Characteristics and prognostic factors of adult poisoning patients in the intensive care unit: A retrospective cohort study using the Japanese intensive care patient database

Abstract

Poisoning presents significant challenges in intensive care, with presentations ranging from mild to life-threatening conditions. Understanding the characteristics and prognostic factors of poisoning patients is crucial for developing treatment strategies and optimizing resource allocation. This study examined the characteristics and outcomes of adult poisoning patients admitted to intensive care units and evaluated the applicability of the Acute Physiology and Chronic Health Evaluation III scoring system. We conducted a multicenter retrospective cohort study using the Japanese Intensive Care Patient Database from April 2015 to March 2023. Adult poisoning patients (aged ≥16 years) from 94 intensive care units were categorized into six groups: medications, psychotropic medications and substances of abuse, alcohols, domestic and industrial chemicals, toxic gases, and natural toxins and envenomations. Clinical characteristics, management, and outcomes were analyzed. Among 1930 patients (median age 45.0 years, 43.1% male), psychotropic medications constituted the largest category (58.5%). Intensive care unit and hospital mortality rates were 2.5% and 4.5%, respectively. Patients with chemical poisoning were significantly older (median 73.0 years) and required the highest mechanical ventilation rates (74.6%). Mechanical ventilation consistently prolonged intensive care unit stay across all categories. Acute Physiology and Chronic Health Evaluation III demonstrated good discriminative performance (area under the receiver operating characteristic curve 0.895) but systematically overestimated mortality risk (standardized mortality ratio 0.426, 95% confidence interval: 0.341–0.526). These findings highlight the limitations of applying general intensive care unit scoring systems to poisoning patients and emphasize the need for poisoning-specific prognostic tools.

Graphical Abstract

Highlights

• •Multicenter study of 1930 ICU poisoning patients across 94 centers in Japan.
• •Ventilated patients had longer ICU stays in all six poisoning categories.
• •APACHE III predicted mortality 10.7% vs observed 4.5% (SMR 0.426, p < 0.001).
• •Chemical poisoning patients: oldest (73 years), highest ventilation (74.6%).
• •Current ICU scoring systems inadequate; poisoning-specific tools needed.

1. Introduction

Poisoning presents a significant challenge in emergency medicine and intensive care medicine, with clinical presentations ranging from mild to life-threatening conditions. While most poisoning cases are not severe and can be managed with outpatient observation or minimal intervention, some patients require intensive treatment such as mechanical ventilation [1], [2]. The indications for intensive care and prognosis of these patients are influenced by multiple factors including age, gender, vital signs, and the specific toxic substances involved, and remain unclear [3], [4], [5], [6], [7].

Understanding the clinical characteristics and prognostic factors of poisoning patients admitted to the intensive care unit (ICU) is crucial for developing appropriate treatment strategies and optimizing resource allocation. Previous efforts to develop prediction models for ICU admission in poisoning patients have shown promise in reducing unnecessary observational admissions [8]. However, previous studies have predominantly been limited to single-center investigations or small sample sizes, with comprehensive nationwide analyses remaining scarce. We utilized the Japanese Intensive Care Patient Database (JIPAD) - the largest ICU patient database in Japan - which provided access to nationwide data from various ICUs [9]. This large-scale database enabled a comprehensive analysis of poisoning patients and their prognostic factors.

This study aimed to examine the characteristics and factors associated with ICU length of stay and outcomes in adult poisoning patients admitted to ICUs, and to evaluate the applicability of conventional ICU prognostic scoring systems to this patient population.

2. Material and methods

2.1. Data collection

We obtained data from the Japanese Intensive care PAtient Database (JIPAD), the largest clinical registry of ICU patients in Japan, established by the Japanese Society of Intensive Care Medicine. As of March 2023, 94 ICUs participate in the JIPAD, and the database includes information on over 320,000 patients. The JIPAD includes the clinical severity information required to calculate the risk of death using the Acute Physiology and Chronic Health Evaluation (APACHE) III model [10], [11].

