Riscuri pentru asistenții medicali implicați în administrarea medicamentelor periculoase. Rolul programelor informatice și al inteligenței artificiale

Introduction

This review presents some issues regarding safety for nurses, particularly related to the toxicity of chemotherapeutic agents, immunotherapy and radiotherapy with radionuclides. Common concrete challenges that drive lapses include: (1) inconsistent training/competency assessment and under-availability of oncology-trained staff; (2) workload/time pressure and interruptions during administration; (1) gaps in adherence to safe administration steps and to disposal practices (including cytotoxic waste and body fluids), even when self-reported compliance is “moderate-good”; and resource constraints (personal protective equipment supply, closed-system transfer devices, spill kits and designated administration areas). Real-world studies in oncology settings consistently find “critical gaps” particularly during administration and disposal steps, and better safety performance in nurses with more experience and prior training⁽²⁾. Broader nursing-reported barriers in cancer care commonly include adverse treatment effects that require intensive symptom management, non-availability of drugs and staffing/education limitations – all of which indirectly increase risk during cytotoxic administration by adding delay, complexity and stress to the workflow⁽³⁾.

Where AI can plausibly reduce harm

Artificial intelligence (AI) can reduce harm in the following domains:

• Order/regimen verification support.
• Closed-loop medication administration augmentation – pairing AI logic with barcode medication administration, smart-pump libraries etc.
• Early toxicity signal detection – AI-assisted surveillance across labs/vitals/symptoms to flag patterns consistent with emerging neutropenia-related deterioration, dehydration/AKI, immune-related adverse events, or sepsis.
• Training and competency maintenance – simulation-based training.
• Occupational safety reinforcement – computer-vision or workflow tools to detect missing personal protective equipment (PPE) steps, prompt spill-response protocols.

Systemic therapy

Systemic cancer therapy includes chemotherapy, hormone therapy, targeted therapy and immune therapy. Cytotoxic chemotherapy agents are classified as⁽¹⁾:

• Alkylating agents (including nitrosoureas), which damage DNA (e.g., busulfan, chlorambucil, cisplatin).
• Antimetabolites, which block DNA (e.g., azacitidine, fluorouracil, methotrexate).
• Antitumor antibiotics (including anthracyclines), which interfere with DNA (e.g., bleomycin, daunorubicin, doxorubicin).
• Mitotic inhibitors, which disrupt cell division (e.g., docetaxel, paclitaxel, vincristine).
• Topoisomerase inhibitors, which stop the activity of enzymes in cancer cells (e.g., etoposide, irinotecan).
• Other chemotherapy agents that do not fit into specific categories (e.g., asparaginase). Generally, cytostatics act in certain phases of the cell cycle.

How can the faulty handling of cytostatics and immunotherapy affect nurses?

The improper handling of cytostatics (“classic” antineoplastics) is associated with relevant occupational exposure in nurses (and pharmacists/physicians), mainly through aerosols/particles, splashes and contamination of surfaces with dermal transfer.

For immunotherapy, the risk depends largely on the type of product (monoclonal antibodies versus antibody-drug conjugates versus cell therapies), but there are still occupational risks (especially through accidental contact and hypersensitivity reactions), and for some products the risk is comparable to that of cytostatics.

1. How does exposure occur (common mechanisms)

In the absence of correct measures (PPE, hood/isolator, CSTD, decontamination procedures):

• Reconstitution/dilution – aerosols when introducing the solvent, expelling air from the syringe, “spray” when disconnecting lines, positive pressure in vials.
• Administration – leaks at connections, extravasation, unprotected disconnection, handling of contaminated bags/syringes.
• Waste and excretions – contact with urine/feces/vomit of treated patients (within the disposal window), linen/urine containers, stoma bags.
• Surface contamination – countertops, pumps, handles, keyboards, trays, carts; transfer to skin/clothing.
• Accidents – punctures, eye splashes, uncontrolled spills.

2. Health effects of the nurse – cytostatics (major hazard)

Cytostatics frequently include mutagenic, genotoxic, teratogenic and/or carcinogenic substances (depending on class). Effects reported/feared by chronic exposure to low doses include:

A) Acute effects (after accidental or repeated exposures)

• skin and eye irritation, contact dermatitis; sometimes, ulcerations upon direct contact;
• headache, dizziness, nausea, airway irritation (especially in aerosol/spill events);
• allergic reactions/sensitization to some substances.

B) Hematological and organ-toxic effects (in repeated exposures)

• signs of bone marrow suppression (especially theoretically and in reports/series: leukopenia/formula alterations), depending on the type of agent and the level of exposure;
• potential hepatic/renal effects (less often documented occupationally, but biologically plausible for certain molecules).

