According to the 2020 National Diabetes Statistics Report, approximately 1 in 3 American adults has impaired glucose tolerance (IGT) or prediabetes, defined as hemoglobin A_1c (HbA_1c) of 5.7 to 6.4 [1]. IGT is often considered an intermediate state of abnormal glucose metabolism that exists between normal glucose homeostasis and diabetes. The natural history of IGT is variable and may include normalization, persistence of impaired tolerance, or progression to diabetes [2-4]. Up to 70% of patients with IGT may ultimately develop diabetes without intervention [5]. Given the variable natural history and the ability to alter disease trajectory with behavioral modifications, individuals with IGT are ideal candidates for diabetes prevention efforts [6,7]. Pregnancy is a time when individuals enter the health care system and may be motivated to focus on their health. As a result, obstetricians can play a role in promoting overall health, impacting both future pregnancy and long-term health outcomes.

As the prevalence of preexisting and gestational diabetes mellitus (GDM) among pregnant people has increased in recent years [8,9], universal HbA_1c screening in early pregnancy has been proposed to identify undiagnosed pre-existing diabetes as well as to screen for early-onset GDM [10,11]. In a retrospective cohort at a single institution, 12% of patients had IGT in early pregnancy but never developed GDM [12]. Patients with IGT in pregnancy are at risk of both adverse pregnancy outcomes and long-term health outcomes, even if they never develop GDM. Although there are relatively few studies of pregnant patients with IGT and without GDM, several studies have demonstrated an increased risk of spontaneous preterm birth less than 37 weeks [13]. A singular cohort study has also shown a potentially increased risk of hypertensive disorders of pregnancy (HDP), major congenital anomalies, shoulder dystocia, large for gestational age births, and perinatal death in this population [13]. The American Diabetes Association recommends that patients with IGT complete HbA_1c screening for the development of diabetes every 1 to 2 years [14]. Pregnant patients diagnosed with GDM receive counseling on dietary and physical activity recommendations, antenatal glucose monitoring, counseling on long-term metabolic risks, and postpartum glucose testing. In contrast, there are no established guidelines for the prenatal or postpartum management of pregnant patients with IGT and without GDM. They may not receive dietary and exercise counseling, counseling on long-term metabolic risks, postpartum glucose testing, or a formal transition of care to a primary care physician or endocrinologist. This can create gaps in care with implications for future pregnancy and future health outcomes. In previously unpublished data from our institution, an exploratory analysis of long-term outcomes for 100 patients with IGT in pregnancy at a single institution demonstrated that 71% of patients did not have a visit with a primary care provider (PCP) in 1 to 3 years after delivery. Of the 29 patients with PCP visits, nearly half (48.3%) did not obtain repeat HbA_1c values to determine if there was persistence of IGT. In half of the cases, this was due to a lack of ordering by the PCP. Of the 15 patients who completed repeat HbA_1c testing, 93.3% remained more than 5.7% 1 to 3 years after delivery. There were also 14 repeat pregnancies in this cohort, 78.6% of which continued to demonstrate IGT during early pregnancy screening.

SMS text messaging support programs have been used effectively during the postpartum period to support transitions of care. A randomized controlled trial of an SMS text messaging support program prompting primary care transitions demonstrated significant improvement in primary care visit completion, as well as fewer postpartum readmissions and increased receipt of services, such as blood pressure assessments and screenings for mood symptoms [15]. Other SMS text messaging support programs have focused on the completion of recommended clinical assessments postpartum for patients with pregnancy complications, such as hypertensive disorders of pregnancy or gestational diabetes, with many programs reporting improved engagement with SMS text messaging support [16-20].

Based on our institutional data demonstrating clinical need, our team developed a program called Better Follow-up of Impaired Glucose Tolerance (BRIDGE) to address postpartum transitions of care for patients diagnosed with IGT during pregnancy. Here, we present a single institution nonrandomized pre-post implementation study, including (1) the design process for BRIDGE and a program evaluation, (2) a comparison of clinical outcomes for BRIDGE patients to a historical population, and (3) a comparative analysis examining differences in outcomes between 2 programmatic strategies used within BRIDGE. We hypothesize that the addition of SMS text messaging support and an IGT-focused postpartum visit will improve the completion of desired postpartum milestones for this population (ie, HbA_1c completion, primary care visit scheduling). Based on the findings noted in this study, we anticipate performing a future randomized trial to definitively assess the impact of this intervention.

