Preventing Severe Hypoglycemia in Adults with Type 2 Diabetes (PHT2): Design, Delivery and Evaluation Framework for a Randomized Controlled Trial

Abstract

Background:

Severe hypoglycemia is a common and feared complication of medications used to lower blood glucose levels in individuals with diabetes. Psycho-educational interventions can prevent severe hypoglycemia in individuals with type 1 diabetes (T1D). We aim to determine the effectiveness of this approach among adults with type 2 diabetes (T2D) at elevated risk for severe hypoglycemia.

Methods:

Preventing Hypoglycemia in Type 2 diabetes (PHT2) is a two-arm, parallel, randomized controlled trial. Participants are eligibile if they are adults with T2D receiving care at an integrated group practice in Washington state and have experienced one or more episodes of severe hypoglycemia in the prior 12 months or have impaired awareness of hypoglycemia (Gold score ≥ 4). Participants are randomized to proactive nurse care management with or without my hypo compass, an evidence-based, psychoeducational intervention combining group and individual self-management training. For this study, my hypo compass was adapted to be suitable for adults with T2D and from an in-person to a virtual intervention over videoconference and telephone. The primary outcome is any self-reported severe hypoglycemia in the 12 months following the start of the intervention. Secondary outcomes include biochemical measures of hypoglycemia, self-reported hypoglycemia awareness, fear of hypoglycemia, and emergency department visits and hospitalizations for severe hypoglycemia. The study includes a process evaluation to assess implementation fidelity and clarify the causal pathway.

Conclusion:

The PHT2 trial will compare the effectiveness of two approaches for reducing severe hypoglycemia in adults with T2D.

Background

Severe hypoglycemia is a blood glucose low enough to require help from another person for recovery and is a common, feared complication of blood glucose-lowering medications. Severe hypoglycemia is associated with accidents,¹ falls,^{2, 3} chronic and acute cardiovascular disease,^4–9 dementia,^10–12 and increased risk of death.^13–16 Older adults are particularly susceptible to injury from severe hypoglycemia due to coexisting frailty and other medical conditions.¹⁷ A systematic review of longitudinal studies found severe hypoglycemia reduces a person’s confidence in their diabetes self-management and perceived health quality, while frequent hypoglycemia reduces energy levels and emotional well-being.¹⁸

Severe hypoglycemia among adults with type 2 diabetes (T2D) is a growing concern for healthcare organizations and policymakers. Of the more than 37 million US adults with diabetes, 90–95% have T2D.¹⁹ Severe hypoglycemia affects 10% of these adults, with minimal difference in relative risk across HbA1c levels. ²⁰ Among older Americans, up to 25% of emergency hospitalizations for adverse drug events involve hypoglycemia related to insulins and oral hypoglycemia agents.²¹

Impaired awareness of hypoglycemia (IAH) and delayed or insufficient treatment of low blood glucoses are major modifiable risk factors for severe hypoglycemia. IAH develops after repeated or sustained exposure to hypoglycemia.²² Individuals with intact awareness can recognize their hormonal and/or neurologic response to low blood glucose, which manifest as symptoms such as weakness, shakiness or sweating. Individuals with IAH have lost or have reduced capacity to recognise the onset of hypoglycemia due to symptoms being reduced in intensity, experienced only at lower blood glucose levels than previously, or (rarely) completely absent. IAH occurs in approximately 10% of individuals managing their T2D taking insulin, and is associated with 17-fold higher risk of severe hypoglycemia compared to those with intact awareness.²² T2D tends to occur in older adults, who are less likely than younger people with T2D to experience typical hypoglycemia symptoms, and they may mistakenly attribute the weakness or unsteadiness to other causes.¹⁷ The National Committee for Quality Assurance, recognizing the importance of severe hypoglycemia risk in this population, added emergency department visits for hypoglycemia in older adults with diabetes as a healthcare quality measure in 2023. ²³

Structured diabetes education and pyscho-education may enable preventive action, avoid severe hypoglycemia, and help restore hypoglycemia awareness in individuals with T2D. However, no established training programs exist for individuals with T2D and IAH that focus on reducing risk of hypoglycemia. In individuals with type 1 diabetes (T1D) at high risk for severe hypoglycemia, structured education, behavioural training and psycho-education programs restores hypoglycemia awareness, facilitates effective action to reverse hypoglycemia and reduces severe hypoglycemia rates.^24–28 This training differs from standard diabetes self-management education, focusing specifically on the challenges of IAH and severe hypoglycemia associated with blood glucose-lowering medications. Many individuals with T2D at high risk for severe hypoglycemia share characteristics with individuals with T1D at high risk such as multiple daily insulin injections, higher IAH rates and a need to balance insulin dose against carbohydrate intake and activity.²⁹ Literature reviews and diabetes care leaders call for prospective trials to prevent severe hypoglycemia in high-risk individuals with T2D.³⁰

Objectives

The aim of PHT2 (The Preventing Severe Hypoglycemia in Type 2 Diabetes) study is to compare the effectivness of two practical, evidence-based approaches to preventing severe hypoglycemia among adults with T2D who had had at least one severe hypoglycaemic event in the prior 12 months and/or have IAH. The study includes a mixed-methods process evaluation to evaluate implementation fidelity of intervention components and to clarify the causal pathway if intervention effects are found.

