"Low-value" glycemic outcomes among older adults with diabetes cared for by primary care nurse practitioners or physicians: A retrospective cohort study.

The study was observational, not randomised. Patients were reassigned after a provider left, and the instrumental variable method relies on assumptions about facility staffing. The adjusted logistic model showed no difference, while the instrumental variable model showed a difference. This means the result depends heavily on the analysis method. The sensitivity analysis included only 1,025 patients prescribed Veterans Health Administration diabetes medications, so it was too small to confirm that prescribing explains the result. Data are from Veterans Health Administration patients, mostly male veterans, and predate later changes that increased outside care, so they may not apply to other populations or current practice. Hypoglycemia risk was identified using administrative diagnosis codes, which may have been too conservative and did not fully capture kidney disease stage or other clinical details. The study did not adjust for provider panel size or prescribing practices, which the authors suggest could influence glycemic targets.

The authors declared no conflicts of interest.

Do not read the 20.38 percentage-point instrumental variable finding as proof that nurse practitioners always provide better diabetes care. The adjusted logistic model showed no difference, the medication sensitivity analysis was not significant, and the study is observational.