Should we be storing police information in records?
Dr Chris Martin points up some particular data anomalies that you can find and fix before the end of March
You have slogged away all year making sure you meet your quality point indicators. You may even have run you own searches to monitor progress. Yet you are still underperforming on some quality indicators. Why?
You have assumed that as you do the work and record it on the computer, your system will demonstrate that fact adequately. Everyone knows that old IT adage 'garbage in, garbage out'. But think a bit harder. To get high-quality data out you must have high-quality data in.
If you have really done nothing to organise your data in advance there isn't much time left to take action. However, there may be some particular data anomalies that you can find and fix before the end of March.
Free text entries are particularly important for a good clinical record that serves the purpose of continuity of care and a good legal record of events.
Unfortunately, free text is hopeless for auditing records. If you have been in the habit of recording important information in free text, you probably have a very good clinical record for the care of a particular patient, but that data is pretty much lost to the Q&O.
Fixing the free text problem will not be easy. An example of this may be 'never smoker' codes. You may have recorded this fact in free text in the early 1990s and have not realised
that the Q&O searches will not pick this up. This is important for the indicators BP 2, Records 10 and Records 16.
These ask for any record of smoking status that has ever been recorded, not just a record in the last 15 months.
In the case of 'never smokers' any record, even from 20 years ago, will count for CHD 3, stroke 3, BP 2,
DM 3, COPD 4 and asthma 4. Between them, these indicators account for 46 points.
If your computer system can support it, run a free text search using the following search format:
·Search for all patients with any free text entries indicating having been a non-smoker
·Remember to include synonyms, like 'never smoked' and every combination of upper and lower case
·Then exclude any patients who have a Read code for 'never smoker' (1371).
You can follow a similar procedure if you feel your disease registers are showing too low a prevalence.
This is important, as the amount of money you receive will be adjusted according to how much above or below the national average prevalence your practice's prevalence for each disease category is.
In these cases it may be more effective to search on medication for asthma, diabetes or antihypertensives, etc, and then check the records for missing or free text diagnoses.
Read codes are meant to be hierarchical. This means that related codes are grouped together under a common root code.
Unfortunately, this hierarchy is not always reliably implemented. One cause of this is historical misinterpretation of codes as a result of not considering their sub codes.
An important example is 'tobacco consumption' coded as '137..'. The sub codes include 'never smoked tobacco' (1371.) and 'current smoker' (137R.).
Unfortunately, code '137..' 'tobacco consumption' has become synonymous with 'current smoker' and this is now enshrined in the Q&O search algorithms.
The consequences are serious for those people or systems that historically recorded 'non-smokers' as '137..' with a number '0' attached to indicate that they consumed no cigarettes and were therefore non-smokers.
This is particularly likely to have occurred with legacy systems that recorded the number of cigarettes smoked, that subsequently converted data to this Read code on a system change.
If you are struggling with your smoking targets, check it out. If you find that you are affected you will need to run a search for all those occurrences of '137.' where there is no record of a sub code subsequently, and then correct them.
Read coding of proteinuria is important for diabetes. The first anomaly to mention here is the Read coding searched for by the Q&O software.
Anyone who read the Blue Book entitled 'The New GMS Contract 2003' issued by the NHS Confederation and the BMA will have noted that the 'preferred code' for proteinuria in the indicator DM 15, was '4678' (proteinuria).
Unfortunately, for some reason, when the Department of Health released the Q&O search algorithms in November 2003, the code '4678' was not included in the DM 15 search.
If you were quick off the mark and have already implemented that code in your diabetes check templates, you had better track them all down and change them to one of the accepted codes like 'R110.' ([D]Proteinuria) or 'R110z.' ([D]Proteinuria NOS), but not 'R1101.' ([D]Bence-Jones proteinuria).
This is another example of where the hierarchical structure of the Read coding system has been broken down by popular interpretation.
A more difficult proteinuria anomaly is the difference between the proteinuria of diabetes indicating nephropathy, and the temporary proteinuria that might be experienced in the context of a UTI, for example.
If you had been in the habit of recording your urinalysis on the computer using any of the proteinuria codes given below, you may need to amend them to one of the codes that the Q&O search does not pick up, given on the right-hand side of the table.
Some of these will be persistent proteinuria but others not, and they will need amending.
It does not matter if you have subsequent entries like '4672' (urine protein test negative), one proteinuria entry is always proteinuria as far as Q&O is concerned, even if the transient proteinuria occurred before a diagnosis of diabetes was made.
Track down those transient proteinuric diabetics and substitute a code like '4678' to keep it both clinically and Q&O meaningful (see left).
Another potential catch is the recording of being fit-free for more than 12 months in the epilepsy 4 indicator.
The only code that the Q&O search will seek is '667F' (seizure free > 12 months). A code '6676' (last fit), '667C' (epilepsy control good) or '667P' (no seizures on treatment) with an attached date more than a year ago will not do. Find them and correct them.
Asthma and ?-blockers
Q&O indicator CHD 10 requires patients with CHD to be taking a ?-blocker. Unfortunately a number of patients will
also have asthma and will be unable to take ?-blockers ever.
The exception code that would seem to be the most appropriate would be
'?-blocker contraindicated'. However, this is an 'expiring' code.
This means it is assumed that the contraindication only applies for this year and must be re-entered every year. The rationale is that some contraindications are temporary only.
For example, somebody may have severe claudication and may have had a bypass graft by the following year.
If you have been using this code, you may need to find your asthmatic patients with ischaemic heart disease who will definitely be permanently unable to take a
?-blocker and substitute a 'persisting' exception code like 'U60B7'
([X]?-adrenoreceptor antagonists causing adverse effects in therapeutic use, not elsewhere classified) which you will only ever need enter once.
COPD and spirometry
The indicators COPD 2, 3 and 6 are about spirometry. Indicators 2 and 3 include reversibility testing and not just a recording of the FEV1/FVC ratio and % predicted FEV1.
If you have assumed that using '339a' (FEV1 before bronchodilation) and '339b' (FEV1 after bronchodilation) is the same thing as recording a reversibility test, then think again.
Not according to the Q&O search it isn't. Again, find them and change them to something like '33G' (spirometry reversibility) or even just '5882' (spirometry).
The other confusion that arises from the spirometry indicators is with COPD 6, which requires that patients with COPD have had an FEV1 in the previous 27 months.
You would think that recording a spirometry code such as '5882' implies that an FEV1 has been done. Well not according to the Great God Q&O, though it will accept the '3398 & 9' codes for pre- and post-bronchodilator FEV1 that it would not accept for reversibility testing.
Confused? Well you should be, though here there is a sudden flash of common-sense in the coding that is so conspicuously absent elsewhere.
This is that even though the COPD 6 code asks for FEV1 recording, most of the codes that it accepts are for the FEV1/FVC ratio such as '3399' (FEV1/FVC ratio abnormal). This makes far more clinical sense.
However, there is one final slap in the face of sanity when you see that it also accepts the code '3398' (FEV1/FVC ratio normal) which is a little odd as the new contract guidance states that an FEV1/FVC ratio of less than 70 per cent is 'required' for a subject to be included on the COPD register.
There is an enduring myth that there must be a record of smoking status for 'never smokers' after the last entry of a Q&O diagnosis like CHD or diabetes. This is not the case.
The Q&O search compares the date of the 'never smoker' entry with the earliest occurrence of the relevant diagnostic code, not the last one. So, no need to get in a stew over that one.
Chris Martin is a GP in Laindon, Essex