Methods

Design:

PHT2 is a parallel-group, individually randomized controlled trial. Two focus groups of adults with T2D and their family members consulted on study design and implementation. Four focus group members are co-investigators. As recommended by the American Diabetes Association (ADA),³¹ the intervention design was guided by the Chronic Care Model to ensure suitability for routine clinical care³² (Figure 1). The study is registered with clinicaltrials.gov and approved by the Kaiser Permanente Washington (KPWA) Institutional Review Board and the study’s Data Safety Monitoring Board (DSMB).

Figure 1: Organizational approach for reducing severe hypoglycemia.

Participants and Setting

The study setting is KPWA, an integrated care-and-coverage system with 34 owned-and-operated clinics for nearly 700,000 members. Nurses trained in diabetes care and working within the KPWA healthcare organization deliver the study interventions with separate nurses for each study arm.

Eligibility Criteria

Potentially eligible participants are identified using the electronic health record (EHR) and then called to invite to join the study and further assess eligibility. Individuals eligible for further telephone assessment must meet all inclusion criteria at baseline: aged ≥18 years, T2D diagnosis, insulin treatment or intermediate-to-high risk for a severe hypoglycemia episode based on a hypoglycemia risk-stratification tool,^33–35 and receiving primary care at a KPWA medical center. Individuals must report that they plan to continue a KPWA health plan for the next 6 months and they had a severe hypoglycemia episode in the prior 12 months or have IAH (based on Gold score ≥ 4).^{36, 37} The original trial protocol was restricted to individuals at least 50 years; we expanded criteria to 18 years or older to increase the recruitment pool and generalizability of study results.

Exclusion based on EHR data is for severe psychiatric conditions with psychosis, severe cognitive impairment, current nursing home residence or hospice care, current continuous glucose monitor (CGM) use, or need for an interpreter. Exclusion at phone contact is for inability to give informed consent or speak or read English; inability or unwillingness to attend or complete online or telephone training sessions, follow-up calls, or outcome assessments; or pregnancy or plans to become pregnant.

Participant Screening and Recruitment

Recruitment started in January 2022. To better represent the diversity of the U.S. population, we oversample those whose EHR shows self-identified race other than white or ethnicity as Hispanic. Potential participants identified using the EHR are mailed an invitation letter, study brochure, $2 preincentive and study consent form. They are then called to receive an explanation of the study, assess their interest, obtain verbal consent, and collect baseline self-report data. Those agreeing to participate complete either an electronic or a mailed consent form and receive a letter summarizing next steps.

To maximize retention, participants receive a $100 incentive for complete data collection at baseline and 14 months ($30 for survey and $70 for 10 days of CGM) and $30 for self-report-only measures at 6 and 10 months, with an additional $25 for having an HbA1c for the study at 14 months.

Interventions

Proactive Care Management:

KPWA nursing staff trained in diabetes care deliver an evidence-based care intervention to prevent hypoglycemia based on ADA guidelines.^{38, 39} This intervention is designed to be similar to care from diabetes nurses but delivered preventively rather than in response to a recent severe hypoglycemia event. For participants with intensive blood glucose targets (e.g., HbA1c <7% with history of severe hypoglycemia or multiple medical conditions), or IAH (Gold score ≥ 4),^{36, 37} nurses work with physicians to adjust blood glucose targets, deintensify hypoglycemic medications, enhance blood glucose self-monitoring, provide glucagon kits and training, or refer for consideration of CGM.^{31, 40} For participants on blood glucose-lowering agents who need help with self-monitoring, nurses provide basic coaching on hypoglycemia symptoms and actions such as safely raising blood glucose, referring to dieticians as needed.^{39, 41} Nurses ensure standard primary care at KPWA including regular assessment and follow-up for mental health and social needs that may contribute to severe hypoglycemia risk.³⁰

Proactive Care Plus my hypo compass:

This intervention group receives proactive care management plus an adapted version of my hypo compass, an evidence-based, psychoeducational intervention combining group and individual self-management training that is associated with restoring hypoglycemic awareness and reducing severe hypoglycemia in individuals with T1D.^{24, 42} For this study, the my hypo compass protocol was adapted to adults with T2D and for delivery over videoconferencing and phone. To reduce potential contamination across comparator groups, different nurses with the same training and experience perform proactive outreach management for the two trial arms, Proactive Care Management and Proactive Care plus my hypo compass. The my hypo compass curriculum is delivered by diabetes nurse educators. The curriculum is standardized in facilitator and participant handbooks and uses behavioral intervention techniques to facilitate discussions on strict avoidance of hypoglycemia. The curriculum uses four key principles around compass points, North, South, East and West (Appendix A): Never delay appropriate treatment of hypoglycemia; recognize hypoglycemia by Subtle symptoms; Establish times of Extra risk; and be Watchful to detect and prevent hypoglycemia. Emphasis is on detection, including asymptomatic nocturnal hypoglycemia, with advice to perform a weekly 8-point blood glucose profile including a 4 a.m. check.

