AI in the Health Industry: Investigative Analysis and Ethical Considerations

AI in the Health Industry: Investigative Analysis and Ethical Considerations

Author: R. Khouja | Critical Observation for Digital Transformation

Executive Summary

The digital transformation of Europe’s healthcare sector through Artificial Intelligence (AI) promises enhanced diagnostics, predictive analytics, and operational efficiency. However, this evolution must be scrutinized for ethical, legal, and financial compliance—especially in the context of public procurement. This report presents a strategic and investigative framework to detect potential misuse of AI projects and public funds in the health domain.

1. Strategic and Legal Overview

1.1 Objectives

- Identify irregularities in AI-related tenders.

- Ensure alignment between innovation, legality, and public interest.

1.2 Key Questions

- Are AI projects clinically justified or technologically hyped?

- Is public procurement being used ethically and effectively?

2. Investigative Strategy

Step 1: Taxonomy – Structured Filtering

Goal: Use formal classification systems for filtering and analyzing tenders.

CPV Codes:

- 72000000: IT services

- 73100000: R&D services

- 85100000: Health services

Ontology Tree:

Digitalization → Automation → AI

Subdomains: NLP, Computer Vision, Predictive Models

Normalization: Map "AI" to synonyms like "machine learning," "intelligent systems," etc.

Temporal Tags: Label data by year to track trends and funding cycles.

Step 2: Folksonomy – Pattern Discovery

Goal: Extract real-world language used in tenders.

- Use NLP tools to extract keywords and phrases.

- Cluster terms using BERTopic or DBSCAN.

- Create graphs linking folk terms to CPV codes.

- Flag semantic mismatches or buzzwords lacking technical grounding.

Step 3: Investigative Triggers – Legal Instruments

Key EU Frameworks:

- Article 325 TFEU: Protects EU financial interests.

- Regulation (EU) 2017/1939: Establishes EPPO.

- Directive 2014/24/EU: Governs procurement.

- Regulation 2020/2092: Rule-of-law conditionality.

- Directive 2019/1937: Whistleblower protection.

Red Flags to Investigate:

- Buzzwords without substance

- Copy-pasted tender specs

- Repeated winners

- CPV and deliverable mismatch

Reporting Channels:

- OLAF

- EPPO

- National audit institutions

Step 4: Data Science & Intelligence

Tools & Techniques:

- Preprocessing TED XML/JSON tender data

- Topic modeling via LDA / BERTopic

- Network analysis for authority-vendor patterns

- Anomaly detection in price or delivery times

- NER (Named Entity Recognition)

Starter Python Script

from docx import Document

import pandas as pd

import xml.etree.ElementTree as ET

import re

import networkx as nx

import nltk

from nltk.corpus import stopwords

from sklearn.feature_extraction.text import TfidfVectorizer

from sklearn.decomposition import LatentDirichletAllocation

nltk.download('stopwords')

def parse_ted_xml(file_path):

    tree = ET.parse(file_path)

    root = tree.getroot()

    tenders = []

    for notice in root.findall('.//FORM_SECTION'):

        title = notice.findtext('.//TITLE')

        description = notice.findtext('.//SHORT_DESCR')

        cpv_code = notice.findtext('.//CPV_MAIN_CODE')

        tenders.append({'title': title, 'description': description, 'cpv': cpv_code})

    return pd.DataFrame(tenders)

def preprocess_text(text):

    if not text:

        return ''

    text = re.sub(r'[^a-zA-Z0-9 ]', ' ', text)

    tokens = text.lower().split()

    tokens = [t for t in tokens if t not in stopwords.words('english')]

    return ' '.join(tokens)

def topic_modeling(df, n_topics=5):

    df['text'] = df['title'].fillna('') + ' ' + df['description'].fillna('')

    df['clean_text'] = df['text'].apply(preprocess_text)

    vectorizer = TfidfVectorizer(max_df=0.95, min_df=2, stop_words='english')

    tfidf = vectorizer.fit_transform(df['clean_text'])

    lda = LatentDirichletAllocation(n_components=n_topics, random_state=42)

    lda.fit(tfidf)

    words = vectorizer.get_feature_names_out()

    for i, topic in enumerate(lda.components_):

        print(f"Topic {i}:")

        print([words[i] for i in topic.argsort()[:-11:-1]])

3. Legal-Ethical Balance

- Technically sound

- Financially justified

- Clinically needed

- Legally compliant

Accountability and innovation are not opposites—they’re codependent.

4. Real Case Studies

- Romania (2023): >€2M in fraud for inflated smart health system costs (EPPO).

- Italy (2022): Tender rigging in digital health (OLAF).

- Spain (2021–2024): AI health projects flagged for underdelivery and high cost.

