Differences In Predictive Performance Research Articles

Prioritising deteriorating patients using time-to-event analysis: prediction model development and internal–external validation

BackgroundBinary classification models are frequently used to predict clinical deterioration, however they ignore information on the timing of events. An alternative is to apply time-to-event models, augmenting clinical workflows by ranking patients by predicted risks. This study examines how and why time-to-event modelling of vital signs data can help prioritise deterioration assessments using lift curves, and develops a prediction model to stratify acute care inpatients by risk of clinical deterioration.MethodsWe developed and validated a Cox regression for time to in-hospital mortality. The model used time-varying covariates to estimate the risk of clinical deterioration. Adult inpatient medical records from 5 Australian hospitals between 1 January 2019 and 31 December 2020 were used for model development and validation. Model discrimination and calibration were assessed using internal–external cross validation. A discrete-time logistic regression model predicting death within 24 h with the same covariates was used as a comparator to the Cox regression model to estimate differences in predictive performance between the binary and time-to-event outcome modelling approaches.ResultsOur data contained 150,342 admissions and 1016 deaths. Model discrimination was higher for Cox regression than for discrete-time logistic regression, with cross-validated AUCs of 0.96 and 0.93, respectively, for mortality predictions within 24 h, declining to 0.93 and 0.88, respectively, for mortality predictions within 1 week. Calibration plots showed that calibration varied by hospital, but this can be mitigated by ranking patients by predicted risks.ConclusionTime-varying covariate Cox models can be powerful tools for triaging patients, which may lead to more efficient and effective care in time-poor environments when the times between observations are highly variable.

Whole-tumor histogram analysis of multiple non-Gaussian diffusion models at high b values for assessing cervical cancer.

To assess the diagnostic potential of whole-tumor histogram analysis of multiple non-Gaussian diffusion models for differentiating cervical cancer (CC) aggressive status regarding of pathological types, differentiation degree, stage, and p16 expression. Patients were enrolled in this prospective single-center study from March 2022 to July 2023. Diffusion-weighted images (DWI) were obtained including 15 b-values (0 ~ 4000s/mm2). Diffusion parameters derived from four non-Gaussian diffusion models including continuous-time random-walk (CTRW), diffusion-kurtosis imaging (DKI), fractional order calculus (FROC), and intravoxel incoherent motion (IVIM) were calculated, and their histogram features were analyzed. To select the most significant features and establish predictive models, univariate analysis and multivariate logistic regression were performed. Finally, we evaluated the diagnostic performance of our models by using receiver operating characteristic (ROC) analyses. 89 women (mean age, 55 ± 11 years) with CC were enrolled in our study. The combined model, which incorporated the CTRW, DKI, FROC, and IVIM diffusion models, offered a significantly higher AUC than that from any individual models (0.836 vs. 0.664, 0.642, 0.651, 0.649, respectively; p < 0.05) in distinguishing cervical squamous cell cancer from cervical adenocarcinoma. To distinguish tumor differentiation degree, except the combined model showed a better predictive performance compared to the DKI model (AUC, 0.839 vs. 0.697, respectively; p < 0.05), no significant differences in AUCs were found among other individual models and combined model. To predict the International Federation of Gynecology and Obstetrics (FIGO) stage, only DKI and FROC model were established and there was no significant difference in predictive performance among different models. In terms of predicting p16 expression, the predictive ability of DKI model is significantly lower than that of FROC and combined model (AUC, 0.693 vs. 0.850, 0.859, respectively; p < 0.05). Multiple non-Gaussian diffusion models with whole-tumor histogram analysis show great promise to assess the aggressive status of CC.

Development of a Machine Learning-Based Model for Predicting the Incidence of Peripheral Intravenous Catheter-Associated Phlebitis.

