SEARCH · PubMed Health

Results for “predictive modelling”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

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

At least 199 records · Page 11Linked to original sources

PubMed2006

A preoperative and intraoperative predictive model of prolonged intensive care unit stay for valvular surgery.

BACKGROUND AND AIM OF THE STUDY: In developing countries, the costs of intensive care unit (ICU) stay are very high for patients after valve surgery. In addition, patients with a prolonged ICU stay have a poor prognosis compared to those with a short ICU stay. The study aim was to develop a specific risk model and to use a logistic EuroSCORE model to predict prolonged ICU stay after valve surgery. METHODS: A total of 507 consecutive patients undergoing valve surgery were studied using univariate and multivariate analyses. Prolonged ICU stay was defined as five days or more. Stepwise logistic regression analysis was used to identify the risk factors for prolonged ICU stay. These variables were then used to calculate a prognostic score (S) and a predicted probability (P) for prolonged ICU stay. A receiver operating characteristic (ROC) curve was calculated to measure the prognostic value of the new risk model and logistic EuroSCORE model. Sensitivity and specificity analysis were used for evaluation. RESULTS: Multivariate logistic regression analysis showed that age > or = 65 years, left ventricular ejection fraction (LVEF) < or = 50%, cardiothoracic ratio (CTR) > or = 0.68, previous cardiac surgery, maximal voluntary ventilation (MVV) observed/predicted < 71% and repeat cardiopulmonary bypass (CPB) during surgery were risk factors. Mitral valve surgery reduced the risk of prolonged ICU stay. Observed probabilities compared well with predicted probabilities. The ROC curve produced an area under the curve (AUC) value of 0.81 for prolonged ICU stay. Based on predicted probability, patients were classified as low-risk (0 < or = P < 10%), intermediate-risk (10% < or = P < 20%), high-risk (20% < or = P < 40%) and very high-risk (> or = 40%) groups. A P-value > or = 40% was used as a cut-off point for the prognostic test. The specificity of this test was 97%, sensitivity 32%, positive predictive value 62%, negative predictive value 89%, positive likelihood ratio 10.67, and negative likelihood ratio 0.70. The ROC curve of a logistic EuroSCORE model gave an AUC value of 0.66 for prolonged ICU stay. CONCLUSION: The study results showed that individual patients undergoing valve surgery could be stratified according to their risk factors for prolonged ICU stay. High-risk patients may require more careful preoperative and postoperative management to reduce postoperative mortality, morbidity, the length of ICU stay, and therefore the cost of valve surgery.

Cardiac Surgical Procedures↗

PubMed2003

External validity of predictive models: a comparison of logistic regression, classification trees, and neural networks.

BACKGROUND AND OBJECTIVE: The utility of predictive models depends on their external validity, that is, their ability to maintain accuracy when applied to patients and settings different from those on which the models were developed. We report a simulation study that compared the external validity of standard logistic regression (LR1), logistic regression with piecewise-linear and quadratic terms (LR2), classification trees, and neural networks (NNETs). METHODS: We developed predictive models on data simulated from a specified population and on data from perturbed forms of the population not representative of the original distribution. All models were tested on new data generated from the population. RESULTS: The performance of LR2 was superior to that of the other model types when the models were developed on data sampled from the population (mean receiver operating characteristic [ROC] areas 0.769, 0.741, 0.724, and 0.682, for LR2, LR1, NNETs, and trees, respectively) and when they were developed on nonrepresentative data (mean ROC areas 0.734, 0.713, 0.703, and 0.667). However, when the models developed using nonrepresentative data were compared with models developed from data sampled from the population, LR2 had the greatest loss in performance. CONCLUSION: Our results highlight the necessity of external validation to test the transportability of predictive models.

Classification↗

PubMed1997

A predictive model for area under the concentration versus time curve of cyclosporin A using several routine monitoring results in renal transplant patients.

