In long-term care (LTC) homes in the province of Ontario, implementation of the Minimum Data Set (MDS) assessment and The Braden Scale for predicting pressure ulcer risk were occurring simultaneously. The purpose of this study was, using available data sources, to develop a bedside MDS-based scale to identify individuals under care at various levels of risk for developing pressure ulcers in order to facilitate targeting risk factors for prevention.
Data for developing the interRAI Pressure Ulcer Risk Scale (interRAI PURS) were available from 2 Ontario sources: three LTC homes with 257 residents assessed during the same time frame with the MDS and Braden Scale for Predicting Pressure Sore Risk, and eighty-nine Ontario LTC homes with 12,896 residents with baseline/reassessment MDS data (median time 91 days), between 2005-2007. All assessments were done by trained clinical staff, and baseline assessments were restricted to those with no recorded pressure ulcer. MDS baseline/reassessment samples used in further testing included 13,062 patients of Ontario Complex Continuing Care Hospitals (CCC) and 73,183 Ontario long-stay home care (HC) clients.
A data-informed Braden Scale cross-walk scale using MDS items was devised from the 3-facility dataset, and tested in the larger longitudinal LTC homes data for its association with a future new pressure ulcer, giving a c-statistic of 0.676. Informed by this, LTC homes data along with evidence from the clinical literature was used to create an alternate-form 7-item additive scale, the interRAI PURS, with good distributional characteristics and c-statistic of 0.708. Testing of the scale in CCC and HC longitudinal data showed strong association with development of a new pressure ulcer.
interRAI PURS differentiates risk of developing pressure ulcers among facility-based residents and home care recipients. As an output from an MDS assessment, it eliminates duplicated effort required for separate pressure ulcer risk scoring. Moreover, it can be done manually at the bedside during critical early days in an admission when the full MDS has yet to be completed. It can be calculated with established MDS instruments as well as with the newer interRAI suite instruments designed to follow persons across various care settings (interRAI Long-Term Care Facilities, interRAI Home Care, interRAI Palliative Care).
The area under the receiver operating characteristics curve (AUC of ROC) is a widely used measure of discrimination in risk prediction models. Routinely, the Mann–Whitney statistics is used as an estimator of AUC, while the change in AUC is tested by the DeLong test. However, very often, in settings where the model is developed and tested on the same dataset, the added predictor is statistically significantly associated with the outcome but fails to produce a significant improvement in the AUC. No conclusive resolution exists to explain this finding. In this paper, we will show that the reason lies in the inappropriate application of the DeLong test in the setting of nested models. Using numerical simulations and a theoretical argument based on generalized U-statistics, we show that if the added predictor is not statistically significantly associated with the outcome, the null distribution is non-normal, contrary to the assumption of DeLong test. Our simulations of different scenarios show that the loss of power because of such a misuse of the DeLong test leads to a conservative test for small and moderate effect sizes. This problem does not exist in cases of predictors that are associated with the outcome and for non-nested models. We suggest that for nested models, only the test of association be performed for the new predictors, and if the result is significant, change in AUC be estimated with an appropriate confidence interval, which can be based on the DeLong approach.
AUC; DeLong test; logistic regression; U-statistics; discrimination; risk prediction
It has become commonplace to use receiver operating curve (ROC) methodology to evaluate the incremental predictive accuracy of new markers in the presence of existing predictors. However, concerns have been raised about the validity of this practice. We have evaluated this issue in detail.
Simulations have been used that show clearly that use of risk predictors from nested models as data in subsequent tests comparing areas under the ROC curves of the models leads to grossly invalid inferences. Careful examination of the issue reveals two major problems: (1) the data elements are strongly correlated from case to case; and (2) the model that includes the additional marker has a tendency to interpret predictive contributions as positive information regardless of whether observed effect of the marker is negative or positive. Both of these phenomena lead to profound bias in the test.
