Clinical predictors of medication error severity in the pediatric intensive care: the role of pharmaceutical interventions as complexity markers

Purpose: Medication errors (MEs) in Pediatric Intensive Care Units (PICUs) are driven by clinical complexity, yet identifying patients at highest risk for severe outcomes remains challenging. We investigated clinical predictors associated with ME severity, specifically evaluating whether the frequency of pharmaceutical interventions (PIs) functions as a marker for risk. Methods: This retrospective cohort study was conducted in a tertiary academic PICU in São Paulo, Brazil (June 2019–May 2021). The cohort included pediatric patients (29 days–19 years) with a length of stay (LOS) >24 hours and at least one recorded PI. MEs were stratified by severity using the NCC MERP index. Multivariate Poisson and multinomial regression models were employed to identify independent predictors of error severity. Results: Among 565 patients, the frequency of PIs emerged as the strongest independent predictor of error severity, surpassing traditional metrics. In the adjusted model, patients requiring >6 PIs exhibited a markedly higher relative risk for both moderate (Type C: RR=6.91, 95% CI 3.03–15.77) and severe (Type D/E: RR=16.53, 95% CI 8.85–30.89) errors compared to those with fewer interventions. While LOS was associated with severity in univariate analysis, this association lost statistical significance after adjusting for PI volume. Conclusions: The volume of PIs serves as a superior marker for clinical complexity and error severity compared to length of stay. These findings suggest that high PI frequency should be utilized not merely as a productivity metric, but as a proactive indicator for risk stratification and targeted safety surveillance in critically ill pediatric populations. Clinical predictors of medication error severity in the pediatric intensive care: the role of pharmaceutical interventions as complexity markers Predictors of medication errors severity in pediatric ICU Mônica Cristina Dutra Rodrigues, MSc – Rodrigues, MCD Msc. School of Pharmaceutical Sciences of Ribeirão Preto, University of São Paulo, Av. Bandeirantes, 3900, Monte Alegre, 14040-906, São Paulo, Ribeirão Preto, Brazil e-mail: [email protected] ORCID: https://orcid.org/0000-0002-1784-968X Lattes: http://lattes.cnpq.br/7202844137071015 Victor KaneKo Matsuno, MSc - Matsuno, VK MSc. School of Pharmaceutical Sciences, University of São Paulo, Av. Professor Lineu Prestes, 580, Cidade Universitária, 05508-000, São Paulo, São Paulo, Brazil e-mail: [email protected] ORCID: https://orcid.org/0000-0003-2079-4568 Lattes: http://lattes.cnpq.br/8695813261484377 Leonardo Régis Leira Pereira, PhD - Pereira, LRL PhD. School of Pharmaceutical Sciences, University of São Paulo, Av. Bandeirantes, 3900, Monte Alegre, 14040-906, São Paulo, Ribeirão Preto, Brazil e-mail: [email protected] ORCID: https://orcid.org/0000-0002-8609-1390 Lattes: http://lattes.cnpq.br/1728145468467999 CORRESPONDING AUTHOR Rodrigues, MCD; [email protected] FUNDING This work was supported by the Coordination for the Improvement of Higher Education Personnel (CAPES) for the financial support through a scholarship grant. Funding Code 001 (88887.718452/2022-00).

Purpose: Medication errors (MEs) in Pediatric Intensive Care Units (PICUs) are driven by clinical complexity, yet identifying patients at highest risk for severe outcomes remains challenging. We investigated clinical predictors associated with ME severity, specifically evaluating whether the frequency of pharmaceutical interventions (PIs) functions as a marker for risk. Methods: This retrospective cohort study was conducted in a tertiary academic PICU in São Paulo, Brazil (June 2019–May 2021). The cohort included pediatric patients (29 days–19 years) with a length of stay (LOS) by severity using the NCC MERP index. Multivariate Poisson and multinomial regression models were employed to identify independent predictors of error severity. Results: Among 565 patients, the frequency of PIs emerged as the strongest independent predictor of error severity, surpassing traditional metrics. In the adjusted model, patients requiring >6 PIs exhibited a markedly higher relative risk for both moderate (Type C: RR=6.91, 95% CI 3.03–15.77) and severe (Type D/E: RR=16.53, 95% CI 8.85–30.89) errors compared to those with fewer interventions. While LOS was associated with severity in univariate analysis, this association lost statistical significance after adjusting for PI volume. Conclusions: The volume of PIs serves as a superior marker for clinical complexity and error severity compared to length of stay. These findings suggest that high PI frequency should be utilized not merely as a productivity metric, but as a proactive indicator for risk stratification and targeted safety surveillance in critically ill pediatric populations.

