We investigated the prognostic value of cardiac myosin-binding protein C (cMyC), a novel cardiospecific marker, both independently and in combination with N-terminal pro-B-type natriuretic peptide (NT-proBNP), for predicting 6-month all-cause mortality in patients without acute coronary syndrome (ACS) treated at medical (nonsurgical) cardiac intensive care units (CICUs). Admission levels of cMyC, high-sensitivity cardiac troponin T (hs-cTnT), and NT-proBNP were measured in 1032 consecutive patients (mean age; 70 years) without ACS hospitalized acutely in medical CICUs for the treatment of cardiovascular disease. Serum cMyC was closely correlated with hs-cTnT and moderately with NT-proBNP (r = 0.92 and r = 0.49, respectively, p < 0.0001). During the 6-month follow-up period after admission, there were 109 (10.6%) all-cause deaths, including 72 cardiovascular deaths. Both cMyC and NT-proBNP were independent predictors of 6-month all-cause mortality (all p < 0.05). Combining cMyC and NT-proBNP with a baseline model of established risk factors improved patient classification and discrimination beyond any single biomarker (all p < 0.05) or the baseline model alone (both p < 0.0001). Moreover, patients were divided into nine groups using cMyC and NT-proBNP tertiles, and the adjusted hazard ratio (95% confidence interval) for 6-month all-cause mortality in patients with both biomarkers in the highest vs. lowest tertile was 9.67 (2.65-35.2). When cMyC was replaced with hs-cTnT, similar results were observed for hs-cTnT. In addition, the C-indices for addition of cMyC or hs-cTnT to the baseline model were similar (0.798 vs. 0.800, p = 0.94). In conclusion, similar to hs-cTnT, cMyC at admission may be a potent, independent predictor of 6-month all-cause mortality in patients without ACS treated at medical CICUs, and their prognostic abilities may be comparable. Combining cMyC or hs-cTnT with NT-proBNP may substantially improve early risk stratification of this population.

Severe primary graft dysfunction (PGD) is the leading cause of early mortality following heart transplant (HT). This study analyzed the temporal trends and mortality associated with severe PGD, identified risk factors, and developed a predictive model based on a contemporary cohort.

A total of 2029 HT performed between 2010 and 2020 in 14 Spanish centers were retrospectively analyzed. Patients with and without severe PGD were compared. Logistic regression was used to identify predictors of severe PGD and to generate a risk score. Model performance was assessed in terms of calibration and discrimination.

The incidence of severe PGD was 10%, with an increase observed over the last 5 years (8% vs 11%). However, 30-day and 1-ear mortality declined significantly (59.1% vs 38.8% and 69.7% vs 58.8%, respectively). Independent predictors of severe PGD included extracorporeal membrane oxygenation (OR, 2.79), pretransplant ventricular assist devices (OR, 2.11), donor-to-recipient weight ratio <0.8 (OR, 2.11), and congenital heart disease (OR, 2.11). A risk score was created, showing good calibration but limited discriminative ability.

Despite a rising incidence of severe PGD, mortality showed a marked decrease. Predictors of severe PGD included congenital heart disease, a donor-to-recipient weight ratio <0.8, and the use of extracorporeal membrane oxygenation or pretransplant ventricular assist devices. The predictive model showed good calibration but only moderate discriminative performance.

Patients with stroke have high rates of all-cause readmission and case fatality. Limited information is available on how to predict these outcomes.

We aimed to assess whether adding the initial National Institutes of Health Stroke Scale (NIHSS) score or modified Rankin scale (mRS) score at discharge improved predictive models of 30-day nonelective readmission or 30-day mortality poststroke.

Using a cohort of patients with ischemic stroke in a large multiethnic integrated health care system from June 15, 2018, to April 29, 2020, we tested 2 predictive models for a composite outcome (30-day nonelective readmission or death). The models were based on administrative data (Length of Stay, Acuity, Charlson Comorbidities, Emergency Department Use score; LACE) as well as a comprehensive model (Transition Support Level; TSL). The models, initial NIHSS score, and mRS scores at discharge, were tested independently and in combination with age and sex. We assessed model performance using the area under the receiver operator characteristic (c-statistic), Nagelkerke pseudo-R2, and Brier score.

The study cohort included 4843 patients with 5014 stroke hospitalizations. Average age was 71.9 (SD 14) years, 50.6% (2537/5014) were female, and 52.1% (2614/5014) were White. Median initial NIHSS score was 4 (IQR 2-8). There were 538 (10.7%) nonelective readmissions and 150 (3.9%) deaths within 30 days. The logistic models revealed that the best performing models were TSL (c-statistic=0.69) and TSL plus mRS score at discharge (c-statistic=0.69).

We found that neither the initial NIHSS score nor the mRS score at discharge significantly enhanced the predictive ability of the LACE or TSL models. Future efforts at prediction of short-term stroke outcomes will need to incorporate new data elements.

Identification of early predictors of postoperative pancreatic fistula (POPF) related postpancreatectomy hemorrhage (PPH) on contrast-enhanced computed tomography (CT) may help tailoring management after pancreaticoduodenectomy (PD) although no model has been validated so far.

A bicentric analysis of consecutive PD performed between 2017 and 2022 was performed. A recently reported CT-based score (CTS) was externally validated. Sensitivity refinements were proposed through a modified-CTS which was internally (development cohort, n=182) and externally validated (validation cohort, n=62). Bootstrap corrected Areas under the curve (AUCs), Sensitivity (Se) and Positive Predictive Value (PVV) were used to evaluate and compare the two scores.

A total of 244 patients (55.1% women; median age: 68 years [IQR: 58.0-75.0], clinically relevant (cr)-POPF: 25.4%, cr-PPH: 13.9%) were included. CTS accurately predicted a cr-PPH with an AUC of 0.83 (1000-boostrap 95% CI: 0.76-0.89). The modified-CTS, made available online (https://stylianostzedakis.shinyapps.io/pph_risk_calculator/), included CTS with 2 supplementary variables selected from a multivariable backward-stepwise regression: Perianastomotic air bubbles, posterosuperior pancreaticojejunal (PJ) anastomosis collection, posterior PJ defect, PJ collection in contact with hepatic or gastroduodenal artery stump and arterial wall irregularities. When compared with the CTS, although modified-CTS AUC [95%CI] were similar in the validation cohort (0.81 [0.62-0.95] vs. 0.87 [0.56-0.96], DeLong p=0.7), Se and PPV for early PPH detection were significantly higher (0.82 [0.75-0.92] vs. 0.71 [0.35-0.75] and 0.95 [0.83-0.99] vs. 0.33 [0.12-0.62], McNemar's p = 0.03).

