Miscoding, misclassification and misdiagnosis of diabetes in primarycare
To determine the effectiveness of self-audit tools designed to detect
miscoding, misclassification and misdiagnosis of diabetes in primary
care.We developed six searches to identify people with diabetes with
potential classification errors. The search results were automatically
ranked from most to least likely to have an underlying problem. Eight
practices with a combined population of 72,000 and diabetes prevalence
2.9\% (n?=?2340) completed audit forms to verify whether additional
information within the patients' medical record confirmed or refuted
the problems identified.The searches identified 347 records, mean
42 per practice. Pre-audit 20\% (n?=?69) had Type?1 diabetes, 70\%
(n?=?241) had Type?2 diabetes, 9\% (n?=?30) had vague codes that
were hard to classify, 2\% (n?=?6) were not coded and one person
was labelled as having gestational diabetes. Of records, 39.2\% (n?=?136)
had important errors: 10\% (n?=?35) had coding errors; 12.1\% (42)
were misclassified; and 17.0\% (59) misdiagnosed as having diabetes.
Thirty-two per cent (n?=?22) of people with Type?2 diabetes (n?=?69)
were misclassified as having Type?1 diabetes; 20\% (n?=?48) of people
with Type?2 diabetes (n?=?241) did not have diabetes; of the 30 patients
with vague diagnostic terms, 50\% had Type?2 diabetes, 20\% had Type?1
diabetes and 20\% did not have diabetes. Examples of misdiagnosis
were found in all practices, misclassification in seven and miscoding
in six.Volunteer practices successfully used these self-audit tools.
Approximately 40\% of patients identified by computer searches (5.8\%
of people with diabetes) had errors; misdiagnosis is commonest, misclassification
may affect treatment options and miscoding in omission from disease
registers and the potential for reduced quality of care.
Published in 2012.