.00%, 100.00% and 100.00% respectively (for B-NHL), Table 2. Conclusions: The Bengaluru Score is probably the first predictive prognostic score for clinical efficacy outcomes and is based on multivariant statistical analysis in the discovery cohort of patients in the IMAGINE study integrating clinical, laboratory and translational data with a very high sensitivity, specificity and accuracy. The score highlights the role of overall T cell expansion in the final product for B-ALL (but not the T N+CM subsets on its own) and for B-NHL, disease burden and haematological reserve. This suggests that earlier use of varnimcabtagene autoleucel may positively influence efficacy outcome but needs to be proven in the real-world setting. The Bengaluru Score can guide informed treatment decisions and evolve clinical strategies, such as reinfusion, CAR-T targeting a different antigen or allogeneic transplant (if applicable), to reduce the risk of relapse / disease progression in high-risk patients. The Bengaluru Score needs validation in a larger cohort of patients following treatment with varnimcabtagene autoleucel and other CD19 targeted CAR-T cell therapies.

Prediction model for cognitive impairment in maintenance hemodialysis patients

PurposeTo explore the risk factors for cognitive impairment in patients undergoing maintenance hemodialysis (MHD) and construct a predictive model for cognitive impairment.MethodsA total of 146 patients with end-stage renal disease (ESRD) undergoing MHD were recruited at our hospital between December 2021 and April 2022. Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA), and scores of < 26 were considered indicative of cognitive impairment. Risk factors were identified using a multivariate logistic regression model, and a receiver operating characteristic curve was applied to construct the prediction model. Cognitive impairment risk was categorized using a multifactorial prediction model based on the weight of evidence.Results46 patients with cognitive impairment were identified, with a prevalence of 31.5% in ESRD patients undergoing MHD. Multivariate logistic regression analyses indicated that the following factors were associated with an increased risk of cognitive impairment in patients undergoing MHD: aged 55.0–64.0 years (OR:6.24; 95%CI:1.81–21.48; P = 0.001), aged 65.0–74.0 years (OR:16.10; 95%CI:4.03–64.37; P < 0.001), aged ≥ 75.0 years (OR:90.22; 95%CI:16.86-482.86; P < 0.001), duration of dialysis ≥ 5 years (OR:3.99; 95%CI:1.58–10.04; P = 0.003), and current smoker (OR:4.61; 95%CI:1.46–14.57; P = 0.009). The predictive value of the constructed model based on the aforementioned factors for cognitive impairment was 84% (95%CI,77-91%). The prevalence of cognitive impairment for patients at low, moderately low, moderately high, and high risk was 0% (95%CI:0-17%), 10% (95%CI:3-22%), 32% (95%CI:16-52%), and 65% (95%CI:50-78%), respectively.ConclusionsThis study constructed a multifactorial prediction model with a high predictive value for cognitive impairment in patients with ESRD undergoing MHD.

Predicting unipolar and bipolar depression using inflammatory markers, neuroimaging and neuropsychological data: a machine learning study