With a robust prediction model, early CT-scan after PD seems a valid tool for early identification of high-risk cr-PPH patients.

The role of inflammation in shaping death risk in diabetes is still unclear.

To study whether inflammation is associated with and helps predict mortality risk in patients with type 2 diabetes. To explore the intertwined link between inflammation and tryptophan metabolism on death risk.

There were 2 prospective cohorts: the aggregate Gargano Mortality Study (1731 individuals; 872 all-cause deaths) as the discovery sample, and the Foggia Mortality Study (490 individuals; 256 deaths) as validation sample. Twenty-seven inflammatory markers were measured. Causal mediation analysis and in vitro studies were carried out to explore the link between inflammatory markers and the kynurenine to tryptophan ratio (KTR) in shaping mortality risk.

Using multivariable stepwise Cox regression analysis, interleukin (IL)-4, IL-6, IL-8, IL-13, RANTES, and interferon gamma-induced protein-10 (IP-10) were independently associated with death. An inflammation score (I score) comprising these 6 molecules is strongly associated with death in both the discovery and the validation cohorts HR (95% CI) 2.13 (1.91-2.37) and 2.20 (1.79-2.72), respectively. The I score improved discrimination and reclassification measures (all P < .01) of 2 mortality prediction models based on clinical variables. The causal mediation analysis showed that 28% of the KTR effect on mortality was mediated by IP-10. Studies in cultured endothelial cells showed that 5-methoxy-tryptophan, an anti-inflammatory metabolite derived from tryptophan, reduces the expression of IP-10, thus providing a functional basis for the observed causal mediation.

Adding the I score to clinical prediction models may help identify individuals who are at greater risk of death. Deeply addressing the intertwined relationship between low-grade inflammation and imbalanced tryptophan metabolism in shaping mortality risk may help discover new therapies targeting patients characterized by these abnormalities.

The application of artificial intelligence (AI) in laboratory medicine will revolutionize predictive modeling using clinical laboratory information. Machine learning (ML), a sub-discipline of AI, involves fitting algorithms to datasets and is broadly used for data-driven predictive modeling in various disciplines. The majority of ML studies reported in systematic reviews lack key aspects of quality assurance. In clinical laboratory medicine, it is important to consider how differences in analytical methodologies, assay calibration, harmonization, pre-analytical errors, interferences, and physiological factors affecting measured analyte concentrations may also affect the downstream robustness and reliability of ML models. In this article, we address the need for quality improvement and proper validation of ML classification models, with the goal of bringing attention to key concepts pertinent to researchers, manuscript reviewers, and journal editors within the field of pathology and laboratory medicine. Several existing predictive modeling guidelines and recommendations can be readily adapted to the development of ML models in laboratory medicine. We summarize a basic overview of ML and key points from current guidelines including advantages and pitfalls of applied ML. In addition, we draw a parallel between validation of clinical assays and ML models in the context of current regulatory frameworks. The importance of classification performance metrics, model explainability, and data quality along with recommendations for strengthening journal submission requirements are also discussed. Although the focus of this article is on the application of ML in laboratory medicine, many of these concepts extend into other areas of medicine and biomedical science as well.

Background/Objectives: Proper confidence interval estimation of the area under the receiver operating characteristic curve (AUC), the net reclassification index (NRI), and the integrated discrimination improvement (IDI) is an area of ongoing research. The most common confidence interval estimation methods employ asymptotic theory. However, developments demonstrate that degeneration of the normal distribution assumption under the null hypothesis exists for measures such as the change in AUC (ΔAUC) and IDI, and confidence intervals estimated under the normal distribution assumption may be invalid. We aim to study the performance of confidence intervals derived assuming asymptotic theory and those derived with non-parametric bootstrapping methods. Methods: We examine the performance of ΔAUC, NRI, and IDI in both the logistic and survival regression context. We explore empirical distributions and compare coverage probabilities of asymptotic confidence intervals with those produced from bootstrapping methods through simulation. Results: The primary finding in both the logistic framework and the survival analysis framework is that the percentile CIs performed well regarding coverage, without compromise to their width; this finding was robust in most scenarios. Conclusions: Our results suggest that the asymptotic intervals are only appropriate when a strong effect size of the added parameter exists, and that the percentile bootstrap interval exhibits at least a reasonable coverage while maintaining the shortest width in nearly all simulated scenarios, making this interval the most reliable choice. The intent is that these recommendations improve the accuracy in the estimation and the overall assessment of discrimination improvement.

Venous thromboembolism (VTE) is a common complication in patients with traumatic brain injury (TBI). This study aimed to assess the predictive ability of the Caprini score, Risk Assessment Profile for Thromboembolism (RAPT), and Trauma Embolic Scoring System(TESS) for VTE risk assessments in TBI patients. A retrospective analysis of 460 TBI patients was conducted, categorizing them into VTE and non-VTE groups based on imaging results. The three scales were applied to assess VTE risk, and their performance was compared using receiver operating characteristic(ROC) curves and area under the curve(AUC) values. The VTE incidence was 31.7%. The RAPT scale demonstrated the highest AUC (0.826) and optimal cutoff (9.5) with balanced sensitivity (0.753) and specificity (0.771). The Caprini and TESS scales also showed moderate to high predictive value but had lower AUCs. All three scoring scales showed medium to high predictive value for the risk of VTE in patients with TBI. Among them, the RAPT scoring scale offered the highest predictive value for VTE risk in TBI patients, with fewer items, making it easier for clinical implementation. It stands as the most appropriate VTE risk assessment scale for TBI patients at present.

