Explainability Framework — SM LDLT Risk Assessor v2.0
Author: Dr. Sharad Maheshwari • Purpose: Transparent mapping of rules, thresholds and rationale used to compute the DRPSS
1. Overview
This document provides a complete, auditable explanation of how each sub-index and the composite Donor–Recipient Profile Safety Score (DRPSS) is computed. It is written for liver transplant surgeons, hepatologists, transplant coordinators and thesis reviewers.
| Domain | Index | Purpose | Default Weight |
|---|---|---|---|
| Donor Technical Complexity | DTC | Anatomic & technical feasibility | 0.4 |
| Recipient Resilience | RRI | Physiologic tolerance & recovery potential | 0.3 |
| Volumetric Safety | VSI | Quantitative graft/remnant safety | 0.2 |
| Socioeconomic Readiness | SRS | Support systems & logistics | 0.1 |
Note: All thresholds and point values are fully traceable in the sections below. We recommend local validation and weight tuning before institutional use.
2. Donor Technical Complexity Index (DTC)
Purpose: Quantifies anatomical and reconstructive complexity using established classifications (Huang, Nakamura, Michels, Cheng/Essen).
Subcomponents and rules
| Sub-component | Classification | Rule (example values) |
|---|---|---|
| BCI (Biliary) | Huang A1–A5 | A1 (single duct) = 100 pts; A2–A5 = 50–70 pts (multi-ducts worse) |
| PVCI (Portal Vein) | Nakamura A–E | A = 100; B = 75; C = 55; D = 50; E (single PV supply) = 0 (absolute contraindication) |
| ACI (Arterial) | Michels I–X | I = 100; variants range 85→40 depending on complexity |
| VCI (Venous) | Cheng/Essen | Type I = 100; complex venous patterns = 45–85 depending on reconstruction need |
Computation
DTC is the simple mean of the four subcomponent scores:
DTC = (BCI + PVCI + ACI + VCI) / 4
Interpretation
>80 Favorable • 70–80 Acceptable • 50–69 High complexity • ≤50 Extreme risk / contraindicated
3. Volumetric Safety Index (VSI)
Purpose: Ensure donor safety (adequate remnant) and recipient graft sufficiency.
Inputs & Rules
| Variable | Calculation | Scoring rule (points) |
|---|---|---|
| GRWR (%) | (Graft volume / Recipient weight) / 10 | ≥0.8 = 20; 0.6–0.79 = 10; <0.6 = 0 |
| RLV (%) | ((TLV – Graft) / TLV) × 100 | ≥40 = 20; 35–39 = 10; 30–34 = 5; <30 = 0 |
| Steatosis (%) | Biopsy or MRI-PDFF | <5 = 20; 5–10 = 10; 10–20 = 5; >20 = 0 |
VSI is normalized to 0–100 using:
VSI = ((GRWR_pts + RLV_pts + Steatosis_pts) / 60) × 100
Hard stops
RLV < 30% → Donation contraindicated (hard stop).
VSI = 0 → Automatic fail trigger (critical volumetric/graft concerns).
4. Recipient Resilience Index (RRI)
Purpose: Objectively measure recipient physiologic reserve and post-op recovery potential.
| Variable | Rule | Rationale |
|---|---|---|
| MELD-Na | ≤15 = 20; 16–25 = 10; >25 = 0 | Higher MELD correlates with worse outcomes in LDLT |
| Liver Frailty Index (LFI) | <3.2 = 20; 3.2–4.4 = 10; >4.4 = 0 | Objective frailty metric linked to mortality |
| Dialysis | No = +10; On dialysis = −10 | Dialysis increases perioperative risk |
| BMI | 18.5–40 = +10; otherwise 0 | Extremes of BMI increase complications |
| Psychosocial support | Strong = 10; Limited = 5; None = 0 | Adherence and recovery depend on support |
Normalization formula:
RRI = ((Raw_pts + 10) / 80) × 100
Raw_pts are allowed to be negative (e.g., dialysis = −10). The normalization shifts range [-10,70] → [0,100].