IntroductionAbout 60% of bipolar disorder (BD) cases are initially misdiagnosed as major depressive disorder (MDD), preventing BD patients from receiving appropriate treatment. An urgency exists to identify reliable biomarkers for improving differential diagnosis (DD). Machine learning methods may help translate current knowledge on biomarkers of mood disorders into clinical practice by providing individual-level classification. No study so far has combined biological data with clinical data to provide a multifactorial predictive model for DD.ObjectivesDefine a predictive algorithm for BD and MDD by integrating structural neuroimaging and inflammatory data with neuropsychological measures (NM). Two different algorithms were compared: multiple kernel learning (MKL) and elastic net regularized logistic regression (EN).MethodsIn a sample of 141 subjects (70 MDD; 71 BD), two different models were implemented for each algorithm: 1) structural neuroimaging measures only (i.e. voxel-based morphometry (VBM), white matter fractional anisotropy (FA), and mean diffusivity (MD)); 2) VBM, FA, and MD combined with NM. In a subsample of 71 subjects (36 BD; 38 MDD), two similar models were implemented: 1) VBM, FA, and, MD combined with only NM; 2) VBM, FA, and MD combined with NM and peripheral inflammatory markers. Finally, the best model was selected for comparison with healthy controls (HC).ResultsOverall, the EN model based on all the modalities achieved the highest accuracy (AUC = 90.2%), outperforming MKL (AUC=85%). EN correctly classified BD and MDD with a diagnostic accuracy of 78.3%, sensitivity of 75%, and specificity of 81.6%. The most significant predictors of BD (variable inclusion probability (VIP) &gt; 80%) were the parahippocampal cingulate, interleukin 9, chemokine CCL5, posterior thalamic radiation, and internal capsule, whereas MDD was best predicted by chemokine CCL23, the anterior cerebellum, and the sagittal stratum. In contrast, NM did not help to differentiate between MDD and BD. However, they help to distinguish patients from HC. Psychomotor coordination and speed of information processing discriminated between MDD and HC (VIP&gt;90%), whereas fluency, working memory, and executive functions differentiated between BD and HC (VIP&gt;80%).ConclusionsIn summary, BD was predicted by a strong proinflammatory profile, whereas MDD was identified by structural neuroimaging data. A multimodal approach offers additional instruments to improve personalized diagnosis in clinical practice and enhance the ability to make DD.Disclosure of InterestNone Declared

Development and Validation of a Multiple Factor-Based Prognostic Score System of Thrombosis in Polycythemia Vera (MFPS-PV)

Introduction: Thrombosis is an important cause of death in patients with polycythemia vera (PV). The conventional two-tiered risk stratification is based on the Polycythemia Vera Study Group (PVSG) criteria and considered only age≥60 years and previous thrombosis as thrombotic risk factors. In the current study, we integrated genetic and clinical information into thrombosis risk stratification and aimed to improve the prediction power of thrombosis in the 2016 WHO-defined PV and propose risk-adapted treatment strategies according to the new model. Methods: Patients were eligible for inclusion if 18 years or older with a confirmed diagnosis of PV according to the 2016 WHO criteria. Clinical and laboratory data were obtained from institutional databases and direct contact with patients. Bone marrow mononuclear cells were collected from the included patients and sent for next-generation sequencing (NGS). A multivariate Cox proportional hazard regression analysis was conducted to identify independent risk factors for thrombosis-free survival (TFS). Hazard ratio (HR)-based risk score allocation was employed to obtain a multifactorial prediction model. Results: The study involved 372 patients in the training cohort and another 195 patients in the external validation cohort. The median follow-up of the training cohort was 43 months (range 3-454) (Table 1). The median age at diagnosis was 53 years (range 18-86) and 59.4% were male. Altogether, 136 patients (36.6%) suffered from thrombosis, including 52 patients (14.0%) with thrombosis before diagnosis, 68 patients (18.3%) at diagnosis, and 68 patients (18.3%) after diagnosis. Mutations in genes encoding for epigenetic modifiers were more frequent in patients with thrombosis than in patients without thrombosis after diagnosis (45.6% vs. 20.1%, p<0.001). Specifically, significant differences were noted in the frequencies of ASXL1 (16.2% vs. 5.9%, p=0.004), DNMT3A (25.0% vs. 4.6%, p<0.001) and BCOR/ BCORL1 (13.2% vs. 4.6%, p=0.017) mutations between patients with and without thrombosis after diagnosis. Multivariate analysis indicated that age≥60 years (HR 2.493, p=0.001), cardiovascular risk factors (HR 4.041, p<0.001), at least one high-risk mutation for thrombosis (mutations in DNMT3A, ASXL1, and BCOR/BCORL1) (HR 4.049, p<0.001) and previous thrombosis (HR 5.576, p<0.001) were independent risk factors for thrombosis. After assigning HR-weighted scores to each risk factor mentioned above (1, 1.5, 1.5 and 2 points respectively), a multiple factor-based prognostic score system of thrombosis was developed, classifying patients into low-risk (≤1 point), intermediate-risk (1.5-2.5 points), and high-risk (≥3 points) groups. Patients in the three groups had significantly different TFS (p<0.001) and remained consistent during external validation (Figure 1). Finally, a risk-adapted therapeutic strategy was established based on the MFPS-PV. TP and low-dose antiplatelet therapy are recommended for all patients to achieve a targeted HCT level lower than 45%. Low-risk patients are not mandatory for cytoreduction and only require low-dose antiplatelet therapy. Intermediate-risk patients are advised for antiplatelet therapy, with cytoreduction added if they have two thrombotic risk factors, previous thrombosis, or age≥60 years. High-risk patients are recommended antiplatelet and cytoreductive combination therapy. Conclusion: The MFPS-PV, integrating genetic and clinical characteristics for the first time, can significantly predict thrombosis and has great therapeutic implications. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Prediction Model for Urinary Tract Infection in Pediatric Urological Surgery Patients