Data were either manually entered by physicians, nurses, and medical assistants or automatically extracted from electronic medical records. Regardless of the method, the data were validated by local software and uploaded to the master database. These uploaded data were routinely checked thereafter by the JIPAD Working Group. This study was approved by the Institutional Review Board of Toranomon Hospital (Approval number: 2647; February 18, 2025) and conducted in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975. Given the de-identified nature of the data, the requirement for informed consent was waived.

2.2. Study population and variable selection

We included patients aged 16 years or older admitted to the ICU between April 2015 and March 2023. Exclusion criteria included non-poisoning cases, severe trauma, brain injury, or burns. Patients fulfilling all inclusion criteria were included in the analysis.

Poisoning cases were extracted based on disease codes and text descriptions in the database. These cases were then broadly classified into six categories based on the disease text descriptions. Antihypertensive agents, antiarrhythmic agents, caffeine, antidiabetic medications, and anti-inflammatory drugs such as acetaminophen and nonsteroidal anti-inflammatory drugs (NSAIDs) were categorized as ‘Medications’. Sedatives, antidepressants, lithium, opioids, amphetamines, and cases involving mixed substances from two or more drugs were classified as ‘Psychotropic medications and substances of abuse’. Ethanol and other toxic alcohols were categorized as ‘Alcohols’. Acids and alkalis, household detergents, bleach, and agricultural pesticides were classified as ‘Domestic and industrial chemicals’. Carbon monoxide poisoning (with or without smoke inhalation) and other gases were categorized as ‘Toxic gases’. Mushroom and plant toxins, puffer fish, and envenomations from snakes, spiders, and other animals were classified as ‘Natural toxins and envenomations’. As the disease text descriptions were not specifically designed for poisoning classification, the categorization was determined by three physicians specializing in emergency medicine and toxicology.

To examine the clinical characteristics and outcomes of poisoning patients, we collected the following variables: age, gender, Glasgow Coma Scale score, mean arterial pressure, PaO2/FiO2 ratio, pH, lactate levels, and return of spontaneous circulation status. We also collected severity scores, including Sequential Organ Failure Assessment (SOFA) [12] and Acute Physiology and Chronic Health Evaluation (APACHE) III scores, along with APACHE III predicted mortality. ICU management data included the use of catecholamines, mechanical ventilation, intermittent hemodialysis, continuous renal replacement therapy, and extracorporeal membrane oxygenation. Outcome measures comprised mechanical ventilation days, ICU length of stay, ICU mortality, hospital length of stay, and hospital mortality. We abstracted each patient's APACHE III scores to predict hospital mortality, as well as vital status at ICU and hospital discharge.

2.3. Calculations and statistics

We calculated clinical characteristics and outcomes for adult poisoning patients admitted to ICUs. Multiple linear regression analysis was performed to examine the association between poisoning categories and ICU length of stay, adjusted for mechanical ventilation status.

To evaluate the performance of the APACHE III scoring system in poisoning patients, we calculated the area under the receiver operating characteristic curve (AUC) for the overall patient cohort and by poisoning category. The standardized mortality ratio (SMR) was calculated as the ratio of observed to expected deaths based on APACHE III predicted mortality. Calibration assessment was performed using locally weighted scatterplot smoothing (LOESS) to visualize the relationship between predicted and observed mortality rates [13].

Descriptive data are summarized as the mean (SD), median (interquartile range, IQR), or number (percentage). A P value < 0.05 was considered statistically significant. All statistical analyses were performed using R version 4.4.2 (2024 The R Foundation for Statistical Computing, Vienna, Austria).

3. Results

A total of 1930 adult poisoning patients from 94 ICUs across Japan were included in the final analysis after applying the inclusion and exclusion criteria (Fig. 1). The median age was 45.0 years (IQR: 29.0–62.0), and 832 patients (43.1%) were male. The median ICU length of stay was 1.76 days (IQR: 0.92–3.38). The overall ICU mortality rate was 2.5% (49 patients), while the hospital mortality rate was 4.5% (87 patients).

