Audit and statistics
Details of statistical tests will not be
covered here but general comments are required so that statistics are not abused
when performing an audit. Descriptive statistics are useful and acceptable —
distribution of ages, operation type, etc. Many surgeons will quote a ‘mean’
and ‘standard deviation’ for items such as age; this assumes that the data
are normally distributed which is not necessarily the case. It is often better
to use median and range for ages, time after surgery, etc. It is unlikely that
incidence or prevalence can be calculated from an audit as the size of the
overall population (the denominator) will not be known. The use of tests of
significance is not acceptable in audit data. Performing tests of significance
requires a comparison between two groups, one of which represents a control
group. A control group is one which is identical to the test group except in the
one aspect that is being examined. Therefore, tests of probability can only be
used where patients are entered into a prospective randomised clinical trial.
The
presentation of data in a publication is a source of concern in many audit
papers. Often the scenario arises that a surgeon will find that a certain number
of patients underwent a specific procedure some years before. The surgeon will
then try and find the details from records of those patients but for various
reasons it is only possible to find, for example, 75 per cent of the original
records. The surgeon will next try and contact this 75 per cent of patients to
find out the ‘outcome’ of the operation (usually using an invalidated questionnaire or score). Only, for example, 50 per cent of those contacted will
reply to an enquiry (i.e. 50 per cent of 75 per cent of the original number).
Some of the remainder may have died but many simply will not be traceable. In
this process if the original number was 100, the assessment will be undertaken
for only about
37 patients. Many authors will then make
comments about the data that can only be applied to the 37 who were assessed but
the report is written as if it refers to the total number.