Risk Prediction Score Research Articles

Development and validation of a disulfidptosis-related scoring system to predict clinical outcome and immunotherapy response in acute myeloid leukemia by integrated analysis of single-cell and bulk RNA-sequencing.

Background: Disulfidptosis is a metabolically relevant mode of cell death, and its relationship with acute myeloid leukemia (AML) has not been clarified. In this study, disulfidptosis scores were computed to examine the potential biological mechanisms. Methods: Consensus clustering was applied to detect disulfidptosis-related molecular subtypes. The least absolute shrinkage and selection operator (LASSO) regression analysis was used to construct a DRG prognostic model. Results: DRGs are upregulated in AML and associated with poor prognosis. The higher the disulfidptosis activity score, the worse the clinical outcome for patients, accompanied by increased immune checkpoint expression and tumor marker pathway activity. The two molecular subtypes exhibited distinct prognoses and tumor microenvironment (TME) profiles. A prognostic risk score model was established using six DRGs, and the AML cohort was divided into high- and low-risk score groups. Patients in the high-risk group experienced significantly worse prognosis, which was validated in seven AML cohorts. Receiver Operating Characteristic (ROC) curve analysis indicated that the area under the curve values for risk score prediction of 1-, 3-, and 5-year survival were 0.779, 0.714, and 0.778, respectively. The nomogram, in conjunction with clinicopathological factors, further improved the accuracy of prognosis prediction. The high-risk score group exhibited a higher somatic mutation frequency, increased immune-related signaling pathway activity, and greater immune checkpoint expression, suggesting a certain degree of immunosuppression. Patients with advanced age and higher cytogenetic risk also had elevated risk scores. According to drug prediction and AML anti-PD-1 therapy cohort analysis, the low-risk score group displayed greater sensitivity to chemotherapy drugs like cytarabine and midostaurin, while the high-risk score group was more responsive to anti-PD-1 therapy. Finally, clinical samples were collected for sequencing analysis, confirming that the progression of myeloid leukemia was associated with a higher risk score and a negative disulfidptosis score, suggesting that the poor prognosis of AML may be associated with disulfidptosis resistance. Conclusion: In conclusion, a systematic analysis of DRGs can help to identify potential disulfidptosis-related mechanisms and provide effective new biomarkers for prognosis prediction, TME assessment, and the establishment of personalized treatment plans in AML.

Analysis across Taiwan Biobank, Biobank Japan, and UK Biobank identifies hundreds of novel loci for 36 quantitative traits

Genome-wide association studies (GWASs) have identified tens of thousands of genetic loci associated with human complex traits. However, the majority of GWASs were conducted in individuals of European ancestries. Failure to capture global genetic diversity has limited genomic discovery and has impeded equitable delivery of genomic knowledge to diverse populations. Here we report findings from 102,900 individuals across 36 human quantitative traits in the Taiwan Biobank (TWB), a major biobank effort that broadens the population diversity of genetic studies in East Asia. We identified 968 novel genetic loci, pinpointed novel causal variants through statistical fine-mapping, compared the genetic architecture across TWB, Biobank Japan, and UK Biobank, and evaluated the utility of cross-phenotype, cross-population polygenic risk scores in disease risk prediction. These results demonstrated the potential to advance discovery through diversifying GWAS populations and provided insights into the common genetic basis of human complex traits in East Asia.

Comparison of risk prediction scores for cancer-associated thrombosis in ambulatory cancer patients not receiving thromboprophylaxis: A prospective cohort study in a South-East Asian population

IntroductionTo date, there is limited data regarding the incidence and risk prediction of cancer-associated thrombosis among South-East Asian patients who do not receive thromboprophylaxis. Materials and methodsThis was a prospective cohort study conducted at a tertiary medical center from June 2020 to December 2021 in Thailand. We enrolled cancer patients aged ≥ 18 years, with ECOG score ≤ 1, scheduled to receive the first cycle of chemotherapy. We measured incidence of venous thromboembolism (VTE), all-cause mortality and performance of risk prediction scores. ResultsA total of 457 patients were included with a mean age of 58.18 ± 12.60 years. By the end of 6 months period, VTE had occurred in 30 patients (6.56 %, 95%CI 4.36–9.21). The median time to the first thrombosis was 1.94 months (IQR 0.26–3.19). Cancer associated thrombosis incidence was 14.58 % for Khorana score ≥ 3, 6.67 % for scores 1–2 and 2.13 % for score 0. C-statistics were 0.50 (95%CI 0.41–0.60) for Khorana score cut-off ≥ 2, 0.57 (95%CI 0.49–0.65) for Khorana score ≥ 3, 0.55 (95%CI 0.46–0.65) for PROTECHT score ≥ 3, and 0.57 (95%CI 0.49–0.65) for CONKO score ≥ 3. Classifying cholangiocarcinoma as very-high-risk increased the Khorana score cut-off ≥ 3's C-statistic to 0.62 (95%CI 0.53–0.71). ConclusionsA significant proportion of ambulatory South-East Asian cancer patients without thromboprophylaxis developed VTE. Further prospective studies investigating the benefit of thromboprophylaxis in high-risk patients with active cancer are warranted.

