SEARCH · PubMed Health

Results for “Model performance”

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 451 records · Page 25Linked to original sources

PubMed2007

VaxiJen: a server for prediction of protective antigens, tumour antigens and subunit vaccines.

BACKGROUND: Vaccine development in the post-genomic era often begins with the in silico screening of genome information, with the most probable protective antigens being predicted rather than requiring causative microorganisms to be grown. Despite the obvious advantages of this approach--such as speed and cost efficiency--its success remains dependent on the accuracy of antigen prediction. Most approaches use sequence alignment to identify antigens. This is problematic for several reasons. Some proteins lack obvious sequence similarity, although they may share similar structures and biological properties. The antigenicity of a sequence may be encoded in a subtle and recondite manner not amendable to direct identification by sequence alignment. The discovery of truly novel antigens will be frustrated by their lack of similarity to antigens of known provenance. To overcome the limitations of alignment-dependent methods, we propose a new alignment-free approach for antigen prediction, which is based on auto cross covariance (ACC) transformation of protein sequences into uniform vectors of principal amino acid properties. RESULTS: Bacterial, viral and tumour protein datasets were used to derive models for prediction of whole protein antigenicity. Every set consisted of 100 known antigens and 100 non-antigens. The derived models were tested by internal leave-one-out cross-validation and external validation using test sets. An additional five training sets for each class of antigens were used to test the stability of the discrimination between antigens and non-antigens. The models performed well in both validations showing prediction accuracy of 70% to 89%. The models were implemented in a server, which we call VaxiJen. CONCLUSION: VaxiJen is the first server for alignment-independent prediction of protective antigens. It was developed to allow antigen classification solely based on the physicochemical properties of proteins without recourse to sequence alignment. The server can be used on its own or in combination with alignment-based prediction methods. It is freely-available online at the URL: http://www.jenner.ac.uk/VaxiJen.

Algorithms↗

PubMed2005

Prediction of death for extremely low birth weight neonates.

OBJECTIVE: To compare multiple logistic regression and neural network models in predicting death for extremely low birth weight neonates at 5 time points with cumulative data sets, as follows: scenario A, limited prenatal data; scenario B, scenario A plus additional prenatal data; scenario C, scenario B plus data from the first 5 minutes after birth; scenario D, scenario C plus data from the first 24 hours after birth; scenario E, scenario D plus data from the first 1 week after birth. METHODS: Data for all infants with birth weights of 401 to 1000 g who were born between January 1998 and April 2003 in 19 National Institute of Child Health and Human Development Neonatal Research Network centers were used (n = 8608). Twenty-eight variables were selected for analysis (3 for scenario A, 15 for scenario B, 20 for scenario C, 25 for scenario D, and 28 for scenario E) from those collected routinely. Data sets censored for prior death or missing data were created for each scenario and divided randomly into training (70%) and test (30%) data sets. Logistic regression and neural network models for predicting subsequent death were created with training data sets and evaluated with test data sets. The predictive abilities of the models were evaluated with the area under the curve of the receiver operating characteristic curves. RESULTS: The data sets for scenarios A, B, and C were similar, and prediction was best with scenario C (area under the curve: 0.85 for regression; 0.84 for neural networks), compared with scenarios A and B. The logistic regression and neural network models performed similarly well for scenarios A, B, D, and E, but the regression model was superior for scenario C. CONCLUSIONS: Prediction of death is limited even with sophisticated statistical methods such as logistic regression and nonlinear modeling techniques such as neural networks. The difficulty of predicting death should be acknowledged in discussions with families and caregivers about decisions regarding initiation or continuation of care.

Female↗

PubMed2000

Estimation of youth smoking behaviours in Canada.