Early and accurate identification of high-risk patients with peripheral intravascular catheter (PIVC)-related phlebitis is vital to prevent medical device-related complications. This study aimed to develop and validate a machine learning-based model for predicting the incidence of PIVC-related phlebitis in critically ill patients. Four machine learning models were created using data from patients ≥ 18 years with a newly inserted PIVC during intensive care unit admission. Models were developed and validated using a 7:3 split. Random survival forest (RSF) was used to create predictive models for time-to-event outcomes. Logistic regression with least absolute reduction and selection operator (LASSO), random forest (RF), and gradient boosting decision tree were used to develop predictive models that treat outcome as a binary variable. Cox proportional hazards (COX) and logistic regression (LR) were used as comparators for time-to-event and binary outcomes, respectively. The final cohort had 3429 PIVCs, which were divided into the development cohort (2400 PIVCs) and validation cohort (1029 PIVCs). The c-statistic (95% confidence interval) of the models in the validation cohort for discrimination were as follows: RSF, 0.689 (0.627-0.750); LASSO, 0.664 (0.610-0.717); RF, 0.699 (0.645-0.753); gradient boosting tree, 0.699 (0.647-0.750); COX, 0.516 (0.454-0.578); and LR, 0.633 (0.575-0.691). No significant difference was observed among the c-statistic of the four models for binary outcome. However, RSF had a higher c-statistic than COX. The important predictive factors in RSF included inserted site, catheter material, age, and nicardipine, whereas those in RF included catheter dwell duration, nicardipine, and age. The RSF model for the survival time analysis of phlebitis occurrence showed relatively high prediction performance compared with the COX model. No significant differences in prediction performance were observed among the models with phlebitis occurrence as the binary outcome.

Utility of pneumonia severity assessment tools for mortality prediction in healthcare-associated pneumonia: a systematic review and meta-analysis

Accurate prognostic tools for mortality in patients with healthcare-associated pneumonia (HCAP) are needed to provide appropriate medical care, but the efficacy for mortality prediction of tools like PSI, A-DROP, I-ROAD, and CURB-65, widely used for predicting mortality in community-acquired and hospital-acquired pneumonia cases, remains controversial. In this study, we conducted a systematic review and meta-analysis using PubMed, Cochrane Library (trials), and Ichushi web database (accessed on August 22, 2022). We identified articles evaluating either PSI, A-DROP, I-ROAD, or CURB-65 and the mortality outcome in patients with HCAP, and calculated the pooled sensitivities, specificities, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and the summary area under the curves (AUCs) for mortality prediction. Additionally, the differences in predicting prognosis among these four assessment tools were evaluated using overall AUCs pooled from AUC values reported in included studies. Eventually, 21 articles were included and these quality assessments were evaluated by QUADAS-2. Using a cut-off value of moderate in patients with HCAP, the range of pooled sensitivity, specificity, PLR, NLR, and DOR were found to be 0.91–0.97, 0.15–0.44, 1.14–1.66, 0.18–0.33, and 3.86–9.32, respectively. Upon using a cut-off value of severe in those patients, the range of pooled sensitivity, specificity, PLR, NLR, and DOR were 0.63–0.70, 0.54–0.66, 1.50–2.03, 0.47–0.58, and 2.66–4.32, respectively. Overall AUCs were 0.70 (0.68–0.72), 0.70 (0.63–0.76), 0.68 (0.64–0.73), and 0.67 (0.63–0.71), respectively, for PSI, A-DROP, I-ROAD, and CURB-65 (p = 0.66). In conclusion, these severity assessment tools do not have enough ability to predict mortality in HCAP patients. Furthermore, there are no significant differences in predictive performance among these four severity assessment tools.

Comparative near Infrared (NIR) spectroscopy calibrations performance of dried and undried forage on dry and wet matter bases

The application of Near Infrared (NIR) spectroscopy for analyzing wet feed directly on farms is increasingly recognized for its role in supporting harvest-time decisions and refining the precision of animal feeding practices. This study aims to evaluate the accuracy of NIR spectroscopy calibrations for both undried, unprocessed samples and dried, ground samples. Additionally, it investigates the influence of the bases of reference data (wet vs. dry basis) on the predictive capabilities of the NIR analysis. The study utilized 492 Corn Whole Plant (CWP) and 405 High Moisture Corn (HMC) samples, sourced from various farms across Italy. Spectral data were acquired from both undried, unground and dried, ground samples using laboratory bench NIR instruments, covering a spectral range of 1100 to 2498 nm. The reference chemical composition of these samples was analyzed and presented in two formats: on a wet matter basis and on a dry matter basis.The study revealed that calibrations based on undried samples generally exhibited lower predictive accuracy for most traits, with the exception of Dry Matter (DM). Notably, the decline in predictive performance was more pronounced in highly moist products like CWP, where the average error increased by 60–70%. Conversely, this reduction in accuracy was relatively contained (10–15%) in drier samples such as HMC. The Standard Error of Cross-Validation (SECV) values for DMres, Ash, CP, and EE were notably low, at 0.39, 0.30, 0.29, 0.21% for CWP and 0.49, 0.14, 0.25, 0.14% for HMC, respectively. These results align with previous studies, indicating the reliability of NIR spectroscopy in diverse moisture contexts.The study attributes this variance to the interference caused by water in 'as is' samples, where the spectral features predominantly reflect water content, thereby obscuring the spectral signatures of other nutrients. In terms of calibration development strategies, the study concludes that there is no significant difference in predictive performance between undried calibrations based on either 'dry matter' or 'as is' basis. This finding emphasizes the potential of NIR spectroscopy in diverse moisture contexts, although with varying degrees of accuracy contingent upon the moisture content of the analyzed samples. Overall, this research provides valuable insights into the calibration strategies of NIR spectroscopy and its practical applications in agricultural settings, particularly for on-farm forage analysis.