Fig. 1.

Study population flowchart, ICU, intensive care unit.

3.1. Patient characteristics and clinical management

The poisoning cases were categorized into six groups based on the causative substances (Table 1). Psychotropic medications and substances of abuse constituted the largest category with 1129 patients (58.5%), followed by medications with 295 patients (15.3%). Toxic gases accounted for 188 patients (9.7%), domestic and industrial chemicals for 138 patients (7.2%), alcohols for 103 patients (5.3%), and natural toxins and envenomations for 59 patients (3.1%). Within the psychotropic medications category, mixed substances from two or more drugs were most common, affecting 859 patients (44.5% of the total cohort).

Table 1.

Categories of suspected poisoning substances.

Group of poisoning substance (example)	Number (%)
Medications
Antihypertensive agent, antiarrhythmic agent	45 (2.3%)
Caffeine, theophylline	87 (4.5%)
Oral antidiabetic agent, insulin	37 (1.9%)
Anticholinergic agent (antihistamine agent)	18 (0.9%)
Cholinergic agent	11 (0.6%)
Anti-inflammatory agent (acetaminophen, NSAIDs, colchicine)	72 (3.7%)
Other medications	25 (1.3%)
Psychotropic medications and substances of abuse
Mixed substances from two or more categories	859 (44.5%)
Sedatives, hypnotics, antiepileptic agent	96 (5.0%)
Antidepressants, antipsychotics (TCAs, SSRI)	40 (2.1%)
Lithium	90 (4.7%)
Opioids	11 (0.6%)
Local anesthetics	13 (0.7%)
Amphetamines	20 (1.0%)
Alcohols
Ethanol	88 (4.6%)
Other toxic alcohols (methanol, ethylene glycol)	15 (0.8%)
Domestic and industrial chemicals
Acids and alkalis, household detergents, bleach	42 (2.2%)
Agricultural pesticides (organic phosphorus, glyphosate)	96 (5.0%)
Toxic gases
Carbon monoxide without smoke inhalation	116 (6.0%)
Carbon monoxide with smoke inhalation	49 (2.5%)
Other gases (chlorine, hydrogen sulfide)	23 (1.2%)
Natural toxins and envenomations
Natural toxins (mushroom, plants, puffer fish)	17 (0.9%)
Animal bite (snake, spider, tick)	40 (2.1%)
Others	2 (0.1%)
Unspecified	18 (0.9%)
Total	1930 (100%)

* NSAIDs;Nonsteroidal anti-inflammatory drugs, TCAs;tricyclic antidepressants, SSRI;selective serotonin reuptake inhibitor

Significant differences in patient characteristics were observed across poisoning categories (Table 2). Patients with domestic and industrial chemical poisoning were significantly older (median age 73.0 years) compared to other categories (P < 0.001). Male predominance was most pronounced in alcohol poisoning (82.5%) and toxic gas exposure (68.6%) (both P < 0.001), while psychotropic medication poisoning showed female predominance (65.7% female, P < 0.001). Patients with psychotropic medication poisoning had the lowest median Glasgow Coma Scale score (9.0, P < 0.001), while those with natural toxin exposure had the highest (15.0, P < 0.001).

Table 2.

clinical characteristics and outcomes of poisoning patients by substance category.