REVEAL Score performance in patients with WHO Group 2 pulmonary hypertension

Abstract Introduction Pulmonary hypertension (PH) secondary to left heart disease, also known as World Health Organization (WHO) Group-2 pulmonary hypertension (WHO-2 PH) is the most common form of PH worldwide. Various risk prediction scores have been derived and validated for other forms of PH, including the REVEAL 2.0 and REVEAL Lite 2 scores. However, whether these scores accurately predict outcomes in patients with WHO-2 PH remains to be investigated. Methods We retrospectively analyzed patients referred to the PH Clinic at the Univ. of Cincinnati from 2016-2022. Only patients with WHO-2 PH were studied. REVEAL 2.0 and REVEAL lite 2 scores were calculated for all patients at the first outpatient visit (OV). Longitudinal assessment was only available for the REVEAL lite 2 score up to 36-48 months. Patients were stratified as low, intermediate, and high-risk and Kaplan-Meier and Cox proportional hazards statistics were used to assess survival based on risk stratification. Concordance index was calculated to evaluate the performance of each independent score in discriminating survivors vs. non-survivors at one year from the first OV (OV-1). Results Total of 214 patients were studied. Mean age was 64.7 years (13.4 SD), and 66.4 % were female. Patients demographic and clinical characteristics are shown in Table 1. At the OV-1, 15% of patients were stratified as low-risk, 27% as intermediate, and 57% as high-risk (Figure 1). Estimated survival (95% confidence intervals [CIs]) at 1 year was 100% [CI 99-100], 96% [CI 91-100], and 89% [CI 84-95] for low, intermediate, and high-risk patients, respectively. Median survival for high-risk patients was 4.6 years [CI 3.33-NA]. Only 98 patients were alive or had sufficient data to calculate REVEAL Lite 2 score. After OV-1, 13% (n=13) of patients changed to a low-risk strata, 57% (n=56) remained at the same level and 30% (n=29) changed to a high-risk strata. Total of 28% (n=68) of patients died within study period. Cox proportional hazards analysis showed that patients classified as high-risk based on REVEAL 2.0 (³9) or REVEAL Lite 2 (³8) had a statistically higher risk of death (HR 5.32 [CI 2.27-12.46], p<0.001 and HR 4.09 [CI 1.62-10.27], p= <0.01 compared to low-risk patients. The C-indices for REVEAL 2.0 and REVEAL lite 2 were 0.67 (p<0.001) and 0.65 (p<0.001). We found no difference in the risk of death between patients stratified as intermediated when compared to low-risk using either score. Conclusions In this single center, retrospective study, the REVEAL 2.0 and REVEAL Lite 2 risk prediction scores were able to differentiate patients at low vs high-risk of death. However, the scores failed to predict mortality in patients categorized at intermediate risk. Importantly, most of our patient population fell within this category at first visit and frequently remained at intermediate risk during the study period. Revisions to these scores are warranted to account for differences in this patient population.Table 1Figure 1

The added prognostic value of stress induced hyperglycemia to the GRACE risk score for the prediction of 1-year major adverse cardiovascular events in patients with ST-elevation myocardial infarction

Abstract Background There is growing evidence suggesting that Stress Induced Hyperglycemia (SIH) is a risk factor of poor prognosis in acute coronary syndrome (ACS)[1–3]. However, this parameter is not included in risk prediction scores, applied in clinical practice. Ιmportantly, there is no universally accepted definition for SIH especially in the setting of ACS. Across the existing evidence base, SIH is defined either based on admission blood glucose (ABG) levels or by calculating the glycemic gap and the stress hyperglycemia ratio (SHR)[4]. Purpose Considering the variability in SIH definitions and the need for further investigation on the long-term prognostic impact of SIH, we present herein original data concerning the prognostic role of SIH for long-term risk of major adverse cardiovascular and cerebrovascular events (MACCEs) in patients presenting with ST-elevation myocardial infarction (STEMI) on top of a traditional prediction risk score such as the GRACE 2.0 risk score. Methods Our data were derived from a post-hoc analysis of a prospective study (NCT04580173). Every trial procedure was conducted according to the principles set by the declaration of Helsinki. Each participant provided written informed consent before being enrolled in the study. Our analysis included 309 consecutive STEMI patients who were diagnosed with SIH if ABG was > 140 mg/dl and a fasting period of at least 8 hours was confirmed. Each participant was followed-up for a median 1-year period. Baseline characteristics of the included patients are presented in Table 1. The records of all enrolled patients were retrospectively analyzed to calculate the GRACE 2.0 risk score for each patient. Results Cox regression analysis adjusted for age, gender, body mass index (BMI), hypertension, smoking, and SIH was performed but did not identify GRACE risk score>140 as an independent predictor of MACCE (aHR=1.408, 95% CI:0.764-2.594 p=0.273, Figure 1). On the other hand, a similar model adjusted for the same covariates identified the combination of GRACE risk score>140 and SIH as a significant predictor of MACCE occurrence (aHR=2.149, 95% CI:1.224-3.771, p=0.008). Conclusions Our results suggest that SIH combined with the GRACE 2.0 risk score was associated with improved prediction of the risk for 1-year MACCE in STEMI patients as compared to the GRACE 2.0 risk score alone. In this study, SIH was diagnosed based on the levels of admission blood glucose (after 8-hours of fasting) and not based on the SHR. Despite the heterogeneity in SIH definitions, stress hyperglycemia in the setting of STEMI might be an underrated yet robust prognostic marker not only for short term but also for long term adverse clinical events, especially when integrated in existing clinical risk scores.Survival curveBaseline characteristics

Post-menopause women show increased levels in risk prediction scores with age as dependent variable

