Skip to main content

​
Overview

VIGIA integrates with DIGEMID (Dirección General de Medicamentos, Insumos y Drogas), Peru’s national regulatory authority, to scrape safety alerts, perform product lookups, and generate compliant RAM (Reacción Adversa a Medicamentos) export reports.

​
Key Capabilities

Alert Scraping

Automatically retrieves safety alerts and modifications from DIGEMID’s public portal

Product Lookup

Searches product registry by name, sanitario number, or active ingredient

RAM Export

Generates PDF reports in official DIGEMID format (Modelo A & B)

Data Enrichment

Optional AI-powered field enhancement using Gemini

​
Alert Scraping

The DIGEMID scraper extracts safety alerts from the official website and parses structured data including alert numbers, affected products, and reasons for action.

​
Data Structure

Each scraped alert contains: 
{
  "n_alerta": "ALERTA DIGEMID Nº 015-2024",
  "producto": "Atorvastatina 40mg; Losartán 50mg",
  "motivo": "Detección de impurezas fuera de especificación",
  "enlace": "https://www.digemid.minsa.gob.pe/...",
  "fecha_publicada": datetime(2024, 8, 5, tzinfo=UTC)
}

​
Implementation

The scraper handles multiple layouts: 
def scrape_digemid_alerts(index_url: str, max_posts: int = 100) -> List[Dict]:
    """
    Scrapes DIGEMID alerts from index page.
    
    Returns:
        List of dicts with: n_alerta, producto, motivo, enlace, fecha_publicada
    """
    html = _get_html(index_url)
    soup = _soup(html)
    
    # Try table format first
    rows = _parse_table_index(soup, index_url)
    if rows:
        return rows
    
    # Otherwise, collect post cards with dates
    post_pairs = _collect_posts_with_dates_from_index(soup, index_url)
    
    out: List[Dict] = []
    for link, fecha_from_card in post_pairs[:max_posts]:
        try:
            detail_html = _get_html(link)
            data = _parse_detail_page(detail_html, link)
            
            if not data.get("fecha_publicada"):
                data["fecha_publicada"] = fecha_from_card
            
            out.append(data)
        except Exception:
            continue
    
    return out

​
Date Parsing

DIGEMID uses Peruvian timezone (America/Lima) with Spanish month names: 
TZ_PE = ZoneInfo("America/Lima")
MESES = {
    "ene": 1, "feb": 2, "mar": 3, "abr": 4, "may": 5, "jun": 6,
    "jul": 7, "ago": 8, "set": 9, "sep": 9, "oct": 10, "nov": 11, "dic": 12
}

def _parse_fecha_publicada_from_card(card: BeautifulSoup) -> Optional[datetime]:
    text = " ".join(card.get_text(" ").split())
    m = re.search(r"\b(\d{1,2})\s*(Ene|Feb|Mar|Abr|May|Jun|Jul|Ago|Set|Sep|Oct|Nov|Dic)\b", text, re.I)
    if not m:
        return None
    day = int(m.group(1))
    mon = MESES[m.group(2).lower()[:3]]
    year = _parse_year_from_context(card, text) or datetime.now(TZ_PE).year
    local_dt = datetime(year, mon, day, 12, 0, 0, tzinfo=TZ_PE)
    return local_dt.astimezone(timezone.utc)

​
RAM Export (ICSR Reports)

VIGIA generates official DIGEMID-compliant PDF reports in two formats:

​
Modelo A: Evaluation Report (FRT-OP-017)

Internal causality assessment using the Karch-Lasagna algorithm. 
from app.services.icsr_digemid_export import generate_model_a_evaluacion_pdf

# Generate evaluation report
pdf_path = generate_model_a_evaluacion_pdf(
    db=db,
    icsr_id=12345,
    out_path="/storage/icsr_exports/modelo_a_evaluacion_12345.pdf"
)
Sections included:
  • Client/Institution reporting
  • Product description (presentation, lot, sanitary registration)
  • Communication description (narrative)
  • Causality analysis (Karch-Lasagna criteria)
  • Conclusions and recommendations
  • Signatures (analyst and operations manager)

