Author

Data Source

Diagnosis/Treatment Codes and Algorithms

Study Cohort

Validation Methodology

Comments

Blanchard et al. (1996)

Country: Canada

Source: Physician billing claims and hospital separation records from Manitoba Health

Fiscal years: 1986-1991

Codes: ICD-9-CM 250

Algorithms: at least two separate physician claims for diabetes within 2 years of each other or at least one hospital separation record with a diagnosis of diabetes.

25 years or older

Diabetes Education Resource (DER) database which includes all contacts with DER program clients, including clinical, service-related, and demographic information.

Specificity, sensitivity and predictive value are not reported.

Ascert. rate > 95%.

Morgan et al. (2000)

Country: Wales , UK

Source: Inpatient dataset (1991-1997), outpatient dataset (1991-1996), diabetes clinic dataset (1993 to present), Office of National Statistics mortality dataset (1993-1997)

Codes: ICD-9 250

ICD-10 E10-E14

Algorithms: An inpatient diagnosis of diabetes, attendance at an outpatient clinic coded as diabetic, inclusion on the diabetic clinic dataset or cause of death coded as diabetes on the ONS mortality dataset.

Age of cohort was not specified.

A general practice audit database.

Specificity, sensitivity and predictive value are not calculated.

“This study combines primary and secondary care data sources to estimate the prevalence of diagnosed diabetes…” p.143

Martens et al. (2002)

Country: Canada

Source: Physician billing claims and hospital separation records from Manitoba Health

Years: 1997

Codes: ICD-9 250

Algorithms: One or more hospitalizations or two or more physician claims with a diabetes diagnosis in a three-year period

20 to 79 years of age

Manitoba First Nations Regional Health Survey (1998)

Sens = 76.0%

Spec =

Wilson et al. (2001)

Country: USA

Sources: Indian Health Service patient registration databases

Years: Does not provide

Codes: ICD-9 250.00-250.93

Algorithm: Four sets of criteria: 1) at least one ICD-9 code, 2) at least two separate ICD-9 codes, 3) a pharmacy prescription entry for sulfonylurea, metformin, acarose, thiazolidindione, or insulin, 4) at least two separate glucose values ≥200 mg/dl.

15 years of age and older

American Indian and Alaskan Native people

Medical chart review

Max. Sen = 92%

Max. Spec = 99

PPV = 95%

“The specificity of a single 250.00 to 250.93 ICD-9 code was nearly the same as the use of two 250.00 to 250.93 ICD-9 codes. The use of two 250.00 to 250.93 ICD-9 codes resulted in significant loss of sensitivity.” p.48-49

Hux et al. (2002)

Country: Canada

Source: CIHI discharge abstracts and Ontario Health Insurance Plan for physician service claims

Fiscal years: 1991-1999

Codes: ICD-8 or ICD-9-CM 250.x

Algorithm: One hospital separation or 1 or 2 physician service claims in a two-year period. Examined all 16 diagnosis fields in hospital data.

Age of cohort was not specified.

Ontario Drug Benefit Program database (for individuals 65+ years).

National Population Health survey (National Population Health Survey; NPHS)

Physician office charts

Max Sens = 94%

Max PPV = 98%

Wilchesky M et al. (2004)

Country: Canada

Source: Medical claims from Quebec

Year: 1995-1996

Codes: ICD-9-CM 250.0-250.9

Definitions: All medical claims with a relevant diagnosis

66 years of age and older

Medical charts

Max Sens = 64%

Max Spec = 98%

Borzecki et al. (2004)

Country: USA

Source: Out-patient clinic (OPC) file for 1998 and 1999, and patient treatment file (1999), from a National Department of Veterans Affairs electronic database.

Codes: ICD-9 250

Algorithms: Varied the minimum required number of claims with a given OPC diagnosis from 1 to 2 and varied the number of years of data from 1 to 2.

Outpatients receiving primary care at 10 different sites across the country.

Age of cohort was not specified.

Electronic clinicians’ notes (i.e., medical outpatient charts)

Max Sens = 97%

Max Spec = 96%

Max Kappa = 0.92

Rector et al. (2004)

Country: USA

Source: Medicare+Choice health plan claims data: Physician, hospital, and pharmaceutical claims

Years: 1999, 2000

Codes: ICD-9-CM

250.xx, 357.1x, 352.0x, 355.41 in physician claims in one of up to four diagnosis fields and hospital claims in one of up to nine diagnosis fields.

National Drug Codes were not specified for the pharmaceutical claims.

Algorithms: 38 different algorithms were examined.

Age of cohort was not specified.

Survey data collected from health plan members

Max. Sens: 95%

Max. Spec = 100%

“Diabetes was the only condition where an algorithm had a specificity and sensitivity greater than 0.90.” p.1852