Acute myocardial infarction is a condition in which a part of the heart muscle cannot receive enough blood due to the narrowing and blockage of the vessels feeding the heart over time. Noticing this situation lately and failing to intervene immediately may cause death and some permanent damage to individuals. The ST-segment elevation MI (STEMI) is one of the most serious and fatal types of acute myocardial infarction which requires urgent diagnosis and intervention. Artificial intelligence-based applications used in health have become widespread, paving the way for early diagnosis and treatment. In modern medicine, it is vital that STEMI patients are identified and treated accurately and quickly. Determining the risk of death of patients in advance plays a major role in making clinical decisions. Traditional risk assessment methods are often time-consuming and subjective processes and rely on manual analysis of clinical data. In this respect, this study is expected to provide clinical decision support in the management of STEMI patients and contribute to improving the quality of healthcare services. In the proposed work, death risk analysis and in-hospital mortality risk prediction are carried out using some selected machine learning (ML) algorithms, such as SVM, RF, RT, k-NN, LMT, and MLP, that are proven to be effective in medical classification tasks. The conducted test results indicate that the proposed method outperforms similar studies in the literature, achieving a superior performance of over 99 % in all metrics, i.e., accuracy, recall, precision, sensitivity, F-score, and AUC. Moreover, the same competitive results have been obtained with even much fewer predictors.

In this study, we used a dataset including 42 individual bottlenose dolphins (Tursiops spp.) to determine the reliability of lung function testing as a method for assessing respiratory health. Each dolphin was trained to beach voluntarily, allowing researchers to measure respiratory flow in a controlled, beached state. From the collected respiratory flow data, alongside timing parameters, we extracted 18 specific variables, supplemented by additional factors such as body mass, age, and sex. These variables were hypothesized to serve as potential variables for identifying respiratory compromise. A model was developed that reduced the number of predictive variables, showing that 4 specific variables were particularly effective, yielding an accuracy of 88.4% in determining whether a dolphin was free from respiratory disease. This high level of accuracy underscores the potential of lung function testing as a diagnostic tool in the context of stranded dolphins, where rapid, non-invasive methods are crucial for assessing health status. These results suggest that lung function testing provides a non-invasive and efficient method for evaluating respiratory health in stranded dolphins and supports the use of lung function assessments in wildlife management and conservation. By enabling early detection of respiratory issues, this approach can enhance the success of rehabilitation efforts, potentially improving the survival rates of dolphins that have stranded, which is often a critical concern in marine conservation initiatives.

Objective: Internet-delivered cognitive behavior therapies (iCBT) are effective and scalable treatments for depression and anxiety. However, treatment adherence remains a major limitation that could be further understood by applying machine learning methods to during-treatment messages. We used machine learned topics to predict drop-out risk and symptom change in iCBT. Method: We applied topic modeling to naturalistic messages from 18,117 patients of nationwide iCBT programs for depression and generalized anxiety disorder (GAD). We used elastic net regression for outcome predictions and cross-validation to aid in model selection. We left 10% of the data as a held-out test set to assess predictive performance. Results: Compared to a set of reference covariates, inclusion of the topic variables resulted in significant decrease in drop-out risk prediction loss, both in between-patient and within-patient session-by-session models. Quantified as partial pseudo-R2, the increase in variance explained was 2.1-6.8 percentage units. Topics did not improve symptom change predictions compared to the reference model. Conclusions: Message contents can be associated with both between-patients and session-by-session risk of drop-out. Our topic predictors were theoretically interpretable. Analysis of iCBT messages can have practical implications in improved drop-out risk assessment to aid in the allocation of additional supportive interventions.

The polygenic risk score (PRS) has been perceived as advantageous in predicting the risk of complex diseases compared to other measures. We aimed to systematically evaluate the influence of PRS on disease outcome and to explore its predictive value.

We comprehensively assessed the relationship between PRS and 32 complex diseases in the UK Biobank. We used Cox models to estimate the effects of PRS on the incidence risk. Then, we constructed prediction models to assess the clinical utility of PRS in risk prediction. For 16 diseases, we further compared the disease risk and prediction capability of PRS across early and late-onset cases.

Higher PRS led to greater incident risk, with hazard ratio (HR) ranging from 1.07 (95% confidence interval (CI) = 1.06-1.08) for panic/anxiety disorder to 4.17 (95% CI = 4.03-4.31) for acute pancreatitis. This effect was more pronounced in early-onset cases for 12 diseases, increasing by 52.8% on average. Particularly, the early-onset risk of heart failure associated with PRS (HR = 3.02; 95% CI = 2.53-3.59) was roughly twice compared to the late-onset risk (HR = 1.48; 95% CI = 1.46-1.51). Compared to average PRS (20-80%), individuals positioned within the top 2.5% of the PRS distribution exhibited varying degrees of elevated risk, corresponding to a more than five times greater risk on average. PRS showed additional value in clinical risk prediction, causing an average improvement of 6.1% in prediction accuracy. Further, PRS demonstrated higher predictive accuracy for early-onset cases of 11 diseases, with heart failure displaying the most significant (37.5%) improvement when incorporating PRS into the prediction model (concordance index (C-index) = 0.546; standard error (SE) = 0.011 vs. C-index = 0.751; SE = 0.010, P = 2.47 × 10-12).

As a valuable complement to traditional clinical risk tools, PRS is closely related to disease risk and can further enhance prediction accuracy, especially for early-onset cases, underscoring its potential role in targeted prevention for high-risk groups.

In the cancer early detection field, logistic regression (LR) is a frequently used approach to establish a combination rule that differentiates cancer from noncancer. However, the application of LR relies on a maximum likelihood approach, which may not yield optimal combination rules for maximizing sensitivity at a clinically desirable specificity and vice versa. In this article, we have developed an improved regression framework, sensitivity maximization at a given specificity (SMAGS), for binary classification that finds the linear decision rule, yielding the maximum sensitivity for a given specificity or the maximum specificity for a given sensitivity. We additionally expand the framework for feature selection that satisfies sensitivity and specificity maximizations. We compare our SMAGS method with normal LR using two synthetic datasets and reported data for colorectal cancer from the 2018 CancerSEEK study. In the colorectal cancer CancerSEEK dataset, we report 14% improvement in sensitivity at 98.5% specificity (0.31 vs. 0.57; P value <0.05). The SMAGS method provides an alternative to LR for modeling combination rules for biomarkers and early detection applications. Prevention Relevance: This study introduces a new machine learning methodology that identifies the optimal features and combination rules to maximize sensitivity at a fixed specificity, making it applicable to many existing biomarker prevention studies.

Accurate prediction of perioperative major adverse cardiovascular events (MACEs) is crucial, as it not only aids clinicians in comprehensively assessing patients' surgical risks and tailoring personalized surgical and perioperative management plans, but also for information-based shared decision-making with patients and efficient allocation of medical resources. This study developed and validated a machine learning (ML) model using accessible preoperative clinical data to predict perioperative MACEs in stable coronary artery disease (SCAD) patients undergoing noncardiac surgery (NCS).