5. Conclusion

- Institutional vigilance

- Public interest protection

- Innovation integrity

Let’s ensure AI in health delivers real value, not just promises.

6. Presumption of Innocence and Ius Publicum Europaeum

This report does not constitute an accusation, either against individuals or institutions. It is a preventive analysis tool designed to enhance digital governance in the public sector.

Presumption of Innocence

Every entity involved in public procurement is entitled to the fundamental right to the presumption of innocence, as enshrined in Article 48 of the Charter of Fundamental Rights of the European Union.

Ius Publicum Europaeum

At the same time, Member States and their public administrations are bound to comply with the principles of European public law, particularly in regard to:

- Transparency

- Good administration

- Lawful public spending

- Prevention of fraud and corruption

Necessary Balance

Institutional vigilance, technological innovation, and the protection of the public interest are not mutually exclusive. In fact, trust in artificial intelligence in healthcare depends on maintaining that balance between legality, ethics, and innovation.

Glossary of Key Concepts and Terminology

Artificial Intelligence (AI)

A branch of computer science focused on creating systems capable of performing tasks that typically require human intelligence, such as learning, reasoning, and problem-solving.

Natural Language Processing (NLP)

A subfield of AI that enables computers to understand, interpret, and generate human language, facilitating interactions between humans and machines.

Computer Vision

An area of AI that trains computers to interpret and process visual information from the world, such as images and videos, emulating human sight.

Predictive Models

Statistical or machine learning models used to predict future outcomes based on historical data, commonly applied in healthcare for forecasting disease progression or treatment responses.

CPV Codes (Common Procurement Vocabulary)

A standardized classification system used in the European Union to describe the subject of public procurement contracts, facilitating transparency and efficiency.

Ontology

In information science, a structured framework that categorizes and defines the relationships between concepts within a particular domain, aiding in data organization and retrieval.

Folksonomy

A system of classification derived from the collective tagging and labeling of information by users, reflecting the way people categorize content in practice.

Named Entity Recognition (NER)

A process in NLP that identifies and classifies key elements in text into predefined categories such as names of people, organizations, locations, and more.

Latent Dirichlet Allocation (LDA)

A generative statistical model that allows sets of observations to be explained by unobserved groups, helping to discover abstract topics within a collection of documents.

BERTopic

An advanced topic modeling technique that leverages BERT embeddings and clustering algorithms to create easily interpretable topics from textual data.

DBSCAN

A clustering algorithm that groups together points that are closely packed together, marking as outliers the points that lie alone in low-density regions.

TED (Tenders Electronic Daily)

The online version of the 'Supplement to the Official Journal' of the EU, dedicated to European public procurement, providing access to business opportunities.

OLAF (European Anti-Fraud Office)

An EU body responsible for investigating fraud against the EU budget, corruption, and serious misconduct within European institutions.

EPPO (European Public Prosecutor’s Office)

An independent EU body with the authority to investigate and prosecute crimes affecting the EU's financial interests, such as fraud and corruption.

Article 325 TFEU

A provision in the Treaty on the Functioning of the European Union obligating member states to counter fraud and any other illegal activities affecting the financial interests of the EU.

Regulation (EU) 2017/1939

A regulation establishing the EPPO, enhancing the EU's capacity to combat crimes affecting its financial interests.

Directive 2014/24/EU

An EU directive that sets out the rules for public procurement by contracting authorities, aiming to ensure transparency and competition.

Regulation 2020/2092

A regulation on the general regime of conditionality for the protection of the EU budget, linking EU funding to the respect for the rule of law in member states.

Directive 2019/1937

An EU directive providing protection for persons who report breaches of Union law, commonly known as the Whistleblower Protection Directive.

Anomaly Detection

The identification of unusual patterns or outliers in data that do not conform to expected behavior, often used in fraud detection.

Network Analysis

A methodological approach to analyzing the relationships and structures within networks, useful for uncovering patterns in complex systems such as procurement processes.

Buzzwords

Trendy or fashionable words or phrases that are often used to impress or persuade, but may lack substantive meaning or clarity in context.

Tender Rigging

A form of collusion where competing parties agree on the outcome of a tender process, undermining fair competition and often leading to inflated prices.

Whistleblower

An individual who exposes information or activity within an organization that is deemed illegal, unethical, or not correct, often protected by specific laws.

Public Procurement

The process by which public authorities, such as government departments or local authorities, purchase work, goods, or services from companies.

Taxonomy

A hierarchical classification system used to organize concepts or items into categories and subcategories based on shared characteristics. In the context of AI and healthcare, taxonomy helps structure knowledge domains (e.g., medical diagnostics, legal frameworks, procurement systems), 

enabling clearer analysis and information retrieval.