HALP Score Fails to Predict Acute Heart Failure Outcomes

A retrospective cohort study published in Frontiers in Cardiovascular Medicine found that the hemoglobin–albumin–lymphocyte–platelet (HALP) score does not effectively predict in-hospital or long-term outcomes in patients with acute heart failure. The research, conducted by Mehmet Zafer Aydın and colleagues, analyzed 197 consecutive patients hospitalized with acute heart failure at a tertiary center in Türkiye. The study aimed to externally validate the HALP score, a metric that integrates nutritional, inflammatory, and hematologic status, which has previously been proposed as a prognostic marker in this clinical context. The primary finding indicates that while the score reflects disease severity, it lacks the discriminatory power necessary to serve as a stand-alone risk-stratification tool.

The study design involved calculating the HALP score at admission for each patient and assessing its ability to discriminate between outcomes using the area under the receiver-operating-characteristic curve (AUC). The pre-specified outcomes included in-hospital mortality, prolonged hospital stay exceeding seven days, all-cause mortality over the follow-up period, and heart-failure rehospitalization. The cohort had a median age of 71 years, with 30.5% of participants being female. The median follow-up duration was 15.5 months. During this period, in-hospital mortality was recorded at 12.7%, prolonged stay at 19.4%, and all-cause mortality at 31.0%. These baseline characteristics provide the context for evaluating the predictive performance of the HALP score against established clinical benchmarks.

The results demonstrated that the HALP score did not discriminate any of the primary outcomes. For in-hospital mortality, the AUC was 0.53 with a 95% confidence interval of 0.41 to 0.65. For prolonged hospital stay, the AUC was 0.47 with a 95% confidence interval of 0.36 to 0.59. For all-cause mortality, the AUC was 0.57 with a 95% confidence interval of 0.48 to 0.66. In all three instances, the confidence intervals included 0.50, which indicates that the score’s predictive ability is statistically indistinguishable from chance. This lack of discrimination suggests that the HALP score cannot reliably differentiate between patients who will experience adverse outcomes and those who will not.

Further statistical analysis revealed that an apparent association between the HALP score and mortality was likely an artifact of data-driven threshold optimization. When a Youden-optimized threshold was used, logistic regression suggested an association with mortality, yielding an adjusted odds ratio of 2.02 with a 95% confidence interval of 1.02 to 4.00. However, this signal disappeared when pre-specified thresholds were applied. Using a median-split threshold, the adjusted odds ratio dropped to 1.23 with a 95% confidence interval of 0.67 to 2.28. Time-to-event analysis using Cox regression also failed to show a significant association, with an adjusted hazard ratio per one-standard-deviation decrease of 1.04 (95% CI 0.78–1.38) and a log-rank p-value of 0.15 across quartiles. This discrepancy highlights the risk of cutoff-optimization bias in prognostic marker validation.

In contrast to the HALP score, established risk scores demonstrated significant predictive value in the same patient cohort. The EHMRG score achieved an AUC of 0.71, and the MAGGIC score achieved an AUC of 0.65, with both confidence intervals excluding 0.50. This comparison underscores that while the HALP score may correlate with disease severity, it does not offer the same level of prognostic accuracy as validated clinical tools.

The HALP score did show correlations with markers of disease severity, although these associations were modest. The study reported a negative correlation between the HALP score and NYHA class (rho = −0.31), as well as with N-terminal pro-B-type natriuretic peptide (NT-proBNP, rho = −0.16) and C-reactive protein (CRP, rho = −0.17). These findings suggest that lower HALP scores are associated with more severe heart failure and higher levels of inflammatory and cardiac stress markers. However, the study concludes that reflecting disease severity is not equivalent to predicting specific clinical outcomes such as mortality or rehospitalization. The distinction is critical for clinical application, as a marker that tracks severity may not provide the forward-looking predictive information needed for risk stratification.

The authors conclude that in this independent cohort, the HALP score should not be used as a stand-alone risk-stratification tool in acute heart failure. The study’s limitations include its retrospective design and the single-center nature of the cohort, which may limit the generalizability of the findings to other populations. The study contributes to the growing body of literature evaluating novel prognostic markers in heart failure, emphasizing the need for rigorous external validation and the avoidance of data-driven threshold biases.

Subject of Research: External validation of the HALP score for predicting in-hospital and long-term outcomes in acute heart failure: a retrospective cohort study

Article Title: External validation of the HALP score for predicting in-hospital and long-term outcomes in acute heart failure: a retrospective cohort study

Article References: Aydın, M. Z., Gök, M., Teber, M. K., Kuzu, Z., & Şimşek, Z. (2026). External validation of the HALP score for predicting in-hospital and long-term outcomes in acute heart failure: a retrospective cohort study. Frontiers in Cardiovascular Medicine, 13, Article 1958343. https://doi.org/10.3389/fcvm.2026.1958343

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1958343

Keywords: External, validation, HALP, score, predicting, in-hospital, long-term, outcomes, acute, heart, failure, retrospective