We recommend strongly against the use of ROC methods derived from risk predictors from nested regression models to test the incremental information of a new marker.
This pilot study compared the risk predictive value of preoperative physiological capacity (PC: defined by gas exchange measured during cardiopulmonary exercise testing) with the ASA physical status classification in the same patients (n=32) undergoing major abdominal cancer surgery.
Uni- and multivariate logistic regression models were fitted to measurements of PC and ASA rank data determining their predictive value for postoperative morbidity. Receiver operating characteristic (ROC) curves were used to discriminate between the predictive abilities, exploring trade-offs between sensitivity and specificity.
Individual statistically significant predictors of postoperative morbidity included the ASA rank [P=0.038, area under the curve (AUC)=0.688, sensitivity=0.630, specificity=0.750] and three newly identified measures of PC: PAT (% predicted anaerobic threshold achieved, <75% vs ≥75%), ΔHR1 (heart rate response from rest to the anaerobic threshold), and HR3 (heart rate at the anaerobic threshold). A two-variable model of PC measurements (ΔHR1+PAT) was also shown to be statistically significant in the prediction of postoperative morbidity (P=0.023, AUC=0.826, sensitivity=0.813, specificity=0.688).
Three newly identified PC measures and the ASA rank were significantly associated with postoperative morbidity; none showed a statistically greater association compared with the others. PC appeared to improve predictive sensitivity. The potential for new unidentified measures of PC to predict postoperative outcomes remains unexplored.
assessment, preanaesthetic; complications, morbidity; measurement techniques, gas exchange metabolic; metabolism, oxygen consumption; oxygen uptake; risk; surgery, postoperative
Multiple regression models are used in a wide range of scientific disciplines and automated model selection procedures are frequently used to identify independent predictors. However, determination of relative importance of potential predictors and validating the fitted models for their stability, predictive accuracy and generalizability are often overlooked or not done thoroughly.
Using a case study aimed at predicting children with acute lymphoblastic leukemia (ALL) who are at low risk of Tumor Lysis Syndrome (TLS), we propose and compare two strategies, bootstrapping and random split of data, for ordering potential predictors according to their relative importance with respect to model stability and generalizability. We also propose an approach based on relative increase in percentage of explained variation and area under the Receiver Operating Characteristic (ROC) curve for developing models where variables from our ordered list enter the model according to their importance. An additional data set aimed at identifying predictors of prostate cancer penetration is also used for illustrative purposes.
Age is chosen to be the most important predictor of TLS. It is selected 100% of the time using the bootstrapping approach. Using the random split method, it is selected 99% of the time in the training data and is significant (at 5% level) 98% of the time in the validation data set. This indicates that age is a stable predictor of TLS with good generalizability. The second most important variable is white blood cell count (WBC). Our methods also identified an important predictor of TLS that was otherwise omitted if relying on any of the automated model selection procedures alone. A group at low risk of TLS consists of children younger than 10 years of age, without T-cell immunophenotype, whose baseline WBC is < 20 × 109/L and palpable spleen is < 2 cm. For the prostate cancer data set, the Gleason score and digital rectal exam are identified to be the most important indicators of whether tumor has penetrated the prostate capsule.
Our model selection procedures based on bootstrap re-sampling and repeated random split techniques can be used to assess the strength of evidence that a variable is truly an independent and reproducible predictor. Our methods, therefore, can be used for developing stable and reproducible models with good performances. Moreover, our methods can serve as a good tool for validating a predictive model. Previous biological and clinical studies support the findings based on our selection and validation strategies. However, extensive simulations may be required to assess the performance of our methods under different scenarios as well as check their sensitivity to a random fluctuation in the data.
To assess the value of psychosocial risk factors in discriminating between individuals at higher and lower risk of coronary heart disease, using risk prediction equations.
Prospective observational study.
5191 employed men aged 35 to 64 years and free of coronary heart disease at study enrolment
Main outcome measures
Area under receiver operating characteristic (ROC) curves for risk prediction equations including different risk factors for coronary heart disease.