This study estimated the prevalence of current smoking and smoking initiation among Canadian youth. Logistic regression was used to relate socio-demographic predictors to the occurrence of the smoking indicators among youth (15-24 years) in the 1994/95 National Population Health Survey (NPHS). Models were then applied to provincial youth populations in the 1996/97 NPHS and the 1996 census of Canada. Model-generated estimates were compared with direct estimates obtained from NPHS data. The models accurately predicted provincial rates of current youth smoking for 1994/95. When applied to the 1996/97 NPHS, the current smoking models performed reasonably well, but were less predictive when applied to 1996 census data. Modelling of youth smoking initiation was not successful. This suggests that although simple estimation models of youth smoking can be derived, these models may not be portable across different populations or time periods.

Adolescent↗

PubMed2005

Is the Charlson Comorbidity Index useful for predicting trauma outcomes?

BACKGROUND: Inclusion of a measure of comorbidity in trauma scoring has been suggested due to the potential for preexisting conditions to impact on patient outcomes, but studies have reported varied results. The Charlson Comorbidity Index (CCI) includes 19 diseases weighted on the basis of their association with mortality, and can be extrapolated from International Classification of Diseases, Ninth Revision (ICD-9) codes for administrative databases. OBJECTIVES: To evaluate the CCI as a predictor of trauma outcome. METHODS: Major trauma patient data from the Victorian State Trauma Registry (VSTR) were used to evaluate the CCI (n = 2,819). The CCI was scored from ICD-10 codes through modification of a previous method of mapping ICD-9 codes to the CCI. Logistic regression was used to determine the association between the CCI and mortality, the effect of adding the CCI to the Trauma and Injury Severity Score (TRISS) methodology, and the impact of adding the CCI to a modification of the TRISS methodology. Model performance was assessed through discrimination and calibration. RESULTS: The CCI was associated with death (p < 0.001), but adding the CCI to TRISS [area under the receiver-operating characteristic curve (AUC) 0.86; 95% CI = 0.84 to 0.88] did not result in improved discrimination over TRISS alone (AUC 0.83; 95% CI = 0.81 to 0.86). Modifying TRISS methodology, with age left as a continuous variable, performed better than the original TRISS (AUC 0.91; 95% CI = 0.89 to 0.92), but the addition of the CCI did not further improve this model (AUC 0.91; 95% CI = 0.89 to 0.92). CONCLUSIONS: While the CCI can be extrapolated from ICD codes and provides a measure of comorbid condition severity and was associated with mortality, addition of the CCI to prediction models did not result in a substantial improvement in performance.

Adolescent↗

PubMed2000

Improving health-based payment for Medicaid beneficiaries: CDPS.

This article describes the Chronic Illness and Disability Payment System (CDPS), a diagnostic classification system that Medicaid programs can use to make health-based capitated payments for TANF and disabled Medicaid beneficiaries. The authors describe the diversity of diagnoses and different burdens of illness among disabled and AFDC Medicaid beneficiaries. Claims from seven States are analyzed, and payment weights are provided that States can use when adjusting HMO payments. The authors also compare the taxonomy and statistical performance of CDPS to other leading diagnostic classification systems and find that the new model performs better in a number of respects.

Adult↗

PubMed1991

Three-dimensional anatomy and renal concentrating mechanism. II. Sensitivity results.

A mathematical model has been developed to simulate hypertonic urine formation in the renal medulla. The model uses published values of membrane transport parameters, as have other models, but is unique in its representation of the three-dimensional anatomy of the medulla. The model successfully predicts measured fluid flows, osmolarities, and NaCl and urea concentrations. The model results are presented in the companion to this paper [A. S. Wexler, R. E. Kalaba, D. J. Marsh. Am. J. Physiol. 260 (Renal Fluid Electrolyte Physiol. 29): F368-F383, 1991.]. In this paper we provide tests of the sensitivity of model performance to variations in the description of the anatomy and in membrane transport parameters. From these studies we conclude that 1) strict counterflow arrangements are required in the outer stripe to prevent loss of NaCl to the systemic circulation, 2) the radial organization in the inner stripe materially improves performance of the inner medulla, 3) radial organization of the inner medulla is essential to hypertonic urine formation there, 4) the model is most sensitive to variation in collecting duct parameters, and 5) reabsorption of urea in the distal tubule improves system performance. The results support the claim that the three-dimensional structure, as captured in the model, provides a crucial framework for the production of hypertonic urine.

