Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “predictive modelling”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,423 records · Page 79Linked to original sources
PubMed2005

Influence of internal biogas production on hydrodynamic behavior of anaerobic fluidized-bed reactors.

Predictive models for describing the hydrodynamic behavior (bed-expansion and bed-pressure gradient) of a three-phase anaerobic fluidized bed reactor (AFBR) was developed according to wake theory together with more realistic dynamic bed-expansion experiments (with and without internal biogas production). A reliable correlation equation for the parameter k (mean volume ratio of wakes to bubbles) was also established, which is of help in estimating liquid hold up of fluidized beds. The experimental expansion ratio of three-phase fluidized beds (E(GLS)) was approximately 18% higher than that of two-phase fluidized beds (E(LS)); whereas the experimental bed-pressure gradient of the former [(-DeltaP/H)(GLS)] was approximately 9.3% lower than that of the latter [(-DeltaP/H)(LS)]. Both the experimental and modeling results indicated that a higher superficial gas velocity (u(g)) gave a higher E(GLS) and a higher E(GLS) to E(LS) ratio as well as a lower (-DeltaP/H)(GLS) and a lower (-DeltaP/H)(GLS) to (-DeltaP/H)(LS) ratio. As for the operation stability of the AFBR, the sensitivity of u(g) to expansion height (H(GLS)) and (-DeltaP/H)(GLS) is between the sensitivity of superficial liquid velocity and biofilm thickness. The model predictions of E(GLS), (-DeltaP)(GLS), and (-DeltaP/H)(GLS) agreed well the experimental measurements. Accordingly, the predictive models accounting for internal biogas production described fairly well the hydrodynamic behavior of the AFBR.

Bacteria, Anaerobic↗
PubMed2000

Modelling and prediction of soil sorption coefficients of non-ionic organic pesticides by molecular descriptors.

Soil sorption coefficients (K(OC)) of 185 non-ionic organic heterogeneous pesticides have been studied searching for quantitative structure-property relationships (QSPRs). The chemical description of pesticide structure has been made in terms of some molecular descriptors: count descriptors, topological indices, information indices, fragment-based descriptors and weighted holistic invariant molecular (WHIM) descriptors; these last are statistical indices describing size, shape, symmetry and atom distribution of molecules in the three-dimensional space. Three new topological indices derived from the electrotopological state indices of Kier and Hall were proposed. Multiple linear regression analysis was performed after previous selection of the descriptors mostly correlated to the response by Genetic Algorithms. The obtained results confirm the capability of the proposed approach to give predictive models for one of the most important partition properties, such as soil sorption coefficient (K(OC)).

Algorithms↗
PubMed1997

3D-modelling and prediction by WHIM descriptors. Part 8. Toxicity and physico-chemical properties of environmental priority chemicals by 2D-TI and 3D-WHIM descriptors.

In environmental and ecotoxicological problems, the need for more general predictive models is one of the aims of the science. To reach this objective, the models, even if capable of producing more approximate values, must be applicable to the huge number of chemicals put on the market that are discharged in the environment, at least for setting priority lists among potentially hazardous substances. Theoretically, the problem is not insurmountable if one assumes that all properties of a chemical (physico-chemical, biological, toxicological) are related to its structural features. The challenge is to describe and represent, on a quantitative basis, all molecular structural features in order to use them as independent variables in a quantitative structure-activity model. In this paper, structural, topological, and 3D-WHIM descriptors are used for modelling toxicological and physico-chemical properties of compounds that are included in the European Union's so-called "List 1" of priority chemicals dangerous for the aquatic environment.

Animals↗
PubMed1986

A model for predicting nephrotoxicity in patients treated with aminoglycosides.

We have previously identified risk factors for nephrotoxicity in 204 patients receiving aminoglycosides. We added data on an additional 134 patients from the University of Chicago who were treated for longer periods (13.1 +/- 9.4 days) to determine if duration of therapy was also a risk factor. In the combined populations, nephrotoxicity developed in 59 (17.5%) of 338 patients, and duration of therapy was the factor most strongly associated with nephrotoxicity (P less than .0001). A new predictive model was developed that was then tested prospectively on 175 consecutive medical and surgical patients who had plasma levels of aminoglycoside measured. The new model accurately identified 14 of 15 patients with nephrotoxicity (sensitivity = 93%) and 106 of 160 without nephrotoxicity (specificity = 66%). A bedside scoring system accurately identified high-risk and low-risk patients. We believe these data add to the understanding of the clinical correlates of aminoglycoside nephrotoxicity and aid in the identification of high-risk patients.

