- Research article
- Open Access
- Open Peer Review
Patterns of gestational diabetes diagnosis inside and outside of clinical guidelines
BMC Pregnancy and Childbirthvolume 17, Article number: 11 (2017)
Hospital discharge codes are often used to determine the incidence of gestational diabetes mellitus (GDM) at state and national levels. Previous studies demonstrate substantial variability in the accuracy of GDM reporting, and rarely report how the GDM was diagnosed. Our aim was to identify deliveries coded as gestational diabetes, and then to determine how the diagnosis was assigned and whether the diagnosis followed established guidelines.
We identified which deliveries were coded at discharge as complicated by GDM at the Brigham and Women’s Hospital in Boston, MA for the year 2010. We reviewed medical records to determine whether the codes were appropriately assigned.
Of 7883 deliveries, coding for GDM was assigned with 98% accuracy. We identified 362 cases assigned GDM delivery codes, of which 210 (58%) had oral glucose tolerance test (OGTT) results available meeting established criteria. We determined that 126 cases (34%) received a GDM delivery code due to a clinician diagnosis documented in the medical record, without an OGTT result meeting established guidelines for GDM diagnosis. We identified only 15 cases (4%) that were coding errors.
Thirty four percent of women assigned GDM delivery codes at discharge had a medical record diagnosis of GDM but did not meet OGTT criteria for GDM by established guidelines. Although many of these patients may have met guidelines if guideline-based testing had been conducted, our findings suggest that clinician diagnosis outside of published guidelines may be common. There are many ramifications of this approach to diagnosis, including affecting population-level statistics of GDM prevalence and the potential impact on some women who may be diagnosed with GDM erroneously.
Hospital discharge data are commonly used to estimate the prevalence of pregnancy complications including gestational diabetes mellitus (GDM). These prevalence data have the potential for far-reaching impact, since they are often used at local and national levels to influence maternal and child health programs, monitor trends, and determine allocation of resources . Therefore, understanding the patterns by which GDM diagnoses are given is important. Previous studies demonstrate variable accuracy for the diagnosis of GDM during pregnancy as reported by hospital discharge data [2–4]. Of note, many studies use documentation of GDM in the medical record as the “gold standard,” but few studies have examined how the GDM diagnosis was made [5, 6].
In 2010, the obstetric services at the Brigham and Women’s Hospital (BWH) were using the American College of Obstetricians and Gynecologists (ACOG) criteria to diagnose GDM. These criteria were first established in 2001 and then reaffirmed in 2013 [7, 8]. These guidelines recommend that all pregnant women be screened for diabetes with a 50-g glucose load test (GLT) between 24 and 28 weeks of gestation. The recommendations state that women with a GLT result of 130 or 140 mg/dL (7.2 or 7.8 mmol/L) or greater should undergo a three hour 100-g oral glucose tolerance test (OGTT), with a diagnosis of GDM given for two or more abnormal values based on Carpenter-Coustan criteria [7, 9]. The degree to which GDM diagnosis is made outside of these published guidelines has not been quantified nor well-described, yet may have significant impact on the integrity of reported data. We therefore sought to validate the hospital discharge diagnosis of GDM by comparing it to the medical record diagnosis of GDM. We then compared the medical record diagnosis to ACOG criteria to determine whether the diagnosis of GDM was made according to established guidelines.
The BWH is the largest obstetrical service in New England with over 6000 deliveries per year, and approximately 5% of these pregnancies are complicated by GDM. In 2010, at discharge, a delivery complicated by GDM was assigned codes 648.80 (Diabetes mellitus of mother complicating pregnancy childbirth or the puerperium unspecified as to episode of care), 648.81 (Diabetes mellitus of mother with delivery) and/or 648.83 (Antepartum diabetes mellitus), according to the International Classification of Diseases, Ninth Revision (ICD-9). We used the Research Patient Data Registry (RPDR), a centralized database that houses clinical data from patient medical records within the Partners Healthcare system, to obtain lists of patients discharged between January 1, 2010 and December 31, 2010 with at least one of the ICD-9 discharge codes associated with a diagnosis of GDM. In addition, we obtained a case list of all women discharged with at least one of the ICD-9 codes associated with a diagnosis of type 1 diabetes or type 2 diabetes (ICD-9 codes 648 and 250) so that we could check for any deliveries to women with GDM that may have been miscoded. We then used the RPDR to obtain a separate list for lab values of all 3-h OGTTs performed at BWH between March 1, 2009 and December 31, 2010. Since the 3-h OGTT test is used only during pregnancy to diagnose GDM, identifying these tests enabled us to create an additional case list to compare with the other lists. The institutional review board at BWH approved this project and a waiver of consent was received from the Partners Human Research Committee to conduct the medical record review.