Medication errors; pediatric intensive care units; pharmaceutical interventions; risk stratification; pharmacoepidemiology; patient safety. • Pharmaceutical intervention (PI) frequency is a robust independent predictor of medication error severity in the PICU, outperforming traditional clinical metrics. • High intervention density (>6 PIs per patient) is associated with a significantly increased risk for severe medication errors (RR=16.53), signaling extreme clinical vulnerability. • PI volume serves as a superior surrogate marker for clinical complexity compared to length of stay, which loses predictive significance when adjusted for intervention frequency. • Integrating pharmaceutical intervention monitoring into risk-stratification protocols allows for the early identification of high-risk patients requiring intensified safety surveillance. • Transitioning PI data from administrative productivity metrics to proactive clinical safety indicators enhances error prevention in high-complexity pediatric critical care settings. Plain language summary Medication errors in Pediatric Intensive Care Units (PICUs) represent a significant safety challenge due to patient complexity, making it difficult to identify those at highest risk for severe outcomes. This study investigated whether the frequency of pharmaceutical interventions (PIs)—actions taken by pharmacists to optimize therapy—serves as a reliable indicator for serious medication errors. By analyzing the medical records of 565 pediatric patients over a two-year period in Brazil, researchers used statistical models to compare PI volume against traditional risk factors like length of hospital stay. The results demonstrated that PI frequency was the most accurate predictor of error severity; specifically, patients requiring more than six interventions faced a markedly higher risk of severe errors. Crucially, length of stay lost its statistical significance when adjusted for intervention volume, suggesting that clinical complexity is a more critical risk factor than time spent in the hospital. These findings imply that high intervention frequency should be viewed as a clinical ”red flag” for patient safety rather than merely a productivity metric. By monitoring PI volume, healthcare teams can proactively identify vulnerable children and implement targeted surveillance to prevent life-threatening medication errors. Medication errors (MEs) pose a significant challenge in Pediatric Intensive Care Units (PICUs) due to high clinical complexity and unique pediatric physiological demands. 1,2 Clinical pharmacists function as vital safety barriers, utilizing pharmaceutical interventions (PIs) to intercept these risks. 3 Despite ME prevalence rates reaching 25% globally and up to 20% in Brazil, the relationship between PI frequency, acceptance, and error severity remains poorly understood. 3-7 Emerging evidence suggests a ”vigilance paradox,” where high intervention rates reflect extreme patient complexity and risk exposure rather than care failure. 8,9 However, few studies have systematically integrated clinical predictors with the prospective role of PIs in signaling clinical deterioration. 10,11 This study aimed to identify clinical and care-related predictors of ME occurrence and severity in a PICU cohort. We hypothesized that while a higher burden of accepted PIs is associated with reduced error severity, clinical factors such as surgery and prolonged length of stay (LOS) act as independent predictors of increased risk.

Study design A longitudinal, analytical cohort study was conducted with retrospectively collected data from June 1, 2019, to May 31, 2021, in the PICU of a high-complexity public university hospital in São Paulo, a national reference center for tertiary and quaternary pediatric care. Sample selection Although a census of 864 admissions occurred during the period, the effective study population comprised the 565 patients who experienced at least one PI. To determine the precision of the frequency estimates within this specific cohort, a post hoc margin of error ( d ) calculation was performed using the single population proportion formula with a 95% confidence level (z=1.96). d = Z α/2 · √p · (1-p) n Adopting a conservative proportion (p) of 0.50 (assuming maximum statistical variance) and the effective sample size (n) of 565: √0.50 · 0.50 d = 1.96 · ______________~ 0.0412 (4.12%) 565 The calculation yields a margin of error of ±4.12%. Therefore, the sample of 565 patients was considered statistically robust, falling well within the clinically acceptable margin of error (24 hours and ≥ PI triggered by a medication error. To mitigate retrospective detection bias, patients without recorded PIs were excluded, as the absence of a PI often reflects a lack of detection rather than error absence. Given the observed 1:1 PI-to-ME ratio, the cohort focused on severity predictors among confirmed events rather than general prevalence estimation. This approach ensures data reliability by restricting the analysis to established medication errors within an actively monitored population. Study setting The study was conducted in a 20-bed, high-complexity university PICU in São Paulo, a national reference center for tertiary and quaternary pediatric care. The unit manages a high-acuity population—primarily liver transplant recipients, who account for over half of all admissions—through an integrated multidisciplinary model. Within this framework, specialized clinical pharmacy services are a fundamental component of the unit’s strategy to prevent medication errors. Pharmaceutical services and routine Specialized pediatric pharmacists integrate with the multidisciplinary team via daily rounds and bedside discussions. Core responsibilities include prescription review, medication reconciliation, dose validation, and safety monitoring. Detected errors are resolved in real-time and documented in standardized electronic databases. While the service lacks 24-hour coverage, its structured integration ensures intervention traceability and supports continuous quality and safety improvements within the PICU. Variables and data collection Data were collected from electronic health records, including demographics, PIM score, surgery, LOS, and clinical outcomes. Although a major determinant of MEs, the number of medications was excluded due to its confounding relationship with LOS and PI frequency—a recognized limitation. Total PI count was utilized as a clinical complexity marker, while acceptance analysis was deferred to separate research. MEs were classified via the NCC MERP index and categorized into three analytical groups: intercepted errors (Types A/B, serving as the ”no error” reference), Type C, and Types D/E. 12 The cohort specifically targeted patients under active pharmaceutical surveillance; thus, all participants had ≥ recorded PI. A strict 1:1 correspondence between PIs and MEs was observed, meaning each intervention represented a detected error. This approach defines the study population by detected risk, focusing the analysis on independent predictors of ME severity rather than general PICU prevalence. Patient age classification Patients were stratified into three age groups—(29 days to 5 years and 11 months), childhood (6 to 9 years and 11 months), and adolescence (10 to 19 years and 11 months)—excluding neonates (less than 29 days of age). This classification was based on developmental stages and