BackgroundUrinary tract infection (UTI) is a common complication in pediatric urological surgery patients and is associated with long-term sequelae, including subsequent recurrent infections and renal scarring. In this study, we aimed to explore the risk factors for UTI in pediatric urological surgery patients and construct a predictive model for UTI.Materials and MethodsA total of 2,235 pediatric patients who underwent urological surgery at a tertiary hospital between February 2019 and January 2020 were included. A multivariate logistic regression model was applied to identify the predictive factors, and a predictive model was constructed using a receiver operating characteristic curve. A multifactorial predictive model was used to categorize the risk of UTI based on the weight of the evidence.ResultsA total of 341 patients with UTI were identified, which corresponded to a prevalence of 15.26% in pediatric urological surgery patients. Multivariate analysis identified six significant risk factors for UTI, including age <12.0 months, upper urinary tract disease, not using an indwelling drainage tube, hospital stay ≥10 days, administration of two or more types of antibiotics, and stent implantation. A combination of the aforementioned factors produced an area under the curve value of 88.37% for preventing UTI in pediatric urological surgery patients. A multifactorial predictive model was created based on the combination of these factors.ConclusionsThe constructed multifactorial model could predict UTI risk in pediatric urological surgery patients with a relatively high predictive value.

30-day mortality in patients treated for brain metastases: extracranial causes dominate

BackgroundEstablished prognostic models, such as the diagnosis-specific graded prognostic assessment, were not designed to specifically address very short survival. Therefore, a brain metastases-specific 30-day mortality model may be relevant. We hypothesized that in-depth evaluation of a carefully defined cohort with short survival, arbitrarily defined as a maximum of 3 months, may provide signals and insights, which facilitate the development of a 30-day mortality model.MethodsRetrospective analysis (2011–2021) of patients treated for brain metastases with different approaches. Risk factors for 30-day mortality from radiosurgery or other primary treatment were evaluated.ResultsThe cause of death was unrelated to brain metastases in 61%. Treatment-related death (grade 5 toxicity) did not occur. Completely unexpected death was not observed, e.g. accident, suicide or sudden cardiac death. Logistic regression analysis showed 9 factors associated with 30-day mortality (each assigned 3–6 points) and a point sum was calculated for each patient. The point sum ranged from 0 (no risk factors for death within 30 days present) to 30. The results can be grouped into 3 or 4 risk categories. Eighty-three percent of patients in the highest risk group (> 16 points) died within 30 days, and none survived for more than 2 months. However, many cases of 30-day mortality (more than half) occurred in intermediate risk categories.ConclusionExtracranial tumor progression was the prevailing cause of 30-day mortality and few, if any deaths could be considered relatively unexpected when looking at the complete oncological picture. We were able to develop a multifactorial prediction model. However, the model’s performance was not fully satisfactory and it is not routinely applicable at this point in time, because external validation is needed to confirm our hypothesis-generating findings.

Prediction Model for Bronchopulmonary Dysplasia in Preterm Newborns.

