Women who meet eligibility criteria based on our screening of electronic health record (EHR) data will receive an invitation letter. We will replicate our successful recruitment approach from prior work in SDM for diabetes prevention [36–38], first confirming patient eligibility with their physician, followed by outgoing patient intake calls, a brief touch-base call with a study principal investigator (PI), and virtual informed consent. All participants will receive a weight scale, which will be used for measuring their weight at baseline and at each follow-up. They will be instructed to weigh themselves without shoes and with emptied pockets. We use a block randomization approach stratified by health system 1:1 to the intervention versus control arm (Fig. 2). Participants will be scheduled for a face-to-face or virtual visit based on their preference, where baseline weight and survey responses will be recorded, and where, for those in the intervention arm, SDM will be performed (Fig. 2. Participants will not be blinded to the study assignment, however the lead investigators and individuals collecting follow-up data (e.g. surveys and EHR reviews) will. We will track recruitment on a monthly basis, including the numbers of patients contacted, enrolled, and randomized (Fig. 2).

Pharmacists or registered nurses trained in SDM will meet face-to-face in a clinical setting or virtually over video with each participant. During the SDM visit, the pharmacist or registered nurse will help the participant move through the steps of the decision process, using the prediabetes DA from Healthwise^®, the health education nonprofit with whom we contracted. The clinician will present a 100-person chart based on data from the DPP trial which demonstrates the relative benefits of metformin and ILI in the form of a DPP for T2DM prevention among women with a history of GDM (Fig. 3). While DPP and metformin were not tested in combination, women will have the option of choosing DPP, metformin, both options, or taking no action. Participants will also receive handouts on diabetes prevention from the National Diabetes Education Program (NDEP).

After the initial SDM visit, the pharmacist or registered nurse will communicate the participant’s choice of diabetes prevention strategy to their designated primary provider, through a note in the EHR. If the participant chooses DPP, they will receive a list of in-person or online providers registered with the CDC Diabetes Prevention Recognition Program, including the DPP providers most likely to be covered by their insurance. The research team will follow up in 2–3 weeks with each participant who chose lifestyle change to help them navigate initial DPP enrollment. If a CDC-recognized DPP is not covered by a participant’s insurance, they will be provided with self-pay DPP options to enroll in if they are amenable. These self-pay DPP options vary in cost, averaging $290 for the 12-month lifestyle change program. If the participant chooses metformin, pharmacists can prescribe it with the approval of the patient’s physician, and registered nurses will propose the prescription request to the patient’s physician. The goal/prescribed dose will be 2000 mg per day of immediate release metformin. Although 850 mg twice daily of immediate release metformin was used in the original DPP trial, 1000 mg twice daily is a more commonly used dose in clinical practice and more aligned with our pragmatic approach. The UCLA pharmacists and Intermountain registered nurses will also follow up with patients who chose metformin in 2–3 weeks to assess any side effects, which they can help the patient manage by temporarily reducing the metformin dose or using other effective mitigation strategies, such as switching to 1500 mg per day of extended release metformin. For those who cannot tolerate the goal metformin dosage, we will encourage patients to take the maximum tolerated dose.

Participants randomized to the control arm will receive an in-person or virtual visit with a research study team member, who will review the NDEP print materials with the patient and measure their weight. The same amount of time is spent with patients in the control and intervention group. Though we describe the control group as receiving “usual care,” their receipt of diabetes prevention materials and meeting with a research study team member is actually above what is typically considered usual care, since they are also presumably receiving prediabetes counseling from the PCP or physician who ordered the labs that resulted in a prediabetes diagnosis. It is primarily the responsibility of each participant’s ordering clinician to discuss prediabetes management (including lifestyle change and/or metformin) with their patient. Our study team’s interactions with the control group are meant to supplement that information.

We will use approaches recommended by the NIH Behavior Change Consortium to ensure that the SDM intervention is being delivered and received as intended [39]. These include the training of all pharmacists and registered nurses about DPP findings and SDM, which has already taken place as part of the broader diabetes prevention implementation study. We will schedule additional training modules specific to GDM that include role-playing scenarios. Our physicians will lead didactics on GDM including the pathophysiology, diagnostic criteria, and treatment of GDM during pregnancy as well as recommended post-partum care. The pharmacists and registered nurses will use an electronic template that becomes a note in the EHR for every patient visit. This template includes check boxes for each step in the structured decision-making framework.

We will provide each participant with a $35 gift card after their baseline visit, a $15 gift card after their 6-month telephone follow-up, and $35 gift cards after their 12 month and 24 month follow ups.

