data: evidence base builder + literature methods doc (refs #237, refs #244)

- Collection hierarchy: 20 sub-collections under "Skin Substitutes" (Clinical Evidence, CMS Policy, Enforcement, etc.) - COI/funding enrichment: 2,862 PubMed articles tagged with coi:single-product (2,003), stance:skeptical (598), stance:favorable (510), coi:industry-linked (11) - All 22 expected tag types verified present - Methods documentation: search queries, PRISMA flow, tag schema, collection hierarchy, limitations
2026-03-25 11:33:32 -04:00
parent b822aa9443
commit c1f60e16ef
2 changed files with 553 additions and 0 deletions
--- a/dev/docs/skin_subs_literature_methods.md
+++ b/dev/docs/skin_subs_literature_methods.md
@@ -0,0 +1,167 @@
+# Skin Substitutes Literature Review: Methods
+
+## Search Strategy
+
+### Databases
+- **PubMed** (MEDLINE) via NCBI E-utilities (esearch + efetch XML)
+- **Grey literature**: OIG, CMS, GAO, MedPAC, DOJ, court filings, MAC LCDs, industry
+
+### Date Range
+- PubMed: inception through 2026-03-25
+- Grey literature: 2015-current (enforcement), 2012-current (CMS rules)
+
+### Search Date
+2026-03-25
+
+## PubMed Search Queries
+
+### Domain 1: Clinical Efficacy
+```
+("skin substitutes"[MeSH] OR "biological dressings"[MeSH]
+ OR "tissue scaffolds"[MeSH] OR "bioengineered skin"[tiab]
+ OR "cellular tissue product"[tiab] OR "wound matrix"[tiab]
+ OR "skin substitute"[tiab] OR "skin substitutes"[tiab]
+ OR "Apligraf"[tiab] OR "EpiFix"[tiab] OR "Grafix"[tiab]
+ OR "DermACELL"[tiab] OR "Dermagraft"[tiab])
+AND ("wound healing"[MeSH] OR "treatment outcome"[MeSH]
+ OR "clinical trial"[pt] OR "randomized controlled trial"[pt]
+ OR "systematic review"[pt] OR "meta-analysis"[pt]
+ OR "efficacy"[tiab] OR "effectiveness"[tiab])
+```
+**Result: 7,448 records**
+
+### Domain 2: Cost-Effectiveness
+```
+("skin substitutes"[MeSH] OR "biological dressings"[MeSH]
+ OR "skin substitute"[tiab] OR "skin substitutes"[tiab]
+ OR "cellular tissue product"[tiab])
+AND ("costs and cost analysis"[MeSH]
+ OR "cost-benefit analysis"[MeSH]
+ OR "health care costs"[MeSH] OR "medicare"[MeSH]
+ OR "reimbursement"[tiab] OR "cost-effectiveness"[tiab]
+ OR "economic"[tiab] OR "spending"[tiab]
+ OR "average sales price"[tiab] OR "ASP"[tiab])
+```
+**Result: 123 records**
+
+### Domain 3: Fraud, Waste, and Abuse
+```
+("skin substitutes"[MeSH] OR "biological dressings"[MeSH]
+ OR "skin substitute"[tiab] OR "skin substitutes"[tiab]
+ OR "cellular tissue product"[tiab] OR "wound care"[tiab])
+AND ("fraud"[MeSH] OR "waste"[tiab] OR "abuse"[tiab]
+ OR "inappropriate"[tiab] OR "overutilization"[tiab]
+ OR "billing"[tiab] OR "enforcement"[tiab]
+ OR "compliance"[tiab] OR "medically unnecessary"[tiab]
+ OR "upcoding"[tiab] OR "kickback"[tiab])
+```
+**Result: 358 records**
+
+## PRISMA Flow
+
+```
+Records identified through PubMed:
+  Clinical efficacy:     7,448
+  Cost-effectiveness:      123
+  Fraud/waste/abuse:       358
+  ─────────────────────────────
+  TOTAL identified:      7,929
+
+Duplicates removed:        112
+Unique articles:         7,817
+
+By publication type:
+  Meta-analyses:            85
+  Systematic reviews:    1,395
+  RCTs:                    455
+  Other (observational,
+   case reports, etc.):  5,882
+```
+
+## Grey Literature Sources (20 documents)
+
+| Source    | Count | Description                                          |
+|-----------|------:|------------------------------------------------------|
+| OIG       |     3 | Sept 2025 payment trends, SAB, audit reports         |
+| CMS       |     4 | OPPS/PFS final rules (2024-2026), Benefit Policy     |
+| DOJ       |     1 | National healthcare fraud enforcement action          |
+| Court     |     3 | Jenson (S.D. Tex.), Gehrke/King (D. Ariz.), Vohra   |
+| GAO       |     1 | Part B biologicals spending (GAO-23-105537)          |