OBJECTIVE: To develop a multifactorial model that allows the prediction of bronchopulmonary dysplasia (BPD) in preterm newborns. MATERIALS AND METHODS: A single-center retrospective study of infants born below 32 + 0 weeks gestational age. We created a receiver operating characteristic curve to assess the multifactorial BPD risk and calculate the BPD risk accuracy using the area under the curve (AUC). BPD risk was categorized using a multifactorial predictive model based on the weight of the evidence. RESULTS: Of the 278 analyzed preterm newborns, 127 (46%) developed BPD. The significant risk factors for BPD in the multivariate analysis were gestational age, number of red blood cell concentrate transfusions, number of surfactant administrations, and hemodynamically significant patent ductus arteriosus. The combination of these factors determined the risk of developing BPD, with an AUC value of 0.932. A multifactorial predictive model based on these factors, weighted by their odds ratios, identified four categories of newborns with mean BPD risks of 9%, 59%, 82%, and 100%. CONCLUSION: A multifactorial model based on easily available clinical factors can predict BPD risk in preterm newborns and inform potential preventive measures.

A beetle antennae search improved BP neural network model for predicting multi-factor-based gas explosion pressures

A gas explosion in an underground structure may cause serious damage to the human body and ground buildings and may result in huge economic losses. The pressure of the gas explosion is an important parameter in determining its severity and designating an emergency plan. However, existing empirical and computational fluid dynamics (CFD) methods for pressure prediction are either inaccurate or inefficient when considering multiple influencing factors and their interrelationships. Therefore, for a more efficient and reliable prediction, the present study developed a multifactorial prediction model based on a beetle antennae search (BAS) algorithm improved back propagation (BP) neural network. A total of 317 sets of data which considered factors of geometry, gas, obstacle, vent, and ignition were collected from previous studies. The results showed that the established model can predict pressures accurately by low RMSE (43.4542 and 50.7176) and MAPE (3.9666% and 4.9605%) values and high R2 (0.7696 and 0.7388) values for training and testing datasets, respectively. Meanwhile, the BAS algorithm was applied to improve both the calculation efficiency and the accuracy of the proposed model by enabling a more intelligent hyperparameter tuning method. Furthermore, the permutation importance of input variables was investigated, and the length (L) and the ratio of length and diameter (L/D) of geometry were found to be the most critical factors that affect the explosion pressure level.

Response prediction to oxaliplatin plus 5-fluorouracil chemotherapy in patients with colorectal cancer using a four-protein immunohistochemical model.

The response of cancer patients to oxaliplatin combined with 5-fluorouracil (5-FU) is difficult to predict. It has been reported that carcinoma-associated fibroblasts (CAFs) could induce AKT and ERK phosphorylation, and upregulate survivin expression in colorectal cancer (CRC) cells, which could lead to oxaliplatin plus 5-FU resistance. A total of 71 patients with advanced CRC (aCRC) treated with oxaliplatin plus 5-FU were included in the present study. These patients comprised 46 chemotherapy responders and 25 non-responders. The expression levels of α-smooth muscle actin (α-SMA), phosphorylated (p)-AKT, p-ERK and survivin were determined by immunohistochemical evaluation of paraffin-embedded samples from patients. A predictive model was established using a Probabilistic Neural Network model. The high expression of α-SMA, p-AKT and survivin in patients with aCRC were associated with oxaliplatin plus 5-FU resistance (P<0.001, P=0.023 and P=0.001, respectively). Furthermore, patients with stage IV CRC exhibiting high expression levels of α-SMA and survivin experienced a reduced progression-free survival time compared with patients with low expressions of α-SMA and survivin (5.5 vs. 15.0 months; 5.5 vs. 15.0 months; P=0.005 and P=0.001, respectively). Stage IV CRC and high survivin expression predicted a reduced overall survival time compared with that for patients with stage IV CRC and low survivin expression (50.0 vs. 15.0 months; P<0.001). Patients with α-SMA, p-AKT, p-ERK and survivin overexpression were more likely to present with intrinsic resistance to the oxaliplatin plus 5-FU regimen (the accuracies of modeling, validation and prediction were 83.7, 92.9 and 85.7%, respectively). In conclusion, the multifactorial predictive biomarker model of α-SMA, p-AKT, p-ERK and survivin expression for patients with aCRC to predict intrinsic resistance to oxaliplatin plus 5-FU regimens is of great efficiency and accuracy. Patients with high expression of this predictive model may be intrinsically resistant to the oxaliplatin and 5-FU regimen.

