The World Health Organization (WHO) (World Health Organization, 2023) defines pharmacovigilance as science and activities related to the detection, evaluation, understanding, and prevention of adverse events related to the use of medicines (adverse drug events, ADEs), such as adverse drug reactions (ADRs). Pharmacovigilance plays a fundamental role in ensuring the safety and effectiveness of medicines after they have been marketed (World Health Organization, 2009).

According to the WHO, an ADR is defined as any harmful effect caused by a drug when used in doses typically administered to humans for the prevention, diagnosis, or treatment of a disease or physiological alteration (Gupta et al., 2015). To effectively manage these reactions, it is crucial to apply defined and structured methods (Piccirillo and Parish, 2022).

Pharmacovigilance uses various methods to identify, evaluate, and understand the safety of medicines. These methods are generally divided into two categories: passive surveillance, which primarily relies on spontaneous reporting, and active surveillance, which involves a systematic search for ADRs using retrospective triggers in medical records (Chatterjee and Aparasu, 2023). One widely used active surveillance method is the Global Trigger Tool (GTT), which is based on an adaptation of the medical record review methodology developed by Harvard University (Leape et al., 1991).

Although spontaneous reporting is the most widely used method worldwide, studies have shown that the ADR notification rates are low, ranging from 9% to 15% (Lazarou et al., 1998; Chan et al., 2016; Impicciatore et al., 2001; Prashanthi et al., 2019). Several factors contribute to underreporting in spontaneous notification systems, such as overburdened health professionals and a lack of appreciation for this type of reporting (García-Abeijon et al., 2023). A systematic review by Varallo et al. (2014) identified ignorance, insecurity, and indifference towards the patients as the main causes of underreporting by health professionals.

Critically ill patients in intensive care units (ICUs) are more susceptible to ADRs owing to various factors, reflecting the clinical complexities and fragility of these patients. These patients are often subjected to polypharmacy and non-physiological conditions, generating a state of stress that can alter the body’s response to drugs (Kane-Gill and Devlin, 2006). In ICUs, ADRs are of extreme importance because of not only clinical factors but also their impact on hospitalization duration. A previous study indicated that ICU stay is prolonged in patients with ADRs, which leads to an increase in hospitalization costs (Sendekie et al., 2023).

Therefore, seeking new reporting strategies and other pharmacovigilance actions is essential to ensure the safe use of medicines and the quality of care provided to these patients (Trifirò and Crisafulli, 2022). The primary objective of this study was to use an adapted GTT to determine the prevalence of ADRs in critically ill patients admitted to two adult ICUs: one general ICU and the other dedicated to COVID-19 patients at a general university hospital in the state of São Paulo, Brazil. The secondary aims were to characterize patients in both ICUs; identify factors associated with the occurrence of ADRs; assess the performance of triggers in detecting ADRs; describe ADRs in terms of severity, mechanism, causality, and suspected drugs; and compare the trigger tool methodology with spontaneous reporting.

Regarding the demographic characteristics of ICU patients, there was a slight predominance of males. A study published in 2011 (Vezzani et al., 2011) showed a male predominance among the total number of inpatients, with males comprising 64.0% of all patients admitted to adult ICU. Similarly, Marques et al. (Marques et al., 2020) showed that 55.8% of critically ill patients were male. Concerning the average age, one study analyzing 156 critically ill patients reported an average age of 54.9 years (Brito and Guirardello, 2012). Another study of critically ill patients found an average age of 51.2 years, ranging from 15 to 85 years (Eulálio et al., 2016). Regarding the average length of stay for critically ill patients, a study conducted in Rio de Janeiro in 2011 (Roque et al., 2016) reported an average of 8.9 days. However, our results showed a slightly higher average age and length of hospitalization than those in the aforementioned studies.