This is a pre-post implementation study of BRIDGE, an SMS text messaging support program promoting postpartum transitions of care for patients with IGT during pregnancy. A program evaluation was performed assessing differences in outcomes between the BRIDGE population and a historical population, as well as between 2 groups enrolled in 2 different programmatic strategies of BRIDGE, as described previously in the Overview of Program Design section.

Patients were eligible for the program if they had an HbA_1c value of 5.7 to 6.4 prior to 20 weeks of gestation, completed third-trimester glucose testing, did not receive a diagnosis of GDM, had a singleton pregnancy, delivered at our hospital, and were English-speaking. BRIDGE was implemented at a single tertiary care institution, with approximately 4000 deliveries per year, in the northeastern United States. Opt-out enrollment of eligible patients was performed manually by the study team at the time of their delivery. Patients received a text message with a description of the program, recommended milestones, and the option to decline participation.

For our historical population, we selected all patients who met identical eligibility criteria to BRIDGE participants and delivered between June 2018 and December 2019. We selected a continuous 19-month time period (identical to the BRIDGE program duration) that preceded the COVID-19 pandemic in March 2020 (given the disruptions to typical care during that time).

The 2 primary outcomes for the BRIDGE program were (1) completion of an HbA_1c test and (2) scheduling of a visit with a PCP by 12 weeks postpartum. Given the paucity of national recommendations regarding follow-up for IGT identified in early pregnancy, these postpartum transition milestones were determined based on expert opinions from our maternal fetal medicine, primary care, and endocrinology teams. When completing the comparative analysis with the historical population, HbA_1c completion was assessed by 1 year postpartum rather than by 12 weeks postpartum to provide an appropriate comparison with the standard of care during the historical period. For our analysis comparing the 2 programmatic strategies in BRIDGE− and BRIDGE+, we assessed HbA_1c completion by 12 weeks postpartum, which is the programmatic goal, as well as by 1 year postpartum, which is aligned with national recommendations from some organizations such as the American Diabetes Association [24]. To account for potential bias caused by missing data, because the authors were unable to use retrospective chart abstraction to identify data outside of the health care system, reporting of these outcomes was stratified among patients with an identified PCP within the network versus patients with an identified PCP outside of the network. In addition, the evaluation was adjusted for demographic and clinical characteristics that differed between the study groups.

Secondary outcomes included other process outcomes, such as PCP visit attendance and the inclusion of IGT counseling during PCP visits (for in-network patients where these data were available in our EMR). We also reported descriptive data regarding iPPV scheduling and attendance for BRIDGE+ participants only. We stratified all outcomes related to PCP visit scheduling and attendance by patients who had in-network PCPs versus all patients due to the insurance and workforce capacity challenges faced by patients who did not have established PCPs within our health care system, as well as considerations regarding potential missing data.

The data were collected by the study team on participant demographics and clinical characteristics, as well as primary and secondary outcomes, through an EMR review.

With a sample size of 246 patients, we would be powered to detect an improvement in PCP visit scheduling from an assumed baseline of 30% based on prior chart review at our institution as well as existing literature [25] to 45% with 80% power and an α level of .05. With an annual birth volume of 4000 patients per year and baseline data at the institution demonstrating that 12% of the population has IGT in early pregnancy, the study anticipated the need to collect data for at least 7 months to accrue a sufficient sample size. All analyses were performed using Stata version 17.0 (StataCorp LLC).

For this nonrandomized study, the CONSORT (Consolidated Standards of Reporting Trials) extension for pilot and feasibility trials was used to report the findings including the abstract extension [26,27].

Race and ethnicity were self-reported. For the primary analysis comparing the BRIDGE population to a historical population, bivariate analyses comparing clinical and demographic characteristics between our BRIDGE− or BRIDGE+ and historical populations were performed using 2-tailed t tests for continuous variables and chi-squared tests for categorical variables. We also assessed 95% CIs for point estimates of these characteristics in each group using the Wilson score intervals. Multivariable logistic regressions comparing the odds of all assessed outcomes in the BRIDGE versus historical population included all variables with P<.20 in the bivariate analyses. Backward stepwise regression was used to create parsimonious models using P>.20 for elimination. An identical strategy was used to develop models for comparative analysis assessing differences in outcomes between our BRIDGE− and BRIDGE+ populations. For both analyses, models comparing outcomes across populations ultimately only included a history of IGT in prior pregnancy as a covariate. Multiple imputation analysis was used to address missing data at random for primary care visit scheduling and attendance.

