Development and Internal Validation of a Nomogram for Predicting 28-Day Mortality in Elderly Hip Fracture Patients Admitted to the Intensive Care Unit: A Retrospective Cohort Study Based on MIMIC-IV

Technical Summary: Development and Internal Validation of a Nomogram for Predicting 28-Day Mortality in Elderly Hip Fracture Patients Admitted to the ICU

Problem Statement
Elderly patients with hip fractures requiring Intensive Care Unit (ICU) admission represent a clinically vulnerable population with high short-term mortality. While serum lactate is a recognized marker for tissue hypoperfusion and shock, its independent prognostic value in this specific orthopedic-ICU cohort remains to be fully characterized alongside other metabolic factors. Furthermore, the relationship between pre-existing diabetes and mortality in critically ill orthopedic patients is complex; while traditionally viewed as a risk factor, some literature suggests a "diabetes paradox" where diabetes may be associated with lower short-term mortality after adjustment. There is a need for a transparent, interpretable prediction tool that integrates these variables to facilitate early risk stratification and individualized perioperative care.

Methodology
This study employed a retrospective cohort design utilizing the Medical Information Mart for Intensive Care IV (MIMIC-IV) database.

• Study Population: The cohort included 337 patients aged 65 years or older admitted to the ICU with a primary or secondary diagnosis of hip fracture (femoral neck, intertrochanteric, or subtrochanteric).
• Inclusion/Exclusion: Patients were included if initial serum lactate and renal function markers (BUN, creatinine) were available within the first 24 hours of ICU admission.
• Variables: Baseline data included demographics, initial laboratory values (lactate, BUN, creatinine), and comorbidities (hypertension, diabetes, heart failure). Lactate was categorized as normal (≤2.0 mmol/L) or high (>2.0 mmol/L).
• Outcome: The primary outcome was 28-day all-cause mortality.
• Statistical Analysis:
  • Univariable and multivariable Cox proportional hazards regression were used to identify independent prognostic factors.
  • A nomogram was constructed based on the final multivariable model to visualize individualized mortality risk.
  • Validation: Internal validation was performed using 1,000 bootstrap resamples to correct for optimism.
  • Performance Metrics: Model discrimination was assessed via the C-index; calibration was evaluated using calibration curves; and clinical utility was quantified using Decision Curve Analysis (DCA).

Key Results

• Cohort Characteristics: The overall 28-day mortality rate was 33.2% (112/337). Patients with high initial lactate (>2.0 mmol/L) were significantly older, had worse renal function (higher BUN and creatinine), and a higher prevalence of diabetes compared to the normal lactate group.
• Independent Predictors: Multivariable Cox regression identified four independent predictors of 28-day mortality:
  1. Initial Hyperlactatemia: High lactate (>2.0 mmol/L) was strongly associated with increased mortality (HR = 2.28, 95% CI: 1.56–3.32; P < 0.001).
  2. Elevated BUN: Higher blood urea nitrogen levels were independently associated with increased mortality (HR = 1.02 per mg/dL; P < 0.001).
  3. Age: Older age was a significant risk factor.
  4. Diabetes Paradox: Contrary to traditional expectations, pre-existing diabetes was independently associated with a reduced risk of 28-day mortality (HR = 0.48, 95% CI: 0.30–0.77; P = 0.002).
• Model Performance: The resulting nomogram demonstrated acceptable discrimination with a C-index of 0.700. Calibration curves indicated satisfactory agreement between predicted and observed mortality. Decision Curve Analysis showed that the nomogram provided a greater net clinical benefit than "treat-all" or "treat-none" strategies across clinically relevant threshold probabilities.

Significance and Claims
The authors claim that this study successfully develops and internally validates a clinically interpretable nomogram for predicting 28-day mortality in elderly hip fracture patients admitted to the ICU. The significance of the work lies in three main areas:

1. Prognostic Utility: The model confirms that initial hyperlactatemia and impaired renal function (elevated BUN) are critical predictors of short-term mortality in this population, supporting the use of these readily available markers for early risk assessment.
2. Observation of the Diabetes Paradox: The study highlights the presence of a "diabetes paradox" in this specific cohort, where pre-existing diabetes was associated with a survival advantage. The authors frame this as a hypothesis-generating finding, suggesting it may reflect context-specific metabolic adaptations or medication effects (e.g., metformin use) rather than a protective effect of the disease itself.
3. Clinical Application: The proposed nomogram integrates age, lactate group, BUN, and diabetes status into a single visual tool. The authors assert that this tool can support early risk stratification and individualized perioperative critical care decision-making.

Limitations and Future Directions
The authors explicitly note that the study is limited by its retrospective design, reliance on a single-center database (MIMIC-IV), and the lack of external validation. They acknowledge that granular variables such as pre-injury frailty, exact injury-to-admission time, and specific pre-admission medication usage (e.g., HbA1c, hypoglycemic agents) were not fully captured. Consequently, the authors state that the nomogram requires external validation in independent multicenter cohorts before it can be widely adopted, and the mechanisms behind the observed diabetes paradox require prospective investigation.