SEARCH · PubMed Health

Results for “Prognostic model”

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 19 recordsLinked to original sources

PubMed2000

What do we mean by validating a prognostic model?

Prognostic models are used in medicine for investigating patient outcome in relation to patient and disease characteristics. Such models do not always work well in practice, so it is widely recommended that they need to be validated. The idea of validating a prognostic model is generally taken to mean establishing that it works satisfactorily for patients other than those from whose data it was derived. In this paper we examine what is meant by validation and review why it is necessary. We consider how to validate a model and suggest that it is desirable to consider two rather different aspects - statistical and clinical validity - and examine some general approaches to validation. We illustrate the issues using several case studies.

Accidental Falls↗

PubMed1999

An evolutionary approach to constructing prognostic models.

A prognostic model is sought to determine whether or not patients suffering from an uncommon form of cancer will survive. Given a set of case histories, we attempt to find the relative weightings of the different variables that are used to describe the cases. Our first innovation is to use a diffusion genetic algorithm (DGA) to find weightings which will give optimal survival predictions. The DGA enables a number of criteria to be satisfied simultaneously, making it particularly suitable for model building. A further innovation is a method of representing synergies between interacting factors. The evolved model correctly predicts 90% of the survivors and 87% of deaths, an improvement over the current model. More significantly, the method enables a simple model to be evolved, one that produces well-balanced predictions, and one that is relatively easy for clinicians to use. The method was validated by running it on a training set made up of 90% of the original database and then studying the performance of the generated models on a test set consisting of the remaining 10% of the cases.

Adolescent↗

PubMed2002

Validity of prognostic models: when is a model clinically useful?

Prognostic models combine patient characteristics to predict medical outcomes. Unfortunately, such models do not always perform as well for other patients as those from whose data the models were derived. Therefore, validity of prognostic models needs to be assessed in new patients. Predicted probabilities can be calculated with the model and compared with the actually observed outcomes. We may distinguish several aspects of validity: (1) agreement between predicted probabilities and observed probabilities (calibration), (2) ability of the model to distinguish subjects with different outcomes (discrimination), and (3) ability of the model to improve the decision-making process (clinical usefulness). We discuss those aspects and show some measures by using models for testicular and prostate cancer. We conclude that good calibration and discriminative ability are not sufficient for a model to be clinically useful. Application of a prognostic model is sensible, if the model is able to provide useful additional information for clinical decision making.

Decision Support Techniques↗

PubMed2005

Survival following the development of ascites and/or peripheral oedema in primary biliary cirrhosis: a staged prognostic model.

OBJECTIVE: Current prognostic models in primary biliary cirrhosis (PBC) have low precision, partly due to the restricted inclusion criteria of some cohorts used for modelling but also because of the prolonged natural course of the disease. It is hypothesized that better precision could be achieved with a staged model, using ascites or peripheral oedema as a new starting-point for prediction. MATERIAL AND METHODS: The study was based on an established database of 289 consecutive patients, followed between 1977 and 1998. Stepwise Cox regression was used to construct a staged model based on 143 patients who first developed ascites (n=111) or peripheral oedema (n=32) at entry or during subsequent follow-up. The model was compared with published models using graphical methods and receiver operating characteristics (ROCs). RESULTS: Mean time from clinical diagnosis of ascites or peripheral oedema to death was 3.1 years. The following variables had independent prognostic significance: log10(bilirubin) (p<0.001), albumin (p<0.001), age (p<0.001) and history of encephalopathy (p<0.001). Goodness of fit showed that the survival probabilities predicted by the Ascites Stage Model fitted well with the observed data. The Ascites Stage Model (ROC 0.8324 (SE 0.0348)) was a better predictor of survival than the Mayo long-term model (ROC 0.7833 (SE 0.0397)), the Mayo Repeated Patient Visits Model (ROC 0.7779 (SE 0.0399)) and the Royal Free PBC Prognostic Model (ROC 0.7785 (SE 0.0396)). CONCLUSIONS: The Ascites Stage Model gives a better survival estimate for PBC patients once they have developed ascites or peripheral oedema compared with the current models, and demonstrates an advantage of staged models in diseases with a prolonged natural history.

Ascites↗

PubMed2005

Performance of prognostic models in critically ill cancer patients - a review.

