Digital algorithm-guided insulin therapy in home healthcare for elderly persons with type 2 diabetes: A proof-of-concept study

**Background:** Type 2 diabetes in elderly persons is associated with higher mortality, institutionalization risk, and comorbidities including heart failure, coronary artery disease, renal and liver disease, and dementia. Hypoglycemia risk is elevated in this population. Basal insulin is recommended to simplify management, but home healthcare settings often lack diabetes specialist support. Clinical decision support systems and evidence-based algorithms have been developed to facilitate optimized glycemic control. This study investigated user acceptance, safety, and efficacy of a digital algorithm-guided system (GlucoTab@MobileCare) for basal and basal-plus insulin therapy in home healthcare.

**Methods:** This open-label, single-centre, non-controlled proof-of-concept study was conducted at two Austrian Red Cross home healthcare sites in Graz. Nine participants (5 female, age 77±10 years, BMI 28±5 kg/m²) with type 2 diabetes on insulin therapy receiving home healthcare were enrolled. Care dependency levels were: 22% level 1 (>65 hours/month), 22% level 2 (>95 hours/month), 33% level 3 (>120 hours/month), 22% level 4 (>160–180 hours/month). Glycemic control targets were: 11% good (FBG 90–130 mg/dL), 67% moderate (FBG 90–150 mg/dL), 22% frail (FBG 100–180 mg/dL). Participants received basal or basal-plus insulin therapy (insulin glargine U300, Toujeo® Solostar®; insulin glulisine Apidra® Solostar®) once daily as suggested by the algorithm-guided digital system over three months. The basal insulin dose was reduced by 2 units if any BG was <20 mg/dL below target and by 4 units if >20 mg/dL below target. The bolus component was initially estimated at 20% of the basal dose. BG corrections used short-acting insulin for very high values (≥300 mg/dL for tight/moderate control; ≥350 mg/dL for frail control). Eight participants completed the study; one withdrew due to hospitalization (unrelated). Nurses completed user acceptance questionnaires on a six-point Likert scale pre- and post-study; participants/relatives completed satisfaction questionnaires at study-end.

**Key Results:** User acceptance was high: 95% of all suggested tasks were performed according to the digital system. The suggested initial basal insulin dose was accepted for one participant (11.1%) and changed for eight (88.9%). Adherence to specific suggestions: 94.9% for BG measurement frequencies, 95.9% for basal insulin titration doses, 99.7% for basal insulin injection doses, 97.9% for bolus insulin injection doses, and 100% for titration time points. All nurses (n=9) and all participants/relatives (n=9) reported positive satisfaction. Mean morning BG across all treatment days was 155±53 mg/dL (n=761); mean total BG was 167±62 mg/dL (n=994). Mean morning BG decreased from 171±68 mg/dL (month 1) to 150±46 mg/dL (month 2) to 145±35 mg/dL (month 3). Glycemic variability (SD) decreased by 33 mg/dL from month 1 to month 3. HbA1c decreased from 60±13 mmol/mol (baseline) to 57±12 mmol/mol (study end). No hypoglycemic episode <54 mg/dL occurred. Morning BG distributions: <70 mg/dL (0.3%), 70–180 mg/dL (77.0%), >180 mg/dL (22.7%), >300 mg/dL (2.6%). Daily mean initial insulin dose was 24±13 IU; daily mean dose on last treatment day was 38±31 IU (range 2–107 IU). Insulin was adjusted on average nine times per participant (44 adjustments in first 4 weeks, 41 in weeks 5–12). Seven adverse events occurred; one serious adverse event (hospitalization due to dyspnea, cough, chest pressure; death) was not study drug- or device-related. One AE (hypoglycemia 58 mg/dL) had probable causality with study drug.

**Clinical Implications:** This proof-of-concept study demonstrates that a digital algorithm-guided insulin therapy system can be safely and effectively used by non-specialist nurses in home healthcare for elderly patients with type 2 diabetes. The high adherence rates (94.9–100%) and positive user satisfaction suggest the system is feasible for real-world implementation. The absence of severe hypoglycemia (<54 mg/dL) and the reduction in mean morning BG and glycemic variability indicate improved glycemic control. The study is limited by its small sample size (n=9) and lack of control group. Larger-scale studies are needed to confirm these findings in more heterogeneous populations and routine use by general practitioners. The digital system could potentially improve quality and safety of basal insulin therapy in home care settings, reduce hospitalizations and emergency visits, and help maintain functional status in elderly patients.