Association between psoas major muscle mass and CPET performance and long-term survival following major colorectal surgery: A retrospective cohort study

ObjectivesTo evaluate whether computed tomography (CT)-derived psoas major muscle measurements could predict preoperative cardiopulmonary exercise testing (CPET) performance and long-term mortality in patients undergoing major colorectal surgery and to compare predictive performance of psoas muscle measurements using 2D approach and 3D approach. MethodsA retrospective cohort study compliant with STROCSS standards was conducted. Consecutive patients undergoing major colorectal surgery between January 2011 and January 2017 following CPET as part of their preoperative assessment were included. Regression analyses were modelled to investigate association between the CT-derived psoas major muscle mass variables [total psoas muscle area (TPMA), total psoas muscle volume (TPMV) and psoas muscle index (PMI)] and CPET performance and mortality (1-year and 5-year). Discriminative performances of the variables were evaluated using Receiver Operating Characteristic (ROC) curve analysis. ResultsA total of 457 eligible patients were included. The median TPMA and TPMV were 21 ​cm2 (IQR: 15–27) and 274 ​cm3 (IQR: 201–362), respectively. The median PMI measured via 2D and 3D approaches were 7 ​cm2/m2 (IQR: 6–9) and 99 ​cm3/m2 (IQR: 76–120), respectively. The risks of 1-year and 5-year mortality were 7.4% and 27.1%, respectively. Regression analyses showed TPMA, TPMV, and PMI can predict preoperative CPET performance and long-term mortality. However, ROC curve analyses showed no significant difference in predictive performance amongst TPMA, TPMV, and PMI. ConclusionRadiologically-measured psoas muscle mass variables may predict preoperative CPET performance and may be helpful with informing more objective selection of patients for preoperative CPET and prehabilitation.

Explaining inter-individual differences in habitat relationships among wildcat hybrids in Scotland

Little is known about the factors that drive nonstationarity and inter-individual differences in realized habitat niches and species-environment relationships. We explored this topic by developing individual habitat selection models for 14 wildcat hybrids distributed across Scotland, and assessed how differences in their predicted probabilities of occurrence were related to factors including (1) geographic distance, (2) multivariate ecological distance, (3) difference in degree of hybridization and (4) difference in sex (male vs female). We found that the individual models were exceptionally effective in predicting the habitat use and occurrence of the particular individuals on whose data they were trained, but were generally highly divergent and not transferable among individuals. We conducted a reciprocal validation approach where we calculated the AUC for each individual model, predicting the occurrence patterns of the 13 other individuals. We then fit regression and nonparametric splines to evaluate the impacts of geographical distance, ecological distance, hybridization distance and difference in the sex of individuals in the ability of individual wildcat hybrid habitat models to predict the occurrences of other individuals. We found that, of the four factors assessed, ecological distance was supported as being inversely related to ability of a model from one individual to predict occurrence of another individual. The other three factors were not strongly related to differences in reciprocal model predictive ability. This suggests that ecological differences where individual wildcat hybrids reside drive differences in their habitat selection, but that geographical distance, degree of genetic hybridization and difference in the sex of individuals are not consistently associated with differences in model prediction or reciprocal validation performance. These results highlight the effect of ecological limiting factors, and the importance of nonstationary limiting factors in determining the habitat they select, their expressed species-environment relationship and the description of their realized habitat niches.

