For the last few years we have operated a dual coding convention to facilitate the process of creating and maintaining CHD and Diabetes registers. The beauty of this is that it provides rapid identification of erroneously or un-coded patient records and a ratchet mechanism for building up a register over time. Practices that have been using it have shown steady improvement in their disease registers - a good preparation for GMS2. It has also meant that we can be confident in Leicestershire that these registers are of high quality, along with the audit data that flows from them.
Why bother with housekeeping, if it's no longer a requirement for audit?
Won't all the dual coding we've done have been a waste of time?
Under GMS2 practices will need to maintain a number of registers simultaneously using the codes that have been defined for the purpose. Extending a dual coding system to all these registers involves a lot of work for practices and the Read code hierarchy is not conducive to this for all disease areas. So we have been looking for an easier way. There is little doubt that some form of housekeeping checks on the accuracy of the register will be necessary; indeed practices are expected to have a system for doing so under GMS2. The challenge is to do this without dual coding.
We need to ensure that the computer record accurately reflects the current status of the patient vis a vis the disease in question.
There are two kinds of error that can arise in disease registers, each with 2 subtypes
1) Patient currently has the disease but the codes in record do not indicate it (false negative) because
a) The codes indicating the condition have never been entered
OR
b) The codes indicating the condition have been entered but other codes indicate that currently the condition has resolved.
2) Patient currently does not have the disease but the codes in the record indicate they do (false positive) because
a) The patient never had the disease but the codes were entered in error.
OR
b) The patient did have the disease and the codes were correctly entered but the record has not been updated to indicate that the condition has resolved.
Up until now we coped with 1a and 2a with dual coding and 1b and 2b with mutually exclusive coding (ie the latest of two mutually exclusive code groups eg asthma and asthma resolved indicates the current status). Provision for the latter exists under GMS2, but unfortunately it does not make any useful suggestions about 1a & 2a. 1a is easy but dual coding was invented to cope with 2a. The problem is how to distinguish correctly coded records from erroneously coded ones when the codes in them are the same! With dual coding it was easy because dual coded entries are much less likely to arise in error and so we only had to search for records where the dual coding convention was not met. This also got round a tricky problem in EMIS where some practices were having trouble completely removing erroneous codes which meant that these patients kept getting picked up in searches they weren't supposed to.
The trick is to look at the pattern and temporal distribution of codes in the records. The new housekeeping searches take all of the diagnostic codes along with the related codes for a condition and generate a score for that record. The greater the diversity of diagnostic and supporting codes and the better they are spread out over time, the higher the score and the more likely that the coding is correct. The housekeeping spreadsheet then highlights those records with a low score as the ones that need to be checked.
This approach is based on the observation that false positive identifications tend to arise as isolated coded entries unsupported by corroborative codes or result from single episodes of care where a wrong diagnosis was made and entered, which because it was wrong and the patient recovered, did not generate ongoing coded entries. (Without having used dual coding it would have been hard to make this observation). I have named this approach 'rich coding', because we are relying on the richness or otherwise of records to spot potentially wrong entries.
There are several reasons why it is in practices' interests to housekeep their registers.
1) GMS2 requires that the practice 'has systems in place to maintain a high quality register.' PCOs will be expected to check this as part of the verification process. We anticipate that demonstrating that housekeeping is occurring will meet this requirement.
2) Both kinds of error in registers above will result in a reduction in the possible earning capacity through the quality and outcomes framework. False positive registrations are unlikely to have had all the criteria met and false negatives do not even count even if they have had all the items of care delivered. High quality registers should therefore be good for practice cash flow.
3) If the framework functions as intended this should also be good for patients, ensuring the correct patients are called and their care delivered.
4) High quality registers should reduce medico-legal risk incurred by inaccurate data. False positive code entries may find their way into electronically generated reports for insurance companies resulting in inappropriately loaded premiums or being turned down for cover. A patient who was not called for care because they were not put on a disease register and suffered harm might claim that the practice failed in its duty of care under the new contract to maintain an accurate disease register.
5) The NHS generally will benefit from high quality data as information is shared more widely electronically in the NHS.
6) PCTs will be better able to commission care for their populations if they have reliable information about disease prevalence.
