Closing care gaps with AI-driven referrals

Patient referral leakage is more than a money matter for hospitals and health systems. While revenue losses are real, so are the care gaps that stem from manual referral workflows, fragmented information exchange, and appointments that simply go unscheduled. Here, artificial intelligence (AI) is emerging as an effective way to help health systems refer patients to the right providers and help providers manage the referrals they receive.  

Unscheduled appointments, limited insights

Referral leakage presents a twofold problem to providers. One is when appointments never get scheduled. It’s been estimated that up to 50% of outpatient specialty referrals aren’t completed, and as many as 58% of referrals from telehealth and 65% of referrals from primary care aren’t completed. The clinical implications of referral leakage are clear and can include care gaps, communication breakdowns, delayed diagnoses, and patient safety risks.  

The other challenge is when referrals are made outside a health system’s network. Here, the problem is mainly financial – WebMD estimated up to 30% of hospital revenue can be lost through this type of referral leakage, adding up to $500 million per year – but there are also clinical implications. Definitive Healthcare noted out-of-network providers may lack access to complete medical records, and the referring provider may lack visibility into the next steps of the patient journey. 

The link between poor interoperability and care gaps is well documented. A team of Partners Healthcare (now Mass General Brigham) researchers described the problem back in 2016, noting that missing or incorrect data in referrals contributes to limited support for care coordination as well as inefficient information-gathering processes. Even in 2024, researchers highlighted the prevalence of fax and phone communication, due in part to the limitations of health information exchange (HIE) technology.   

Improved scheduling, patient retention, and workload management

As with many aspects of healthcare delivery, AI tools appear poised to modernize the referral process. It starts with breaking down information silos: Aggregating and normalizing data from disparate sources to get a more complete view of the patient record.  

The immediate benefit is a better understanding of the patient’s condition, which helps ensure they see the best specialist to address their concerns. From there, AI tools integrated with the EHR can search for and recommend in-network providers. Advanced tools can even automatically schedule the appointment and send the patient a reminder before their visit is over, setting the stage for short-term gains (an appointment that’s kept) and long-term gains (a patient that remains within the network). 

The downstream benefit of aggregated data is more comprehensive data exchange. This helps the new provider understand a patient’s condition and make an informed diagnostic decision and/or treatment plan, even if they don’t have the same EHR system in place. Care teams can also act quickly; one Canadian hospital saw a 48% reduction in time from referral to initial treatment for lung cancer by integrating and automating care coordination and diagnostic pathways, which included AI-driven referrals.  

AI can also support providers in managing the referrals they receive. For example, researchers in Brazil demonstrated higher specificity and lower sensitivity using AI models compared to human “gatekeepers” to distinguish which referrals to specialty care required authorization or additional information.  

With such a model in place, researchers concluded patients needing immediate attention would be more likely to get it, reducing wait times and ensuring resources were allocated where needed. At the same time, a gatekeepers’ workload would shift to referrals requiring clinical review – a job the researchers found humans did better than AI models.  

Stay ahead of cutting-edge AI in healthcare — engage with the innovators, operators, and researchers shaping what’s next in the AI Zone @ViVE 2026.