The individual-level causal model for the my hypo compass intervention is informed by the Biopsychobehavioral Model of severe hypoglycemia,⁴³ Social Cognitive Theory, ⁴⁴ and Self Determination Theory (Figure 2).⁴⁵

Figure 2: PHT2 Study Model for my hypo compass intervention.

The my hypo compass intervention facilitates reflection on past cognitions and behaviors ,and generation of an individualized self-management plan tailored for a person's risk factors. This activity increases awareness of hypoglycemia and acknowledgement of the importance of hypoglycemia self-monitoring, enhancing behavioral skills for monitoring and management. The gained knowledge is intended to increase confidence, motivation and self-efficacy to self-manage blood glucose to avoid severe hypoglycemia events. We hypothesize these improvements will lead to better management of severe hypoglycemia through recognition of low glucose and early and appropriate self-treatment, resulting in fewer severe hypoglycemia episodes (primary outcome). A process evaluation will determine the elements and pathway of the causal model.

My hypo compass intervention activities include a blood glucose/hypoglycemia diary and two group behavioral-awareness training sessions interspersed with two one-on-one nurse calls for review and reflection (Figure 3). Two diabetes nurse educators trained on the my hypo compass protocol lead group training of participants. A research staff member attends the group sessions to provide technical support. A nurse care manager also attends the sessions to document additional clinical information relevant for patient care. The nurse care managers are trained on the my hypo compass protocol and deliver the one-on-one nurse calls for review and reflection. Hard-copy handbooks with full course content and personal worksheets are mailed to all participants before the first group training. Details of the group training sessions, one-on-one calls and fidelity monitoring are in Appendix A; Facilitator Handbook is in Appendix B; and Participant Handbook is in Appendix C.

Figure 3: my hypo compass self-management support activities.

The original my hypo compass protocol, developed by members of the research team (Shaw, Speight), was designed for individuals with T1D. Recommended changes for adapting to adults with type 2 diabetes and a virtual format were sendorsed by the clinicians and investigators who developed and validated my hypo compass with T1D and are detailed in the intervention activities description (Appendix A).

Randomization

After identifying potentially eligible individuals based on EHR data, verifying eligibility via phone, documenting written consent, and collecting baseline self-report data, participants are randomized to one of two trial arms using a 1:1 allocation ratio. Eligible individuals are randomized based on sequential assignments from randomization sequences generated by the study biostatistician and stored on a computer inaccessible to study staff. The randomization list uses permuted blocks of randomized size 4 or 8, stratified by age (<75 versus 75+ years) and risk score for severe hypoglycemia (moderate versus high risk), to ensure balance of these important risk factors across arms. Participants in the moderate risk category include those with a hypoglycemia risk stratification³⁴ score of moderate or taking insulin regardless of score.

Masking

Research staff collecting self-report and CGM data will not know participants' treatment assignment. Participants, nurses, and physicians engaged in the interventions, and study staff conducting fidelity monitoring or collecting qualitative data cannot reasonably be masked to the study groups.

The study programmer and biostatisticians have access to unmasked data to conduct data summaries by intervention group as necessary for reporting to the DSMB. The principal investigator and other study investigators will remain masked until all study data for the relevant analyses are collected.

Data Collection

During the recruitment call, staff collect baseline patient-reported outcome measures (Table 1) by telephone following oral consent. Following return of written consent, we mail participants a single-use sensor-transmitter CGM unit (Dexcom G6 Pro). A study nurse calls the participant to help place the device and checks in by telephone during a 10-day recording period. Participants mail the device back at the end of the measurement period. If a participant’s CGM sensor does not have at least 70% of scheduled measures (every 5 minutes) over the 10-day period, we ask the participant to repeat the monitoring and we mail a new sensor-transmitter unit. CGM measures of blood glucose levels are masked to participants until the end of the study.