healthcare needs, according to the criteria of the World Health Organization (WHO, 2024) 13 . Pediatric Index of Mortality (PIM) Score The PIM 2 score, in its version adapted for the Brazilian context, classifies pediatric patients admitted to ICUs into severity categories based on the estimated probability of death. The stratification considers the ranges: 0–1% (low risk), >1–5%, risk), where each category represents an increasing risk of mortality. This tool is applied at the time of admission and utilizes specific clinical and laboratory variables to predict the outcome 14,15 . Medication Error classification and error severity according to the NCC MERP Medication errors (MEs) were classified using the NCC MERP index and grouped into three analytical categories: intercepted (Types A/B), moderate (Type C), and severe (Types D/E) 12, where severity was determined by clinical impact and the need for corrective interventions. To ensure reliability, two independent external pediatric pharmacists validated all classifications of medication errors and MEs, obtaining substantial agreement (Cohen’s kappa κ=0.702; p < 0.001). 16,17 Pharmaceutical intervention classification The number of PI was collected as a continuous variable and categorized (6) and with cut-off points defined based on the sample distribution. Outcomes The study outcomes were defined in two categories. The primary outcome was the association between the quantity of PIs and the occurrence and severity of MEs. As a secondary outcome, the relationship between clinical variables—including undergoing surgery, ICU LOS, outcome, and PIM score—and the severity of MEs was evaluated. Ethical aspects All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This study received approval from the Research Ethics Committee of the Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo (HCFMUSP) under CAAE 50570421.6.0000.0068. Statistical analysis Statistical analyses were performed in two stages. Initially, a descriptive analysis of the demographic and clinical characteristics of the sample was conducted, presenting categorical variables as absolute and relative frequencies, and continuous variables as measures of central tendency and dispersion, according to data distribution. To assess the association between clinical/demographic variables and the occurrence of MEs, Pearson’s chi-squared test was used. Subsequently, Poisson regression with robust variance was employed to estimate incidence rate ratios (IRR) and their respective 95% confidence intervals, considering the cumulative incidence of MEs. The severity of MEs, categorized according to the NCC MERP classification, was analyzed using multinomial regression models, estimating relative risk ratios (RRR) for the different severity levels compared to the no-error group. Univariate and, subsequently, multivariate analyses were performed, including in the final model the variables with p<0.20 in the initial analyses in order to identify independent associated factors. The significance level adopted was 5% (p<0.05) in all analyses. The analyses were performed using SPSS® software. Missing data management All collected data were sourced from electronic health records and structured pharmaceutical intervention registries. No missing data were identified in the main variables analyzed, as cases with incomplete information were excluded during the chart review process. Thus, no imputation or specific modeling techniques for missing data were applied in this study.

Sample selection Over 729 days (9,512 patient-days), 3,087 PIs were initially identified. Validation resulted in 16 exclusions (8 duplicates; 8 missing records) and 20 additions, yielding a final dataset of 3,091 PIs (0.32 PIs/patient-day). From 864 total admissions, 299 patients (34.6%) without recorded PIs were excluded. The final analytical cohort comprised 565 patients (65.4%) experiencing ≥ 1 ME (Figure 1). MEs identified through active pharmaceutical surveillance were classified by severity according to the NCC MERP index, with the full distribution of categories detailed in Supplementary Table 1. Characteristics of the population according to the occurrence of ME Next, in Table 1, demonstrated that demographics (age, sex) and clinical outcomes (discharge vs. death) were not significantly associated with ME frequency (p>0.05). Conversely, surgical patients significantly increased error rates (54.4% vs. 33.1%; p<0.001). Both PI volume and LOS showed a strong, progressive association with ME occurrence (p6 PIs and 64.5% for hospitalizations exceeding 16 days. While higher PIM scores trended toward increased ME rates, the association was not statistically significant (p=0.358). Occurrence and severity of medication errors In the analysis of ME severity according to clinical and demographic variables (Table 2), it was observed that there were no significant differences in relation to age (p=0.162), sex (p=0.257), and PIM mortality risk score (p=0.838). Conversely, undergoing surgery was associated with greater ME severity (p<0.001), with a higher proportion of events classified as categories D and E among patients submitted to a surgical procedure (54.4% with MEs, with 38.7% in category D and 15.7% in category C). Similarly, the LOS in the ICU significantly associated with the severity of MEs (p<0.001), progressively increasing the proportion of more severe events: from 21.2% in category D in hospitalizations shorter than 8 days, to 51.1% in those longer than 16 days. Cumulative incidence of medication errors Consistent with the findings on ME severity, the cumulative incidence analysis (Table 3) demonstrated that sex, age, and PIM-estimated mortality risk did not show a statistically significant association with the occurrence of MEs. In contrast, undergoing surgery remained an associated factor, with a 64% increase in the incidence of MEs among patients who underwent surgical procedures (IRR=1.64; 95% CI 1.35–2.00; p<0.001). ICU LOS also showed a significant and progressive association: hospitalizations of 8–16 days had a 43% higher risk of error compared to those of less than 8 days (IRR=1.43; 95% CI 1.12–1.83; p=0.004), while those over 16 days doubled the risk (IRR=2.00; 95% CI 1.62–2.47; p<0.001). No significant difference was observed according to hospital outcome (discharge vs. death; p=0.484). Univariate multinomial regression analysis Univariate analysis (Table 4) identified surgery, prolonged LOS, and PI frequency as significant predictors of ME severity. Surgery more than doubled the relative risk (RR) for both Type C (RR=2.78; 95% CI 1.60–4.81) and Types D/E errors (RR=2.28; 95% CI 1.57–3.32). LOS severe errors (95% CI 2.94–7.44). Notably, PI frequency demonstrated the strongest association; patients requiring >6 interventions faced a 7.66-fold risk for Type C and an 18.47-fold risk for Types D/E errors ($p < 0.001$). Conversely, age, sex, PIM score, and clinical outcomes showed no significant associations. Multivariate analysis and multinomial regression In the adjusted model (Table 5), PI frequency remained the primary independent predictor of ME severity. Compared to 6 PIs, the risk surged to RR=6.91 and RR=16.53, respectively (p < 0.001). While surgery remained an independent predictor for Type C errors (RR=2.07; p = 0.013), it lost significance for severe errors. Notably, after adjusting for PI volume, ICU LOS was no longer a significant predictor for any error category.

