A shared language for clinical care

Eucalyptus is an Australian digital health company that runs online digital clinics across the globe — each focused on a specific area of care. Patients get assessments, prescriptions, and ongoing treatment remotely, with oversight from certified practitioners, rather than visiting a physical clinic.

All information about the patient’s profile, medical history and treatment details is housed in an Electronic Medical Record (EMR). The record can be accessed by different people including doctors, medical support team and customer support teams — all with different permissions.

As the business is growing rapidly at a global scale, there is a need to keep accuracy and safety intact.

The current method for the team to add information like allergies, medications and medical conditions is free-text. The fields are populated by patients from their initial medical quiz answers which is then reviewed and edited by the doctor or medical support team. Given the submitted and edited data is not collected in a standardised way, it introduces high rates of inconsistency, such as +10 ways of spelling the same condition like Polycystic Ovarian Syndrome.

SNOMED CT (Systematised Nomenclature of Medicine – Clinical Terms) is a digital clinical database used across healthcare systems globally. It contains standardised clinical concepts, symptoms and medications that can be accurately documented inside the EMR. Adopting it would align Eucalyptus’s records with a global standard as the business scales across markets.

First and foremost, swapping a free text field to multiple choice answers immediately reduces the cognitive load for patients. No longer do they need to think of what to write or how to answer the question, the answers shown are also linked to the Snomed database. This allows the internal teams to review quiz answers with confidence and reduces error rates by a high margin.

A doctor is usually required to review, add and remove any medical details during and after a consultation. To provide a seamless diagnostic experience while they are in a call, changes to the edit flow included:

• A new call to action button: clearly identifying how to add one or multiple conditions, medications or allergies.
• Considered search behaviour: with over thousands of conditions and medical types, pre-filling the search results after 4 key strokes immediately helps a doctor narrow down the correct problem.
• Modifying any previous or new logs: regardless of user type (doctor or medical support team), details such as condition status, notes and deleting can be easily amended through one form over the course of a patient’s treatment journey.

The previous experience had the ‘Edit’ button visible to all support teams regardless of their editing permissions. This became a misleading and frustrating experience for users like customer support team who only have read permissions. Those users were only clicking into something to only find out they couldn’t action anything. By changing the ‘Edit’ button to be viewed conditionally — visible for medical support and hidden for customer support team, this allows everyone to understand their available actions at a glance and decrease frustration. Applying the same UI edit drawer for both platforms reduces cognitive load especially for the medical team who have to juggle patient cases across two platforms.

Standardising the input experience delivered measurable gains across patient, doctor and support teams alike.

The new changes in the initial quiz significantly improved the top-of-funnel patient experience. The conversion rate of users completing their initial screening quiz increased by 31% after full implementation.

This suggests that patients had a reduction of cognitive load at the medical section, and were able to quickly find their conditions swiftly.

The new search experience greatly improved the doctor’s experience during online consultations. The richness of the data provided meant they didn’t have to rely on their memory to make any logs, improving consultation accuracy as a whole.

Without SNOMED, the triage process was relying 100% on doctors to manually review and reject ineligible cases. With the new implementation, rejection rates specifically for patients who may or have multiple conditions (co-morbidity risk) shown in the initial quiz answers dropped by about 66%. This result indicates that the triage process is now more effective at filtering users at the initial quiz stage based on structured data, saving doctors and patients time and frustration.