Table 1:

Study Outcomes, Measurement Tools and Time Points

Outcome	Measurement Tool	Assessment Timepoints
		0m	6m	10m	14m
Primary
Severe hypoglycemia events in prior 12 months*	Self-report via single item from Diabetes Care Profile47	X			X
Secondary
Percent time glucose below range <54 mg/dL*	Continuous Glucose Monitor, masked	X			X
Percent time glucose below range <70mg/dL	Continuous Glucose Monitor, masked	X			X
Number of Level 1 hypoglycemic events (glucose: 54 – 69 mg/dL) duration ≥15 minutes	Continuous Glucose Monitor, masked	X			X
Number of Level 2 hypoglycemic events (glucose <54 mg/dL) duration ≥15 minutes	Continuous Glucose Monitor, masked	X			X
Number of nocturnal (between midnight and 6 am) hypoglycemic events (<70 mg/dL) duration ≥15 minutes	Continuous Glucose Monitor, masked	X			X
Standard deviation of glucose	Continuous Glucose Monitor, masked	X			X
Coefficient of variation for glucose (%CV)	Continuous Glucose Monitor, masked	X			X
Glucose management indicator	Continuous Glucose Monitor, masked	X			X
Percent time glucose in range: 70 to 180 mg/dL	Continuous Glucose Monitor, masked	X			X
Percent time glucose above range >180 mg/dL	Continuous Glucose Monitor, masked	X			X
Percent time glucose above range >250 mg/dL	Continuous Glucose Monitor, masked	X			X
Average blood glucose, mg/dL	Continuous Glucose Monitor, masked	X			X
Severe hypoglycemia events in prior 4 months	Self-report via single items from Diabetes Care Profile47	X	X	X	X
Number of symptomatic hypoglycemia events in prior 4 weeks	Self-report via Diabetes Care Profile47	X	X	X	X
Number of nocturnal hypoglycemia events (defined as “during sleep”) in past 6 months	Self-report via Hypoglycemia Awareness Questionnaire (HypoA-Q)51	X	X	X	X
Impaired awareness of hypoglycemia	Self-report via Gold Survey (score ≥ 4) 36, 37	X	X	X	X
Fear of hypoglycemia (total and subscales of worry, avoidance, and maintenance)	Self-report via Hypoglycemia Fear Survey II, short form52	X	X	X	X
Diabetes knowledge	Self-report via Diabetes Knowledge Test53	X	X	X	X
Diabetes self-efficacy	Self-report via Diabetes Self-efficacy Scale54	X	X	X	X
Hemoglobin A1c in past 90 days, %	Electronic health record	X			X
Hypoglycemia-related healthcare utilization		X
Outpatient visits	Electronic health record	X			X
Emergency department/ Urgent Care	Electronic health record	X			X
Hospitalizations	Electronic health record	X			X

m=months following randomization

^*study powered to detect significant difference

Figure 4 shows the data collection points and their relation to the intervention activities. Follow-up self-reported data are collected by telephone at 6, 10, and 14 months after baseline (4, 8, and 12 months after intervention start) to identify potential early and later intervention effects.⁴² At the 14-month time point, we gather the primary outcome data, encompassing a 12-month measurement window that excludes the 2 months after randomization but before intervention start. This approach captures complete treatment effect and prevents bias in treatment effect estimation. The timing of the 6-month interim measure is informed by studies showing that hypoglycemia symptoms return as soon as 3 to 4 weeks after avoiding hypoglycemia⁴⁶ and the my hypo compass intervention has early impacts on the percent time with blood glucose < 54 mg/dL in individuals with T1D.²⁴ At 14 months after baseline, participants collect 10-day CGM measurements using the same procedure as baseline measurements. Additional details on data sources and data collection methods are in Appendix D.

Figure 4: Outcome Measures and Study Activities.

Outcomes

The primary outcome is self-report of severe hypoglycemia occurrence in the 12 months following initiation of the intervention. This is assessed using a validated question from the Diabetes Care Profile: “In the last 12 months, have you had a low blood glucose that resulted in passing out or needing help from someone else? (For example, you were unable to treat yourself, were unconscious or were given glucagon or intravenous glucose).”^47–49 If the participant answers, “Yes” to this question, we then ask, “How many times did this happen in the last 12 months?” Secondary outcomes are assessed through CGM,⁵⁰ EHR, and self-report surveys by phone (Table 1).

Analysis Plan

Comparative effectiveness analyses will apply intent-to-treat principles, with study participants analyzed based on randomization group, regardless of intervention received. We will summarize demographic characteristics and responses to baseline surveys by group. The primary outcome, occurrence of severe hypoglycemia, is assessed as a binary indicator for whether or not the participant self-reported having a severe hypoglycemia event in the prior 12 months. We will estimate the relative risk of severe hypoglycemia in the Proactive Care Management group versus the Proactive Care plus my hypo compass group by fitting a modified Poisson regression model with the binary self-report measure as the dependent variable and randomization group as the independent variable.⁵⁵ Models will adjust for age, sex, risk score, and number of self-reported severe hypoglycemic events in the 12 months prior to baseline. Models will be fit using generalized estimating equations (GEE) and robust variance estimation to account for correlation due to patient clustering in clinics⁵⁶ and the misspecified mean-variance relationship in using Poisson regression for a binary outcome.⁵⁵

Prior to conducting primary outcome analyses, we will look at the baseline distribution for the number of severe hypoglycemic events in the 12 months prior to baseline. If more than 60% of enrolled participants report 2 or more events in this time period, we will revise the primary outcome to be defined as the self-reported number of severe hypoglycemic events in the 12 months following intervention (i.e., a count rather than a binary outcome), and fit a Poisson regression model to estimate the rate ratio for severe hypoglycemia events for the my hypo compass group relative to the care management only group. We will assess model fit, and consider zero-inflated models as an alternative if the number of participants reporting zero events violates distributional assumptions of the Poisson model. If primary analyses use the binary outcome measure (i.e., because fewer than 60% of participants report 2 or more events in the 12 months prior to baseline), but 40% or more of respondents report 2 or more events at 14 months, we will conduct a secondary analysis with the count variable as the outcome.