Once we identified cases of potential GDM either by ICD-9 delivery code or by laboratory results, we reviewed medical records to obtain age, race, parity, history of prior GDM, use of insulin during pregnancy, mode of delivery, gestational age at delivery, pre-pregnancy body mass index (BMI), and infant birth weight, as well as results of GLT and OGTT testing. For cases where laboratory results for the GLT or OGTT were not available within the BWH system, we reviewed results in clinic notes and outside laboratory reports when available.
Once we obtained all of the ICD-9 coded cases as well as all available GLT and OGTT values, we compared the laboratory values to the Carpenter-Coustan ACOG guidelines [8, 9]. We then placed cases into one of five categories: 1) diagnosed by guidelines, 2) diagnosed outside of guidelines, 3) clinician error, 4) coding error, or 5) no laboratory values available (Table 1). A case met the criteria for diagnosis by guidelines if the 3-h OGTT had at least two abnormal values based on the Carpenter-Coustan thresholds and the GDM ICD-9 code was properly assigned. If the pre-pregnancy or first trimester HbA1c met criteria for type 2 diabetes, we considered these cases to have preexisting diabetes and therefore defined as a clinician error. For cases not meeting the Carpenter-Coustan laboratory criteria, we reviewed the medical record to determine whether and how a clinician made a diagnosis of GDM. A case qualified as “GDM diagnosed outside of guidelines” if the laboratory results in the medical record did not meet criteria, but the clinician provided treatment for GDM, including a prescribed diet, oral medication, and/or insulin. Two physicians independently reviewed the cases not meeting diagnosis criteria by guidelines, as well as the records of cases assigned ICD-9 codes for T1DM (n = 13) or T2DM (n = 29) to ensure that none of these cases should have been coded as GDM deliveries.
We compiled descriptive tables of cases in the five different categories, and then compared those GDM cases diagnosed by guidelines with those diagnosed outside of guidelines using univariate and multivariate analyses to determine predictors of diagnosis outside of guidelines, using a significance level of p < .05. Univariate analyses included t-tests, chi-squared tests, Fisher’s Exact tests, and Wilcoxon tests, as appropriate. Multivariable logistic regression analyses using variables with p values less than .20 were also performed. Analyses were conducted using JMP 10 Pro.
Out of 7883 total deliveries, 362 (4.6%) cases of GDM were identified by ICD-9 discharge coding (Fig. 1). We validated 210 cases (57% of all GDM cases) as appropriately coded and as diagnosed within guidelines based on laboratory values meeting Carpenter-Coustan criteria. An additional 126 GDM cases (34%) were diagnosed outside of guidelines. Three of these cases were determined to be clinician error. In these cases, the GDM code was assigned appropriately by the coders since the diagnosis was clearly specified in the medical record; however the GDM was incorrectly diagnosed by the clinician. In addition, we identified 9 cases (0.1% of all deliveries) where the OGTT met criteria but the ICD-9 delivery code for GDM was not assigned. We identified 6 cases (0.08% of all deliveries) where a GDM diagnosis was assigned but either there were no laboratory data meeting criteria, or the patient had documented Type 2 diabetes. There were 20 cases (5.5% of GDM discharge codes) where laboratory data were unavailable from the medical record and we could not assess the validity of the assigned delivery code. All deliveries coded as type 1 diabetes and type 2 diabetes were appropriately assigned.
As noted, clinical diagnosis of GDM outside of published guidelines was common (N = 126, 34% of GDM cases). There were multiple reasons for clinicians to diagnose women with GDM outside guidelines (Fig. 1). Of the 126 cases diagnosed outside of guidelines, 101 had some form of glucose tolerance testing (80%). An elevated GLT value alone led to diagnosis of GDM in 47 cases (37%), of which 28 (22%) of these cases had a GLT result ≥ 200 mg/dL (11.1 mmol/L). Nineteen of the 126 (15%) were treated as GDM due to an elevated GLT and only one abnormal value on the OGTT. In 25 cases no glucose tolerance testing was performed. Twelve of these women had a history of GDM and the clinician initiated treatment without any formal glucose testing.