The project was designated as a quality improvement project and deemed exempt from review by the University of Pennsylvania Institutional Review Board, as (1) this intervention did not guide any clinical decisions regarding care and was intended solely to improve care delivery within the health care system, and (2) participants were not randomized to the intervention and were intended to obtain direct benefits from the intervention. Informed consent was not required for this quality improvement initiative. Deidentified data were collected for retrospective review and stored on secure institutional servers. Participants were not compensated for their participation in this initiative. There is no identification of any individual participants in any manuscript images or supplementary materials.

The recruitment of participants occurred between October 2021 and April 2023 and was stopped due to funding constraints, although the target sample size was not achieved. A convenience sampling of patients meeting the eligibility criteria for the intervention and presenting to prenatal clinics within the health care system over the study time period was performed. Providers had the option of directly enrolling patients from the clinic. In addition, on a weekly basis, the study team received an automated report from the health care system analytics team with all newly eligible participants. After reviewing the medical records to confirm eligibility, the study team manually enrolled all eligible patients. The team also reviewed the EMR of any participants added directly by clinicians and removed participants deemed ineligible. A participant flow diagram was included to demonstrate the number of individuals enrolled and ultimately included in the program evaluation (Figure 2).

In total, 161 participants were enrolled in the BRIDGE program, compared to 342 participants in the historical population (Table 1). A higher proportion of the BRIDGE population had a personal history of IGT and polycystic ovarian syndrome compared to the historical population (Table 1). There were 82 participants in the BRIDGE− arm and 79 participants in the BRIDGE+ arm (Table 2). The BRIDGE− and BRIDGE+ populations had similar demographic and clinical characteristics.

BRIDGE participants had increased odds of HbA_1c completion by 1 year postpartum (39.8% vs 12.5%; adjusted odds ratio [aOR] 4.28, 95% CI 2.71‐6.78) and PCP visit scheduling (31.0% vs 12.0%; aOR 9.58, 95% CI 4.39‐20.9) compared to the historical population (Table 3). The difference in PCP visit scheduling remained statistically significant in the multiple imputation analysis (aOR 9.49, 95% CI 4.60‐19.6; Multimedia Appendix 1). This was also true among the subgroup of patients with in-network PCPs (27.3% vs 12.0%; aOR 8.30, 95% CI 3.77‐18.2). Importantly, in the historical population, every patient who scheduled a PCP visit had an established in-network PCP during pregnancy; in contrast, in the BRIDGE population, 6 out of the 50 (12%) patients who scheduled a PCP visit did not already have an established in-network PCP.

There were remarkably high levels of attendance of PCP visits among those who scheduled PCP visits, across both the historical and BRIDGE populations (95% CI 92.7%‐96% of scheduled visits). However, rates of IGT counseling during PCP visits were relatively low in both the historical and BRIDGE populations (95% CI 39.5%‐54.2% of visits).

In comparing the BRIDGE− to BRIDGE+ arms of the program, the BRIDGE+ population had slightly improved odds of HbA_1c completion by 12 weeks postpartum (36.7% vs 19.5%; aOR 2.31, 95% CI 1.12‐4.74), though this positive effect using the BRIDGE+ strategy did not persist by 1 year postpartum (48.1% vs 31.7%; aOR 1.88, 95% CI 0.98‐3.61; Table 4). There was no difference in PCP visit scheduling between the BRIDGE− and BRIDGE+ program strategies (32.9% vs 29.3%; aOR 0.46, 95% CI 0.11‐2.00), which remained consistent in multiple imputation analysis (aOR 0.49, 95% CI 0.10‐2.25; Multimedia Appendix 1). This was similar when looking at PCP scheduling among patients with established in-network PCPs specifically (29.1% vs 25.6%; aOR 0.47, 95% CI 0.11‐2.09). A similar proportion of patients who did not have established in-network PCPs were able to schedule a PCP visit in both the BRIDGE− and BRIDGE+ populations.

Once again, there were remarkably high attendance rates for scheduled PCP visits (91.6%-100%). There were relatively low levels of IGT counseling during PCP visits (53.8%-54.5%). In the BRIDGE+ population, a larger proportion of patients scheduled an iPPV than a PCP visit (59.5% vs 29.1%), but attendance rates at scheduled iPPV visits were much lower than those for PCP visits (48.9% vs 100%).