Generalized linear regression models estimated with GEE and robust sandwich errors will also be used to estimate intervention effects on secondary outcomes, with link function and error distribution appropriate for each outcome. For secondary outcomes measured at multiple follow-up time points (6-, 10-, and 14-months post baseline), we will fit a single regression model that includes outcome measures from all follow-up time points. Models will include indicator variables for time, intervention group, and interaction terms between these variables, to estimate intervention effects at each time point. Models will adjust for age, sex, risk score, and the baseline measure of the outcome. For hypoglycemia-related healthcare utilization outcomes (urgent care and emergency department visits, hospital admissions), we will follow a similar analytic plan as the primary outcome, including the process to determine whether to model these outcomes as binary or count variables.

Outreach with repeated calls maximizes the response rates of follow-up telephone surveys and minimizes missing outcome data. If attrition is greater than 10%, statistical methods will account for potential missing data bias. We will compare baseline characteristics of responders and nonresponders and consider multiple imputation or inverse probability weighting methods to account for potential bias due to missing data.

Exploratory analyses will assess whether treatment effects vary by patient characteristics including age (under 75 years vs. 75 years and older), sex, race and ethnicity, comorbidities, insulin and/or sulfonylurea use, baseline presence of IAH, and baseline severe hypoglycemia history.

If significant differences (p<0.05) between intervention groups are found for the primary outcome, we will perform mediation analysis to test the hypothesis that the possible causal pathway for reducing the occurrence of severe hypoglycemia is through improvements in hypoglycemia knowledge, hypoglycemia awareness, self-efficacy of hypoglycemia management, blood glucose testing, and hypoglycemic medication adjustment (Figure 2). Potential mediators will be defined as the change in each measure between baseline and the 6-month survey. Mediation analyses will use a regression-based approach within the causal mediation framework,⁵⁷ separating the total intervention effect into a natural indirect effect (effect occurring through the mediator), and a natural direct effect (effect through all other pathways). To estimate these effects, we will use either a log-linear or Poisson model for the outcome model (depending on whether the primary outcome is defined as a binary or count variable), and linear regression for the mediator model. The effect of each mediator will be estimated in separate models.

A mixed methods process evaluation will evaluate the fidelity of implementation of intervention components, document the implementation process, assess patient experience, and clarify the causal pathway if intervention effects are found. We will use the Patient Centered Outcomes Research Institute’s five domains for randomized controlled trials to guide the process evaluation: fidelity, dose, mechanism, reach and context. Qualitative analysis will complement quantitative, including the mediation analysis described above, by providing a diverse set of perspectives and an understanding of the operation of the intervention in the real world setting of the healthcare system. Details are in appendix E.

Statistical Power

We are powered to identify a meaningful 20% reduction in the rate of severe hypoglycemia between proactive care and proactive care plus my hypo compass arms.⁵⁸ Power calculations assume a type 1 error rate of 0.05 and are based on a test of two independent proportions. Assuming 60% of participants in the Proactive Care Management group self-report a severe hypoglycemia event in 12-month look-back, a trial with 204 participants (102 per group) will have 90% power to detect a 26.6% absolute difference between groups (i.e., 37.4% severe hypoglycemia rate in the my hypo compass arm) and 80% power for a 19.7% difference for the primary outcome. To account for 20% loss to follow-up, we are enrolling 256 participants. We are also powered to identify a 20% reduction in the secondary outcome of percent time with biochemical hypoglycemia (<54 mg/dL), with 90% power to detect a 1.6% difference and 80% power to detect a 1.4% difference between groups. Other studies evaluating my hypo compass in individuals with T1D reported similar or larger effects on severe hypoglycemia reductions^24–28 and biochemical hypoglycemia. ^{25, 43}

For subgroups comprising approximately 30% of the study sample (e.g., minority racial and ethnic backgrounds), we have 90% power to detect a 39% difference and 80% power to detect a 35% difference in the rate of severe hypoglycemia between interventions within this subgroup. For subgroups comprising approximately 50% of participants (e.g., female, or age ≥75), we have 90% power to detect a 31% difference and 80% power to detect a 27% difference in the primary outcome. All these effect sizes are smaller than those associated with the my hypo compass intervention among individuals with T1D,⁴² indicating PHT2 should have adequate power to detect both average and subgroup treatment effects.