Characteristics of the cases diagnosed by guidelines and the cases diagnosed outside guidelines are shown in Table 2. Age, race, and pregnancy outcomes did not differ significantly between the two groups. Women diagnosed by guidelines were more likely to be nulliparous (N = 108, 51%) than those diagnosed outside of guidelines (N = 40, 32%, P = .0004). Similarly, women diagnosed by guidelines were less likely to have a history of GDM than women diagnosed outside of guidelines (N = 31 (15%) vs. 57 (45%), P < .0001). Among the 88 women with a history of GDM, only 31 (35%) were diagnosed by guidelines (as compared to 57% overall). In a logistic regression analysis, a history of GDM led to 4.7 (95% CI 2.8–7.9) increased odds of a diagnosis outside of guidelines. Insulin use was not retained in the multivariable model and nulliparity was removed from the model due to collinearity with “history of GDM.”
In this study, we validated cases of deliveries assigned ICD-9 GDM discharge codes and found that 356/362 (98%) of cases assigned a GDM ICD-9 discharge code had GDM specified in the medical record. We identified only nine cases of deliveries with laboratory tests meeting criteria for GDM that should have been assigned a GDM ICD-9 code that were not. Of interest, we noted a large proportion of the GDM cases (34%) were diagnosed outside of established clinical guidelines. A history of GDM in a previous pregnancy demonstrated increased odds for a diagnosis outside of guidelines.
The high rate of accuracy (98%) seen in our study for GDM delivery codes may be in part due to the fact that GDM codes were assigned based on a thorough review of the medical record by experienced coders. In comparison, a medical record review study of deliveries in Washington State reported a true positive rate for GDM delivery discharge codes of 81% when compared to a review of the medical record . In a medical record review study conducted among high volume hospitals in California, the sensitivity was 78% for GDM ICD-9 codes when compared to a review of the medical record .
In our study, we explored the way in which women were given a diagnosis of GDM and notably one-third of the GDM cases were diagnosed outside of established clinical guidelines. Of these, more than half had some abnormality in their GLT and/or OGTT, including 22% with a GLT greater than or equal to 200 mg/dl. Many of these women may have met criteria if guideline-based diagnosis had been conducted. However, a previous study revealed that, even with a cut-off of greater than or equal to 200 mg/dl on the 50 g GLT, the likelihood of GDM on a 3-h OGTT was only 69% . Assigning a diagnosis of GDM to women with lesser degrees of glucose intolerance may reflect the clinician’s belief that these women would benefit from treatment. In the observational Hyperglycemia and Adverse Pregnancy Outcomes study, women with hyperglycemia not meeting criteria for GDM were still at increased odds of multiple adverse outcomes, including large-for-gestational-age (LGA) birthweight, high cord-blood serum C-peptide levels, primary cesarean delivery, neonatal hypoglycemia, premature delivery, shoulder dystocia or birth injury, intensive neonatal care, hyperbilirubinemia, and preeclampsia . Despite the compelling observational data however, there is still no clear evidence that treatment of women with these milder glucose abnormalities will decrease rates of complications . For this reason, at the time of the study and to date, the BWH has elected to use two-step testing for the diagnosis of GDM following Carpenter-Coustan criteria.
Diagnosis of GDM outside of guidelines may lead to falsely inflated estimates of recurrence risk, and also increase variability of administrative datasets. Consequently, the overall reliability of these datasets is decreased and may affect public health efforts and fund allocation. In addition, a diagnosis of GDM has many ramifications for women, including the need for frequent fingersticks, dietary restriction, a more medicalized pregnancy, and possibly increased psychological distress [13, 14]. Some studies suggest that just carrying a diagnosis of GDM increases the rate of cesarean delivery [8, 15, 16]. In addition, over-diagnosis of GDM is associated with system-wide costs. A pregnancy affected by GDM has an 18–34% increased cost over a normal pregnancy, at an average of $3305, an annual cost of $636 million .
