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 379 records · Page 21Linked to original sources

PubMed2004

Application of a prediction model for identification of individuals at diabetic risk.

As a follow-up to our preceding paper, we attempted to extract features of health risk progression for diabetes in Sequential Multi Layered Perceptron (SMLP) via inverse processing of the learned structure. The time-varying risk progress was assessed with risk trajectory and conditional mixture model. Overall risk cut along with the prediction was stable over time and high body mass index (BMI) tops the health behavioral risks predicting the onset of diabetes. For the initial prediction, high BMI (obesity), high blood pressure (BP), high cholesterol, and diet in fatty food were significant. Over time, variations in trajectory were due to changes in BMI, stress, BP, cholesterol, and fatty food intake. We tested the effectiveness of identifying prediabetics by the SMLP by applying the implemented SMLP to a test population of employees from a large manufacturing company, where an early worksite health promotion was initiated (1984). This resulted in a potential sensitivity (71.4%) although there were issues like mapping corresponding risks and large time lags. A secondary test on the similar population as in the previous paper showed a promising sensitivity (86.5%) over 3 years. When combining with targeted screening such as impaired glucose tolerance test only for those predicted to be diabetics, the presented prediction model and extracted features can be used in implementing an effective disease prevention and management program.

Diabetes Mellitus, Type 2↗

PubMed2001

Analysis and assessment of comparative modeling predictions in CASP4.

We describe here the results of our analysis of the comparative modeling predictions submitted to the fourth round of Critical Assessment of Structure Prediction (CASP4). On the basis of a numerical evaluation of the models, we assessed their ability to predict the overall fold correctly, the relative orientation of domains in multidomain proteins, the conformation of the side chains, the loop regions, and the biologically important residues of the targets. We also discuss the performance of automatic prediction servers and compare the results of CASP4 with those obtained in CASP3.

Automation↗

PubMed2003

Predictive modelling of growth and enzyme production and activity by a cocktail of Pseudomonas spp., Shewanella putrefaciens and Acinetobacter sp.

The possibility was examined of developing a predictive model that combined microbial growth (increase in cellular number) with extracellular lipolytic and proteolytic enzyme activity of a cocktail of four strains of Pseudomonas spp. and one strain each of Acinetobacter sp. and Shewanella putrefaciens. Environmental conditions within the following matrix of conditions were examined: temperature 2-20 degrees C, pH value 4.0-7.5 and water activity (a(w)) 0.95-0.995 and a model was constructed, which predicted growth based on increase in cell number. Data on lipase production and protease activity were generated and will be available as a database, but no function could be identified, which was a good fit to these data, since most enzymatic production and activity occurred, as expected, during transition from exponential to stationary phase. Even at lower cell numbers, in more unfavourable conditions, hydrolysing effects were detectable, which made it difficult to construct a model combining both microbiological and enzymatic data.

Acinetobacter↗

PubMed1997

A predictive model for ADL dependence in community-living older adults based on a reduced set of cognitive status items.

OBJECTIVE: To develop and validate a simple tool, based on a reduced set of Mini-Mental State Examination (MMSE) items, that can be used to predict the onset of ADL dependence, and to compare the predictive accuracy of this new tool with that of the MMSE. DESIGN: Two prospective, population-based cohort studies, in tandem. The predictive model developed in the initial cohort was subsequently validated in a separate cohort. SETTING: General community in New Haven, Connecticut. PARTICIPANTS: For the development cohort, 775 community-living persons, 72 years of age and older, who were independent at baseline in their ADI, function. For the validation cohort, 1038 comparable subjects. MEASUREMENTS: All subjects underwent a baseline interview and cognitive assessment in their homes by a trained research nurse using standard instrument. Self-reported ADLs were ascertained at 1 year and 3 years for the development cohort and at 1 year and 21/2 years for the validation cohort. RESULTS: ADL dependence developed in 221 (28.5%) subjects in the development cohort. Although the rate of ADL dependence increased within each MMSE domain as the number of incorrect items increased, only orientation and short-term memory remained significantly associated with ADL dependence in multivariable analysis. A predictive model, based on the presence of impairments in these two domains, was developed that stratified subjects into three risk groups. Rates of ADL dependence were 22% (neither domain impaired), 44% (one domain impaired), and 68% (both domains impaired) (P < .001). The corresponding rates in the validation cohort, in which 191 (18.4%) subjects developed ADL dependence, were 15%, 26%, and 45% (P < .001). The area under the ROC curves for the MMSE and the reduced item strategy were nearly identical at 0.63 and 0.62, respectively. CONCLUSIONS: A simple and valid six-item strategy, based on the presence of impairments in orientation and short-term memory, predicts the onset of ADL dependence as effectively as does the 30-item MMSE. This new tool may be useful as part of a more comprehensive assessment when determining an older person's risk for developing ADL dependence.