Proposal of a methodology for prediction of heavy metals concentration based on PM2.5 concentration and meteorological variables using machine learning

In this study, we developed a prediction model for heavy metal concentrations using PM2.5 concentrations and meteorological variables. Data was collected from five sites, encompassing meteorological factors, PM2.5, and 18 metals over 2 years. The study employed four analytical methods: multiple linear regression (MLR), random forest regression (RFR), gradient boosting, and artificial neural networks (ANN). RFR was the best predictor for most metals, and gradient boosting and ANN were optimal for certain metals like Al, Cu, As, Mo, Zn, and Cd. Upon evaluating the final model’s predicted values against the actual measurements, differences in the concentration distribution between measurement locations were observed for Mn, Fe, Cu, Ba, and Pb, indicating varying prediction performances among sites. Additionally, Al, As, Cd, and Ba showed significant differences in prediction performance across seasons. The developed model is expected to overcome the technical limitations involved in measuring and analyzing heavy metal concentrations. It could further be utilized to obtain fundamental data for studying the health effects of exposure to hazardous substances such as heavy metals.

Radiomics analysis of contrast-enhanced computerized tomography for differentiation of gastric schwannomas from gastric gastrointestinal stromal tumors

PurposeTo assess the performance of radiomics-based analysis of contrast-enhanced computerized tomography (CE-CT) images for distinguishing GS from gastric GIST.MethodsForty-nine patients with GS and two hundred fifty-three with GIST were enrolled in this retrospective study. CT features were evaluated by two associate chief radiologists. Radiomics features were extracted from portal venous phase images using Pyradiomics software. A non-radiomics dataset (combination of clinical characteristics and radiologist-determined CT features) and a radiomics dataset were used to build stepwise logistic regression and least absolute shrinkage and selection operator (LASSO) logistic regression models, respectively. Model performance was evaluated according to sensitivity, specificity, accuracy, and receiver operating characteristic (ROC) curve, and Delong’s test was applied to compare the area under the curve (AUC) between different models.ResultsA total of 1223 radiomics features were extracted from portal venous phase images. After reducing dimensions by calculating Pearson correlation coefficients (PCCs), 20 radiomics features, 20 clinical characteristics + CT features were used to build the models, respectively. The AUC values for the models using radiomics features and those using clinical features were more than 0.900 for both the training and validation groups. There were no significant differences in predictive performance between the radiomic and clinical data models according to Delong’s test.ConclusionA radiomics-based model applied to CE-CT images showed comparable predictive performance to senior physicians in the differentiation of GS from GIST.

A unified deep learning framework for water quality prediction based on time-frequency feature extraction and data feature enhancement

Deep learning methods exhibited significant advantages in mapping highly nonlinear relationships with acceptable computational speed, and have been widely used to predict water quality. However, various model selection and construction methods resulted in differences in prediction accuracy and performance. Hence, a unified deep learning framework for water quality prediction was established in the paper, including data processing module, feature enhancement module, and data prediction module. In the established model, the data processing module based on wavelet transform method was applied to decomposing complex nonlinear meteorology, hydrology, and water quality data into multiple frequency domain signals for extracting self characteristics of data cyclic and fluctuations. The feature enhancement module based on Informer Encoder was used to enhance feature encoding of time series data in different frequency domains to discover global time dependent features of variables. Finally, the data prediction module based on the stacked bidirectional long and short term memory network (SBiLSTM) method was employed to strengthen the local correlation of feature sequences and predict the water quality. The established model framework was applied in Lijiang River in Guilin, China. The maximum relative errors between the predicted and observed values for dissolved oxygen (DO), chemical oxygen demand (CODMn) were 12.4% and 20.7%, suggesting a satisfactory prediction performance of the established model. The validation results showed that the established model was superior to all other models in terms of prediction accuracy with RMSE values 0.329, 0.121, MAE values 0.217, 0.057, SMAPE values 0.022, 0.063 for DO and CODMn, respectively. Ablation tests confirmed the necessity and rationality of each module for the established model framework. The established method provided a unified deep learning framework for water quality prediction.