+| MedPAC    |     2 | June 2024, March 2025 Reports to Congress            |
+| MAC LCD   |     4 | Noridian, CGS, First Coast, Palmetto                 |
+| Industry  |     2 | Alliance position statement, WHS guidelines          |
+
+## Conflict of Interest / Stance Enrichment
+
+PubMed articles were enriched with automated COI and stance detection:
+
+| Tag                   | Count | Method                                          |
+|-----------------------|------:|-------------------------------------------------|
+| coi:single-product    | 2,003 | Only one brand name mentioned in title/abstract |
+| stance:skeptical      |   598 | Negative outcome language in abstract           |
+| stance:favorable      |   510 | Positive outcome language in title              |
+| coi:industry-linked   |    11 | Funding/employment patterns in extra metadata   |
+
+**Enrichment coverage:** 2,862 / 7,817 PubMed articles (36.6%)
+
+## Tag Schema
+
+All items tagged with `module:skin-subs` plus:
+
+- **Source tags:** `source:pubmed`, `source:oig`, `source:cms`, `source:court`,
+  `source:doj`, `source:gao`, `source:medpac`, `source:mac-lcd`, `source:industry`
+- **Type tags:** `type:clinical`, `type:economic`, `type:fraud`, `type:rct`,
+  `type:review`, `type:meta-analysis`, `type:report`, `type:rule`, `type:filing`,
+  `type:lcd`, `type:position`, `type:press-release`, `type:manual`
+- **Year tags:** `year:YYYY`
+- **COI tags:** `coi:industry-linked`, `coi:single-product`
+- **Stance tags:** `stance:skeptical`, `stance:favorable`
+- **Entity tags:** `entity:oig`, `entity:doj`, `entity:gao`, `entity:medpac`,
+  `entity:noridian`, `entity:cgs`, `entity:first-coast`, `entity:palmetto`
+- **Case tags:** `case:jenson`, `case:gehrke-king`, `case:vohra`
+- **Rule tags:** `rule:cy2026-opps`, `rule:cy2025-opps`, `rule:cy2024-pfs`
+
+## Collection Hierarchy
+
+```
+Skin Substitutes/
+  Clinical Evidence/
+    RCTs                    (455 items)
+    Systematic Reviews    (1,395 items)
+    Meta-Analyses            (85 items)
+    Observational Studies (5,547 items)
+  CMS Policy/
+    Final Rules (OPPS/PFS)    (3 items)
+    Benefit Policy Manuals    (1 item)
+    ASP Pricing Files
+  OIG Reports                 (3 items)
+  GAO & MedPAC                (3 items)
+  Enforcement/
+    DOJ Press Releases        (1 item)
+    Court Filings             (3 items)
+  MAC LCDs                    (4 items)
+  Market Data
+  Industry & Societies        (2 items)
+  Cost-Effectiveness         (70 items)
+  Fraud & Abuse Literature  (265 items)
+```
+
+## Scripts
+
+| Script | Purpose |
+|--------|---------|
+| `dev/scripts/search_pubmed_skin_subs.py` | PubMed E-utilities search, XML parsing, bib.sqlite storage |
+| `dev/scripts/collect_grey_lit_skin_subs.py` | Curated grey literature catalogue, bib.sqlite storage |
+| `dev/scripts/build_skin_subs_evidence_base.py` | Collection hierarchy, COI enrichment, verification |
+
+## Limitations
+
+1. **No full-text screening**: Articles included based on PubMed metadata only; no manual
+   title/abstract screening for relevance (full PRISMA would require human review)
+2. **COI detection is heuristic**: Based on keyword patterns in abstracts and metadata,
+   not full-text disclosure sections
+3. **Grey literature is curated, not systematic**: Known key documents captured; no
+   systematic search of OIG/GAO/PACER databases
+4. **No Cochrane Library**: Only PubMed searched for journal literature
+5. **No citation network analysis**: Forward/backward snowball not yet performed (#237)
+6. **No quantitative meta-analysis**: Study data extraction and pooling not yet done (#237)
--- a/dev/scripts/build_skin_subs_evidence_base.py
+++ b/dev/scripts/build_skin_subs_evidence_base.py
@@ -0,0 +1,386 @@
+"""Build the Skin Substitutes evidence base in bib.sqlite.