INTRODUCTION: Prognostic models, such as the Acute Physiology and Chronic Health Evaluation (APACHE) II or III, the Simplified Acute Physiology Score (SAPS) II, and the Mortality Probability Models (MPM) II were developed to quantify the severity of illness and the likelihood of hospital survival for a general intensive care unit (ICU) population. Little is known about the performance of these models in specific populations, such as patients with cancer. Recently, specific prognostic models have been developed to predict mortality for cancer patients who are admitted to the ICU. The present analysis reviews the performance of general prognostic models and specific models for cancer patients to predict in-hospital mortality after ICU admission. METHODS: Studies were identified by searching the Medline databases from 1994 to 2004. We included studies evaluating the performance of mortality prediction models in critically ill cancer patients. RESULTS: Ten studies were identified that evaluated prognostic models in cancer patients. Discrimination between survivors and non-survivors was fair to good, but calibration was insufficient in most studies. General prognostic models uniformly underestimate the likelihood of hospital mortality in oncological patients. Two versions of a specific oncological scoring systems (Intensive Care Mortality Model (ICMM)) were evaluated in five studies and showed better discrimination and calibration than the general prognostic models. CONCLUSION: General prognostic models generally underestimate the risk of mortality in critically ill cancer patients. Both general prognostic models and specific oncology models may reliably identify subgroups of patients with a very high risk of mortality.

APACHE↗

PubMed2001

Effects of case removal in prognostic models.

Constructing and updating prognostic models that learn from training cases is a time-consuming task. The more compact, and yet informative, the training sets are, the faster one can build and properly evaluate such models. We have compared different regression diagnostic methods for selection and removal of training cases in prognostic models. Univariate determinations were performed using classical regression diagnostic statistics. Multivariate determinations were performed using (1) a sequential "backward" selection of cases, and (2) a non-sequential genetic algorithm. The genetic algorithm produced final models that kept few cases and retained predictive capability. A genetic algorithm approach to case selection may be better suited for guiding removal of cases in training sets than a univariate or a sequential multivariate approach, possibly because of its ability to detect sets of cases that are influential en bloc but may not be sufficiently influential when considered in isolation.

Algorithms↗

PubMed2002

Statistical characteristics of area under the receiver operating characteristic curve for a simple prognostic model using traditional and bootstrapped approaches.

Prognostic models are increasingly common in the biomedical literature. These models are frequently evaluated with respect to their ability to discriminate between those with and without an outcome. The area under the receiver-operating curve (AROC) is often used to assess discrimination. In this study, we introduce a bootstrap method, and, using Monte Carlo simulation, we compare three different bootstrap approaches with four commonly used methods in their ability to accurately estimate 95% confidence intervals (CIs) around the AROC for a simple prognostic model. We also evaluated the power of a bootstrap method and the commonly used trapezoid rule to compare different prognostic models. We show that several good methods exist for calculating 95% CIs of AROC, but the maximum likelihood estimation method should not be used with small sample sizes. We further show that for our simple prognostic model a bootstrap z-statistic approach is preferred over the trapezoidal method when comparing the AROCs of two related models.

Aged↗

PubMed2001

Systematic review of prognostic models in patients with acute stroke.

Prognostic models in stroke may be useful in clinical practice and research. We systematically reviewed the methodology and results of studies that have identified independent predictors of survival, independence in activities of daily living, and getting home in patients with acute stroke. Eligible studies (published in full in English) included at least 100 patients in whom at least 3 predictor variables were assessed within 30 days of stroke onset and who were followed up for at least 30 days. We recorded 25 indicators of the validity and practicality of each model and identified variables that were consistent independent predictors of each outcome. Eighty-three separate prognostic models were found but most had potentially serious deficiencies in internal and statistical validity, many had limited generalisability, and none had been adequately validated. Only 4 studies met 8 simple quality criteria. Over 150 different predictor variables have been analysed but most were assessed in only 1 or 2 models. None of the existing prognostic models have been sufficiently well developed and validated to be useful in either clinical practice or research. Better quality models must be produced to enable, for example, adequate case-mix correction when comparing outcome among different groups of stroke patients.

Humans↗

PubMed1990

Validation of prognostic models in primary biliary cirrhosis.