During the first 10 years of follow up, 203 men died of coronary heart disease and a further 200 were admitted to hospital with this diagnosis. Area under the ROC curve for the standard Framingham coronary risk factors was 74.5%. Addition of “vital exhaustion” and psychological stress led to areas under the ROC curve of 74.5% and 74.6%, respectively. Addition of current social class and lifetime social class to the standard Framingham equation gave areas under the ROC curve of 74.6% and 74.9%, respectively. In no case was there strong evidence for improved discrimination of the model containing the novel risk factor over the standard model.
Consideration of psychosocial risk factors, including those that are strong independent predictors of heart disease, does not substantially influence the ability of risk prediction tools to discriminate between individuals at higher and lower risk of coronary heart disease.
cardiovascular disease; risk assessment; Framingham risk score; primary prevention; psychosocial factors
Prediction of the binding ability of antigen peptides to major histocompatibility complex (MHC) class II molecules is important in vaccine development. The variable length of each binding peptide complicates this prediction. Motivated by a text mining model designed for building a classifier from labeled and unlabeled examples, we have developed an iterative supervised learning model for the prediction of MHC class II binding peptides.
A linear programming (LP) model was employed for the learning task at each iteration, since it is fast and can re-optimize the previous classifier when the training sets are altered. The performance of the new model has been evaluated with benchmark datasets. The outcome demonstrates that the model achieves an accuracy of prediction that is competitive compared to the advanced predictors (the Gibbs sampler and TEPITOPE). The average areas under the ROC curve obtained from one variant of our model are 0.753 and 0.715 for the original and homology reduced benchmark sets, respectively. The corresponding values are respectively 0.744 and 0.673 for the Gibbs sampler and 0.702 and 0.667 for TEPITOPE.
The iterative learning procedure appears to be effective in prediction of MHC class II binders. It offers an alternative approach to this important predictionproblem.
The objective of this study was to compare and contrast two techniques of modeling mortality in a 30 bed multi-disciplinary ICU; neural networks and logistic regression. Fifteen physiological variables were recorded on day 3 for 422 consecutive patients whose duration of stay was over 72 hours. Two separate models were built using each technique. First, logistic and neural network models were constructed on the complete 422 patient dataset and discrimination was compared. Second, the database was randomly divided into a 284 patient developmental dataset and a 138 patient validation dataset. The developmental dataset was used to construct logistic and neural net models and the predictive power of these models was verified on the validation dataset. On the complete dataset, the neural network clearly outperformed the logistic model (sensitivity and specificity of 1 and .997 vs. .525 and .966, area under ROC curve .9993 vs. .9259), while both performed equally well on the validation dataset (area under ROC of .82). The excellent performance of the neural net on the complete dataset reveals that the problem is classifiable. Since our dataset only contained 40 mortality events, it is highly likely that the validation dataset was not representative of the developmental dataset, which led to a decreased predictive performance by both the neural net and the logistic regression models. Theoretically, given an extensive dataset, the neural network should be able to perform mortality prediction with a sensitivity and a specificity approaching 95%. Clinically, this would be an extremely important achievement.(ABSTRACT TRUNCATED AT 250 WORDS)
Objective: The aim of this study was to create a classifier for automatic detection of chest radiograph reports consistent with the mediastinal findings of inhalational anthrax.
Design: The authors used the Identify Patient Sets (IPS) system to create a key word classifier for detecting reports describing mediastinal findings consistent with anthrax and compared their performances on a test set of 79,032 chest radiograph reports.
Measurements: Area under the ROC curve was the main outcome measure of the IPS classifier. Sensitivity and specificity of an initial IPS model were calculated based on an existing key word search and were compared against a Boolean version of the IPS classifier.
Results: The IPS classifier received an area under the ROC curve of 0.677 (90% CI = 0.628 to 0.772) with a specificity of 0.99 and maximum sensitivity of 0.35. The initial IPS model attained a specificity of 1.0 and a sensitivity of 0.04.