We collected data from 9171 adult SCAD patients who underwent NCS and extracted 64 preoperative variables. First, the optimal data imputation, resampling, and feature selection methods were compared and selected to deal with missing data values and imbalances. Then, nine independent machine learning models (logistic regression (LR), support vector machine, Gaussian Naive Bayes (GNB), random forest, gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), light gradient boosting machine, categorical boosting (CatBoost), and deep neural network) and a stacking ensemble model were constructed and compared with the validated Revised Cardiac Risk Index's (RCRI) model for predictive performance, which was evaluated using the area under the receiver operating characteristic curve (AUROC), the area under the precision-recall curve (AUPRC), calibration curve, and decision curve analysis (DCA). To reduce overfitting and enhance robustness, we performed hyperparameter tuning and 5-fold cross-validation. Finally, the Shapley additive interpretation (SHAP) method and a partial dependence plot (PDP) were used to determine the optimal ML model.

Of the 9,171 patients, 514 (5.6 %) developed MACEs. 24 significant preoperative features were selected for model development and evaluation. All ML models performed well, with AUROC above 0.88 and AUPRC above 0.39, outperforming the AUROC (0.716) and AUPRC (0.185) of RCRI (P < 0.001). The best independent model was XGBoost (AUROC = 0.898, AUPRC = 0.479). The calibration curve accurately predicted the risk of MACEs (Brier score = 0.040), and the DCA results showed that XGBoost had a high net benefit for predicting MACEs. The top-ranked stacking ensemble model, consisting of CatBoost, GBDT, GNB, and LR, proved to be the best (AUROC 0.894, AUPRC 0.485). We identified the top 20 most important features using the mean absolute SHAP values and depicted their effects on model predictions using PDP.

This study combined missing-value imputation, feature screening, unbalanced data processing, and advanced machine learning methods to successfully develop and verify the first ML-based perioperative MACEs prediction model for patients with SCAD, which is more accurate than RCRI and enables effective identification of high-risk patients and implementation of targeted interventions to reduce the incidence of MACEs.

Autoantibodies against apolipoprotein A-1 (AAA1) are elicited by SARS-CoV-2 infection and predict COVID-19 symptoms persistence at one year in adults, but whether this applies to children is unknown. We studied the association of SARS-CoV-2 exposure with AAA1 prevalence in children and the association of AAA1 seropositivity with symptom persistence.

Anti-SARS-CoV-2 and AAA1 serologies were examined in 1031 participants aged 6 months to 17 years old from the prospective SEROCOV-KIDS cohort and recruited between 12.2021 and 02.2022. Four SARS-CoV-2 serology-based groups were defined: "Infected-unvaccinated (I+/V-)", "Uninfected-vaccinated (I-/V+)", "Infected-Vaccinated (I+/V+)", and "Naïve (I-/V-)". Reported outcomes were collected using online questionnaires. Associations with study endpoints were assessed using logistic regression.

Overall, seropositivity rates for anti-RBD, anti-N, and AAA1 were 71% (736/1031), 55% (568/1031), and 5.8% (60/1031), respectively. AAA1 showed an inverse association with age but not with any other characteristics. The I+/V- group displayed higher median AAA1 levels and seropositivity (7.9%) compared to the other groups (p ≤ 0.011), translating into a 2-fold increased AAA1 seroconversion risk (Odds ratio [OR]: 2.11, [95% Confidence Interval (CI)]: 1.22-3.65; p=0.008), unchanged after adjustment for age and sex. AAA1 seropositivity was independently associated with a 2-fold odds of symptoms persistence at ≥ 4 weeks (p ≤ 0.03) in the entire dataset and infected individuals, but not ≥ 12 weeks.

Despite the limitations of the study (cross-sectional design, patient-related outcomes using validated questionnaires), the results indicate that SARS-CoV-2 infection could elicit an AAA1 response in children, which could be independently associated with short-time symptoms persistence.

Recent studies suggest that the metabolic score for insulin resistance (MetS-IR) is an effective indicator of metabolic disorders. However, evidence on the relationship between MetS-IR and metabolic syndrome (MetS) among the Chinese middle-aged and older adult population is limited.

This cohort study aims to assess the associations of MetS-IR levels with MetS risk and its components.

Data used in this study from the National Basic Public Health Service Project Management System (2020-2023). Multivariable Cox proportional hazards model and restricted cubic spline (RCS) were employed to evaluate the associations of baseline MetS-IR levels with MetS risk and its components, receiver operating characteristic (ROC) curves were further utilized to assess the efficacy of MetS-IR in predicting the risk of MetS and its component.

Of 1,498 subjects without MetS at baseline, 392 incident MetS cases were observed during a median of 27.70 months of follow-up. The adjusted multivariable Cox regression analysis indicated an elevated 15% risk of developing MetS for 1-SD increment of MetS-IR [hazard ratios (HRs) and 95% confidence intervals: 1.16 (1.13-1.18)]. Compared to the first tertile of MetS-IR, the HRs of the third tertile and second tertile were 6.31 (95% CI 4.55-8.76) and 2.72 (95% CI 1.92-3.85), respectively. Consistent findings were further detected across subgroups. Moreover, nonlinear associations were observed between MetS-IR and the risk of MetS, abdominal obesity, and reduced high-density lipoprotein concentration (HDL-C) (P nonlinear < 0.01), with the cutoff of MetS-IR was 32.89. The area under the curve for MetS-IR in predicting MetS was 0.740 (95% CI 0.713-0.768), which was better than those of other indicators.

Our cohort study indicates a positive nonlinear association between MetS-IR with incident MetS, abdominal obesity, and reduced HDL-C, but positive linear associations of MetS-IR and elevated blood pressure (BP), elevated fasting blood glucose (FBG), elevated triglycerides (TG) in middle-aged and older adult people, more studies are warranted to verify our findings.

Accurate bleeding risk stratification after percutaneous coronary intervention is important for treatment individualization. However, there is still an unmet need for a more precise and standardized identification of patients at high bleeding risk. We derived and validated a novel bleeding risk score by augmenting the Predicting Bleeding Complications in Patients Undergoing Stent Implantation and Subsequent Dual Antiplatelet Therapy (PRECISE-DAPT) score with the Academic Research Consortium for High Bleeding Risk (ARC-HBR) criteria.