Aminoglycosides↗
PubMed2001

Predictive spatial modelling of alternative control strategies for the foot-and-mouth disease epidemic in Great Britain, 2001.

A spatial simulation model of foot-and-mouth disease was used in March and early April 2001 to evaluate alternative control policies for the 2001 epidemic in Great Britain. Control policies were those in operation from March 20, 2001, and comprised a ban on all animal movements from February 23, 2001, and a stamping-out policy. Each simulation commenced with the known population of infected farms on April 10, 2001, and ran for 200 days. For the control policy which best approximated that actually implemented from late March, the model predicted an epidemic of approximately 1800 to 1900 affected farms, and estimated that the epidemic would be eradicated between July and October 2001, with a low probability of continuing beyond October 2001. This policy included the slaughter-out of infected farms within 24 hours, slaughter of about 1.3 of the surrounding farms per infected farm within a further 48 hours, and minimal interfarm movements of susceptible animals. Delays in the slaughter of animals on infected farms beyond 24 hours after diagnosis slightly increased the epidemic size, and failure to achieve pre-emptive slaughter on an adequate number of at-risk farms substantially increased the expected size of the epidemic. Vaccination of up to three of the most outbreak-dense areas carried out in conjunction with the adopted control policy reduced the predicted size of the epidemic by less than 100 farms. Vaccination of buffer zones (designed to apply available vaccine and manpower as effectively as possible) carried out in place of the adopted control policy allowed the disease to spread out of control, producing an epidemic involving over 6000 farms by October 2001, with no prospect of immediate eradication.

Animals↗
PubMed2000

Mid-trimester beta-hCG levels incorporated in a multifactorial model for the prediction of severe pre-eclampsia.

Pre-eclampsia remains a major cause of perinatal morbidity and mortality worldwide. Proposed predicting tests for early detection of pregnant women destined to develop pre-eclampsia remain unsatisfactory. The aim of this study was to investigate the clinical utility of combining mid-trimester maternal serum beta-human chorionic gonadotrophin (MShCG) levels with selected clinical determining factors as a multifactorial predictive test for pre-eclampsia. Thirty-nine cases with mild pre-eclampsia and 56 with severe pre-eclampsia were recruited as the study groups. Normotensive women (957) were enrolled as controls. Potential determining risk factors for severe pre-eclampsia were selected using a multiple logistic regression to build various combined prediction models. A receiver-operator characteristic curve was employed to assess the performance of each prediction test for pre-eclampsia. The prediction efficacy of each test was examined by the area under the curve (AUC). Our data show that mid-trimester MShCG levels significantly correlated with severity of pre-eclampsia (Spearman rank correlation coefficient=0.195, p<0.001). Women with mild pre-eclampsia had a 2.61-times greater chance, while women with severe pre-eclampsia had a 6.13-times greater chance of having MShCG exceeding 2.0 multiples of the median than did women with a normal pregnancy. A combined prediction model composed of MShCG levels, body mass index (BMI), parity, and age as a predictive test for severe pre-eclampsia was superior to MShCG levels alone (AUC 0.765 versus 0.648). The integrated multifactorial model could identify women at risk early on for developing severe pre-eclampsia, with a sensitivity of 70% and a specificity of 71%. Thus, we demonstrate a potentially effective and convenient method by which women at risk for developing severe pre-eclampsia can be identified early, based on a multifactorial predictive model composed of midtrimester MShCG levels, BMI, parity, and age.

Adult↗
PubMed2005

Lung cancer rate predictions using generalized additive models.