This study provides a comprehensive analysis of ME severity in a Brazilian PICU, highlighting the predictive value of clinical complexity. Findings confirm that care-related factors—specifically surgery and prolonged LOS—independently predict higher ME risk, whereas demographics do not. This reinforces that ME vulnerability is driven by therapeutic intensity rather than patient-specific traits. Notably, the cohort reflects a population under active pharmaceutical surveillance. This methodological framework identifies PI frequency as a robust marker of clinical complexity and heightened detection rather than a causal factor for errors. The association between surgery and a higher frequency of MEs was consistent with previous studies in North American and Asian pediatric ICUs, which identify the perioperative period as a critical time, with multiple therapeutic adjustments and exposure to high-risk medications 18,19,20 . Similarly, a LOS exceeding 16 days doubled the risk of MEs, in line with data from European cohorts that identified longer medication exposure time, team fatigue, and communication failures as predictors of adverse events 21 . As demonstrated by Sikora et al. (2022), the risk of MEs progressively increases in patients with prolonged ICU stays, due to polypharmacy and care overload, which reinforces our findings 22,23 . Likewise, a 2021 systematic review from Brazil showed that long ICU stays, associated with polypharmacy, therapeutic complexity, and multidisciplinary team fatigue, create a fertile ground for errors, with the pharmacist’s role being essential to mitigate them 24 . These results reinforce the need for systematic pharmacotherapeutic monitoring strategies in long-stay patients, with periodic prescription reviews. Another relevant finding was the strong association between the number of PIs and the severity of MEs. Patients with more than six PIs had a relative risk up to 16 times higher for severe MEs (types D/E). Although this result might suggest causality, the appropriate interpretation is that the frequency of PIs functions as a marker of clinical complexity, since more critically ill patients receive more medications and thus accumulate more opportunities for both the detection of failures and the occurrence of errors. This “vigilance paradox” has already been described in multicenter studies in the US and Europe 25,26,27, where an increase in interventions reflects intensified surveillance in higher-risk contexts, rather than a failure of pharmaceutical care. Thus, the number of PIs should be interpreted as an early warning indicator and a marker of care complexity. This finding suggests the feasibility of using the number of PIs as an institutional indicator of pharmacotherapeutic complexity, serving as a risk stratification tool to direct resources, such as more frequent clinical rounds or intensified multidisciplinary reviews in critically ill patients. The lack of a significant association between age, sex, and PIM score with the occurrence or severity of MEs is in line with national and international literature 28,29, which suggests that isolated demographic factors play a secondary role in the face of therapeutic complexity. On the other hand, previous studies demonstrate that the acceptance of PIs is a determinant for the reduction of MEs, reaching rates above 90% in European countries and the USA 30,31, whereas in some Brazilian settings this rate is similar or lower, possibly due to cultural and communication barriers 32,33 . This suggests that, in addition to the quantity of PIs, the quality of the pharmacist’s integration into multidisciplinary decisions is crucial for transforming recommendations into concrete clinical outcomes. Multivariate analysis revealed that the association between prolonged LOS and ME severity, though significant univariately, vanished after adjusting for PI frequency. Conversely, PI volume remained the most robust independent predictor for both moderate and severe errors. These findings indicate that the effect of LOS is mediated by accumulated pharmacotherapeutic complexity, positioning PI frequency as a superior and more direct surrogate for risk exposure than hospitalization time alone.

