Of all the challenges I hear as a healthcare technology consultant, there's one that persistently worries doctors and administrators alike: patient readmission. In particular, post-operative readmissions seem avoidable. If only the patient followed discharge instructions, right?

The motivation here seems obvious—the risk of ignoring wound care instructions is sepsis. But over the past 15 years of building, reviewing, and revising patient journeys, I've found that when it comes to post-operative care, following instructions is rarely that simple.

Finally we have a new tool to deliver critical support at a vulnerable moment, and offer timely follow-up care. No, it's not a scalpel or a phone. Generative AI can tailor care instructions to each patient's needs and follow up regularly along their recovery journey, tracking progress and adapting directions as needed to improve overall health outcomes.

But as we say in product inserts: For best results, consult your medical professional. AI needs the guidance and oversight of healthcare and medtech professionals, working as a team. Here are five ways we can work together to reduce readmissions.

1. Identify relevant KPIs AI can track.

Readmitting patients post-surgery puts an unexpected strain on them, their families, their providers—and their healthcare system. When patient readmission rates go up, healthcare quality ratings go down, directly impacting CMS reimbursement. That means fewer resources to go around for everyone, including other patients in need.

To make the post-operative discharge process work better for everyone, we need to identify key performance indicators (KPIs) that can help us track success over time. We can deploy AI to gather key metrics in three categories: core readmission metrics, patient satisfaction, and operational efficiency.

Core readmission metrics:

30-day readmission rates. AI can help calculate the percentage of patients readmitted within 30 days of discharge for conditions targeted by CMS, including heart failure, AMI, pneumonia, and COPD. Process improvements should result in significant decreases compared to pre-launch benchmarks.
Excess readmission ratio (ERR). AI can help monitor the hospital's ERR for each CMS-tracked condition. The goal here is to reduce ERR scores below 1, demonstrating better-than-expected performance.

Patient satisfaction:

Patient satisfaction scores (HCAHPS). AI can assist with collecting data on patient satisfaction with the discharge process, and their interactions with AI. This ongoing research can help track improvements in relevant HCAHPS (Hospital Consumer Assessment of Healthcare Providers and Systems) survey measures, including communication with doctors and nurses, responsiveness of hospital staff, and clarity of discharge instructions.

Operational efficiency:

Discharge processing time. AI can help track the average time between the decision to discharge and the patient's actual discharge. Extra time spent on discharge procedures is not only an inefficient use of essential healthcare resources, but an avoidable stress on patients anxious to go home—and other patients waiting to be admitted.
Staff time allocation. AI can track time saved on administrative tasks due to automation, and redirect it toward patient care. This helps improve care and reduce burnout for purpose-driven healthcare providers, who would rather focus on life-changing care than paperwork.

2. Apply deep human understanding to improve processes with AI.

If you wonder why patients don't follow aftercare instructions, try talking to them after surgery. The last thing many anxious, groggy post-op patients want to talk about is wound care. Consider other touchpoints the patient's journey where patients might be more responsive to post-op care instructions.

Using generative AI with a human-in-the-middle approach can help pinpoint physical and digital processes where healthcare systems can improve communication, enhance patient support, and reduce readmissions. With empathetic human oversight, AI can help in four key areas:

Simplified discharge instructions and educational resources. AI tools can analyze patient records to identify potential medication conflicts, health conditions, language preferences, and lifestyle factors. Taking these factors into consideration, AI can tailor instructions for each patient in language they can easily understand.
Appointment reminders & proactive follow-up. An AI voice assistant can personalize communication, proactively manage appointments, and initiate follow-up calls to address potential issues early.
Personalized questionnaires. AI-enhanced chatbots can access relevant parts of the EHR, including medical history, current medications, and recent lab results. This allows the chatbot to ask the patient targeted questions. How is the wound healing? Are they feeling any pain, and how would they describe it? Are they taking their medication, and if not, why not? AI can generate and upload detailed reports directly into the EHR, and also summarize them for clinical staff review.
Addressing diverse needs. There's more to patient care than wound care. AI can provide multilingual care instructions, offer basic mental health support, assist with transportation coordination, and offer tailored support to caregivers.

3. Anticipate cultural and operational hurdles with AI solutions.

Change management is key to any effective digital transformation, but especially in healthcare. New technologies are understandably met with rigorous skepticism by dedicated healthcare providers, for whom "do no harm" is a foundational principle. When integrating new technologies like AI into post-operative care, build training and culture shifts into your adoption strategies and timelines. Key strategies for successful implementation include:

Addressing staff concerns proactively. A snappy AI explainer video isn't enough to convince busy care teams to adopt it into demanding routines. In-depth training must explicitly demonstrate how generative AI tools support clinicians, augmenting their capacity rather than adding to their workloads. One message deserves repeating: AI is here to help, but the human-in-the-middle approach remains optimal for patient safety and quality of care.
Seamless workflow integration. AI aims to help with to-do lists, not add to them. AI tools must be meticulously designed for integration, streamlining processes without disrupting clinical workflows. This requires intensive collaboration between IT and clinical teams from the project's inception.

4. Ensure IT and data infrastructure can handle emergencies.

Emergencies are a constant in healthcare, not a contingency. Before you deploy a solution, you need to pressure-test it to consistently meet—and ideally exceed—urgent patient needs and rigorous regulatory standards in adverse circumstances. To build, scale and sustain AI-driven post-op care, build these requirements into your solution:

Robust data systems. Secure, integrated systems are essential for storing patient preferences, communication history, and accessed resources. AI must safely access data in compliance with HIPAA to adapt post-op information to patient needs.
Interoperability. AI solutions must seamlessly interact with existing EHRs. Standards like FHIR (Fast Healthcare Interoperability Resources) facilitate these exchanges.

5. Address ethical and regulatory challenges first, not last.

AI represents an opportunity to extend and improve post-op care for patients, but it requires human oversight to ensure safe, quality care for all. So before you build your solution, strategize how it might comply with and improve upon medical ethics, accessibility and regulatory standards. Any robust, future-proof solution must consider:

Algorithmic bias. Build in monitoring and failsafe systems for AI models to help prevent unfair patient treatment based on factors like race, gender, ability, or socioeconomic status.
Regulatory compliance. Adherence is mandatory to HIPAA (USA), GDPR (EU), and other applicable regulations, including emerging state-level privacy laws. To future-proof your solution, consider how you might improve upon current standards in anticipation of their ongoing expansion.

Summary

Healthcare leaders are right to worry about post-op readmissions. They put strain on patients, families, providers, and healthcare systems—and they're often avoidable. AI has potential to transform post-op processes, improving outcomes for patients while relieving pressure on dedicated care teams.

From the 15 years I've spent improving patient care through technology for healthcare leaders from USCF to Roche, I've distilled a five-step framework to leverage AI for enhanced efficiency and improved patient experiences. I'm sharing it here in the spirit of open innovation—and if you'd like help applying it, let's talk.

With AI, we can provide discharge instructions patients can actually follow–and deliver effective, personalized follow-up care we can track with relevant health metrics. Success requires ongoing collaboration between technologists, clinicians, and patients. But consider the benefit: If we can reduce readmissions, we all get to go home from the hospital with a brighter outlook on the future.