AI models predict ESKD risk
Medscape highlighted AI models that predict end‑stage kidney disease risk in hemodialysis patients and paired those predictions with nurse‑led interventions aimed at avoiding admissions. The coverage frames predictive models as tools that can trigger targeted nursing actions to prevent deterioration. (x.com)
For people on hemodialysis, new artificial intelligence tools are being used to flag who is most likely to be hospitalized within a week. In one 2023 Medicare analysis, nurse-led follow-up after those alerts was linked to fewer admissions. (medscape.com) Hemodialysis is the blood-filtering treatment used when kidneys can no longer do the job. In the United States, 433,396 patients were receiving in-center hemodialysis as of March 31, 2025, according to the National Forum of End-Stage Renal Disease Networks. (esrdnetworks.org) The models in the new study worked like early-warning scores. One looked for signs of fluid overload, another for infection, and patients with a score of 0.64 or higher were flagged as at risk of hospitalization within seven days. (academic.oup.com) Nurses then reviewed the flagged cases on a dashboard that showed the main reasons for the score and the patient’s clinical data. They could respond with actions such as adjusting dry weight targets, arranging extra dialysis, reviewing antibiotics or diuretics, making referrals, or setting up transportation and home health support. (renalandurologynews.com) Researchers studied 10,294 adult Medicare patients with end-stage kidney disease treated in 1,786 integrated kidney care clinics across the United States in 2023. The analysis included 83,928 risk-score observations, and 13,988 of them led to a case review and intervention. (academic.oup.com) (europesays.com) Those interventions were associated with an 8% reduction in the odds of hospitalization within seven days, with similar results whether remote nurses or clinic staff managed the response. In the subgroup with scores between 0.75 and 0.85, the reduction was 12%. (medscape.com) (europesays.com) The gains were smaller at the very top of the risk scale. Patients with final scores above 0.85 had higher hospitalization rates, and the interventions did not significantly change seven-day admissions in that group. (renalandurologynews.com) The study also found that patients with persistently high scores had 53% higher odds of hospitalization than patients whose scores spiked only occasionally. Older age and multiple hospital admissions in the prior year also tracked with higher risk. (renalandurologynews.com) Kidney specialists have been testing prediction models in dialysis for years, but most earlier work focused on longer windows such as 12-month hospitalization risk. A 2021 study of a dialysis hospitalization reduction program reported that machine-learning-directed interventions were associated with lower hospital admission rates at the clinic level. (pubmed.ncbi.nlm.nih.gov) The current report was retrospective and observational, not a randomized trial, and the authors said they did not have 30-day hospitalization data to show whether admissions were prevented or only delayed. Even so, the work points to a practical use for artificial intelligence in dialysis clinics: not replacing nurses, but telling them which patient may need attention first. (renalandurologynews.com)