Implementing BRIDGE, an SMS text messaging support program with educational and behavioral prompts focusing on postpartum transitions for patients with IGT during pregnancy, resulted in increased completion of an HbA_1c test within 1 year postpartum and increased scheduling of a PCP visit by 12 weeks postpartum. The addition of an iPPV to the SMS text messaging support improved the completion of HbA_1c testing by 6 weeks postpartum, but this effect did not persist at 1 year postpartum and did not impact PCP visit scheduling.

To our knowledge, this is the first study examining the impact of an intervention on postpartum transition milestones for patients diagnosed with IGT in pregnancy, who often do not receive tailored support during pregnancy and the immediate postpartum period through specialized programming given to patients with HDP or GDM (eg, transition clinics, remote blood pressure monitoring programs, and diabetes in pregnancy programs). This is an important population to target as these patients experience higher rates of negative interpregnancy and long-term health outcomes [28-30].

While there is a paucity of literature focusing on patients with IGT, postpartum transitions for patients with GDM have been well studied. Postpartum glucose screening rates within 12 weeks postpartum for patients with GDM (primarily assessed via completion of a 2-hour glucose tolerance test rather than HbA_1c completion by 6 weeks postpartum, based on guidelines) range from 7% to 59% in the literature, which are similar to the 6-week HbA_1c completion rate in historical and BRIDGE populations [31-34]. The PCP visit attendance rates for both the historical population and BRIDGE population were similar to published rates, which range from 22% to 60% [15,35]. Multiple systematic reviews assessing the impact of reminder interventions (eg, emails, phone calls, and mail) on postpartum glucose screening rates for patients with GDM have shown potential benefits [36-38]. A randomized controlled trial of a text message intervention targeting primary care transitions for patients with GDM and HDP improved PCP visit completion by 18.7%, which is similar to the difference in PCP visit attendance between the historical and BRIDGE populations [15]. Overall, this consistency with existing literature provides reassurance that the BRIDGE intervention may be generalizable to other settings.

This program evaluation demonstrated that the integration of iPPV increased HbA_1c completion in the first 12 weeks postpartum, but this improvement did not persist until 1 year postpartum and did not impact PCP visit scheduling. Patient engagement may play a role in this transient impact of iPPV integration on desired postpartum milestones. While the iPPV scheduling rate was nearly double the PCP visit scheduling rate, the number of individuals who attended the iPPV was similar to the number of individuals who attended the PCP visit. Although the authors were insufficiently powered to assess the relationship between iPPV and PCP visit attendance, it is also possible that patients who completed an iPPV perceived a decrease in the usefulness of PCP visits in the first year postpartum because they had already received IGT-counseling and screening via HbA_1c completion from a prenatal care provider.

This program evaluation highlights areas for improvement in the quality of the postpartum transition for patients with complications of pregnancy, as a large proportion of PCP visits in both the historical and BRIDGE populations did not include counseling regarding IGT. Per existing literature, potential contributing factors to this gap in primary care transitions may be related to communication breakdowns between the obstetric and primary care teams regarding pregnancy complications that may influence long-term counseling or management, awareness and knowledge of GDM guidelines among PCPs, and patients’ recall of providers’ counseling regarding long-term risks, among other factors [39,40]. Potential solutions include enhanced support for care coordination (eg, expedited referral pathways to primary care and patient navigation services), improved handoffs between obstetric and PCP teams (eg, standardized documentation of pregnancy complications), and reminder systems [41].

Our efforts to effectively implement BRIDGE highlighted many deep-rooted challenges with primary care transitions for pregnant patients. From a health care system standpoint, PCP access was limited by both health insurance specifications and reduced capacity in the primary care workforce. This limited the ability to successfully schedule PCP visits for postpartum patients without targeted efforts in this area. In addition, patients reported negative social drivers of health (eg, lack of childcare support, transportation support, or challenges with newborn care such as a city-wide formula shortage) as well as physical or mental health challenges that served as barriers or competing priorities for patients attempting to complete postpartum transition milestones. These challenges have been well characterized in both quantitative and qualitative analyses within the literature and highlight the need for a multilevel approach to optimizing postpartum transitions of care, addressing individual, health care system, and societal barriers [42-44].