Performance analysis of machine learning methods for short-term traffic prediction

With the rapid development of urbanization and the rapid increase of the number of motor vehicles, the problem of urban traffic congestion has become increasingly prominent. The accurate prediction of short-term traffic flow is considered as a promising solution, which can provide a key decision-making basis for route planning and traffic flow scheduling, so that can greatly alleviate or even prevent congestion. Researchers have used many machine learning methods to predict traffic flow, but few people pay attention to the boundaries of different machine algorithms. In this paper, we use AdaBoost, Random Forest, SVM and BP neural network to predict short-term traffic flow in California, which aims to compare the differences in prediction performance of different algorithms and analyze their potential reasons. The results show that, the integration methods such as AdaBoost and Random Forest are quite appropriate to solve the short-term traffic flow, which can obtain an accuracy more than 95%, while prediction made by SVM is less precise than the two aforementioned methods with a 79% accuracy. And BP neural network may be inappropriate if the parameters remain default. The different results are due to the periodicity of the database. Integration methods can recognize the periodicity while the SVM and BP neural network fail to do it. When employing the SVM and BP neural network, the datasets need to be divided within a period to avoid being disturbed by cyclically. Besides, the precise of BP neural network can be improved when adjusting the parameters to the optimal.

Prediction models of suicide and non-fatal suicide attempt after discharge from a psychiatric inpatient stay: A machine learning approach on nationwide Danish registers.

To develop machine learning models capable of predicting suicide and non-fatal suicide attempt as separate outcomes in the first 30 days after discharge from a psychiatric inpatient stay. Prospective cohort study using nationwide Danish registry data. We included individuals who were 18 years or older, and all discharges from psychiatric hospitalizations in Denmark from 1995 to 2018. We trained predictive models using 10-fold cross validation on 80% of the data and did testing on the remaining 20%. The best model for predicting non-fatal suicide attempt was an ensemble of predictions from gradient boosting (XGBoost) and categorical boosting (catBoost). The ROC-AUC for predicting suicide attempt was 0.85 (95% CI: 0.84-0.85). At a risk threshold of 4.36%, positive predictive value (PPV) was 11.0% and sensitivity was 47.2%. The best model for predicting suicide was an ensemble of predictions from random forest, XGBoost and catBoost. For suicide, the ROC-AUC was 0.71 (95% CI: 0.70-0.73). At a risk threshold of 0.15%, PPV was 0.34% and sensitivity was 56.0%. The most contributing predictors differed when predicting suicideand suicide attempt, indicating that separate models are needed. The ensemble model was fair across sex and age, and more so than the penalized logistic regression model. We achieved good performance for predicting suicide attempts and demonstrated a clinical application of ensemble models. Our results indicate a difference in predictive performance for models predicting suicide and suicide attempt, respectively. Thus, we recommend that suicide and suicide attempt are treated as two separate endpoints, in particular for clinical application. We demonstrated that the ensemble model is fairer across sex and age compared with a penalized logistic regression, and therefore we recommend the use of well-tested ensembles despite a more complex explainability.

Warming increases the differences among spring phenology models under future climate change.

Phenological models are built upon an understanding of the influence of environmental factors on plant phenology, and serve as effective tools for predicting plant phenological changes. However, the differences in phenological model predictive performance under different climate change scenarios have been rarely studied. In this study, we parameterized thirteen spring phenology models, including six one-phase models and seven two-phase models, by combining phenological observations and meteorological data. Using climatic data from two Shared Socioeconomic Pathways (SSP) scenarios, namely SSP126 (high mitigation and low emission) and SSP585 (no mitigation and high emission), we predicted spring phenology in Germany from 2021 to 2100, and compared the impacts of dormancy phases and driving factors on model predictive performance. The results showed that the average correlation coefficient between the predicted start of growing season (SOS) by the 13 models and the observed values exceeded 0.72, with the highest reaching 0.80. All models outperformed the NULL model (Mean of SOS), and the M1 model (driven by photoperiod and forcing temperature) performed the best for all the tree species. In the SSP126 scenario, the average SOS advanced initially and then gradually shifted towards a delay starting around 2070. In the SSP585 scenario, the average SOS advanced gradually at a rate of approximately 0.14 days per year. Moreover, the standard deviation of the simulated SOS by the 13 spring phenology models exhibited a significant increase at a rate of 0.04 days per year. On average, two-phase models exhibited larger standard deviations than one-phase models after approximately 2050. Models driven solely by temperature showed larger standard deviations after 2060 compared to models driven by both temperature and photoperiod. Our findings suggest investigating the release mechanisms of endodormancy phase and incorporating new insights into future phenological models to better simulate the changes in plant phenology.