Predictions of lung cancer incidence and mortality are necessary for planning public health programs and clinical services. It is proposed that generalized additive models (GAMs) are practical for cancer rate prediction. Smooth equivalents for classical age-period, age-cohort, and age-period-cohort models are available using one-dimensional smoothing splines. We also propose using two-dimensional smoothing splines for age and period. Variance estimation can be based on the bootstrap. To assess predictive performance, we compared the models with a Bayesian age-period-cohort model. Model comparison used cross-validation and measures of predictive performance for recent predictions. The models were applied to data from the World Health Organization Mortality Database for females in five countries. Model choice between the age-period-cohort models and the two-dimensional models was equivocal with respect to cross-validation, while the two-dimensional GAMs had very good predictive performance. The Bayesian model performed poorly due to imprecise predictions and the assumption of linearity outside of observed data. In summary, the two-dimensional GAM performed well. The GAMs make the important prediction that female lung cancer rates in these countries will be stable or begin to decline in the future.

Adult↗
PubMed1999

Dichotomization of continuous measurements using generalized additive modelling--application in predicting intrapartum caesarean delivery.

In prediction model development, continuous variables are often dichotomized at empirically chosen thresholds to simplify calculations and facilitate decision making. However, these choices are often made in the absence of estimated covariate effects and may be inaccurate, thus weakening the models. To improve this approach, generalized additive modelling that allows non-parametric estimation of the true covariate effects is used for threshold selection. In this study, this approach is illustrated by development of prediction models for intrapartum Caesarean deliveries. The prediction performance of the models thus developed is significantly better than that developed using empirically chosen thresholds for dichotomization.

Adolescent↗
PubMed2003

Thermal modeling and performance analysis of a thermoacoustic refrigerator.

A heat-driven thermoacoustic refrigerator has been designed and tested. A detailed thermal model of the device is presented. Energy balances within the system are discussed using external, heat exchanger, and stack control volumes in order to clarify the relationships of work and heat fluxes below and above onset. Thermal modeling is discussed as a tool for performance analysis as well as for determining system heat losses and finding input heat flows required by a thermoacoustic code. A method of using the control volume balance equations to find stack work and device efficiencies is presented. Experimental measurements are compared to DELTAE thermoacoustic modeling predictions. Modeling results show that viscous losses within the system have a significant impact on the device performance as well as on the ability of DELTAE to accurately predict performance. Modeling has led to an understanding of system performance and highlighted loss sources that are areas for improvement in a redesign.

Acoustics↗
PubMed2001

Pediatric emergency assessment tool (PEAT): a risk-adjustment measure for pediatric emergency patients.

OBJECTIVE: To develop a multivariable model predicting the level of care required by pediatric patients for use as a risk-adjustment tool in the evaluation of emergency medical services for children. METHODS: A random 10% sample of records of all visits over a 12-month period to a suburban, university-affiliated pediatric emergency department (PED) was selected and abstracted. The outcome variable, level of care received, was categorized in three levels: routine care only (R); diagnostic or therapeutic procedures performed in the ED but patient not admitted (EDT); and admission to hospital (ADM). Predictor variables included information routinely elicited and recorded at the time of triage. Using multinomial logistic regression, a predictive model was derived from a subset of 70% of the selected visits, and was validated in the remaining 30%. RESULTS: The total sample included 2,287 visits. The overall rate of each outcome was R-37%, EDT-53%, and ADM-10%. The final regression model included the following predictors significantly associated with the outcome: age, past medical history, temperature, abnormal respiratory rate or pulse oximetry in triage, chief complaint, and triage level (model likelihood ratio chi-square, 14 df = 332, p < 0.00001, R(2) = 0.14). The number of outcomes was well predicted by the model in both subsamples. Analysis of variance showed a significant association between Pediatric Emergency Assessment Tool (PEAT) score (weighted sum of the predicted probabilities of EDT and ADM) and both ED charges and time spent in the ED (p < 0.001). CONCLUSIONS: A model based on easily and routinely measured variables can accurately predict the level of care rendered in the PED. The predicted probabilities from such a model correlate well with other outcomes of care and may be useful in adjusting for differences in risk when evaluating quality of care.

Child, Preschool↗
PubMed1998

[Mathematical model for the predictive value of a test in critically ill patients studies according to APACHE II score and pathology at admission].