	Total	Medications		Psychotropic medications and substances of abuse		Alcohols		Domestic or industrial chemicals		Toxic gases		Natural toxins and envenomations
	Number (%) or median (IQR)	Number (%) or median (IQR)	P values	Number (%) or median (IQR)	P values	Number (%) or median (IQR)	P values	Number (%) or median (IQR)	P values	Number (%) or median (IQR)	P values	Number (%) or median (IQR)	P values
Number of patients	1930 (100%)	295 (15.3%)		1129 (58.5%)		103 (5.3%)		138 (7.2%)		188 (9.7%)		59 (3.1%)
ICU admission
Age, years	45.0 (29.0–62.0)	38.0 (23.0–61.0)	< 0.001	43.0 (29.0–54.0)	< 0.001	48.0 (29.5–61.5)	0.957	73.0 (60.0–81.0)	< 0.001	50.0 (31.5–66.3)	0.029	69.0 (52.0–79.0)	< 0.001
male	832 (43.1%)	120 (40.7%)	0.394	387 (34.3%)	< 0.001	85 (82.5%)	< 0.001	63 (45.7%)	0.591	129 (68.6%)	< 0.001	40 (67.8%)	< 0.001
Glasgow Coma Scale score	11.0 (6.0–12.0)	14.0 (11.0–15.0)	< 0.001	9.0 (5.0–14.0)	< 0.001	12.0 (5.5–14.0)	0.266	13.0 (6.0–15.0)	0.293	12.5 (6.0–15.0)	0.250	15.0 (14.0–15.0)	< 0.001
Mean arterial pressure	67.0 (58.0–77.0)	66.0 (54.3–77.0)	0.306	68.0 (59.0–77.0)	< 0.001	63.0 (56.5–74.0)	0.046	61.0 (52.0–70.6)	< 0.001	66.0 (57.5–76.0)	0.783	69.5 (54.3–79.0)	0.591
PaO2/FiO2 ratio	375.0 (265.0–472.9)	431.4 (339.9–519.1)	< 0.001	359.4 (255.0–457.3)	< 0.001	393.6 (287.7–485.9)	0.149	337.5 (238.4–424.6)	0.002	387.2 (262.1–495.3)	0.161	377.5 (281.4–438.3)	0.802
pH	7.38 (7.34–7.43)	7.40 (7.34–7.44)	0.036	7.39 (7.34–7.43)	0.982	7.37 (7.29–7.43)	0.044	7.38 (7.32–7.43)	0.050	7.38 (7.32–7.43)	0.420	7.40 (7.37–7.44)	0.126
Lactate	2.0 (1.2–3.6)	2.9 (1.6–5.9)	< 0.001	1.8 (1.1–3.0)	< 0.001	3.8 (2.1–6.4)	< 0.001	2.3 (1.3–4.6)	0.128	2.0 (1.1–2.9)	0.053	1.6 (1.1–2.8)	0.154
Return of spontaneous circulation	59 (3.1%)	18 (6.1%)	0.002	18 (1.6%)	< 0.001	6 (5.8%)	0.167	6 (4.3%)	0.310	9 (4.8%)	0.175	1 (1.7%)	1.000
Severity score
SOFA	5.0 (3.0–7.0)	4.0 (1.0–7.0)	< 0.001	5.0 (3.0–7.0)	0.610	5.0 (3.0–8.0)	0.052	6.0 (3.0–8.0)	0.004	5.0 (3.0–7.0)	0.507	5.0 (2.0–8.0)	0.721
APACHE III	57.0 (36.0–81.0)	49.0 (33.0–78.5)	0.016	58.0 (35.0–79.0)	0.044	60.0 (39.5–81.0)	0.610	73.0 (52.0–100.8)	< 0.001	56.5 (39.8–82.0)	0.409	56.0 (39.0–78.5)	0.757
APACHE III mortality	2.82 (0.78–10.9)	1.59 (0.57–13.5)	0.032	2.68 (0.69–8.30)	< 0.001	4.34 (1.09–18.5)	0.037	6.06 (1.95–25.1)	< 0.001	4.00 (1.49–16.1)	0.005	4.19 (1.50–21.7)	0.043
ICU management
Use of catecholamines	482 (26.7%)	72 (24.4%)	0.876	276 (24.4%)	0.723	24 (23.3%)	0.828	39 (28.3%)	0.514	52 (27.7%)	0.619	15 (25.4%)	0.879
Use of mechanical ventilation	981 (50.8%)	97 (32.9%)	< 0.001	583 (51.6%)	0.425	50 (48.5%)	0.707	103 (74.6%)	< 0.001	119 (63.3%)	< 0.001	20 (33.9%)	0.012
Intermittent hemodialysis	193 (10.0%)	61 (20.7%)	< 0.001	98 (8.7%)	0.027	8 (7.8%)	0.543	10 (7.2%)	0.331	4 (2.1%)	< 0.001	9 (15.3%)	0.252
Continuous renal replacement therapy	191 (9.9%)	51 (17.3%)	< 0.001	99 (8.8%)	0.058	10 (9.7%)	1.000	15 (10.9%)	0.658	4 (2.1%)	< 0.001	9 (15.3%)	0.179
Extracorporeal membrane oxygenation	25 (1.3%)	12 (4.1%)	< 0.001	9 (0.8%)	0.025	0 (0%)	0.641	0 (0%)	0.253	3 (1.6%)	0.729	1 (1.7%)	0.542
Outcomes
Mechanical ventilation days	1.30 (0.65–3.08)	1.58 (0.64–3.95)	0.373	1.09 (0.62–2.08)	< 0.001	0.85 (0.47–2.56)	0.036	3.14 (1.08–7.11)	< 0.001	2.03 (0.93–5.12)	< 0.001	1.43 (0.55–3.11)	0.729
ICU length of stay	1.76 (0.92–3.38)	1.75 (0.95–3.35)	0.847	1.63 (0.89–2.85)	< 0.001	1.56 (0.69–3.04)	0.048	3.42 (1.73–7.40)	< 0.001	2.18 (1.15–5.63)	< 0.001	1.40(0.70–3.92)	0.373
ICU mortality	49 (2.5%)	14 (4.7%)	0.009	17 (1.5%)	0.001	0 (0%)	0.092	8 (5.8%)	0.012	7 (3.7%)	0.277	3 (5.1%)	0.207
Hospital length of stay	5.0 (2.00–17.0)	5.0 (2.0–17.5)	0.403	4.0 (2.0–12.0)	< 0.001	4.0 (1.0–12.0)	0.013	22.0 (8.0–46.5)	< 0.001	11.5 (3.0–26.0)	< 0.001	12.0 (5.0–38.0)	< 0.001
Hospital mortality	87(4.5%)	26 (8.8%)	< 0.001	29 (2.6%)	< 0.001	1 (1.0%)	0.075	16 (11.6%)	< 0.001	12 (6.4%)	0.192	3 (5.1%)	0.829

