Reducing Requirements for Insulin Pump Referral Improves Pump Initiation for Publicly Insured Children and Youth With Type 1 Diabetes Free

Children’s Hospital Los Angeles (CHLA) is a large, urban tertiary care pediatric hospital that receives referrals from Southern California. CHLA provides care to ∼1,800 children and youth with type 1 diabetes. The patient population is diverse in terms of race/ethnicity and socioeconomic status (58% publicly insured; 26% identify as White, 3% as Asian, 5% as Black, and 30% as Hispanic). English and Spanish are the most common spoken languages at home (85 and 12%, respectively).

Continuous subcutaneous insulin infusion pumps (hereafter referred to as insulin pumps) deliver rapid-acting insulin through an external device. Insulin pumps can be used alone to replace multiple daily insulin injections (manual mode) or interface with a continuous glucose monitoring (CGM) system with variable insulin delivery based on sensor glucose levels (automated insulin delivery [AID]). AID systems are well recognized for their ability to improve glucose profiles and reduce the burden of diabetes care and are recommended in people of all ages with type 1 diabetes (1).

Insulin pump utilization has become more prevalent, although its uptake is not equitable across insurance types or racial/ethnic groups (2,3). Insulin pump use is less common in minoritized races and ethnicities and among people with public insurance (2). These disparities were seen at CHLA. Among patients with type 1 diabetes who initiated insulin pump therapy between 2019 and 2022, 72% were privately insured (compared with private insurance status of 42% in our entire type 1 diabetes cohort). Self-described race categories were 39% White, 4% Black, 3% Asian, 0.6% American Indian or Alaska Native (AIAN), and 53% other or unknown (compared with our type 1 diabetes cohort demographic of 22% White, 5% Black, 3% Asian, 0.1% AIAN, and 70% other or unknown). Fourteen percent identified as Latino (compared with 52% identifying as Latino in the full type 1 diabetes cohort).

In 2021, the pump utilization rate was 36% among patients with type 1 diabetes at CHLA (610 of 1,739 patients), which also fell below the T1D Exchange benchmark of 65% (4). Our previous insulin pump referral process included glycemic optimization prior to pump initiation, which can prolong the time before pump starts. This process entailed weekly phone calls with the nurse care manager to review glucose log data (including carbohydrate intake and insulin doses) and insulin dose adjustments. The process can be difficult to complete for families who cannot readily initiate and answer phone calls because of their work schedule. With AID systems becoming the standard of care, this initiative was undertaken to reduce barriers to insulin pump initiation for patients with type 1 diabetes.

Data were extracted from the hospital’s electronic medical record (Cerner) and an internal flowsheet tracking insulin pump referrals. A root cause analysis was performed, and a fishbone diagram was generated to identify the most common barriers to transitions to an insulin pump (Supplementary Figure S1).

Baseline data were collected from 2019 to 2022. At baseline, of 315 patients referred, 161 (51%) successfully initiated insulin pump therapy, with a median duration of 143 days from referral to pump start. Among those who transitioned successfully to pump therapy, 71 and 38% had private and public insurance, respectively. Comparable proportions were observed among English-speaking patients (72 and 37% for private and public insurance, respectively, with a relative risk of 0.524 [95% CI 0.421–0.647] for publicly insured patients).

This 2-year QI project began in November 2021 with an initial meeting of key stakeholders to identify barriers to transitioning patients to insulin pump therapy. Redesign of the pump referral process occurred over the next 9 months, and a pilot workflow was implemented in September 2022. We implemented additional changes over the next year, with the final cohort entering the pump referral process in August 2023. The last date for data entry was 24 February 2024, when all pump referrals were reviewed for completion.

CHLA is a participating site in the T1D Exchange Quality Improvement Collaborative (5). The clinical diabetes director served as the project leader. She has oversight of the clinical diabetes workflow and knowledge of the practice culture and actively collaborates with clinicians, nursing staff, and financial counselors. The team also includes the nursing supervisor, clinical service manager, a nurse scientist, six nurse care managers, three physicians, two nurse practitioners (NPs), and two QI project coordinators. One nurse care manager was designated for the pilot process.

The standard insulin pump referral process encompassed the following steps.

• 1. Patients referred to clinic meet individually with a clinic nurse to review general concepts of insulin pump therapy. They are given instructions to record 4 weeks of diabetes logs, including blood glucose levels, insulin doses, and carbohydrate intake.

• 2. Patients’ caregivers contact nurse care managers to discuss insulin pump options and review the diabetes logs. Insulin dosing may be adjusted with weekly phone follow-up for 2–4 weeks. This process may be prolonged if diabetes logs are not submitted.

• 3. Nurse care managers coordinate pump training, obtaining orders for insulin pumps and insulin prescriptions from patients’ clinicians.