C) Reproduction and pregnancy (most important practical concern)

• association in occupational literature with spontaneous abortion, subfertility, premature birth and congenital malformations (risk varies; difficult to quantify precisely, but consistent enough that guidelines treat exposure as “high concern”);
• therefore, many institutional policies recommend avoiding direct handling of pregnant/conceiving personnel, especially during reconstitution and spill management.

D) Carcinogenesis/genotoxicity

• some cytostatics are recognized as carcinogens; occupational exposure is associated with biomarkers of DNA damage (in biological monitoring studies) and concerns about increased long-term risk.

3. Immunotherapy – the occupational risk depends on the product

The term is heterogeneous. It is useful to separate:

A) “Naked” monoclonal antibodies (e.g., anti-PD-1/PD-L1, anti-CTLA-4)

They are generally not genotoxic like classical cytostatics. However, there may be:

• hypersensitivity reactions (skin/mucous contact, eye splash) and sensitization;
• theoretical risk of immunological effect upon accidental parenteral exposure (e.g., needle stick) – data are limited; it is treated as significant exposure, and it is managed according to the biological/chemical accident protocol;
• risk during pregnancy – many mAbs (IgG) cross the placenta, especially in the second or third trimester; although occupational exposure is much lower than the therapeutic dose, safety policies treat it cautiously.

B) Antibody-drug conjugates (ADC) and immunotoxins

They actually contain a cytotoxic “payload” (e.g., microtubule inhibitors, calicheamicin etc.) occupational profile similar to cytostatics or even more problematic in case of spill. Here “mishandling” can generate the same risks: genotoxicity, teratogenicity, irritation etc.

C) Cell therapies (e.g., CAR-T)/live biological products

The main risk is biohazard (biological material) and requires specific procedures; it is not the same type of hazard as with cytostatics, but it still requires strict exposure control.

4. Practical indicators of handling problems (red flags)

Repeated spills, odor/irritation in preparation area, leaky connections.

• Lack or incorrect use of cytostatic-certified gloves, impervious gowns, eye protection.
• Preparation outside the hood/isolator, disconnections without safety devices.
• Cleaning “with alcohol” as the only measure (does not adequately decontaminate many antineoplastics).
• Absence of surface contamination monitoring programs.

5. What measures reduce the risk (essential for hospital/oncology practice)

Engineering controls: vertical laminar flow hood/biological safety cabinet or isolator + adequate ventilation; CSTD for transfer and administration.

Correct PPE: two pairs of cytostatic gloves, impervious gown, eye protection/visor for splash risk, mask/respirator according to local policy when aerosols are present.

Procedures: spill kit, extravasation protocols, safe disconnection, dedicated containers, excretion management (with defined intervals), drug-specific decontamination according to procedures.

Training and auditing: initial training + refresher, documented competencies, compliance audit.

Pregnancy/breastfeeding and conception policies: re-assignment of tasks with direct exposure, especially during preparation/spill.

Post-event medical reporting and assessment: eye splash, sting, major spill, acute symptoms^(2,5,6).

First step in the introduction of AI in nurse practice

AI and machine learning

Artificial intelligence (AI) and machine learning (ML) are often used interchangeably, but they are distinct concepts within the field of computer science. Understanding their differences and connections is crucial for leveraging their potential in various applications.

Artificial intelligence

AI is a broad field that encompasses the development of systems capable of performing tasks that typically require human intelligence. These tasks include understanding natural language, recognizing patterns, making decisions and solving problems. AI systems can work with various types of data, including structured, semi-structured and unstructured data.

Artificial intelligence aims to create intelligent systems that can perform complex tasks autonomously. It uses technologies such as logic, decision trees, and self-correcting algorithms to mimic human decision-making processes.

Machine learning

ML is a subset of AI that focuses on enabling machines to learn from data and improve their performance over time without explicit programming. Machine learning algorithms analyze large datasets to identify patterns and make predictions or decisions based on the insights gained.

Machine learning models are trained using data, and their performance improves as they are exposed to more data. The goal of ML is to build systems that can learn from data to increase the accuracy of their outputs.

Key differences

While AI encompasses the broader concept of creating intelligent systems, machine learning specifically deals with the ability of machines to learn from data. AI systems can perform a wide range of tasks, whereas ML systems are designed to perform specific tasks and improve their accuracy over time.

For example, an AI system might be designed to simulate human intelligence and solve complex problems, while an ML system might be trained to analyze traffic data and predict traffic flow⁽⁸⁾.

