This BPA-TECH study will use mixed methods and nonrandomized controlled trial designs (ie, controlled before-after study) to implement and evaluate a BPA aimed at reducing inequities in the use of ADTs among children and adults with T1D. Phase 1 will be a qualitative mixed methods study to develop, implement, and refine the BPA. Phase 2 will be a prospective nonrandomized matched pair intervention design to evaluate the effectiveness of the BPAs in promoting prescribing and thus patient use of ADTs. The study organizing committee will be represented jointly by principal investigators in the T1DX-QI coordinating center and Johns Hopkins University, while the T1DX-QI will be responsible for the data management and analysis. We adhered to the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) guidelines (Multimedia Appendix 1) in the development of this study protocol and all items from the World Health Organization Trial registration dataset are included [31,32]. Protocol amendments will be communicated during monthly meetings between the organizing committee and the participating clinical sites, and updates will be posted on ClinicalTrials.gov. The timeline for this 3-year study is shown in Figure 1.

In 2016, through financial support from the Leona M. and Harry B. Helmsley Charitable Trust, T1D Exchange established the T1DX-QI, which initially included 10 diabetes and endocrinology centers focused on population health by advancing quality improvement, health equity, and real-world studies. The network has since expanded to include 62 T1D and 20 type 2 diabetes (T2D) centers, along with partners in primary care.

T1DX-QI manages more than 10 multicenter projects annually, with 4-10 centers participating per project, and has expanded its scope to include equitable access to diabetes technology and devices for both T1D and T2D populations.

In phase 1 of the study, we will recruit health care providers specializing in diabetes management from academic hospitals affiliated with the T1DX-QI. People with T1D or their caregivers will be recruited from the T1DX online registry, which includes individuals across the United States who may receive care in community clinics or academic hospitals. Phase 2 of the study will also encompass participating clinical sites affiliated with the T1DX-QI, specifically academic medical centers.

Using a nonrandomized matched pair-intervention design (ie, controlled before-after study), we will compare the use of ADT following the BPA intervention among non-Hispanic Black and Hispanic people with T1D at 7 participating T1DX-QI centers. These will be matched with control non-Hispanic Black and Hispanic people with T1D at nonintervention T1DX-QI centers over a 12-month period. Both control and intervention centers have completed the EMR data mapping to the T1DX-QI, allowing for consistent outcome assessment across both groups. The study will consist of 2 distinct periods:

The before-after matched-pair intervention design allows for a real-world assessment of the BPA’s impact on ADT use while controlling for key confounders. Comparing ADT prescribing before and after BPA implementation within intervention sites and using matched controls from nonintervention centers, the design also accounts for site-specific prescribing patterns and secular trends. This quasi-experimental approach is practical for an EMR-based intervention and minimizes selection bias by matching key clinical and demographic variables.

Inclusion and exclusion criteria are listed in Textbox 1.

People with T1D meeting the eligibility criteria at one of the 7 intervention centers where the BPAs are deployed will be assigned to the intervention group. We estimate that there are approximately 11,496 eligible people with T1D in the intervention centers.

Matched control people with T1D will be selected from the pool of nonintervention clinical centers sharing EMR data with the T1DX-QI coordinating center with over 80,000 people with T1D data available in the database.

To accurately match people with T1D between the intervention and control groups, we will identify and include a comprehensive set of covariates that may influence the likelihood of receiving the BPA. These covariates will include both patient-level and clinical center-level characteristics. Patient-level characteristics will include demographics (age, sex, race, and ethnicity), socioeconomic status (insurance type and zip code–based income), and clinical characteristics (duration of diabetes, comorbidities, HbA_1c levels, etc). Clinical center-level characteristics will include clinic size and resources (number of people with T1D seen annually, number and type of clinical full-time equivalents, geographic location, patient population demographics, and historical usage rates of CGMs and AID systems. Intervention and control groups will be matched using the propensity score method.

Before propensity score matching, any missing baseline covariate data will be imputed using predictive modeling techniques. Multiple imputation by chained equations will be used to generate multiple complete datasets, which will then be averaged to create a single dataset for the matching process. This approach ensures that the imputed values are based on observed data, maintaining the integrity and representativeness of the dataset.

Propensity scores will be estimated using logistic regression, where the outcome variable is the likelihood of receiving the BPA intervention. The model will include the identified patient-level and clinical center–level covariates as predictors. The estimated propensity score represents the predicted probability that a given patient would receive the BPA based on their covariate profile. Patients in the intervention group will be matched to patients in the control group using a 1:4 matching ratio. Nearest neighbor matching without replacement will be used, ensuring that each control patient is only matched once. The matching process will aim to minimize the distance between propensity scores of matched pairs, thus creating comparable groups. The quality of the matching process will be assessed by comparing the distribution of covariates between the matched intervention and control groups. Standardized mean differences will be calculated for each covariate to ensure balance, with a standardized mean difference of less than 0.1 indicating adequate balance. In addition, a visual inspection of propensity score distributions before and after matching will be conducted to confirm the success of the matching process.

To assess the robustness of the results, sensitivity analyses will be performed using different matching algorithms (eg, caliper matching and kernel matching) and varying the matching ratio (eg, 1:3). In addition, the impact of imputation on the results will be examined by comparing outcomes using datasets with and without imputed values.

For nonstatic matching variables (eg, laboratory results), the last available result in the baseline period of 12 months before the BPA deployment date in the intervention centers will be used.

The primary outcome will be progression in ADT use (as documented in the EMR) during the 12-month study period. For both intervention and control centers, a participant’s baseline ADT status will be ascertained in the 12-month period before deployment of the BPA intervention. The primary outcome will be defined as positive for an individual if any of the following transitions from the baseline period to the implementation period occur with respect to ADT:

The T1DX-QI database includes variables that will be used to evaluate the study outcome (eg, medication list and flowsheet data). We have extensively published on the T1DX-QI data source and its validation, and have used the database for several real-world studies [36-42].