Activities of Daily Living↗

PubMed2006

Radiation rescue for biochemical failure after surgery for prostate cancer: predictive parameters and an assessment of contemporary predictive models.

OBJECTIVES: To determine pretreatment prognostic variables that predict outcome of radiotherapy for biochemical failure after prostate cancer surgery and evaluate contemporary clinical decision tools for patient selection. METHODS: Fifty patients were identified with failure after rescue radiation was defined as a confirmed rise in PSA, distant metastases, prostate cancer death, or initiation of hormonal therapy. Univariate analysis and multivariate Cox models were constructed. Outcome was compared with decision tree and recursive partitioning predictive models. RESULTS: The median preradiation PSA (pre-RT PSA) was 1.2 ng/mL and the median dose of radiation was 66.6 Gy; median follow-up was 39.6 months. Overall, the estimated 3-year failure free survival was 54%, 95%CI [43,74]. Seminal vesicle involvement (SVI) (P = 0.003) and preradiation PSA Doubling Time (PSADT) <10 months (P = 0.01) were both significant predictors for treatment failure whereas pre-RT PSA was of borderline significance (P = 0.07). On multivariate analysis a pre-RT PSA of >1 and SVI were associated with hazard ratios of 6.2 and 7.3 (P = 0.01 and P = 0.004), respectively. An additional Cox model constructed for 31 patients for whom pre-RT PSADT could be calculated showed PSADT and SVI to be independent prognostic parameters. Two predictive models, a decision tree analysis, and a recursive partitioning model were moderately accurate in predicting outcome in this series, however, high-risk patients experienced less treatment failures than predicted. CONCLUSIONS: Pre-RT PSA <1 ng/mL, longer PSADT (>10 months) and no SVI are associated with improved outcome after rescue radiation. Contemporary clinical prediction tools are imperfect predictors of outcome for rescue radiation therapy.

Decision Trees↗

PubMed2005

A model predictive control based scheduling method for HIV therapy.

Recently developed models of the interaction of the human immune system and the human immunodeficiency virus (HIV) suggest the possibility of using interruptions of highly active anti-retroviral therapy (HAART) to simulate a therapeutic vaccine and induce cytotoxic lymphocyte (CTL) mediated control of HIV infection. We have developed a model predictive control (MPC) based method for determining optimal treatment interruption schedules for this purpose. This method provides a clinically implementable framework for calculating interruption schedules that are robust to errors due to measurement and patient variations. In this paper, we discuss the medical motivation for this work, introduce the MPC-based method, show simulation results, and discuss future work necessary to implement the method.

AIDS Vaccines↗

PubMed1986

Local recurrence and the EEA stapler--examination of a predictive model.

The overall local recurrence rate following resection of colorectal cancer with restoration of continuity with staples in Wellington was 24%. Nine of 11 patients with local recurrence following resection of rectal tumours had distant metastases at the time of diagnosis of their local recurrence. Using a predictive model to retrospectively estimate the probability of local recurrence it was found that nine of these 11 patients would have been expected to have had a lower local recurrence rate had they undergone abdominoperineal resection of the rectum initially. Since local recurrence is simply a local manifestation of systemic disease in 90% of patients, however, it is suggested that patients would prefer restoration of bowel continuity in preference to rectal excision and stoma formation, there being such little survival advantage for the latter procedure. The utility of the predictive model is therefore questioned.

Adult↗

PubMed2002

Prediction models in the design of neural network based ECG classifiers: a neural network and genetic programming approach.

BACKGROUND: Classification of the electrocardiogram using Neural Networks has become a widely used method in recent years. The efficiency of these classifiers depends upon a number of factors including network training. Unfortunately, there is a shortage of evidence available to enable specific design choices to be made and as a consequence, many designs are made on the basis of trial and error. In this study we develop prediction models to indicate the point at which training should stop for Neural Network based Electrocardiogram classifiers in order to ensure maximum generalisation. METHODS: Two prediction models have been presented; one based on Neural Networks and the other on Genetic Programming. The inputs to the models were 5 variable training parameters and the output indicated the point at which training should stop. Training and testing of the models was based on the results from 44 previously developed bi-group Neural Network classifiers, discriminating between Anterior Myocardial Infarction and normal patients. RESULTS: Our results show that both approaches provide close fits to the training data; p = 0.627 and p = 0.304 for the Neural Network and Genetic Programming methods respectively. For unseen data, the Neural Network exhibited no significant differences between actual and predicted outputs (p = 0.306) while the Genetic Programming method showed a marginally significant difference (p = 0.047). CONCLUSIONS: The approaches provide reverse engineering solutions to the development of Neural Network based Electrocardiogram classifiers. That is given the network design and architecture, an indication can be given as to when training should stop to obtain maximum network generalisation.

