21 lines
3.3 KiB
Markdown
21 lines
3.3 KiB
Markdown
```markdown
|
||
# Clinical Refactor — SOFA / GCS Interview Questions
|
||
|
||
## Q1: Why did you replace SIRS with SOFA for sepsis detection?
|
||
|
||
SIRS (Systemic Inflammatory Response Syndrome) uses four criteria — temperature, heart rate, respiratory rate, white blood cell count — to screen for sepsis. The problem is non-specificity: a patient who just exercised, is anxious, or has a mild viral infection can meet ≥ 2 criteria. The Sepsis-3 consensus (2016) replaced SIRS with SOFA (Sequential Organ Failure Assessment) because SOFA measures actual organ dysfunction across six systems. A delta SOFA ≥ 2 from baseline in the presence of suspected infection is both more specific and more clinically actionable than SIRS. We kept qSOFA as a bedside screening tool that recommends ordering SOFA labs, matching the Sepsis-3 two-tier approach.
|
||
|
||
## Q2: How does your system handle the fact that SOFA requires lab values that aren't continuously available?
|
||
|
||
Three strategies: **carry-forward with staleness** — the most recent lab value is cached in Redis with a configurable TTL (default 24 hours) and classified as CURRENT (< 12h), STALE (12–24h), or EXPIRED (> 24h). Stale values are still used for scoring but flagged in `sofa_scores.staleness_flags` JSON so clinicians know the score is based on older data. Second, **SpO₂/FiO₂ fallback** — when arterial blood gas isn't available (common on general wards), we use the SpO₂/FiO₂ ratio as a proxy for the respiratory component, per Rice et al. (2007). Third, **partial scoring** — SOFA baseline is only established when ≥ 4 of 6 organ systems have data, preventing false low baselines that would inflate the delta.
|
||
|
||
## Q3: Why did you keep qSOFA after removing SIRS, and how does the screening workflow differ from the old SIRS alert?
|
||
|
||
qSOFA is a validated bedside screening tool — three simple criteria (respiratory rate, blood pressure, mental status) that any nurse can assess without labs. The key change is clinical positioning: under SIRS, meeting ≥ 2 criteria immediately declared sepsis and triggered the treatment bundle. Under the new workflow, qSOFA ≥ 2 creates a WARNING-level screening alert (`QSOFA_SCREEN`) that recommends ordering SOFA labs (PaO₂/FiO₂, platelets, bilirubin, creatinine). Only when those labs confirm organ dysfunction — SOFA delta ≥ 2 — does the system declare sepsis (`SOFA_SEPSIS`) and trigger the bundle. This prevents false-positive bundle activations that occurred with SIRS.
|
||
|
||
## Q4: How did you handle the GCS → SOFA → NEWS2 → qSOFA dependency chain?
|
||
|
||
GCS feeds into three downstream systems: SOFA CNS scoring (GCS → 0–4 organ score), NEWS2 consciousness (GCS 15 = score 0, GCS < 15 = score 3), and qSOFA altered mentation (GCS < 15 = criterion met). Architecturally, the GCS Kafka consumer computes the total and caches it in Redis. Downstream consumers (SOFA, NEWS2, qSOFA) read GCS from Redis when triggered. The resolution order is **GCS-first with AVPU-fallback** — if GCS components exist, they take priority; if only AVPU is available (legacy data), the old mapping still works. This avoided a breaking migration while making GCS the preferred consciousness assessment going forward.
|
||
```
|
||
|
||
Also update `docs/clinical-scoring-refactor-summary.md` — add a "Phase 29 scenarios" section listing the three new scenario IDs and what each validates. |