​
Modelo B: Healthcare Professional Format

Official format for regulatory submission based on DOCX template. 
from app.services.icsr_digemid_export import generate_model_b_prof_salud_pdf

# Generate professional health format
pdf_path = generate_model_b_prof_salud_pdf(
    db=db,
    icsr_id=12345,
    out_path="/storage/icsr_exports/modelo_b_prof_salud_12345.pdf",
    extra_ctx={
        "firma_nombre": "Dr. Juan Pérez",
        "firma_cargo": "Pharmacovigilance Manager",
        "firma_img_path": "/path/to/signature.png",
    }
)
Template Processing:
  1. Reads DOCX template from backend/app/templates/docs/Formato Profesionales Salud.docx
  2. Renders with docxtpl using Jinja2 context
  3. Converts to PDF using docx2pdf (Windows/Mac) or LibreOffice (Linux)
  4. Saves filled DOCX copy to storage/icsr_exports/ICSR_{id}_ProfSalud_Relleno.docx

​
Context Builder

Both models use a shared context builder: 
from app.services.icsr_export_ctx import build_ctx_for_digemid

ctx = build_ctx_for_digemid(db, icsr_id)
# Returns dict with:
# - paciente_iniciales, edad, sexo
# - producto_sospechoso, presentacion, lote
# - evento_adverso, fecha_inicio
# - karch_criterios (temporalidad, conocimiento_previo, etc.)
# - clasificacion_qf/clasificacion_sys

​
Configuration

​
Environment Variables

# Template path (optional, defaults to backend/app/templates/docs/)
DIGEMID_TEMPLATE=/path/to/Formato\ Profesionales\ Salud.docx

# Export storage directory
ICSREXPORT_DIR=/path/to/storage/icsr_exports

# AI enhancement (optional)
GEMINI_API_KEY=your_gemini_key_here

​
DOCX to PDF Conversion

The system attempts conversion in this order:
  1. docx2pdf (requires Microsoft Word on Windows/Mac)
    • Windows: Initializes COM with pythoncom.CoInitialize()
    • Mac: Uses Word automation
  2. LibreOffice headless (cross-platform fallback) 
    soffice --headless --convert-to pdf --outdir /output /input.docx
If both fail, the system returns the filled DOCX file.

​
API Endpoints

​
Search Alerts

GET /api/v1/sources/digemid?q={query}&start={date}&end={date}&max_pages={int}
Response: 
{
  "source": "digemid",
  "ok": true,
  "query": "atorvastatina",
  "period": {
    "from": "2024-01-01",
    "to": "2024-12-31"
  },
  "hits": 3,
  "items": [
    {
      "n_alerta": "ALERTA DIGEMID Nº 015-2024",
      "producto": "Atorvastatina 40mg",
      "motivo": "Impurezas detectadas",
      "enlace": "https://...",
      "fecha_publicada": "2024-08-05T16:00:00Z"
    }
  ],
  "traces": []
}

​
Export ICSR to DIGEMID Format

POST /api/v1/icsr/{id}/export/digemid
Content-Type: application/json

{
  "modelo": "B",
  "firma_nombre": "Dr. Juan Pérez",
  "firma_cargo": "Pharmacovigilance Manager"
}
Response: PDF binary with Content-Disposition: attachment; filename="modelo_b_prof_salud_12345.pdf"

​
Error Handling

DIGEMID’s website structure changes occasionally. The scraper includes multiple fallback strategies:
  • Table-based index parsing
  • Card-based post extraction
  • PDF text extraction for missing fields
  • AI-powered field enrichment

FDA Integration

US FDA FAERS database access

EMA Integration

European safety alerts

VigiAccess

WHO global database

Build docs developers (and LLMs) love

Get started for freeTalk to us