CKD Risk Stratification, Payment Models, LLMs Aim to Close Care Gaps

Chronic kidney disease (CKD) risk stratification that flagged high-risk patients in the electronic medical record, paired with quarterly prescribing feedback to clinicians, was associated with greater uptake of guideline-directed medical therapy (GDMT) and albuminuria testing in a large US nephrology practice, according to a poster presented at American Society of Nephrology Kidney Week 2026 in Denver, Colorado.¹ The largest gains in several drug classes occurred among patients the Klinrisk model classified as high risk.

The study was among the posters presented in a Kidney Week session on clinical prediction, decision support, and digital kidney care, where other researchers reported on a Medicare value-based payment model and large language models (LLMs) used to draft prior authorization (PA) letters.^2,3 Each examined a different approach to identifying or managing patients with CKD, from risk prediction and payment incentives to administrative automation.

Can Risk Prediction Improve Low GDMT Uptake in Chronic Kidney Disease?

CKD increases the risk of cardiovascular events, kidney failure, and death and raises costs, but uptake of GDMT for patients with CKD and diabetes remains low, the authors wrote.¹ The quality improvement initiative used the Klinrisk model, which predicts CKD progression from routine laboratory data and has been validated in more than 4 million patients, according to the authors. It included all patients with available laboratory data in a practice with more than 100 clinicians and more than 50,000 patients. The study was funded by Klinrisk and AstraZeneca.

Patients with a 5-year risk of progressive CKD—a 40% decline in estimated glomerular filtration rate or kidney failure—of at least 25% were flagged as high risk, and clinicians received quarterly reports benchmarking their urine albumin-creatinine ratio (UACR) testing and GDMT use against practice-wide rates. Researchers compared a baseline period spanning November 2023 to October 2024 with a follow-up period of March 2025 to February 2026, separated by a 4-month implementation washout.

The analysis included 51,097 patients at baseline and 54,257 at follow-up, of whom 33.2% and 31.6%, respectively, were classified as high risk. Among high-risk patients, use of every GDMT class increased after implementation (all P < .001). SGLT2 [sodium-glucose cotransporter 2] inhibitor use rose from 30.7% to 37.9% (OR, 1.38; 95% CI, 1.33-1.43), glucagon-like peptide-1 receptor agonist use from 14.5% to 18.2% (OR, 1.31; 95% CI, 1.25-1.38), mineralocorticoid receptor antagonist use from 13.5% to 15.9% (OR, 1.22; 95% CI, 1.16-1.28), and renin-angiotensin-aldosterone system inhibitor use from 61.2% to 63.5% (OR, 1.11; 95% CI, 1.07-1.15).

UACR testing across all patients increased from 43.0% to 50.7% (OR, 1.36; 95% CI, 1.33-1.39; P < .001). Increases in SGLT2 inhibitor (P = .028), mineralocorticoid receptor antagonist (P = .021), and renin-angiotensin-aldosterone system inhibitor (P = .040) use were greater among high-risk patients than among low- and intermediate-risk patients, whereas gains in glucagon-like peptide-1 receptor agonist use were similar across risk strata.

“Risk score reporting and provider feedback were associated with increased GDMT uptake and albuminuria testing, with greater uptake among patients classified as high-risk by the Klinrisk model,” the authors concluded. “These findings suggest risk stratification can direct kidney-targeted therapies to those most likely to benefit and provide a scalable template for value-based CKD care.”

CKCC Model Linked to Better Vascular Access, Lower Costs

A separate retrospective cohort study examined the first 2 years of the Comprehensive Kidney Care Contracting (CKCC) option of the CMS Kidney Care Choices Model.² The voluntary model, which began in 2022, aims to delay dialysis and encourage transplant for Medicare beneficiaries with stage 4 or 5 CKD or end-stage renal disease; under the CKCC option, kidney contracting entities that include nephrologists and transplant providers take on varying levels of financial risk for total cost of care.⁴