Identifying women at risk for sustained postpartum anxiety

IntroductionTo describe the prevalence of sustained postpartum anxiety and to develop a multifactorial predictive model to assist in targeted screening procedures. MethodsIn a population-based cohort in a health region near Vancouver, Canada, 522 mothers completed a mailed questionnaire at 1, 4, and 8 weeks postpartum measuring socio-demographic, biological, pregnancy-related, life stressor, social support, obstetric, and maternal adjustment factors. We undertook a sequential logistic regression analysis to develop a multifactorial predictive model of sustained postpartum anxiety, as measured by a State Trait Anxiety Inventory (STAI) score >40 at 1 week and/or 4 weeks, and 8 weeks postpartum. ResultsThe prevalence of sustained postpartum anxiety was 12.6% (95% CI 9.6–16.2). In the multivariable model, predictors of sustained anxiety in the postpartum period were perceived stress at 1 week (1 SD increase; aOR 3.74, 95% CI 2.17-6.44) and partner social support at 1 week (1 SD increase; aOR 0.59, 95% CI 0.40–0.85). Depression symptomatology at 1 week, child care stress, and maternal self-esteem were non-significant. LimitationsSingle women and women from ethnic minority backgrounds were underrepresented in the sample. ConclusionsA large proportion of women experience sustained postpartum anxiety. High perceived stress and low partner social support can be used to facilitate early identification of women likely to experience persistent anxiety in the postpartum period and suggest the need for urgent access to psychotherapeutic services for these women. These factors may also be potential targets for individual or couples therapy to treat postpartum anxiety.

THE INFLUENCE OF ECOLOGICAL RISK FACTORS ON MORTALITY IN MOSCOW POPULATION. RISK EVALUATION AND PREDICTION

Aim. To assess the influence of temperature, humidity, air pollution on the mortality ofMoscowinhabitants in 2007-2014, to evaluate the possibility of mortality prediction by mathematic modelling. Material and methods. For each day of 2007-2014 years, the data was analyzed forMoscowinhabitants mortality, temperature, humidity and air pollution (РМ10). Number of deaths was 956436 (48,1% of men and 51,9% women), of those from cardiovascular disorders (CVD) — 536625. Multifactorial regression analysis was done with generalized logistic regression, evaluation of risks ratios (RR), 95%, confidence analysis. The prognostic modelling applied, based on non-linear multifactorial analysis. Results. The factor “heat wave” increased RR for the death from all cases and from CVD by 1,43 and 1,65 times. Heat waves mostly increased all-cause mortality, and from CVD in women (p<0,005), the increase of air pollution (РМ10) — in men (р<0,005). The highest increase of death risk was found for those 70 years and older, higher humidity worsened it. For the summer 2010 prediction was finer than for 2014 (correlation 90,1% and 45,1%), e. g. for the period of more significant anomalies of temperature and air pollution. Applying multifactorial prediction model for 2007-2014 period with additional evaluation of weather factors influence for last 4 days and all-cause mortality for the last month, correlation was higher for death risk prediction for 1-3 days (88,0%), than for month (45,2%). Conclusion. Heat waves, increase of air temperature and pollution do increase the risk of all-cause mortality, especially from CVD, in those 70 y. o. and older. Anomal heat increases death risk for women higher than for men. Prediction of mortality by 1-3 days with multifactorial model show better correlation than prediction for month.

Externally validated HPV-based prognostic nomogram for oropharyngeal carcinoma patients yields more accurate predictions than TNM staging