Prior physical illness predicts death better than acute physiological derangement on intensive care unit admission in COVID-19: A Swedish registry study.

COVID-19 is associated with prolonged intensive care unit (ICU) stay and considerable mortality. The onset of persistent critical illness, defined as when prior illness predicts death better than acute physiological derangement, has not been studied in COVID-19. This national cohort study based on the Swedish Intensive Care Registry (SIR) included all patients admitted to a Swedish ICU due to COVID-19 from 6 March 2020 to 9 November 2021. Simplified Acute Physiology Score-3 (SAPS3) Box 1 was used as a measure of prior illness and Box 3 as a measure of acute derangement to evaluate the onset and importance of persistent critical illness in COVID-19. To compare predictive capacity, the area under receiver operating characteristic (AUC) of SAPS3 and its constituent Box 1 and 3 was calculated for 30-day mortality. In 7 969 patients, of which 1 878 (23.6%) died within 30 days of ICU admission, the complete SAPS3 score had acceptable discrimination: AUC 0.75 (95% CI 0.74 to 0.76) but showed under prediction in low-risk patients and over prediction in high-risk patients. SAPS3 Box 1 showed markedly better discrimination than Box 3 (AUC 0.74 vs 0.65, P<0,0001). Using custom logistic models, the difference in predictive performance of prior and acute illness was validated, AUC 0.76 vs AUC 0.69, p<0.0001. Prior physical illness predicts death in COVID-19 better than acute physiological derangement during ICU stay, and the whole SAPS3 score is not significantly better than just prior illness. The results suggests that COVID-19 may exhibit similarities to persistent critical illness immediately from ICU admission, potentially because of long median ICU length-of-stay. Alternatively, the variables in the acute physiological derangement model may not adequately capture the severity of illness in COVID-19.

Development and External Validation of Individualized Prediction Models for Pain Intensity Outcomes in Patients With Neck Pain, Low Back Pain, or Both in Primary Care Settings.

The purpose of this study was to develop and externally validate multivariable prediction models for future pain intensity outcomes to inform targeted interventions for patients with neck or low back pain in primary care settings. Model development data were obtained from a group of 679 adults with neck or low back pain who consulted a participating United Kingdom general practice. Predictors included self-report items regarding pain severity and impact from the STarT MSK Tool. Pain intensity at 2 and 6months was modeled separately for continuous and dichotomized outcomes using linear and logistic regression, respectively. External validation of all models was conducted in a separate group of 586 patients recruited from a similar population with patients' predictor information collected both at point of consultation and 2 to 4weeks later using self-report questionnaires. Calibration and discrimination of the models were assessed separately using STarT MSK Tool data from both time points to assess differences in predictive performance. Pain intensity and patients reporting their condition would last a long time contributed most to predictions of future pain intensity conditional on other variables. On external validation, models were reasonably well calibrated on average when using tool measurements taken 2 to 4weeks after consultation (calibration slope = 0.848 [95% CI = 0.767 to 0.928] for 2-month pain intensity score), but performance was poor using point-of-consultation tool data (calibration slope for 2-month pain intensity score of 0.650 [95% CI = 0.549 to 0.750]). Model predictive accuracy was good when predictors were measured 2 to 4weeks after primary care consultation, but poor when measured at the point of consultation. Future research will explore whether additional, nonmodifiable predictors improve point-of-consultation predictive performance. External validation demonstrated that these individualized prediction models were not sufficiently accurate to recommend their use in clinical practice. Further research is required to improve performance through inclusion of additional nonmodifiable risk factors.