Several studies have used the SOFA score as a predictor of mortality in ICUs (Lambden et al., 2019; Raith et al., 2017; Polok et al., 2023). The SOFA score was initially developed to describe the morbidity of populations in various hospitalization environments such as ICUs. Today, after several validations, it is used as a determining factor in the diagnosis of sepsis (Lambden et al., 2019). Small variations in SOFA scores have been associated with changes in patient mortality rates. For instance, an increase in the SOFA score upon ICU admission is related to an increased mortality rate among these patients (Raith et al., 2017; Polok et al., 2023). In this study, critically ill patients had a mean SOFA score of approximately eight points throughout hospitalization, with a mean maximum SOFA score of approximately 12 points, and a mortality rate of approximately 70%. This aligns with the results of previous studies, which indicated that a higher SOFA score correlates with increased patient mortality. It is also consistent with the study by Anami et al. (2010), in which patients with a mean SOFA score of 6.01–7 had a mortality rate exceeding 50%, while those with a mean SOFA score above 10 had a 98.6% mortality rate.

In this study, less than half of the patients (40.7%) experienced at least one ADR. The proportion of ADRs found in medical records varies significantly across studies owing to various factors, including the methodologies and tools used and the number of researchers involved. A 2013 study conducted in a tertiary hospital with 128 patients reported that 15.6% of patients experienced at least one ADR during hospitalization (Rozenfeld et al., 2013), while other studies have indicated an ADR prevalence ranging from 2% to 24% (Franklin et al., 2010; Classen et al., 1991; Nilsson et al., 2012).

This study showed an association between length of hospitalization, age, and occurrence of ADRs. Aging and the concomitant use of multiple medications by elderly patients increase the likelihood of ADR occurrence (Rodrigues and de Oliveira, 2016). Furthermore, studies have indicated an association between the occurrence of ADEs, such as ADRs, and longer hospitalization (Roque and Melo, 2011; Classen et al., 1997). This study found that most patients in the G-ICU and C19-ICU experienced ADRs between the second and third quartiles of their hospitalization time; most ADRs were detected approximately halfway through the hospitalization period for each patient. Although no other studies have reported this association, this finding is important for directing pharmacovigilance actions and intensifying monitoring during periods of hospitalization that are most likely to result in ADR.

Patients admitted to the C19-ICU were found to be 2.4 times more likely to develop an ADR. Several factors may explain this, including the severity of COVID-19, which can lead to multiorgan dysfunction, such as in the liver and kidneys. These organs are crucial for drug metabolism and excretion, and their dysfunction can result in drug accumulation in the body, thereby increasing the likelihood of ADRs (Melo et al., 2021; Gérard et al., 2020).

Among the triggers used in this study, those with the highest PPV were PTT >50, skin rash, protamine, and hydroxyzine, all at 100%. This implies that each time these triggers appear, they identify an ADR. However, this result should not be relied upon solely because of the PPV, as protamine and hydroxyzine only appeared once in the medical records, and each identified a single ADR. Therefore, caution is advised when claiming that triggers with 100% PPV are the best for detecting ADRs. For example, blood glucose <50 mg/dL, despite having a PPV of 26.4%, appeared more frequently in medical records and was associated with a greater number of ADRs. Similarly, the “abrupt medication stop” trigger, with a PPV of 44.8% and a frequency of 49.6/100 medical records, identified 22.2 ADRs/100 medical records. Notably, more frequent triggers do not always identify ADR and are not necessarily associated with higher PPV. For instance, sodium <135 mEq/L (105.9/100 records) and potassium >5.5 mmol/L (96.3/100 records) have PPV of 0.7 and 0.8, respectively.

The results indicated that, in both ICUs, most ADRs were moderate, Type A, and classified as possible. These findings differ from those of previously published studies due to methodological differences. For instance, a 2020 study on critical patients in an emergency department reported that, according to the WHO classification, the majority of ADRs were classified as probable (Pandya et al., 2020). However, the results were similar regarding the mechanism, with most patients presenting Type A ADRs according to the Rawlins and Thompson classification, and regarding severity, with most patients presenting moderate ADRs according to the modified Hartwig and Seigel severity scale (Pandya et al., 2020). Another study, involving patients admitted to the ICU, found that most ADRs were mild according to a severity algorithm score, and the causality was classified as probable according to the WHO classification (Joshua et al., 2009).

