SEARCH · PubMed Health

Results for “Predictive modeling”

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 253 records · Page 14Linked to original sources

PubMed2001

Prognostic significance of pathologic features in localized prostate cancer treated with radical prostatectomy: implications for staging systems and predictive models.

PURPOSE: Although predicting outcome for men with clinically localized prostate cancer (PC) has improved, the staging system and nomograms used to do this are based on results from the North American health system. To be internationally applicable, these models require testing in cohorts from a variety of different health systems based on the predominant PC case identification methods used. PATIENTS AND METHODS: We studied 732 men with localized PC treated with radical prostatectomy and no preoperative therapy between 1986 and 1999 at one Australian institution to determine the effect of clinicopathologic features on disease-free survival. RESULTS: Preoperative serum prostate-specific antigen (PSA) concentration, Gleason score, pathologic stage, and year of surgery were independent predictors of outcome. Although margin status demonstrated only a trend toward significance in multivariate modeling overall, it proved to be independent in subgroups based on later year of surgery (1986 to 1994 v 1995 to 1998), preoperative PSA of less than 10 ng/mL, and Gleason score > or = 7. Adjuvant radiation therapy improved disease-free survival rates in patients with multiple surgical margin involvement. CONCLUSION: This work confirms the prognostic significance of pathologic stage, Gleason score, and preoperative serum PSA. In the context of a contemporaneous screening effect in Australia, these findings may have implications for methods that predict outcome following surgery as screening becomes more prevalent in a population. The independent prognostic effect of margin status may alter with an increase in the proportion of screening-identified PCs. Staging systems and nomograms that predict outcome following surgery require validation in cohorts with different health practices before being universally applied.

Adenocarcinoma↗

PubMed2006

Prospective breast cancer risk prediction model for women undergoing screening mammography.

BACKGROUND: Risk prediction models for breast cancer can be improved by the addition of recently identified risk factors, including breast density and use of hormone therapy. We used prospective risk information to predict a diagnosis of breast cancer in a cohort of 1 million women undergoing screening mammography. METHODS: There were 2,392,998 eligible screening mammograms from women without previously diagnosed breast cancer who had had a prior mammogram in the preceding 5 years. Within 1 year of the screening mammogram, 11,638 women were diagnosed with breast cancer. Separate logistic regression risk models were constructed for premenopausal and postmenopausal examinations by use of a stringent (P<.0001) criterion for the inclusion of risk factors. Risk models were constructed with 75% of the data and validated with the remaining 25%. Concordance of the predicted with the observed outcomes was assessed by a concordance (c) statistic after logistic regression model fit. All statistical tests were two-sided. RESULTS: Statistically significant risk factors for breast cancer diagnosis among premenopausal women included age, breast density, family history of breast cancer, and a prior breast procedure. For postmenopausal women, the statistically significant factors included age, breast density, race, ethnicity, family history of breast cancer, a prior breast procedure, body mass index, natural menopause, hormone therapy, and a prior false-positive mammogram. The model may identify high-risk women better than the Gail model, although predictive accuracy was only moderate. The c statistics were 0.631 (95% confidence interval [CI] = 0.618 to 0.644) for premenopausal women and 0.624 (95% CI = 0.619 to 0.630) for postmenopausal women. CONCLUSION: Breast density is a strong additional risk factor for breast cancer, although it is unknown whether reduction in breast density would reduce risk. Our risk model may be able to identify women at high risk for breast cancer for preventive interventions or more intensive surveillance.

Adult↗

PubMedDate not supplied

Novel control system for blood glucose using a model predictive method.

