Pairing Physician Education with Patient Previsit Handouts Doubles Diabetes Deprescribing Rates in Adults Over 75

There is a well-established problem in how we treat type 2 diabetes in older adults: we frequently keep doing things that no longer make sense. An 82-year-old with an A1c of 7.3% on insulin and a sulfonylurea is at meaningful risk of hypoglycemia — the thing that causes falls, hospitalizations, and death — while gaining almost nothing from tight glycemic control at that stage of life. The evidence for aggressive glucose lowering in patients over 75 with limited life expectancy is thin. The harms, particularly hypoglycemia and the cognitive impairment it triggers, are real. Yet deprescribing — reducing or stopping these medications — is remarkably rare in clinical practice.[1]

The JAMA Internal Medicine randomized clinical trial by Grant RW and colleagues set out to test a scalable deprescribing intervention in a large integrated health system. Conducted at Kaiser Permanente Northern California from September 2020 through March 2024, the trial enrolled 450 patients (mean age 79.9 years, 49.6% female) treated by 211 primary care physicians. All patients were 75 or older, had an A1c of 8.0% or lower — meaning their diabetes was already reasonably controlled — and were being treated with insulin, a sulfonylurea, or both. These are the patients most vulnerable to hypoglycemia and least likely to derive benefit from the current regimen intensity.[1]

All participating primary care physicians received academic detailing (AD) — an evidence-based educational outreach method where a trained specialist meets individually with physicians to review guidelines and discuss specific patient cases. Patients were then randomly assigned to one of two arms: those in the intervention arm received a previsit activation handout before their PCP appointment that explained why deprescribing diabetes medications is sometimes appropriate and invited them to discuss it with their doctor. Those in the control arm received a healthy lifestyle handout of the same length — an attention-matched control designed to ensure any difference came from the deprescribing content specifically, not simply from receiving any handout at all.[1]

At 6 months, patients in the combined AD plus previsit activation arm had significantly higher rates of diabetes medication deprescribing: 15.8% versus 9.0% in the AD-only arm (adjusted risk difference 7.5%; 95% CI 1.5% to 13.6%; P=0.01). The effect was sustained at 12 months: 22.8% versus 16.3% (adjusted RD 7.9%; 95% CI 0.4% to 15.5%; P=0.04). Critically, there was no significant increase in patient-reported severe hypoglycemia between groups at 6 months (4.7% versus 6.5%; P=0.34) — a reassurance that reducing medication intensity in this population does not appear to increase the most dangerous short-term complication.[1]

Clinical Context

The case for deprescribing in older adults with type 2 diabetes rests on a straightforward asymmetry: the benefits of tight glycemic control accumulate over decades, while harms from hypoglycemia occur immediately. A 79-year-old with multiple comorbidities may not live long enough to see the microvascular benefits of an A1c of 6.5% versus 8.0%, but they can absolutely fall, fracture a hip, and die from a sulfonylurea-induced hypoglycemic episode this week. Sulfonylureas stimulate insulin secretion regardless of blood glucose level, making hypoglycemia a persistent pharmacologic risk. Insulin carries similar hazards. In patients whose A1c is already at or below 8% — meaning medication intensity is working — reducing that intensity does not suddenly mean abandoning good care. It means recalibrating the risk-benefit ratio for a different patient than who was enrolled in the landmark glucose-lowering trials.[2]

Guidelines from the American Geriatrics Society and the American Diabetes Association now explicitly recommend less stringent A1c targets for older frail adults and for those with limited life expectancy. Despite these recommendations, real-world rates of deprescribing remain persistently low. Part of the problem is structural: the system is designed to prescribe and renew, not to proactively reconsider. Part of it is conversational: the deprescribing discussion is time-consuming, requires gauging patient values, and can feel to a busy PCP like opening a can of worms in a 20-minute visit. This trial directly addresses the structural and conversational barriers with two scalable tools.[3]

Why It Matters Clinically

The magnitude of effect here — roughly 7 percentage points more deprescribing — may sound modest, but it's the kind of difference that is detectable, reproducible, and achievable without any new technology, extra staff, or system redesign. It requires a one-page handout and the physician outreach session the health system already delivers. Scaled to a large integrated system's population of older adults on sulfonylureas and insulin, a 7-point shift in deprescribing rates translates to thousands of patients removed from unnecessary hypoglycemia risk annually.[1]

The trial was conducted entirely virtually, including the academic detailing sessions, due to the COVID-19 pandemic — meaning these results likely understate what is achievable with in-person physician engagement. The researchers also note that both arms received academic detailing, making the true comparator not "usual care" but rather "usual care plus physician education." The absolute deprescribing rate in the control arm (9.0%) likely exceeds what happens in real-world practices without any intervention. The incremental benefit of the patient handout is being measured against a somewhat-activated control, which means in an average practice without existing deprescribing culture, the combined intervention might produce larger absolute gains.[1]

Limitations

Both trial arms received physician academic detailing, so there is no true usual-care control. This limits conclusions about the absolute effect of AD alone versus nothing. The study was conducted in a single integrated health system (Kaiser Permanente Northern California), which has unusually strong infrastructure for care coordination and may not generalize to fragmented primary care settings. The trial was conducted virtually during the pandemic, limiting in-person engagement. Sample size was modest at 450 patients, and the trial was not powered to detect differences in hard clinical outcomes such as hypoglycemia hospitalizations, falls, or mortality — only in deprescribing rates and self-reported severe hypoglycemia.