A 2023 cross-sectional study of 340,509 Medicare fee-for-service beneficiaries aged 65 or older with Alzheimer's disease and related dementias (ADRD) found that greater hospital adoption of patient-related AI/ML tools was associated with lower odds of frequent hospitalization, 30-day readmission, and preventable acute hospitalizations. The analysis, linking inpatient claims, the Medicare Beneficiary Summary File, and the American Hospital Association IT Supplement, also found that inpatient risk-prediction tools were associated with lower total Medicare spending, while treatment-recommendation tools were associated with higher beneficiary out-of-pocket costs.

Why It Matters To Your Practice

• Older adults with ADRD are at high risk for repeat admissions and potentially avoidable hospitalizations, making this a useful test case for whether AI changes real-world utilization.

• The strongest signals were tied to AI functions clinicians already recognize: inpatient risk prediction and identification of high-risk outpatients.

• For clinicians facing pressure on readmissions and care transitions, the study suggests some hospital AI deployments may improve utilization metrics without raising overall spending.

Clinical Implications

• Not all AI tools performed the same: risk-prediction and high-risk identification tools were more consistently linked to lower inpatient utilization than treatment-recommendation tools.

• If your organization is evaluating AI, prioritize the clinical workflow it supports — triage, surveillance, and care management may deliver different value than recommendation engines.

• Ask operational questions before rollout: who acts on the alert, how quickly, and whether the model changes discharge planning, follow-up, or escalation decisions.

• Watch the financial tradeoffs: some tools were linked to lower Medicare spending, but treatment-recommendation tools were associated with higher patient out-of-pocket spending.

Insights

• This was an observational, cross-sectional study, so it shows association, not causation.

• The population was limited to Medicare fee-for-service beneficiaries with ADRD who had at least one hospitalization in 2023, so results may not generalize to Medicare Advantage, younger patients, or other conditions.

• The findings reinforce that AI should not be treated as a single intervention; outcomes varied by tool type and intended function.

The Bottom Line

• Hospital AI adoption — especially for risk prediction and identifying high-risk patients — was linked to fewer acute, potentially preventable admissions and lower readmission-related utilization in older adults with ADRD.

• For practice leaders, the takeaway is practical: evaluate AI by the specific job it performs, and track both utilization outcomes and patient cost burden after implementation.