1 table omitted
Type 2 diabetes is increasing among young American Indians (AIs) and
other populations,1- 4 and
accurate surveillance is important to monitor trends in diabetes prevalence.
The Indian Health Service (IHS) patient database has been used to identify
cases of diabetes and estimate diabetes prevalence among AIs aged ≥15 years.5 However, limited studies have assessed the use
of health databases to ascertain diabetes cases in young persons. The Montana
Department of Public Health and Human Services (MDPHHS), in collaboration
with the Billings Area IHS, conducted a study to assess use of the IHS patient
database to identify AIs aged <20 years with diabetes in Montana and Wyoming.
This report summarizes the results of that study, which found that diabetes
cases were identified more accurately by using at least two patient visits
for diabetes rather than only one patient visit. To reduce misclassification
of diabetes, health-care systems and managed care organizations that use patient
databases for diabetes surveillance should evaluate the accuracy of case ascertainment
periodically and ensure adequate training for staff responsible for coding
health-care visits and database entry.
During 2000-2002, AIs aged <20 years with at least one outpatient
visit or hospitalization coded for diabetes (i.e., using International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM-CM] codes 250.0-250.9) at one of six IHS facilities were
identified from the IHS database. Medical records of each person were reviewed
to confirm the diagnosis and classify the type of diabetes.6 MDPHHS
collected demographic and clinical data and assessed the diagnoses of diabetes.
A case of confirmed diabetes was defined as a case with documented diagnostic
blood glucose values7 or a record of treatment
with antidiabetic therapies (e.g., insulin or oral medication). To assess
the accuracy of case ascertainment, the study compared the percentage of false
positives (i.e., for which persons were determined not to have diabetes) for
cases based on only one health-care visit with the percentage for cases based
on at least two health-care visits during 2000-2002. Diagnostic codes or reason-for-visit
narratives that might have led to case misclassification were identified for
the false-positive cases.
The study identified 93 persons classified with diabetes based on one
coded health-care visit. Assessment of the diagnoses by MDPHHS found that
40 persons (43%) did not have diabetes. No statistically significant differences
by sex or by mean age were found when confirmed cases were compared with false
positives. Wide variation was observed in the proportion of false-positive
cases across the six clinical facilities: 0%, 25%, 27%, 50%, 67%, and 89%.
Among the false-positive cases, the most common reason (15 cases out of 40)
for a health-care visit was diabetes screening or a school health assessment;
for 19 of the cases, no specific reason was identified (Table). On the basis
of the 93 database cases with one coded health-care visit, the prevalence
of diabetes in young AIs was 4.0 per 1,000 population (estimated population
of AIs aged <20 years = 23,035),8 and
2.3 per 1,000 population by using only the 53 confirmed cases.
On the basis of two health-care visits, the study identified 61 persons
classified with diabetes; 12 (20%) were false positives. Once again, no statistically
significant differences by sex or by mean age were found. Of the 12 persons
with false-positive cases, seven had been referred for a health-care visit
through diabetes screening or a school health assessment. On the basis of
the 61 database cases with at least two coded visits, the prevalence of diabetes
in young AIs was 2.9 per 1,000 population, and 2.1 per 1,000 population by
using only the 49 confirmed cases.
KR Moore, MD, Billings Area Indian Health Svc, Billings; TS Harwell,
MPH, JM McDowall, CS Oser, MPH, SD Helgerson, MD, D Gohdes, MD, Montana Dept
of Public Health and Human Svcs. NR Burrows, MPH, Div of Diabetes Translation,
National Center for Chronic Disease Prevention and Health Promotion, CDC.
Accurate surveillance of type 1 and type 2 diabetes in young persons
is important to monitor trends in prevalence and incidence. The findings in
this report suggest that using only one ICD-9-CM coded visit during a 3-year
period to ascertain diabetes cases among young AIs was accurate in only 57%
of cases; therefore, the number of cases was probably overestimated by approximately
40%. The use of at least two ICD-9-CM coded visits for case ascertainment
was substantially more accurate (80%). Because of the low national prevalence
of diabetes in young AIs (less than five cases per 1,000 persons),3 an increase in false-positive cases has little
effect on the estimated rates; however, the number of affected young persons
will be overestimated.
The findings in this report are subject to at least one limitation.
This analysis included only six IHS facilities. The accuracy of case ascertainment
in other IHS areas and facilities might vary by facility and by the prevalence
of disease in young persons.
Patient databases have been useful for monitoring diabetes care in adults
and can be helpful in monitoring diabetes prevalence in adolescents.5,9,10 This report
highlights the importance of evaluating the use of patient databases for ascertaining
diabetes cases among young persons and emphasizes the need to update and maintain
case registries based on patient databases. In addition, adequate training
of staff responsible for coding and database entry of patient diagnoses, particularly
related to diabetes screening and school health assessments, probably will
reduce misclassification of diabetes in young persons.
Diabetes Among Young American Indians—Montana and Wyoming, 2000-2002. JAMA. 2004;291(5):555-556. doi:10.1001/jama.291.5.555