Can a machine learning model find hidden diabetic kidney disease while I am hospitalized?

Diabetic kidney disease (DKD) often starts silently. Standard tests like eGFR and urine albumin may stay normal in early stages, so the condition can be missed during a hospital stay. Researchers have developed machine learning models that use routine lab results and clinical information to spot this 'occult' DKD. One study found that a logistic regression model using eight common variables could identify hidden DKD in hospitalized patients with good accuracy ². This means your hospital data might already contain clues that a computer can use to catch kidney disease early.

What the research says

A 2025 multicenter study developed a machine learning model specifically to detect occult DKD in hospitalized patients with type 2 diabetes ². The model used eight routine variables: hemoglobin, HbA1c, hypertension status, uric acid, sex, microvascular complications, cardiovascular disease, and albumin-to-globulin ratio. It performed well in both the derivation and external validation cohorts, suggesting it could be a practical screening tool during hospitalization ².

Another study tested a series of machine learning models that added layers of information — from basic lab results to social determinants of health — and found that discrimination improved as more data was included ³. This supports the idea that combining multiple data sources can enhance detection of DKD risk.

A separate study integrated urine physical characteristics (like foam stability) with metabolic markers to create a fusion model that achieved high accuracy for early-stage DKD detection ⁶. While this approach is more specialized, it shows that innovative methods can uncover hidden kidney disease.

These models are not yet standard in hospitals, but they demonstrate that machine learning can analyze existing data to flag patients who might otherwise be missed. If you are hospitalized, your doctors may already have the information needed to assess your risk, even if standard tests look normal.

What to ask your doctor

• Could my routine lab results (like hemoglobin, HbA1c, and uric acid) be used to screen for hidden kidney disease?
• Are there any machine learning tools or risk calculators available at this hospital to detect early diabetic kidney disease?
• If my standard kidney tests are normal, is there still a chance I have early kidney damage, and how would we know?
• What additional tests or monitoring would you recommend to catch kidney disease early?
• Should I be concerned about my blood pressure variability or other factors that might affect my kidneys?