We created a predictive model for the area under the concentration versus time curve (AUC) of cyclosporin A (CsA) using routine monitoring results, and examined its clinical utility. Based on 48 clinical time courses accumulated from renal transplant patients, the AUC predictive model was created. An estimate of the AUC0-8 (integrated from time zero to 8 h) was then given as follows: AUC0-8 = 5673.1 x log(TL) + 9342.8 x log(OB) + 64.1 x Dprd x 869.4 x DTK - 168.9 x HCT - 161.2 x SCr - 11.3 x GPT + 3.0 x PL - 588.6 x SEX - 24794.5. In this model, the AUC0-8(ng.h/ml) is given as a function of the CsA through levels (TL, ng/ml), obesity (OB, %), daily dose of prednisolone (Dprd, mg/d), donor type of kidney (DTK), hematocrit (HCT, %), serum creatinine (SCr, mg/dl), glutamate-pyruvate transaminase activity (GPT, IU/l), plasma lipids (PL, mg/dl) and sex distinction (SEX). The Statistical significance of the multiple regression was p < 0.00001 (R2 = 0.862, n = 48), and the day after transplantation, neither the administered oral dose of CsA, or the patient's age had any contribution to the regression. The predictive performance of this model was almost equal to that of the existing method which used 3-point data on the concentration versus time curve. In clinical adaptation for renal transplant patients, the steady-state concentration of CsA (Css) based on the AUC0-8 predictive model was significantly decreased during acute gastroenteritis or before acute rejection, whereas nephrotoxicity was increased, even though CsA trough levels were within a normal therapeutic range (100-200 ng/ml). These findings suggest that the created AUC0-8 predictive model using routine monitoring results, i.e., the trough level of CsA, biochemical tests, a daily dose of predorinsolone (PRD), and basic patient information, is convenient as a monitoring device for CsA therapy, and is satisfactory in clinical practice.

Adult↗

PubMed1993

Variability of physiologically based pharmacokinetic (PBPK) model parameters and their effects on PBPK model predictions in a risk assessment for perchloroethylene (PCE).

When used in the risk assessment process, the output from physiologically based pharmacokinetic (PBPK) models has usually been considered as an exact estimate of dose, ignoring uncertainties in the parameter values used in the model and their impact on model predictions. We have collected experimental data on the variability of key parameters in a PBPK model for tetrachloroethylene (PCE) and have used Monte Carlo analysis to estimate the resulting variability in the model predictions. Blood/air and tissue/blood partition coefficients and the interanimal variability of these data were determined for tetrachloroethylene (PCE). The mean values and variability for these and other published model parameters were incorporated into a PBPK model for PCE and a Monte Carlo analysis (n = 600) was performed to determine the effect on model predicted dose surrogates for a PCE risk assessment. For a typical dose surrogate, area under the blood time curve for metabolite in the liver (AUCLM), the coefficient of variation was 25% and the mean value for AUCLM was within a factor of two of the maximum and minimum values generated in the 600 simulations. These calculations demonstrate that parameter uncertainty is not a significant potential source of variability in the use of PBPK models in risk assessment. However, we did not in this study consider uncertainties as to metabolic pathways, mechanism of carcinogenicity, or appropriateness of dose surrogates.

Administration, Inhalation↗

PubMed1994

Assessment of predictive models for binary outcomes: an empirical approach using operative death from cardiac surgery.

Predictive models in medical research have gained popularity among physicians as an important tool in medical decision making. Eight methodological strategies for creating predictive models are compared in a large, complex data base consisting of preoperative risk and operative outcome data on 12,712 patients undergoing coronary artery bypass grafting and entered into the Department of Veterans Affairs Cardiac Surgery Risk Assessment Program between April 1987 and March 1990. The models under consideration were developed to predict operative death (any death within 30 days following the surgical procedure or later if the result of a perioperative complication). The two strategies with the best predictive power among the eight examined were stepwise logistic regression alone and data reduction by cluster analysis combined with clinical judgement followed by a logistic regression model. The additive model based on unadjusted relative risks, the model based on Bayes' Theorem, and the logistic model using all candidate variables were good alternatives. Whether or not we imputed values did not have a significant impact on the predictive power of the models.

Aged↗

PubMed1984

Model prediction of plasma volume change induced by hemodialysis.

We simulated the change in plasma volume by hemodialysis by combining two models for transcapillary and transcellular fluid exchange. Model predictions of the plasma volume change were found to be in good agreement with the measured values in five patients who were studied during 4-hour hemodialysis at a constant ultrafiltration rate of 0.5 L/hr using three different dialysate sodium concentrations: 7% below, 7% above, and equal to the predialytic serum concentration. The measurements and model predictions indicate that the decrease in plasma volume is smaller in high sodium dialysis than in normal and low sodium dialyses. Because model predictions were consistent with measurements, this model should be useful in clinical practice to quantitatively analyze the change in plasma volume during hemodialysis and to relate it with hypovolemic hypotension.

Blood Proteins↗

PubMed1997

Variables that influence response to different interferon schedules in chronic hepatitis C and predictive models.