Prognostic models in primary biliary cirrhosis have been validated for large population groups but the predictive value for individual patients has not been tested. We used data from ten deceased patients with primary biliary cirrhosis to test three prognostic models: the Shapiro model (bilirubin); the Christensen model (age, bilirubin, albumin, presence of cirrhosis or cholestasis, azathioprine treatment); and the Dickson model (age, bilirubin, albumin, prothrombin time, oedema). The predictive value of each model for individual patients was determined by assessing whether it would have accurately predicted appropriate timing of liver transplantation in patients prior to death. The Dickson model predicted that four of nine cases would have been considered for liver transplantation one year before death and one of seven cases two years before death. The Christensen model predicted that this procedure would have been considered in three of seven cases two years before death. The Shapiro model was demonstrated to be the least predictive of the three tested. Although none of the three models assessed was found to accurately predict survival, no model predicted a worse survival than actually occurred. Liver transplantation is indicated in those cases with a poor predicted survival.

Aged↗

PubMed1999

Prognostic models: evidence-based approach to predicting disease outcome.

BACKGROUND: A thorough understanding of the techniques underlying prognostic modelling, the potential shortcomings of prognostic models, and their applicability to dermatology are important for both clinicians and researchers. OBJECTIVE: This article distinguishes between prognostic and explanatory (causal) models, discusses some of the techniques used in developing a prognostic model, and addresses the importance of model generalizability. CONCLUSION: Prognostic models may be useful to the clinician, but must be used with care. It is important to critically appraise prognostic models and to assure that they are relevant to the population of interest. The ability to critically appraise prognostic models is predicated on a thorough understanding of the techniques used in their development and evaluation.

Decision Support Techniques↗

PubMed2006

Long-term survival after non-small cell lung cancer surgery: development and validation of a prognostic model with a preoperative and postoperative mode.

OBJECTIVE: At present, there is no prognostic model that is specific for prediction of survival after non-small cell lung cancer surgery. We aimed to develop a prognostic model that can be used to estimate the postoperative survival of individual patients. METHODS: A total of 766 patients underwent resection for primary non-small cell lung cancer. Comorbid conditions were scaled according to the Charlson comorbidity index (CCI). Cox proportional hazard analyses were used to determine risk factors for survival. A prognostic model for survival with a preoperative and postoperative mode was established. Performance of the prognostic model, the CCI, and pathologic tumor stage were quantified by a concordance statistic to indicate discriminative ability. RESULTS: The factors associated with an impaired survival were male sex, age, chronic obstructive pulmonary disease, congestive heart failure, any prior tumor, moderate-to-severe renal disease (preoperative and postoperative mode), clinical tumor stage (preoperative mode), type of resection, and pathologic tumor stage (postoperative mode). The discriminative performance was poor for the CCI (c = 0.55), better for pathologic tumor stage (c = 0.60) and for the preoperative mode (c = 0.61), and best for the postoperative mode (c = 0.65). The discriminative performance of the postoperative mode was better than the discriminative performance of the CCI (P < .0001), the preoperative mode (P < .0002), and pathologic tumor stage (P < .0001). The discriminative performance of the preoperative mode was better than the discriminative performance of the CCI (P < .0001) and similar (P = .90) to a model that only included pathologic tumor stage. CONCLUSIONS: The prognostic model, particularly the postoperative mode, successfully estimates long-term survival of individual patients and could help clinicians in clinical decision-making and treatment tailoring.

Aged↗

PubMed2004

The importance of accounting for the uncertainty of published prognostic model estimates.

OBJECTIVES: Reported is the importance of properly reflecting uncertainty associated with prognostic model estimates when calculating the survival benefit of a treatment or technology, using liver transplantation as an example. METHODS: Monte Carlo simulation techniques were used to account for the uncertainty of prognostic model estimates using the standard errors of the regression coefficients and their correlations. These methods were applied to patients with primary biliary cirrhosis undergoing liver transplantation using a prognostic model from a historic cohort who did not undergo transplantation. The survival gain over 4 years from transplantation was estimated. RESULTS: Ignoring the uncertainty in the prognostic model, the estimated survival benefit of liver transplantation was 16.7 months (95 percent confidence interval [CI], 13.5 to 20.1), and was statistically significant (p < .001). After adjusting for model uncertainty using the standard errors of the regression coefficients, the estimated survival benefit was 17.5 months (95 percent CI, -3.9 to 38.5) and was no longer statistically significant. An additional adjustment for the correlation between regression coefficients widened the 95 percent confidence interval slightly: the estimated survival benefit was 17.0 months (95 percent CI: -4.6 to 38.6). CONCLUSIONS: It is important that the precision of regression coefficients is available for users of published prognostic models. Ignoring this additional information substantially underestimates uncertainty, which can then impact misleadingly on policy decisions.