Regular insulin, morphine, heparin, and enoxaparin were the drugs most frequently suspected to cause ADRs. The drug classes most frequently associated with ADRs in ICU environments include analgesics, sedatives, insulins, and anticoagulants (Esfahani et al., 2016). Hypoglycemia resulting from insulin use was the most commonly identified ADR in the ICUs studied. Cases of glycemic variation in critically ill patients are well documented in the literature (Kwan et al., 2020; Wang et al., 2021; Kwan et al., 2022). Hermanides et al. (Hermanides et al., 2010) reported that among the 5,961 patients evaluated, 4.6% experienced at least one episode of hypoglycemia, with insulin being the primary drug suspected to cause this ADR. Furthermore, hypoglycemia associated with insulin use has been linked to increased mortality in these patients. Hypoglycemia is the primary ADR associated with the continuous infusion of regular insulin owing to various factors such as excess insulin administration, hormonal deficiencies, concomitant medications, non-physiological conditions such as intubation, and variations in nutritional support (Almagthali et al., 2024).

Bleeding was also a frequent ADR observed in this study. It is common in ICUs, resulting from the use of anticoagulants and antiplatelet drugs, as well as invasive procedures such as tracheostomy and intubation, as reported in previous studies (Jonsson et al., 2021; Hariri et al., 2024). Additionally, cases of excessive drowsiness and delirium are notable, likely because of the concomitant use of central nervous system (CNS) depressants commonly used in ICUs (Wong et al., 2020). Opioids administered via infusion pumps for analgesia and sedation, combined with other CNS depressants, can lead to delirium and reduced consciousness associated with hypotension (Wu et al., 2023).

Finally, the use of triggers is essential for detecting ADRs that are not reported through spontaneous reporting. This technique has been proven to be highly effective in identifying ADRs, as it allows for the detection of a significantly higher number of ADRs and ADEs than spontaneous reporting methods (Classen et al., 2011). However, implementing this method can be challenging for institutions because it requires trained professionals with dedicated time for pharmacovigilance activities (Martins et al., 2018). The underreporting rate identified in this study was 100.0%, which aligns with the rate reported in a review of 37 studies by Hazell et al. (Hazell and Shakir, 2006). The high workload of the staff during the pandemic, as they prioritized patient care to ensure survival, was the primary factor contributing to this underreporting.

This study has several limitations. Evaluating physical medical records is time consuming, especially when assessing the presence of triggers. Laboratory test requests are generally described only in online medical records, whereas results are often found only in physical records. The retrospective nature of the study also posed a limitation, as it was not possible to discuss the 2020 cases with the G-ICU and C19-ICU staff owing to various factors, such as the high workload of health professionals. Consequently, the analysis relied solely on descriptions in medical records and nursing notes, which were sometimes handwritten. Another limitation is that only one researcher assessed the medical records, as no additional staff at the HES were available owing to the high workload, which may have affected the evaluation results of both physical and electronic records. Additionally, the influence of baseline clinical conditions on hospitalization duration was not analyzed, which is a critical omission, as these conditions may confound the interpretation of ADR prevalence. Finally, the possibility that some of the events observed in this study were medication errors and not ADRs cannot be ruled out.

The patients evaluated in this study had a mean age of approximately 61 years, were mostly male, spent an average of 13 days in hospital, and had a high SOFA score, which could be directly correlated with the high mortality rate observed. This study also revealed a high prevalence of ADRs in ICU patients. ADRs were associated with age and length of hospitalization. Additionally, patients admitted to the C19-ICU were 2.4 times more likely to develop an ADR.

Triggers such as “PTT >50” and “skin rash” proved to be effective in detecting ADRs, with a PPV of 100%. Most ADRs were moderate, Type A, and had possible causality. The most common ADRs were hypoglycemia and bleeding, which are often associated with the use of regular insulin and antithrombotic agents.

ADRs were identified more frequently using the adapted GTT, highlighting a significant under-reporting rate within the institution. Underreporting represents a major challenge in detecting ADRs, indicating the need to implement more effective active surveillance methods, such as trigger tools, in institutions. In conclusion, the use of triggers is essential for improving the detection of ADRs, significantly contributing to patient safety and the effectiveness of pharmacovigilance actions.