Secondary Logo

Journal Logo

Contents: Original Research

Diabetes Screening Reminder for Women With Prior Gestational Diabetes

A Randomized Controlled Trial

Zera, Chloe A. MD, MPH; Bates, David W. MD, MSc; Stuebe, Alison M. MD, MSc; Ecker, Jeffrey L. MD; Seely, Ellen W. MD

Author Information
doi: 10.1097/AOG.0000000000000883
  • Free

Gestational diabetes (GDM) affects 3–8% of pregnancies in the United States.1 A diagnosis of GDM is associated with an increased risk for future type 2 diabetes; as many as 50% of women with GDM will develop diabetes within 5 years of delivery.2 Consequently, the American College of Obstetricians and Gynecologists and the American Diabetes Association recommend that women diagnosed with GDM undergo screening at 6–12 weeks postpartum to detect type 2 diabetes and continued screening for type 2 diabetes at least every 3 years.3,4 Despite these recommendations, screening rates remain suboptimal even in the immediate postpartum period.5–8

Previous interventions to increase diabetes screening have targeted obstetric providers in the postpartum period9,10; however, we are unaware of efforts targeting primary care providers. Health care provider-level barriers to ongoing diabetes screening among women with a history of GDM include lack of knowledge as well as poor communication between obstetricians and primary care providers about pregnancy complications.11 Clinical decision support utilizing the electronic health record (EHR) has shown promise as a system improvement that may improve quality of care.12

In a cluster randomized trial of EHR reminders conducted in our system, reminders improved rates of screening for hyperlipidemia in adults with chronic disease.13 We therefore sought to study whether an EHR reminder in a primary care setting would increase screening for diabetes among women with a history of GDM.


We performed a cluster randomized trial of a reminder within the EHR to evaluate the effect on diabetes screening in women with a history of GDM. Because our intervention targeted health care providers, we chose a cluster randomized design to avoid contamination by colleagues within participating practices. We included primary care sites from within the Partners HealthCare System, a nonprofit network of outpatient and inpatient facilities founded by Brigham and Women's Hospital and Massachusetts General Hospital based in Boston, Massachusetts. All of the participating sites utilize an internally developed EHR called the Longitudinal Medical Record. These 23 clinics included community health centers, hospital-based practices, and off-site practices and have consented to participate in Longitudinal Medical Record-based research projects. The Partners institutional review board approved this study. The trial is registered at (study identifier NCT01288144).

We randomized the sites stratified by primary hospital affiliation (Brigham and Women's Hospital or Massachusetts General Hospital) and practice type (women's health center, community health center, or off-site practice) to balance health care provider and patient characteristics between the intervention and control groups. One site in which health care providers are housed in separate suites was divided into four clusters for randomization, resulting in 26 clusters in total. Within each stratum, we used SAS to generate a random sequence, assigning 13 clusters to the intervention group and 13 to the control group.

The Longitudinal Medical Record allows health care providers to maintain problem, medication, and allergy lists; to view laboratory and radiology results; and to generate medication prescriptions and laboratory orders. The first screen visible when a clinician accesses the electronic record for a patient is called the summary screen and is viewed each time the patient record is opened. The system includes several clinical reminders including cancer screening, chronic disease management, and smoking cessation, which are displayed on the summary screen when indicated. The diabetes screening reminder text read “Patient has a history of gestational diabetes and should be screened for type 2 diabetes.” For clinicians at intervention sites, the reminder was displayed within the summary screen along with any other clinical reminders each time the record was accessed. The diabetes screening reminder was not visible to health care providers practicing at control sites. At intervention sites, a dropdown menu with coded responses was displayed with the reminder. Clicking on the reminder displayed health care provider education about the rationale and recommended screening strategies in women with a history of GDM as well as links to further information online. The reminder content was approved by an interdisciplinary committee of health care providers who use the Longitudinal Medical Record.

