Executive Summary
JIL Sovereign's Documentation Integrity Index computes, for each of a plurality of workers, a composite raw score from component metrics derived from that worker's own sealed source documentation. Rather than benchmarking every worker against one blended organization-wide distribution, the system assigns each worker to a peer cohort defined by the combination of discipline and visit or engagement type, then z-scores the worker's composite raw score against the distribution within that cohort only - before mapping the result onto a bounded 0-100 index.
Every emitted index record carries a structural guardrail: a machine-readable flag asserting that a human reviewer retains final determination authority, and full traceability of each component metric back to the specific sealed source record that produced it. The guardrail is built into the output's data structure, not just stated in an accompanying policy document.
Problem Statement
Documentation-quality scorecards in healthcare, particularly in home health and hospice settings, commonly compare all clinical staff against one blended benchmark regardless of discipline or visit type. That produces systematically distorted results: a hospice registered nurse handling complex end-of-life visits, scored against a benchmark dominated by routine aide visits, will look artificially weak, and the reverse distortion applies in other combinations. Separately, these scorecards are frequently repurposed by management as a de facto performance-management or termination input even when that was never the intended use, because nothing in the software itself prevents that repurposing.
Why Existing Approaches Are Insufficient
- Blended organization-wide benchmarks: Compare workers across incompatible disciplines and visit types, producing scores that reflect cohort mix as much as individual documentation quality.
- Generic quality-scorecard products (e.g., Press Ganey-style tools): Typically compare across roles without discipline/visit-type-specific peer cohorts, and carry no structural restriction on downstream use.
- Policy-only use restrictions: An HR policy stating "do not use this score for termination decisions" constrains nothing in the software itself and is easy to disregard in practice.
Technical Architecture
| Stage | Computation |
|---|---|
| 1. Component metrics | Derived per worker from that worker's sealed source documentation records |
| 2. Composite raw score | Combination of the component metrics into a single raw score per worker |
| 3. Cohort assignment | Worker assigned to a peer cohort keyed on (discipline, visit_type) |
| 4. Normalization | z = (raw_score - cohort_mean) / cohort_stddev, computed within the assigned cohort only |
| 5. Index mapping | Normalized z-score mapped onto a bounded 0-100 index |
Structural (Not Policy) Guardrails
The distinction between a written HR policy against misuse and a structural guardrail is the difference between a rule a person can choose to ignore and a property built into the data itself. Every emitted score record in JIL's index carries a machine-readable flag asserting that a human reviewer retains final determination authority, and each component metric remains glass-box decomposable back to the specific sealed source record that produced it. Any adverse use of the index therefore requires a human to consciously open and review the underlying documentation rather than act on the numeric index alone.
Prior Art Differentiation
| Approach | Peer-Cohort Normalized? | Traceable to Source Records? | Structural Use Guardrail? |
|---|---|---|---|
| Blended organization-wide scorecards | No | Rarely | No |
| Generic HR / quality-management dashboards | Sometimes, by role only | Rarely | Policy only |
| JIL Sovereign Documentation Integrity Index | Yes, discipline x visit-type cohort | Yes, every component metric | Yes, embedded in output schema |