Conclusion: The IPS system is a useful tool for helping domain experts create a statistical key word classifier for textual reports that is a potentially useful component in surveillance of radiographic findings suspicious for anthrax.
The aim of the present study was to compare the ability to predict difficult visualization of the larynx from the following preoperative airway predictive indices, in isolation and combination: modified Mallampati test (MMT), the ratio of height to thyromental distance (RHTMD) and the Upper-Lip-Bite test (ULBT).
We collected data on 603 consecutive patients scheduled for elective surgery under general anesthesia requiring endotracheal intubation and then evaluated all three factors before surgery. An experienced anesthesiologist, not informed of the recorded preoperative airway evaluation, performed the laryngoscopy and grading (as per Cormack and Lehane's classification). Sensitivity, specificity, and positive and negative predictive value, Receiver operating characteristic (ROC) Curve and the area under ROC curve (AUC) for each airway predictor in isolation and in combination were determined.
Difficult laryngoscopy (Grade 3 or 4) occurred in 41 (6.8%) patients. The main endpoint of the present study, the AUC of the ROC, was significantly lower for the MMT (AUC, 0.511; 95% CI, 0.470–0.552) than the ULBT (AUC, 0.709; 95% CI, 0.671–0.745, P=0.002) and the RHTMD score (AUC, 0.711; 95% CI, 0.673–0.747, P=0.001). There was no significant difference between the AUC of the ROC for the ULBT and the RHTMD score. By using discrimination analysis, the optimal cutoff point for the RHTMD for predicting difficult laryngoscopy was 21.06 (sensitivity, 75.6%; specificity, 58.5%).
The RHTMD is comparable with ULBT for prediction of difficult laryngoscopy in general population.
Difficult laryngoscopy; endotracheal intubation; RHTMD; thyromental distance; ULBT
This study evaluates a clinical pathway currently being employed at a large single-center pediatric cardiology practice. The dataset includes 1,997 pediatric patients with the primary complaint of chest pain. A logistic regression model was developed to predict cardiac disease and identify strong indicators of cardiac pathology. The area under the ROC curve was 0.73 and the Matthews correlation coefficient was 0.23. Given the low incidence of pathology disease, this study was unable to identify strong predictors of major cardiac pathology. The analysis did support syncope, palpitations and the onset of chest pain in the past 2–7 days as predictors of minor cardiac disease. However, the model indicated exertional chest pain is negatively associated with cardiac disease. This data should be evaluated with caution as some of the results are contrary to most clinical cardiologists’ views. The majority of the results support the cardiac disease predictors in the clinical pathway.
Current guidelines recommend completion axillary lymph node dissection (cALND) in case of a sentinel lymph node (SLN) metastasis larger than 0.2 mm. However, in 50%–65% of these patients, the non-SLNs contain no further metastases and cALND provides no benefit. Several nomograms and scoring systems have been suggested to predict the risk of metastases in non-SLNs. We have evaluated the Tenon score.
Patients and Methods
In a retrospective review of the Swedish Sentinel Node Multicentre Cohort Study, risk factors for additional metastases were analysed in 869 SLN-positive patients who underwent cALND, using uni- and multivariate logistic regression models. A receiver operating characteristic (ROC) curve was drawn on the basis of the sensitivity and specificity of the Tenon score, and the area under the curve (AUC) was calculated.
Non-SLN metastases were identified in 270/869 (31.1%) patients. Tumour size and grade, SLN status and ratio between number of positive SLNs and total number of SLNs were significantly associated with non-SLN status in multivariate analyses. The area under the curve for the Tenon score was 0.65 (95% CI 0.61–0.69). In 102 patients with a primary tumour <2 cm, Elston grade 1–2 and SLN metastases ≤2 mm, the risk of non SLN metastasis was less than 10%.