IQR, interquartile range; ICU, intensive care unit; SOFA, Sequential Organ Failure Assessment; APACHE, Acute Physiology and Chronic Health Evaluation

Catecholamine use showed no significant differences between categories. Mechanical ventilation requirements varied significantly by poisoning category. Domestic and industrial chemical poisoning had the highest rate of mechanical ventilation (74.6%, P < 0.001), followed by toxic gas exposure (63.3%, P < 0.001). Intermittent hemodialysis and continuous renal replacement therapy were most frequently required in medication poisoning (20.7%, P < 0.001 and 17.3%, P < 0.001). The multivariate analysis revealed that patients requiring mechanical ventilation had substantially longer ICU stays across all categories (Fig. 2). Among mechanically ventilated patients, domestic and industrial chemical poisoning had the longest mean ICU stay (6.26 days).

Fig. 2.

ICU length of stay by poisoning category and mechanical ventilation. Legend: Box plots showing ICU length of stay (days) across six poisoning categories, stratified by mechanical ventilation status (red: no mechanical ventilation, blue: mechanical ventilation). The boxes represent the interquartile range (25th-75th percentiles), horizontal lines within boxes represent medians, and dots represent outliers. Mechanical ventilation consistently prolonged ICU stay across all poisoning categories. Med, medications; Psy, psychotropic medications and substances of abuse; Alc, alcohols; Che, domestic and industrial chemicals; Gas, toxic gases; Nat, natural toxins and envenomations.

