Validating PbR
The Audit Commission’s report on the pilot study to validate PbR is something I missed earlier.
Between April and August 2006, the Audit Commission piloted the external clinical coding audit and benchmarking components of the PbR assurance framework in two health economies – South Yorkshire and Avon, Gloucestershire and Wiltshire (AGW).
Key findings
The pilots show a relatively high level of clinical coding error, which is leading to inaccurate payments under PbR. The HRG error rate, averaged across the 12 participating acute trusts, is 11.9 per cent. There is also considerable variation across trusts, ranging from an overall Healthcare Resource Group (HRG) error rate of 3.5 per cent to 28 per cent, and across specialities. Diagnosis coding has a higher error rate than procedure coding and the error is more likely to impact on payments. There is greater and more material error among non-elective episodes.
With error rates of nearly 12% in the limited sample above, is this system fit for purpose?
The coding error has resulted in both overpayment and underpayment to trusts for individual episodes – underpayment for seven trusts and overpayment for five trusts. The absolute impact of the error on individual trust payments ranges from £7,091 to £67,698 for the sampled 200 episodes, representing between 5 per cent and 14 per cent of the total sample value. Although it is difficult to extrapolate the impact on accuracy of payments to a trust overall with any confidence, we can say that clinical coding error is clearly impacting negatively on payment accuracy and that mechanisms should be in place to address this.
I am not aware of too many organisations that can stomach this level of payment instability without going out of business.
Both coder and non-coder errors are contributing to the high error rate. Quality of clinical documentation and clinician involvement are some of the non-coder issues that are contributing to the error rate; while low numbers of accredited coders, insufficient training, and insufficient guidance on coding practice locally (for example, absence of a complete and up-to-date policy and procedure document) are the main causes of coder error. Error rates are noticeably higher where there is a low proportion of accredited coders or a poor coding process (for example, coders are not coding from case notes).
There still isn’t an easy way to get detailed coding information to be completed by the clinicians & there will not be until there is full-scale computerised record keeping. The use of various smaller consumables is not recorded & is only estimated. Most coding is done away from the coal-face by non clinical staff who have to infer details from hospital records not meant for this purpose.
There is evidence of overcoding (recording more diagnosis/procedure codes than is necessary) in some trusts, but in some cases this has been a historic practice, and without access to prior audit reports it is difficult to attribute to the financial incentives under PbR. There was also evidence of trusts actively working to optimise their coding to maximise income, but within existing coding rules (that is, the practice is not fraudulent). There is some concerning practice emerging, for example, where the clinical coding team reports to the director of finance, and is involved in discussions about the financial impact of coding changes on a regular basis. This increases the likelihood of manipulation and it will be important to have a protocol in place to refer cases of suspected fraud to the NHS Counter Fraud and Security Management Service (CFSMS).
There is some gaming of the system but not in a fraudulent manner.
Coding auditors have also raised concerns about the appropriateness of admissions in some trusts, where it appeared from the case notes that patients should have been treated as outpatients and there was no obvious reason for admission. This is a big issue in both pilots. However, current data definitions are ambiguous and their application is subjective, leading to different local practices in the treatment of day cases/outpatients – until this is resolved it will be difficult to address this area.
Seems like a value judgement & second-guessing of clinical decisions to me, I wonder whether there were any clinicians actually on this audit team?
While there is no agreed standard or target for clinical coding error in the NHS, the reported level of HRG error seems intuitively too high, particularly in a system where clinical coding is the primary determinant of payment. The level and nature of clinical coding error is not so high as to destabilise the PbR regime, but it does undermine it, raising concerns about the accuracy and fairness of funding flows. In addition, it has major data quality, epidemiological and clinical implications.
I agree, though I am more concerned about the stability of the system, especially given the other pressures acting on the NHS. It will be interesting to see whether this has filtered through to the decision makers & if PbR still goes ahead as planned on the current timetables.
The pilots demonstrate that there is a need for a centrally co-ordinated external clinical coding audit programme, which ensures objectivity and independence of results. Even where there are strong local monitoring arrangements in place at commissioner and provider level, there is a need for an external and independent audit to supplement and provide additional impetus to improve data quality. Both pilot sites have commented on the value of the audits in highlighting data quality and payment issues from an independent and objective perspective, and in opening dialogue about this between commissioners and providers and across the health economy.
An audit programme will help, but accurate coding capture can only happen at the point of treatment, not after the fact & certainly not at time and distance removed. If we are to go down this road, a well-designed EPR system that facilitates clinical participation is a must & is one we are sorely lacking.
In addition to the local benefits, a comprehensive audit programme has an obvious national benefit in improving the quality of data in the NHS and refining the policy framework (for example, the audits have highlighted inconsistencies in coding rules and/or the HRG grouping algorithm). There is also a benefit for patients. It is the first time that this kind of information has been made available, and various stake-holders have commented on its value and the opportunity to share lessons learnt and good practice.
Yes, but given that the report goes on to describe a shortage of accredited clinical coding auditors even to complete a pilot study which meant that the number of cases reviewed had to be cut & does not make any mention whatsoever of having clinicians on hand to interpret the case notes suggests that there is an awful lot more to do before any of this can be trusted.
April 20th, 2007 at 9:30 am
[...] Something that I have been saying for a while. It is not just the safety data however but also the rest of the data infrastructure. It is not a solid foundation to rely on for the development of an internal NHS economy. [...]