Abstract Background Men and women differ in outcome after percutaneous coronary intervention (PCI). However, frequently used scores for ischemic risk prediction in patients with cardiovascular disease do not include sex related factors. In this study, we aimed to analyze sex related differences in risk score prediction in patients after PCI in dependence from age-determined menopause. Methods In this mono-center analysis, we compared frequently used risk prediction scores for post PCI patients (GRACE [Global Registry of Acute Coronary Events], PRECISE-DAPT [PREdicting bleeding Complications In patients undergoing Stent implantation and subsEquent Dual Antiplatelet Therapy], NCDR-mortality and PARIS [Patterns of Non-Adherence to Anti-Platelet Regimen in Stented Patients]-thrombotic score) in male versus female post PCI patients in dependence from menopause (age determined) and age itself. Three-year follow-up was conducted to assess the incidence of MACCE (major adverse cerebro- and cardiovascular events) including all-cause mortality, myocardial infarction and stroke. Receiver operating characteristic (ROC) analysis was conducted to analyze predictive capacity of each score. Results 1,001 patients post PCI were included (38% chronic coronary syndrome [CCS], 61% acute coronary syndrome [ACS]). Out of these, 765 were male and 306 were female. Women were slightly older (68.1±11.5 vs. 73.3±11.3 years, p<0.0001). Male and female patients did not differ in achieved levels of risk score below the age of 55. In contrast, in the subgroups >55 years, GRACE, PD-DAPT and NCDR Mortality scores were higher in woman compared to men (GRACE – 120.5±27.89 vs. 128.5±29.97, p<0.0001; PD-DAPT – 26.46±13.13 vs. 30.66±12.76, p<0.001; NCDR Mortality – 345.79±18.33 vs. 40.8±20.62, p<0.0001). Interestingly, PARIS thrombotic score was reduced (4.02±2.28 vs. 3.6±2.3, p=0.027). After Propensity score matching for age, GRACE, PD-DAPT and NCDR Mortality did no longer differ whereby PARIS thrombotic was reduced in female patients in this analysis as well. ROC revealed a moderate predictive capacity for MACCE for all scores in this post PCI cohort with a numerically higher predictive capacity for women (GRACEAUC – 0.625 vs. 0.662; PD-DAPTAUC 0.589 vs. 0.660, NCDR-mortalityAUC 0.625 vs. 0.667; PARIS-thromboticAUC 0.594 vs. 0.638). Conclusion In this single center analysis of post PCI patients with ACS and CCS, we could reveal that the GRACE, PD-DAPT and NCDR-mortality risk scores show higher values in post-menopause women compared to men whereby the PARIS-thrombotic score was reduced. We revealed age as major driver for this discrepancy. Moreover, women showed a numerically higher predictive accuracy for the occurrence of MACCE in line with the demonstrated worse outcome after PCI in several cohorts.

Carotid plaque score for stroke and cardiovascular risk prediction in a middle-aged cohort from the general population: the akershus cardiac examination 1950 study

Abstract Background Assessment of carotid atherosclerosis with ultrasound may have an unfilled potential in predicting stroke and cardiovascular events. Purpose We aimed to explore the predictive value of the carotid plaque score compared to the Systematic COronary Risk Evaluation (SCORE) 2 risk prediction algorithm, on incident stroke and major adverse cardiovascular events (MACE), and establish a prognostic cut-off for the carotid plaque score. Methods In the prospective Akershus Cardiac Eexamination (ACE) 1950 cohort study, carotid plaque score was calculated with ultrasound at inclusion in 2012-2015. The largest plaque diameter in each extra cranial segment of the carotid artery both sides was measured and scored from 0-3 points. The sum of each segment provided the carotid plaque score. Incident stroke and composite endpoint MACE (nonfatal stroke, nonfatal myocardial infarction, and cardiovascular death) was assessed by linkage to national registries throughout 2020. Results Carotid plaque score was available in 3650 (98.5%) of the participants, with mean age at inclusion was 63.9±0.64 years. Only 462 (12.7%) were free of plaque and the median carotid plaque score was 2 (IQR 1-4). By the end of 2020, 42 (1.2%) subjects had experienced a stroke and 145 (4.0%) a MACE. Carotid plaque score was a predictor of incident stroke (HR 1.25, 95%CI 1.15-1.36) and MACE (OR 1.21, 95%CI 1.14-1.27) after adjustment for SCORE2, and outperformed SCORE2 in predicting stroke (p=0.001). The best cut-off value for carotid plaque score determined by receiver operator characteristics area under the curve was ≥4 with a positive predictive value 2.5% and negative predictive value 99.3%. Conclusion Carotid plaque score is a strong predictor of incident stroke and MACE and outperforms SCORE2 for risk prediction in a middle-aged cohort recruited from the general population. A cut-off score of ≥4 seems to be suitable to identify high-risk subjects.Graphical AbstractCarotid plaque score distribution

The association between HbA1c variability and strokes in patients receiving anticoagulants