Discussion

The PHT2 study findings will inform future implementation of evidence-based strategies for preventing severe hypoglycemia among those with T2D most at risk. Currently, no interventions for preventing severe hypoglycemia have been tested in individuals with T2D. Severe hypoglycemia results in significant and possibly preventable negative impacts on quality of life, morbidity, mortality, healthcare utilization and costs. Adoption of an effective intervention would have positive health benefits for individuals with T2D and improved quality of care for healthcare systems.

Structured psycho-education, such as my hypo compass, is recommended for preventing severe hypoglycemia in at-risk individuals with T1D, including those with impaired awareness of hypoglycemia.⁵⁹ All effective programs share educational components, though my hypo compass is briefer and, as implemented in this trial, is all virtual, focusing specifically on successful prevention of severe hypoglycemia.⁴² The protocol used in this trial has only two brief group intervention sessions with telephone follow-up in between and after sessions. Given the relatively modest commitment for both adults with type 2 DM and health professionals, my hypo compass has potential for wider adoption in healthcare settings.

The Proactive Care Management comparison group also fills a care gap for individuals with T2D at risk for severe hypoglycemia. This group receives evidence-based care that healthcare providers usually give after an acute severe hypoglycemia episode. This evidence-based standard follows ADA guidelines and while widely supported, frequently does not happen according to current evidence.^{38, 60} Since most severe hypoglycemia events are unreported to healthcare providers,⁴⁹ most individuals with severe hypoglycemia and associated IAH never receive preventive education and care. To address this gap, we provide recommended care proactively.³⁰ This comparison arm spreads the current standard of care to individuals often unrecognized as having risk for severe hypoglycemia.

The study has several limitations. Our intervention is limited to those with T2D who are willing and able to connect to group training sessions through either video conference or telephone. Generalizability of findings is also limited because we are currently unable to include individuals who do not speak or read English, are planning to become or who are currently pregnant, use a CGM at baseline, or have psychosis or cognitive impairment. Participants are limited to those receiving care within an integrated group practice. Independent practices may have more difficulty implementing the intervention.

PHT2’s novel, evidence-based protocols have potential to significantly reduce severe hypoglycemia-related morbidity and mortality of individuals with T2D. If my hypo compass is more effective than proactive care management alone, it will offer healthcare systems and providers a practical approach to effectively reducing severe hypoglycemia in at-risk individuals with T2D. If proactive care management results in similar severe hypoglycemia prevention as my hypo compass, we will have a tool and evidence-based protocol for preemptively providing this standard of care to individuals with T2D most at risk for severe hypoglycemia.

Supplementary Material

Appendix B

Appendix B: my hypo compass Facilitator Handbook

Appendix C

Appendix C: my hypo compass Participant Handbook

Appendix A: My hypo compass intervention activities

My hypo compass intervention activities are described below. Appendix B has the Facilitator Handbook and Appendix C has the Participant Handbook.

Baseline Outreach Assessment and Engagement in Structured Self-Management Training Sessions:

Clinic nurses specializing in diabetes call individuals and coordinate care for severe hypoglycemia risk factors as in the proactive care group. Nurses elicit and address barriers for attending my hypo compass sessions.^61–63 We send individuals the my hypo compass Glucose/Hypoglycemia diary and Participant Handbook through US mail in preparation for training sessions. The diary supports participant reflection on patterns of blood glucoses and episodes of hypoglycemia. In the diary, participants record scheduled blood glucose checks (4 times a day for 4 weeks; 8 times a day one day a week for 4 weeks), biochemical hypoglycemia (blood glucose < 70 mg/dL) and hypoglycemia episodes with symptoms even if blood glucose was not checked.

Structured Self-Management Training Session 1, Virtual:

Two diabetes nurse educators trained on the my hypo compass protocol lead group sessions independently over Microsoft Teams video conferencing (Figure 3). Participants unable to use video conferencing join by telephone. Sessions are a maximum of 8 individuals for 3 hours, and are based on the Facilitator Guide (Appendix B) and Participant Handbook (Appendix C). The diabetes nurse educators encourage participants to reflect on personal hypoglycemia events using the Glucose/Hypoglycemia diary. The session promotes patient understanding and self-management strategies by group discussion of the my hypo compass points (Figure 5).

Sessions are also attended by the nurse care managers (who are different from the nurse educators) who make the follow-up one-on-one telephone calls. In the training sessions, the nurse care managers document in patients’ charts any clinical concerns raised in the sessions that may be relevant for the patients' overall care and any topics that may need further discussion in the telephone sessions. Diaries are completed over the first 4 weeks of the study. Daily 4-point and weekly 8-point blood glucose profiles determine times of increased hypoglycemia risk including at night, and allow reflection on possible prevention strategies.

Figure 5: The Hypo Compass.