Electrocardiography↗

PubMed2004

Predictive models for protein crystallization.

Crystallization of proteins is a nontrivial task, and despite the substantial efforts in robotic automation, crystallization screening is still largely based on trial-and-error sampling of a limited subset of suitable reagents and experimental parameters. Funding of high throughput crystallography pilot projects through the NIH Protein Structure Initiative provides the opportunity to collect crystallization data in a comprehensive and statistically valid form. Data mining and machine learning algorithms thus have the potential to deliver predictive models for protein crystallization. However, the underlying complex physical reality of crystallization, combined with a generally ill-defined and sparsely populated sampling space, and inconsistent scoring and annotation make the development of predictive models non-trivial. We discuss the conceptual problems, and review strengths and limitations of current approaches towards crystallization prediction, emphasizing the importance of comprehensive and valid sampling protocols. In view of limited overlap in techniques and sampling parameters between the publicly funded high throughput crystallography initiatives, exchange of information and standardization should be encouraged, aiming to effectively integrate data mining and machine learning efforts into a comprehensive predictive framework for protein crystallization. Similar experimental design and knowledge discovery strategies should be applied to valid analysis and prediction of protein expression, solubilization, and purification, as well as crystal handling and cryo-protection.

Bayes Theorem↗

PubMed2003

Prediction model to identify patients with Staphylococcus aureus bacteremia at risk for methicillin resistance.

OBJECTIVES: To identify institution-specific risk factors for MRSA bacteremia and develop an objective mechanism to estimate the probability of methicillin resistance in a given patient with Staphylococcus aureus bacteremia (SAB). DESIGN: A cohort study was performed to identify institution-specific risk factors for MRSA. Logistic regression was used to model the likelihood of MRSA. A stepwise approach was employed to derive a parsimonious model. The MRSA prediction tool was developed from the final model. SETTING: A 279-bed, level 1 trauma center. PATIENTS: Between January 1, 1999, and June 30, 2001, 494 patients with clinically significant episodes of SAB were identified. RESULTS: The MRSA rate was 45.5%. Of 18 characteristics included in the logistic regression, the only independent features for MRSA were prior antibiotic exposure (OR, 9.2; CI95, 4.8 to 17.9), hospital onset (OR, 3.0; CI95, 1.9 to 4.9), history of hospitalization (OR, 2.5; CI95, 1.5 to 3.8), and presence of decubitus ulcers (OR, 2.5; CI95, 1.2 to 4.9). The prediction tool was derived from the final model, which was shown to accurately reflect the actual MRSA distribution in the cohort. CONCLUSION: Through multivariate modeling techniques, we were able to identify the most important determinants of MRSA at our institution and develop a tool to predict the probability of methicillin resistance in a patient with SAB. This knowledge can be used to guide empiric antibiotic selection. In the era of antibiotic resistance, such tools are essential to prevent indiscriminate antibiotic use and preserve the longevity of current antimicrobials.

Adult↗

PubMed2003

Autologous peripheral blood stem cell transplantation in first remission adult acute myeloid leukaemia--an intention to treat analysis and comparison of outcome using a predictive model based on the MRC AML10 cohort.

The role of autologous peripheral blood stem cell transplantation (APBSCT) in acute myeloid leukaemia (AML) remains controversial. The current study evaluated the application of APBSCT in a large consecutive series of patients with untreated AML, and compared outcome with a predictive model based on MRC AML10 data. Of 148 evaluable patients, 118 patients entered complete remission (CR) after induction therapy comprising three cycles of daunorubicin, cytosine arabinoside and oral 6-thioguanine. Of these patients, 68 (57%) proceeded to consolidation therapy with two courses of intermediate dose cytosine arabinoside, and stem cell mobilisation, and 40 of these patients (34%) underwent the APBSCT procedure after high dose busulphan conditioning. Harvest quality was the main factor precluding APBSCT. Five-year event-free survival (EFS) in patients who achieved CR was 38% and in APBSCT patients was 57%. There were no transplant-related deaths. No significant differences were demonstrated between observed and expected outcomes at 1 and 2 years, based on the predictive model derived from the MRC AML10 study. These data therefore indicate that only a third of eligible adult patients will undergo APBSCT. However, the results demonstrate favourable survival in such patients, with no transplant-related mortality.