The derivation cohort comprised 29 188 patients undergoing percutaneous coronary intervention, of whom 1136 (3.9%) had Bleeding Academic Research Consortium (BARC) 3 or 5 bleeding at 1 year, from 4 contemporary real-world registries and the XIENCE V USA trial. The PRECISE-DAPT score was refitted with a Fine-Gray model in the derivation cohort and extended with the ARC-HBR criteria. The primary outcome was BARC 3 or 5 bleeding within 1 year. Independent predictors of BARC 3 or 5 bleeding were selected at multivariable analysis (P<0.01). The discrimination of the score was internally assessed with apparent validation and cross-validation. The score was externally validated in 4578 patients from the MASTER DAPT trial (Management of High Bleeding Risk Patients Post Bioresorbable Polymer Coated Stent Implantation With an Abbreviated Versus Prolonged DAPT Regimen) and 5970 patients from the STOPDAPT-2 (Short and Optimal Duration of Dual Antiplatelet Therapy-2) total cohort.

The PRECISE-HBR score (age, estimated glomerular filtration rate, hemoglobin, white blood cell count, previous bleeding, oral anticoagulation, and ARC-HBR criteria) showed an area under the curve (AUC) for 1-year BARC 3 or 5 bleeding of 0.73 (95% CI, 0.71-0.74) at apparent validation, 0.72 (95% CI, 0.70-0.73) at cross-validation, 0.74 (95% CI, 0.68-0.80) in MASTER DAPT, and 0.73 (95% CI, 0.66-0.79) in STOPDAPT-2, with superior discrimination compared with PRECISE-DAPT (cross-validation: ΔAUC, 0.01; P=0.02; MASTER DAPT: ΔAUC, 0.05; P=0.004; STOPDAPT-2: ΔAUC, 0.02; P=0.20) and other risk scores. In the derivation cohort, a cutoff of 23 points identified 11 414 patients (39.1%) with a 1-year BARC 3 or 5 bleeding risk ≥4%. An alternative version of the score, including acute myocardial infarction on admission instead of white blood cell count, showed similar predictive ability.

The PRECISE-HBR score is a contemporary, simple 7-item risk score to predict bleeding after percutaneous coronary intervention, offering a moderate improvement in discrimination over multiple existing scores. Further evaluation is required to assess its impact on clinical practice.

This review compares non-laboratory-based and laboratory-based cardiovascular disease (CVD) risk prediction equations in populations targeted for primary prevention.

Systematic review.

We searched five databases until 12 March 2024 and used prediction study risk of bias assessment tool to assess bias. Data on hazard ratios (HRs), discrimination (paired c-statistics) and calibration were extracted. Differences in c-statistics and HRs were analysed.

PROSPERO (CRD42021291936).

Nine studies (1 238 562 participants, 46 cohorts) identified six unique CVD risk equations. Laboratory predictors (eg, cholesterol and diabetes) had strong HRs, while body mass index in non-laboratory models showed limited effect. Median c-statistics were 0.74 for both models (IQR: lab 0.77-0.72; non-lab 0.76-0.70), with a median absolute difference of 0.01. Calibration measures between laboratory-based and non-laboratory-based equations were similar, although non-calibrated equations often overestimated risk.

The discrimination and calibration measures between laboratory-based and non-laboratory-based models show minimal differences, demonstrating the insensitivity of c-statistics and calibration metrics to the inclusion of additional predictors. However, in most reviewed studies, the HRs for these additional predictors were substantial, significantly altering predicted risk, particularly for individuals with higher or lower levels of these predictors compared with the average.

Calibration and discrimination indicators alone are insufficient for evaluating the clinical usefulness of prediction models, as they do not account for the cost of misclassification errors. This study aimed to modify the Geriatric Trauma Outcome Score (GTOS) and assess the clinical utility of the modified model using net benefit (NB) and decision curve analysis (DCA) for predicting in-hospital mortality.

The Trauma Quality Improvement Program (TQIP) 2017 was used to identify geriatric trauma patients (≥ 65 years) treated at Level I trauma centers. The outcome of interest was in-hospital mortality. The GTOS was modified to include additional patient, injury, and treatment characteristics identified through machine learning methods, focusing on early risk stratification. Calibration and discrimination indicators, along with NB and DCA, were utilized for evaluation.

Of the 67,222 admitted geriatric trauma patients, 5.6% died in the hospital. The modified GTOS score included the following variables with associated weights: initial airway intervention (5), Glasgow Coma Scale ≤13 (5), packed red blood cell transfusion within 24 h (3), penetrating injury (2), age ≥ 75 years (2), preexisting comorbidity (1), and torso injury (1), with a total range from 0 to 19. The modified GTOS demonstrated a significantly higher area under the curve (0.92 vs. 0.84, p < 0.0001), lower misclassification error (4.9% vs. 5.2%), and lower Brier score (0.036 vs. 0.042) compared to the original GTOS. DCA showed that using the modified GTOS for predicting in-hospital mortality resulted in higher NB than treating all, treating none, and treating based on the original GTOS across a wide range of clinician preferences.

The modified GTOS model exhibited superior predictive ability and clinical utility compared to the original GTOS. NB and DCA offer valuable complementary methods to calibration and discrimination indicators, comprehensively evaluating the clinical usefulness of prediction models and decision strategies.

An artificial intelligence (AI)-enabled electrocardiogram (ECG) model has been developed in a healthy adult population to predict ECG biological age (ECG-BA). This ECG-BA exhibited a robust correlation with chronological age (CA) in healthy adults and additionally significantly enhanced the prediction of aging-related diseases' onset in adults with subclinical diseases. The model showed particularly strong predictive power for cardiovascular and non-cardiovascular diseases such as stroke, coronary artery disease, peripheral arterial occlusive disease, myocardial infarction, Alzheimer's disease, osteoarthritis, and cancers. When combined with CA, ECG-BA improved diagnostic accuracy and risk classification by 21% over using CA alone, notably offering the greatest improvements in cancer prediction. The net reclassification improvement significantly reduced misclassification rates for disease onset predictions. This comprehensive study validates ECG-BA as an effective supplement to CA, advancing the precision of risk assessments for aging-related conditions and suggesting broad implications for enhancing preventive healthcare strategies, potentially leading to better patient outcomes.