We developed a novel blood glucose control system, using a model predictive method, to achieve optimal control of the blood glucose level in severely diabetic or pancreatectomized patients. This system is designed to predict glucose level changes in advance, considering delayed response time and the administered doses of insulin. This method is also designed to calculate the most appropriate insulin infusion rate by considering differences in individual response to insulin. In this study, we compared our system with a conventional proportional and differential controller (PD controller) to determine whether the new system could regulate the glucose level efficiently in pancreatectomized dogs. The model predictive control method resulted in a significant reduction of mean insulin infusion rate compared with the conventional PD controller (0.71 mU/kg per min vs. 1.81 mU/kg per min, p = 0.0005), when the glucose level in both methods reached the planned target level (100 mg/dl). The new system also tended to have a reduced mean glucose infusion rate for compensating for overshooting of the glucose level compared with the PD controller (0.7 mg/kg per min vs. 1.1 mg/kg per min, p = 0.16). These results indicate that the new system should be a useful tool for regulating the glucose level in severely diabetic patients.

Animals↗

PubMed1992

Combining logistic regression and neural networks to create predictive models.

Neural networks are being used widely in medicine and other areas to create predictive models from data. The statistical method that most closely parallels neural networks is logistic regression. This paper outlines some ways in which neural networks and logistic regression are similar, shows how a small modification of logistic regression can be used in the training of neural network models, and illustrates the use of this modification for variable selection and predictive model building with neural networks.

Algorithms↗

PubMed2005

A fully predictive model for one-dimensional light attenuation by Chlamydomonas reinhardtii in a torus photobioreactor.

The light attenuation in a photobioreactor is determined using a fully predictive model. The optical properties were first calculated, using a data bank of the literature, from only the knowledge of pigments content, shape, and size distributions of cultivated cells which are a function of the physiology of the current species. The radiative properties of the biological turbid medium were then deduced using the exact Lorenz-Mie theory. This method is experimentally validated using a large-size integrating sphere photometer. The radiative properties are then used in a rectangular, one-dimensional two-flux model to predict radiant light attenuation in a photobioreactor, considering a quasi-collimated field of irradiance. Combination of this radiative model with the predictive determination of optical properties is finally validated by in situ measurement of attenuation profiles in a torus photobioreactor cultivating the microalgae Chlamydomonas reinhardtii, after a complete and proper characterization of the incident light flux provided by the experimental set-up.

Animals↗

PubMed2001

Performance of a predictive model for streptococcal pharyngitis in children.

CONTEXT: Group A beta-hemolytic streptococcus (GABHS) pharyngitis is a common childhood illness. The clinical diagnosis is difficult to determine and laboratory tests have limitations; hence, the condition is generally overdiagnosed and overtreated. Several clinical pediatric-specific predictive models have been published but none have been prospectively studied. OBJECTIVE: To test the performance of a previously published predictive model for GABHS pharyngitis in children in different clinical settings and during different seasons. DESIGN: Prospective cohort study. SETTINGS: Pediatric emergency department and 2 pediatric outpatient clinics. PATIENTS: Children aged between 1 and 18 years with pharyngitis on initial examination at study sites between April 1, 1999, and March 31, 2000. INTERVENTIONS: Recording of clinical features during initial evaluation using a standardized form and recovery of GABHS from patients' throats using reference standard methods. MAIN OUTCOME MEASURES: Posttest probability for GABHS positive throat culture associated with the model's positive predictors (moderate to severe tonsillar swelling, cervical lymphadenopathy [moderate to severe tenderness and enlargement of cervical lymph nodes], scarletiniform rash, and the absence of coryza) and the models' negative predictors (absence of the above signs and the presence of coryza). RESULTS: Of 587 patients analyzed, 218 (37%) had a positive throat culture for GABHS. Forty-nine percent were boys. Mean +/- SD age was 6.7 +/- 3.9 years. There was no difference between the subsets within the sample. The posttest probability values for a positive throat culture associated with positive and negative predictors of the model were 79% and 12%, respectively. CONCLUSIONS: A pediatric predictive model for GABHS pharyngitis performed better than physicians' subjective estimates for a positive throat culture and was comparable with a rapid antigen detection test. The model performed consistently well in different populations and across seasons. It can be useful if reliable microbiological testing and/or follow-up are not attainable.

Adolescent↗

PubMed2001

Validation of a prediction model and its predictors for the histology of residual masses in nonseminomatous testicular cancer.