3.2. APACHE III performance in poisoning patients

The APACHE III scoring system demonstrated good discriminative performance for the overall poisoning cohort, with an area under the receiver operating characteristic (ROC) curve of 0.895 (Fig. 3, left panel). However, when examined by poisoning category, the AUC varied: medications (0.903), psychotropic medications (0.929), alcohol poisoning (1.000), domestic/industrial chemicals (0.717), toxic gases (0.938), and natural toxins (0.926).

Fig. 3.

APACHE III discriminative performance and calibration in poisoning patients. Legend: (a) Receiver operating characteristic (ROC) curve showing the discriminative performance of APACHE III for predicting hospital mortality in the overall poisoning cohort (AUC = 0.895). The table shows category-specific AUC values. (b) Calibration plot comparing predicted versus observed mortality rates using locally weighted scatterplot smoothing (LOESS). The diagonal dashed line represents perfect calibration. The solid curve shows systematic overestimation of mortality risk (SMR = 0.426, 95% CI: 0.341–0.526) with poor calibration. (Hosmer-Lemeshow test P < 0.001). The table shows category-specific standardized mortality ratios (SMR). AUC, area under the curve; SMR, standardized mortality ratio; CI, confidence interval; Med, medications; Psy, psychotropic medications and substances of abuse; Alc, alcohols; Che, domestic and industrial chemicals; Gas, toxic gases; Nat, natural toxins and envenomations.

The calibration analysis revealed significant systematic overestimation of mortality risk by the APACHE III system in poisoning patients (Fig. 3, right panel). The overall standardized mortality ratio (SMR) was 0.426 (95% CI: 0.341–0.526), indicating that observed deaths were less than half of those predicted. The Hosmer-Lemeshow test showed poor calibration (χ² = 118.72, P < 0.001). Category-specific SMRs varied widely: medications (0.696), psychotropic medications (0.308), alcohol poisoning (0.064), domestic/industrial chemicals (0.708), toxic gases (0.470), and natural toxins (0.336) with all categories showing significant overestimation of mortality risk compared to actual outcomes. The overall hospital mortality rate of 4.5% was substantially lower than the mean APACHE III predicted mortality of 10.7%.

4. Discussion

Our key findings demonstrate that mechanical ventilation consistently prolonged ICU stay across all poisoning categories. Additionally, the APACHE III scoring system overestimated mortality risk in poisoning patients, with considerable variation among categories.

Our analysis revealed clinical patterns across the six poisoning categories that reflect the unique pathophysiology of different toxic exposures. Patients with domestic and industrial chemical poisoning were significantly older (median age 73.0 years) and required the highest rates of mechanical ventilation (74.6%), resulting in the longest ICU stays among ventilated patients (6.26 days). This finding aligns with the known severity of caustic injuries and pesticide poisoning, which cause multi-organ dysfunction, including respiratory failure, gastrointestinal injury, and cardiovascular instability requiring intensive support [14], [15], [16].

In contrast, psychotropic medication poisoning, the largest category (58.5% of cases), predominantly affected younger females with lower consciousness levels (median Glasgow Coma Scale score 9.0) but had relatively favorable outcomes with lower mortality rates (2.6%). The female predominance (65.7%) and younger age profile are consistent with intentional overdose patterns reported in previous studies [17], [18]. The predominance of mixed substance poisoning within this category (44.5% of the total cohort) reflects the complex nature of intentional overdoses in clinical practice, where patients often consume multiple medications simultaneously [19].

Male predominance in alcohol (82.5%) and toxic gas exposure (68.6%) likely reflects occupational and lifestyle exposure patterns. The high mechanical ventilation rate in toxic gas exposure (63.3%) is consistent with the respiratory complications associated with carbon monoxide poisoning and smoke inhalation injury [20], [21], [22], [23].

Our multivariate analysis revealed that patients requiring mechanical ventilation had substantially longer ICU stays across all categories. Additionally, the variation in ventilation requirements by category (ranging from 32.9% in medications to 74.6% in chemicals) suggests that poisoning-specific triage protocols could improve resource utilization. Unlike general ICU patients, where mechanical ventilation often reflects underlying chronic conditions, poisoning patients typically require ventilation for acute, potentially reversible pathophysiology, which may explain the generally favorable outcomes despite intensive support requirements.

