Development and Preclinical Validation of DiaBuddy, a Point-of-Care Decision Support Tool for Children with Type 1 Diabetes: A Pilot Clinical Study

This pilot study demonstrates that DiaBuddy, a guideline-aligned mobile decision support tool, significantly outperforms family recommendations in preclinical accuracy for insulin dosing and sick-day management while showing preliminary improvements in glycemic control and quality of life for children with Type 1 diabetes, supporting the need for larger randomized controlled trials.

The Gist

Imagine managing Type 1 Diabetes for a child is like trying to drive a high-performance race car through a storm. You have to constantly adjust the speed (insulin) based on the road conditions (food), the weather (illness), and the car's current speed (blood sugar). If you guess wrong, the car could stall (low blood sugar) or crash (diabetic ketoacidosis, a life-threatening condition).

For years, parents have been the drivers, trying to make these split-second calculations in their heads, often without a map or a co-pilot. This study introduces DiaBuddy™, a new "smart co-pilot" app designed to help these parents navigate the storm safely.

Here is a breakdown of how the study worked and what it found, using simple analogies:

1. The Problem: Guessing in the Dark

Managing diabetes requires precise math. Parents have to figure out:

• How much insulin to eat? (Based on carbs in food).
• How much insulin to fix a high? (Based on current blood sugar).
• What to do when the child is sick? (This is the trickiest part; getting it wrong is the #1 cause of hospital visits).

Without help, parents often make mistakes, just like a driver guessing the speed limit in a foggy storm.

2. The Solution: The "Smart Co-Pilot" (DiaBuddy™)

The researchers built an app called DiaBuddy. Think of it as a GPS for diabetes that follows the strictest traffic rules (medical guidelines). It has two main features tested in this study:

• The "Insulin Wizard": It calculates exactly how much insulin to give for meals and corrections.
• The "Sick-Day Guide": It acts like a triage nurse, telling parents exactly when to give extra insulin, what fluids to drink, and crucially, when to rush to the hospital.

3. Phase 1: The "Mock Exam" (Preclinical Validation)

Before letting the app help real kids, the researchers gave it a test.

• The Setup: They created 40 difficult "what-if" scenarios (like "The child ate pizza, has a fever, and their blood sugar is high").
• The Contest: They asked 37 families to solve these scenarios. Then, they asked the app to solve them. Finally, a top-tier expert doctor (the "Gold Standard") solved them to see who was right.
• The Result:
  • The Families: Like students who forgot their math formulas, they were often way off. Their insulin guesses were frequently too high or too low (errors of up to 45%).
  • The App: It was like a calculator that never makes a mistake. It matched the expert doctor's answers almost perfectly (over 90% accuracy).
  • The Big Win: In the "Sick-Day" scenarios, the app would have prevented 94.5% of the mistakes families made. Most importantly, it would have caught every single case where a family failed to send a child to the hospital when they needed to.

4. Phase 2: The "Real Road Trip" (Pilot Clinical Study)

Next, they let 20 families use the app in real life for three months.

• The Journey: Families used the app to manage their child's diabetes daily.
• The Destination (Results):
  • Blood Sugar Control: The average blood sugar level (HbA1c) dropped significantly. It's like the car finally found a smooth lane and started cruising at the right speed.
  • Quality of Life: This was the biggest surprise. The families reported feeling much less stressed and more confident. Imagine the relief of a driver who finally has a GPS that speaks up before you take a wrong turn. Their "happiness score" jumped by a huge margin.
  • Satisfaction: The families loved it. 96% said, "We want to keep using this!"

5. The Catch (Limitations)

The researchers are honest about the study's limits:

• Small Group: It was a small "test drive" with only 20 families. We need a bigger test with more drivers to be sure.
• No Control Group: They didn't have a group of families not using the app to compare against. It's possible the families improved just because they were paying more attention to the study (the "Hawthorne effect"), not just because of the app.
• Short Time: Three months is a short trip. We need to see if the app keeps working well over years.

The Bottom Line

DiaBuddy™ is like a safety net and a navigation system rolled into one.

In the test phase, it proved to be far more accurate than parents guessing on their own, especially in dangerous sick-day situations. In the real-world test, it helped lower blood sugar levels and, perhaps even more importantly, took a huge weight off the shoulders of parents, making them feel more confident and less anxious.

While more research is needed to confirm it works for everyone, this study suggests that giving parents a "smart co-pilot" could be a game-changer for keeping children with Type 1 Diabetes safe and happy.

Technical Summary

1. Problem Statement

Type 1 Diabetes (T1D) management in children requires precise, real-time decision-making regarding insulin dosing and sick-day protocols. Current challenges include:

• Suboptimal Glycemic Control: Families often struggle with complex calculations for insulin-to-carbohydrate ratios (ICR) and insulin sensitivity factors (ISF), leading to hyperglycemia or hypoglycemia.
• Sick-Day Management Errors: Inappropriate management during intercurrent illness is the leading cause of Diabetic Ketoacidosis (DKA), a life-threatening complication.
• Lack of Integrated Tools: Existing bolus calculators often lack integrated sick-day logic, and manual carbohydrate counting is prone to error and attrition over time.
• Resource Limitations: There is a critical need for accessible, guideline-aligned decision support in settings where specialist access and structured education are limited.

2. Methodology

The study employed a two-phase design: a preclinical vignette-based validation followed by a prospective single-arm pilot clinical study.

