|
Currently trusts will be considering what has been learnt from the results
of audits arising from the five National Patient Safety Agency alerts
in the Safe Medication Practice Work programme for 2007/08. There have
also been recent reaudits of the safe use of diamorphine and methotrexate.1–3
Robust audits are ideal for gathering good evidence of practice but only
if they are undertaken in a statistically valid manner. Furthermore,
if the audit is not well designed it is a waste of time that could better
be devoted to patient care. Discussing the audit with a trained statistician
beforehand will prevent many pitfalls.
It might be more appropriate in future if audits are designed centrally
with professional statistical advice that may not be easily available
to many trusts. Perhaps pilot audits of the guidance might be undertaken
first, to provide an evidence base for recommendations. Early implementer
sites could provide estimates of the costs and future savings attributable
to some of the audits.
Surveys and risk assessments may be more cost-effective undertaken at
a regional or national level rather than duplicating this work in individual
trusts. This is, perhaps, best
illustrated with respect to the NPSA Alert 20 promoting safer use of
injectable medicines. This requires a risk assessment to be undertaken
on the selection, preparation and
administration of injectables in all clinical areas.
It partly involves
assessing the risk of products. Pharmacies in different trusts have worked
together to produce an agreed
common categorisation of all injections into high, medium and mild risks,4,
5 eliminating the requirement for each trust to develop its
own.
Unlike guidance from the National Institute for Health and Clinical Excellence,
NPSA audits do not come with a costing template. Trusts will have done
what they can from existing resources, but without a common set of tools,
they will differ in their approach, preventing pooling of data and duplicating
the work across the nation. Perhaps a more serious consequence may be
the temptation to cut corners and undertake a small survey, which would
not produce reliable results.
When undertaking audits, it may seem appropriate that no breaches of
practice should be found that threaten patient safety. This might be
achievable if risks can be eliminated, such as dispensaries not holding
stocks of methotrexate 10mg — stocking only 2.5mg tablets would
eliminate potentially fatal errors caused by selecting the wrong strength.
Auditing
such compliance, however, presents difficulties. When a deviation of
less than 5 is expected, no matter what the sample size, it is difficult
to design a valid audit. In these cases the sample size has to be 100
per cent or the advice of a statistician is essential.
Most audits are concerned with measuring good practice and 100 per
cent compliance is unrealistic. Some idea of the expected result is
needed
before designing an audit. In many cases an idea of the results can
be gained from the literature. For example, Cousins et al identified
risks associated with intravenous drug preparation in a multicentre
European audit.6
They observed 824 doses prepared and 798 administrations. The
most common error observed was administering doses at the wrong rate
(48 per cent in UK hospitals and 40 per cent overall). This error rate
may be above what would be wished for at my trust, but we must check
how we comply before altering procedures and reauditing.
It might,
therefore, be unrealistic to use a desired standard for the first audit.
Where there
are no data in the literature a proportion of 50 per cent should
be assumed because this will require the largest sample size. Precision
The precision required in the results must also be decided. The smaller
the size of the sample audited, the greater is the possible spread
of results and the observed value will be less precise. If a spread
of + or –10 per cent is deemed acceptable, a smaller sample will
be required than if + or –5 per cent is chosen.
An estimate of
how reliable the results are is also needed. Usually 95 per cent probability
levels are chosen, meaning that, on average, in one in a series of
20 audits the true value lies outside the defined range.
Free statistical packages are now available through the internet, giving
worked examples of how to calculate sample size. Unfortunately, the most
user friendly ones are deep within the websites of universities. For
example, sample size calculation D-8.1, from St
George’s University
of London website, offers a more detailed explanation than presented
in this article, but is under the heading of how to apply for grants.
Once the audit is completed we should check the confidence interval chosen
is what is attained. This is achieved by rearranging the formula in the
website used above. Free web-based programs are again available, for
example, CIPROPORTION,
an Excel spreadsheet written by Robert Newcombe.
This allows the spreadsheet data to be overwritten with local data to
yield a confidence interval.
Not only does the sample size have to be sufficient, but also the way
the audit is conducted must be appropriate. For example, to ensure that
high risk injectables are given appropriately, an audit of their use
rather than the use of injectables in all clinical areas is required.
Alert
21 (“Safer practice with epidural injections and infusions”)
suggests a consecutive five-day audit. You may wish to ask why five and
why days, rather than a proportion of the total epidural use. Trusts
may wish to propose audits that examine these aspects as part of next
year’s plan.
Having gained a measure of performance the next step is to consider whether
we should attempt to change practice — there are also risks associated
with any change in an organisation or service. For example, there is
a growing, strong base of evidence that patients are more likely to take
their medicines if they are prescribed as once daily and taken at the
same time each day, rather than several times during the day or on alternate
days.
It may also be preferable to provide whole dosage units, rather
asking patients to halve tablets. The NPSA guidance that 0.5mg warfarin
tablets should be made available should at first sight find support
from the profession. However, some localities have standardised on only
one
strength (3mg) and use both alternate day administration and half tablet
regimens to adjust clotting time.
Where the people are used to this
system, introduction of the full range of strengths could introduce
other risks,
such as confusing 0.5mg with 5mg tablets. As we do not have access
to the evidence on which the recommendation to use all strengths of
warfarin is based, a trust may decide to keep its current system. If
this is
the
conclusion the trust will have to record in its risk register why
it is not adhering to guidance.
This decision should not be founded on
isolated case reports from several years ago, but on sound evidence,
provided
by a robust audit of current practice.
References
1. National Patient Safety Agency. Safe Medication Practice Work programme
for 2007–2008. London: The Agency; 2007.
2. National Patient Safety Agency. Safe Practice Notice ensuring safer
practice with high dose ampoules of morphine and diamorphine. London:
The Agency; 2006.
3. National Patient Safety Agency. Improving compliance with oral methotrexate
guidelines. London: The Agency; 2006.
4. Beaney AM, Black A, Dobson CR, Williamson S, Robinson M. Development
and application of a risk assessment tool to improve safety of patients
receiving injectable medicines. Hospital Pharmacist 2005;12:150 –4
(PDF 90K)
5. Hardy L, Mellor L. Risk
assessment of parenteral product presentation across secondary care acute
trusts in the north of England. Hospital
Pharmacist 2007;14:58–64.
6. Cousins DH, Sabatier B, Begue D, Schmitt C and Hoppe-Tichy T. Medication
errors in intravenous drug preparation and administration: a multicentre
audit in the UK, Germany and France. Quality and Safety in Health Care
2005:14:190–5.
|
The
Pharmaceutical Journal