The study compared Medicare fee-for-service beneficiaries in CKCC practices with geographically matched beneficiaries outside the model over 8 quarters in 2022 and 2023 across 9 outcomes in patients with stage 4 to 5 CKD and those with dialysis-dependent end-stage renal disease.² In difference-in-differences analysis, CKCC practices showed a statistically significant differential quarterly improvement in permanent vascular access at hemodialysis initiation (β = 1.44 percentage points per quarter; 95% CI, 0.24-2.65; P = .023). Preemptive kidney transplant rates increased significantly over time among CKCC-aligned beneficiaries with CKD (0.021 per 100 person-years per quarter; P = .014) but not among comparators (P = .318), and CKCC-aligned beneficiaries with stage 4 to 5 CKD had lower costs (mean difference, –$3876 per year; P < .001).

The author, Ohm S. Tripathi, of the University of Connecticut, wrote that the findings support extending voluntary, risk-bearing kidney care payment models and “identify permanent vascular access planning as a priority for broader dissemination.” The analysis differs in scope from CMS’ independent evaluation of the full Kidney Care Choices Model, which found no statistically significant effect on total Medicare Parts A and B payments in 2023 and an increase of $304.8 million in net Medicare spending, driven mostly by incentive payments to participants.⁵

LLM-Drafted Prior Authorization Letters Still Need Review

A Mayo Clinic–led team compared OpenEvidence and ChatGPT-5 for drafting prior authorization letters for nephrology medications across 28 clinical scenarios.³ “Writing prior authorization letters is a time-consuming task that can delay treatment initiation and increase clinician burden,” the authors wrote. “Large language models may help facilitate PA workflows, but the quality of AI-generated letters should be evaluated.”

False statements appeared in 3.6% of ChatGPT-5 letters and 7.1% of OpenEvidence letters (P > .99), and International Classification of Diseases, Tenth Revision (ICD-10) coding was correct in 78.6% and 67.9% of letters, respectively (P = .25). Both tools produced valid references in 92.9% of letters, and clinical reasoning was rated strong in 89.3% of ChatGPT-5 letters and 78.6% of OpenEvidence letters (P = .45).

None of the comparisons reached statistical significance, but the authors reported different error patterns. OpenEvidence produced longer letters that often exceeded the 350-word target, which the authors linked to more chances for inaccuracies alongside added clinical detail. ChatGPT-5 produced more concise drafts with fewer high-impact mistakes, and both tools generated hallucinations.

“LLMs could generate clinically usable PA letters, although error rates were similar across models, with differing error profiles,” the authors concluded. “These findings support the use of LLM-generated letters as clinician-reviewed drafts prior to clinical use.”

References

1. Tangri N, Ferguson TW, Bhanushali GK, et al. Applying risk stratification to improve guideline-directed medical therapy uptake and albuminuria testing in a large US nephrology practice. Poster presented at: American Society of Nephrology Kidney Week 2026; October 22-25, 2026; Denver, Colorado. Abstract FR-PO0018.

2. Tripathi OS. Effectiveness of the Comprehensive Kidney Care Contracting (CKCC) model on kidney treatment outcomes and Medicare expenditures. Poster presented at: American Society of Nephrology Kidney Week 2026; October 22-25, 2026; Denver, Colorado. Abstract FR-PO0019.

3. Aiumtrakul N, Thongprayoon C, Kookanok C, Miao J, Cheungpasitporn W. Assessment of OpenEvidence vs ChatGPT-5 for prior authorization letter generation in nephrology. Poster presented at: American Society of Nephrology Kidney Week 2026; October 22-25, 2026; Denver, Colorado. Abstract FR-PO0031.

4. Kidney Care Choices (KCC) Model. CMS. Updated September 21, 2026. Accessed October 23, 2026. https://www.cms.gov/priorities/innovation/innovation-models/kidney-care-choices-kcc-model

5. Kidney Care Choices (KCC) Model executive summary: second evaluation report, performance year 2023. CMS. February 2026. Accessed October 23, 2026. https://www.cms.gov/priorities/innovation/data-and-reports/2026/kcc-2nd-annual-report-execsum