We will collect data at baseline, 6 months (by telephone), 12 months, and 24 months (a total of 3 in-person or virtual visits, plus a telephone visit) for participants in both the intervention and control arms of the evaluation. A baseline questionnaire was developed to ask participants about their risk awareness of T2DM and experience with preventative measures, as well as to assess patient activation and lifestyle choices around nutrition/physical activity. At the initial visit, the intervention participants will meet with pharmacists or registered nurses while the control participants will meet with research assistants. At 6 months, all participants will receive a telephone call from a research assistant to assess uptake of lifestyle change (yes/no) and/or uptake of metformin (yes/no). Attendance of a minimum of 9 of 16 weekly lifestyle change sessions will be the threshold for uptake of lifestyle change for both groups, as this reflects the CDC’s standard for DPP recognition from 2011 to 2020. Since some patients will not be able to tolerate metformin at the goal total daily dose of 2000 mg, we will encourage patients to take the maximum dose that they can tolerate and will count any metformin dose toward the yes/no outcome of “taking metformin.” We will backfill with EHR data on medication reconciliation as needed, since nurses and medical assistants at both systems verify and document current medications at each ambulatory and inpatient visit. At the 12-month and 24-month follow-ups, all participants will conduct an in-person or virtual follow-up visit with research assistants for weight measurement and follow-up surveys. The primary outcome (proportion of participants with ≥5% weight loss at 12 months) will be determined from weights recorded at study-related visits, but we will backfill missing weights from the EHR. We will include a variable in our analysis indicating whether weight was measured based on the patient’s home scale provided by the study or backfilled from the EHR. Additional outcomes are shown in Table 1.

Hypothesis: The intervention will result in a greater proportion of participants achieving ≥ 5% weight loss at 12 months as compared to patients receiving usual care.

For the primary analysis, we will use logistic regression to compare the proportion of participants with ≥5% weight loss between the two study arms at 12 months. The model will include indicator variables for study arm, SDM provider type (UCLA pharmacist vs. Intermountain registered nurse), and pre-selected baseline patient characteristics such as age and race/ethnicity. Given potential differences in intervention effectiveness across sites, we will conduct preliminary analyses within each site. Effect estimates for the intervention and relevant baseline covariates, if any, from the two models will be compared. Those with relatively large differences indicate potential interaction effects with site (i.e., site is an effect modifier). We will not draw final conclusions from these preliminary analyses, but in the formal analysis which combines data from the two sites, we can test interaction effects between site and intervention, and between site and other covariates as indicated by the preliminary analyses.

We used data from prior work in SDM for diabetes prevention to estimate the effect size of the proposed intervention [36–38], in which 31% of intervention participants and 13% of control participants lose ≥5% body weight at 12 months. Assuming 20% attrition, we will need at least 152 participants per arm to provide 90% power using a two-sided exact test at a significance level of 0.05. For this individually-randomized two-site study, we determined that it was not necessary to consider the intraclass correlation in the power calculation if we include a dummy variable for site (UCLA vs. Intermountain) in our analyses. We will set our enrollment goal at 310 participants (divided between 155 participants at both UCLA and Intermountain).

Hypotheses: The intervention will result in: 1) a greater proportion of participants achieving ≥ 5% weight loss at 24 months, 2) increased uptake of lifestyle change and/or metformin at 6 months, 3) better patient-reported outcomes (physical activity, eating patterns, patient activation, health-related quality of life), 4) annual follow-up screening for progression to T2DM.

We will use the same analytic plan as described above for the primary outcome to compare intervention and usual care in terms of the rates of ≥5% weight loss at 24 months, uptake of lifestyle change and/or metformin use at 6 months, and rates of screening for T2DM. Linear regression will be used to examine patient-reported outcomes at 6, 12, and 24 months, and we will be able to measure whether patients attended a minimum of 9 of 16 weekly lifestyle change sessions even if there was a delay in enrollment. Annual follow-up screening for progression to T2DM will include any ADA-recommended screening test for the detection of T2DM, including measuring fasting plasma glucose or HbA_1c level or an oral glucose tolerance test. The model will include the same set of predictors outlined above for the logistic regression and also adjust for baseline patient-reported outcomes.

Hypothesis: The intervention will result in lower rates of GDM in a subsequent pregnancy.

We will use the same analytic plan as described above for the primary outcome to compare the rates of GDM in a subsequent pregnancy. GDM diagnosis will be based on ICD10 codes or positive 1-step or 2-step oral glucose tolerance test. The analysis will be restricted to participants who are not pregnant at baseline or at the 6 month follow up and who are no >3 months pregnant as the 12 month follow up. This analysis will be adjusted for baseline covariates that show differential distributions between the two groups. However, this will be an exploratory outcome, and our study is not powered to assess its significance.

We will not test a hypothesis with this aim, but will rather conduct a descriptive analysis of costs related to implementation of the intervention [45]. We will express results as the mean implementation cost to the healthcare system per patient who loses 5% of their body weight.

Study data will be examined and summarized by descriptive statistics including median, mean, and standard deviation for quantitative variables and by frequencies and proportions for categorical ones. All variables of interest will be summarized for the whole sample, by clinic sites, and by study arms. We will control for baseline covariates and use either logistic or linear regression models, depending on the type of outcome variable. Graphic displays, such as boxplots and histograms, will be used to inspect skewness and normality for quantitative variables and to identify possible outliers. All randomized subjects will be included and the intention to treat principle will be applied.

Missing data will be handled using multiple imputation techniques that assume data are missing at random [46]. Analysis will be done using the complete, imputed datasets and then combined to generate the final result. If strong evidence suggests that missing data due to loss to follow-up may be nonrandom and informative, we can perform alternative statistical modeling approaches to incorporate modeling of the dropout process. We will utilize joint models for outcomes of interest and event times (e.g., time to study dropout) in order to adjust for the possibility of non-ignorable missing data [47].