Cohort Studies↗

PubMed2004

Predicting outcome after acute ischemic stroke: an external validation of prognostic models.

OBJECTIVE: To externally validate two prognostic models predicting functional outcome and survival 100 days after acute ischemic stroke. METHODS: Using prospectively collected data from 1,470 patients, the authors evaluated two previously developed models. Model I predicts incomplete functional recovery (Barthel Index <95) vs complete functional recovery with 11 variables, whereas model II predicts mortality vs survival with 3 variables. On admission to a participating hospital, patients were registered prospectively and included according to defined criteria. Within 72 hours, predictive variables under investigation were assessed. Follow-up was performed 100 days after the event. RESULTS: Model I correctly predicted 68.1% of the patients who had incompletely recovered or had died and 85.7% of the completely recovered patients, model II 46.9% of the patients who had died and 95.9% of the surviving patients. Both models performed better than the treating physicians' predictions made within 72 hours after admission. CONCLUSION: The resulting prognostic models are useful to correctly stratify treatment groups in clinical trials and to accurately predict the distribution of endpoint variables.

Aged↗

PubMed2004

What defines intermediate-risk prostate cancer? Variability in published prognostic models.

PURPOSE: To assess the efficacy of a variety of prognostic models in the definition of intermediate-risk prostate cancer and to compare them to our own empiric model. METHODS AND MATERIALS: Two hundred fifty-six consecutive men with prostate adenocarcinoma treated with external beam radiotherapy alone were studied. Biochemical failure (defined as 3 consecutive PSA rises or the initiation of androgen deprivation therapy) was examined using univariate, multivariate, and recursive partitioning analyses. The risk classification model used in our department was then compared to a number of published models to assess the relative performance of each in discriminating risk groups. RESULTS: At a median follow-up of 62.4 months, the 5-year Biochemical failure-free survival (bFFS) was 46.8% for the overall group. This relates to 5-year bFFS of 77.8%, 51.1%, and 33.8% based on our institutional criteria for low-, intermediate-, and high-risk features, respectively. All the models examined showed an outcome group with a comparatively similar poor outcome when applied to our data. Large variation was seen in the intermediate-risk groups, with 5-year bFFS ranging from 38.1% to 51.1%. Good risk categories had similar large variations. All published models showed inability to delineate three significantly different outcome groups. Recursive partitioning analysis derived categories based on combinations of PSA (with cutpoints at 42.4, 20, and 10.6 ng/mL) and Gleason score (with cutpoints at 2-6 and 7-10) only. CONCLUSIONS: Large variations in the relative performance of a number of prognostic models are shown when applied to our local data. The prognostic efficacy of PSA and biopsy Gleason score is reiterated, although other factors will need to be explored to further improve the performance of prognostic models, particularly in defining the intermediate-risk subset of prostate cancer.

Adult↗

PubMed2004

Development and validation of a prognostic model for survival time data: application to prognosis of HIV positive patients treated with antiretroviral therapy.

The process of developing and validating a prognostic model for survival time data has been much discussed in the literature. Assessment of the performance of candidate prognostic models on data other than that used to fit the models is essential for choosing a model that will generalize well to independent data. However, there remain difficulties in current methods of measuring the accuracy of predictions of prognostic models for censored survival time data. In this paper, flexible parametric models based on the Weibull, loglogistic and lognormal distributions with spline smoothing of the baseline log cumulative hazard function are used to fit a set of candidate prognostic models across k data sets. The model that generalizes best to new data is chosen using a cross-validation scheme which fits the model on k-1 data sets and tests the predictive accuracy on the omitted data set. The procedure is repeated, omitting each data set in turn. The quality of the predictions is measured using three different methods: two commonly proposed validation methods, Harrell's concordance statistic and the Brier statistic, and a novel method using deviance differences. The results show that the deviance statistic is able to discriminate between quite similar models and can be used to choose a prognostic model that generalizes well to new data. The methods are illustrated by using a model developed to predict progression to a new AIDS event or death in HIV-1 positive patients starting antiretroviral therapy.