Absorption↗

PubMed1988

Neural model of adaptive hand-eye coordination for single postures.

A neural network model has been developed that achieves adaptive visual-motor coordination of a multijoint arm, without a teacher. The model learns to position an arm so that it reaches a cylinder arbitrarily positioned in space. The model uses a new neural architecture and a new algorithm for modifying neural-connection strengths. Computer simulations show that the model performs with an average position error of 4% of the arm's length and with an average orientation error of 4 degrees. The model is designed to be generalized for coordinating any number of topographic sensory inputs with limbs of any number of joints.

Humans↗

PubMed2026

Dataset Readiness Assessment With Large Language Model (DRAFT-LLM): A Multi-Axis Audit Guided by LLM.

This article details the Dataset Readiness Assessment for Training (DRAFT), a systematic method for determining whether a high-dimensional biological dataset is suitable for developing reliable, equitable (i.e., the extent to which model performance, error patterns, and potential benefits or harms are evaluated and found to be acceptably distributed across relevant demographic, biological, clinical, and contextual subgroups), and scientifically meaningful machine-learning models, and DRAFT Large Language Model (DRAFT-LLM), its optional human-in-the-loop extension for calibrating study-specific audits through structured, critically reviewed LLM guidance. Standard model validation often fails to detect when apparent performance is driven by spurious correlations, technical artifacts, or hidden stratification, leading to irreproducible and inequitable findings. DRAFT-LLM addresses this gap by shifting the focus from model tuning to structured dataset auditing, organized around Support Protocols 1 to 4 that capture the scientific intent, data structure, and governance constraints of a given study. These Support Protocols: (1) elicit and formalize investigator input into a study intake and dataset card; (2) compute standardized dataset statistics and structural summaries suitable for downstream analysis and LLM context; (3) configure the language model using form-based responses, safety guardrails, and governance rules; and (4) generate personalized instructions, prompts, and code templates for running DRAFT audits. Basic Protocols 1 to 3 are instantiated from this support layer for generalization, equity, and stability: they are reusable execution patterns whose concrete behavior is determined by the cards, statistics, and configurations defined in the Support Protocols. DRAFT-LLM and DRAFT are demonstrated in this article through an end-to-end case study on The Cancer Genome Atlas (TCGA). © 2026 Wiley Periodicals LLC. Support Protocol 1: Study intake and dataset card construction Support Protocol 2: Dataset structure and advanced summary statistics for LLM context Support Protocol 3: LLM configuration using structured form responses Support Protocol 4: Generation of personalized instructions for DRAFT audits Basic Protocol 1: Generalization audit Basic Protocol 2: Equity audit Basic Protocol 3: Stability audit.

Large Language Models↗

PubMed1993

[Multifocal intraocular lenses--an assessment of current status].

UNLABELLED: Besides the diffractive multifocals, which produce a second focus for near vision by means of diffraction rings, there are different refractive multifocal IOL types with 2-7 refractive zones or an aspheric/spherical construction principle. Long-term results: 2 years after implantation of diffractive multifocal IOLs, the corrected distance and near acuities were unchanged compared to the 3-month results. The uncorrected distance acuity was, however, slightly decreased due to a minus shift of refraction to -1.2 D. The contrast sensitivity was improved after 2 years. Multi- versus monofocal IOLs: After diffractive multifocal IOL implantation, the near acuity with distance correction only was markedly improved compared to monofocal IOLs. All other acuity data did not differ between multi- or monofocal lenses. The contrast sensitivity (at low contrasts and high spatial frequencies) and mesopic visual acuity (without and with glare) were reduced compared to monofocal pseudophakic eyes. Near aniseikonia and binocular functions: In unilateral multifocal pseudophakia (monofocal IOL in fellow eye), a near aniseikonia up to 8% was found. The width of fusion was significantly lower than in bilateral multifocal pseudophakia, whereas the stereopsis showed no difference. Determinants of bifocal function: In 7.1% of our cases, no bifocal function (BFF) was present after implantation of diffractive multifocal IOLs. These patients exhibited a significantly higher age as well as higher pre- and postoperative astigmatism, when compared to patients with good BFF. Optical performance of different multifocal IOLs: By means of an optical system, described by Reiner, images of intraocular lenses can be projected into the eye ("optical implantation"); thus, the optical performance of IOLs can be judged subjectively. Using this method, the refractive 2- and 3-zone models performed best within the multifocal group (contrast sensitivity not significantly worse than that of monofocal IOL), when viewing a low-contrast chart (Regan 4%). All other multifocal lenses (diffractive, aspheric/spherical, refractive 5- and 7-zone models) were significantly inferior to the monofocal IOL. CONCLUSIONS: Implantation of multifocal IOLs should presently be restricted to special indications, particularly to the distinct patient request to dispense with wearing near or bifocal glasses, if possible. Because of the reduction in contrast sensitivity and mesopic vision and the increased glare sensibility, multifocal IOLs should not be implanted especially in professional car drivers. There are, however, differences in optical performance between the various multifocal IOL types. Further improvements, in particular concerning lens technology, will presumably extend the present spectrum of indications.