PURPOSE: We validated a prediction model for histology of residual retroperitoneal masses, either benign or tumor, in patients treated with chemotherapy for metastatic nonseminomatous testicular cancer. MATERIALS AND METHODS: We studied 276 patients treated with chemotherapy before retroperitoneal lymph node dissection at Indiana University Medical Center between 1985 and 1999. A previously developed prediction model was modified to provide predictions for the Indiana population based on 5 predictors. For these predictors, including teratomatous elements in the primary tumor, pre-chemotherapy tumor markers (alpha-fetoprotein and human chorionic gonadotropin), size of the residual mass and reduction in mass size, univariate and multivariate odds ratios were determined. The modified model was evaluated by calculating the concordance statistic and studying model reliability. RESULTS: All odds ratios from univariate and multivariate analyses were in the expected directions. The modified model had good discriminative ability (concordance statistic 0.79). However, the predicted probabilities for benign tissue were generally too high due to the low prevalence of benign tissue (76 of 276 cases or 28%). CONCLUSIONS: This study confirms the predictive ability of formerly identified predictors for the histology of residual retroperitoneal masses in testicular cancer. However, the previously developed prognostic model must be adjusted for the local overall ratio of benign versus tumor histology to provide reliable predictions in the Indiana population.

Antineoplastic Agents↗

PubMed2003

Application of the CFU-GM assay to predict acute drug-induced neutropenia: an international blind trial to validate a prediction model for the maximum tolerated dose (MTD) of myelosuppressive xenobiotics.

In a previous study of prevalidation, a standard operating procedure (SOP) for two independent in vitro tests (human and mouse) had been developed, to evaluate the potential hematotoxicity of xenobiotics from their direct and the adverse effects on granulocyte-macrophages (CFU-GM). A predictive model to calculate the human maximum tolerated dose (MTD) was set up, by adjusting a mouse-derived MTD for the differential interspecies sensitivity. In this paper, we describe an international blind trial designed to apply this model to the clinical neutropenia, by testing 20 drugs, including 14 antineoplastics (Cytosar-U, 5-Fluorouracil, Myleran, Thioguanine, Fludarabine, Bleomycin, Methotrexate, Gemcitabine, Carmustine, Etoposide, Teniposide, Cytoxan, Taxol, Adriamycin); two antivirals (Retrovir, Zovirax,); three drugs for other therapeutic indications (Cyclosporin, Thorazine, Indocin); and one pesticide (Lindane). The results confirmed that the SOP developed generates reproducible IC90 values with both human and murine GM-CFU. For 10 drugs (Adriamycin, Bleomycin, Etoposide, Fludarabine, 5-Fluorouracil, Myleran, Taxol, Teniposide, Thioguanine, and Thorazine), IC90 values were found within the range of the actual drug doses tested (defined as the actual IC90). For the other 10 drugs (Carmustine, Cyclosporin, Cytosar-U, Cytoxan, Gemcitabine, Indocin, Lindane, Methotrexate, Retrovir, and Zovirax) extrapolation on the regression curve out of the range of the actual doses tested was required to derive IC90 values (extrapolated IC90). The model correctly predicted the human MTD for 10 drugs out of 10 that had "actual IC90 values" and 7 drugs out of 10 for those having only an extrapolated IC90. Two of the incorrect predictions (Gemcitabine and Zovirax) were within 6-fold of the correct MTD, instead of the 4-fold range required by the model, whereas the prediction with Cytosar-U was approximately 10-fold in error. A possible explanation for the failure in the prediction of these three drugs, which are pyrimidine analogs, is discussed. We concluded that our model correctly predicted the human MTD for 20 drugs out of 23, since the other three drugs (Topotecan, PZA, and Flavopiridol) were tested in the prevalidation study. The high percentage of predicitivity (87%), as well as the reproducibility of the SOP testing, confirm that the model can be considered scientifically validated in this study, suggesting promising applications to other areas of research in developing validated hematotoxicological in vitro methods.

Acute Disease↗

PubMed2005

Accuracy of prediction models in the context of disease management.

