{
  "abstract": "Introduction The Global Diastology Index (GDi) is a novel, machine learning-derived echocardiographic score, validated against invasive pressure measurements, which summarises diastolic function on a continuous scale from normal physiology to advanced dysfunction. We aimed to calibrate a normal threshold for GDi in large-scale, real-world data, and confirm clinical utility based on prognostic significance for mortality and heart failure (HF) admissions.Methods Routinely collected anonymised echocardiogram data were analysed from the EchoVision study (19/HRA/2068), which comprises structured quantitative measurements and free-text comments from scans performed between January 2017 and July 2020. Linked outcome data through to October 2020 were available from hospital electronic records, including all-cause mortality and hospital admissions based on ICD-10 codes.A phenotypically normal reference group was defined using guideline criteria, including normal left ventricular size, structure and systolic function, normal left atrial size, absence of significant valvular disease, and normal diastolic parameters. Using this reference group, a normal GDi threshold was derived by receiver operating characteristic (ROC) analysis, maximising Youden’s J statistic. Prognostic validation was based on Kaplan–Meier and Cox proportional hazards models.Results 101,522 echocardiograms, from 64,630 patients, were categorised as either normal (7206; 7.1%) or abnormal (94,266; 92.9%). The optimal GDi threshold separating these groups was 0.23 (95% CI 0.223–0.232). ROC analysis demonstrated excellent discrimination between groups (AUC 0.89; sensitivity 0.74; specificity 0.93) at the selected threshold.10,854 patients died, with median time to death of 564 days (IQR 189–1173 days) after echocardiography. GDi >0.23 was associated with significantly higher all-cause mortality (HR 2.88, 95% CI 2.76–3.02; p<0.001).4,169 studies (from 2,675 patients) had a subsequent HF admission (4.1%). GDi >0.23 was associated with first HF admission (HR 6.76, 95% CI 6.01–7.60; p<0.001) (figure 1). In secondary analyses, a pragmatic GDi threshold of 0.20 was tested, to simplify clinical interpretation, which provided high specificity (0.97) with moderate sensitivity (0.67). Stratification into four GDi risk categories (<0.20, 0.20–0.5, 0.5–0.75, 0.75–1.0) demonstrated a graded increase in HF admission risk, with hazard ratios of 5.22, 12.58 and 19.37 respectively compared with <0.20 (figure 2).Conclusions In a large real-world echocardiography dataset, a clinically meaningful threshold for normal GDi can be identified with excellent diagnostic performance. Furthermore, the GDi threshold effectively stratifies long-term risk of mortality and future HF admission, with a graded association between increasing GDi and risk of future events. GDi offers the possibility of a simple, automated, scalable metric for diastolic function, with prognostic value.Kaplan–Meier curves of heart-failure admission–free survival following echocardiography stratified by Global Diastology Index (GDi). Patients with GDi >0.23 (orange) had a significantly higher risk of heart-failure admission compared with those with GDi ≤0.23 (blue) (log-rank p<0.001). Deaths were treated as censoring events. Shaded areas represent 95% confidence intervals, and numbers at risk are shown below the x-axis.Kaplan–Meier curves of heart-failure admission–free survival following echocardiography stratified by Global Diastology Index (GDi). Patients were stratified into four GDi risk categories (<0.20, 0.20–0.5, 0.5–0.75, and 0.75–1.0). Deaths were treated as censoring events. Shaded areas represent 95% confidence intervals, and numbers at risk are shown below the x-axis.Abstract 351 Figure 1Abstract 351 Figure 2",
  "authors": [
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Samuel Krasner"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom",
        "Department of Cardiac Physiology, Royal Papworth Hospital NHS Foundation Trust, Cambridge, United Kingdom"
      ],
      "name": "Andrew Fletcher"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom",
        "Healthcare Engineering Innovation Group, Department of Biomedical Engineering, Khalifa University of Science and Technology, Abu Dhabi, United Arab Emirates"
      ],
      "name": "Mohanad Alkhodari"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Casey L Johnson"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Natalie Savage"
    },
    {
      "affiliations": [
        "Institute for Medical Engineering and Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States"
      ],
      "name": "Maryam Alsharqi"
    },
    {
      "affiliations": [
        "Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Adam J Lewandowski"
    },
    {
      "affiliations": [
        "Oxford Centre for Clinical Magnetic Resonance Research, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Stefan Neubauer"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Winok Lapidaire"
    },
    {
      "affiliations": [
        "Cardiovascular Clinical Research Facility, RDM Division of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom"
      ],
      "name": "Paul Leeson"
    }
  ],
  "title": "351 The global diastology index: calibrating normal thresholds and association with mortality and heart failure admission in a large real-world echocardiography dataset",
  "uid": "1ff30cae-3d52-5ac1-99b1-efa35baa7865"
}