Abstract Background Diabetes, higher glycaemic level, is associated with an increased risk of stroke in patients with atrial fibrillation (AF) and is integrated into the CHA2DS2-VASc score for stroke risk prediction. However, the association between glycaemic variability and stroke risk in AF patients remains unclear. Objective To evaluate the association between glycaemic variability and stroke among diabetic and non-diabetic patients with nonvalvular AF receiving anticoagulants. Methods Newly diagnosed AF from 2013-2018 (n=8,790) were divided by their diabetes status. Variability in glycaemic, represented by coefficient of variation (CV) on HbA1c (haemoglobin A1c), was assessed by at least three HbA1c measurements following enrollment to the end of the follow-up period. Participants with ischaemic stroke (IS)/systemic embolism (SE), haemorrhagic stroke (HS) or all-cause mortality at baseline or within 1 year of enrollment were excluded. Multivariate Cox regression was performed to evaluate the association between variability in HbA1c and subsequent incident IS/SE, HS and all-cause mortality. Sensitivity analyses included defining HbA1c variability using average successive variability (ASV) and variability independent of the mean (VIM). Results Diabetes status significantly modified the relationship between HbA1c variability and risk of IS/SE (Pinteraction = 0.36)and all-mortality (Pinteraction <0.01). Greater incidence rates were observed among individuals with diabetes, compared to those without diabetes (3.74 vs. 2.41 for IS/SE, 0.54 vs. 0.33 for HS and 4.89 vs. 2.42 for all-cause mortality per 100 person years). Among those with diabetes, greater HbA1c variability was significantly associated with increased risk of IS/SE (hazard ratio [HR] = 1.21, 95% confidence interval [CI]: 1.11-1.31), all-cause mortality (HR = 1.13, 95%CI: 1.05-1.20) and HS (HR = 1.10, 95%CI: 1.02-1.23). Similar patterns were found in individuals without diabetes (IS/SE: HR = 1.28, 95%CI: 1.08-1.52; all-cause mortality: HR = 1.50, 95%CI: 1.29-1.75). However, no significant association was observed between glycemia variability and HS among those without diabetes. Results were consistent with alternative measures of HbA1c variability. Conclusion Variability in HbA1c was independently associated with the risk of IS/SE and all-cause mortality among patients with AF, regardless of diabetes status.

Population prevalence of ECG abnormalities and risk of incident CVD outcomes: 5-year follow-up of 25,000 Chinese adults

Abstract Background and purpose Little is known about the utility of automated reporting of electrocardiograms (ECGs) for prediction of cardiovascular disease (CVD) types in low- and middle-income countries. We compared derived ECG phenotypes and a clinical risk score for prediction of incident major CVD types in a population study of Chinese adults. Methods CVD risk factors were collected and 12-lead ECGs performed with automated interpretation using Mortara’s VERITAS™ algorithm on 25,239 adults in the China Kadoorie Biobank resurvey in 2013-2014. Incident CVD outcomes were recorded prospectively over a median follow-up of 5 years. We examined the prevalence and prognosis of broad ECG phenotypes developed from automated outputs: (i) atrial fibrillation/flutter (AFib); (ii) ischaemia (possible/probable or definite); (iii) left ventricular hypertrophy (LVH); and (iv) normal ECG. Participants were stratified by modified CHA2DS2-VA score (<3 vs ≥3) and by age at resurvey. Kaplan-Meier curves were plotted to display cumulative CVD event rates during follow-up for ECG phenotypes. Results The mean (SD) age at resurvey was 59.5 (10.2) years, 62% were women and 57% were from rural areas. Overall, 44.3% had a normal ECG and 1.2% had AFib (0.4%, 0.9%, and 3.9% at ages <55, 55-69, and 70+). A total of 28.1% had ischaemia, and 13.6% had LVH (Table). The prevalence of all pathological ECG phenotypes increased with age and prior CVD. AFib was more common in individuals with CHA2DS2-VA ≥3 vs <3 (3.6% vs 0.8%). CHA2DS2-VA ≥3 was associated with higher HRs (95% CI) for stroke (1.91; 1.67-2.19), ischaemic heart disease (IHD) (1.79; 1.52-2.11), and heart failure (HF) (1.80; 1.23-2.62), respectively. Compared to CHA2DS2-VA ≥3, HRs for the AFib phenotype were greater for IHD (2.62; 1.97-3.50) and HF (3.79; 2.21-6.49) and similar for stroke (1.88; 1.44-2.77). LVH was most strongly associated with HF (1.83; 1.31-2.55) and positive associations with each CVD type were greater for definite vs probable/possible ischaemic phenotypes. After adjustment for age, 5-year cumulative incidence rates of CVD were 36.8% in individuals with AFib, 19.4% in those with ischaemia (all), 18.0% in those with LVH, and 14.1% in those with normal ECGs. Individuals with CHA2DS2-VA ≥3 had a 3-fold greater cumulative incidence of CVD than those with scores <3 (40.9% vs 13.9%). Conclusions All pathological ECG phenotypes were associated with higher risks of CVD outcomes, but AFib was most strongly predictive and comparable to moderate-high CHA2DS2-VA scores. However, absolute risks of subsequent CVD events were similar in those with ischaemia, LVH and a normal ECG. When combined with clinical risk scores, automated ECG interpretation is a useful adjunct in low-resource settings for CVD risk prediction in population studies and in clinical practice.