In chronic hepatitis C (HCV), standard interferon therapy with 3 MU three times weekly for 6 months is associated with sustained response in about 10-20% of patients while another 10-15% respond only when higher dosages or/and longer periods of treatment are used. Different variables have been described that are associated with sustained response and may also identify patients requiring low- or high-dose regimens. We have analysed a large data base of 442 patients with chronic hepatitis C treated with interferon-alpha to define rates of sustained response in different patient subgroups treated with different schedules. The rate of sustained response was increased with higher dose regimens in most patient categories, defined according to age, pre-treatment liver histology and HCV genotype, while the amount of interferon per one sustained response remained the same or was reduced. The use of higher dose regimen was particularly cost-effective in patients with cirrhosis. Using the same data base, different models of prediction of sustained response in the individual patient were developed and compared. Inclusion of the HCV genotype in these models was found to increase significantly specificity and sensitivity, confirming that this parameter has a major influence on sustained response to interferon therapy in chronic HCV.

Adolescent↗

PubMed2006

[Study on the predictive model for Yersinia enterocolitica growth at different temperatures].

OBJECTIVE: A predictive model for Yersinia enterocolitica growth in MTSB at different temperatures was studied. According to this model, if food has been contaminated by a few Yersinia enterocolitica, the growth rate, generation time, lag time and total bacterial number at any time and any effective temperature can be predicted. Under the same condition, the parameters for Yersinia enterocolitica serotype O:3 and Yersinia enterocolitica O:9 were compared, respectively. RESULTS: 1. The minimum growth temperature for Yersinia enterocolitica serotype O:3 and O:9 were--17.8 degrees C and--10.3 degrees C, respectively. 2. In the range of 4 degrees C and 30 degrees C, the growth rate of Yersinia enterocolitica serotype O:3 or O:9 increase with temperature going up. 3. In the range of 4 degrees C and 15.7 degrees C, the growth rate of Yersinia enterocolitica serotype O:3 is higher than that of Yersinia enterocolitica serotype O:9. In the range of 15.7 degrees C and 30 degrees C, the growth rate of Yersinia enterocolitica serotype O:9 is higher than that of Yersinia enterocolitica serotype O:3. 4. We get a predictive model for Yersinia enterocolitica in MTSB: Yersinia enterocolitica serotype O:3, square root of M = 0.0134 (T-255.1866). Yersinia Enterocolitica serotype O:9, square root of M = 0.0173 (T-262.7457).

Food Microbiology↗

PubMed2004

A predictive model for heat inactivation of Listeria monocytogenes biofilm on stainless steel.

Heat treatment of potential biofilm-forming sites is sometimes used for control of Listeria monocytogenes in food processing plants. However, little information is available on the heat treatment required to kill L. monocytogenes present in biofilms. The purpose of this study was to develop a predictive model for the heat inactivation of L. monocytogenes in monoculture biofilms (strains Scott A and 3990) and in biofilms with competing bacteria (Pseudomonas sp. and Pantoea agglomerans) formed on stainless steel in the presence of food-derived soil. Biofilms were produced on stainless steel coupons with diluted tryptic soy broth incubated for 48 h at 25 degrees C. Duplicate biofilm samples were heat treated for 1, 3, 5, and 15 min at 70, 72, 75, 77, and 80 degrees C and tested for survivors using enrichment culture. The experiment was repeated six times. A predictive model was developed using logistic regression analysis of the fraction negative data. Plots showing the probability of L. monocytogenes inactivation in biofilms after heat treatment were generated from the predictive equation. The predictive model revealed that hot water sanitation of stainless steel can be effective for inactivating L. monocytogenes in a biofilm on stainless steel if time and temperature are controlled. For example, to obtain a 75% probability of total inactivation of L. monocytogenes 3990 biofilm, a heat treatment of 80 degrees C for 11.7 min is required. The model provides processors with a risk management tool that provides predicted probabilities of L. monocytogenes inactivation and allows a choice of three heat resistance assumptions. The predictive model was validated using a five-strain cocktail of L. monocytogenes in the presence of food soil.

Biofilms↗

PubMed2003

Equilibrium temperature in aerated basins--comparison of two prediction models.

This note presents and compares two models to predict the equilibrium temperature in aerated basins. They differ by their degree of complexity and therefore by the input data they require. Both models were able to estimate the temperature of an industrial aerated lagoon, the more complex model giving, in addition, a complete breakdown of the heat exchanges.

Bioreactors↗

PubMed2006

Total body protein mass: validation of total body potassium prediction model in children and adolescents.