Contrast Sensitivity↗

PubMed1998

Lumbar muscle force estimation using a subject-invariant 5-parameter EMG-based model.

The use of electromyographic measures, in concert with modeled or empirical representations of muscle physiology, is a common approach for estimation of muscle force. Existing models of the lumbar musculature have allowed model parameters to vary for an individual subject. While this approach improves apparent predictive ability, it loses some degree of construct validity since parameter variability may not be physiologically justifiable. An EMG-based five-parameter model, adapted and generalized from earlier reports, is presented here. Inherent in the model is the requirement of subject-invariant modeling parameters. As a practical analysis tool was desired, the model relies on relatively few calibration constants whose determination is described. Empirical evaluation was undertaken using a database of 398 experimental trials involving lifting and transferring objects of moderate mass. Model performance, evaluated by comparison of measured and predicted lumbar moments, was comparable to earlier models, with r2 mean (S.D.) values of 0.76(0.15) for sagittal plane moments, and rms mean (S.D.) errors of 14.1(7.4), 9.7(5.3), and 8.6(3.6) Nm in the sagittal, frontal, and horizontal planes, respectively. These empirical results and the argument of physiological veracity support the use of a subject-invariant model.

Adult↗

PubMed2006

Semi-supervised learning via penalized mixture model with application to microarray sample classification.

MOTIVATION: It is biologically interesting to address whether human blood outgrowth endothelial cells (BOECs) belong to or are closer to large vessel endothelial cells (LVECs) or microvascular endothelial cells (MVECs) based on global expression profiling. An earlier analysis using a hierarchical clustering and a small set of genes suggested that BOECs seemed to be closer to MVECs. By taking advantage of the two known classes, LVEC and MVEC, while allowing BOEC samples to belong to either of the two classes or to form their own new class, we take a semi-supervised learning approach; for high-dimensional data as encountered here, we propose a penalized mixture model with a weighted L1 penalty to realize automatic feature selection while fitting the model. RESULTS: We applied our penalized mixture model to a combined dataset containing 27 BOEC, 28 LVEC and 25 MVEC samples. Analysis results indicated that the BOEC samples appeared to form their own new class. A simulation study confirmed that, compared with the standard mixture model with or without initial variable selection, the penalized mixture model performed much better in identifying relevant genes and forming corresponding clusters. The penalized mixture model seems to be promising for high-dimensional data with the capability of novel class discovery and automatic feature selection.

Algorithms↗

PubMed1996

Secondary diagnoses as predictive factors for survival or mortality in Medicare patients with acute pneumonia.