Serum proteomic profiling of carotid atherosclerosis: a population outcome study

Abstract Background Circulating proteins reflecting atherosclerosis may improve cardiovascular (CV) disease prediction. We applied feature selection and unsupervised clustering on proteomic data to identify proteins associated with carotid atherosclerosis and construct a protein-based classification system for prediction of CV events. Methods 491 community-dwelling participants (mean age, 58±11 years; 51% women) underwent carotid ultrasonography and proteomic profiling (CVD II panel, Olink Proteomics). Subsequently, CV outcome was collected (median follow-up time: 10.2 years). We applied partial least squares (PLS) to identify proteins linked to carotid intima-media thickness (cIMT). Next, we assessed the association between future CV events and protein-based phenogroups derived by unsupervised clustering (Gaussian Mixture modelling) based on proteins influential in PLS. Results PLS marked 19 proteins as important, which were all associated with cIMT in multivariable-adjusted linear regression. 8 of the 19 proteins were excluded from the clustering due to high collinearity. Based on the 11 remaining proteins, the clustering algorithm subdivided the cohort into two phenogroups. Compared to the first (n=177), participants in the second phenogroup (n=314) presented: i) a more unfavourable lipid profile with higher total cholesterol and triglycerides and lower HDL cholesterol (P≤0.014); ii) higher cIMT (P=0.0020); and iii) a significantly higher risk for future CV events (multivariable-adjusted hazard ratio (95% CI) versus cluster 1: 2.05 (1.26 to 3.52); P=0.0093). The protein-based phenogrouping supplemented the ACC/AHA 10-year ASCVD risk score for prediction of a first CV event. Conclusions Phenogrouping based on a focused set of proteins related to carotid atherosclerosis identified individuals at high risk for future CV events. This approach may complement current CV risk stratification strategies.Summary of workflow and findingsCV event prediction

Improving risk stratification of recurrent myocardial infarction in a large real-world dataset using machine learning

Abstract Background Patient (pt) risk stratification for cardiovascular (CV) events is key to determine the appropriate approach to secondary prevention. Several risk scores have been developed (SCORE2, EUROASPIRE, Thrombolysis in Myocardial Infarction [TIMI] risk score for secondary prevention [TRS 2°P]); these are based on few predictors and have low discriminatory power. Machine learning methods using real-word data (RWD) from electronic health records (EHR) databases and capable of handling a multitude of commonly available variables can serve as powerful tools to improve the discriminatory power and generate efficient and clinically relevant risk scores. Purpose This study used a non-linear machine learning model to build an improved risk score for secondary CV events from a large Israeli EHR database. Methods Records of pts with a first record of atherosclerotic CV disease (ASCVD) admission or coronary procedure and available follow-up data (mean: 1052 days) from the period of 01/2005–11/2020 were considered (n=77432 [67% coronary artery disease, 21% carotid artery disease, 12% peripheral artery disease], a stratified 1:1 randomization by cause of admission yielding in training set [n=38716] and test set [n=38716]). The model’s performance was assessed by comparison with TRS 2°P score on the test set independent of the training set. Twenty-six covariates commonly used in clinical routine and widely available in EHR have been selected and a gradient boosting trees algorithm was applied to the pt data at baseline after the original event to predict recurrent myocardial infarction (MI). The ability of the model to discriminate pts was evaluated using the concordance index (c-index), which expresses how well a predicted risk score describes the observed events (1 implies perfect concordance, 0.5 implies random concordance). Results Prediction accuracy of our model showed a c-index for a future MI of 0.74 (95% confidence interval [CI]: 0.72, 0.76) which is significantly higher than 0.57 (95% CI: 0.56, 0.58) obtained with TRS 2°P. Our 2-tier risk score has successfully discriminated a low-risk group (74% of cohort) with a 3-year MI prevalence of 5% from a high-risk group (26% of cohort) that had a 5-fold higher prevalence (26.6%) (Figure 1). In comparison, the 3-tier TRS 2°P yielded an unbalanced distribution of group sizes with a disproportionate intermediate risk group (22% low, 67% intermediate, 11% of high-risk) with a limited clinical discriminatory power (7.1%, 10,2% and 15.6% of observed events, respectively) (Figure 2). Conclusions Our machine learning based risk score model created on a large set of RWD with a long follow-up seems to offer substantial gain of performance compared to the TRS 2°P. The present 2-tier model shows significant gain of accuracy and clinical discriminatory power paving the way for more individualized prevention strategies.Figure 1Figure 2

Abstract 12476: QT Prolongation Predicts All-Cause Mortality Above and Beyond a Validated Risk Score

Introduction: QT prolongation is a risk factor for life-threatening arrhythmias and sudden cardiac death. In large cohorts, QT interval was associated with all-cause mortality after adjusting for cardiovascular risk factors, but these analyses may contain residual confounding. Whether the QT interval provides prognostic information above and beyond a validated mortality risk score is unknown. Hypothesis: QT interval on ECG will independently predict mortality after adjustment for the Care Assessment Needs (CAN) score, which was validated to predict mortality nationwide at the Veterans Administration (VA) (c-index 0.86). Methods: Patients receiving primary care at the Minneapolis VA from 2012 to 2016, with at least one ECG were included in this retrospective cohort study. QT intervals were corrected for heart rate (QTc) using the Bazett’s formula. CAN score, calculated within 1-week of the ECG, was obtained from the VA Corporate Data Warehouse. Results: Of the 36,240 patients, 1526 (4.2%) had QTc ≥ 500 ms, 3249 (9.0%) had QTc 470-500 ms, and 31,465 (86.8%) had QTc < 470 ms. A total of 2,189 (6.0%) patients died at 1-year follow-up and 7898 (21.8%) died at 5-year follow-up. Mortality was higher in patients with longer QTc (Table 1). Compared to those with QTc < 470 ms, CAN-adjusted odds of 1-year and 5-year mortality were significantly higher in those with longer QTc (Table 2). C-index for CAN score predicting 1-year and 5-year mortality (0.829 and 0.826, respectively) improved marginally after adding QTc data (0.834 and 0.832, respectively; p <0.0001 for both). Conclusions: QT prolongation predicts all-cause mortality independently of a validated mortality risk prediction score.