Protein is an important body component for monitoring growth, development, and nutritional status. We previously developed a total body potassium (TBK, in mmol) and bone mineral (Mo, in kg) model for predicting total body protein (TBPro, in kg) in adults (TBPro = 0.00252 x TBK + 0.732 x Mo). However, the applicability of the TBK-Mo model for children is unknown. The study aims were to develop a TBK-independent 6-component (6-C) TBPro approach as the criterion, and then to validate the TBK-Mo model in children. The following measurements were made in adolescents and children (n = 62, 38 boys and 24 girls, aged 5-17 y): body weight (BW, in kg), body volume (BV, in liters) by air displacement plethysmography, total body water (TBW, in kg) by 2H2O dilution, Mo by dual-energy X-ray absorptiometry, and TBK by whole-body counting. A 6-C model was derived as TBPro = 2.922 x BW - 0.301 x TBW - 2.039 x Mo - 2.632 x BV. The TBPro estimates did not differ between the 6-C and TBK-Mo models (mean +/- SD, 0.20 +/- 0.86 kg). There was a significant correlation between TBPro by the 6-C and TBK-Mo models (r = 0.94, P < 0.001). Bland-Altman analysis indicated that the differences between TBPro by 6-C and TBK-Mo models were not significantly correlated with the mean TBPro estimates by the 2 models (r = 0.032, P > 0.05). The TBK-Mo model can thus be used to estimate TBPro in healthy adults, adolescents, and children > 5 y old.

Absorptiometry, Photon↗

PubMed1998

[A quantitative analysis of a predictive model of ambulatory blood pressure monitoring integrating physical activity recording].

OBJECTIVE: To determine how much of the variations of blood pressure during a 24 hour period could be accounted for by a change in activity using an accelerometer to detect the physical activity and establish a predictive model. MATERIALS AND METHODS: 18 healthy subjects (mean age 25 +/- 2 yrs) were studied during daily life (24 hours) twice one week apart. The systolic and diastolic blood pressure, heart rate (HR), and time of measure were recorded by ambulatory monitoring using Spacelabs (4 measures per hour). A portable digital memory device was designed for the 24 hours ambulatory monitoring of HR (ECG) and physical activity. This device consists of an ECG Holter (ELA medical model Cinesis with digital memory) and a three piezoresistive type accelerometer sensors (prototype ELA research) able to record physical activity in the 3 space dimension. RESULTS: The data of the first recording were compared to the predicated values from the application of a logarithmic model of activity to the second recording. The model then predicted 53 +/- 19% of the systolic BP values of the test day. The mean individual difference for a given time period of one hour between the measured and the predicted systolic BP from the model was 1.45 +/- 3.1 mmHg with a range of [-6.9; 3.4 mmHg]. The mean individual systolic BP difference for the same given time period of one hour but without predictive model was 1.29 +/- 10 mmHg with a range of [-28; 43 mmHg]. CONCLUSION: This study show that 3 D accelerometer is an easy tool to program individual model of ambulatory blood pressure variability. The introduction of this qualitative method seems logical in therapeutic trial.

Adult↗

PubMed1988

The use of prediction models for eliminating effects due to regression-to-the-mean in road accident data.

In recent years, various methods have been proposed for estimating the true accident level, i.e. the expected number of accidents m when a total of x accidents have been observed at a junction, road section, etc., during a certain period of time. One such method has been named the Empirical Bayes Method (EB method). A description is given of a variant of the EB method utilizing prediction models for the number of accidents. Input data to the prediction models may consist, for example, of traffic flows in a junction. According to empirical comparisons of accidents in junctions, this variant of the EB method may be preferable in certain cases to the conventional EB method. However, it has not yet been determined how this variant of the EB method should generally take into account the precision of the prediction models. This means, for example, that in a nonexperimental before-and-after study of the effect of a particular action, varying results may be obtained according to the assumptions made concerning the precision of the prediction model.

Accident Prevention↗

PubMed1988

Predictive model of conjugative plasmid transfer in the rhizosphere and phyllosphere.

