Nomogram Predicts Short-Term Palliative Survival Using CRP, eGFR, and Functional Status

Refining Survival Estimates in Palliative Care

Accurately estimating the remaining lifespan of patients in palliative care is a fundamental component of high-quality end-of-life management, yet it remains one of the most challenging tasks for practicing clinicians. While prognostic accuracy is highly valued by patients and their families, systematic reviews indicate that physicians and nurses frequently struggle to provide reliable survival timelines [1]. Even widely used screening tools, such as the "Surprise Question," demonstrate significant variability in accuracy across different clinical settings [2]. Identifying appropriate candidates for palliative interventions often relies on a complex assessment of disease trajectory, physical symptoms, and functional performance [3]. As the population of patients with advanced, life-limiting illnesses continues to grow, the need for objective tools to guide these difficult clinical conversations has become increasingly urgent [4]. To address this gap, researchers recently investigated whether combining specific objective biomarkers with functional assessments could improve the precision of short-term mortality predictions.

Identifying Objective Predictors of Mortality

The researchers aimed to develop a simple, interpretable nomogram (a graphical calculating device used to estimate individual clinical risk) by integrating objective biomarkers and functional status for end-stage patients. To construct this tool, the investigators conducted a retrospective review of 167 patients who had recently engaged with a palliative care team. This cohort was randomly partitioned into a 7:3 ratio, yielding a training set of 117 patients and an internal validation set of 50 patients. The team initially evaluated a pool of 31 variables for their prognostic significance in predicting short-term mortality. To refine the model, independent prognostic factors were identified using least absolute shrinkage and selection operator (LASSO) Cox regression, a statistical method that isolates the most important predictors while minimizing model complexity. For continuous measures, the researchers determined optimal thresholds via maximally selected rank statistics, a technique used to find the most clinically significant cut-off point for a variable. This rigorous selection process identified three core components for the final predictive model. C-reactive protein (CRP) emerged as an independent risk factor for mortality (p < 0.001), alongside estimated glomerular filtration rate (eGFR) (p = 0.001) and Barthel Activities of Daily Living (BADL) scores (p = 0.01). By combining systemic inflammation, renal function, and a standardized measure of physical independence, the resulting tool offers clinicians a multifaceted snapshot of a patient's physiological decline.

Evaluating Model Discrimination and Calibration

The researchers evaluated the predictive accuracy of the nomogram using the area under the curve (AUC), a metric where a value of 1.0 represents perfect discrimination and 0.5 indicates performance no better than chance. In the training set of 117 patients, the model demonstrated a 30-day AUC of 0.764, a 60-day AUC of 0.716, and a 90-day AUC of 0.705. These results remained consistent or improved when applied to the internal validation set of 50 patients, which yielded a 30-day AUC of 0.770, a 60-day AUC of 0.748, and a 90-day AUC of 0.788. These figures suggest that the combination of CRP, eGFR, and BADL scores provides a stable and reliable basis for distinguishing between patients with different survival probabilities across a three-month follow-up period. Beyond its ability to discriminate between survivors and non-survivors, the model was assessed for its calibration, a measure of how closely the predicted probabilities of death match the actual observed outcomes. Calibration plots showed excellent agreement between predicted and observed survival, indicating that the nomogram does not systematically overestimate or underestimate mortality risk. Furthermore, the researchers utilized decision curve analysis, a method used to determine the practical utility of a model by weighing the benefits of true-positive predictions against the harms of false-positive results. This decision curve analysis indicated a clinical net benefit across commonly used risk thresholds. For practicing physicians, this means that using the nomogram to guide palliative care interventions yields better clinical outcomes than either treating all patients uniformly or treating none based on assumed risk levels.

Clinical Utility and Diagnostic Accuracy

The diagnostic performance of the nomogram was further characterized in the validation set of 50 patients, where it demonstrated high accuracy in identifying individuals at risk of short-term mortality. At the 30-day mark, the tool showed a sensitivity of 69.05% and a specificity of 75.00%. The predictive power of the model increased as the time horizon extended. At 60 days, the nomogram reached a sensitivity of 83.33% and a specificity of 87.50%. By the 90-day interval, the sensitivity improved to 88.10% while maintaining a specificity of 87.50%. These metrics suggest that the tool is particularly robust at identifying patients unlikely to survive the three-month period, providing clinicians with an objective framework to supplement their bedside intuition. To assist physicians in the practical application of these findings, the researchers calculated likelihood ratios, which represent the change in the odds of an outcome based on a test result. The positive likelihood ratio (LR+), a measure of how much a positive result increases the probability of mortality, was 2.762 at 30 days, 6.666 at 60 days, and 7.048 at 90 days. Conversely, the negative likelihood ratio (LR-), which indicates how much a negative result decreases the probability of mortality, was 0.413 at 30 days, 0.191 at 60 days, and 0.136 at 90 days. In clinical terms, an LR+ above 7.0 at 90 days suggests a strong ability to rule in short-term mortality, while an LR- as low as 0.136 provides significant evidence to rule it out, allowing for more confident prognostic assessments. For the practicing clinician, these data points offer a pathway toward more precise resource optimization and end-of-life decision-making. By integrating objective laboratory values with functional assessments, the nomogram provides a standardized method for identifying patients who may require immediate palliative interventions, hospice transitions, or more intensive family counseling. This objective approach reduces the reliance on subjective clinical impressions, ensuring that end-of-stage care is tailored to the actual physiological status and survival probability of the individual patient.

References

1. White N, Reid F, Harris AJL, Harries P, Stone P. A Systematic Review of Predictions of Survival in Palliative Care: How Accurate Are Clinicians and Who Are the Experts?. PLoS ONE. 2016. doi:10.1371/journal.pone.0161407

2. White N, Kupeli N, Vickerstaff V, Stone P. How accurate is the ‘Surprise Question’ at identifying patients at the end of life? A systematic review and meta-analysis. BMC Medicine. 2017. doi:10.1186/s12916-017-0907-4

3. Hui D, Meng Y, Bruera S, et al. Referral Criteria for Outpatient Palliative Cancer Care: A Systematic Review. The Oncologist. 2016. doi:10.1634/theoncologist.2016-0006

4. Siegel RL, DeSantis C, Virgo KS, et al. Cancer treatment and survivorship statistics, 2012. CA A Cancer Journal for Clinicians. 2012. doi:10.3322/caac.21149