The Tenon score performed inadequately in our material and we could, based on tumour and SLN characteristics, only define a very small group of patients in which negative non-sentinel nodes could be predicted.
breast cancer; sentinel node; metastases
Blood stasis syndrome (BSS) in traditional Asian medicine has been considered to correlate with the extent of atherosclerosis, which can be estimated using the cardioankle vascular index (CAVI). Here, the diagnostic utility of CAVI in predicting BSS was examined. The BSS scores and CAVI were measured in 140 stroke patients and evaluated with respect to stroke risk factors. Receiver operating characteristic (ROC) curve analysis was used to determine the diagnostic accuracy of CAVI for the diagnosis of BSS. The BSS scores correlated significantly with CAVI, age, and systolic blood pressure (SBP). Multiple logistic regression analysis showed that CAVI was a significant associate factor for BSS (OR 1.55, P = 0.032) after adjusting for the age and SBP. The ROC curve showed that CAVI and age provided moderate diagnostic accuracy for BSS (area under the ROC curve (AUC) for CAVI, 0.703, P < 0.001; AUC for age, 0.692, P = 0.001). The AUC of the “CAVI+Age,” which was calculated by combining CAVI with age, showed better accuracy (0.759, P < 0.0001) than those of CAVI or age. The present study suggests that the CAVI combined with age can clinically serve as an objective tool to diagnose BSS in stroke patients.
To understand enzyme functions, identifying the catalytic residues is a usual first step. Moreover, knowledge about catalytic residues is also useful for protein engineering and drug-design. However, to experimentally identify catalytic residues remains challenging for reasons of time and cost. Therefore, computational methods have been explored to predict catalytic residues. Here, we developed a new algorithm, L1pred, for catalytic residue prediction, by using the L1-logreg classifier to integrate eight sequence-based scoring functions. We tested L1pred and compared it against several existing sequence-based methods on carefully designed datasets Data604 and Data63. With ten-fold cross-validation, L1pred showed the area under precision-recall curve (AUPR) and the area under ROC curve (AUC) of 0.2198 and 0.9494 on the training dataset, Data604, respectively. In addition, on the independent test dataset, Data63, it showed the AUPR and AUC values of 0.2636 and 0.9375, respectively. Compared with other sequence-based methods, L1pred showed the best performance on both datasets. We also analyzed the importance of each attribute in the algorithm, and found that all the scores contributed more or less equally to the L1pred performance.
Context: The Heart Disease Program (HDP) is a novel computerized diagnosis program incorporating a computer model of cardiovascular physiology. Physicians can enter standard clinical data and receive a differential diagnosis with explanations.
Objective: To evaluate the diagnostic performance of the HDP and its usability by physicians in a typical clinical setting.
Design: A prospective observational study of the HDP in use by physicians in departments of medicine and cardiology of a teaching hospital. Data came from 114 patients with a broad range of cardiac disorders, entered by six physicians.
Measurements: Sensitivity, specificity, and positive predictive value (PPV). Comprehensiveness: the proportion of final diagnoses suggested by the HDP or physicians for each case. Relevance: the proportion of HDP or physicians' diagnoses that are correct. Area under the receiver operating characterist (ROC) curve (AUC) for the HDP and the physicians. Performance was compared with a final diagnosis based on follow-up and further investigations.
Results: Compared with the final diagnoses, the HDP had a higher sensitivity (53.0% vs. 34.8%) and significantly higher comprehensiveness (57.2% vs. 39.5%, p < 0.0001) than the physicians. Physicians' PPV and relevance (56.2%, 56.0%) were higher than the HDP (25.4%, 28.1%). Combining the diagnoses of the physicians and the HDPs, sensitivity was 61.3% and comprehensiveness was 65.7%. These findings were significant in the two collection cohorts and for subanalysis of the most serious diagnoses. The AUCs were similar for the HDP and the physicians.