+
+Creates collection hierarchy, assigns items to sub-collections,
+enriches PubMed articles with conflict-of-interest and funding
+analysis, and verifies tag schema coverage.
+
+Addresses issues #244 (Zotero evidence base) and #237 (partial —
+COI/funding enrichment of PubMed results).
+
+Usage:
+    uv run python dev/scripts/build_skin_subs_evidence_base.py
+    uv run python dev/scripts/build_skin_subs_evidence_base.py --dry-run
+"""
+
+from __future__ import annotations
+
+import argparse
+import re
+from datetime import datetime
+
+from bib.store import Store
+
+# ---------------------------------------------------------------------------
+# Collection hierarchy for skin-subs research
+# ---------------------------------------------------------------------------
+
+SKIN_SUBS_COLLECTIONS = {
+    "Skin Substitutes": {
+        "Clinical Evidence": {
+            "RCTs": {},
+            "Systematic Reviews": {},
+            "Meta-Analyses": {},
+            "Observational Studies": {},
+        },
+        "CMS Policy": {
+            "Final Rules (OPPS/PFS)": {},
+            "Benefit Policy Manuals": {},
+            "ASP Pricing Files": {},
+        },
+        "OIG Reports": {},
+        "GAO & MedPAC": {},
+        "Enforcement": {
+            "DOJ Press Releases": {},
+            "Court Filings": {},
+        },
+        "MAC LCDs": {},
+        "Market Data": {},
+        "Industry & Societies": {},
+        "Cost-Effectiveness": {},
+        "Fraud & Abuse Literature": {},
+    },
+}
+
+# ---------------------------------------------------------------------------
+# Known manufacturer names for COI detection
+# ---------------------------------------------------------------------------
+
+MANUFACTURERS = [
+    "organogenesis", "mimedx", "smith nephew", "smith & nephew",
+    "integra", "solsys", "amnioexcel", "derma sciences",
+    "healthpoint", "shire", "acelity", "kci", "3m",
+    "molnlycke", "medline", "hollister", "coloplast",
+    "stryker", "zimmer biomet", "wright medical",
+    "solventum", "apria", "anika", "musculoskeletal transplant",
+    "surmodics", "tissue regenix", "nuo therapeutics",
+    "sanara medtech", "kerecis", "human bioprocessing",
+    "alphatec", "biosig technologies",
+]
+
+# Brand names that indicate manufacturer-linked studies
+BRAND_NAMES = [
+    "apligraf", "dermagraft", "epifix", "grafix", "amnioexcel",
+    "dermacell", "oasis", "primatrix", "integra", "graftjacket",
+    "dermapure", "affinity", "biovance", "cytal", "endoform",
+    "kerecis omega3", "novafix", "puraply", "restorigin",
+    "surgicraft", "theraskin", "amnioburn", "clarix",
+    "epicord", "genesis", "grafix core", "grafix prime",
+    "innovamatrix", "nushield", "stravix", "woundex",
+]
+
+# Patterns suggesting industry funding
+FUNDING_PATTERNS = [
+    r"funded by .{0,50}(organogenesis|mimedx|smith|integra|solsys|amnio)",
+    r"grant from .{0,50}(organogenesis|mimedx|smith|integra|solsys)",