Comparison between P-POSSUM and NELA risk score for patients undergoing emergency laparotomy in Egyptian patients

Background and aimsThe Portsmouth-Physiological and Operative Severity Score for the enumeration of Mortality and Morbidity (P-POSSUM) is one of the scores that is used most frequently for determining the likelihood of mortality in patients undergoing emergency laparotomy. National Emergency Laparotomy Audit (NELA) presents a novel and validated score. Therefore, we aimed to compare the performance of the NELA and P-POSSUM mortality risk scores in predicting 30-day and 90-day mortality in patients undergoing emergency laparotomy.MethodsBetween August 2020 and October 2022, this cohort study was undertaken at Menoufia University Hospital. We compared the P-POSSUM, preoperative NELA, and postoperative NELA scores in patients undergoing emergency laparotomy. All variables needed to calculate the used scores were collected. The outcomes included the death rates at 30 and 90 days. By calculating the area under the curve (AUC) for every mortality instrument, the discrimination of the various methods was evaluated and compared.ResultsData from 670 patients were included. The observed risk of 30-day and 90-day mortality was 10.3% (69/670) and 13.13% (88/670), respectively. Concerning 30-day mortality, the AUC was 0.774 for the preoperative NELA score, 0.763 for the preoperative P-POSSUM score, and 0.780 for the postoperative NELA score. Regarding 90-day mortality, the AUCs for the preoperative NELA score, preoperative P-POSSUM score, and postoperative NELA score were 0.649 (0.581–0.717), 0.782 (0.737–0.828), and 0.663 (0.608–0.718), respectively. There was noticeable difference in the three models' capacity for discrimination, according to pairwise comparisons.ConclusionsThe probability of 30-day and 90-day death across the entire population was underestimated by the NELA and P-POSSUM scores. There was discernible difference in predictive performance between the two scores.

299 Predictive Performance of NELA Versus P-POSSUM Mortality Scores; Are We Underestimating the Risk of Mortality Following Emergency Laparotomy?

Abstract Aim To compare the performance of NELA and P-POSSUM mortality risk scores in predicting 30-day and 90-day mortality in patients undergoing emergency laparotomy. Method A retrospective cohort study was conducted to compare predicative performance of pre-operative NELA, post-operative NELA and P-POSSUM predicted mortality scores in patients undergoing emergency laparotomy between 2014 and 2021. The outcomes of interest included the observed 30-day and 90-day mortality. The discrimination of the mortality tools was assessed and compared by determining the Area Under the Curve (AUC) for each tool using Receiver Operating Characteristic curve analysis. Results A total of 681 patients were included. The observed risk of 30-day and 90-day mortality were 10.4% (71/681) and 14.2% (97/681), respectively. In terms of 30-day mortality, the AUC was 0.791 (0.727-0.855) for pre-operative NELA score, 0.784 (0.721-0.848) for pre-operative P-POSSUM score, and 0.761 (0.699-0.824) for post-operative NELA score. In terms of 90-day mortality, the AUC was 0.765 (0.708-0.821) for pre-operative NELA score, 0.749 (0.692-0.807) for pre-operative P-POSSUM score, and 0.745 (0.691-0.800) for post-operative NELA score. Observed/Expected ratio for 30-day and 90-day mortality were calculated as 3.25 and 4.43 for pre-operative NELA, 2.81 and 3.84 for pre-operative P-POSSUM, and 2.17 and 2.96 for post-operative NELA, respectively. Pairwise comparisons showed no statistically significant difference in discrimination among the three models. Conclusions Pre-operative NELA, post-operative NELA and P- POSSUM scores underestimated the risk of 30-day and 90-day mortality for patients undergoing emergency laparotomy. No significant difference in predictive performance was found among the three models.

Individualized Dynamic Prediction Model for Patient-Reported Voice Quality in Early-Stage Glottic Cancer.

Early-stage glottic cancer (ESGC) is a malignancy of the head and neck. Besides disease control, preservation and improvement of voice quality are essential. To enable expectation management and well-informed decision-making, patients should be sufficiently counseled with individualized information on expected voice quality. This study aims to develop an individualized dynamic prediction model for patient-reported voice quality. This model should be able to provide individualized predictions at every time point from intake to the end of follow-up. Longitudinal cohort study. Tertiary cancer center. Patients treated for ESGC were included in this study (N = 294). The Voice Handicap Index was obtained prospectively. The framework of mixed and joint models was used. The prognostic factors used are treatment, age, gender, comorbidity, performance score, smoking, T-stage, and involvement of the anterior commissure. The overall performance of these models was assessed during an internal cross-validation procedure and presentation of absolute errors using box plots. The mean age in this cohort was 67 years and 81.3% are male. Patients were treated with transoral CO2 laser microsurgery (57.8%), single vocal cord irradiation up to (24.5), or local radiotherapy (17.5%). The mean follow-up was 43.4 months (SD 21.5). Including more measurements during prediction improves predictive performance. Including more clinical and demographic variables did not provide better predictions. Little differences in predictive performance between models were found. We developed a dynamic individualized prediction model for patient-reported voice quality. This model has the potential to empower patients and professionals in making well-informed decisions and enables tailor-made counseling.

