Measuring accuracy
Each month, NHS Prescription Services (NHS PS) takes a random sample of prescription items it has priced and reimbursed contractors for in that month and reprocesses them to check for any errors.
Up until May 2010, NHS PS reprocessed a small number (between seven and 12) of complete contractor bundles per month, which led to around 50,000 items being checked. Then, in June 2010, the bureau changed its sampling method and began selecting around 10,000 prescriptions items for reprocessing from a range of contractor accounts each month.
The move followed a review by the NHS PS statistician "to ensure that the correct methodology is used to produce the most accurate and robust results", NHS PS said.
The batches being reprocessed are checked by NHS PS staff against scanned images of the prescriptions that are taken by the CIP automated payment system. For each of these monthly samples, NHS PS can compare how it originally priced the prescriptions (and reimbursed contractors) with how the prescriptions should have been priced according to the accuracy check. It records the number of errors found each month, along with the overpayments and underpayments to contractors that those errors caused.
C+D has obtained copies of the monthly error records and of the underpayment and overpayment figures using the Freedom of Information Act.
The figures are reported on a 12-month rolling basis, which makes them more difficult to analyse in detail. But over the two years since April 2009, NHS PS reported that it found monetary errors (errors that caused a payment inaccuracy) on around 2.7 per cent of the items it checked on average. And in every single month for which there are detailed records, the paymaster reported that its rolling underpayment total was higher than the rolling overpayment total (see the two examples to the right).
Once NHS PS has the under- and overpayment figures, it uses them to calculate the "net cash variance" – the accuracy measure it reports back to the Department of Health, and which should be over 99.8 per cent. To calculate it, NHS PS takes the difference between the over- and underpayment values to work out the payment errors it has made. Because the over- and underpayments can balance each other out, this figure ends up being quite small. So when NHS PS works out what percentage – of the total amount it paid out for all the items it checked – that error value is, accuracy comes out at over 99 per cent (see row I, right). That said, while accuracy has always been well over 99 per cent, NHS PS has missed its 99.8 per cent targets in 10 out of the past 26 months, with accuracy dropping as low as 99.67 per cent in March 2010.
And while these figures represent the accuracy of the total amount that NHS PS has paid out (and therefore that the NHS has paid the sector), they give no indication of the accuracy of contractors' individual payments.
If you instead take the overall value of the errors made (by adding the under- and overpayments made, rather than using them to cancel each other out) then accuracy looks rather worse, as C+D has shown in the examples on the right. Accuracy expressed using this methodology is consistently less than 99 per cent. Over the 15 months it was 98.33 per cent on average, and in March 2010 fell as low as 97.86 per cent (see row L, right).
As C+D reported last week, these figures are only estimates of the accuracy of the total amount paid out by NHS PS every month and do not represent the accuracy of payments to individual contractors. NHS PS has not assessed the accuracy of individual payments.
