CKD Alert Design May Limit Clinician Action: Antoine Créon, MD

Chronic kidney disease (CKD) alerts did not translate into more clinician action at any stage of care in a Stockholm primary care trial, and alert fatigue and interface design may help explain why, according to Antoine Créon, MD, a PhD student at Karolinska Institutet.

In an interview with The American Journal of Managed Care^®, Créon discussed the limits of what the trial can show, why alerts may not have changed clinician behavior, and the wide variation in testing between centers.

What the Trial Can and Cannot Show

The trial did not establish that the alert improved earlier detection, so Créon said it cannot be said that earlier detection led to more referral or treatment. At every stage of management, including screening, retesting, treatment, and referral, the alerts did not translate into more clinician action. Population-level measures also cannot determine whether treatment or referral was appropriate for an individual patient or why it was not done.

The approach still has potential for scale because it is embedded in existing electronic health records. Once integrated, it can identify care gaps across many patients without manual review of every record. Regional deployment within the local record system was feasible.

“The problem is that successful deployment does not guarantee improved care,” Créon said.

Créon described the clinical logic as transferable but said effectiveness requires local evaluation. The trial involved one region, one health record software, and one user interface, so it cannot establish whether results would differ in another environment.

Why Alerts May Not Prompt Action

Créon called alert fatigue plausible. “Clinicians may receive many messages, and the CKD alert also competes sometimes with the results of the consultation,” Créon said. Other conditions and administrative tasks may take priority at the moment an alert appears, so even an appropriate recommendation may not arrive when it is easy to follow.

Interface design probably matters a great deal. “Recommendations need to appear at the moment where someone can act and is not distracted by something else,” Créon said, and the physician should be able to act with minimal effort. Créon noted that this is work the team must now pursue.

What Else Stands Between an Alert and Completed Care

Identifying a task is different from completing it. Testing requires an order but also patient participation and review of the results. Follow-up treatment may require discussion and monitoring, and referral depends on specialist capacity and on whether the patient attends the appointment.

Testing rates also differed widely between centers. Creatinine testing ranged from approximately 27% to 84%, which suggests that local organization and routines are important areas to investigate. The team is still analyzing usage data to understand how the system was used in practice and what may have contributed to the negative findings.