We wished to determine if a claims-based method for severity adjustment would predict mortality or survival in pneumonia based on age, gender, and secondary diagnoses. We used a discriminant analysis model of severity of illness developed from Medicare Part A claims data. Our data base was taken from a hospitalized population age 65 years or older coded as DRG 89 (pneumonia with complications/comorbidities). There were 35,677 cases with a mortality = 11.2% in the derivation cohort from 1989 to 1990, and 19,915 cases with a mortality = 9.8% in the validation cohort from 1991. In the derivation cohort, 98% of patients predicted to live, lived, whereas 18% of patients predicted to die, died. Of the three variables, secondary diagnoses had greatest explanatory power. Receiver operating characteristic curves showed that the model performed best at 40% survival. Results were confirmed with the 1991 validation cohort. The model could be applied to hospitals with as few as 172 discharges. This simple, claims-based method can predict survival in pneumonia. It may be useful in selecting medical records for intensified review of medical quality.

Aged↗

PubMedDate not supplied

Implementation of sulfate adsorption in the SAFE model.

An SO4(2-) adsorption submodel has been implemented in the dynamic soil chemistry model SAFE. The submodel calculates pH-dependent SO4(2-) and H+ adsorption to the soil, as well as the net surface charge development due to uneven adsorption of SO4(2-) and H+, using the empirical equations derived from an electrostatic model (Extended Constant Capacitance Model, ECCM) of SO4(2-) adsorption. The resulting new SAFE model was applied on a roof experiment plot in the Norway spruce [Picea abies (L.) H. Karst.] stand at Solling, Germany, where atmospheric S and N deposition was artificially reduced by the roof construction. The model performance was compared with the previous versions that used a pH-independent Freudlich model of SO4(2-) adsorption or assumed no SO4(2-) adsorption. With the ECCM-based SO4(2-) adsorption submodel, SAFE simulated soil solution SO4(2-) concentration and base saturation better, in comparison with measured data, than with the previous SO4(2-) adsorption formulations. Through the model application, also, need of additional improvement was suggested, such as calibration of mass transfer coefficients.

Adsorption↗

PubMed2002

Genotype x environment interaction for protein yield in Dutch dairy cattle as quantified by different models.

Variance components and breeding values for protein yield were estimated with REML without and with correction for heterogeneity of variances. Three different sire models were applied, which all accounted for genotype x environment (G x E) interaction. The first model included a sire x herd-year-season subclass (HYS) interaction. The second model divided all records in four different types of management groups, based on estimated HYS subclass effect. The third model, the reaction norm model, performed a random linear regression on the estimated HYS effect. For comparison, a standard model that did not take G x E interaction into account was also applied. Data consisted of 102,899 305-d first-lactation protein records of Holstein Friesians of 1,000 ofthe largest Dutch dairy herds. All animals calved in 1997, 1998, or 1999. Estimated breeding values (EBV) for 2,150 bulls with at least five daughters were calculated. The interaction model detected an interaction variance of 2.5% of the phenotypic variance. The EBV showed a correlation of 1.00 with those of the standard model without interaction. The model with the division in groups showed correlations between groups ranging from 0.73 to 0.86. The EBV showed correlations from 0.84 to 0.91 with the EBV of the standard model. The reaction norm model calculated EBV that had a correlation of 1.00 with the EBV of the standard model. The reaction norm model was not able to detect significant variance of the slope for the protein data corrected for heterogeneity of variances.

Analysis of Variance↗

PubMed2005

An information maximization model of eye movements.

We propose a sequential information maximization model as a general strategy for programming eye movements. The model reconstructs high-resolution visual information from a sequence of fixations, taking into account the fall-off in resolution from the fovea to the periphery. From this framework we get a simple rule for predicting fixation sequences: after each fixation, fixate next at the location that minimizes uncertainty (maximizes information) about the stimulus. By comparing our model performance to human eye movement data and to predictions from a saliency and random model, we demonstrate that our model is best at predicting fixation locations. Modeling additional biological constraints will improve the prediction of fixation sequences. Our results suggest that information maximization is a useful principle for programming eye movements.

Algorithms↗

PubMed2007

Air quality assessment for Portugal.