It is not anticipated that the intervention will pose any additional risk of harm to people with T1D, particularly given the established benefits of AIDs and CGMs in enhancing glycemic control and reducing the risk of hypoglycemia. Nevertheless, the study will closely monitor rates of serious hypoglycemic events and DKA events, using existing data mapped from control and intervention sites to the T1DX-QI database. These rates will be compared between the intervention and control sites to ensure that there are no unexpected or unintended consequences of the intervention on adverse glycemic outcomes.

This intervention study was registered on ClinicalTrials.gov (NCT06931275) and received approval from the WCG institutional review board on April 26, 2024 (institutional review board approval number 45614667). For the qualitative research proposed in phase 1 of the study, oral informed consent will be obtained from all participants (patients, caregivers, and health care providers) for the focus groups and interviews, and written informed consent will be obtained from all participants (patients, caregivers, and health care providers) for the surveys. However, due to the nature of this implementation study, informed consent cannot be obtained from participants during the observation phase (phase 2), which will evaluate the impact of a before-and-after design on clinical outcomes.

The overarching goal of this study will be to promote the use of ADT and enhance health equity in ADT use for diabetes management in people with T1D. While previous research has illustrated the many barriers and racial disparities in ADT use, the focus of this initiative is to develop a comprehensive and evidence-based solution using HIT to address and mitigate ongoing disparities in ADT use and outcomes. We expect that clinical centers implementing the BPA will demonstrate higher rates of progression to ADT use compared with matched controlled sites, as well as a smaller gap in the use of these technologies between White and non-White patients.

In a 2-phased approach, this protocol aims to develop a BPA with input from key stakeholders, and then evaluate the effectiveness of the BPA in reducing disparities in ADT use among people with T1D. The goal of this intervention is to increase the progression of ADT use overall, while reducing health disparities and improving outcomes in all people with T1D. In addition, reasons for not prescribing ADT in response to the BPA will be collected. If successful, the build for this BPA can be disseminated to other hospital systems for implementation to improve outcomes on a larger scale.

To our knowledge, this is one of the first studies to develop and use a CDS initiative in the EMR to promote equitable discussion and prescribing of ADT in people with T1D. In addition, to increase the generalizability of our findings, a diverse sample of People with T1D will be included from multiple centers across the United States to account for geographical and systems-level (ie, state-level and insurance-based) variabilities.

Previous initiatives to improve the use of diabetes technologies in people with T1D have focused on patient-level interventions. Early initiation of CGM soon after diagnosis of T1D, and point-of-care CGM placement and education during routine diabetes care visits have led to increased uptake of CGM and improved glycemic outcomes [45-47]. A current randomized control trial deploying a diabetes navigator will determine if the help of a technology liaison can improve uptake and sustained use of ADT in people with T1D [48].

The use of a BPA as a CDS tool in this protocol serves as an intervention at both the provider and system level, enabling greater scalability to create lasting improvements in ADT uptake and glycemic outcomes. There is previous evidence supporting the successful implementation of computer-based decision support in diabetes care [49-51]. One cluster randomized clinical trial consisting of 14 primary care centers (66 primary care physicians and 697 patients with T2D on insulin therapy) in Madrid, Spain successfully implemented a computer application designed to help primary care physicians make decisions about insulin therapy [52]. The design of the algorithm included the recommendation to change the insulin dose and the insulin regimen when necessary. The people receiving care at the 7 centers randomly assigned to use the algorithm had a significant reduction in HbA_1c compared with the people at the 7 centers in the control group [52]. Another meta-analysis of 70 randomized controlled trials assessing the critical features of CDS systems that improve clinical practice highlighted 4 initiatives strongly associated with a successful center decision support system: (1) CDS provided automatically as part of clinical workflow, (2) CDS delivered at the time and location of decision-making, (3) actionable recommendations provided such as prescriptions, and (4) computer-based [53]. In this protocol, the BPA is designed to incorporate these 4 critical features and thus we expect to see improvement in ADT uptake at the intervention sites.

This study leverages the T1DX-QI consortium, which has a robust quality improvement and health equity framework to evaluate the impact of an EMR-based BPA on ADT use and racial disparities. By implementing the intervention at 7 motivated sites with strong research and QI expertise, and strong IT infrastructure, the study ensures effective troubleshooting and optimization of the BPA. This real-world, multicenter approach enhances generalizability and supports the potential for scalability to other centers if successful. In addition, integrating health equity principles strengthens the study’s ability to assess disparities and inform broader systemic improvements in diabetes care.

Although this study is limited to centers using the Epic EMR system to streamline the development of the BPAs, the BPA prototypes developed in this study could be adapted for other EMR systems in the future. Further, this study is limited to centers participating in the T1DX-QI, where there is likely to be greater motivation to improve outcomes and close care gaps, and thus findings may not be generalizable to all centers providing diabetes care. However, the use of BPAs allows standardization of practice and can serve as an equalizer across different practices.

The findings will be published in diabetes, health technology, and health equity–focused medical journals, and will be presented at major national and international conferences. Patient-friendly summaries will be developed and shared with patients and caregivers, while educational materials and training programs will be created for health care providers to facilitate clinical implementation on a larger scale. Authorship eligibility for publications resulting from this study will adhere to the International Committee of Medical Journal Editors (ICMJE) guidelines [54].

The research approach outlined in this protocol aims to leverage health IT to reduce racial disparities in prescribing and use of ADTs, thereby promoting equity, improving disease management, and enhancing outcomes in people with T1D.