OBJECTIVE: To find a predictive model for mortality at four different days from the admission for critically ill patients. DESIGN: Retrospective study on two consecutive series of critically ill patients admitted in ICU. SUBJECTS: 1254 critically ill patients, subdivided into two series of 813 (561 survivors and 252 non survivors) and 441 patients (291 survivors and 150 non survivors), respectively. INTERVENTIONS: None. MEASUREMENTS: All patients had APACHE II calculated within the first 24 hours from the admission in ICU and, if the patient was still in ICU, also at the 5th, 10th and 15th day from the admission. Casistics was subdivided into two unequal series, ratio 2:1, with a random selection made on each of the 6 considered years. On the 1st series, in 1st, 5th, 10th and 15th day, for mathematical predictive models were made, using stepwise logistic regression (BMDP, Los Angeles). In the 1st day the following independent variables were utilized: APACHE II score, the specific diagnosis at admission, fitted following Knaus' diagnostic criteria, united in 6 principal categories, while for the other 3 days the variation % of APACE II score as regards the previous day. RESULTS: For each of the considered day four mathematical models have been made. These models have been validated in both series in calibration from the Hosmer-Lemeshow Goodness-of-fit test and in discrimination from the ROC curves. For each day Y (Prob.% to die) = eLogit/1 + eLogit, where Logit = beta 0 (constant) + beta 1*APACHE II + beta 2*Variat.%APACE II (difference between actual APACHE II - APACHE II of the previous day/actual APACHE II) + beta k, (coefficient pertinent to pathology). CONCLUSIONS: The mathematical model, as other models do, stratifies enough the casistics according to the risk of death. Waiting for further studies to make more precise prognostic mathematical models, this one and others can help the clinical assessment in single patient evaluation.

APACHE↗
PubMed2006

ROC curves predicted by a model of visual search.

In imaging tasks where the observer is uncertain whether lesions are present, and where they could be present, the image is searched for lesions. In the free-response paradigm, which closely reflects this task, the observer provides data in the form of a variable number of mark-rating pairs per image. In a companion paper a statistical model of visual search has been proposed that has parameters characterizing the perceived lesion signal-to-noise ratio, the ability of the observer to avoid marking non-lesion locations, and the ability of the observer to find lesions. The aim of this work is to relate the search model parameters to receiver operating characteristic (ROC) curves that would result if the observer reported the rating of the most suspicious finding on an image as the overall rating. Also presented are the probability density functions (pdfs) of the underlying latent decision variables corresponding to the highest rating for normal and abnormal images. The search-model-predicted ROC curves are 'proper' in the sense of never crossing the chance diagonal and the slope is monotonically changing. They also have the interesting property of not allowing the observer to move the operating point continuously from the origin to (1, 1). For certain choices of parameters the operating points are predicted to be clustered near the initial steep region of the curve, as has been observed by other investigators. The pdfs are non-Gaussians, markedly so for the abnormal images and for certain choices of parameter values, and provide an explanation for the well-known observation that experimental ROC data generally imply a wider pdf for abnormal images than for normal images. Some features of search-model-predicted ROC curves and pdfs resemble those predicted by the contaminated binormal model, but there are significant differences. The search model appears to provide physical explanations for several aspects of experimental ROC curves.

Algorithms↗
PubMed1999

A novel mechanism of anode-break stimulation predicted by bidomain modeling.

-Anodal stimulation by external pacemakers has been explained on the basis of bidomain models of cardiac tissue. Bidomain models predict that anodal stimuli will hyperpolarize the underlying tissue while adjacent regions become depolarized (virtual cathodes), initiating excitation. We investigated the contribution of active cellular properties to anode-break stimulation. A bidomain model was implemented in which each cell contained realistic ionic currents, including those recruited by hyperpolarization. Simulations reveal that anode-break excitation can originate at the site of stimulation itself and not only from adjacent regions of induced depolarization. The threshold for initiating excitation at the site of stimulation is lower than that for stimulation initiating from adjacent depolarized regions. Thus, incorporation of active cellular properties into a bidomain model predicts a novel mechanism for anode-break stimulation of the heart. The results will improve our understanding of anodal pacing and its risks and benefits in patients.