Limited health literacy (HL) leads to poor health outcomes, psychological stress, and misutilization of medical resources. Although interventions aimed at improving HL may be effective, identifying patients at risk of limited HL in the clinical workflow is challenging. With machine learning (ML) algorithms based on readily available data, healthcare professionals would be enabled to incorporate HL screening without the need for administering in-person HL screening tools.

Develop ML algorithms to identify patients at risk for limited HL in spine patients.

Between December 2021 and February 2023, consecutive English-speaking patients over the age of 18 and new to an urban academic outpatient spine clinic were approached for participation in a cross-sectional survey study. HL was assessed using the Newest Vital Sign and the scores were divided into limited (0-3) and adequate (4-6) HL. Additional patient characteristics were extracted through a sociodemographic survey and electronic health records. Subsequently, feature selection was performed by random forest algorithms with recursive feature selection and five ML models (stochastic gradient boosting, random forest, Bayes point machine, elastic-net penalized logistic regression, support vector machine) were developed to predict limited HL.

Seven hundred and fifty-three patients were included for model development, of whom 259 (34.4%) had limited HL. Variables identified for predicting limited HL were age, Area Deprivation Index-national, Social Vulnerability Index, insurance category, Body Mass Index, race, college education, and employment status. The Elastic-Net Penalized Logistic Regression algorithm achieved the best performance with a c-statistic of 0.766, calibration slope/intercept of 1.044/-0.037, and Brier score of 0.179.

Elastic-Net Penalized Logistic Regression had the best performance when compared with other ML algorithms with a c-statistic of 0.766, calibration slope/intercept of 1.044/-0.037, and a Brier score of 0.179. Over one-third of patients presenting to an outpatient spine center were found to have limited HL. While this algorithm is far from being used in clinical practice, ML algorithms offer a potential opportunity for identifying patients at risk for limited HL without administering in-person HL assessments. This could possibly enable screening and early intervention to mitigate the potential negative consequences of limited HL without taxing the existing clinical workflow.

Estimating the genetic risk of coronary artery disease (CAD) is now possible by aggregating data from genome-wide association studies (GWAS) into polygenic risk scores (PRS). Combining multiple PRS for specific circulating blood lipids could improve risk prediction. Here, we sought to evaluate the performance of PRS derived from CAD and blood lipids GWAS to predict the incidence of CAD.

This study included individuals aged between 40 and 69 from UK Biobank. We conducted GWAS for blood lipids measured by nuclear magnetic resonance in individuals without lipid-lowering treatments (n = 73,915). Summary statistics were used to derive PRS in the remaining participants (n = 318,051). A PRSCAD was derived using the CARDIoGRAMplusC4D GWAS. Hazard ratios (HR) for CAD (n = 9017 out of 301,576; median follow-up: 12.6 years) were calculated per standard deviation increase in each PRS. Models' discrimination capacity and goodness-of-fit were evaluated.

Out of 30 PRS, 27 were significantly associated with the incidence of CAD (p < 0.0017). The optimal combination of PRS included PRS for CAD, VLDL-C, total cholesterol and triglycerides. Discriminative capacities were significantly increased in the model including PRSCAD and clinical risk factors (CRF) (C-statistic = 0.778 [0.773-0.782]) compared to the model with CRF only (C-statistic = 0.755 [0.751-0.760], difference = 0.022 [0.020-0.025]). Although the C-statistic remained similar when independent lipids PRS were added to the model with PRSCAD and CRF (C-statistic = 0.778 [0.773-0.783]), the goodness-of-fit was significantly increased (chi-square test statistic = 20.18, p = 1.56e-04).

Although independently associated with CAD incidence, blood lipids PRS provide modest improvement in the predictive performance when added to PRSCAD.

Post-thrombotic syndrome (PTS) is a common complication of deep vein thrombosis (DVT) that occurs in 20-50% of patients and results in a decreased quality of life. Even with the progressive identification of PTS risk factors, clinically useful predictors of PTS continue to be limited, unobjective, and ill-defined. The neutrophil-to-lymphocyte ratio (NLR) is an emerging prognostic biomarker used in a variety of diseases that reflects acute systemic inflammation. This pilot study aimed to evaluate the utility of the NLR at the time of iliofemoral DVT diagnosis in predicting PTS incidence in patients.

A retrospective chart review was performed on patients identified with iliofemoral DVT at the University of Maryland Medical Center between 2020 and 2022. Patients with at least one follow-up visit between 3 and 6 months after initial DVT diagnosis were included. Diagnosis of PTS was determined based on Villalta Score. The Youden index with receiver operating characteristic curve analysis was used to determine the NLR cut-off value that may be predictive of PTS. A multivariable logistic regression model was then performed to assess the utility of this NLR cut-off value and other common clinical markers in predicting the presence of PTS symptoms.

Four hundred and eighteen patients with positive iliofemoral DVT venous duplex ultrasounds were screened for eligibility. One hundred and eighteen patients were eligible with a mean age of 53.18 ± 15.45 years. A total of 43 patients (36.44%) were found to have PTS. An NLR cut-off of 7.71 was determined with an area under the receiver operating characteristic curve (area under the curve) of 0.63 (P = 0.046). When the NLR was assessed jointly with other clinical markers at the time of DVT diagnosis, NLR was a statistically significant positive predictor, measured using odds ratio (1.83; 95% confidence interval, 1.20-2.78; P = 0.005).

Our study found that when stratified by a determined cutoff value, the NLR at the time of DVT diagnosis was significantly associated with the development of PTS in patients with iliofemoral DVT. This result is consistent with one prior research finding yet is novel in its specificity for iliofemoral DVTs and its acute lab collection for NLR calculation. The NLR should be further investigated as a potential inexpensive prognostic tool to aid in the improvement of treatment and prophylactic strategies for PTS.

The suitability of the DOAC score for assessing bleeding risk in Chinese patients with atrial fibrillation (AF) who are receiving non-vitamin K antagonist oral anticoagulants (NOACs) remains unclear. We compared the DOAC score to the HAS-BLED and ORBIT scores in Chinese patients in a real-world retrospective study.

The efficacy of these scores was assessed by a comparison study that measured their discrimination, calibration, net reclassification index (NRI), and decision curve analysis (DCA) over a 1-year follow-up period.

