Inpatient insulin starter

Start or intensify insulin for persistent glucose ≥180 mg/dL (10.0 mmol/L) confirmed twice within 24 h. Aim for 100–180 mg/dL (5.6–10.0) in non-critically ill patients if achievable without significant hypoglycemia, and 140–180 mg/dL (7.8–10.0) in most critically ill patients, where continuous IV insulin is recommended. Up to 250 mg/dL may be acceptable for terminal illness, advanced kidney failure or dialysis, high hypoglycemia risk or labile glucose. A fasting glucose <100 mg/dL predicts hypoglycemia in the next 24 h.

Poor or no oral intake: basal or basal + correction. Adequate intake: basal + prandial + correction. Correction insulin alone ("sliding scale") is discouraged for most patients (ADA 16.10) — the exception is type 2 diabetes or stress hyperglycemia with mild hyperglycemia that stays <180 mg/dL. Endocrine Society 2022 suggests starting with correction insulin alone when there is no prior diabetes (R10.1), correction or scheduled insulin for diet- or non-insulin-treated diabetes (R10.2), and continuing a home scheduled insulin regimen modified for nutrition and illness severity (R10.3). In type 1 diabetes, correction alone is contraindicated and basal insulin is never held.

Weight (insulin-naive): ADA 2026 gives 0.3–0.6 U/kg/day as the usual start. The tiers come from Endocrine Society 2012 Table 1 (derived from RABBIT 2; not replaced in the 2022 update): 0.4 U/kg/day for glucose 140–200 mg/dL, 0.5 U/kg/day for 201–400, and 0.2–0.3 U/kg/day for age ≥70 y and/or GFR <60. The reduced tier also applies for creatinine ≥2.0 mg/dL (RABBIT 2 Surgery: 0.3), frailty (2024 crisis consensus: ≈0.3) and hypoglycemia earlier in the admission (ADA 16.13 regimen review). In eGFR <45, 0.25 vs 0.5 U/kg/day gave the same glucose with half the hypoglycemia (Baldwin 2012). If eGFR is blank it is calculated from creatinine with CKD-EPI 2021 (needs age and sex). Above 400 mg/dL the tiers were not studied — rule out DKA/HHS first. Half basal, half prandial over three meals.

Home insulin: each component is scaled separately and the home basal:prandial ratio is kept — never re-split. ADA 2026 allows a starting dose of 80% of the home dose, weighing adherence, nutrition and kidney function; the 2024 crisis consensus and ES 2022 R10.3 support continuing the usual dose, adjusted. The tool shows 80–100% of each component, and only the 80% end when hypoglycemia risk is flagged. It does not show any cut below 80% outside the NPO/procedure rule (analog 75–80%, NPH one-half; ADA notes this may be inappropriate for some people with type 1 diabetes).

Where the tool stops and asks for the diabetes team: concentrated insulin at home (U-200/U-300/U-500 — U-500 users' inpatient needs can fall by up to 75% versus home dosing; ADA 2026); a home total or estimated total above 200 U/day (a limit of this tool, not a clinical threshold); and type 1 diabetes already on insulin selected on the weight method (use the home regimen). The weight method is offered in type 1 only for a new diagnosis, with a specialist caveat — the tiers come from type 2 trials.

IV infusion: stable rate over the last 6–8 h × 24, then about 60% to allow for glucotoxicity (ADA 2026); half basal, half nutritional; give basal 2 h before stopping the infusion. Rates above 15 U/h are treated as implausible input.

Units of rapid-acting or regular insulin per dose; half dose at bedtime. Column choice, in order: (1) sensitive if not eating, age ≥70, kidney dysfunction (eGFR <60 or creatinine ≥2.0), frailty, or any glucose <100 mg/dL / hypoglycemia this admission (ES 2012; ADA 16.13); (2) resistant for glucocorticoids or a home dose >80 U/day (ES 2012); (3) otherwise the ADA 2026 tier from the estimated daily dose including steroid NPH — low <40 → sensitive, 40–80 → usual, >80 → usual (ES "usual" is the default for eating patients, so the tool does not escalate on the tier alone). ADA starts correction at 140–150 mg/dL.

Enteral/parenteral (ADA 2026): orders must cover basal, nutritional and correction needs; continue basal; nutritional insulin ≈1 U per 10–15 g carbohydrate; do not cover all needs with basal. Continuous feeds: rapid-acting every 4 h, regular every 6 h or NPH every 8–12 h (regular-based plans gave better control with fewer injections); correction regular every 6 h or rapid-acting every 4 h; if feeds stop, IV dextrose for the duration of active insulin. Bolus feeds: ≈1 U rapid-acting per 10–15 g before each feed. Nocturnal feeds: NPH at the start of the feed. Avoid pushing glucose below 140 mg/dL during continuous nutrition. PN: regular insulin in the bag, 1 U per 10 g dextrose (particularly if >20 U correction in 24 h), adjusted daily. ES 2022 R6.1: NPH-based or basal-bolus regimens for enteral nutrition.

Glucocorticoids: NPH with once- or twice-daily prednisone/prednisolone, in addition to basal-bolus or oral therapy (ADA 2026; ES 2022 R2.1: NPH-based or basal-bolus). Starting NPH by steroid dose (Suh & Park 2017 Table 3, after Clore 2009): prednisone ≥40 mg (dexamethasone ≥8 mg) 0.4 U/kg; 30 (6) 0.3; 20 (4) 0.2; 10 (2) 0.1 U/kg/day. Doses between rows use the lower row; below 10 mg prednisone (2 mg dexamethasone) no NPH dose is defined. Dexamethasone and multi-dose or continuous steroids may need long-acting basal; higher doses often need 40–60% more prandial and correction insulin. Adjust daily — a sudden taper causes hypoglycemia. Steroid NPH counts toward the daily total used for the correction tier.

NPO / procedures (ADA 2026): basal ± correction every 4–6 h with glucose every 4–6 h while not eating; every 2–4 h perioperatively. Before surgery reduce long-acting analog to 75–80% and NPH to one-half of the usual dose — individualized, as the reduction may not be appropriate for some people with type 1 diabetes.

Level 1: 54–69 mg/dL; level 2: <54 mg/dL; level 3: altered mental or physical function needing assistance. Treat any glucose <70 mg/dL: 15 g fast-acting carbohydrate if able to swallow and not NPO, otherwise IV glucose or glucagon; recheck every 15 min and repeat until >70. Review and change the regimen whenever a value <70 mg/dL is documented.