A chatbot (often known as a smartbot, talkbot, chatterbot, IM bot, conversational interface, or conversational artificial entity) is a program or form of artificial intelligence that can carry out various conversations, as well as text^(1,2). Such programs are often designed to realistically simulate the behavior of a conversation partner, passing the Turing test^(9,10).

Preparing nurses for the age of AI requires both formal education and ongoing professional development. Despite the growing presence of AI in healthcare, many nursing programs still lack dedicated content on this topic. While 59.9% of nursing students were aware of AI in healthcare in one study, they reported a lack of available educational resources (35.8%) and mentorship (41.4%)^(11-13). Graduates often enter the workforce without the foundational knowledge needed to engage confidently with AI technologies; however, AI has provided powerful opportunity to transform nursing education. Integrating AI into undergraduate curricula, covering how algorithms function, what data they rely on, and how to interpret their outputs are essential for the future nursing education and practice. For example, generative AI tools like ChatGPT are being used by students to brainstorm research questions and refine literature reviews⁽¹⁵⁾. While these tools offer new opportunities for learning, they also raise concerns about academic integrity and the accuracy of generated content⁽¹⁶⁾. To address these challenges, Gosak et al.⁽¹⁴⁾ advocate for careful consideration on the overreliance and ethical AI use across all levels of nursing education⁽¹¹⁾.

How could a computer program help prepare infusion solutions, both in terms of shortening working time and reducing personnel contact with toxic substances?

A computer program integrated into hospital/oncology pharmacy flow can significantly reduce preparation time for infusion solutions and minimize staff contact with toxic substances, especially cytostatics. Moreover, automation enhances safety and efficiency – it reduces medication errors and improves traceability, optimizing inventory management and workload for healthcare professionals.

Altai Oncology Suite: this system streamlines chemotherapy management with smart automation and built-in safeguards, ensuring accuracy and safety throughout the treatment process.

Robotic systems for intravenous compounding: these systems detect potential medication errors, enhancing quality and safety in chemotherapy compounding.

Intravenous room technology-assisted workflow: this system standardizes intravenous admixture operations while incorporating patient safety measures, improving workflow efficiency^(15,16).

Overall, these advancements contribute to safer and more efficient healthcare practices in oncology.

How could a computer program protect nurses who prepare infusion solutions?

A computer program could protect nurses preparing infusion solutions by implementing smart infusion pumps and data-based program management. These technologies offer automated infusion programming and documentation, which can improve patient safety and workflow efficiency. They also provide bidirectional interoperability, allowing for continuous monitoring of intravenous smart pump compliance, which can help reduce programming errors and medication administration errors. Additionally, computer programs can enhance patient and clinician safety, streamline clinical workflows and provide near real-time infusion data⁽¹⁷⁾.

The application of intelligent electronic systems in the practice of intravenous infusions is also realized through intelligent electronic pumps with interoperability.

Bidirectional intravenous smart pump interoperability for i.v. infusion

Bidirectional intravenous smart pump interoperability refers to the integration of smart infusion pumps with electronic health records (EHR) systems, enabling two-way communication between the two. This interoperability allows for automated infusion programming and documentation, which can significantly improve patient safety and workflow efficiency. It involves real-time, continuous communication between the smart pump and the EHR, allowing the pump to automatically program medications and the EHR to automatically document infusion information. This bidirectional integration addresses the gap between programmed pump rates and actual medication orders, reducing the risk of medication administration errors⁽⁷⁾.

Conclusions

A computer program (integrated into the hospital/oncology pharmacy flow) can reduce both the preparation time of infusion solutions and the contact of staff with toxic substances (especially cytostatics) through standardization and automation of calculation.

The therapeutic scheme chosen by the doctor, perhaps in consultation with the AI, will be burdened by the need to comply with the therapeutic protocols of the insurance company. Therefore, it may be good that the software for the electronic cytostatic infusion system be developed in collaboration with the specialists of the insurance company who provide the necessary funds for the procurement of medication.

Real-world adoption of generative artificial intelligence is often already occurring without formal oversight (“shadow AI”), and clinicians cite concerns about erroneous outputs (“hallucinations”) and confidentiality as major barriers – both directly relevant to medication safety and hazardous-drug handling documentation⁽¹⁸⁾.

We must take into consideration that even nongenerative digital tools in oncology frequently run into alert burden, workflow misalignment and interoperability problems, which can negate safety gains if not engineered carefully⁽¹⁹⁾. No less important is the development of an educational program for the nurses who will work with these operational systems.

Corresponding author: Alexandru-Călin Grigorescu E-mail: alexgrigorescu2004@yahoo.com

Conflict of interests: none declared.

Financial support: none declared.

This work is permanently accessible online free of charge and published under the CC-BY.