A computer simulation model was used to predict the dynamics of survival and conjugation of Pseudomonas cepacia (carrying the transmissible recombinant plasmid R388:Tn1721) with a nonrecombinant recipient strain in simple rhizosphere and phyllosphere microcosms. Plasmid transfer rates were derived for a mass action model, and donor and recipient survival were modeled as exponential growth and decay processes or both. Rate parameters were derived from laboratory studies in which donor and recipient strains were incubated in test tubes with a peat-vermiculite solution or on excised radish or bean leaves in petri dishes. The model predicted donor, recipient, and transconjugant populations in hourly time steps. It was tested in a microcosm planted with radish seeds and inoculated with donor and recipient strains and on leaf surfaces of radish and bean plants also growing in microcosms. Bacteria were periodically enumerated on selective media over 7 to 14 days. When donor and recipient populations were 10(6) to 10(8) CFU/g (wet weight) of plant or soil, transconjugant populations of about 10(1) to 10(4) were observed after 1 day. An initial rapid increase and a subsequent decline in numbers of transconjugants in the rhizosphere and on leaf surfaces were correctly predicted.

Computer Simulation↗

PubMed1995

HLA prediction model for extended family matches.

This article derives a probabilistic model for predicting HLA matches from a population of extended family members. (Extended family members are defined as either first cousins or blood-related aunts or uncles). The model uses family pedigree information and haplotype frequency data to estimate the likelihood of a match. Results are given for many ethnic groupings. A case study is also described. This technique is most applicable when all family members are of the same ethnic origin, which increases the likelihood of a match among a small number of family members, and the patient possesses a haplotype having a frequency exceeding 10%, within the patient's ethnic population. Under such conditions, an extended family search can frequently approximate the success rate of a single sibling search.

Adult↗

PubMed2004

Atmospheric optical turbulence over land in middle east coastal environments: prediction modeling and measurements.

Beam intensity scintillations, characterized by a refractive-index structure parameter and caused by variations of macrometeorological features of the coastal atmosphere such as air temperature, wind speed and direction, and relative humidity, are examined theoretically and experimentally. In our theoretical analysis we present two well-known models considered separately for over-water and over-land atmospheric optical communication or imaging channels. By means of comparison with our experiments carried out in midland coastal environments in southern and northern Israel, we show the limitations of the models to predict the refractive-index structure Cn2 parameter for both daytime and nighttime turbulent atmospheres in different coastal zone meteorological conditions. We also present an extension of an existing model with two different practical applications that, as is shown experimentally, can be a good predictor of Cn2 for optical atmospheric paths over midland coastal zones.

Journal Article↗

PubMed2002

Cross-institutional evaluation of BI-RADS predictive model for mammographic diagnosis of breast cancer.

OBJECTIVE: Given a predictive model for identifying very likely benign breast lesions on the basis of Breast Imaging Reporting and Data System (BI-RADS) mammographic findings, this study evaluated the model's ability to generalize to a patient data set from a different institution. MATERIALS AND METHODS: The artificial neural network model underwent three trials: it was optimized over 500 biopsy-proven lesions from Duke University Medical Center or "Duke," evaluated on 1,000 similar cases from the University of Pennsylvania Health System or "Penn," and reoptimized for Penn. RESULTS: Trial A's Duke-only model yielded 98% sensitivity, 36% specificity, area index (A(z)) of 0.86, and partial A(z) of 0.51. The cross-institutional trial B yielded 96% sensitivity, 28% specificity, A(z) of 0.79, and partial A(z) of 0.28. The decreases were significant for both A(z) (p = 0.017) and partial A(z) (p < 0.001). In trial C, the model reoptimized for the Penn data yielded 96% sensitivity, 35% specificity, A(z) of 0.83, and partial A(z) of 0.32. There were no significant differences compared with trial B for specificity (p = 0.44) or partial A(z) (p = 0.46), suggesting that the Penn data were inherently more difficult to characterize. CONCLUSION: The BI-RADS lexicon facilitated the cross-institutional test of a breast cancer prediction model. The model generalized reasonably well, but there were significant performance decreases. The cross-institutional performance was encouraging because it was not significantly different from that of a reoptimized model using the second data set at high sensitivities. This study indicates the need for further work to collect more data and to improve the robustness of the model.

Breast Neoplasms↗

PubMed1991

Preliminary development of two predictive models for DNR patients in intensive care.

The purpose of this study was to identify which variables are the best predictors of a do-not-resuscitate (DNR) classification and develop a model to predict the nursing care required by DNR patients in the ICU. Data collected on DNR and non-DNR patients included nursing care requirements, severity of illness, resource allocation and sociodemographic characteristics. One model identified the best predictors of a DNR classification in intensive care as the origin of admission and the severity of illness score on the day of admission to intensive care. The second model identified the best predictors of nursing care requirements for DNR patients in intensive care as the number of days spent in intensive care prior to the DNR order, the average daily resource allocation points after the DNR order, and the severity of illness score on the day the DNR order was designated.

Aged↗