Conclusions: The heart disease program has the potential to improve the differential diagnoses of physicians in a typical clinical setting.
ROC analysis occupies an increasingly important role in technology assessment. ROC curves allow one to compare a set of ordinal estimates over the entire range of estimates. Sources of such estimates may include subjective probabilities, mathematical prediction models and empiric prediction models (like the APGAR score). The area under the ROC curve measures the ability of the estimation method to discriminate between two states (usually disease and non-disease). This paper discusses how one constructs ROC curves, what the area under the curve means, and how and why one compares two ROC curves. The computer program (ROC ANALYZER) allows easy performance of these analyses on MS-DOS compatible machines.
Pre-symptomatic prediction of disease and drug response based on genetic testing is a critical component of personalized medicine. Previous work has demonstrated that the predictive capacity of genetic testing is constrained by the heritability and prevalence of the tested trait, although these constraints have only been approximated under the assumption of a normally distributed genetic risk distribution.
Here, we mathematically derive the absolute limits that these factors impose on test accuracy in the absence of any distributional assumptions on risk. We present these limits in terms of the best-case receiver-operating characteristic (ROC) curve, consisting of the best-case test sensitivities and specificities, and the AUC (area under the curve) measure of accuracy. We apply our method to genetic prediction of type 2 diabetes and breast cancer, and we additionally show the best possible accuracy that can be obtained from integrated predictors, which can incorporate non-genetic features.
Knowledge of such limits is valuable in understanding the implications of genetic testing even before additional associations are identified.
Diagnostic accuracy studies address how well a test identifies the target condition of interest.Sensitivity, specificity, predictive values and likelihood ratios (LRs) are all different ways of expressing test performance.Receiver operating characteristic (ROC) curves compare sensitivity versus specificity across a range of values for the ability to predict a dichotomous outcome. Area under the ROC curve is another measure of test performance.All of these parameters are not intrinsic to the test and are determined by the clinical context in which the test is employed.High sensitivity corresponds to high negative predictive value and is the ideal property of a “rule-out” test.High specificity corresponds to high positive predictive value and is the ideal property of a “rule-in” test.LRs leverage pre-test into post-test probabilities of a condition of interest and there is some evidence that they are more intelligible to users.
To compare the general adiposity index (BMI) with abdominal obesity indices (waist circumference (WC), waist-to-hip ratio (WHR) and waist-to-height ratio (WHtR)) in order to examine the best predictor of cardiometabolic risk factors among Hispanics living in Puerto Rico.
Secondary analysis of measurements taken from a representative sample of adults. Logistic regression models (prevalence odds ratios (POR)), partial Pearson’s correlations (controlling for age and sex) and receiver-operating characteristic (ROC) curves were calculated between indices of obesity (BMI, WC, WHR and WHtR) and blood pressure, HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), total cholesterol (TC):HDL-C, TAG, fasting blood glucose, glycosylated Hb, high-sensitivity C-reactive protein (hs-CRP), fibrinogen, plasminogen activator inhibitor-1 (PAI-1) and an aggregated measure of cardiometabolic risk.
Household study conducted between 2005 and 2007 in the San Juan Metropolitan Area in Puerto Rico.
A representative sample of 858 non-institutionalized adults.
All four obesity indices significantly correlated with the cardiometabolic risk factors. WHtR had the highest POR for high TC:HDL-C, blood pressure, hs-CRP, fibrinogen and PAI-1; WC had the highest POR for low HDL-C and high LDL-C and fasting blood glucose; WHR had the highest POR for overall cardiometabolic risk, TAG and glycosylated Hb. BMI had the lowest POR for most risk factors and smallest ROC curve for overall cardiometabolic risk.
The findings of the study suggest that general adiposity and abdominal adiposity are both associated with cardiometabolic risk in this population, although WC, WHR and WHtR appear to be slightly better predictors than BMI.
