How one kidney care company built a solid AI foundation

Tom Hawkes, chief technology officer at Strive Health

By the time many health systems begin experimenting with artificial intelligence, they quickly discover the technology itself is rarely the biggest obstacle. Disconnected data, limited interoperability and clinician workflows that are already overloaded often prevent promising AI projects from moving beyond pilots.

Those were the problems Strive Health set out to solve as it expanded AI across its value-based kidney care organization. Rather than starting with algorithms, the organization concentrated on creating a stronger clinical data foundation that could give care teams a more complete picture of patients with chronic kidney disease.

The approach has produced measurable clinical and operational gains.

Among them: a reported 77.2% reduction in disease progression for Stage 3b chronic kidney disease patients, a 65.2% reduction for Stage 4 patients, up to a 30% reduction in 30-day readmissions among transitional care management patients completing medication reconciliation, a 32% reduction in documentation time and nearly a 20% productivity improvement across roughly 500 clinicians.

Patients with chronic kidney disease often receive care from primary care physicians, nephrologists, hospitals, laboratories and other providers. Each encounter generates valuable information, but those data frequently remain isolated in different systems.

"There are dozens of signals that can indicate whether their condition is stable or beginning to change, but those signals rarely appear in one place," Hawkes said.

Instead of treating each data source independently, Strive focused on connecting those signals to provide care teams with a more holistic patient view.

Equally important, the information has to arrive quickly enough to affect clinical decisions. Hawkes noted that symptoms reported separately to different specialists may seem insignificant in isolation but can reveal meaningful deterioration when analyzed together. The same applies to changing laboratory values, missed prescriptions and other indicators that a patient's condition is worsening.

By improving visibility into those events as they occur, care teams have greater opportunity to intervene before avoidable complications become hospitalizations or other high-acuity events.

Even accurate insights can fail if they create more work for clinicians.

Hawkes said Strive intentionally designed AI to fit existing workflows instead of requiring clinicians to monitor another application or dashboard.

"In an environment already saturated with alerts, dashboards and point systems, new technology must be user-friendly," he said.

The organization's AI largely operates behind the scenes by organizing clinical information, highlighting what deserves attention and reducing administrative burden while leaving clinical decision-making to physicians and care teams.

Improving medication reconciliation

Medication reconciliation became one of the organization's most important AI-supported workflows following patient discharge from the hospital.

Medication changes, discharge instructions and follow-up information often remain spread across hospitals, physician offices and specialists. To bridge those gaps, Strive combines health information exchange data with AI-generated patient summaries that consolidate medications and relevant clinical history before clinicians begin reconciliation.

Ambient documentation technology further supports clinicians by capturing medication-related information during patient conversations and integrating it into workflows.

Together with broader data aggregation tools, these capabilities reduced documentation time by 32% while increasing clinician productivity by nearly 20%.

Most importantly, Hawkes said the organization observed up to a 30% reduction in 30-day readmissions among transitional care management patients who completed medication reconciliation.

Despite those gains, clinicians remain responsible for reviewing medications, confirming discrepancies, discussing treatment plans with patients and determining appropriate interventions.

"The role of AI is to improve the workflow, not replace clinical judgement," Hawkes said.

Strive's Care Multiplier platform analyzes laboratory values, medication histories, diagnoses and broader care patterns through machine learning models.

Rather than relying only on health plan classifications, the platform independently stages patients based on available clinical evidence, helping identify individuals whose kidney disease may be more advanced than previously recognized.

The platform also prioritizes patients according to disease trajectory, readmission risk, medication gaps and comorbidities, while showing clinicians the evidence supporting each recommendation.

Those insights allow care teams to escalate nephrology referrals sooner, strengthen medication management and increase outreach to high-risk patients before disease progression accelerates.

According to Hawkes, clinicians always see the laboratory values, medication history and care patterns underlying AI recommendations, allowing them to apply clinical judgment before modifying care plans.

Looking across the organization's AI initiatives, Hawkes believes their greatest value comes from helping clinicians spend more time caring for patients and less time searching for information or documenting encounters.

He argues that governance and trust must come before widespread AI deployment. Organizations with fragmented or incomplete data are unlikely to move beyond isolated pilots because clinicians will not trust inconsistent outputs.

Strong governance provides confidence that AI recommendations are based on reliable information while ensuring appropriate oversight, he said.

Equally important is involving frontline clinicians throughout implementation so new technology clearly improves daily work instead of creating additional friction.

For provider CIOs, the broader message is that successful AI adoption depends less on sophisticated models than on disciplined work involving interoperability, data quality, governance and thoughtful workflow integration, Hawkes advised.

Once those fundamentals are established, AI can become a practical clinical tool that supports earlier intervention, strengthens care coordination, reduces administrative burden and ultimately improves outcomes for patients with chronic kidney disease, he concluded.

Follow Bill's health IT coverage on LinkedIn: Bill SiwickiEmail him: [email protected]Healthcare IT News is a HIMSS Media publication.

WATCH NOW: For Duke, AI strategy is business strategy