A performance evaluation of drug response prediction models for individual drugs.

Drug response prediction is important to establish personalized medicine for cancer therapy. Model construction for predicting drug response (i.e., cell viability half-maximal inhibitory concentration [IC50]) of an individual drug by inputting pharmacogenomics in disease models remains critical. Machine learning (ML) has been predominantly applied for prediction, despite the advent of deep learning (DL). Moreover, whether DL or traditional ML models are superior for predicting cell viability IC50s has to be established. Herein, we constructed ML and DL drug response prediction models for 24 individual drugs and compared the performance of the models by employing gene expression and mutation profiles of cancer cell lines as input. We observed no significant difference in drug response prediction performance between DL and ML models for 24 drugs [root mean squared error (RMSE) ranging from 0.284 to 3.563 for DL and from 0.274 to 2.697 for ML; R2 ranging from -7.405 to 0.331 for DL and from -8.113 to 0.470 for ML]. Among the 24 individual drugs, the ridge model of panobinostat exhibited the best performance (R2 0.470 and RMSE 0.623). Thus, we selected the ridge model of panobinostat for further application of explainable artificial intelligence (XAI). Using XAI, we further identified important genomic features for panobinostat response prediction in the ridge model, suggesting the genomic features of 22 genes. Based on our findings, results for an individual drug employing both DL and ML models were comparable. Our study confirms the applicability of drug response prediction models for individual drugs.

Gastrointestinal Toxicity Prediction Not Influenced by Rectal Contour or Dose-Volume Histogram Definition

Rectal dose delivered during prostate radiation therapy is associated with gastrointestinal toxicity. Treatment plans are commonly optimized using rectal dose-volume constraints, often whole-rectum relative-volumes (%). We investigated whether improved rectal contouring, use of absolute-volumes (cc), or rectal truncation might improve toxicity prediction. Patients from the CHHiP trial (receiving 74 Gy/37 fractions [Fr] vs 60 Gy/20 Fr vs 57 Gy/19 Fr) were included if radiation therapy plans were available (2350/3216 patients), plus toxicity data for relevant analyses (2170/3216 patients). Whole solid rectum relative-volumes (%) dose-volume-histogram (DVH), as submitted by treating center (original contour), was assumed standard-of-care. Three investigational rectal DVHs were generated: (1) reviewed contour per CHHiP protocol; (2) original contour absolute volumes (cc); and (3) truncated original contour (2 versions; ±0 and ±2 cm from planning target volume [PTV]). Dose levels of interest (V30, 40, 50, 60, 70, 74 Gy) in 74 Gy arm were converted by equivalent-dose-in-2 Gy-Fr (EQD2α/β= 3 Gy) for 60 Gy/57 Gy arms. Bootstrapped logistic models predicting late toxicities (frequency G1+/G2+, bleeding G1+/G2+, proctitis G1+/G2+, sphincter control G1+, stricture/ulcer G1+) were compared by area-undercurve (AUC) between standard of care and the 3 investigational rectal definitions. The alternative dose/volume parameters were compared with the original relative-volume (%) DVH of the whole rectal contour, itself fitted as a weak predictor of toxicity (AUC range, 0.57-0.65 across the 8 toxicity measures). There were no significant differences in toxicity prediction for: (1) original versus reviewed rectal contours (AUCs, 0.57-0.66; P=.21-.98); (2) relative- versus absolute-volumes (AUCs, 0.56-0.63; P=.07-.91); and (3) whole-rectum versus truncation at PTV ± 2 cm (AUCs, 0.57-0.65; P=.05-.99) or PTV ± 0 cm (AUCs, 0.57-0.66; P=.27-.98). We used whole-rectum relative-volume DVH, submitted by the treating center, as the standard-of-care dosimetric predictor for rectal toxicity. There were no statistically significant differences in prediction performance when using central rectal contour review, with the use of absolute-volume dosimetry, or with rectal truncation relative to PTV. Whole-rectum relative-volumes were not improved upon for toxicity prediction and should remain standard-of-care.