Objective. First, to test the validity of self-reported quality of care and
treatment measures compared with medical records and administrative data f
or: eye examinations, hemoglobin A(1C) tests, and use of insulin and oral a
gents for adult patients with diabetes; and secondly to assess the consiste
ncy between medical record information and administrative data for the same
measures plus microalbumin testing.
Design. Cross-sectional study using data from telephone survey primary care
medical and eye records, and administrative claims.
Setting. Statewide health maintenance organization in Minnesota, USA, 1995.
Study participants, Four hundred and forty adults with diabetes, aged 31-64
years.
Main outcome measures. Validity of self-reported diabetes quality of care m
easures compared with a criterion standard combining information from prima
ry care and eve records with information from administrative data; and reli
ability of medical record information compared with administrative data.
Results, Although the sensitivity of self-reported eye examination was high
(89%), the specificity was low (65%). Self-report of hemoglobin A(1C) also
had high sensitivity (99%) and a lower specificity than that of eye examin
ation (28%). The two information sources (medical records and claims) used
in the criterion standard each contained complementary and nonoverlapping i
nformation. Reliability was highest for microalbumin testing (kappa, 0.75)
and lowest for eye examination (kappa, 0.37).
Conclusions. Quality of care measures for diabetes are often drawn from a v
ariety of sources. To the extent that data sources are biased, the measures
can be misleading. Self-report is likely to lead to an overestimate of eye
screening and the measurement of hemoglobin A(1C) Reported rates of qualit
y of care should be inspected carefully. The 'same' rate taken from differe
nt sources may vary.