Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

VA Health Systems Research

Go to the VA ORD website
Go to the QUERI website

SHP 08-144 – HSR Study

SHP 08-144
Accuracy of Initial Diagnosis of Diabetes in VHA
Lawrence S Phillips, MD
Atlanta VA Medical and Rehab Center, Decatur, GA
Decatur, GA
Funding Period: April 2008 - September 2008
Early recognition and appropriate management are the keys to longterm glycemic control - limiting the inflammation and oxidative stress which adversely impact beta cell mass and function, and lead to a rise in glucose levels over time; individuals with new-onset diabetes are candidates for intensive control since they are largely free of microvascular disease and major comorbidity. However, the VA has no programs for systematic detection and early management, and our preliminary data showed that the disease is recognized late and not managed aggressively. All program development and evaluation, including assessment of potential disparities in care, must begin with diagnosis, but the initial diagnosis of diabetes in the VA had never been validated - the focus of this proposal.

We tested the hypothesis that the first outpatient use of the diabetes ICD-9 code 250.xx by a VA primary care provider would be accurate in (i) reflecting diagnostic antecedent hyperglycemia, and (ii) predicting subsequent prescription of diabetes-specific medications or high A1c. Our goal was to show that the addition of diabetes to problem lists by primary care providers at outpatient visits would be sufficiently accurate to permit use in real time as both (a) a signal to prompt initiation of standardized care paradigms aimed at control of both glycemia and risk factors for cardiovascular disease, and (b) a marker to permit evaluation of care provided during a critical period in the natural history of the disease. If our project succeeded as expected, future work was planned to include proposals aimed at both objectives. Since better control helps preserve beta-cell function and reduces cost, the end results should be lower longterm A1c, complications, and costs.

AIM #1. ACCURACY IN DIAGNOSIS: To test diagnostic accuracy, we used abstraction of text in Atlanta VAMC CPRS records by diabetes-trained reviewers, allowing analysis to determine whether use of the diabetes diagnostic code 250.xx is justified by antecedent hyperglycemia. Comparison of such initial primary care diagnosis "case" patients (randomly selected from the Corporate Data Warehouse) with randomly selected "control" patients - who are similar in age, sex, BMI, and race but lack use of the code - and with patients who met VA Diabetes Epidemiology Cohort criteria (DEpiC - any use of the 250.xx code twice, and/or prescription of a diabetes medication) allowed determination of positive and negative predictive value, and ascertainment of the proportion of code-(+) patients in the population of VA patients with diabetes allowed estimation of sensitivity and specificity using Bayes' rule. AIM #2. ACCURACY IN PREDICTION: To test predictive accuracy, we utilized VISN 7 Corporate Data Warehouse deidentified data to determine the extent to which initial use of the diagnostic code was associated with subsequent prescription of diabetes-specific medications or high A1c, with both tests for proportions and Kaplan-Meier analysis. This also provided preliminary data for subsequent national VA database studies which can determine the extent to which the level of predictive accuracy found in VISN 7 is reliably generalizable to other regions.

"Case" veterans with initial primary care use of 250.xx (index date) were compared to "controls" without use of 250.xx (matched for age, BMI, gender, race/ethnicity, and index date), and veterans meeting VA Diabetes Epidemiology Cohort criteria (any use of 250.xx twice and/or diabetes drug Rx, "DEpiCs"). We conducted CPRS chart review of randomly selected Atlanta VAMC "case", control, and DEpiC pts (n = 100 each) who had follow-up > = 2 yr before and > = 3 yr after the index date, and examined comparable groups in the VISN 7 Corporate Data Warehouse database.
Patients averaged age 63 years and had BMI 30 kg/m2, and were 2% female, 32% white, and 21% Black (race and/or ethnicity were unknown for many subjects). In the Atlanta VAMC, diagnostic accuracy based only on computerized analyses of glucose and A1c levels was correct for 294 out of 300 "case", "DEpiC", and control patients (discrepancies mainly involved issues such as diabetes being diagnosed outside the VA, etc.). In the Atlanta VAMC, diabetes diagnostic criteria (two or more of fasting plasma glucose >=126 mg/dl, random plasma glucose >=200 mg/dl, or A1c >=6.5%, or any value twice) were fulfilled in 74% of "case" patients vs. 1% of controls (p < 0.001), and "case" diagnostic accuracy was noninferior to that of DEpiC (76%, p = 0.9). Based on Bayes' rule, initial primary care use of the 250.xx diabetes ICD-9 code provided sensitivity 86%, specificity 98%, positive predictive value 74%, and negative predictive value 99%.
In Kaplan-Meier analyses of 3,081 "case" patients and 13,407 matched controls across VISN 7, initial primary care use of the 250.xx code was followed within 3 yr by A1c > 7.0% or prescription of a diabetes medication in 74% of "case" patients vs. 2% of controls (p < 0.001). In VISN 7, "case" predictive accuracy was lower than that in 10,270 DEpiC pts (91%, p < 0.001). However, the period between fulfillment of diabetes diagnostic criteria and initial "case" primary care diagnosis averaged 25 months across 8 VISN 7 facilities (range 21-32 months), whereas DEpiC criteria were met after an average of 35 months (range 31-38 months, p < 0.001).

Diabetes is a VA health problem of epidemic proportions - a major challenge to both quality of care and financial stability. Many veterans have diabetes control which is suboptimal, even in patients who lack mitigating comorbidities. Diabetes control is easiest to achieve and most cost-effective early in the natural history of the disease, but we have no well-validated basis to act upon recognition of the disease by providers. Our study targeted this problem, and our findings should enable both evaluation of care, and development of programs to facilitate management during this critical period. The work involved the priority areas of Complex, Chronic Condition Care, Equity and Health Disparities, and Implementation and Management Research; systematic management of early diabetes should be efficacious and cost-effective - improving the health of individual veterans, reducing health care resource use and costs, and helping to spare VA funds for management of other disorders.
The findings show that the initial primary care diagnosis of diabetes could be used programmatically to trigger CPRS reminders aimed to facilitate management. However, since the diagnosis is substantially delayed and the interval prior to diagnosis varies considerably across VISN 7 facilities, systematic screening and provider education may need to be implemented to optimize early recognition of diabetes in the VA.

External Links for this Project

Dimensions for VA

Dimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.

Learn more about Dimensions for VA.

VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address.
    Search Dimensions for this project


Journal Articles

  1. Fraser LA, Twombly J, Zhu M, Long Q, Hanfelt JJ, Narayan KM, Wilson PW, Phillips LS. Delay in diagnosis of diabetes is not the patient's fault. Diabetes Care. 2010 Jan 1; 33(1):e10. [view]
  2. Phillips LS, Ziemer DC, Kolm P, Weintraub WS, Vaccarino V, Rhee MK, Chatterjee R, Narayan KM, Koch DD. Glucose challenge test screening for prediabetes and undiagnosed diabetes. Diabetologia. 2009 Sep 1; 52(9):1798-807. [view]

DRA: Health Systems
DRE: Epidemiology, Diagnosis
Keywords: Diabetes, Informatics
MeSH Terms: none

Questions about the HSR website? Email the Web Team

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.