Action Potentials↗
PubMed2004

Investigation of biomechanical factors affecting rowing performance.

It was hypothesized that a crew's rowing performance was predictable based on their total propulsive power, synchrony (a real-time comparison of rower propulsive force magnitudes) and total drag contribution (a measure of the rowers' effect on shell drag forces during the recovery), quantities calculated from individual rower's force-time profiles and recovery kinematics. A rowing pair was equipped with transducers to gather shell velocity, propulsive blade force, oar angular position and seat displacement. Eight subjects (four port, four starboard) participated in two rounds of data collection. The first round pairings were random, while the second round pairings were assigned based on Round 1 results. Regression analysis and ANCOVA were used to test the validity of assumptions inherent in the predictive model and, if applicable, explore a linear model predicting rowing performance based on total propulsive power, synchrony and total drag contribution. Total propulsive power, synchrony and total drag contribution were correlated and further were affected by pairing, violating assumptions inherent in the linear model. The original hypothesis was not supported based on these violations. Important findings include (1) performance cannot be predicted using the simple linear model proposed, (2) rowers' force-time profiles are repeatable between trials, with some but not all rowers adapting their force-time profile dependent on their pair partner, presumably in an effort to increase the level of synchrony between the two, and (3) subtle biomechanical factors may play a critical role in performance.

Adult↗
PubMedDate not supplied

Predicting EuroQoL EQ-5D preference scores from the SF-12 Health Survey in a nationally representative sample.

PURPOSE: To predict the EuroQoL EQ-5D utility index from the SF-12 Health Survey for a US national sample of adults. METHODS: The authors used the 2000 Medical Expenditure Panel Survey to examine the relationship between instruments. Linear regression was used to predict EQ-5D scores from Physical Component Summary (PCS) and Mental Component Summary (MCS) scores of the SF-12. A prediction model was derived in one half of the sample and validated in the other half. RESULTS: Complete responses to both measures were available for 14,580 adults; 7313 (50.2%) surveys were used for the derivation set. The 2-variable model predicted 61% of the variance in EQ-5D scores and provided reasonable ability to predict mean EQ-5D scores from mean PCS and MCS scores. Confidence intervals are dependent on sample size and variance of PCS and MCS scores. CONCLUSIONS: EQ-5D scores can be reasonably predicted from the SF-12. This model allows researchers to estimate utility data for use in decision and cost-utility analyses.

Adolescent↗
PubMedDate not supplied

Creatinine clearance in continuous peritoneal dialysis: dialysis dose required for a minimal acceptable level.

OBJECTIVES: To identify the most advantageous formula for estimating creatinine clearance (CCr) and to establish a dose of dialysis that will ensure minimal acceptable levels of creatinine clearance in patients on continuous peritoneal dialysis (CPD). DESIGN: Analysis of all CCr studies performed in CPD patients over 40 months. SETTING: All four dialysis units following CPD patients in one city. One dialysis unit is government-owned, one is university-affiliated, and two are community based. PARTICIPANTS: One hundred and ninety-four patients representing almost the entire CPD population in Albuquerque. INTERVENTIONS: Creatinine and urea clearance studies were performed in 24-hour urine and drained dialysate samples. Creatinine clearance (peritoneal plus urinary) was normalized to either 1.73 m2 body surface area (CCr) or body water estimated by the Watson formulas (KT/VCr). CCr and KT/VCr were either corrected by averaging urinary creatinine and urea clearances or were not corrected. Two dialysis units were designated as the training set (92 patients, 143 clearance studies) and the other two units as the validation set (102 patients, 181 clearance studies). MAIN OUTCOME MEASURES: Minimal acceptable creatinine clearance levels were determined in the training set by computing the creatinine clearance value corresponding to 1.70 weekly KT/V urea by linear regression. Logistic regression models predicting low creatinine clearance were developed in the training set and were tested in the validation set. RESULTS: The following weekly creatinine clearance values corresponded to 1.70 KT/V urea: corrected CCr 52.0 L/1.73 m2, uncorrected CCr 54.4 L/1.73 m2, corrected KT/VCr 1.46, uncorrected KT/VCr 1.53. Logistic regression identified as predictors of low creatinine clearance low daily urine volume (UV) and low daily dialysate drain volume/body water (DV/V) for all four creatinine clearance formulas, plus low/low-average peritoneal solute transport (only for uncorrected CCr) and serum creatinine (for both KT/VCr formulas). In the validation set, the predictive models produced an area under the receiver operating characteristic (ROC) curve between 0.835 and 0.919 indicating very good predictive accuracy. For corrected CCr and anuria, the regression model produced a minimal normalized drain volume (DV/V) value consistent with minimal acceptable CCr equal to 0.305 L/L per 24 hours. This DV/V cutoff detected low corrected CCR in validation set anuric subjects (n = 55) with a sensitivity of 85% and a specificity of 71%. For uncorrected CCR and anuria, DV/V cutoffs were 0.273 L/L per 24 hours (high/high-average peritoneal solute transport) and 0.420 L/L per 24 hours (low/low-average transport). Sensitivity and specificity of these cutoffs in validation set anuric subjects were 87% and 85%, plus 86% and 33%, respectively. CONCLUSIONS: The uncorrected CCr appears to be the most advantageous creatinine clearance formula in CPD, because it allows the use of peritoneal solute transport type in the calculation of the minimal required normalized drain volume. The minimal acceptable uncorrected CCr is 54.4 L/1.73 m2 weekly. To achieve this uncorrected CCr in anuria, the required minimal normalized drain volume is 0.273 L per liter of body water daily if peritoneal solute transport is high or high-average and around 0.420 L per liter of body water daily if peritoneal solute transport is low or low-average. The required total daily drain volume is computed by multiplying the required normalized drain volume by body water.