+    r"financial support.{0,50}(organogenesis|mimedx|smith|integra)",
+    r"supported by .{0,50}(organogenesis|mimedx|smith|integra|solsys)",
+    r"sponsored by .{0,50}(organogenesis|mimedx|smith|integra|solsys)",
+    r"employee of .{0,50}(organogenesis|mimedx|smith|integra|solsys)",
+    r"consultant.{0,30}(organogenesis|mimedx|smith|integra|solsys)",
+    r"speaker.{0,30}(organogenesis|mimedx|smith|integra|solsys)",
+    r"advisory board.{0,30}(organogenesis|mimedx|smith|integra|solsys)",
+    r"honorari.{0,30}(organogenesis|mimedx|smith|integra|solsys)",
+    r"conflict.{0,50}interest",
+    r"disclosur.{0,80}(stock|equity|consult|employ|honorar|speaker|grant)",
+]
+
+# Patterns suggesting independent/skeptical perspective
+SKEPTICAL_PATTERNS = [
+    r"no (significant )?difference",
+    r"lack.{0,20}evidence",
+    r"insufficient evidence",
+    r"low.{0,15}quality evidence",
+    r"limited evidence",
+    r"no.{0,15}superiority",
+    r"similar (outcome|result|efficacy|effectiveness)",
+    r"(waste|fraud|abuse|overutiliz|unnecessary|inappropriate)",
+    r"(overspend|excessive.{0,15}cost|cost concern)",
+    r"(marketing|promotional|commercial bias)",
+]
+
+
+# ---------------------------------------------------------------------------
+# Collection assignment rules
+# ---------------------------------------------------------------------------
+
+
+def assign_collection(tags: list[str], title: str, abstract: str) -> str:
+    """Return the most specific sub-collection name for an item."""
+    tag_set = set(tags)
+
+    # Grey literature — by source tag
+    if "source:oig" in tag_set:
+        return "OIG Reports"
+    if "source:gao" in tag_set or "source:medpac" in tag_set:
+        return "GAO & MedPAC"
+    if "source:doj" in tag_set:
+        return "DOJ Press Releases"
+    if "source:court" in tag_set:
+        return "Court Filings"
+    if "source:mac-lcd" in tag_set:
+        return "MAC LCDs"
+    if "source:industry" in tag_set:
+        return "Industry & Societies"
+    if "source:cms" in tag_set:
+        for t in tag_set:
+            if t.startswith("type:rule"):
+                return "Final Rules (OPPS/PFS)"
+            if t.startswith("type:manual"):
+                return "Benefit Policy Manuals"
+        return "CMS Policy"
+
+    # PubMed — by type tag
+    if "type:meta-analysis" in tag_set:
+        return "Meta-Analyses"
+    if "type:review" in tag_set:
+        return "Systematic Reviews"
+    if "type:rct" in tag_set:
+        return "RCTs"
+    if "type:economic" in tag_set:
+        return "Cost-Effectiveness"
+    if "type:fraud" in tag_set:
+        return "Fraud & Abuse Literature"
+    if "type:clinical" in tag_set:
+        return "Observational Studies"
+
+    return "Skin Substitutes"  # fallback to root
+
+
+def detect_coi(extra: str, abstract: str, title: str) -> list[str]:
+    """Detect conflict-of-interest indicators, return enrichment tags."""