• 4. Patients and families attend an all-day, in-person pump training class, which consists of technical training in the morning (i.e., insulin pump initiation), followed by medical management training in the afternoon.

• 1. The requirement of completed diabetes logs with the intention of optimizing glycemia before pump initiation. Only a subset of patients referred for an insulin pump submit the logs.

• 2. A high patient-to-nurse care manager ratio, which limited the ability to initiate phone calls to families to complete their pre-pump assessment. In addition to adjusting insulin doses, nurse care managers also spend time discussing features of various insulin pump systems and assessing diabetes self-management skills.

• 3. A lack of uniform criteria to define when a patient is ready to initiate insulin pump therapy

• 4. A lack of flexibility in training dates. These classes were offered twice per month on the same day of the week. This schedule sometimes interfered with a patient’s existing plans.

• 1. We initiated a pilot monthly Introduction to Pump class with in-person and virtual attendance options, providing information on the benefits of and precautions regarding insulin pump therapy. After attending the class, families who are ready to proceed with pump therapy respond to a Microsoft Form questionnaire, which also serves as an assessment of the class curriculum. Only English-speaking families were included in the pilot.

• 2. We still required evidence of blood glucose monitoring (CGM reports sufficed), but families were no longer required to document carbohydrate intake and insulin doses on logs.

• 3. The nurse care manager had two scheduled phone calls with each family to review diabetes safety (e.g., when to check for ketones and whom to call), current diabetes self-care plans, and the choice of an insulin pump. We eliminated repeated weekly phone calls (generally two to four calls) to review insulin doses. These two calls were scheduled to reduce the frequency of missed phone calls between the nurse and family.

• 4. Families were sent a video recording of the insulin pump medical management presentation, which historically was given to patients who attended the in-person pump training. Families completed a knowledge assessment after watching the video.

• 5. NPs conducted virtual visits before and after pump initiation. They wrote the pump setting orders, sent prescriptions for insulin vials, and coordinated pump training, all of which offloaded work previously performed by the nurse care managers. NPs also addressed any gaps in knowledge that families had before and after pump initiation.

• 6. Pump training was switched from in-person group classes to virtual training by vendors to offer more flexibility.

The most impactful changes were:

• 1. Eliminating our internal requirement for families to provide detailed diabetes logs. Some families are unable to provide detailed logs of carbohydrate consumption and insulin doses, especially when school nurses provided diabetes care at school and the details were not readily available to patients’ caregivers.

• 2. Changing pump training from in-person to virtual training. Families coordinated a time with the pump trainer that worked for their schedule, and multiple family members could attend the virtual training at the same time.

• Plan. Prior to PDSA cycle 5, nurse care managers needed to initiate phone calls to families to ensure that they were sharing CGM data with the clinic and to conduct pre-pump assessments. Typically, multiple phone calls were needed because of the nurse and families missing each other’s calls. In PDSA cycle 5, a clinical care assistant called families after the Introduction to Pump class to confirm that they were sharing their CGM data with the clinic and scheduled two phone calls with the nurse care manager.

• Do. This process was implemented for PDSA cycle 5.

• Study. PDSA cycles 4 and 5 each had six patients who wished to proceed with pump initiation. The mean time to completing the pre-pump assessment was 36 ± 11 days in cycle 4 and 12 ± 4 days in cycle 5 (Student t test P <0.0001).

• Act. As a result of the improved efficiency of this process, scheduled phone calls were adopted as a routine practice for all subsequent cohorts. The next modification included implementing Introduction to Pump classes as a standing monthly class.

We compared pump completion data from patients referred to the pump pilot process with those referred to the standard process between 1 September 2022 and 31 August 2023. We chose to use a contemporaneous control group instead of baseline data because of factors that could influence insulin pump therapy over time, including the availability of AID systems in commercially available pumps, insurance coverage of these devices, and staffing shortages.

Only English-speaking families were included in this analysis. All patients used CGM at the time of pump referral. Pump initiation occurred for 39% of standard care versus 55% of pump pilot patients (P = 0.080). We noted a significant disparity in pump completion rate by insurance status. As shown in Supplementary Table S1, among publicly insured patients, pump transition occurred for 51 and 20% of patients enrolled in the pump pilot and standard care, respectively (P = 0.004). The corresponding pump completion rates were 68 and 76% for publicly and privately insured patients, respectively (P = 0.898).

Despite the differences in overall pump completion rates, the time between referral for pump initiation was unchanged between the two arms. The median times to pump initiation were 83 and 74 days, for the pilot and standard care groups, respectively (Mann-Whitney U test, P = 0.991).

The standard process entailed in-person training, whereas the pilot process included an asynchronous medical management video and a synchronous technical pump training session. Retention of the insulin pump medical management presentation was assessed using a 15-question knowledge assessment. There was no difference in percentage of correct answers between families who attended the pump class in person and those who watched the video asynchronously (standard 76% [n = 16], pilot 74% [n = 41], P = 0.459).

