Prognostic factors for mortality in age-stratified elderly sepsis patients: A retrospective cohort study based on the MIMIC-IV database
Clinical Snapshot
PICO Framework
| P — Population | Elderly sepsis patients aged ≥60 years from MIMIC-IV database, stratified into younger-old (60-74 years) and older-old (75-90 years) groups |
| I — Intervention | Age-stratified prognostic assessment using clinical and biochemical markers |
| C — Comparator | Younger-old versus older-old age groups |
| O — Outcomes | 28-day mortality, prognostic accuracy of biomarkers including LDH, PaO2/FiO2 ratio, urea nitrogen, albumin, and urine output |
Bottom Line
This large retrospective cohort study of 13,461 elderly sepsis patients identifies age-specific prognostic factors that could enhance clinical decision-making. In patients aged 60-74 years, lactate dehydrogenase and PaO2/FiO2 ratio were key mortality predictors, while urea nitrogen and hypoalbuminemia predominated in those aged 75-90 years. Oliguria emerged as a common predictor across age groups, conferring a 2.5-fold mortality increase in older patients. The findings suggest that age-stratified prognostic approaches may be superior to one-size-fits-all models in elderly sepsis management. However, the study's retrospective design and lack of external validation limit immediate clinical implementation. Australian intensivists should consider these age-specific markers alongside established sepsis scores when assessing prognosis in elderly patients, while awaiting prospective validation studies.
Key Findings
P Value: Not specified in abstract
Effect Size: Oliguria associated with 2.5-fold increase in mortality risk in older patients
Primary Outcome: 28-day mortality in age-stratified elderly sepsis patients
Nnt Or Sensitivity: Age-specific prognostic factors: LDH and PaO2/FiO2 ratio in younger-old; urea nitrogen and hypoalbuminemia in older-old groups
Confidence Interval: Not specified in abstract
Clinical Application
Highly feasible as uses readily available clinical and biochemical markers in ICU setting Directly applicable to Australian ICUs; aligns with ANZICS sepsis guidelines and could inform risk stratification in elderly patients under Medicare-funded intensive care Elderly sepsis patients aged 60+ years in intensive care settings
Abstract
Sepsis represents a major cause of death among elderly patients; however, prognostic factors specific to different age strata within this population remain insufficiently characterized. This retrospective cohort study aimed to identify early predictors of mortality in elderly sepsis patients using the MIMIC-IV database. We included 13,461 sepsis patients aged ≥60 years, stratified into younger-old (60-74 years; n = 7618) and older-old (75-90 years; n = 5843) groups. Propensity score matching was applied to balance baseline characteristics, yielding 4657 matched patients per group. Multivariable logistic regression, receiver operating characteristic analysis, and Kaplan-Meier survival curves were employed to assess clinical and biochemical markers for 28-day mortality and to identify age-specific prognostic indicators. These findings underscore the importance of age-stratified prognostic models to guide targeted clinical interventions and potentially improve outcomes in elderly patients with sepsis. Lactate dehydrogenase (LDH) and the PaO2/FiO2 ratio emerged as key mortality predictors in the younger-old group, whereas urea nitrogen and hypoalbuminemia were predominant in the older-old group. Urine output served as a common predictor across both strata, with oliguria associated with a 2.5-fold increase in mortality risk among older patients. While a combination of predictors enhanced prognostic accuracy in younger-old patients, this was not observed in the older-old group, suggesting a greater influence of frailty and comorbidities in advanced age.
References
- 1.Chen, J., Lin, J., Wang, H., Gu, C., Lu, X., & Sun, C. (2026). Prognostic factors for mortality in age-stratified elderly sepsis patients: A retrospective cohort study based on the MIMIC-IV database. Medicine, Article 0000000000048209. https://doi.org/10.1097/MD.0000000000048209
Related Research
Journal of the American Medical Informatics Association : JAMIA
Explainable machine learning in healthcare: methods, interpretation, and applications for clinical research
2 Aug 2026
Neurosurgical review
Development and external validation of a machine learning model for predicting postoperative hydrocephalus in 1,073 posterior fossa tumor patients
31 July 2026
JAMA
Prone Positioning in Infants With Acute Bronchiolitis: The PROPOSITIS Randomized Clinical Trial
30 July 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service