Patients and physicians were enrolled on the first occasion during the study period (November 25, 2010, to December 1, 2012) that a 1) health care provider at a participating site opened a patient chart within the Longitudinal Medical Record and 2) a reminder was generated. Women were therefore enrolled in the study at the time their health care provider accessed their chart. Each time that a clinician opened a patient chart within the Longitudinal Medical Record, an algorithm was run to identify women with a history of GDM and exclude women who were recently screened, were less than 3 months postpartum, or had already been diagnosed with diabetes. The algorithm was developed with the input of the Partners electronic medical record clinical content committee and prioritized specificity of GDM diagnosis over sensitivity. The algorithm to identify women with a history of GDM searched all laboratory results as well as the problem list. Females age 12–55 years who had either failed a 100-g oral glucose tolerance test (OGTT) by Carpenter-Coustan criteria or had GDM in the problem list were identified as patients with probable GDM. To exclude women who had already been screened, the algorithm then excluded those who had a 75-g OGTT within the past year. Of note, at this point in time, both Brigham and Women's Hospital and Massachusetts General Hospital use the two-step screening method for GDM and therefore patients receiving prenatal care within either system would not have undergone a 75-g OGTT during pregnancy. The algorithm excluded women less than 3 months postpartum by searching for a weight documented within the obstetric record within the previous 3 months. Finally, it excluded women with a coded diagnosis of diabetes in the problem list. Reminders were generated within the system and logged electronically but were not visible to health care providers outside of the participating sites.

Baseline patient characteristics including age, self-reported race, and insurance status were collected from the EHR. Body mass index (BMI, calculated as weight (kg)/[height (m)]2) was calculated from the last documented height and weight before the index visit at which intervention status was assigned. Health care provider characteristics including age and gender were collected from administrative databases. We used laboratory databases to identify the date, time, and result of the first testing performed. We defined our primary outcome of diabetes screening as performance of a hemoglobin A1C, 2-hour, 75-g OGTT, or fasting glucose during the study period.

We compared patient and health care provider characteristics at enrolled sites using t tests for continuous covariates and χ2 for categorical covariates. We compared unadjusted rates of screening in the intervention group compared with the control group using a χ2. We used a mixed-effects model (PROC GENMOD in the SAS statistical package) to account for clustering of patients within clinical sites. The model was then adjusted for patient age, race, BMI, and insurance status, health care provider age and sex, and whether GDM was included as a coded problem in the electronic problem list. We performed sensitivity analyses to estimate the effect of the reminder within a group of women without demographic risk factors for diabetes by comparing screening rates in the population restricted to normal-weight women and also in women who did not have GDM in the electronic problem list.

We estimated that a sample size of 1,000 participants would be needed to give us 80% power to detect a 15% absolute difference in screening rates between the control and intervention arms when adjusting for an intracluster correlation coefficient of 0.05 and a mean cluster size of 50 patients.14 Based on our actual mean cluster size of 33 patients, a sample size of 850 participants gave us greater than 80% power to detect a 15% absolute difference in screening rates between the control and intervention arms.


There were 6,439 women with a history of GDM identified during the study period, of whom 847 visited an eligible primary care provider at a participating clinic (Fig. 1). Of those patients, 376 (44.4%) were first seen at a control site and 471 (55.6%) were first seen at an intervention site. Patients were a mean of 41.0±7.3 years old, and 47% were white. Women seen at intervention sites were less likely to have private insurance (65% compared with 72%, P=.03) and more likely to have GDM as a coded problem within the Longitudinal Medical Record problem list than women at control sites (86% compared with 75%, P<.001) but were otherwise similar to women at the control sites. Health care providers at intervention sites were younger (mean age 42.9±10.6 years compared with 46.6±11.7 years, P<.001) than health care providers at control sites (Table 1).