Abstract 16327: Machine Learning-Based Prediction of Clinical Frailty Scale Based on Gait Analysis in Elderly Patients With Heart Failure

Background: Although frailty assessment is recommended for guiding treatment strategies and outcome prediction in elderly patients with heart failure (HF), most frailty scales are subjective, and the scores vary among raters. We sought to develop a machine learning-based automatic rating method/system/model of the clinical frailty scale (CFS) for patients with HF. Methods: We prospectively examined 358 elderly (≥75 years) with symptomatic chronic HF patients (mean age 82.3±5.2 years, left ventricular ejection fraction 58 [IQR 45-67]%) from six sites between January 2019 and September 2022. The CFS was evaluated based on the modified Delphi method by 10 independent trained cardiologists as the gold standard. We obtained body-tracking motion data using OpenPose ® , a deep learning-based pose estimation library, on a smartphone camera. Predicted CFS was calculated from 128 key features, including gait parameters, using the extremely randomized trees (ERT) model. To evaluate the performance of this model, we calculated Cohen’s weighted kappa and intraclass correlation coefficient (ICC) between the predicted and actual CFSs. The studied patients were divided into derivation (from a university hospital) and validation (from other five sites) cohorts. Results: In the derivation (n = 181) and validation (n = 177) cohorts, the ERT model showed sufficient agreement between the actual and predicted CFSs (kappa 0.864, ICC 0.864 and kappa 0.796, ICC 0.796, respectively) ( Figure ). During a median follow-up period of 421 (IQR 280-616) days, the higher predicted CFS was independently associated with a higher risk of all-cause death (HR 1.97, 95% CI 1.21-3.20) adjusted for the MAGGIC mortality risk prediction score, N-terminal pro-brain natriuretic peptide, and serum albumin levels. Conclusions: Machine learning-based algorithms that automatically rate the CFS are feasible, and the predicted CFS is associated with the risk of all-cause death in elderly patients with HF.

Derivation of the Screening of Nutritional Risk in Intensive Care risk prediction score: A secondary analysis of a prospective cohort study.

For patients who are critically ill, the recommended nutrition risk screening tools are the Nutrition Risk in the Critically Ill (NUTRIC) and the Nutritional Risk Screening 2002 (NRS-2002) have limitations. To develop a new screening tool, the Screening of Nutritional Risk in Intensive Care (SCREENIC score), and assess its predictive validity. A secondary analysis of a prospective cohort study was conducted. Variables from several nutritional screening and assessment tools were considered. The high nutrition risk cutoff point was defined using mNUTRIC as a reference. Predictive validity was evaluated using logistic regression and Cox regression. The study included 450 patients (64 [54-71] years, 52.2% men). The SCREENIC score comprised six questions: (1) does the patient have ≥2 comorbidities (1.3 points); (2) was the patient hospitalized for ≥2 days before intensive care unit (ICU) admission (0.9 points); (3) does the patient have sepsis (1.0 point); (4) was the patient on mechanical ventilation upon ICU admission (1.2 points); (5) is the patient aged >65 years (1.2 points); and (6) does the patient exhibit signs of moderate/severe muscle mass loss according to the physical exam (0.6 points). The high nutrition risk cutoff point was set at 4.0. SCREENIC demonstrated moderate agreement (κ = 0.564) and high accuracy (0.896 [95% CI, 0.867-0.925]) with mNUTRIC. It predicted prolonged ICU (odds ratio [OR] = 1.81 [95% CI, 1.14-2.85]) and hospital stay (OR = 2.15 [95% CI, 1.37-3.38]). The SCREENIC score comprises questions with variables that do not require nutrition history. Further evaluation of its applicability, reproducibility, and validity in guiding nutrition therapy is needed using large external cohorts.

Abstract 18140: Is Socioeconomic Status Associated With a Novel Clinical Risk Prediction Score for Sudden Cardiac Death?

Introduction: Lower socioeconomic status (SES) is associated with higher risk of sudden cardiac death (SCD), but SES has not been used for SCD risk stratification. We have recently developed a novel risk score (VFRisk) for avertable SCD (8 clinical factors, 4 ECG markers, and 1 echocardiographic marker). We evaluated the association of the clinical components of the VFRisk score with SES. Hypothesis: Higher VFRisk scores will associate with lower SES. Methods: Individuals with SCD (2002-2019), were identified from a large ongoing study in a northwest US metro area (pop ~1 million). For each SCD case, a partial VFRisk score was calculated limited to the 8 clinical components (VFRisk-clinical; diabetes, HF, stroke, atrial fibrillation, COPD, history of MI, seizure, and syncope). The VFRisk-clinical score was categorized as low (<25 th percentile), moderate (25 th -75 th ), and high (>75 th percentile). SES was defined for each SCD case by residential census tract median household income and categorized into quintiles. Results: Among 4,412 SCD cases, data for SES and VFRisk-clinical was available for 3,606 (82%) with a median VFRisk of 2.8 (min 0 - max 13.4). There was a weak negative linear correlation between median household income quintile and VFRisk-clinical score (r = -0.05, p=0.002) and mean VFRisk-clinical score was slightly lower with higher SES quintile (Figure). However, there was no significant trend in the proportion of SCD cases with a high VFRisk-clinical score (>75 th percentile) by median income quintile (p=0.13). In a logistic regression model, income quintile was not significantly associated with a high VFRisk-clinical score (p=0.60). Conclusions: No clear association was observed between the VFRisk-clinical score and SES, possibly due to involvement of SES early in the causal pathway of these clinical conditions, or due to effects of SES on risk pathways for SCD separate from the components of the VFRisk score.