The study had several key limitations. As a nonrandomized pre-post implementation study, our analysis cannot account for unmeasured confounders. The BRIDGE population had higher-risk metabolic comorbidities than the historical population (eg, history of IGT in prior pregnancy), but the overall cohort size was much smaller. As only 8 participants declined enrollment in BRIDGE, the study does not suspect that this difference in cohort size is reflective of selection bias but may be due to other unmeasured factors contributing to decreased rates of IGT in the contemporary population. The COVID-19 pandemic also introduced constraints to the preimplementation or postimplementation design that increase the risk of temporal bias. For example, the historical population in this study was not subject to the systemic changes in health care post–COVID-19, such as the rising use of telemedicine [45], which may have improved access to care [46]. Finally, the study was unable to recruit the target number of participants despite extending the planned enrollment period due to decreases in the prevalence of IGT in the population. The smaller sample size raises some concern regarding the precision and robustness of estimates in the primary analysis, as well as with the multiple imputation analysis. A larger, randomized trial will be required to ascertain the definitive impact of this intervention, though this study provides positive preliminary results.

There are also some scalability concerns with the BRIDGE program. Specifically, the iteration of the BRIDGE program used in this study was resource intensive, involving the manual enrollment of participants by study staff using EMR review; however, if proven effective, these resource intensive steps could be automated, as has been done with other W2H programs in our institution [47]. In addition, other institutions may not have access to the staffing capacity required to support other aspects of the program, such as social work support or IGT-focused postpartum visits. Finally, the study enrolled only English-speaking patients at a single tertiary care institution in the northeastern United States, which limits generalizability.

The study may also be prone to bias due to missing data. For patients with out-of-network PCPs, the study had extremely limited data on PCP visit scheduling and attendance. Phone interviews were attempted for BRIDGE patients to obtain self-reported data but had a low response rate (n=10). Phone interviews were deferred for the historical population because of concerns about recall bias. It is possible, therefore, that the PCP visit scheduling and attendance rates among patients with out-of-network PCPs in the historical population were under-reported. This would bias the results toward programmatic impact. However, reassuringly, the beneficial impact of BRIDGE on PCP visit scheduling persists when looking at patients with in-network PCPs alone, where complete access to primary care data is present.

The study also had many strengths. First, the development of the BRIDGE program used robust input from interdisciplinary teams spanning behavioral health, digital health, and communications expertise [48]. In addition, the preimplementation and ongoing implementation work with prenatal care sites allowed the authors to optimize the effectiveness of the program using flexible and varied implementation strategies [49], based on an expansive understanding of the barriers faced by patients and providers attempting to achieve a successful postpartum transition. Finally, the population was racially and socioeconomically diverse with varied experiences with PCP care entering pregnancy, which may make their experiences more generalizable to other populations.

Future directions for research could include additional assessments of implementation outcomes (eg, fidelity and acceptability) to ascertain contributing factors to the difference in outcomes between BRIDGE− and BRIDGE+ [50]. Following this work, a formal implementation trial of BRIDGE where groups are randomized to the intervention, which would account for any unmeasured confounding in this study [51]. Planning for that trial should include phone or in-person surveys or other strategies to obtain complete information on primary care transitions for individuals with out-of-network PCPs. Consideration should also be given to adapting BRIDGE to be applicable to more diverse or resource-limited settings [52]; potential strategies could include the automation of participant enrollment, integration of iPPV content into routine postpartum visits (thus eliminating the need for additional staffing), the conversion of iPPV from an in-person to telehealth visit (thus eliminating the need for additional clinic capacity), and the cocreation of adapted messaging in other languages alongside community partners.

In this nonrandomized pre-post implementation study, SMS text messaging support improved postpartum transitions of care for patients diagnosed with IGT in pregnancy, though the quality of the transitions to primary care remained low with poor rates of IGT counseling by PCPs. SMS text messaging support with educational and behavioral prompts tripled completion rates of HbA_1c screening within 1 year postpartum and doubled the scheduling rate for PCP visits by 12 weeks postpartum. While attendance at scheduled PCP visits was very high, less than 60% of PCP visits included IGT counseling, highlighting key areas for improvement in the quality of postpartum transitions to primary care. This study also highlighted a complex array of individual, systemic, and societal barriers to achieving desired postpartum milestones. A larger randomized trial is needed to definitively ascertain the impact of this SMS text messaging intervention. Optimizing postpartum transitions of care for populations with longitudinal metabolic risk after pregnancy will require a multilevel approach, of which text message support could be an important component.