Among 2532 patients with non-valvular AF (mean age, 71.7 ± 11.3 years, 58.5% men), major bleeding (MB) occurred in 91 patients (3.59%/year): 44 intracranial haemorrhage (ICH) events (1.74%/year) and 49 gastrointestinal bleeding (GB) events (1.94%/year). The best predictor for MB was the HAS-BLED score (area under the receiver operating characteristic curve [AUC], 0.674). HAS-BLED score ≥3 provided the best prediction for MB (AUC, 0.642), followed by DOAC score ≥8 and ORBIT score ≥4 (AUCs of 0.615 and 0.583, respectively). The DOAC and HAS-BLED scores did not differ significantly in discriminating MB events and risk reclassification. The calibration performance of the HAS-BLED score was superior to that of the other two scores. Decision curve analysis showed that using the HAS-BLED score to predict MB and ICH is clinically beneficial. However, there were no significant distinctions among the three models in forecasting GB.

In a non-valvular AF Chinese patients receiving NOACs, the HAS-BLED score showed an ability to predict MB comparable to that of the DOAC score and superior to that of the ORBIT score. The DOAC score does not seem to be more suitable for Chinese patients than the HAS-BLED score.

The atherogenic index of plasma (AIP) is a brand-new lipid parameter that has been used to assess various cardiovascular events. This study aimed to investigate the prognostic value of AIP in patients with pulmonary hypertension (PH).

This retrospective study was conducted at Shanghai Jiao Tong University School of Medicine affiliated Renji Hospital, and included data from 125 PH patients treated during 2014-2018. The endpoint events of this study were clinical worsening outcomes. PH patients include those from group 1 and group 4. AIP was determined as the logarithm of the blood triglycerides ratio to high-density lipoprotein cholesterol.

The 1-year, 3-year, and 5-year incidence rates of clinical worsening outcomes in PH patients in this study were 20.0, 44.8, and 54.4%, respectively. The median age of the PH patients was 38.00 years, with females accounting for 90.4%. After controlling for multivariable factors, the results of Cox regression analysis indicated that AIP was an independent predictor of adverse outcomes with a hazard ratio and 95% confident interval (CI) of 2.426 (1.021-5.763). The positive linear relationship of AIP was evaluated using restricted cubic spline analysis. Kaplan-Meier curves showed a significantly higher events rate in patients with AIP ≥ 0.144 compared to those with AIP < 0.144 (p = 0.002). Four potential prognostic variables, including AIP, were identified by LASSO regression to construct a nomogram. Compared to the model minus AIP, the AUC of the nomogram displayed a non-significant improvement (0.749 vs. 0.788, p = 0.298). In contrast, the results of net reclassification improvement (0.306, 95% CI: 0.039-0.459, p < 0.001) and integrated discrimination improvement (0.049, 95% CI: 0.006-0.097, p = 0.020) demonstrated significant enhancements in the predictive ability of the model when AIP was added to the clinical model.

AIP is an independent predictor of long-term clinical worsening in PH patients, and its inclusion in prognostic models could improve risk stratification and management.

To identify cardiovascular (CV) risk factors in Asian elderly aged 75 years and older and subsequently develop and validate a sex-specific five-year CV risk assessment tool for this population.

This study included 12,174 patients aged ≥ 75 years without a prior history of cardiovascular disease at a single hospital in Taiwan. Electronic health records were linked to the National Health Insurance Research Database and the National Death Registry to ensure comprehensive health information. Eligible patients were randomly divided into derivation (80%) and validation (20%) cohorts. A sex-specific CV risk assessment tool was developed to predict major adverse cardiovascular events (MACE) using Cox regression modeling.

During a median follow-up period of 8.6 years for men and 8.5 years for women in the derivation cohort, MACE occurred in 3.62% of men and 3.02% of women. Predictors for men comprised advanced age, smoking, non-HDL-C levels > 160 mg/dL, metastatic cancer, and aspirin usage. Predictors for women included advanced age, smoking, atrial fibrillation, cancer, dementia, osteoarthritis, systemic lupus erythematosus, use of antihypertensives, and use of oral anticoagulants. In the validation cohort, the sex-specific risk assessment tool demonstrated fair discriminative power (AUC: men, 0.64; women, 0.68). Model calibration demonstrated good performance for women but was less optimal for men.

This sex-specific CV risk assessment tool shows fair discriminative capability in estimating risk of cardiovascular disease among elderly Asians, potentially enabling targeted interventions in this vulnerable population.

Currently, there are no standardized guidelines for graft allocation in heart transplants (HTxs), particularly when considering organs from marginal donors and donors after cardiocirculatory arrest. This complexity highlights the need for an effective risk analysis tool for primary graft dysfunction (PGD), a severe complication in HTx. Existing score systems for predicting PGD lack superior predictive capability and are often too complex for routine clinical use. This study sought to develop a user-friendly score integrating variables from these systems to enhance the efficacy of the organ allocation process.

Severe PGD was defined as the need for mechanical circulatory support and/or death from an unknown etiology within the first 24 hours following HTx. We used a meta-analytical approach to create a derivation cohort to identify risk factors. We then applied a logistic regression analysis to generate an equation predicting severe PGD risk. We used our previous experience in HTx to create a validation cohort. Subsequently, we implemented the formula in a smartphone application.

The meta-analysis comprising six studies revealed a 10.5% ( 95% confidence interval (CI): 5.3-12.4) incidence rate of severe PGD and related 30-day mortality of 38.6%. Eleven risk factors were identified: female donors, female donor to male recipient, undersized donor, donor age, recipient on ventricular assist device support, recipient on amiodarone treatment, recipient with diabetes and renal dysfunction, re-sternotomy, graft ischemic time, and bypass time. An equation to predict the risk, including the 11 parameters (GREF-11), was created using logistic regression models and validated based on our experience involving 116 patients. In our series, 29 recipients (25%) required extracorporeal membrane oxygenation support within 24 hours post-HTx. The overall 30-day mortality was 4.3%, 3.4%, and 6.8% in the non-PGD and severe PGD groups, respectively. The area under the receiver operating characteristic (AU-ROC) curve of the model in the validation cohort was 0.804.

The GREF-11 application should offer HTx teams several benefits, including standardized risk assessment and bedside clinical decision support, thereby helping minimize the risk of severe PGD post-HTx.