Abstract 14129: Identification of Highly Correlated Sub-Networks of Cardiovascular Ageing Imaging Markers With 10-Year Predicted Risk Scores

Introduction: While Framingham 10-year risk scores (FRS) estimates cardiovascular disease (CVD) risk in older adults, few studies have benchmarked these estimates against age-related changes in cardiac structure and function, which are complex and highly inter-related. We investigated associations between cardiovascular (CV) imaging markers and FRS, vascular age, calculated regression, and determined connectivity between them in a network analysis. Methods: Participants recruited from a community-based cohort without CVD completed standardized questionnaires that collected baseline clinical factors, from which FRS, vascular age and calculated regression were derived. Cardiac structural and functional parameters were obtained via echocardiography. Results: 202 participants (46.0% women, mean age 70.2 SD8.8) were included in this analysis. CV imaging markers correlated across a spectrum with FRS, vascular age and calculated regression. Predictive features and their associations by pearson correlation (nodes containing the variables; edges denoting links) (Fig 1) show significant associations. In ageing, this network showed that E/A ratio is inversely linked to FRS, vascular age and calculated regression while LVEF has a weak association with FRS and vascular age, without direct relationship with E/A ratio. While LAVI is associated with FRS and vascular age, LAVI is associated with E/E1 and not E/A ratio nor LVMI. Multivariate regression confirmed independent associations between FRS and vascular age with E/A ratio (b=-7.03, 95% CI -10.55, -3.52, p < 0.001; b= -15.13, 95% CI -19.33, -10.93, p < 0.001) and E/E prime ratio (b= 0.33, 95% CI 0.03 - 0.63, p = 0.033; b=0.66, 95% CI 0.30 - 1.02, p < 0.001). Conclusion: This structural collation of associations depicted different sub networks of cardiovascular imaging features into context. FRS, vascular age and calculated regression were associated with imaging markers of CV ageing at various network levels.

Abstract 17060: Preliminary Assessment of Artificial Intelligence Guided Echocardiographic Screening of Cardiac Amyloidosis

Background: Cardiac amyloidosis (CA) diagnosis is often overlooked. Novel CA treatments underscore the importance of early identification, as untreated CA has a poor prognosis. EchoNet-Screening is an echocardiogram-based artificial intelligence model to identify CA; previously refined by echocardiograms from patients with known CA. The goal of our study is to prospectively evaluate the accuracy of EchoNet-Screening across a large health system. Methods: All echocardiographic studies from our cardiac quaternary care center between 2019-2021 were batch-processed by EchoNet-Screening to produce a CA risk-prediction score (CA-RPS) for each patient based on the highest quality study, which is determined by the model. 336 patients with the highest CA-RPS underwent evaluation by detailed chart review by our CA clinical cardiology team. High risk patients without an alternative diagnosis are subsequently referred for CA specialty evaluation. (ClinicalTrial.gov/NCT05139797) Results: EchoNet-Screening was performed on echocardiographic images from 58,236 patients. Preliminary retrospective assessment of 336 patients with the highest CA-RPS revealed CA in 142 patients; almost half were not known from our advanced heart failure (AHF) registry and newly identified from chart review. 62 patients are pending prospective enrollment for CA subspecialty evaluation to further characterize the performance of EchoNet-Screening. Conclusions: Preliminary results from our trial demonstrate a high accuracy of EchoNet-Screening for detection of patients with known CA, emphasized by high rates of identification of CA patients even on initial chart review. Further work is ongoing to evaluate the performance of EchoNet-Screening to bring high-risk patients to new diagnoses and eventual care.

Predicting Vaso-Occlusive Crises in Sickle Cell Disease through Digital, Longitudinal Tracking of Wearable Metrics and Patient-Reported Outcomes

