Table 3. GIRISH applied to a deteriorating 28-week infant.
AAP: American Academy of Pediatrics; BP: Blood pressure; BPD: Bronchopulmonary dysplasia; BW: Birth weight; CRP: C-reactive protein; CXR: Chest radiograph; DOL: Day of life; FiO2: Fraction of inspired oxygen; GA: Gestational age; GIRISH: Goal, Input, Role, Iterative refinement, Safety verification, and Human accountability; HR: Heart rate; MAP: Mean arterial pressure; NNF: National Neonatology Forum (India); PDA: Patent ductus arteriosus; PMA: Post-menstrual age; RDS: Respiratory distress syndrome; SIMV: Synchronized intermittent mandatory ventilation; WBC: White blood cell count.
| GIRISH Step | Application in this case |
| Scenario | 28w, 950 g male, DOL 7, on SIMV, FiO2 0.55 (up from 0.38 over 6 h). CRP 48 mg/L, WBC 18.4 x 109/L (74% neutrophils), temp 37.9℃, feeds 80 mL/kg/day. Last CXR (DOL 5): moderate RDS. Medications: caffeine, ampicillin day 7 (gentamicin completed day 3, blood cultures negative). |
| G – Goal | “Generate a ranked differential for acute clinical deterioration and recommend evidence-based initial investigations. Do not suggest management yet.” |
| I – Input | GA 28w, BW 950 g, DOL 7. Ventilated (SIMV). FiO2 rising from 0.38 to 0.55 over 6 h. CRP 48, WBC 18.4 x 109/L (74% neutrophils), temp 37.9℃, HR 176, BP 38/22 (MAP 27). Feeds 80 mL/kg/day, tolerating. Last CXR DOL 5: moderate RDS, no new changes. Ampicillin day 7, caffeine. Gentamicin completed day 3 (cultures negative). Echo not yet performed today. Fully anonymized. |
| R – Role | “Act as a neonatal evidence-retrieval assistant. Rank differentials by clinical probability. Flag an evidence grade for each. State what you are uncertain about.” |
| AI response (abbreviated) | (1) Late-onset sepsis: high probability given CRP, WBC, and rising FiO2, with Gram-negative organisms most likely after 7 days of ampicillin. (2) Ventilator-associated pneumonia: moderate; ventilated over 5 days, CXR unremarkable but cannot exclude. (3) Evolving chronic lung-disease trajectory: moderate; FiO2 trend consistent with evolving lung injury at 28w – note that a formal BPD diagnosis requires reassessment at 36 weeks PMA and is not applicable at DOL 7. (4) PDA hemodynamic compromise: lower probability without echo; cannot exclude. |
| I – Iterative refinement | The clinician asks for the evidence grade for adding Gram-negative cover after 7 days of ampicillin in a 28-weeker; the AI references AAP/NNF India guidance and prompts a local antibiogram check. Challenged on terminology (“Can BPD be diagnosed at DOL 7?”), the AI clarifies that BPD requires assessment at 36 weeks PMA or 28 days of age and that “evolving chronic lung disease” or “worsening RDS” is more accurate at DOL 7. The clinician then asks which findings would raise or lower the probability of PDA compromise without echo. |
| S – Safety verification | Gentamicin dose verified against Neofax for GA 28w. Local antibiogram checked for the Gram-negative sensitivity pattern. No hallucination red flags identified. Ibuprofen not suggested (correct, given the low MAP). |
| H – Human accountability | Record: “GIRISH AI-assisted differential generation, information-retrieval adjunct. Clinical decision to commence gentamicin in addition to ampicillin and to perform echo at 06:00 by (Consultant Name), Neonatology.” |