+    tags: list[str] = []
+    text = f"{extra}\n{abstract}\n{title}".lower()
+
+    # Check for manufacturer mentions in author affiliations
+    authors_section = ""
+    for line in extra.split("\n"):
+        if line.startswith("Authors:"):
+            authors_section = line.lower()
+            break
+
+    # Industry funding patterns
+    for pattern in FUNDING_PATTERNS:
+        if re.search(pattern, text, re.IGNORECASE):
+            tags.append("coi:industry-linked")
+            break
+
+    # Single-product studies (often manufacturer-funded)
+    brand_count = sum(1 for b in BRAND_NAMES if b in text)
+    if brand_count == 1:
+        tags.append("coi:single-product")
+
+    # Skeptical / critical perspective
+    for pattern in SKEPTICAL_PATTERNS:
+        if re.search(pattern, text, re.IGNORECASE):
+            tags.append("stance:skeptical")
+            break
+
+    # Pro-product sentiment (positive claims in title)
+    title_lower = title.lower()
+    pro_patterns = [
+        r"(effective|superior|promising|excellent|favorable|beneficial)",
+        r"(accelerat|improv|enhanc|advanc).{0,20}(heal|wound|outcome)",
+        r"(novel|innovative|breakthrough).{0,20}(treatment|therapy|approach)",
+    ]
+    for pattern in pro_patterns:
+        if re.search(pattern, title_lower):
+            tags.append("stance:favorable")
+            break
+
+    return tags
+
+
+# ---------------------------------------------------------------------------
+# Main
+# ---------------------------------------------------------------------------
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Build skin-subs evidence base"
+    )
+    parser.add_argument("--dry-run", action="store_true")
+    args = parser.parse_args()
+
+    print("=" * 70)
+    print("Skin Substitutes Evidence Base Builder")
+    print(f"Date: {datetime.now().strftime('%Y-%m-%d %H:%M')}")
+    print("=" * 70)
+
+    store = Store()
+    con = store._con()
+
+    # --- Step 1: Create collection hierarchy ---
+    print("\n--- Step 1: Creating collection hierarchy ---")
+    col_map = store.ensure_collections(SKIN_SUBS_COLLECTIONS)
+    for name, key in sorted(col_map.items()):
+        print(f"  {key}  {name}")
+    print(f"  Total collections created/verified: {len(col_map)}")
+
+    # --- Step 2: Load all skin-subs items ---
+    print("\n--- Step 2: Loading skin-subs items ---")
+    rows = con.execute(
+        """SELECT DISTINCT i.id, i.key, i.title, i.abstract, i.extra
+           FROM items i
+           JOIN item_tags it ON i.id = it.item_id
+           JOIN tags t ON it.tag_id = t.id
+           WHERE t.name = 'module:skin-subs'"""
+    ).fetchall()
+    print(f"  Total items: {len(rows)}")
+
+    # Pre-load tags for each item
+    item_tags: dict[int, list[str]] = {}
+    tag_rows = con.execute(
+        """SELECT it.item_id, t.name
+           FROM item_tags it
+           JOIN tags t ON it.tag_id = t.id
+           WHERE it.item_id IN (
+               SELECT DISTINCT i.id FROM items i
+               JOIN item_tags it2 ON i.id = it2.item_id
+               JOIN tags t2 ON it2.tag_id = t2.id
+               WHERE t2.name = 'module:skin-subs'
+           )"""
+    ).fetchall()
+    for tr in tag_rows:
+        item_tags.setdefault(tr["item_id"], []).append(tr["name"])
+
+    # --- Step 3: Assign to collections ---
+    print("\n--- Step 3: Assigning items to collections ---")
+    collection_counts: dict[str, int] = {}
+    assignments: list[tuple[int, str]] = []  # (item_id, collection_key)
+
+    for row in rows:
+        tags = item_tags.get(row["id"], [])
+        col_name = assign_collection(
+            tags, row["title"] or "", row["abstract"] or ""
+        )
+        if col_name in col_map:
+            assignments.append((row["id"], col_map[col_name]))
+            collection_counts[col_name] = (
+                collection_counts.get(col_name, 0) + 1
+            )
+
+    print("  Assignment distribution:")