Fig. 1
Fig. 1:
Study flow. *Mean cluster size (n=29); range (n=3–91). Mean cluster size (n=36); range (n=10–94).Zera. Diabetes Screening Reminder. Obstet Gynecol 2015.
Table 1
Table 1:
Patient and Health Care Provider Characteristics Among Enrolled Clinics

A mean of 2.0±1.6 reminders were generated per patient; this did not differ between intervention sites, where reminders were visible, and control sites, where the reminders were not visible to health care providers. During the study period, 265 (56%) of eligible patients at intervention sites were screened compared with 206 (55%) of eligible patients at control sites (P=.67). The majority of women who were screened had a hemoglobin A1C performed (n=242/265 [91%] screened in the intervention group compared with 194/206 [94.1%] screened in the control group; Table 2). There was no significant difference in the number of cases of diabetes identified in patients at intervention sites (n=22/471 [4.7%]) compared with those at control sites (n=15/376 [4%]) (P=.74). There was no significant difference in the proportion of women screened on the same day as the first visit during the study period (22% of intervention participants compared with 21% of control participants, P=.52). Among the patients at an intervention site, the coded response “done” or “done elsewhere” was selected infrequently (n=48 [10.2%]).

Table 2
Table 2:
Diabetes Screening in Participating Clinics

Adjustment for cluster effects and patient and health care provider characteristics did not change the effect estimate of the reminder. The electronic reminder was not associated with diabetes screening (adjusted odds ratio [OR] 1.04, 95% confidence interval [CI] 0.79–1.38). In the final adjusted model, only patient-level predictors were significantly associated with screening. For each unit increase in BMI, there was a 4% increased odds of screening, whereas nonwhite race was associated with a twofold increased odds of screening (Table 3).

Table 3
Table 3:
Multivariate Model: Odds of Diabetes Screening

Among the 321 women with a normal BMI, 82 (45%) of intervention participants were screened compared with 58 (41%) of control participants (P=.49). Among the 164 women without a history of GDM documented in the electronic problem list, 51% of the control participants were screened compared with 60% of the intervention participants (P=.24).


In our study, 55% of women with prior GDM were screened for diabetes. An EHR reminder had no effect on the rate of diabetes screening. Patient race and BMI were significant predictors of screening.

This study has several strengths. We were able to include a diverse cohort of women with prior GDM as well as a broad representation of medical practices. Unlike previous postpartum screening studies, our study evaluated screening after the postpartum period in a primary care setting.

Our results must be interpreted in the context of the limitations of our study design. The algorithm to identify at-risk women included a search for GDM in the problem list; thus, a system-wide change to allow GDM to remain an active problem after delivery was implemented before the study start date. It is therefore possible that allowing health care providers to see GDM in the problem list was in and of itself an intervention in both arms of the study. In a secondary analysis including only women without GDM in the problem list, we found no significant difference in the proportion screened; however, by restricting the population, we were underpowered for small differences in screening rates. Another potential explanation for our findings is that the reminder was passive and designed to have a low effect on health care provider workflow and therefore may have been overlooked. We were unable to assess how many health care providers noticed the reminder but did not acknowledge with a coded response, and because many sites used paper laboratory orders at the time of the study, we were unable to assess the number of women for whom screening was ordered but not done. Although we attempted to limit the intervention to primary care providers, it is also possible that patients were seen for problem visits rather than preventive care visits and therefore that screening was deferred.

Our intervention was directed only at health care providers, and patient-level factors were significantly associated with screening. Fewer women in the intervention arm were privately insured, a patient-level predictor associated with screening in other studies. It is possible that this baseline difference biased our findings toward a null result, although we did not see any association with insurance status and screening in our adjusted model. Patient race and BMI were associated with screening, suggesting that health care providers screened women they were able to easily identify as at risk. Although HbA1C is less sensitive than an OGTT for diabetes screening, it was the test performed for nearly all women who were screened, highlighting the difficulty of obtaining fasting tests that require subsequent visits.

Our screening rates were similar to those reported within the first year postpartum.7 Retrospective studies of quality improvement initiatives to increase postpartum diabetes screening in women with prior GDM that included both health care provider and patient-level interventions suggested that health care provider reminders may be effective.10,15 Clark reported a successful pilot randomized trial of patient and health care provider reminders to improve postpartum screening; however, screening rates were lower in clinical practice (28%) than in the trial (60%).16 Our study extends the existing literature to include a randomized trial in a primary care setting beyond the postpartum period.

