Performance of the MAGGIC heart failure risk score and its modification with the addition of discharge natriuretic peptides

Abstract

AimsPredictive models for heart failure patients are widely used in the clinical practice to stratify patients' mortality and enable clinicians to tailor and intensify their approach. However, such models have not been validated internationally. In addition, biomarkers are now frequently measured to obtain prognostic information, and the implications of this practice are not known. In this study, we aimed to validate the model performance of the Meta‐analysis Global Group in Chronic Heart Failure (MAGGIC) score in a Japanese acute heart failure registry and further explore the incremental prognostic value of discharge B‐type natriuretic peptide (BNP) level.Methods and ResultsIn this study, we evaluated the registered data of 2215 consecutive acute HF patients (with 694 119 person‐years follow‐up) from a prospective multicentre registry (the West Tokyo Heart Failure) conducted in Japan from April 2006 to August 2016. The mean age was 73.0 ± 13.0, and 61.2% were male. The MAGGIC score demonstrated modest discrimination (c‐index = 0.71, 95% confidence interval 0.67–0.74) and good calibration (R 2 value = 0.97); there was constant overestimation for 1 year mortality. However, when the BNP level was added to the original MAGGIC variables, the model demonstrated good discrimination (c‐index = 0.74, 95% confidence interval 0.70–0.78) with adequate calibration (R 2 value = 0.91). The modified MAGGIC BNP score was externally validated in a separate Japanese registry (NaDEF) and demonstrated moderate discrimination (c‐index = 0.69, 95% confidence interval 0.65–0.73) and calibration (R 2 value = 0.85).ConclusionThe original MAGGIC score performed modestly in Japanese patients, but the addition of discharge BNP level enhanced model performance. The addition of objective biomarkers may result in effective modification of preexisting internationally recognized risk models and aid in multinational comparisons of heart failure patients' outcomes.

Improving risk prediction in heart failure using machine learning

“…Numerous risk scores have been developed to predict mortality in patients with HF . These tools generally do not have the discriminatory power of MARKER‐HF.…”

“…Numerous risk scores have been developed to predict mortality in patients with HF. [3][4][5][6][7][8][9][10][11][12][13] performance of previous risk prediction models has often been ascribed to the heterogeneity of patients with HF. However, their modest performance is more likely due to the failure of the statistical methods used in their design to capture prognostically important multi-dimensional correlations between the state of the patient and their outcome.…”

“…The ability to accurately assess outcomes in patients with HF, however, has proven to be a difficult task. Although a number of tools, including biomarkers, risk scores and their combination have been developed for this purpose, most have achieved only modest success, particularly when they are employed in HF populations other than those from which the score was derived . For instance, the Meta‐Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score achieved a C‐statistic for the area under the curve (AUC) of 0.74 for predicting mortality risk in a large cohort of patients followed in the Swedish Heart Failure Registry, but performed less well in patients on a transplant waiting list and in a large population of ambulatory HF patients in the US where the C‐statistics were 0.69 and 0.70, respectively .…”

“…The ability to accurately assess outcomes in patients with HF, however, has proven to be a difficult task. Although a number of tools, including biomarkers, 1,2 risk scores [3][4][5][6][7][8][9][10] and their combination [11][12][13] have been developed for this purpose, most waiting list and in a large population of ambulatory HF patients in the US where the C-statistics were 0.69 and 0.70, respectively. 16,18 The results with previous HF risk scores are likely due to several causes including dependence on variables that are not universally available and temporal dispersion in the collection of variables so that the state of the patient at a discrete point in the course of their disease is not captured.…”

Clinical factors related to morbidity and mortality in high‐risk heart failure patients: the GUIDE‐IT predictive model and risk score

“…One of the most important predictors that was present in all three models was the log of NT‐proBNP. This is an important variable that not only is associated with prediction of all‐cause death, the composite of CV death and HF hospitalization, but HF hospitalization alone . While it has been well known that NT‐proBNP has important prognostic information, the confirmation in this high‐risk population independent of ethnicity and race, many co‐morbidities and the congestion score, provides additional important information to the clinician .…”