Israeli AI Tool Helps Type 1 Diabetics Dose Insulin via Smartphone

A study from Israel has tested a promising artificial intelligence decision-support system (AI-DSS) that delivers personalized insulin dosing recommendations directly to patients with Type 1 diabetes — right on their smartphones. Conducted at the Schneider Children's Medical Center of Israel in Petah Tikva, this proof-of-concept research could point toward a new era of smarter self-management for the millions of Americans living with Type 1 diabetes who rely on multiple daily injections (MDI).

What the Israeli Researchers Did

The study enrolled 16 young adults with Type 1 diabetes (average age 25 years; 56% female) who were already using continuous glucose monitors (CGM) and managing their insulin through multiple daily injections rather than a pump. Participants first completed a 4-week baseline period, then entered a 12-week phase during which an algorithm — delivered via a mobile app — provided updated insulin dosing recommendations every two weeks.

Researchers then compared CGM data from the final two weeks of the baseline period against the final two weeks of the active intervention.

Key Findings

The results were encouraging across the board:

• HbA1c dropped significantly — from a median of 7.5% down to 7.1% (p = 0.013). For context, the American Diabetes Association (ADA) recommends an HbA1c below 7.0% for most non-pregnant adults with diabetes.
• Time in Range (TIR) improved by 3.5% — meaning patients spent meaningfully more time with blood glucose between 70–180 mg/dL, the standard target range endorsed by both US and international guidelines.
• Dangerous lows stayed flat — time spent below 54 mg/dL (a critical hypoglycemia threshold) did not increase, a key safety win. It remained under 1.1% throughout, well within the internationally recommended limit of less than 1%.
• High blood sugar trended down — time above 250 mg/dL showed a downward trend (14.3% vs. 12.3%), though this did not reach full statistical significance.
• No severe adverse events were reported during the entire study period.

Why This Matters for US Patients

In the United States, approximately 1.6 million people live with Type 1 diabetes. While insulin pumps with automated delivery systems exist, they are expensive and not accessible to everyone. The majority of Type 1 patients in the US — and globally — still rely on multiple daily injections. An AI-powered app that helps fine-tune insulin dosing without requiring a pump could be a game-changer for this underserved population.

Israel has long been a leader in digital health innovation, and this research reflects a growing international consensus: smart technology can safely extend the reach of endocrinology care beyond the clinic. The ADA's current Standards of Care already endorse CGM use for MDI patients, but personalized algorithmic dosing support remains largely unavailable outside of clinical trials. This Israeli study provides early evidence that such tools are both safe and effective.

It's also worth noting that the participants started with an average HbA1c of 7.6% — a number many US patients would recognize as their own reality — making these findings directly relatable.

Important Caveats

This was a small proof-of-concept study with just 16 participants and no control group. The researchers themselves call for larger, randomized controlled trials before this technology could be considered ready for widespread clinical use. US patients should not attempt to replicate algorithmic insulin dosing without physician supervision.

The Bottom Line

This Israeli research adds important international evidence that AI-assisted insulin dosing via smartphone is feasible, safe, and capable of meaningfully improving blood sugar control for Type 1 diabetes patients on injections. Watch this space — larger trials are the logical next step.

Source: Nimri R, Weizman Demri, Fisch-Shvalb N. "Decision Support Tool for Self-Management of Insulin Dosing in Type 1 Diabetes with Multiple Daily Injections: A Proof-of-Concept Study." Diabetes Technology & Therapeutics, 2026. DOI: 10.1177/15209156251380860. PMID: 40986451.