We usually see coders doing 20 to 30 letters per hour using BetterLetter, even while working on complex clinic letters. This is a 2x increase on the speed at which coders can code comparable letters in EMIS. For an average practice who receives approximately 80 to 100 clinical letters per day, this means that one part time coder can clear all letters within 24 hours.

Without BetterLetter, when pressed for time practices sometimes resort to coding minimally (i.e. only major diagnoses and actions). With BetterLetter, you do not have to compromise clinical data quality as our AI suggestions and one-click coding make it easy to comprehensively code every document.

The change is minimal for most practices: we will automatically collect letters as they arrive at the practice, code and file into your EMR, and workflow to existing groups.
‍
The main change for some practices is that you will now do your coding at the point of filing. This reduces complexity, as letters are processed once when they arrive into the practice, and only those letters with workflow need to be actioned further. Overall, BetterLetter will help to increase efficiency, allowing for more letters to be processed in a shorter amount of time, without compromising the quality of the data that is extracted from the letters.

BetterLetter will collect letters from a Docman folder in your filing area, or you can upload letters directly to the platform. We are working on collecting direct from MESH and EDI feeds, and this will be available in early 2025.

All coding and free text is placed in a consultation in the EMR with the letter attached. The coding is organised with the appropriate headings (e.g. diagnoses, procedure, etc), and linked to problems as a new episode or review as appropriate. For Docman practices, we also file to Docman and generate tasks to existing workflow groups.

The letters filed still have the colour coding and summary page, however the EMR and workflow data otherwise appears to clinicians exactly the same as letters filed the old fashioned way.

Onboarding usually takes a couple of days. Firstly you fill in the Onboarding Form, then we'll hop on a quick call to enable the API and discuss any custom workflow groups you wish to have. We'll then share our training materials with your team, and book a training call to go over everything and do a few letters together. After that, most practices are up and running with no major issues.

Yes, by default BetterLetter comes with four workflow groups (Safeguarding, Reception, Meds Management and GPs), but we can add additional groups to match your practice custom workflow. Please contact us on hello@betterletter.ai and we will get that set up for you.

Yes, we offer full training via Teams, alongside digital materials such as video tutorials, articles and other resources to help you become familiar with the platform.

Although we do not currently work with SystmOne, we are in the IM1 process with TPP, and we are hoping to have the S1 integration live in 2025.

We use two types of AI: supervised machine learning for document classification (letter type and department), and graph enabled information extraction (a patented combination of NLP, knowledge graphs and large language models) for information extraction and SNOMED coding.

Via out-of-sample testing we have shown our latest machine learning models to be over 95% accurate, however it is more difficult to quantify accuracy metrics for information extraction as every practices uses a different coding protocol, and frequently two human coders will disagree about how a certain letter should be coded. For this reason, while our AI is very good, it is not perfect and a human must still check and individually approve each suggestion made by the AI.

Yes, BetterLetter's AI is very good at reading handwriting and understanding abbreviations. Sometimes, however, an abbreviation can mean two different things (for example, RA could be rheumatoid arthritis or right atrium), and therefore a human should always double check what has been suggested by the AI.

Yes, the AI prioritises QOF and IFF codes, and these are always indicated in the BetterLetter app with Q and I icons, so a coder can see at a glance where they have been used.

Yes, the AI will suggest a workflow that is most appropriate based on the contents of the letter. However, if you would like to change the workflow that the AI has suggested to a more appropriate one, you can override the AI’s choice.

Yes, the AI will identify similar problems in a patient's health record and notify the user so that you can either link the problem to an existing one as a review or create a new problem.