Practices that are already doing housekeeping will notice very little difference because all the analysis occurs 'under the bonnet'. You download the Miquest sets and excel spreadsheets to analyse the data exactly as before. The spreadsheets have been simplified a little to reduce the number of worksheets in them, hopefully making them a bit easier to use. Each sheet contains instructions on what to do. The searches and analysis take slightly longer because housekeeping with rich coding is a more complex process.
After analysis each sheet contains lists of patients whose read coding vis a vis the disease register in question has raised some doubts about its accuracy, with information about why that patients record has been flagged for checking. Practices need to process these lists, which can be sorted and distributed by usual doctor, which makes the job easier as the usual doctor will often be able to answer the query without referring to the notes.
To establish a register it's a good idea to run the housekeeping every few months while simultaneously ensuring your team is set up for the job of ongoing data capture. For new registers practices should concentrate on the possible false negative registrations and only worry about false positives once the register is established. For maintenance, once or twice a year should be adequate depending on how good your team is at coding new patients / diagnoses. It would be best to do it well before National Prevalence Day (14 February):
in case you identify significant numbers of new cases and need time to deliver all the care by the 31st March (QOF National Achievement Day) deadline.
Just carry on using the templates already set up with the diagnostic codes specified for GMS2. You don't have to worry about dual coding each patient any more. Using codes such as Ischaemic Heart disease as problem titles for follow-ups not only links those consultations together, making review of the condition easier but also contributes to the richness of coding, confirming the diagnosis and thus aiding the housekeeping process.
Not at all!
1) By using it you will already have weeded out all the erroneous codes from your registers - good preparation for GMS2.
2) Rich coding will still recognise the coding you've done and not flag up these patients again.
3) Dual coding is not just about registers but also data quality. Coding for CHD and diabetes will have been improved by identifying MI patients, Type 1 and 2 diabetes etc.
Yes - Dual coding is really a special case of rich coding and so is compatible. The new system has been designed so that in practices using dual coding, the dually coded patients will score at least 6 points. Practices should therefore concentrate efforts on the false positive possibles whose score is under 6 initially and then expand their efforts as time & experience allow. It is likely that different practices will find that different scores represent their particular cut off above which the yield in genuine false positive registrations is too low to be worth pursuing. The audits will run whether you have dual coded or not. Practices that do not use their computer systems much will find dual coding more useful because their databases are unlikely to be rich enough to allow good discrimination between true and false false positives.
The PCAG templates still support dual coding where this is established. This is useful for another reason too. In following up patients with IHD or cerebrovascular disease, it is a good idea to use the high level code eg G3 or G6 (which we hitherto have used as one of the dual codes) as the problem title. In GMS2 it is particularly important not to inadvertently add new episodes of MI or stroke by using these codes as follow-up problem titles with wrong episode types, otherwise the flag to scan / refer for exercise testing will be triggered inappropriately. The use of high-level codes avoids this.
Dual coding has been particularly useful in such practices and will out perform rich coding in practices with less rich clinical databases. We anticipate that with GMS2 use of computer systems to record chronic care is likely to increase substantially and so this should become less of a problem over time. Practices that do not use their computer systems much will find dual coding more useful for now because their databases are unlikely to be rich enough to allow good discrimination between true and false false positives.
Indeed, because no system can. If it is possible to code accurately it logically follows that it will always be possible to code inaccurately, whatever system you use. Housekeeping is about ensuring that registers will remain within certain ranges of accuracy by preventing the accrual of errors. Our research on CHD registers in Leicester has shown that regular housekeeping is an effective, sustainable way of producing accurate registers. Clinical teams need to be aware that errors may exist and that they should correct them opportunistically too.
The main advantage of the new system is that there is no need to dual code patients' records, resulting in a time saving and removing certain training and template needs
Rich coding also has an advantage over dual coding in that it tends to flag up fewer possible false positives, providing the database is rich enough. This is because a patient only coded with G33 for angina say, but coded repeatedly over time, with other supporting codes is very likely to be a true positive and will score highly on the rich coding system, but will have been flagged up inappropriately by dual coding.
The exact algorithm for scoring the potential false positives is based on several years of experience in handling housekeeping at PCAG and trials in one practice. We anticipate that it may need tweaking to improve its accuracy in identifying real false positives and negatives. If practices come up against certain code patterns in patient records that have resulted in false positives not being picked up we would be grateful for feedback on the details so that we can sharpen this tool