A Machine Learning Algorithm to Improve the Use of Interpreters for Hospitalized Patients with Complex Care Needs


Supporting Health Systems in Advancing Care Delivery


Improving Equity in Healthcare with Digital Healthcare Solutions

A machine learning, predictive analytic intervention has the potential to improve healthcare, making it more equitable for patients with a non-English language preference and complex care needs by supporting timely interpreter use to facilitate decision making and promote patient-centered care.

In-person interpreters improve communication between patients and their families and providers, supporting more equitable care

For hospitalized non-English language preference (NELP) patients, especially those with complex care needs, communication between patients and their families and inpatient healthcare providers is often difficult, and these communication barriers can lead to lower quality of care, poor health outcomes, and increased length of hospital stay. In-person medical interpretation in a patient’s preferred language improves communication, patient satisfaction, and clinical outcomes. Moreover, using in-person interpreters reduces cultural, linguistic, and literacy barriers, supporting more equitable care.

“Interpreters can help patients and clinicians understand each other and relay information, but they can also help inform the clinicians about potential other background issues that might be important, so clinicians have a better understanding of why decisions are being made, or which people in the family are likely to be important for decision making.”- Dr. Amelia Barwise

Complex medical conversations are hard, even among English-language preferred speakers

Using in-person interpreters encounters several barriers, beyond the overall nationwide shortage. Getting an interpreter to the bedside takes substantial coordination and logistics, and some clinical providers are hesitant to engage with that process in busy hospital environments in which they are likely caring for multiple patients. Frequently, providers make do with using family members to interpret or their own limited language skills, but that is not ideal.

“Families can either not understand all of the medical jargon, so there's the accuracy piece, but then they can also withhold information.”- Dr. Amelia Barwise

An interpreter is critical to having these conversations, especially as they relate to treatment, end-of-life care or withdrawal of care, or to any other sensitive discussions that are already difficult among English-language preferred speakers.

Dr. Barwise and her research team in Mayo Clinic Rochester want to find a better way to identify which patients need an interpreter and prioritize the use of interpreters for these patients. The team is developing machine learning algorithms to reliably identify NELP patients with complex care needs early in their hospitalization.

Use of machine learning algorithms will improve appropriateness and timeliness of using interpreters for patients who need it most

The team is developing the tool with input from clinicians and interpreters to build a process to connect identified patients with timely interpreter services. Once developed, they will integrate the tool into the clinical workflow at Mayo Clinic Rochester and study whether use of the tool increases appropriate and timely interpreter use among NELP patients with complex care needs. In addition to using the tool to identify when to provide an interpreter, they also hope to normalize the use of interpreters to reduce cultural, linguistic, and literacy barriers, and in doing so, support more equitable care.