There has been a significantly increased interest in the adoption of prediction modeling by many disease and case management programs to risk stratify members in order to optimize the utilization of available clinical resources. Before adopting any prediction model, it is critical to understand how to evaluate the model's accuracy. This paper explains the basic concepts of prediction accuracy, the relevant parameters, their drawbacks, and their interpretations. It also introduces a new accuracy parameter termed "cost concentration," which indicates the model accuracy more explicitly in the context of disease management.

Disease Management↗

PubMed2002

Triage scoring systems, severity of illness measures, and mortality prediction models in pediatric trauma.

Trauma triage scores, severity of illness measures, and mortality prediction models quantitate severity of injury and stratify patients according to a specified outcome. Triage scoring systems are typically used to assist prehospital personnel determine which patients require trauma center care, but they are not recommended as the sole determinant of triage. Severity of illness measures and mortality prediction models are used in clinical and health services research for risk-adjusted outcomes analyses and institutional benchmarking. As clinicians and researchers, it is imperative that we be knowledgeable of the methodologies and applications of these scoring and risk prediction systems to ensure their quality and appropriate utilization.

Benchmarking↗

PubMed2004

Predictive models of safety belt use: a regression analysis of MVOSS data.

A substantial portion of the U.S. population fails to regularly use their safety belts. The explanations for the differential belt use have addressed, for example, socioeconomics, state law, attitudes, and perceived likelihood of being ticketed. The current analyses create predictive models of safety belt use. Using NHTSA's Motor Vehicle Occupant Safety Surveys (Years 1998 and 2000; N = 9577), variables related to belt use were entered into backward stepwise logistic regressions to produce two predictive models (Demographic and Attitudinal) of safety belt use (Always versus Not always). The results indicated that belt use is a complicated issue as there were several interactions between variables. The Demographic predictive model contained main effects for, law types, socioeconomics, population density, a gender-law type interaction, and a three-way interaction between age, marital status, and vehicle type. The Attitudinal model included perceived effectiveness of the belt, fatalistic attitudes, and an interaction between perceived effectiveness of the belt and perceived risk of being ticketed. These models survived a multinomial logistic regression when belt use was parsed into three categories (Always, Part-time, and Infrequent). In addition to variables that affect belt use, the results suggested that the structure of "belt use" as a psychological/behavioral construct is more complicated than once thought. Specifically, a dichotomous breakdown of belt use (Always and Not always) oversimplifies the construct because the predictor factors sometimes affect "part-time" belt users differently than "infrequent" belt users (compared to "full-time" users). Many of the factors included in the models have been previously shown to impact belt use, but the interaction effects--indicating a more complicated relationship between these variables than previously suggested--may contribute to a better understanding of safety belt use.

Adolescent↗

PubMed2004

Data evaluations and quantitative predictive models for vapor pressures of polycyclic aromatic hydrocarbons at different temperatures.

Polycyclic aromatic hydrocarbons (PAHs) are typical and ubiquitous organic pollutants. Vapor pressures, which can be classified as solid vapor pressure (P(S)) and (subcooled) liquid vapor pressure (P(L)), are key physicochemical properties governing the environmental fate of organic pollutants. It is of great importance to develop predictive models of vapor pressures. In the present study, partial least squares (PLS) regression together with 15 theoretical molecular structural descriptors was used to develop quantitative predictive models for vapor pressures of PAHs at different temperatures. Two procedures were adopted to develop the optimal predictive models by eliminating redundant molecular structural descriptors. The cross-validated Q2(cum) values for the obtained models have been found higher than 0.975, indicating good predictive ability and robustness of the models. It has been shown that the intermolecular dispersive interactions played a leading role in governing the values of log P(L). In addition to dispersive interactions, dipole-dipole interactions also played a secondary role in determining the magnitude of log P(S). In view of the scarceness of chemical standards for some PAHs, the difficulty in experimental determinations, and the high cost involved in experimental determinations, the obtained models should serve as a fast and simple first approximation of the vapor pressure values for PAHs at different environmental temperatures.