+    for name, count in sorted(
+        collection_counts.items(), key=lambda x: -x[1]
+    ):
+        print(f"    {name:30s}: {count:>5}")
+
+    # --- Step 4: COI / funding enrichment ---
+    print("\n--- Step 4: COI / funding / stance enrichment ---")
+    coi_tags_to_add: dict[int, list[str]] = {}
+    coi_counts: dict[str, int] = {}
+
+    for row in rows:
+        tags = item_tags.get(row["id"], [])
+        # Only enrich PubMed items
+        if "source:pubmed" not in tags:
+            continue
+        new_tags = detect_coi(
+            row["extra"] or "", row["abstract"] or "", row["title"] or ""
+        )
+        if new_tags:
+            coi_tags_to_add[row["id"]] = new_tags
+            for t in new_tags:
+                coi_counts[t] = coi_counts.get(t, 0) + 1
+
+    print("  Enrichment tag counts:")
+    for tag, count in sorted(coi_counts.items(), key=lambda x: -x[1]):
+        print(f"    {tag:30s}: {count:>5}")
+    print(f"  Items enriched: {len(coi_tags_to_add)} / {len(rows)}")
+
+    if args.dry_run:
+        print("\n[DRY RUN] Skipping writes")
+        store.close()
+        return
+
+    # --- Step 5: Write collection assignments ---
+    print("\n--- Step 5: Writing collection assignments ---")
+    for item_id, col_key in assignments:
+        col_row = con.execute(
+            "SELECT id FROM collections WHERE key = ?", (col_key,)
+        ).fetchone()
+        if col_row:
+            con.execute(
+                "INSERT OR IGNORE INTO collection_items "
+                "(collection_id, item_id) VALUES (?, ?)",
+                (col_row["id"], item_id),
+            )
+    con.commit()
+    print(f"  Assigned {len(assignments)} items to collections")
+
+    # --- Step 6: Write COI enrichment tags ---
+    print("\n--- Step 6: Writing COI enrichment tags ---")
+    total_tags_added = 0
+    for item_id, new_tags in coi_tags_to_add.items():
+        for tag in new_tags:
+            tag_id = store._ensure_tag(tag)
+            con.execute(
+                "INSERT OR IGNORE INTO item_tags (item_id, tag_id) "
+                "VALUES (?, ?)",
+                (item_id, tag_id),
+            )
+            total_tags_added += 1
+    con.commit()
+    print(f"  Added {total_tags_added} enrichment tags")
+
+    # --- Step 7: Verify ---
+    print("\n--- Step 7: Verification ---")
+
+    # Tag schema coverage
+    expected_tags = [
+        "module:skin-subs",
+        "source:pubmed", "source:oig", "source:cms", "source:court",
+        "source:doj", "source:gao", "source:medpac",
+        "source:mac-lcd", "source:industry",
+        "type:clinical", "type:economic", "type:fraud",
+        "type:rct", "type:review", "type:meta-analysis",
+        "type:report", "type:rule", "type:filing",
+        "type:lcd", "type:position", "type:press-release",
+    ]
+    for tag in expected_tags:
+        count = con.execute(
+            "SELECT count(*) FROM item_tags it "
+            "JOIN tags t ON it.tag_id = t.id WHERE t.name = ?",
+            (tag,),
+        ).fetchone()[0]
+        status = "OK" if count > 0 else "MISSING"
+        print(f"  {status:7s}  {tag:30s}: {count:>5}")
+
+    # Collection item counts
+    print("\n  Collection item counts:")
+    for name, key in sorted(col_map.items()):
+        count = con.execute(
+            "SELECT count(*) FROM collection_items ci "
+            "JOIN collections c ON ci.collection_id = c.id "
+            "WHERE c.key = ?",
+            (key,),
+        ).fetchone()[0]
+        if count > 0:
+            print(f"    {name:30s}: {count:>5}")
+
+    # Total
+    total = con.execute(
+        """SELECT count(DISTINCT i.id) FROM items i
+           JOIN item_tags it ON i.id = it.item_id
+           JOIN tags t ON it.tag_id = t.id
+           WHERE t.name = 'module:skin-subs'"""
+    ).fetchone()[0]
+    print(f"\n  Total module:skin-subs items: {total}")
+
+    store.close()
+    print("\nDone.")
+
+
+if __name__ == "__main__":
+    main()