PurposeDue to the established role of the human papillomavirus (HPV), the optimal treatment for oropharyngeal carcinoma is currently under debate. We evaluated the most important determinants of treatment outcome to develop a multifactorial predictive model that could provide individualized predictions of treatment outcome in oropharyngeal carcinoma patients. MethodsWe analyzed the association between clinico-pathological factors and overall and progression-free survival in 168 OPSCC patients treated with curative radiotherapy or concurrent chemo-radiation. A multivariate model was validated in an external dataset of 189 patients and compared to the TNM staging system. This nomogram will be made publicly available at www.predictcancer.org. ResultsPredictors of unfavorable outcomes were negative HPV-status, moderate to severe comorbidity, T3–T4 classification, N2b–N3 stage, male gender, lower hemoglobin levels and smoking history of more than 30 pack years. Prediction of overall survival using the multi-parameter model yielded a C-index of 0.82 (95% CI, 0.76–0.88). Validation in an independent dataset yielded a C-index of 0.73 (95% CI, 0.66–0.79. For progression-free survival, the model’s C-index was 0.80 (95% CI, 0.76–0.88), with a validation C-index of 0.67, (95% CI, 0.59–0.74). Stratification of model estimated probabilities showed statistically different prognosis groups in both datasets (p<0.001). ConclusionThis nomogram was superior to TNM classification or HPV status alone in an independent validation dataset for prediction of overall and progression-free survival in OPSCC patients, assigning patients to distinct prognosis groups. These individualized predictions could be used to stratify patients for treatment de-escalation trials.

Propensity for biofilm formation by clinical isolates from urinary tract infections: developing a multifactorial predictive model to improve antibiotherapy

A group of biofilm-producing bacteria isolated from patients with urinary tract infections was evaluated, identifying the main factors contributing to biofilm formation. Among the 156 isolates, 58 (37.2%) were biofilm producers. The bacterial species (P<0.001), together with patient's gender (P = 0.022), were the factors with the highest influence for biofilm production. There was also a strong correlation of catheterization with biofilm formation, despite being less significant (P = 0.070) than species or gender. In fact, some of the bacteria isolated were biofilm producers in all cases. With regard to resistance profile among bacterial isolates, β-lactam antibiotics presented the highest number of cases/percentages--ampicillin (32/55.2%), cephalothin (30/51.7%), amoxicillin/clavulanic acid (22/37.9%)--although the carbapenem group still represented a good therapeutic option (2/3.4%). Quinolones (nucleic acid synthesis inhibitors) also showed high resistance percentages. Furthermore, biofilm production clearly increases bacterial resistance. Almost half of the biofilm-producing bacteria showed resistance against at least three different groups of antibiotics. Bacterial resistance is often associated with catheterization. Accordingly, intrinsic (age and gender) and extrinsic (hospital unit, bacterial isolate and catheterization) factors were used to build a predictive model, by evaluating the contribution of each factor to biofilm production. In this way, it is possible to anticipate biofilm occurrence immediately after bacterial identification, allowing selection of a more effective antibiotic (among the susceptibility options suggested by the antibiogram) against biofilm-producing bacteria. This approach reduces the putative bacterial resistance during treatment, and the consequent need to adjust antibiotherapy.

TOP2A Protein by Quantitative Immunofluorescence As a Predictor of Response to Epirubicin in the Neoadjuvant Treatment of Breast Cancer

Anthracyclines are commonly used in breast cancer, although they lack validated predictive biomarkers. We explored the interaction between TOP2A protein by quantitative immunofluorescence (QIF) and anthracycline sensitivity. Patients with estrogen receptor-negative breast cancer received neoadjuvant epirubicin. Pretreatment biopsies were analyzed using AQUA(®). Total, cytoplasmic (C) and nuclear (N) TOP2A protein concentrations were expressed as QIF scores and compared with pathologic complete response (pCR), TOP2A by immunohistochemistry, TOP2A mRNA, TOP2A and HER2 gene status, and Ki-67 level. In total, 76 cases were assessable. C, N, and total scores did not correlate with pCR, or other markers. The N:C ratio differed significantly by HER2 status. No pCRs occurred in patients in the lowest N:C quartile. Although no relevant correlation between TOP2A QIF scores and pCR was found, N:C ratio may have a negative predictive role, and may merit further exploration in a multifactorial predictive model that includes tumor and host factors.

4:34 PM Abstract No. LB04 - A multifactorial predictive model of coiled intracranial aneurysm retreatment

4:34 PM Abstract No. LB04 - A multifactorial predictive model of coiled intracranial aneurysm retreatment