Environmental Pollutants↗

PubMed2005

Establishing a prediction model for coronary angiography based on coronary risk factors.

The aim of the present study was to establish an evidence-based effective prediction model for improving the accuracy and priority for undertaking coronary angiography. The sample population consisted of 2002 coronary angiography patients. Our data were taken from claim forms provided by the Taiwanese Bureau of National Health Insurance. The results were tested using chi-square automatic interaction detection to establish a prediction model using coronary risk factors. We found significant variation across homogeneous groups, with the probabilities of developing coronary heart disease (CHD) varying according to risk factors such as sex, hypertension, diabetes, age, and physical inactivity. The study also explored the influence of interactions among patient characteristics. The sensitivity, specificity, and positive predictive value of our study were 92.0%, 35.4%, and 76.5% respectively, indicating the diagnostic accuracy of the model is at least as high as the treadmill exercise test. The results suggest that the accuracy of a decision concerning the performance of cardiac angiography can be significantly enhanced by an evidence-based effective prediction model that takes interactions between risk factors into account. This model also helps to priortize patients waiting to undergo coronary angiography.

Aged↗

PubMed2006

Do anxiety and depression cluster into distinct groups?: a test of tripartite model predictions in a community sample of youth.

Because of their high comorbidity and strong associations, the distinctiveness of anxiety and depression in youth continues to be debated. In this study we used cluster analysis in a community sample (n=225) of youth to test tripartite model predictions regarding the grouping of individuals based on their levels of anxiety and depression symptoms. Findings were consistent with tripartite model predictions that four groups would emerge (primarily elevated on anxiety symptoms only, elevated on depression symptoms only, elevated on both anxiety and depression symptoms, and a low symptom group). Analyses using specific tripartite model variables and parent report of internalizing symptoms provided additional support for the groupings and tripartite model predictions. Across age groupings, the clustering of anxiety and depression symptoms was consistent with some hypothesized developmental differences in the expression of internalizing symptoms in youth. Findings add support for the tripartite model in youth, and support the idea that anxiety and depression do represent unique syndromes in youth. Depression and Anxiety 23:453-460, 2006. Published 2006 Wiley-Liss, Inc.

Adolescent↗

PubMed2006

The role of predictive models in the formation of auditory streams.

Sounds provide us with useful information about our environment which complements that provided by other senses, but also poses specific processing problems. How does the auditory system distentangle sounds from different sound sources? And what is it that allows intermittent sound events from the same source to be associated with each other? Here we review findings from a wide range of studies using the auditory streaming paradigm in order to formulate a unified account of the processes underlying auditory perceptual organization. We present new computational modelling results which replicate responses in primary auditory cortex [Fishman, Y.I., Arezzo, J.C., Steinschneider, M., 2004. Auditory stream segregation in monkey auditory cortex: effects of frequency separation, presentation rate, and tone duration. J. Acoust. Soc. Am. 116, 1656-1670; Fishman, Y. I., Reser, D. H., Arezzo, J.C., Steinschneider, M., 2001. Neural correlates of auditory stream segregation in primary auditory cortex of the awake monkey. Hear. Res. 151, 167-187] to tone sequences. We also present the results of a perceptual experiment which confirm the bi-stable nature of auditory streaming, and the proposal that the gradual build-up of streaming may be an artefact of averaging across many subjects [Pressnitzer, D., Hupé, J. M., 2006. Temporal dynamics of auditory and visual bi-stability reveal common principles of perceptual organization. Curr. Biol. 16(13), 1351-1357.]. Finally we argue that in order to account for all of the experimental findings, computational models of auditory stream segregation require four basic processing elements; segregation, predictive modelling, competition and adaptation, and that it is the formation of effective predictive models which allows the system to keep track of different sound sources in a complex auditory environment.

Acoustic Stimulation↗

PubMed1996

Precipitating factors for delirium in hospitalized elderly persons. Predictive model and interrelationship with baseline vulnerability.