Acute Disease↗

PubMed2004

Traffic safety assessment and development of predictive models for accidents on rural roads in Egypt.

This paper starts by presenting a conceptualization of indicators, criteria and accidents' causes that can be used to describe traffic safety. The paper provides an assessment of traffic safety conditions for rural roads in Egypt. This is done through a three-step procedure. First, deaths per million vehicle kilometers are obtained and compared for Egypt, three other Arab countries and six of the G-7 countries. Egypt stands as having a significantly high rate of deaths per 100 million vehicle kilometers. This is followed by compiling available traffic and accident data for five main rural roads in Egypt over a 10-year period (1990-1999). These are used to compute and compare 13 traffic safety indicators for these roads. The third step for assessing traffic safety for rural roads in Egypt is concerned with presenting a detailed analysis of accident causes. The paper moves on to develop a number of statistical models that can be used in the prediction of the expected number of accidents, injuries, fatalities and casualties on the rural roads in Egypt. Time series data of traffic and accidents, over a 10 years period for the considered roads, is utilized in the calibration of these predictive models. Several functional forms are explored and tested in the calibration process. Before proceeding to the development of these models three ANOVA statistical tests are conducted to establish whether there are any significant differences in the data used for models' calibration as a result of differences among the considered five roads.

Accident Prevention↗

PubMedDate not supplied

A predictive model for substrates of cytochrome P450-debrisoquine (2D6).

Molecular modeling techniques were used to derive a predictive model for substrates of cytochrome P450 2D6, an isozyme known to metabolize only compounds with one or more basic nitrogen atoms. Sixteen substrates, accounting for 23 metabolic reactions, with a distance of either 5 A ("5-A substrates", e.g., debrisoquine) or 7 A ("7-A substrates", e.g., dextromethorphan) between oxidation site and basic nitrogen atom were fitted into one model by postulating an interaction of the basic nitrogen atom with a negatively charged carboxylate group on the protein. This acidic residue anchors and neutralizes the positively charged basic nitrogen atom of the substrates. In case of "5-A substrates" this interaction probably occurs with the carboxylic oxygen atom nearest to the oxidation site, whereas in the case of "7-A substrates" this interaction takes place at the other oxygen atom. Furthermore, all substrates exhibit a coplanar conformation near the oxidation site and have negative molecular electrostatic potentials (MEPs) in a part of this planar domain approximately 3 A away from the oxidation site. No common features were found in the neighbourhood of the basic nitrogen atom of the substrates studied so that this region of the active site can accommodate a variety of N-substituents. Therefore, the substrate specificity of P450 2D6 most likely is determined by the distance between oxidation site and basic nitrogen atom, by steric constraints near the oxidation site, and by the degree of complementarity between the MEPs of substrate and protein in the planar region adjacent to the oxidation site.(ABSTRACT TRUNCATED AT 250 WORDS)

Astemizole↗

PubMed2002

Application of short-term water demand prediction model to Seoul.

To predict daily water demand for Seoul, Korea, the artificial neural network (ANN) was used. For the cross correlation, the factors affecting water demand such as maximum temperature, humidity, and wind speed as natural factors, holidays as a social factor and daily demand 1 day before were used. From the results of learning using various hidden layers and units in order to establish the structure of optimal ANN, the case of 3 hidden layers and numbers of unit with the same number of input factors showed the best result and, therefore, it was applied to seasonal water demand prediction. The performance of ANN was compared with a multiple regression method. We discuss the representation ability of the model building process and the applicability of the ANN approach for the daily water demand prediction. ANN provided reasonable results for time series prediction.

Forecasting↗

PubMed1992

Mink as a predictive model in toxicology.

This paper reviewed the biomedical and toxicological database concerning the use of mink as a predictive model of human responses. It is concluded that substantial information exists on the mink genetics, physiology, metabolism, nutritional requirements, and susceptibility to infectious disease; and provides a foundation upon which interspecies extrapolation may be considered. In addition, information on the response of mink to several dozen toxic substances revealed that mink respond in a qualitatively and quantitatively similar manner to other more commonly employed species as well as humans. Our conclusion does not infer that mink should be used routinely in toxicological testing for estimation of human responses. However, it indicates that toxicological data from this species may be a useful complement in risk assessment processes based upon data obtained from traditionally employed models such as rats and dogs.

Animals↗

PubMed2006

Advanced illness index: Predictive modeling to stratify elders using self-report data.