Background Sickle Cell Disease (SCD) is a genetic disorder caused by an HBB gene mutation and resulting structural variant haemoglobin (HbS) that polymerises upon deoxygenation, leading to a sickled morphology of red blood cells ( Chakravorty et al., Arch Dis Child 2015). This causes acute pain episodes, known as vaso-occlusive crises (VOC), that can result in severe end-organ damage in the long-term and a shortened lifespan ( Jang et al., J Transl Med 2021). Responsible for an estimated 95% of SCD hospitalisations and as a key predictor of death ( Darbari et al., Eur J Haematol 2020, Ballas et al., Am J Hematol 2005), it is critical to find new ways to drive early detection of VOCs in order to support preventative interventions. Aims This work aimed to develop a machine learning (ML) algorithm capable of predicting the potential onset of a VOC, using physiological data captured longitudinally by a wearable smartwatch, as well as patient-reported outcomes (PROs) entered via a specialised mobile phone application. Methods Following informed patient consent, participants gained access to a mobile phone application, or “digital patient wallet”, encompassing their health data and access to a PRO entry portal ( Summers et al., HemaSphere 2023). This allowed for daily recording of PROs including EQ-5D-5L, pain, mood, and fatigue scores, alongside self-reported VOCs. Participants were also provided with a Withings ScanWatch to be worn day and night, capturing physical activity, sleep quality, and heart rate data in an automated manner. Medical record data including healthcare utilisation, pathology, and demographic data was obtained through participant completion of a Subject Access Request form. Data for a snapshot cohort of 186 patients over a 1-month period (May-June 2023) were analysed through an ensemble of ML models: gradient boosting machine, neural network, and k-means clustering. All collated metrics were fed into the model, with the exclusion of hospitalisation as an 'a posteriori' variable, to ensure predictions were based on appropriate variables and not strengthened by self-confirming factors. The resulting predicted likelihood of a VOC for each participant was calculated on each day and given as a scale from 0-100%, with a likelihood of 75% or higher deemed to be a predicted VOC. This was then assessed against the true incidence of the participants' self-reported VOCs. Results The mean age was 36 (SD 11.6) years, and most patients were female (70%). The HbSS genotype was most common (75%), followed by HbSC (16%), and HbS Beta + Thalassemia (5%). 82,804 datapoints were collated for all participants across this period, encompassing 90 different variables. 69 instances of a VOC were self-reported by patients during the period of analysis, 58 (84%) of which were accurately predicted by the algorithm. 2,366 instances of no VOC were self-recorded, giving a negative predicted value of 83% (1,958/2,366 predicted). Further breakdown by the predicted likelihood score is displayed in Figure 2, highlighting the spread of individually calculated VOC likelihood scores per day against the incidence of patient-reported VOCs. Summary/Conclusions This work provides an important translation of daily trackable metrics through a remote, digital ecosystem, to the real-world prediction of potential VOC events in SCD. The purpose of this algorithm is, at its core, to act as a pre-emptive warning system from which proactive behavioural changes or interventions can be enacted. As such, the prediction of additional potential risk scores above the 75% threshold reflects an approach of ‘over-predicting’ considered to be a preferable strategy for the algorithm, given the severe nature of VOCs and the health impacts of missing a potential event. Our ongoing aim is to further develop the algorithm and its contributing metrics against medical data, in addition to patient self-reported events, to further optimise the predictor. This includes the curation of core metrics requiring minimal patient input, in order to reduce missing datapoints across those that feed into the algorithm and thereby reduce unpredicted VOCs. Ultimately, we hope to provide an early alert system that will help patients to identify their VOCs sooner, and thus action the next steps that may prevent possible hospital admissions.

Validating and Improving Adjusted Clinical Group's Future Hospitalization and High-Cost Prediction Models for Dutch Primary Care.

The rise in health care costs, caused by older and more complex patient populations, requires Population Health Management approaches including risk stratification. With risk stratification, patients are assigned individual risk scores based on medical records. These patient stratifications focus on future high costs and expensive care utilization such as hospitalization, for which different models exist. With this study, the research team validated the accuracy of risk prediction scores for future hospitalization and high health care costs, calculated by the Adjusted Clinical Group (ACG)'s risk stratification models, using Dutch primary health care data registries. In addition, they aimed to adjust the US-based predictive models for Dutch primary care. The statistical validity of the existing models was assessed. In addition, the underlying prediction models were trained on 95,262 patients' data from de Zoetermeer region and externally validated on data of 48,780 patients from Zeist, Nijkerk, and Urk. Information on age, sex, number of general practitioner visits, International Classification of Primary Care coded information on the diagnosis and Anatomical Therapeutic Chemical Classification coded information on the prescribed medications, were incorporated in the model. C-statistics were used to validate the discriminatory ability of the models. Calibrating ability was assessed by visual inspection of calibration plots. Adjustment of the hospitalization model based on Dutch data improved C-statistics from 0.69 to 0.75, whereas adjustment of the high-cost model improved C-statistics from 0.78 to 0.85, indicating good discrimination of the models. The models also showed good calibration. In conclusion, the local adjustments of the ACG prediction models show great potential for use in Dutch primary care.

Risk factor analysis of refracture in the same cemented vertebra after percutaneous kyphoplasty for Kümmell's disease.

The objective of this study was to evaluate the factors that affect refracture in the same cemented vertebra after percutaneous kyphoplasty (PKP) for Kümmell's disease (KD) and establish a risk prediction score. A total of 2932 patients who were treated with PKP for KD between January 2019 and December 2021 were retrospectively reviewed. After inclusion and exclusion criteria were applied, 191 patients were included in the study. According to the criteria for refracture, there were 50 patients in the refracture group and 141 patients in the no-refracture group. Twenty-five factors were analyzed. Patient demographics, medical history, imaging data, surgical data, and postoperative management were reviewed. Multivariate logistic regression modeling was used to identify the independent risk factors for refracture. Receiver operating characteristic (ROC) curve analysis was used to assess and establish a risk score system and further predict the risk of refracture. In this study, 50 (26.2%) patients developed a refracture. Through univariate analysis, bone mineral density (BMD) (p < 0.001), compression rate (p = 0.007), classification (i.e., the stages determined by the compression ratios) (p < 0.001), bone cement volume (p < 0.001), volume fraction (p < 0.001), distribution pattern (p = 0.007), non-PMMA endplate contact (p < 0.001), and anti-osteoporosis therapy (p < 0.001) were found to be significant factors for post-cement vertebral refracture after PKP in patients with KD. Three independent risk factors were found to be significant for refracture: small volume fraction, low BMD, and no anti-osteoporosis therapy. One point was assigned for each factor. The incidence rates of refracture in patients with scores of 0, 1, 2, and 3 were 3.7%, 4.4%, 42.0%, and 100%, respectively. The area under the ROC curve for this risk prediction score was 0.888 (p < 0.001), indicating moderate accuracy. Volume fraction, BMD, and osteoporosis therapy are the main factors influencing the refracture of the same cemented vertebra in KD. On the basis of these factors, the risk prediction score developed in this paper can be used to forecast the incidence of refracture.