AECOPD Mortality Model Uses Hemoglobin and Albumin
Key Takeaways
- In a single-hospital cohort in Wuhu, China, more than 2,000 patients hospitalized with AECOPD were studied, and 5% died during the admission.
- Older age, lower hemoglobin, lower albumin, respiratory failure at admission, and a history of lung cancer were independently associated with higher in-hospital mortality.
- The five-variable model showed good discrimination and favorable calibration for short-term risk estimation in hospitalized AECOPD.
- Decision-curve findings suggested clinical net benefit across practical risk thresholds, but external multicenter validation is still needed.
In the AECOPD in-hospital mortality prediction model, investigators conducted a retrospective cohort study at the Second People’s Hospital of Wuhu from January 1, 2016, through October 21, 2025, among patients aged 40 years or older hospitalized with AECOPD. Major exclusions were severe coexisting pulmonary disease, active malignancy, and death or discharge within 24 hours. Prespecified candidate variables were age, sex, respiratory failure at admission, history of lung cancer, hemoglobin, albumin, blood urea nitrogen, creatinine, and interleukin-6 (IL-6), with baseline data drawn from admission electronic medical records. Variables with 5% or less missingness underwent complete-case analysis, while those with 5% to 30% missingness were handled with MissForest-based multiple imputation; multivariable logistic regression, receiver operating characteristic (ROC) analysis, bootstrap calibration, the Hosmer-Lemeshow test, and decision curve analysis (DCA) were used for internal validation.
In-hospital mortality was 5%, with 109 deaths among 2167 patients. In the final multivariable model, respiratory failure at admission was strongly associated with death (OR 6.002, 95% CI 3.936–9.234, P<0.001), and a history of lung cancer was also associated with higher mortality (OR 4.43, 95% CI 1.815–9.855, P<0.001). Older age, lower hemoglobin, and lower albumin were independently associated in the same model, while interleukin-6, blood urea nitrogen, and creatinine did not remain significant after adjustment.
Model performance in hospitalized AECOPD showed an AUC of 0.811. Calibration was supported by a Hosmer-Lemeshow χ2 of 11.085 with P=0.197 and a bootstrap mean absolute error of 0.005. Decision curve analysis showed greater net benefit than treat-all or treat-none approaches across threshold probabilities of about 5% to 50%, and the nomogram used the same five predictors, with respiratory failure contributing the largest share of total points.
This China-based derivation cohort came from a single hospital and was validated internally only, so broader generalizability remains uncertain until external multicenter testing is completed. Residual confounding could not be excluded because variable availability depended on medical-record completeness, and radiologic findings, pulmonary-function measures such as Global Initiative for Chronic Obstructive Lung Disease stage and forced expiratory volume in 1 second (FEV1), and some acute physiological and inflammatory measures were not included because of substantial missing data or lack of independent predictive value during variable selection. Nutritional status was assessed using two indirect, nutrition-related laboratory indicators—hemoglobin and albumin—rather than more comprehensive measures such as body mass index or muscle mass, and the correlational findings do not show that correcting low hemoglobin, low albumin, or malnutrition alone would reduce mortality.
Investigators developed and internally validated a five-variable in-hospital mortality model for hospitalized AECOPD that combined two routinely available nutrition-related laboratory markers with three clinical characteristics. The authors said external validation in independent multicenter populations is still needed before broader implementation.
Clinician Questions
Which hospitalized AECOPD patients were included in this mortality model, and which patients were excluded?
The model included adults aged 40 years or older who were hospitalized with a primary diagnosis of AECOPD at the Second People’s Hospital of Wuhu. Exclusions were severe coexisting pulmonary disease, including conditions such as tuberculosis or interstitial lung disease, active malignancy, and death or discharge within 24 hours, so applicability is bounded to this single-center hospitalized cohort in China.
How was respiratory failure defined in the AECOPD mortality model?
In hospitalized AECOPD, respiratory failure was defined as respiratory failure present at admission based on the need for noninvasive or invasive ventilatory support and documentation in the admission medical record.
Which candidate variables were considered before the final AECOPD mortality model was reduced to five predictors?
Before reduction to the final five-predictor AECOPD model, investigators evaluated age, sex, respiratory failure at admission, history of lung cancer, hemoglobin, albumin, blood urea nitrogen, creatinine, and IL-6.
What remains uncertain before this AECOPD mortality model can be used more broadly?
Whether the model will perform similarly outside the original derivation setting remains uncertain because it has only internal validation from a single-center retrospective cohort. The authors also noted that residual confounding cannot be excluded and that radiologic, pulmonary-function, and some acute physiologic or inflammatory measures were not consistently available.