An important finding in our study is that a history of GDM in a prior pregnancy was significantly associated with clinicians making a diagnosis of GDM without an OGTT meeting criteria by guidelines or sometimes without performing glucose tolerance testing at all. The tendency not to formally retest women with prior GDM may have a major impact on the estimates of recurrence of GDM as many studies of recurrence rates use a discharge diagnosis of GDM as the documentation for recurrence. Estimates of the rates of recurrence of GDM in a subsequent pregnancy are estimated at 41% from large population studies, with a broader range of 30-84% seen in studies of smaller populations [18, 19]. A recent report of recurrent GDM in Massachusetts looked at recurrent GDM using both discharge diagnosis and birth certificate data found similarly high rates of recurrence whether by birth certificate alone (38%), discharge diagnosis alone (47%) or the combination (48%). Importantly, however, the authors did not evaluate OGTT data . Our study provides important insight into the data on the recurrence of GDM by documenting that at our institution only 35% of women with a history of GDM given a discharge diagnosis of GDM were diagnosed by an OGTT meeting criteria in established guidelines.
Our findings must be interpreted in the context of study design. We may not have identified all cases that should have been coded as GDM but were not, especially if no laboratory data were found in our system. This was a single institution study, and the coding and clinician diagnosis patterns we identified may not be representative of patterns at other hospitals.
Future studies should address the prevalence of GDM diagnosis outside of established guidelines in other settings, including community hospitals in addition to academic centers. Given recent controversy about new diagnostic guidelines for GDM , it is important to note that a large proportion of diagnoses may be made outside of guidelines. Estimates of the impact of the new guidelines on increasing the diagnosis of GDM  may substantially underestimate the true impact of more liberal guidelines on the incidence of GDM if clinical diagnosis outside of guidelines continues to occur. It may be important to clarify clinical coding rules and diagnosis guidelines in conjunction with the implementation of new criteria, particularly if clinicians are using their own judgment about when to apply guideline criteria. In addition, public health agencies that use ICD discharge data to estimate state and national incidence should be aware of the potential contribution of clinician diagnosis to those women receiving a chart diagnosis of GDM and therefore an ICD code of GDM.
In this study, 98% of deliveries assigned ICD-9 gestational diabetes (GDM) discharge codes did have a diagnosis of GDM specified in the medical record. However, 34% of women assigned GDM delivery codes at discharge did not meet OGTT criteria for GDM by established guidelines. Although many of these patients may have met guidelines if guideline-based testing had been conducted, our findings suggest that clinician diagnosis outside of published guidelines may be common. There are many ramifications of this approach to diagnosis, including affecting population-level statistics of GDM prevalence and recurrence, as well as the potential impact on some women who may be diagnosed with GDM erroneously.
Gestational Diabetes Mellitus
Glucose Load Test (50-g)
International Classification of Diseases, Ninth Revision
Oral Glucose Tolerance Test
Gillespie P, Cullinan J, O’Neill C, Dunne F. Modeling the independent effects of gestational diabetes mellitus on maternity care and costs. Diabetes Care. 2013;36:1111–6.
Lydon-Rochelle MT, Holt VL, Cardenas V, Nelson JC, Easterling TR, Gardella C, Callaghan WM. The reporting of pre-existing maternal medical conditions and complications of pregnancy on birth certificates and in hospital discharge data. Am J Obstet Gynecol. 2005;193:125–34.
Devlin HM, Desai J, Walaszek A. Reviewing performance of birth certificate and hospital discharge data to identify births complicated by maternal diabetes. Matern Child Health J. 2009;13:660–6.
Yasmeen S, Romano PS, Schembri ME, Keyzer JM, Gilbert WM. Accuracy of obstetric diagnoses and procedures in hospital discharge data. Am J Obstet Gynecol. 2006;194:992–1001.
Dietz PM, Vesco KK, Callaghan WM, Bachman DJ, Bruce FC, Berg CJ, England LJ, Hornbrook MC. Postpartum screening for diabetes after a gestational diabetes mellitus-affected pregnancy. Obstet Gynecol. 2008;112:868–74.
Hillier TA, Ogasawara KK, Pedula KL, Vesco KK. Markedly different rates of incident insulin treatment based on universal gestational diabetes mellitus screening in a diverse HMO population. Am J Obstet Gynecol. 2013;209:440.e441–449.
ACOG Practice Bulletin. Clinical management guidelines for obstetrician-gynecologists. Number 30, September 2001 (replaces Technical Bulletin Number 200, December 1994). Gestational diabetes. Obstet Gynecol. 2001;98:525–38.
