Understanding the importance of algorithm choices in automated insulin delivery

Clinicians emphasise that knowledge of pump algorithms is crucial for effective diabetes management with automated insulin delivery systems, as new device options and algorithm types offer personalised care but require understanding of their unique features.

At a recent Association of Diabetes Care & Education Specialists meeting, Erin Henry, a nurse practitioner at North Atlanta Endocrinology and Diabetes in Georgia, argued that the growth of automated insulin delivery has made pump algorithms a practical issue for everyday care, not just a technical one. Speaking to Healio, Henry said clinicians who do not understand how a system makes dosing decisions will struggle to advise patients on everything from pre-bolusing to adjusting carbohydrate-to-insulin ratios. Healio reported that her message was straightforward: the algorithm matters as much as the device itself.

Henry drew a clear line between systems that build basal insulin from total daily dose and those that adapt around a set basal profile. In total daily dose systems, the pump generates the basal rate and users cannot simply edit it; in adaptive systems, changing the programmed basal schedule can still influence how the algorithm behaves. She used Omnipod 5 as an example of a system where altering basal settings would not have the intended effect, while pointing to devices such as twiist and Tandem systems as examples where background settings can still matter. Omnipod’s own clinical materials say its algorithm recalculates insulin delivery every five minutes and updates an adaptive rate based on changing total daily insulin needs.

The variety of systems now available in the United States has also widened the choices for patients. A 2025 comparison from Taking Control Of Your Diabetes listed five commercial automated insulin delivery platforms in the U.S., including Tandem Control-IQ+, Medtronic MiniMed 780G, Insulet Omnipod 5, Beta Bionics iLet and Sequel twiist, each with different features and levels of user interaction. Henry told Healio that no single algorithm is best for everyone, because some people want more control while others prefer a more hands-off approach. She said clinicians should also understand extras such as exercise modes, temporary basal rates, extended boluses and preset meal options, because those functions can be decisive in real-life use.

Her advice differs somewhat by diabetes type. For people with type 2 diabetes who already use multiple daily injections, Henry said she would consider pump therapy regardless of diagnosis and often looks to larger reservoirs that hold 300 units or more. For type 1 diabetes, she said smaller dose adjustments are usually safer because of greater insulin sensitivity and a higher risk of rapid diabetic ketoacidosis if a site fails. Practical training, she added, should include teaching all patients to recognise delivery failure, with particular emphasis in type 1 diabetes because they have little or no insulin reserve.

Henry also urged caution when patients start glucagon-like peptide-1 drugs alongside automated insulin delivery. She told Healio that insulin needs can fall sharply when these medicines reduce appetite and food intake, and that she typically lowers settings quickly to avoid hypoglycaemia. In her practice, she said, people with type 1 diabetes may need roughly a 30 per cent reduction in insulin requirements, compared with about 20 per cent for type 2 diabetes. She added that people using total daily dose systems have fewer adjustable levers, so close monitoring is essential. When reviewing data, Henry said the most useful indicators are time spent in automation, bolus behaviour, total daily dose and the balance between basal and bolus insulin. A consensus review in Diabetes Spectrum and guidance in the open-access literature both support that emphasis on reviewing algorithm behaviour, insulin distribution and day-by-day glucose patterns rather than relying on a single summary figure.