News / News analysis: United front on data quality

The NHS Information Centre and Department of Health chose a series of football grounds as venues for its recent series of HRG4 road shows. And delegates could not have escaped the central message – delivered in terrace chant style – from all the main speakers. Data quality is the key issue that needs to be addressed as the NHS develops payment by results.

Of course, data quality should always have been an issue but it has undeniably been thrown into the spotlight in recent years in England as payment by results has taken hold. In the days of block contracts and even crude cost and volume arrangements, counting the right activity and getting the exact coding on a patient might have had very little influence on the money paid or received.

But now the right code or an additional code could have a dramatic impact on the level of payment – even making the difference between being paid the right amount and getting paid at all.

Until recently there have been few incentives to invest in costing and improve accuracy. Now accurate costing is vital, both in terms of giving organisations an understanding of their real cost drivers –whether at service line or patient level – and in terms of providing the data for national tariff prices.

Virginia Jordan (pictured above), head of standards and classification at the NHS IC started the road show chant. While heralding new healthcare resource groups HRG4 as a major step forward, providing an opportunity to more accurately differentiate between routine and more involved care, she acknowledged that it was ‘a more complex product’ than its predecessor (version 3.5). But she said the importance of accurate coding was now ‘non-negotiable’.

Group programme director for NHS Connecting for Health Carol Clarke added her voice, insisting that ‘the benefits [of HRG4] will only be derived if we get better data quality’.

The refrain was picked up by Sarah Butler, deputy director of NHS finance at the Department. She pointed out that quality of data had affected the design of the tariff in 2009/10. The most obvious examples are accident and emergency services, which remain on version 3.5, and outpatient procedures, for which the tariff is non-mandatory. ‘The key going forward is to realise that the benefits of HRG4 are reliant on getting good data in,’ she said. 

In addition to whole areas where data quality was a concern, such as A&E, the initial tariff for 2009/10 revealed more specific indictors of underlying data issues. For instance, some interventions for patients without complications or co-morbidities were initially priced higher than those with. Reference cost audits are being seriously considered.

Tougher grouper

One of the reasons why data and coding quality become even more important this year (2009/10) is that HRG4 has a more robust grouper far less tolerant of missing or inaccurate data. Version 3.5 grouped activity first, then checked for all the necessary fields. Even if data was missing, activity was grouped. In effect this meant there was a tiny chance of the activity receiving a U-code – which translates into a zero payment for the trust involved.

Under HRG4, the grouper checks all data items are valid (not just present) – including age, discharge date, treatment function codes and procedure/diagnosis codes – before processing further.

‘HRG4 initially validates the data and if it is not valid it assigns a U-code. That means a zero payment,’ Paula Monteith, principal casemix consultant at the NHS IC, told the road show. ‘And if there are dump codes you use that aren’t valid as a primary diagnosis, they will all U-code. If you are not adhering to national data standards you will have problems.’

Some organisations are not helping themselves. From 1 April the mandated version of the procedure codes used in England became OPCS 4.5. But even if NHS bodies have not kept up with coding updates, version 4.3 – implemented in April 2006 – is essential for operating HRG4. Without this coding system, much of the additional granularity of HRG4 will be off limits. ‘It would appear that some of you aren’t using OPCS 4.3,’ said Ms Monteith. ‘You won’t be able to unbundle and there are some 70 core HRGs that you simply can’t reach.’

Trusts could be using the correct codes. But if their clinicians are not providing enough detail on patient notes for coders to identify the specific codes, then patients will end up in a catch-all HRG that will not attract the right funding.

The good news for NHS bodies is that the local payment HRG output file provides detailed information as to why U-codes have been assigned. This pinpoints whether the activity couldn’t be grouped because of invalid diagnosis, poorly coded primary diagnosis, age conflicting with diagnosis, diagnosis conflicting with anatomical sites, invalid procedures or poorly coded procedures. This should give organisations an opportunity to fix any process or individual errors being made in the coding system.

It is not just about money – if organisations are not coding properly how can they know what they are doing and plan ahead. But the financial implications of getting it right were spelt out by the Audit Commission in a workshop at the roadshow.

Howard Davis, payment by results benchmarking manager at the commission, gave examples from the commission’s coding audit programme of errors that could change payment from £600 to £6,000. He highlighted that the impact of coding errors could be as much down to the trim points, excess bed days and business rules associated with different tariffs as it was to the tariff itself (see examples above).

The commission’s findings from its 2008/09 coding audit are yet to be published. But early analysis of the first three quarters, shared with the roadshow, shows an overall similar distribution of HRG errors. Outliers have been reduced and there are improvements within specialties (for instance, a 24% improvement in the error rate for rheumatology activity, although this specialty remains the biggest coding challenge).

In 2007/08 the average error rate from the audit sample was 9.4%, contributing to a gross financial error of about £3.5m or 5% of the price of the sample reviewed. In most cases the net financial impact – adding the gains and losses from the different errors – was close to zero. Again, data from the first three months of 2008/09 shows a similar picture.

But the fact that errors often have only limited impact on the overall income of a trust does not mean the errors are unimportant. As the commission said in its report on the coding audit last autumn: ‘A large gross financial error, even when the net impact is small, will have profound implications for the accuracy of service line economics.’

Good decisions can only be taken if informed by good data. But HRG4 and the new grouper clearly ramp up the importance of data quality to financial flows. As the IC’s Virginia Jordan told the road shows: ‘The viability of a number of your services may well depend on your knowledge of HRG4.’