Glycemic status was compared between the pilot and standard care arms using A1C values, glucose management indicator (GMI) data, and CGM 30-day time in range (TIR) values before and after pump initiation. Compared with standard care, patients in the pilot arm had a higher median pre-pump A1C (standard 7.2% [95% CI 6.7–7.5%], pilot 7.9% [95% CI 7.8–8.9%], P = 0.001) and higher median GMI (standard 7.2% [95% CI 7.1–7.7%], pilot 8.0% [95% CI 7.7–8.6%], P = 0.006) and lower median TIR (standard 61% [95% CI 54–70%], pilot 36% [95% CI 34–49%], P = 0.0004) (Supplementary Figure S2). There were no significant differences between the groups in post-pump A1C (standard 7.0% [95% CI 6.6–7.3%], pilot 7.4% [95% CI 6.7–7.9%], P = 0.188), GMI (standard 7.2% [95% CI 6.9–7.4%], pilot 7.2% [7.1–7.6%], P = 0.593), or TIR (standard 66% [95% CI 64–75%], pilot 65% [95% CI 56–69%], P = 0.258). Patients in the pilot arm experienced a greater improvement in TIR (standard 7% [95% CI 2.6–13%], pilot 24% [95% CI 13–27%], P = 0.014). These results suggest that pre-pump glycemic optimization is not necessary because AID systems can rapidly improve glucose profiles.

The major strength of this QI project was its ability to reduce the inequity in pump access for our publicly insured patients. This new process eliminated the requirement of detailed glucose logs and glucose optimization prior to pump initiation. It added an asynchronous training video and knowledge assessment to maximize safety, with comparable scoring compared with in-person training. Virtual pump initiation is scheduled at times convenient for families, minimizing the time families need to take off from school and work. Collectively, this process narrows the gap in pump access between families of different socioeconomic status. Expanding access to AID systems allows people with type 1 diabetes to optimize their TIR while reducing their burden of care, which has the potential to improve their quality of life and reduce long-term comorbidities.

In its current design, the pump QI process did not shorten the duration between pump referral and initiation. We speculate that reasons for the lack of change in duration before pump initiation may include:

• 1. A potentially more protracted process for pump training. Families are asked to watch a pump training video and complete a knowledge assessment. They next have a pre-pump virtual appointment with the NP to address concerns and determine pump settings. Families then individually schedule a training time with the pump vendor, which may sometimes be further into the future than the in-person group training schedule would have been.

• 2. Families who complete the pump transition through the standard process tend to be highly motivated families who submit glucose logs and are proactive in working with the nurse care managers. This underestimates the time for pump initiation because families who have difficulties with this process would never complete a pump transition. The duration of pump transition time of the standard process therefore excluded families who did not submit glucose logs and reflected only the completion time for highly motivated families.

We are developing several changes to adapt the QI process as the new standard of care for pump referrals. We have increased the cadence of the Introduction to Pump class to once a week and have incorporated the entire team of nurse care managers into the process. To address the other inequity of decreased pump usage for our non–English-speaking families, we are in the process of developing a similar workflow for Spanish speakers (the most commonly spoken language in our population after English). We will also develop a workflow that allows families to determine whether they wish to complete the pump training virtually or in person at CHLA’s group class.

We recognized early on that a process change that affects all members of our diabetes team requires the active engagement of all stakeholders. Gathering the perspectives of the team members allowed us to build a workflow that incorporated the most salient aspects of the process, which included providing patient education, ensuring safety, and reducing requirements, during the pump transition redesign. Updating stakeholders with data from the process also helped to keep them engaged. Ultimately, focusing on the purpose of the project—to improve outcomes for patients with diabetes—helped to center the team on creative problem-solving.

The study team thanks the CHLA diabetes nurse care manager team for their feedback and support and Brenda Gallardo and Kathleen Carney for support in the development and implementation of this work.

This work was supported by the T1D Exchange Quality Improvement Collaborative.

No potential conflicts of interest relevant to this article were reported.

B.M. contributed to the study design and extracted and analyzed the data. J.B. contributed to study design and implemented the study. J.C. and J.L.A. implemented the study. J.A. contributed to data extraction and analysis. C.B. and J.F. contributed to the discussion. D.M. contributed to the study design and supervised the implementation. O.O. critically reviewed and edited the manuscript. R.O.L.B.B. contributed to the study design. L.C.C. designed the study, analyzed the data, and wrote the manuscript. B.M. and L.C.C. are the guarantors of this work and, as such, had full access to the data reported and takes responsibility for the integrity of the data and the accuracy of the data analysis.

This work was previously published in abstract form in the Journal of Diabetes 2023;15(Suppl. 1):4–31.