Waist circumference; Waist-to-hip ratio; BMI; Waist-to-height ratio; Cardiometabolic risk
The estimate of a multivariate risk is now required in guidelines for cardiovascular prevention. Limitations of existing statistical risk models lead to explore machine-learning methods. This study evaluates the implementation and performance of a decision tree (CART) and a multilayer perceptron (MLP) to predict cardiovascular risk from real data. The study population was randomly splitted in a learning set (n = 10,296) and a test set (n = 5,148). CART and the MLP were implemented at their best performance on the learning set and applied on the test set and compared to a logistic model. Implementation, explicative and discriminative performance criteria are considered, based on ROC analysis. Areas under ROC curves and their 95% confidence interval are 0.78 (0.75-0.81), 0.78 (0.75-0.80) and 0.76 (0.73-0.79) respectively for logistic regression, MLP and CART. Given their implementation and explicative characteristics, these methods can complement existing statistical models and contribute to the interpretation of risk.
An elevated lactate level reflects impaired tissue oxygenation and is a predictor of mortality. Studies have shown that the anion gap is inadequate as a screen for hyperlactataemia, particularly in critically ill and trauma patients. A proposed explanation for the anion gap's poor sensitivity and specificity in detecting hyperlactataemia is that the serum albumin is frequently low. This study therefore, sought to compare the predictive values of the anion gap and the anion gap corrected for albumin (cAG) as an indicator of hyperlactataemia as defined by a lactate ⩾2.5 mmol/l.
A retrospective review of 639 sets of laboratory values from a tertiary care hospital. Patients' laboratory results were included in the study if serum chemistries and lactate were drawn consecutively. The sensitivity, specificity, and predictive values were obtained. A receiver operator characteristics curve (ROC) was drawn and the area under the curve (AUC) was calculated.
An anion gap ⩾12 provided a sensitivity, specificity, positive predictive value, and negative predictive value of 39%, 89%, 79%, and 58%, respectively, and a cAG ⩾12 provided a sensitivity, specificity, positive predictive value, and negative predictive value of 75%, 59%, 66%, and 69%, respectively. The ROC curves between anion gap and cAG as a predictor of hyperlactataemia were almost identical. The AUC was 0.757 and 0.750, respectively.
The sensitivities, specificities, and predictive values of the anion gap and cAG were inadequate in predicting the presence of hyperlactataemia. The cAG provides no additional advantage over the anion gap in the detection of hyperlactataemia.
correcting anion gap; hyperlactataemia
The present study aimed to develop an artificial neural network (ANN) based prediction model for cardiovascular autonomic (CA) dysfunction in the general population.
We analyzed a previous dataset based on a population sample consisted of 2,092 individuals aged 30–80 years. The prediction models were derived from an exploratory set using ANN analysis. Performances of these prediction models were evaluated in the validation set.
Univariate analysis indicated that 14 risk factors showed statistically significant association with CA dysfunction (P < 0.05). The mean area under the receiver-operating curve was 0.762 (95% CI 0.732–0.793) for prediction model developed using ANN analysis. The mean sensitivity, specificity, positive and negative predictive values were similar in the prediction models was 0.751, 0.665, 0.330 and 0.924, respectively. All HL statistics were less than 15.0.
ANN is an effective tool for developing prediction models with high value for predicting CA dysfunction among the general population.
Cardiovascular autonomic dysfunction; Artificial neural network; Prediction model; Chinese population
BACKGROUND AND OBJECTIVES:
Body mass index (BMI) is the most widely used measure to define obesity and predict its complications, such as diabetes and hypertension, but its accuracy and usefulness in Saudi subjects is unknown. This study aimed to assess the validity of standard BMI cut-point values in the Saudi population.
SUBJECTS AND METHODS:
197 681 adults participated in a cross-sectional study to detect diabetes and hypertension in the Saudi Eastern province in 2004/2005, with blood pressure, fasting blood sugar, height and weight measurements taken. Sensitivities, specificities, areas under the curves, predictive values, likelihood ratios, false positive, false negatives and total misclassification ratios were calculated for various BMI values determined from receiver operating characteristic (ROC) curves. The significance of the association between risk factors and BMI was assessed using regression analysis.