The Intervention: DiaBuddy™

DiaBuddy is a mobile application featuring seven modules, with this study focusing on two core components:

• Insulin Wizard: Calculates basal titration and bolus doses based on ISPAD 2022 guidelines. It uses proprietary algorithms to compute ICR ($500/\text{TDD}$) and ISF ($1,800/\text{TDD}$ for rapid-acting insulin). It incorporates a "CarbCal" database (10,245 foods) for carbohydrate quantification and accounts for Insulin-on-Board (IOB) using a linear decay model.
• Sick-Day Guide: A stepwise algorithm for intercurrent illness management. It inputs symptoms (vomiting, breathing), glucose, and ketone levels to generate recommendations on hospital admission, total daily dose (TDD) modification, supplemental insulin, and fluid intake.

Phase 1: Preclinical Validation

• Design: Comparison of DiaBuddy outputs and family recommendations against an independent "Gold Standard" (a pediatric endocrinologist from a different institution, blinded to the study).
• Participants: 37 families of children with T1D.
• Task: Families and the app responded to 40 standardized vignettes (20 insulin-dosing, 20 sick-day scenarios).
• Metrics: Absolute Relative Deviation (ARD) for dosing accuracy; domain-specific accuracy for sick-day decisions (hospitalization, TDD modification, fluid type, supplemental insulin).

Phase 2: Pilot Clinical Study

• Design: Prospective, single-arm, uncontrolled pilot study.
• Participants: 20 children/adolescents (ages 5–18) with T1D duration >1 year. (Originally 25 enrolled; 5 excluded due to withdrawal or incomplete data).
• Protocol: Structured training on DiaBuddy followed by 3 months of unrestricted home use.
• Outcomes Measured:
  • Primary: HbA1c change.
  • Secondary: Continuous Glucose Monitoring (CGM) metrics (TIR, TAR, TBR, GMI), Quality of Life (PedsQL), and application satisfaction.
• Statistical Analysis: Paired t-tests/Wilcoxon signed-rank tests for pre/post comparisons; Cohen's $d_z$ for effect sizes.

3. Key Contributions

• Integrated Algorithmic Approach: DiaBuddy is one of the few tools that seamlessly integrates basal/bolus dosing with sick-day management logic based on ISPAD 2022 guidelines.
• Independent Gold Standard Validation: The study utilized a blinded, independent expert for validation, mitigating bias from the developers.
• Preclinical vs. Clinical Bridge: The study uniquely connects high-accuracy vignette testing with real-world clinical outcomes in a pediatric population.
• Resource-Limited Context: The tool is designed for scalability in low-resource settings, offering a low-cost digital solution for guideline adherence.

4. Results

Phase 1: Preclinical Validation

• Insulin Dosing Accuracy: DiaBuddy significantly outperformed families.
  • Basal Dose ARD: 5.0% (DiaBuddy) vs. 24.2% (Families).
  • Bolus Dose ARD: 6.9% (DiaBuddy) vs. 45.3% (Families).
  • Combined Dose ARD: 6.3% (DiaBuddy) vs. 39.0% (Families).
  • All comparisons had $P < 0.001$.
  • 95% of DiaBuddy recommendations fell within the clinically acceptable $\pm20\%$ error margin, compared to <50% for families.
• Sick-Day Management:
  • DiaBuddy achieved $\ge90\%$ accuracy across all domains.
  • Family accuracy ranged from 27% to 70%.
  • Family errors totaled 51.8% of all decisions. Adherence to DiaBuddy would have prevented 94.5% of these errors, including 100% of errors regarding hospitalization decisions.

Phase 2: Pilot Clinical Study

• Glycemic Control:
  • Mean HbA1c decreased from 9.18% to 8.48% ($P = 0.049$).
  • Mean reduction: $-0.70\%$ (95% CI: $-1.40$ to $-0.003$).
  • 60% of participants improved; 55% achieved a clinically meaningful reduction ($\ge0.5\%$).
  • Note: CGM metrics (TIR, TAR, TBR) did not show significant changes, likely due to the short 3-month duration and sample size.
• Quality of Life (QoL):
  • PedsQL scores increased significantly from 76.5 to 89.1 ($P < 0.001$).
  • Mean improvement: +12.6 points (Large effect size: Cohen's $d_z = 1.94$).
  • 95% of participants showed QoL improvement.
• User Satisfaction:
  • Mean satisfaction score: 44.1/50.
  • 96% of families wished to continue using the app.
  • Average usage frequency: 66.4% of days.

5. Significance and Conclusion

• Clinical Impact: The study demonstrates that DiaBuddy can significantly reduce insulin dosing errors and improve sick-day decision-making, potentially reducing the risk of DKA and severe hypoglycemia.
• Patient-Centered Outcomes: The substantial improvement in Quality of Life suggests the tool effectively reduces the psychological burden of diabetes management for families.
• Scalability: As a low-cost, mobile-based solution, DiaBuddy offers a viable strategy to bridge the gap in specialist care access, particularly in low-and-middle-income countries.
• Limitations & Future Work: The study is limited by its single-arm design (risk of regression to the mean/Hawthorne effect) and short duration. The authors conclude that larger, multicenter randomized controlled trials (RCTs) with longer follow-up are necessary to confirm causal efficacy and evaluate long-term CGM metric improvements.

Conclusion: DiaBuddy™ represents a validated, guideline-aligned decision support tool that substantially outperforms family-derived decisions in preclinical testing and shows promising preliminary signals for improving glycemic control and quality of life in children with T1D.