Biological Transport↗
PubMed1999

Using computational auditory models to predict simultaneous masking data: model comparison.

In order to develop improved remediation techniques for hearing impairment, auditory researchers must gain a greater understanding of the relation between the psychophysics of hearing and the underlying physiology. One approach to studying the auditory system has been to design computational auditory models that predict neurophysiological data such as neural firing rates [15], [1]. To link these physiologically-based models to psychophysics, theoretical bounds on detection performance have been derived using signal detection theory to analyze the simulated data for various psychophysical tasks [20]. Previous efforts, including our own recent work using the Auditory Image Model, have demonstrated the validity of this type of analysis; however, theoretical predictions often continue to exceed experimentally-measured performance [9], [21]. In this paper, we compare predictions of detection performance across several computational auditory models. We also reconcile some of the previously observed discrepancies by incorporating appropriate signal uncertainty into the optimal detector.

Computer Simulation↗
PubMed1996

Using a mailed survey to predict hospital admission among patients older than 80.

OBJECTIVE: To determine if the results of a questionnaire mailed to older patients can help identify those patients at greatest risk of hospital admission. DESIGN: A longitudinal cohort study. SETTING: A prepaid managed care plan in the Denver metropolitan area. PARTICIPANTS: Of the 4414 eligible patients at least 81 years old, 3745 (84.8%) responded. MEASUREMENTS: We studied the predictive power of self-reported demographic, health status, medical history, health habits, functional status (including Katz' activities of daily living and OARS instrumental activities of daily living), and socioeconomic status data to identify those older adults at greatest risk of hospitalization within 4.5 months of completing the survey. We derived our predictive model on one-half the subjects and tested its validity it on the other half. RESULTS: Univariate analysis revealed 25 variables significantly associated with hospital admission. In a logistic regression model, four significant variables successfully stratified the patients by risk of admission. These four variables are: the presence of heart disease, the presence of diabetes, need for help preparing meals, and limited physical independence (requiring the help of a person or mechanical aid to get around). In addition, there was an antagonistic interaction between the presence of heart disease and limited physical independence. The model stratified patients from low risk (4.5% chance of admission) to high risk (39% chance of admission). As measured by the Hosmer-Lemeshow statistic and the area under a receiver-operator characteristic (ROC) curve, this model fit both the derivation and validation subjects well. CONCLUSIONS: A mailed questionnaire achieved a high response rate, and the information collected produced an effective model predictive of hospitalization in the short term. Four easily ascertained pieces of information identify those patients older than age 81 at increased risk.

Activities of Daily Living↗