For the definition of overweight, ROC curve analysis suggested optimal BMI cut-offs of 28.50 to 29.50 in men and 30.50 to 31.50 in women, but the levels of sensitivity and specificity were too low to be of clinical value and the overall misclassification was unacceptably high across all the selected BMI values (>0.80). The relationship between BMI and the presence of diabetes and/or hypertension was not improved when a BMI of 25 was used. Using regression analyses, the odds ratios for hypertension and/or diabetes increased significantly from BMI values as low as 21-23 with no improvement in the diagnostic performance of BMI at these cutoffs.
In Saudi population, there is an increased risk of diabetes and hypertension relative to BMI, starting at a BMI as low as 21 but overall there is no cutoff BMI level with high predictive value for the development of these chronic diseases, including the WHO definition of obesity at BMI of 30.
Population-based screening for cardiovascular disease (CVD) risk, incorporating blood tests, is proposed in several countries.
The aim of this study was to evaluate whether a simple approach to identifying individuals at high risk of CVD using routine data might be effective.
Design of study
Prospective cohort study (EPIC-Norfolk).
Norfolk area, UK.
A total of 21 867 men and women aged 40–74 years, who were free from CVD and diabetes at baseline, participated in the study. The discrimination (the area under the receiver operating characteristic curve [aROC]), calibration, sensitivity/specificity, and positive/negative predictive value were evaluated for different risk thresholds of the Framingham risk equations and the Cambridge diabetes risk score (as an example of a simple risk score using routine data from electronic general practice records).
During 203 664 person-years of follow-up, 2213 participants developed a first CVD event (10.9 per 1000 person-years). The Cambridge diabetes risk score predicted CVD events reasonably well (aROC 0.72; 95% confidence interval [CI] = 0.71 to 0.73), while the Framingham risk score had the best predictive ability (aROC 0.77; 95% CI = 0.76 to 0.78). The Framingham risk score overestimated risk of developing CVD in this representative British population by 60%.
A risk score incorporating routinely available data from GP records performed reasonably well at predicting CVD events. This suggests that it might be more efficient to use routine data as the first stage in a stepwise population screening programme to identify people at high risk of developing CVD before more time- and resource-consuming tests are used.
cardiovascular disease; diabetes; prediction; primary care; risk assessment
No study has evaluated the performance of BRCA1/2 mutations prediction models in male breast cancer (MBC) series. Although rare, MBC deserves attention because male and female breast cancers share many characteristics, including the involvement of genetic predisposition factors such as BRCA1/BRCA2 mutations. Indeed, the occurrence of MBC is a commonly used criterion to select families for BRCA mutation testing. We evaluated the performance and clinical effectiveness of four different predictive models in a population-based series of 102 Italian MBC patients characterized for BRCA1/2 mutations. Sensitivity, specificity, and positive and negative predictive values (PPV, NPV) were calculated for each risk model at the 10% threshold. The area under the ROC (AUC) curves and its corresponding asymptotic 95% CIs were calculated as a measure of the accuracy. In our study, the BRCAPRO version 5.0 had the highest combination of sensitivity, specificity, NPV and PPV for the combined probability and for the discrimination of BRCA2 mutations. In individuals with negative breast–ovarian cancer family history, BRCAPRO 5.0 reached a high discriminatory capacity (AUC=0.92) in predicting BRCA2 mutations and showed values of sensitivity, specificity, NPV and PPV of 0.5, 0.98, 0.97 and 0.67, respectively, for the combined probability. BRCAPRO version 5.0 can be particularly useful in dealing with non-familial MBC, a circumstance that often represents a challenging situation in genetic counseling.
male breast cancer; BRCAPRO 5.0 prediction model; risk assessment