and strengths

include a retrospective design and partial pharmacist coverage, potentially underreporting errors outside working hours. While excluding patients without PIs precludes general prevalence estimates, the 1:1 PI-to-ME ratio (n=3,091) ensures high internal validity for detected risks. Methodological rigor was maintained through standardized protocols and training, while PI acceptance analysis was deferred to a future publication to ensure thematic clarity. Conversely, this study represents one of the largest Brazilian pediatric cohorts evaluating medication error severity. The findings reinforce the pharmacist’s essential role in surveillance, particularly for surgical and long-stay patients. Furthermore, validating PI frequency as a practical surrogate for clinical complexity provides a robust tool for early risk stratification and targeted safety interventions. These findings constitute a robust evidence base for the refinement of patient safety protocols across diverse clinical environments. However, it is imperative to situate these results within the international landscape; while continuous clinical pharmacy services are institutionalized in several high-income nations, the Brazilian context is characterized by structural asymmetries. Many domestic institutions have yet to operationalize structured clinical pharmacy programs, primarily due to systemic resource constraints, encompassing both human capital and budgetary limitations. Addressing this discrepancy through systematic pharmacotherapeutic monitoring and incremental service expansion—notwithstanding that 24-hour coverage remains an aspirational benchmark for most facilities—is fundamental to mitigating underreporting and maximizing the efficacy of PIs. Subsequent investigations will evaluate PI acceptability rates, providing a more granular understanding of their correlation with patient clinical outcomes

In this PICU cohort, the PI frequency emerged as the primary independent predictor of medication error severity, outperforming length of stay. While surgery independently predicted moderate errors, PI volume proved to be a more robust marker of underlying clinical complexity. These findings suggest that PI data should transition from a productivity metric to a practical risk-stratification tool. By leveraging PI frequency to identify clinically vulnerable patients, healthcare systems can target clinical pharmacy resources more effectively, proactively mitigating the risk of significant preventable harm. ETHICAL ASPECTS This retrospective chart review study involving human participants was in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This study received approval from the Research Ethics Committee of the Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo (HCFMUSP) under CAAE 50570421.6.0000.0068. ACKNOWLEDGMENTS The authors would like to thank the Clinical Pharmacy Team of the Instituto da Criança e do Adolescente (ICr), Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo (HCFMUSP), for their support and contribution to this study.

1. Kwiecień-Jaguś K, et al. Understanding medication errors in intensive care units: a systematic review. J Clin Med . 2025;61(3):369. 2. Bante A, et al. Medication errors and associated factors among pediatric intensive care units. J Pediatr Pharmacol Ther . 2023;28(2):115-121. 3. Alghamdi AA, Keers RN, Sutherland A, Ashcroft DM. Prevalence and nature of medication errors and preventable adverse drug events in paediatric and neonatal intensive care settings: a systematic review. Drug Saf . 2019;42(12):1423-1436. 4. Hannibal G, Vithanage N, Madhushika M, et al. A systematic review of prescription errors in paediatric care. BMC Health Serv Res . 2025;25(1):967. doi:10.1186/s12913-025-13109-6 5. Camarinha BD, Silva BC, Cordeiro BC, Camuzi RC, Brito MA. Associations between high-alert medications and potential drug-drug interactions in a pediatric intensive therapy unit. J Hosp Pharm Health Serv . 2023;14(3):802. https://jhphs.org/sbrafh/article/view/802. Accessed December 2, 2025. 6. Lago AD, et al. Avaliação das intervenções farmacêuticas em uma UTI pediátrica do Hospital Universitário Materno Infantil em São Luís – MA. Res Soc Dev . 2022;11(7):e36311729839. 7. Althomali A, Altowairqi A, Alghamdi A, et al. Impact of clinical pharmacist intervention on clinical outcomes in the critical care unit, Taif City, Saudi Arabia: a retrospective study. Pharmacy (Basel) . 2022;10(5):108. 8. Durand A, Gillibert A, Membre S, Mondet L, Lenglet A, Mary A. Acceptance factors for in-hospital pharmacist interventions in daily practice: a retrospective study. Front Pharmacol . 2022;13:811289. 9. Moges TA, Dagnew SB, Anberbr SS, et al. Clinical pharmacists’ interventions about drug therapy problems and its acceptability by prescribers among pediatric hospitalized patients with infectious diseases in resource-limited settings. BMC Infect Dis . 2025;25(1):629. 