Telephone Call #1, Review:

Nurse care managers trained on the my hypo compass protocol call participants for a 10- to 15-minute review. Nurses discuss progress and review the Glucose/Hypoglycemia diary. They encourage reflection on any hypoglycemia events and self-management strategies. The call occurs 14 days after training sessions. This is a change from 7 days in the original my hypo compass for individuals with T1D, based on patient and caregiver suggestions that 2 weeks is better to see meaningful, actionable patterns in glucose levels and hypoglycemia symptoms.

Structured Self-Management Training Session 2, Virtual:

In a 60-minute follow-up session with participants in their original groups, the same diabetes nurse educator who led the first session facilitates discussion on progress, referring to the 4 points of my hypo compass. A nurse care manager attends this group for clinical documentation. Participants are encouraged to reflect on hypoglycemia symptoms experienced during each event including subtle symptoms with review and discussion of the strategies implemented.

Telephone Call #2, Booster:

Nurse care managers call participants for a 10- to 15-minute review and booster call 3 months after intervention start or 2 months after the last training session. Nurses and participants reflect on any hypoglycemia events and review self-management strategies. This booster, performed by the same nurse as the first telephone review, was not in the original my hypo compass protocol but has been added because focus group participants thought a follow-up call in a “couple of months” was important to sustain the effect of the core intervention.

Training and Monitoring Intervention Fidelity

We support generalizability and relevance to healthcare providers and healthcare systems by applying the Treatment Fidelity Workgroup of the framework endorsed by the National Institutes of Health Behavioral Change Consortium to ensure interventions are delivered as intended.⁶⁴

Training

• Proactive Care Management: All intervention nurses use standard training, tools and workflows that are already in use at KPWA to provide the proactive care management part of both interventions. All KPWA nurses are trained in these protocols by nurses who are certified diabetes care and education specialists. The study provides continuing nurse education training to refresh nurses in these standard protocols for all nurses before enrolling the first study participant.

• My hypo compass Protocol: Study intervention leads (who are PhD health psychologists) train the diabetes nurse educators over 5 sessions in delivery of the my hypo compass protocol with input and assistance from the program originators. The diabetes nurse educators conduct all online virtual training for group sessions. The study intervention leads and the diabetes nurse educators train the my hypo compass intervention nurses for telephone review and booster follow-up calls. Training includes workflow and documentation tools developed for the intervention in the Epic EHR. The my hypo compass facilitator guide and patient handbook are in Appendices B and C, respectively.

Fidelity: We maintain and assess fidelity through the following proactive strategies:

• Use of standardized written nurse and participant materials. In the Facilitator Guide, all concepts and skills that must be discussed are highlighted in bold to ensure ease in protocol delivery and adherence. Study intervention leads and diabetes nurse educators and population care nurses meet regularly to ensure protocol clarity and treatment integrity and fidelity.

• Monitoring for potential drift through audio recording sessions and observation. After 2 months of observation, a Fidelity Monitoring Research Associate randomly reviews 10% of sessions, coding for intervention adherence using a standard protocol and returning results to intervention leads if there is significant drift. If nurses “drift” from protocols, we will add training and supervision and monitor until fidelity is established.

Appendix D: Data Collection and Management

Data collection methods

Study data are collected using self-report surveys at baseline, 6, 10 and 14 months; biochemical measures of blood glucose levels using masked CGM at baseline and 14 months; and EHR data for HbA1c and hypoglycemia-related healthcare utilization at baseline and 14 months.

Self-Report Measures:

Survey questions are administered by telephone at baseline and all follow-up time points (6, 10 and 14 months) by trained Survey Research Program (SRP) staff at Kaiser Permanente Washington Health Research Institute. Telephone data collection is performed using Sawtooth Ci3 and WinCATI computer-assisted telephone interviewing (CATI) software. SRP interviewers are trained in standardized survey interviewing techniques, and use of the CATI system, and are trained in project-specific procedures including item-by-item specifications for each questionnaire. Interviewers' telephone performance, productivity, and response rates are routinely monitored; corrective or remedial training is provided if necessary. Interview quality is continuously assessed using silent monitors installed on all telephones; completed interviews are edited and coded within 48 hours, and interviewer feedback occurs on a regular basis. SRP Interviewers are trained in National Institutes of Health Human Subjects Protections.

CGM Data:

CGM data are collected at baseline and 14 months using the Dexcom G6 Pro CGM integrated sensors and transmitters in participant-masked mode. Dexcom sensors were chosen because of validity and reliability; ability to record blood glucose measurements that are not revealed to the participant; and ability to store up to 10 days of data. At both timepoints we collect 10 days of CGM readings. To allow adequate self-calibration of the Dexcom G6 CGM,⁶⁵ we will use the last 9 of 10 days of the CGM collection window to calculate outcomes and only calculate outcomes if 70% of the CGM measures are present in that 9-day window. CGM outcomes will be shared with participants at the end of the study. Baseline CGM data are collected after collection of baseline self-report data but before beginning any intervention activities.