Practice Bulletin No 137. Gestational diabetes mellitus. Obstet Gynecol. 2013;122:406–16.
Carpenter MW, Coustan DR. Criteria for screening tests for gestational diabetes. Am J Obstet Gynecol. 1982;144:768–73.
Atilano LC, Lee-Parritz A, Lieberman E, Cohen AP, Barbieri RL. Alternative methods of diagnosing gestational diabetes mellitus. Am J Obstet Gynecol. 1999;181:1158–61.
Metzger BE, Lowe LP, Dyer AR, Trimble ER, Chaovarindr U, Coustan DR, Hadden DR, McCance DR, Hod M, McIntyre HD, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med. 2008;358:1991–2002.
National Institutes of Health consensus development conference statement. Diagnosing gestational diabetes mellitus, March 4–6, 2013. Obstet Gynecol. 2013;122:358–69.
Daniells S, Grenyer BF, Davis WS, Coleman KJ, Burgess JA, Moses RG. Gestational diabetes mellitus: is a diagnosis associated with an increase in maternal anxiety and stress in the short and intermediate term? Diabetes Care. 2003;26:385–9.
Hui AL, Sevenhuysen G, Harvey D, Salamon E. Stress and anxiety in women with gestational diabetes during dietary management. Diabetes Educ. 2014;40:668–77.
Naylor CD, Sermer M, Chen E, Sykora K. Cesarean delivery in relation to birth weight and gestational glucose tolerance: pathophysiology or practice style? Toronto Trihospital Gestational Diabetes Investigators. JAMA. 1996;275:1165–70.
Mitrovic M, Stojic S, Tesic DS, Popovic D, Rankov O, Naglic DT, Paro JN, Pejin R, Bulatovic S, Veljic MT, Zavisic BK. The impact of diabetes mellitus on the course and outcome of pregnancy during a 5-year follow-up. Vojnosanit Pregl. 2014;71:907–14.
Chen Y, Quick WW, Yang W, Zhang Y, Baldwin A, Moran J, Moore V, Sahai N, Dall TM. Cost of gestational diabetes mellitus in the United States in 2007. Popul Health Manag. 2009;12:165–74.
Getahun D, Fassett MJ, Jacobsen SJ. Gestational diabetes: risk of recurrence in subsequent pregnancies. Am J Obstet Gynecol. 2010;203:467.e461–466.
Khambalia AZ, Ford JB, Nassar N, Shand AW, McElduff A, Roberts CL. Occurrence and recurrence of diabetes in pregnancy. Diabet Med. 2013;30:452–6.
England L, Kotelchuck M, Wilson HG, Diop H, Oppedisano P, Kim SY, Cui X, Shapiro-Mendoza CK. Estimating the Recurrence Rate of Gestational Diabetes Mellitus (GDM) in Massachusetts 1998–2007: Methods and Findings. Matern Child Health J. 2015;19:2303–13.
Vandorsten JP, Dodson WC, Espeland MA, Grobman WA, Guise JM, Mercer BM, Minkoff HL, Poindexter B, Prosser LA, Sawaya GF, et al. NIH consensus development conference: diagnosing gestational diabetes mellitus. NIH Consens State Sci Statements. 2013;29:1–31.
This study was supported by the Centers for Disease Control (CDC 200-2011-M-3918), and work by the investigators was supported by an Institutional National Research Service Award #T32AT000051 from the National Center for Complementary and Alternative Medicine (NCCAM) at the National Institutes of Health, an NIH BIRCWH K12 HD057022, an NIH NHLBI 1K23HL133604-01, and a K24 from the National Heart Lung and Blood Institute at the National Institutes of Health (9K24HL096141). We acknowledge the contribution from program officers at the Centers for Disease Control who provided conceptualization of ideas, interpreted findings, and review drafts of the manuscript. However, the findings and conclusions are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Availability of data and materials
The data will not be shared due to the presence of protected health information.
Dr. Nicklas takes responsibility for the integrity of the data and the accuracy of the data analysis. J.N., J.L., and E.S. had full access to the data. J.N. and J.L. wrote the manuscript. J.N., C.Z., J.L., and E.S. researched data and reviewed and edited the manuscript. J.N. conducted the statistical analysis. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
The Institutional Review Board at Brigham and Women’s Hospital approved this project and a waiver of consent was received from the Partners Human Research Committee to conduct the medical record review.