Electronic health records represent a promising technology for quality improvement and are widely used. However, our study shows that a passive reminder is not sufficient to change performance. We hypothesize that a multipronged approach, including actionable reminders that facilitate care plans as well as patient-facing technology to engage women in their own care, is needed to improve diabetes screening for women with a history of GDM.


1. Albrecht SS, Kuklina EV, Bansil P, Jamieson DJ, Whiteman MK, Kourtis AP, et al.. Diabetes trends among delivery hospitalizations in the U.S., 1994–2004. Diabetes Care 2010;33:768–73.
2. Kim C, Berger DK, Chamany S. Recurrence of gestational diabetes mellitus: a systematic review. Diabetes Care 2007;30:1314–9.
3. Postpartum screening for abnormal glucose tolerance in women who had gestational diabetes mellitus. ACOG Committee Opinion No. 435. American College of Obstetricians and Gynecologists. Obstet Gynecol 2009;113:1419–21.
4. American Diabetes Association. Standards of medical care in diabetes—2012. Diabetes Care 2012;35(suppl 1):S11–63.
5. Almario CV, Ecker T, Moroz LA, Bucovetsky L, Berghella V, Baxter JK. Obstetricians seldom provide postpartum diabetes screening for women with gestational diabetes. Am J Obstet Gynecol 2008;198:528.e1–5.
6. Dietz PM, Vesco KK, Callaghan WM, Bachman DJ, Bruce FC, Berg CJ, et al.. Postpartum screening for diabetes after a gestational diabetes mellitus-affected pregnancy. Obstet Gynecol 2008;112:868–74.
7. Ferrara A, Peng T, Kim C. Trends in postpartum diabetes screening and subsequent diabetes and impaired fasting glucose among women with histories of gestational diabetes mellitus: A report from the Translating Research Into Action for Diabetes (TRIAD) study. Diabetes Care 2009;32:269–74.
8. Smirnakis KV, Chasan-Taber L, Wolf M, Markenson G, Ecker JL, Thadhani R. Postpartum diabetes screening in women with a history of gestational diabetes. Obstet Gynecol 2005;106:1297–303.
9. Clark HD, Graham ID, Karovitch A, Keely EJ. Do postal reminders increase postpartum screening of diabetes mellitus in women with gestational diabetes mellitus? A randomized controlled trial. Am J Obstet Gynecol 2009;200:634.e1–7.
10. Vesco KK, Dietz PM, Bulkley J, Bruce FC, Callaghan WM, England L, et al.. A system-based intervention to improve postpartum diabetes screening among women with gestational diabetes. Am J Obstet Gynecol 2012;207:283.e1–6.
11. Stuebe A, Ecker J, Bates DW, Zera C, Bentley-Lewis R, Seely E. Barriers to follow-up for women with a history of gestational diabetes. Am J Perinatol 2010;27:705–10.
12. Patient safety and the electronic health record. Committee Opinion No. 472. American College of Obstetricians and Gynecologists. Obstet Gynecol 2010;116:1245–7.
13. Sequist TD, Gandhi TK, Karson AS, Fiskio JM, Bugbee D, Sperling M, et al.. A randomized trial of electronic clinical reminders to improve quality of care for diabetes and coronary artery disease. J Am Med Inform Assoc 2005;12:431–7.
14. Campbell MK, Thomson S, Ramsay CR, MacLennan GS, Grimshaw JM. Sample size calculator for cluster randomized trials. Comput Biol Med 2004;34:113–25.
15. Lega IC, McLaughlin H, Coroneos M, Handley-Derry F, Donovan N, Lipscombe LL. A physician reminder to improve postpartum diabetes screening in women with gestational diabetes mellitus. Diabetes Res Clin Pract 2012;95:352–7.
16. Shea AK, Shah BR, Clark HD, Malcolm J, Walker M, Karovitch A, et al.. The effectiveness of implementing a reminder system into routine clinical practice: does it increase postpartum screening in women with gestational diabetes? Chronic Dis Can 2011;31:58–64.
© 2015 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.