The proper use of model evaluation metrics is important for model evaluation and model selection in binary classification tasks. This study investigates how consistent different metrics are at evaluating models across data of different prevalence while the relationships between different variables and the sample size are kept constant. Analyzing 156 data scenarios, 18 model evaluation metrics and five commonly used machine learning models as well as a naive random guess model, I find that evaluation metrics that are less influenced by prevalence offer more consistent evaluation of individual models and more consistent ranking of a set of models. In particular, Area Under the ROC Curve (AUC) which takes all decision thresholds into account when evaluating models has the smallest variance in evaluating individual models and smallest variance in ranking of a set of models. A close threshold analysis using all possible thresholds for all metrics further supports the hypothesis that considering all decision thresholds helps reduce the variance in model evaluation with respect to prevalence change in data. The results have significant implications for model evaluation and model selection in binary classification tasks.

Infections and cancer are major causes of premature death in organ recipients. However, there have been few advances in personalized immunosuppressive therapy. We previously reported that a cell-based assay measuring CD14+16+tumor necrosis factor-α+ monocytes after peripheral blood mononuclear cell (PBMC) incubation with Epstein-Barr virus peptides has a high sensitivity for detecting over-immunosuppression (OIS) events in kidney recipients in the short term. We aimed to develop a risk score for predicting long-term events.

We studied 551 PBMC samples from 118 kidney recipients. Following random allocation to a testing and training set, we derived a risk function for the delineated tertiles of low, intermediate, and high risk of OIS based on age and CD14+16+tumor necrosis factor-α+ cells.

Patients were followed for a median of 6.3 y (25th-75th percentiles: 3.7-8.3 y). Of these, 40 (34%) experienced an OIS event. The validation indicated that the risk score was well calibrated, with an absolute risk of 21%, 41%, and 61% in the low-, intermediate-, and high-risk categories, respectively (P = 0.03). In sensitivity analyses, the risk score was robust to alternative definitions of OIS ranging from mild to severe. In particular, when restricted to life-threatening OIS, the proportion of events varied from 5% to 27% among the low- and high-risk categories, respectively (P = 0.01).

Using a combination of age and in vitro PBMC response to Epstein-Barr virus peptides allows a substantial shift in the estimated risk of OIS events.

Differences in predictability of ablation success for premature ventricular contractions (PVCs) between earliest isochronal map area (EIA), local activation time (LAT) differences on unipolar and bipolar electrograms (⊿LATBi-Uni), LAT prematurity on bipolar electrograms (LATBi), and unipolar morphology of QS or Q pattern remain unclear. We verified multiple statistical predictabilities of those indicators of ablation success on mapped cardiac surface.

Thirty-five patients with multiple PVCs underwent catheter ablation after LAT mapping using multipolar mapping catheters with unipolar-based annotation. Patients were divided into success and failure groups based on ablation success on mapped cardiac surfaces. Discrimination ability, reclassification table, calibration plots, and decision curve analysis of 10 ms EIA (EIA10ms), ⊿LATBi-Uni, and LATBi were validated. Unipolar morphology was compared between success and failure groups.

Right ventricular outflow tract, aortic cusp, and left ventricle were mapped in 17, 10, and 8 patients, respectively. In 14/35 (40%) patients, successful ablation was performed on mapped cardiac surfaces. Area under the curve of receiver-operating characteristic curve of EIA10ms, ⊿LATBi-Uni, and LATBi were 0.874, 0.801, and 0.650, respectively (EIA10ms vs. LATBi, p =.014; ⊿LATBi-Uni vs. LATBi, p =.278; EIA10ms vs. ⊿LATBi-Uni, p =.464). EIA10ms and ⊿LATBi-Uni demonstrated better predictability, calibration, and clinical utility on reclassification table, calibration plots, and decision curve analysis than LATBi. Unipolar morphology of QS or Q pattern did not correlate with ablation success (p =.518).

EIA10ms and ⊿LATBi-Uni more accurately predict ablation success for PVCs on mapped cardiac surfaces than LATBi and unipolar morphology.

Mass cytometry enables deep profiling of biological samples at single-cell resolution. This technology is more than relevant in cancer research due to high cellular heterogeneity and complexity. Downstream analysis of high-dimensional datasets increasingly relies on machine learning (ML) to extract clinically relevant information, including supervised algorithms for classification and regression purposes. In cancer research, they are used to develop predictive models that will guide clinical decision making. However, the development of supervised algorithms faces major challenges, such as sufficient validation, before being translated into the clinics. In this work, we provide a framework for the analysis of mass cytometry data with a specific focus on supervised algorithms and practical examples of their applications. We also raise awareness on key issues regarding good practices for researchers curious to implement supervised ML on their mass cytometry data. Finally, we discuss the challenges of supervised ML application to cancer research.

Cardiovascular disease (CVD) is the leading cause of mortality in women. We aimed to assess whether adding female-specific risk factors to traditional factors could improve CVD risk prediction.

We used a cohort of women from the UK Biobank Study aged 45 to 69 years, free of CVD at baseline (2006-2010) followed until the end of 2019. We developed Cox proportional hazards models using the risk factors included in 3 contemporary CVD risk calculators: Pooled Cohort Equation - Atherosclerotic Cardiovascular Disease, Qrisk2, and PREDICT. We added each of the following female-specific risk factors, individually and all together, to determine if these improved measures of discrimination and calibration for predicting CVD: early menarche (<11 years), endometriosis, excessive, frequent or irregular menstruation, miscarriage, number of miscarriages, number of stillbirths, infertility, preeclampsia or eclampsia, gestational diabetes (without subsequent type 2 diabetes), premature menopause (<40 years), early menopause (<45 years), and natural or surgical early menopause (menopause <45 years or timing of menopause reported as unknown and oophorectomy reported at age <45).

In the model of 135 142 women (mean age, 57.5 years; SD, 6.8) using risk factors from Pooled Cohort Equation - Atherosclerotic Cardiovascular Disease, CVD incidence was 5.3 per 1000 person-years. The c-indices for the Pooled Cohort Equation - Atherosclerotic Cardiovascular Disease, Qrisk2, and PREDICT models were 0.710, 0.713, and 0.718, respectively. Adding each of the female-specific risk factors did not improve the c-index, the net reclassification index, the integrated discrimination index, the slope of the regression line for predicted versus observed events, and the Brier score or plots of calibration. Adding all female-specific risk factors simultaneously increased the c-index for the Pooled Cohort Equation - Atherosclerotic Cardiovascular Disease, Qrisk2, and PREDICT models to 0.712, 0.715, and 0.720, respectively.

Although several female-specific factors have been shown to be early indicators of CVD risk, these factors should not be used to reclassify risk in women aged 45 to 69 years when considering whether to commence a blood pressure or lipid-lowering medication.