CGM data at 14 months will be collected at the same time as outreach for the 14-month follow-up self-report. In both cases the process is the same: Study staff telephone the participant and schedule a CGM virtual visit. Study staff mail the sensor-transmitter unit (programmed so blood glucose readings are not visible to the participant) with instructions for the participant, with postage-paid packaging to return at the end of the 10-day recording period. At the virtual visit, the study nurse walks the participant through placing the device, answers questions and provides a number to call for questions or issues that may arise. The participant wears the device for 10 days with readings stored on the unit. The study nurse makes one follow-up call near the end of the 10 days to check in, answer questions and confirm that the participant will remove and mail back the single-use sensor-transmitter on the planned date. At the end of the recording period, the participant removes and returns the device using the provided mailing materials. The data are read or downloaded from the sensor-transmitter using the DexCom 6G Pro reader/receiver.

EHR Data:

The study programmer extracts EHR data to assess study eligibility for recruitment and to collect outcome measures including HbA1c and hypoglycemia-related healthcare utilization, including emergency department, urgent care visits and hospitalizations. HbA1c results collected as part of the study are documented in the EHR, so results for these outcomes can be collected from the medical record in the same way as other HbA1c results. The most recent HbA1c measures in the 12 months before baseline will be used for the baseline measure. Most recent measures within the past 3 months will be collected at 14 months. If the participant has not had an HbA1c within the past 3 months at 14 months, they will be asked to have one for the study.

Procedures for Participants Not Completing Study Activities, or Withdrawing Consent:

Participation in this study is voluntary. Even after consent, participants may choose to not do specific study components such as CGM data collection. They may decline survey questions or choose not to participate in intervention activities. Participants are not considered withdrawn from the study unless they state clearly that they wish to withdraw. For participants who do not complete specific activities but do not withdraw consent, we continue to use information they have provided and will extract and use EHR data as described in the consent process. If participants withdraw their consent, we may still use information collected before withdrawal, but not collect any additional information and will destroy any record of their name or other information that may identify them. This is clearly stated in the consent materials and reviewed as part of the informed consent process.

Appendix E: Process Evaluation

A mixed methods process evaluation will evaluate the implementation of intervention components and clarify the causal pathway if intervention effects are found. Our process evaluation will assess whether the intervention was implemented as planned and test the hypothesized causal pathways as shown in the study model (Figure 2 in the main text). These results will be essential to ensure future implementations of severe hypoglycemia prevention are centered on patient needs and are feasible and acceptable for healthcare professionals and systems.

We will assess the implementation process according to the five PCORI domains for RCTs (Table 2): fidelity, dose, mechanism, reach, and context.^66–68 We will use qualitative and quantitative methods for assessing the fidelity, dose and mechanism of the intervention. Qualitative analysis will complement quantitative by providing a diverse set of perspectives and an understanding of the operation of the intervention in the real world setting of the healthcare system. We will concurrently and independently pursue the quantitative and qualitative evaluations in these domains.⁶⁹ We will also use qualitative methods to understand contextual factors influencing the intervention implementation and quantitative methods to assess intervention reach.

Table 2:

Process Assessment Domains, Question, Methods⁶⁶

Evaluation Domain	Research Questions	Research Methods
Fidelity	Are the interventions delivered as intended?	Qualitative analysis of nurse and patient interviews, nurse observations -Quantitative analysis of fidelity checklists and participant contacts
Dose	Is the intended amount of the intervention delivered?	-Qualitative analysis of nurse and patient interviews, nurse observations -Quantitative analysis of fidelity-checklists and participant contacts
Mechanism	Why and how did the intervention work?	-Quantitative mediation analysis of causal pathway -Qualitative analysis of patient and nurse interviews
Reach	Did the intended populations receive the intervention?	Quantitative descriptive analysis of clusters and individuals using administrative and EHR data
Context	What clinic, institutional, state and national factors may moderate intervention?	-Qualitative analysis of interviews with stakeholders

To assess fidelity and dose to individuals, we will observe nurse care managers and diabetes nurse educators working with participants and interview them after interventions.⁷⁰ We will conduct observations of nurses on the two educational calls in the my hypo compass group. We will observe 3 sets of my hypo compass group training sessions facilitated by the two diabetes nurse educators using video recordings of these sessions. After the intervention and collection of outcomes, we will conduct telephone interviews with 8 to 12 participants in each comparator group, the 2 diabetes nurse educators delivering the in-person education sessions and 3–6 nurse care managers in each comparator group. We will ask participants to reflect on their experiences with the intervention and the impact of the intervention on care including unintended consequences. To ensure representativeness in interviews, we will purposively over sample by minority race and ethnicity, gender, and educational level. Guided by the causal model (Figure 2), we will also ask about process evaluation domains (Table 1). For evaluating intervention context, we will interview leaders of the healthcare system (e.g. directors of nursing, endocrinology). We will use Atlas.ti to manage our qualitative data. Our analysis will use a template approach.⁷¹ The code book will focus on our conceptual model, the intervention components and emergent or patient-centered themes that help understand patient experiences and impact.^{72 73}