OBJECTIVES: Develop a prediction model to identify persons who have an increased risk of dying within the next 36 months, in order to focus additional resources and assessment in areas related to advanced care planning. DESIGN: Retrospective study with a 3-year observation period. SETTING: Integrated, not-for-profit managed care organization. PARTICIPANTS: Beneficiaries aged 65-105 responding to an annual survey (n = 4888). MEASUREMENTS: Survey instrument includes physical function, geriatric syndromes, health care utilization, special equipment use, self-care deficits, caregiving responsibilities, and general health problems. RESULTS: An 11-variable model changed the baseline chi2 from 315.71 (df = 1) to 742.511 (df = 11). The percent of subjects correctly classified was 74.3% and the negative predictive value was 92.2%. CONCLUSION: Advanced Illness Index (AII) model is stable. Characteristic variables used are not easily reversed: the 1997 cohort classified as at-risk consistently remained at risk or died in the subsequent years (1998, 92%; and 1999, 96%) and 92% of those not at-risk survived the next 36 months. Persons at high risk should at a minimum be made aware of the types of integrated home and community-based services available to them should it be needed. They also should be targeted for elicitation of treatment preferences, values, designation of health care proxy, planning, and advanced care directives.

Aged↗

PubMed2026

Netupitant versus aprepitant: model-predicted neurokinin-1 receptor occupancy and implications for long-delayed nausea and vomiting prevention.

PURPOSE: Nausea and vomiting beyond 5 days after emetogenic chemotherapy or antibody-drug conjugate (ADC) therapy are common, yet the role of neurokinin-1 (NK1) receptor antagonists in this setting remains underrecognized. We used pharmacokinetic/pharmacodynamic (PK/PD) modeling to estimate the NK1 receptor occupancy (RO), a proxy for clinical efficacy, for up to 20 days after a single 300 mg dose of oral netupitant, 3-day oral aprepitant (125 mg on day 1; 80 mg on days 2-3), or a single 165 mg dose of oral aprepitant. METHODS: Data from previous PK studies were analyzed by compartmental modeling. Positron emission tomography studies assessing striatal NK1 RO were used to develop maximum drug effect PD models, which were fitted to NK1 RO data as a function of plasma concentrations. RESULTS: Model predicted NK1 RO exceeded 90% at 3 h for all treatments. Thereafter, RO declined more gradually with netupitant (76%, 70%, 60%, and 21% on days 5, 7, 10, and 20, respectively) than with 3-day aprepitant (81%, 39%, 3%, and negligible) or single-dose aprepitant (45%, 10%, < 1%, and negligible). The half-life of netupitant was ~ 6.1 times longer than aprepitant's. Netupitant plasma concentration remained above the effective concentration for 50% NK1 RO (EC50) through day 10, whereas aprepitant concentrations fell below the EC50 by ~ days 7 and 5 after repeated and single dosing, respectively. CONCLUSIONS: Single-dose netupitant maintained NK1 RO substantially longer than repeated- and single-dose aprepitant, suggesting greater potential for prolonged prevention of nausea and vomiting in ADC-treated patients. Prospective clinical validation of these model predictions would be beneficial.

Aprepitant↗

PubMed2002

An alternative accident prediction model for highway-rail interfaces.

Safety levels at highway/rail interfaces continue to be of major concern despite an ever-increasing focus on improved design and appurtenance application practices. Despite the encouraging trend towards improved safety, accident frequencies remain high, many of which result in fatalities. More than half of these accidents occur at public crossings, where active warning devices (i.e. gates, lights, bells, etc.) are in place and functioning properly. This phenomenon speaks directly to the need to re-examine both safety evaluation (i.e. accident prediction) methods and design practices at highway-rail crossings. With respect to earlier developed accident prediction methods, the Peabody Dimmick Formula, the New Hampshire Index and the National Cooperative Highway Research Program (NCHRP) Hazard Index, all lack descriptive capabilities due to their limited number of explanatory variables. Further, each has unique limitations that are detailed in this paper. The US Department of Transportation's (USDOT) Accident Prediction Formula, which is most widely, also has limitations related to the complexity of the three-stage formula and its decline in accident prediction model accuracy over time. This investigation resulted in the development of an alternate highway-rail crossing accident prediction model, using negative binomial regression that shows great promise. The benefit to be gained through the application of this alternate model is (1) a greatly simplified, one-step estimation process; (2) comparable supporting data requirements and (3) interpretation of both the magnitude and direction of the effect of the factors found to significantly influence highway-rail crossing accident frequencies.

Accidents, Traffic↗