Anti-HIV Agents↗

PubMed1979

A prognostic model for head injury.

A prognostic model for head injured patients was developed. Patients fall into one of four prognostic categories at the end of the first hospital day: I. Discharge alive on or before the seventh hospital day. II. Discharge alive 8--42 days after admission. III. Discharge alive after 42 days, or dead after five years. IV. Discharge dead before five years. The outcome of 63% of the patients was predicted correctly. The model correctly placed, or missed by only one category, 97% of the patients observed. Errors tended to the optimistic side of observed outcomes. Age and pattern of consciousness were critical prognostic factors.

Adolescent↗

PubMed1998

Validation of a melanoma prognostic model.

BACKGROUND: A "clinically accessible," 4-variable (patient age, patient sex, tumor location, and tumour thickness) prognostic model has been published previously. This model evaluated variables that were commonly available to the clinician. Because models are heuristic, validity of a prognostic model should be evaluated in a population different from the original population. OBJECTIVE: To evaluate the external validity of this 4-variable melanoma prognostic model. DESIGN: To estimate the external validity of this model, we used a population-based cohort of individuals with melanoma. We also evaluated a 1-variable model (tumor thickness). Estimates of the external validity of these logistic regression models were made using the c statistic and the Brier score. SETTINGS AND PATIENTS: A total of 1261 patients with melanoma evaluated in a multispecialty, university-based practice and 650 patients with melanoma from throughout Connecticut. MAIN OUTCOME MEASURE: Death from melanoma within 5 years of diagnosis. RESULTS: The c statistics for the 4-variable model were 0.86 (95% confidence interval [CI], 0.83-0.89) for the university-based practice data set and 0.81 (95% CI, 0.75-0.86) for the Connecticut data set. For thickness alone, the c statistics were 0.83 (95% CI, 0.80-0.86) and 0.79 (95% CI, 0.74-0.85), respectively. Brier scores for the 4-variable model were 0.09 (95% CI, 0.08-0.10) and 0.08 (95% CI, 0.06-0.09) and for the 1-variable model were 0.09 (95% CI, 0.08-0.10) and 0.08 (95% CI, 0.07-0.10), respectively. No significant differences exist between the data sets for the 4- and 1-variable models. CONCLUSIONS: The 4- and 1-variable models are generalizable. The simpler 1-variable model--tumor thickness--can be used with a relatively small loss in accuracy.

Female↗

PubMed2006

Prognostic model for disease-specific and overall mortality in newly diagnosed symptomatic patients with Waldenstrom macroglobulinaemia.

We sought to define prognostic factors for survival in Waldenstrom macroglobulinaemia (WM). Of 585 patients diagnosed with WM and seen at Mayo Clinic between 1960 and 2001, 337 symptomatic patients met the inclusion criteria and were analysed for overall and disease-specific survival. The median survival from the time of diagnosis was 6.4 years. The median disease-specific survival was 11.2 years. Univariate analysis for overall survival identified the following adverse prognostic factors: age >65 years (P < 0.001), organomegaly (P < 0.001), elevated beta2-microglobulin (<0.001), anaemia (Hb < 10.0 g/dl) (P = 0.01), leucopenia (<4.0 x 10(9)/l) (P = 0.03), thrombocytopenia (<150 x 10(9)/l) (P = 0.01), serum albumin <40 g/l (P = 0.001), and quantitative IgM < 0.4 g/l (P = 0.04). On multivariate analysis, age >65 years and organomegaly were associated with poor prognosis. A prognostic model was built based on these two variables. Patients at high risk (1-2 risk factors, median survival 4.2 years) experienced worse survival than patients at low risk (0 risk factors, median survival 10.6 years), P < 0.001. The prognostic model was validated in 204 patients who were not included in the analysis cohort. Beta2-microglobulin > or =4 mg/l was associated with a threefold increase in the risk of death when added to the prognostic model. We describe a simple prognostic model for overall survival for newly diagnosed patients with WM.

Adult↗