10. Smith SE, Shelley R, Newsome AS. Medication regimen complexity vs patient acuity for predicting critical care pharmacist interventions. Am J Health Syst Pharm . 2021. 11. Zhao B, Shen Y, Devlin JW, et al. Prediction of pharmacist medication interventions using medication regimen complexity. medRxiv . Preprint posted online October 23, 2024. doi:10.1101/2024.10.23.24316001 12. National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP). NCC MERP Index for Categorizing Medication Errors . NCC MERP; 2020. https://www.nccmerp.org/sites/default/files/nccmerp-index-medication-errors-2020.pdf. Accessed October 12, 2025. 13. World Health Organization. Adolescent health. Published 2024. Accessed December 1, 2024. https://www.who.int/health-topics/adolescent-health 14. Shann F, Pearson G, Slater A, et al. Paediatric index of mortality (PIM): a mortality prediction model for children in intensive care. Intensive Care Med . 1997;23(2):201-207. 15. Slater A, Shann F, Pearson G; Paediatric Index of Mortality (PIM) Study Group. PIM2: a revised version of the Paediatric Index of Mortality. Intensive Care Med . 2003;29(2):278-285. 16. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas . 1960;20(1):37-46. 17. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics . 1977;33(1):159-174. 18. Liu W, Liu Y, Zhu X, Zhang C, Ye D. Medication errors in pediatric intensive care units in China: a multicenter study. J Clin Pharm Ther . 2019;44(1):102-110. 19. Manias E, Kinney S, Cranswick N, Williams A, Borrott N. Interventions to reduce medication errors in pediatric intensive care: an integrative review. Paediatr Drugs . 2020;22(6):661-679. 20. Assiri GA, Grant L, Aljadhey H, Sheikh A. Investigating the epidemiology of medication errors and adverse drug events in pediatric intensive care units: a systematic review and meta-analysis. Eur J Clin Pharmacol . 2019;75(9):1261-1274. 21. Valdez CA, McBride SM, Kuntz JL, Bellows BK. Drug-related problems in the intensive care unit: prevalence, risk factors, and associated outcomes. Ann Pharmacother . 2015;49(8):771-780. 22. Sikora Newsome A, Catlin J, Dager WE, et al. Medication errors in the intensive care unit: an evidence-based systematic review. Crit Care Med . 2019;47(9):e765-e772. 23. Sikora A, Ayyala D, Rech MA, et al. Impact of pharmacists to improve patient care in the critically ill: a large multicenter analysis using meaningful metrics with the Medication Regimen Complexity-ICU (MRC-ICU) Score. Crit Care Med . 2022;50(9):1318-1328. 24. Lima L, da Silva M, de Sousa F, et al. Errors in the prescription, preparation, and administration of medicines in Pediatric and Neonatal Intensive Care Unit: systematic review. Res Soc Dev . 2021;10(14):e460101422315. 25. Kandaswamy S, Dawson TE, Moore WH, et al. Pharmacist metrics in the pediatric intensive care unit: an exploration of the Medication Regimen Complexity-Intensive Care Unit (MRC-ICU) Score. J Pediatr Pharmacol Ther . 2023;28(8):728-734. 26. Sikora A, Min W, Devlin JW, et al. Effect of comprehensive medication management on mortality in critically ill patients. Crit Care Med . Published online August 6, 2025. 27. Alghamdi A, Keers R, Sutherland A, et al. Incidence and nature of adverse drug events in paediatric intensive care units: a prospective multicentre study. Br J Clin Pharmacol . 2021;88(5):2213-2222. 28. Shirzad-Yazdi N, Taheri S, Vazin A, et al. Drug-related problems among pediatric intensive care units: prevalence, risk factors, and clinical pharmacists’ interventions. BMC Pediatr . 2024;24(1):714. 29. Kuitunen S, Saksa M, Holmström AR. Contributing factors of pediatric medication errors involving high-alert medications: a qualitative content analysis of self-reported medication safety incidents. Res Social Adm Pharm . Published online 2025. 30. Naseralallah LM, Hussain TA, Jaam M, Pawluk SA. Impact of pharmacist interventions on medication errors in hospitalized pediatric patients: a systematic review and meta-analysis. Int J Clin Pharm . 2020;42(4):979-994. 31. Lee H, Ryu K, Sohn Y, et al. Impact on patient outcomes of pharmacist participation in multidisciplinary critical care teams: a systematic review and meta-analysis. Crit Care Med . 2019;47(9):1243-1250. 32. Neves ER, Júlio CD, Viana GD, Pereira JA. Análise das intervenções farmacêuticas clínicas em Unidade de Terapia Intensiva de um hospital de urgência e trauma. Rev Cient Esc Estadual Saude Publica Goias . 2023;9:e90009. 33. Araujo EO, Viapiana M, Domingues EAM, Oliveira GS, Polisel CG. Intervenções farmacêuticas em uma unidade de terapia intensiva de um hospital universitário. Rev Bras Farm Hosp Serv Saude . 