OBJECTIVES: To prospectively develop and validate a predictive model for delirium based on precipitating factors during hospitalization, and to examine the interrelationship of precipitating factors and baseline vulnerability. DESIGN: Two prospective cohort studies, in tandem. SETTING: General medical wards, university teaching hospital. PATIENTS: For the development cohort, 196 patients aged 70 years and older with no delirium at baseline, and for the validation cohort, 312 comparable patients. MAIN OUTCOME MEASURE: New-onset delirium by hospital day 9, defined by the Confusion Assessment Method diagnostic criteria. RESULTS: Delirium developed in 35 patients (18%) in the development cohort. Five independent precipitating factors for delirium were identified; use of physical restraints (adjusted relative risk [RR], 4.4; 95% confidence interval [CI], 2.5 to 7.9), malnutrition (RR, 4.0; 95% CI, 2.2 to 7.4), more than three medications added (RR, 2.9; 95% CI, 1.6 to 5.4), use of bladder catheter (RR, 2.4; 95% CI, 1.2 to 4.7), and any iatrogenic event (RR, 1.9; 95% CI, 1.1 to 3.2). Each precipitating factor preceded the onset of delirium by more than 24 hours. A risk stratification system was developed by adding 1 point for each factor present. Rates of delirium for low-risk (0 points), intermediate-risk (1 to 2 points), and high-risk groups (> or equal to 3 points) were 3%, 20%, and 59%, respectively (P < .001). The corresponding rates in the validation cohort, in which 47 patients (15%) developed delirium, were 4%, 20%, and 35%, respectively (P < .001). When precipitating and baseline factors were analyzed in cross-stratified format, delirium rates increased progressively from low-risk to high-risk groups in all directions (double-gradient phenomenon). The contributions of baseline and precipitating factors were documented to be independent and statistically significant. CONCLUSIONS: A simple predictive model based on the presence of five precipitating factors can be used to identify elderly medical patients at high risk for delirium. Precipitating and baseline vulnerability factors are highly interrelated and contribute to delirium in independent substantive, and cumulative ways.

Aged↗

PubMed2004

Artificial neural networks for predictive modeling in prostate cancer.

Artificial neural networks (ANNs) represent a relatively new methodology for predictive modeling in medicine. ANNs, a form of artificial intelligence loosely based on the brain, have a demonstrated ability to learn complex and subtle relationships between variables in medical applications. In contrast with traditional statistical techniques, ANNs are capable of automatically resolving these relationships without the need for a priori assumptions about the nature of the interactions between variables. As with any technique, ANNs have limitations and potential drawbacks. This article provides an overview of the theoretical basis of ANNs, how they function, their strengths and limitations, and examples of how ANNs have been used to develop predictive models for the management of prostate cancer.

Diagnosis, Differential↗

PubMed2004

Nonlinear model predictive control of glucose concentration in subjects with type 1 diabetes.

A nonlinear model predictive controller has been developed to maintain normoglycemia in subjects with type 1 diabetes during fasting conditions such as during overnight fast. The controller employs a compartment model, which represents the glucoregulatory system and includes submodels representing absorption of subcutaneously administered short-acting insulin Lispro and gut absorption. The controller uses Bayesian parameter estimation to determine time-varying model parameters. Moving target trajectory facilitates slow, controlled normalization of elevated glucose levels and faster normalization of low glucose values. The predictive capabilities of the model have been evaluated using data from 15 clinical experiments in subjects with type 1 diabetes. The experiments employed intravenous glucose sampling (every 15 min) and subcutaneous infusion of insulin Lispro by insulin pump (modified also every 15 min). The model gave glucose predictions with a mean square error proportionally related to the prediction horizon with the value of 0.2 mmol L(-1) per 15 min. The assessment of clinical utility of model-based glucose predictions using Clarke error grid analysis gave 95% of values in zone A and the remaining 5% of values in zone B for glucose predictions up to 60 min (n = 1674). In conclusion, adaptive nonlinear model predictive control is promising for the control of glucose concentration during fasting conditions in subjects with type 1 diabetes.

Blood Glucose↗