Professor David Eyre, Consultant in Infectious Diseases, talks about different types of data in IORD, and why free text data is important for research
The information in IORD is mostly numbers or groups. For example, it might say that on 18 July 2016 at 17:11 someone went to a specific part of the hospital, and at that time their blood had 11.2 grams per decilitre of haemoglobin (which helps the body use oxygen), their body temperature was 37.2 Celsius and they had a microbe called MRSA in their blood. It’s like writing down all the important details about a person’s health using numbers and categories.
Example of IORD data
| iord id | admission date | discharge date | admission source | admission method | primary diagnosis code | secondary diagnosis code | haemoglobin | sex | month / year of birth |
|---|---|---|---|---|---|---|---|---|---|
| 509123 | 29/07/2023 | 31/07/2023 | 19 | 21 | J22 | 150.0 | 12.7 | F | May-33 |
| 43251 | 30/07/2023 | 05/08/2023 | 19 | 22 | A41.9 | E11 | 10.4 | F | Dec-49 |
| 10001762 | 05/06/2023 | 09/06/2023 | 19 | 11 | J15 | J44.9 | 15.6 | M | Jul-65 |
| 501943 | 17/06/2023 | 08/06/2023 | 51 | 12 | L30.3 | 14.5 | F | Jul-43 | |
| 444100 | 12/07/2023 | 31/08/2023 | 19 | 13 | G03.9 | M | Jan-62 | ||
| 49172 | 23/06/2023 | 27/06/2023 | 19 | 21 | K65 | J45 | 10.2 | M | May-55 |
| 10283 | 30/06/2023 | 02/07/2023 | 19 | 21 | L03.9 | E11 | 16.9 | M | Apr-52 |
| 99831 | 10/07/2023 | 17/07/2023 | 19 | 21 | N10 | M | Mar-47 | ||
| 87513 | 17/06/2023 | 20/06/2023 | 19 | 11 | L98.4 | 12.7 | M | Sep-49 |
In IORD, there’s also something called “free text” data. This is what doctors and nurses write down or type in their notes. This kind of information comes from things like:
- Reports of scans, which are detailed pictures of the inside of your body, such as ultrasound, CT, or MRI scans.
- Why doctors gave someone antibiotics.
- What infection specialists, who are experts in fighting infections, have to say about a patient.
So, as well as the numbers and categories, there are these notes that help tell the whole story of a person’s health.
This kind of information is becoming more and more important for working out who is most likely to get different infections and understanding what’s really going on with people in the hospital. This is because the codes that are used to describe what’s wrong with a patient are mainly to make sure the hospital gets paid the right amount by the NHS, and they might not fully show what’s going on with the patient’s health.
Doctors and nurses are careful not to include any details that could identify a patient, like their names, in these reports. There’s a tiny chance, less than 1 in 10,000 times, that they might accidentally include something that could identify a person. It is really tough to get the important information from the report without having the whole thing.
- For example, if we just look for the word “pneumonia” (a serious chest infection) there is a big difference between these three different notes – “definite/probable pneumonia” or “pneumonia is one possibility but so is XXX” or “whilst pneumonia was originally considered, after XXX it is definitely not the cause”.
- We are working to test whether some new artificial intelligence programs could be a good way to find and get rid of accidental things that could identify a person.
Because of this, we are especially careful with this “free text” data to make sure it’s used properly. Right now, only researchers who work for the NHS can use the “free text” from IORD. And if they happen to come across any personal information while they’re looking at the “free text”, they have to let us know right away. It’s important to keep people’s private information safe.