2017;8(3):25-30. Table 1. Frequency of medication errors according to sample characteristics | Category | n | % | p- value | | | 0-5 | 155/350 | 44.3 | || | Age (years) | 6-9 | 27/78 | 34.6 | 0.152 | | >10 | 66/137 | 48.2 | || | Gender | Female | 123/280 | 43.9 | 0.987 | | Male | 125/285 | 43.9 | || | Surgery | No | 92/278 | 33.1 | 1–5% | 109/239 | 45.6 | || | PIM score | >5–15% | 68/157 | 43.3 | 0.358 | | >15–30% | 15/27 | 55.6 | || | >30% (hight risk) | 9/17 | 52.9 | || | <3 | 57/255 | 22.4 | || | PI frequency | 3-6 | 60/148 | 40.5 | 6 | 131/162 | 80.9 | || | <8 | 90/279 | 32.3 | || | Length of PICU stay (days) | 8-16 | 67/145 | 46.2 | <0.001* | | <16 | 91/145 | 64.5 | || | Clinical outcome | Discharge | 216/498 | 43.4 | 0.497 | | Death | 32/67 | 47.8 | || | This table presents the distribution of medication errors (MEs) across different patient subgroups. Data are shown as the number of patients with at least one medication error over the total number of patients in each category (n), and the corresponding percentage (%). The p-value was calculated using the Chi-square test to assess the statistical significance of the association between each variable and the occurrence of MEs. Abbreviation. PIM: Pediatric Index of Mortality, a score used to estimate the risk of mortality upon ICU admission. PI: Pharmaceutical Interventions. PICU: Pediatric Intensive Care Unit. *P-value for Chi-square test : Indicates a statistically significant result (p < 0.05). | Table 2. Severity of medication errors, according to the NCCMERP classification, stratified by clinical and demographic variables. | Category | n | % | n | % | n | % | n | % | || | Age (years) | 0-5 | 195 | 55.7 | 48 | 13.7 | 107 | 30.6 | 0 | 0.0 | | | 6-9 | 51 | 65.4 | 6 | 7.7 | 21 | 26.9 | 0 | 0.0 | 0.162 | | | >10 | 71 | 51.8 | 14 | 10.2 | 51 | 37.2 | 1 | 0.7 | || | Gender | Female | 157 | 56.1 | 40 | 14.3 | 83 | 29.6 | 0 | 0.0 | 0.257 | | Male | 160 | 56.1 | 28 | 9.8 | 96 | 33.7 | 1 | 0.4 | || | Surgery | No | 186 | 66.9 | 23 | 8.3 | 68 | 24.5 | 1 | 0.4 | 1–5% | 130 | 54.4 | 29 | 12.1 | 79 | 33.1 | 1 | 0.4 | || | >5–15% | 89 | 56.7 | 16 | 10.2 | 52 | 33.1 | 0 | 0.0 | 0.838 | | | >15–30% | 12 | 44.4 | 4 | 14.8 | 11 | 40.7 | 0 | 0.0 | || | >30% | 8 | 47.1 | 3 | 17.7 | 6 | 35.3 | 0 | 0.0 | || | Length of PICU stay (days) | <8 | 189 | 67.7 | 31 | 11.1 | 59 | 21.2 | 0 | 0.0 | | | 8-16 | 78 | 53.8 | 19 | 13.1 | 48 | 33.1 | 0 | 0.0 | <0.001* | | | <16 | 50 | 35.5 | 18 | 12.8 | 72 | 51.1 | 1 | 0.7 | || | This table details the distribution of patients according to the severity of Medication Errors (MEs), classified using the NCCMERP scale. The data are presented as the number of patients (n) and their corresponding percentage (%) for each severity category (No ME, Type C, Type D, Type E), stratified by demographic and clinical variables. The p-value was calculated using the Chi-square test to assess the association between each variable and the severity of the ME. Abbreviation: PIM: Pediatric Index of Mortality. NCCMERP: National Coordinating Council for Medication Error Reporting and Prevention. PICU: Pediatric Intensive Care Unit. *: Indicates a statistically significant result (p < 0.05). | Table 3. Poisson regression analysis of factors associated with the cumulative incidence of medication errors. | Gender | Female | 1.00 | ||| | Male | 1.00 | 0.987 | 0.83 | 1.20 | | | 10 | 1.09 | 0.432 | 0.88 | 1.34 | | | 0–1% | ||||| | >1–5% | 1.21 | 0.154 | 0.93 | 1.58 | | | PIM score | >5–15% | 1.15 | 0.336 | 0.86 | 1.54 | | >15–30% | 1.48 | 0.060 | 0.98 | 2.22 | | | >30% | 1.41 | 0.182 | 0.85 | 2.33 | | | Surgery | No | |||| | Yes | 1.64 | <0.001* | 1.35 | 2.00 | | | Clinical outcome | Discharge | |||| | Death | 1.10 | 0.484 | 0.84 | 1.44 | | | 16 | 2.00 | <0.001* | 1.62 | 2.47 | | | This table presents the results of the Poisson regression analysis, showing the Incidence Rate Ratios (IRR) for the occurrence of Medication Errors (MEs). The IRR represents the relative risk of an ME occurring in one group compared to a reference category (IRR = 1.00). A 95% Confidence Interval (CI) is provided for each IRR. Abbreviation: PIM: Pediatric Index of Mortality. ICU: Intensive Care Unit. The p-value was derived from the Poisson regression model. *: Indicates a statistically significant result (p < 0.05). | Table 4. Univariate multinomial regression analysis of factors associated with medication error severity. | ME type C vs no ME | |||||| | 0-5 | 1.00 | ||||| | Age (years) | 6-9 | 0.48 | 0.109 | 0.19 | 1.18 | | | >10 | 0.80 | 0.506 | 0.42 | 1.54 | || | Gender | Female | 1.00 | |||| | Male | 0.69 | 0.165 | 0.40 | 1.17 | || | Surgery | No | 1.00 | |||| | yes | 2.78 | 0.000* | 1.60 | 4.81 | || | 0–1% | 1.00 | ||||| | >1–5% | 1.09 | 0.807 | 0.56 | 2.13 | || | PIM score | >5–15% | 0.88 | 0.733 | 0.41 | 1.87 | | | >15–30% | 1.63 | 0.448 | 0.46 | 5.69 | || | >30% | 1.83 | 0.409 | 0.44 | 7.65 | || | Clinical outcome | Discharge | 1.00 | |||| | Death | 1.23 | 0.606 | 0.56 | 2.69 | || | <8 | |||||| | Length of ICU stay (days) | 8-16 | 1.49 | 0.218 | 0.79 | 2.79 | | | <16 | 2.19 | 0.019* | 1.14 | 4.24 | || | 6 | 7.66 | 0.000* | 3.79 | 15.49 | || | ME type D or E vs no ME | 0-5 | 1.00 | |||| | Age (years) | 6-9 | 0.75 | 0.315 | 0.43 | 1.31 | | | >10 | 1.33 | 0.187 | 0.87 | 2.05 | || | Gender | Female | 1.00 | |||| | Male | 1.15 | 0.464 | 0.79 | 1.65 | || | Surgery | No | 1.00 | |||| | Yes | 2.2 8 | 0.000* | 1.57 | 3.32 | || | 0–1% | 1.00 | ||||| | >1–5% | 1.55 | 0.087 | 0.94 | 2.55 | || | PIM score | >5–15% | 1.47 | 0.161 | 0.86 | 2.52 | | | >15–30% | 2.31 | 0.074 | 0.92 | 5.78 | || | >30% | 1.89 | 0.274 | 0.61 | 5.88 | || | Clinical outcome | Discharge | 1.00 | |||| | Death | 1.18 | 0.562 | 0.67 | 2.07 | || | <8 | 1.00 | ||||| | Length of PICU stay (days) | 8-16 | 1.97 | 0.004* | 1.24 | 3.13 | | | <16 | 4.68 | 0.000* | 2.94 | 7.44 | || | 6 | 18.47 | 0.000* | 10.85 | 31.44 | || | This table presents the results of the univariate multinomial regression analysis, showing the Relative Risk Ratios (RR) for Type C errors and Type D or E errors, with the ’No ME’ group serving as the reference category. The RR indicates how many times more likely a group is to experience a specific type of error compared to the reference group (RR=1.00). The analysis provides the RR, the 95% Confidence Interval (CI), and the corresponding p-value. Abbreviation: PIM: Pediatric Index of Mortality. PI: Pharmaceutical Interventions. ME: Medication Error. PICU: Pediatric Intensive Care Unit. *: Indicates a statistically significant p-value (p < 0.05) from the multinomial regression model. | Table 5. Multivariate multinomial regression analysis: adjusted relative risk ratios for type C and type D or E errors compared to the No ME group. | ME type C vs no ME | 6 | 6.91 | <0.001* | 3.03 | 15.77 | ||| | Surgery | No | 1.00 | ||||| | Yes | 2.07 | 0.013* | 1.16 | 3.68 | ||| | <8 | 1.00 | |||||| | Length of ICU stay (days) | 8-16 | 0.89 | 0.731 | 0.45 | 1.76 | || | <16 | 0.81 | 0.614 | 0.37 | 1.80 | ||| | ME type D or E vs no ME | ||||||| | 6 | 16.53 | <0.001* | 8.85 | 30.89 | ||| | Surgery | No | 1.00 | ||||| | Yes | 1.31 | 0.227 | 0.84 | 2.04 | ||| | <8 | 1.00 | |||||| | Length of PICU stay (days) | 8-16 | 0.96 | 0.874 | 0.55 | 1.65 | || | <16 | 1.07 | 0.829 | 0.58 | 1.96 | ||| | This table displays the final results of the multivariate multinomial regression model, identifying independent predictors of Medication Error (ME) severity. The analysis shows the adjusted Relative Risk Ratios (RR) for experiencing Type C or Type D/E errors compared to the reference group with no errors. The model was adjusted for all variables shown. Abbreviation: PI: Pharmaceutical Interventions. PICU: Pediatric Intensive Care Unit. *: Indicates a statistically significant p-value (p < 0.05) in the final model. | Figure 1. Study flow diagram illustrating data collection and sample selection. The flowchart details the medical record review over a 729-day period (9,512 patient-days). Initial screening identified potential Pharmaceutical Interventions (PI). After excluding duplicates and missing data, a total of 3,091 PIs were included in the final analysis (rate: 0.32 PI/patient-day). The final study population consisted of 565 patients following the application of inclusion and exclusion criteria. Supplementary Table 1. Distribution of detected medication errors by severity category (N=3,091). | Category A | Circumstances or events that have the capacity to cause error (Potential error) | 1,365 | 44.16% | | Category B | An error occurred but did not reach the patient | 860 | 27.82% | | Category C | An error occurred that reached the patient but did not cause harm | 380 | 12.29% | | Category D | An error occurred that reached the patient and required monitoring and/or intervention to preclude harm | 470 | 15.21% | | Category E | An error occurred that resulted in temporary harm and required intervention | 16 | 0.52% | | Category F | An error occurred that resulted in initial or prolonged hospitalization and temporary harm | 0 | 0.00% | | Category G | An error occurred that resulted in permanent patient harm | 0 | 0.00% | | Category H | An error occurred that resulted in a near-death event (e.g., anaphylaxis, cardiac arrest) | 0 | 0.00% | | Category I | An error occurred that resulted in patient death | 0 | 0.00% | | Total | 3,091 | 100.0% | Categories A and B represent intercepted errors; Category C represents errors without harm; Categories D and E represent errors with increased monitoring or temporary harm. No errors of categories F through I were recorded during the study period. Abbreviation: NCC MERP, National Coordinating Council for Medication Error Reporting and Prevention. Information & Authors Information Version history Copyright This work is licensed under a Non Exclusive No Reuse License.

Authors Metrics & Citations Metrics Article Usage 149views 89downloads Citations Download citation Mônica Cristina Dutra Rodrigues – Rodrigues, Victor KaneKo Matsuno, Leonardo Pereira. Clinical predictors of medication error severity in the pediatric intensive care: the role of pharmaceutical interventions as complexity markers. Authorea. 22 December 2025. DOI: https://doi.org/10.22541/au.176643752.20841048/v1 DOI: https://doi.org/10.22541/au.176643752.20841048/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu.