Study: Predictive Mortality Tools Limited in Elderly Advanced CKD Population

In the chronic kidney disease (CKD) space, existing prognostic tools demonstrate below-standard quality performance in a validation cohort of older patients, said study investigators in Internal Medicine Journal. They emphasized that further research is needed to develop a specific tool that is more applicable to the older population with advanced CKD.¹

Image credit: onephoto | stock.adobe.com

Kidney disease occurs when a patient’s kidneys are not functioning properly and cannot filter waste, leading to buildup in the blood. CKD worsens over time and is often caused by high blood pressure and diabetes. Because there is currently no cure for CKD, treatments are dedicated to preserving kidney function as long as possible. In late-stage disease, dialysis or transplant may be required. Approximately 15% of US adults have CKD, according to Cleveland Clinic.²

Common symptoms include the following: frequent urination; loss of appetite; tiredness, weakness, and/or low energy level; puffy eyes; dry and itchy skin; difficulty sleeping and/or concentrating; nausea or vomiting; and swelling of the hands, feet, and ankles. The disease is often caused by hypertension and diabetes but can also be caused or affected by glomerulonephritis, lupus, polycystic kidney disease, membranous nephropathy, or obstructions from the kidney tract (eg, kidney stones, cancer, or enlarged prostate).²

Prior research has demonstrated that patients with advanced CKD want to know their prognosis, and although many predictive tools exist, they are not widely used in clinical practice because their applicability is unclear. This is particularly prevalent for older patients. Investigators conducted a single-center retrospective validation cohort study to determine the applicability and utility of current models that predict mortality in a cohort of older Australian patients with advanced CKD.¹

For this study, 387 patients ages 65 years or older (median age: 80 years; IQR 74 to 85 years) with stage 4 and 5 CKD between March 2009 and August 2018 were enrolled. Patients were considered ineligible if they had either undergone dialysis for acute kidney injury or received a kidney transplantation. The researchers tested the Ivory, Schmidt, and Cohen models in this population.¹

The Ivory model is a 6-month mortality prognostic tool with a point-score system including certain variables (eg, age, sex, race, body mass index, smoking status, disease status, comorbidities, late referral to nephrologist, and primary renal disease) for patients undergoing dialysis. Points are assigned to each category, allowing for prognostic variables to have differing weighted effects on mortality risk depending on the logistic regression analysis of registry data. The 6-month mortality risk is calculated by the addition of the total accumulated points.¹

Alternatively, the Schmidt model is a 12-month mortality prediction tool for patients with stage 4 and 5 CKD who are not yet on dialysis and includes 3 variables (age, “surprise question,” and Karnofsky performance scale index, which is a measure of function). The output of this tool provides a probability of 12-month mortality in percentage form.¹

Lastly, the Cohen 6-month mortality prognostic tool includes 5 variables (“surprise question,” albumin, age, and the presence of dementia and peripheral vascular disease). The output of this tool provides an estimated risk of mortality in 6 months. The authors noted that, for the purposes of the analysis, a mortality risk of 50% was used as a point of delineation for those with high or ‘lower’ risk. Of note, among the 387 patients included in the study, 156 were not on dialysis.¹

For the Ivory model, 231 patients were evaluated. The c-statistic was about 0.617 (95% CI 0.47–0.74), and the Hosmer-Lemeshow test statistic was 0.96 (P = .48) and was considered significant for some values, but poor calibration was indicated. Additionally, the range of predicted mortality probability was about 2% to 40% (median: 7%). Notably, 4 of 13 patients (31%) in the predicted high mortality risk group died within 6 months, as opposed to 19 of 218 (9%) who were not predicted to have a high mortality rate.¹

The Schmidt model was utilized on 156 patients with stage 4 or 5 CKD. The c-statistic was about 0.57 (95% CI 0.43–0.70), and Hosmer–Lemeshow test statistic was 9.81 (P < .001), indicating poor discrimination and calibration. Eighty-two patients (25%) had died within 12 months, noted the authors.¹

A total of 183 patients were evaluated using the Cohen model, of which 21 (11%) had died within 6 months. Notably, the patients who had died during this time were all missing a response to the “surprise question.” Further, 118 of 130 patients (91%) who were alive had a predicted 6-month mortality score of less than 50%, which is defined as a below 50% likelihood of dying in the next 6 months. Only 12 (9%) had a predicted 6-month mortality risk over 50%.¹

Based on these findings, the authors concluded that the clinical utility of existing predictive tools to predict mortality in elderly patients with advanced CKD is limited by poor performance and accuracy. Future studies should prioritize the development of a prognostic tool that is specifically tailored to the elderly advanced CKD population.¹

“[Further research] would enhance the shared decision-making process when deciding the appropriateness of dialysis and would also help to identify individuals who would benefit from early palliative care input,” wrote the authors.¹

REFERENCES

1. Chou A, Farshid S, Brown, MA, Li KC. Can we predict mortality of older patients with advanced chronic kidney disease?. Intern Med J. 2025. doi:10.1111/imj.70089

2. Cleveland Clinic. Chronic kidney disease. Accessed May 20, 2025. https://my.clevelandclinic.org/health/diseases/15096-chronic-kidney-disease