5. Socioeconomic Readiness Score (SRS)
Purpose: Operationalize social determinants of health to flag non-medical risks that impact recovery and follow-up.
| Factor | Points |
|---|---|
| Insurance | Private = 20; Public = 10; Uninsured = 0 |
| Education | Graduate = 15; Some college = 10; High school or less = 5 |
| Employment | Paid leave = 15; No paid leave = 10; Unemployed = 0 |
| Caregiver support | Strong = 20; Limited = 10; None = 0 |
| Residence stability | Stable = 15; Unstable = 5 |
SRS = (Total_pts / 85) × 100
Ethical note: SRS is a mitigation trigger, not an exclusion criterion. Low SRS should prompt social work interventions and logistical support.
6. Composite Score — DRPSS
The DRPSS is a weighted sum of the four indices. Default weights reflect clinical priorities but are tunable for institutional needs.
DRPSS = (0.4 × DTC) + (0.3 × RRI) + (0.2 × VSI) + (0.1 × SRS)
Risk bands & guidance
| DRPSS range | Category | Guidance |
|---|---|---|
| ≥85 | Excellent | Low-risk pairing; proceed |
| 70–84 | Acceptable | Proceed with vigilance |
| 50–69 | Marginal | MDT review required; optimize modifiable risks |
| <50 | Unsafe | Not recommended without exceptional justification |
Hard-stop conditions (immediate contraindication): RLV <30%, VSI = 0, DTC ≤ 50, or Single PV Supply (Nakamura E).
7. Explainability map & audit trace
Every computed value in the DRPSS can be traced to an atomic decision node containing: the input, the rule applied, and the literature/clinical rationale. The app exports a per-case text/PDF summary with this trace.
Example
RLV = 28% → RLV_pts = 0 → VSI = 0 → Hard Stop (RLV <30%)
Visualization (explainability tree)
Each leaf node in this tree has an explicit rule and a literature note embedded in the exported report.
8. Ethical safeguards
- Decision support not decision making: DRPSS informs discussion but does not mandate exclusion.
- SRS triggers mitigation: Low SRS should initiate social work, financial counseling and logistical support.
- Transparency by design: All thresholds, sources and overrides are auditable in the case report.
9. Auditability example (chain of explanation)
The exported per-case report contains the full calculation chain. Example:
Input: Donor: Michels IV, RLV=28%, Steatosis=15% Recipient: MELD=22, LFI=3.8, strong support → DTC = 60 (Complex: Michels IV mapping) → VSI = 0 (RLV < 30%) → RRI = 68 → SRS = 80 → DRPSS = 0.4(60)+0.3(68)+0.2(0)+0.1(80) = 49.2 → Final: UNSAFE (primary driver: RLV <30% [guideline])
10. Core references (for report provenance)
Hiatt JR et al.; Nakamura T et al.; Huang TL et al.; Cheng Y et al.; Yamaoka Y et al.; Kiuchi T et al.; Lai JC et al.; DiMartini A et al.; McCormack L et al.; EASL Clinical Practice Guidelines; WHO Health Systems Resilience Index.
REFERENCES (Core Citations for Transparency)
- Hiatt JR et al. Ann Surg 1994; 220(1):3–9. (Michels classification relevance)
- Nakamura T et al. Transplant Proc 2002; 34(6):2679–2680. (Portal vein variants)
- Huang TL et al. Transplant Proc 1996; 28(3):1669–1670. (Biliary anatomy)
- Cheng Y et al. Transplantation 2010; 90(11):1304–1310. (Venous variants)
- Yamaoka Y et al. Liver Transpl 2015; 21(4):524–534. (Donor RLV thresholds)
- Kiuchi T et al. Transplantation 1999; 67(9):1315–1320. (GRWR thresholds)
- Lai JC et al. Hepatology 2020; 71(4):1219–1229. (Liver Frailty Index)
- DiMartini A et al. Transplantation 2017; 101(2):291–300. (Psychosocial predictors)
- McCormack L et al. Ann Surg 2011; 253(1):78–84. (Steatosis & outcomes)
- EASL Clinical Practice Guidelines, J Hepatol 2023. (Living donor safety)
- WHO Health Systems Resilience Index, 2019. (Socioeconomic determinants)
Comments
Post a Comment