According to the Air Quality Framework Directive, air pollutant concentration levels have to be assessed and reported annually by each European Union member state, taking into consideration European air quality standards. Plans and programmes should be implemented in zones and agglomerations where pollutant concentrations exceed the limit and target values. The main objective of this study is to perform a long-term air quality simulation for Portugal, using the CHIMERE chemistry-transport model, applied over Portugal, for the year 2001. The model performance was evaluated by comparing its results to air quality data from the regional monitoring networks and to data from a diffusive sampling experimental campaign. The results obtained show a modelling system able to reproduce the pollutant concentrations' temporal evolution and spatial distribution observed at the regional networks of air quality monitoring. As far as the fulfilment of the air quality targets is concerned, there are excessive values for nitrogen and sulfur dioxides, ozone also being a critical gaseous pollutant in what concerns hourly concentrations and AOT40 (Accumulated Over Threshold 40 ppb) values.

Air Pollutants↗

PubMed2005

Prediction of liver fibrosis in human immunodeficiency virus/hepatitis C virus coinfected patients by simple non-invasive indexes.

BACKGROUND: Liver biopsy is an invasive technique with associated major complications. There is no information on the validity of five non-invasive indexes based on routinely available parameters, estimated and validated in hepatitis C virus (HCV) monoinfected patients, in human immunodeficiency virus (HIV)/HCV coinfected patients. AIM: To validate these predictive models of liver fibrosis in HIV/HCV coinfected patients. PATIENTS: A total of 357 (90%) of 398 patients from five hospitals were investigated, who underwent liver biopsy and who had complete data to validate all of the models considered. METHODS: The predictive accuracy of the indexes was tested by measuring areas under the receiver operating characteristic curves. Diagnostic accuracy was calculated by estimating sensitivity, specificity, and positive (PPV) and negative (NPV) predictive values. RESULTS: The models performed better when liver biopsies>or=15 mm were used as reference. In this setting, the Forns and Wai indexes, models aimed at discriminating significant fibrosis, showed PPV of 94% and 87%, respectively. Using these models, 27-34% of patients could benefit from exclusion of liver biopsy. If both models were applied sequentially, 41% of liver biopsies could be spared. The indexes aimed at predicting cirrhosis achieved NPV of up to 100%. However, they showed very low PPV. CONCLUSIONS: The diagnostic accuracy of these models was lower in HIV/HCV coinfected patients than in the validation studies performed in HCV monoinfected patients. However, simple fibrosis tests may render liver biopsy unnecessary in deciding anti-HCV treatment in over one third of patients with HIV infection and chronic hepatitis C.

Adult↗

PubMed1993

The predictive performance of a system model for enflurane closed-circuit inhalational anesthesia.

BACKGROUND: Previously, the authors described a system model for closed-circuit inhalational anesthesia, and demonstrated close agreement between end-tidal isoflurane concentrations measured in their clinical study and those predicted by the model. The predictive performance of their model has not, however, been tested for anesthetics featuring nonpulmonary elimination (NPE). METHODS: The authors quantified the predictive performance of two versions (A and C) of the model in 50 patients by comparing the predicted and the measured alveolar concentration-time profiles after bolus injections of liquid enflurane into the expiratory limb of the closed system. Version A did not incorporate NPE, but version C emulated NPE by adopting the irreversible loss of a fraction of the enflurane present in the arterial hepatic blood flow (0.131, derived from a mass balance study performed by others). For each concentration measured by mass spectrometry, the authors used computer simulations of version A and C to calculate a predicted concentration for both versions. For each patient, the authors calculated the bias (indicating systematic over- or underprediction) and the scatter of the prediction errors (indicating typical error size). RESULTS: The authors administered a total of 379 ml of liquid enflurane via 466 injections. A total of 18,432 alveolar concentrations (one per 10-s period; average concentration = 0.96 vol%) were measured. The bias and the scatter, both given as mean (and SD), were 10.0 (13.1)% and 11.8 (3.9)% for version A and -0.8 (11.4)% and 11.4 (2.8)% for C. The bias for version C was closer to zero; the scatters were similar. CONCLUSIONS: Version C incorporating NPE performs better than version A. The accuracy that was obtained should encourage the use of version C for clinical, teaching, research, economic, and ecologic purposes.

Adult↗