The overestimation of mortality risk by APACHE III in poisoning patients (SMR 0.426) represents a limitation of applying general ICU scoring systems to this specialized population. While APACHE III demonstrated good discriminative performance overall (AUC 0.895), the wide variation in category-specific performance (AUC ranging from 0.717 for chemicals to 1.000 for alcohols) indicates that the scoring system fails to capture the unique pathophysiology of different poisoning types. The SMR variation across categories (0.064 for alcohols to 0.708 for chemicals) further supports this conclusion.

This overestimation likely reflects fundamental differences between poisoning patients and the general ICU population used to develop APACHE III. Poisoning patients are often younger, have acute reversible conditions, and may present with profound physiological derangement that rapidly improves with supportive care and specific antidotes. This contrasts with the chronic comorbidities and progressive organ failure common in general ICU populations. Additionally, the APACHE III model was developed using data from 1988 to 1989, a period when intensive care management was less advanced, potentially contributing to the overestimation in contemporary practice [11].

Previous studies have also reported similar findings regarding the poor performance of general ICU scoring systems in poisoned patients [24]. A Spanish multicenter study reported that SAPS3 systematically overestimated mortality in poisoned patients (predicted mortality 26.8% vs observed mortality 6.7%, Hosmer-Lemeshow test: P < 0.001), while APACHE III showed appropriate predictions. This was attributed to the fact that the APACHE system includes specific diagnostic category weights for drug overdose, whereas SAPS3 lacks such adjustment. However, the small sample size may have influenced the results in this study.

The major strengths of this study include the use of the largest ICU database in Japan and being the first to systematically evaluate category-specific performance differences. Data collected from 94 centers over 8 years provided a sufficient sample size for subgroup analyses and included a broad spectrum of patient populations and practice patterns not achievable in single-center studies. The minimal missing data in the database also enhances the reliability of our findings.

However, several limitations should be acknowledged. First, voluntary participation in JIPAD potentially introduces selection bias toward larger academic centers with greater interest in quality improvement and research, which may limit generalizability to smaller community hospitals. Additionally, as JIPAD operates as a facility-level opt-out registry, the number of patients who declined participation cannot be determined, precluding a formal assessment of their potential impact on the results. Second, generalizability to other countries may be limited due to differences in poisoning patterns, healthcare systems, demographics, and clinical practice patterns. Third, the database was not specifically designed for poisoning research, which may have introduced variability in substance categorization despite expert physician review. The broad categories used may not capture important differences within each group, such as specific pesticide types or individual medications. Fourth, we could not assess the impact of specific antidotes, poison control center consultations, or specialized toxicology interventions, which may significantly influence outcomes. Finally, the 8-year study period may have captured temporal changes that could affect the results. These include changes in antidote availability and usage, ICU admission criteria, and management protocols, as well as patient-level factors such as alterations in healthcare-seeking behavior during the Coronavirus Disease 2019 pandemic and shifts in psychotropic medication prescribing trends.

This study demonstrates that poisoning patients admitted to ICUs have distinct characteristics and outcomes that vary significantly by substance category. The systematic overestimation of mortality risk by APACHE III across all categories highlights the urgent need for developing poisoning-specific prognostic scoring systems. The ICU Requirement Score, specifically developed for poisoned patients, has demonstrated excellent negative predictive value (0.99) for excluding ICU requirements, though its positive predictive value remains limited [25], [26]. Future research should focus on developing category-specific prediction tools, as this may be warranted given the substantial differences observed among poisoning types.

5. Conclusion

This multicenter retrospective study demonstrates that poisoning patients admitted to ICUs have distinct characteristics and outcomes that vary significantly by substance category. Mechanical ventilation consistently prolonged ICU stay across all categories, with domestic and industrial chemical poisoning requiring the most intensive support. The systematic overestimation of mortality risk by APACHE III highlights the limitations of general ICU scoring systems in poisoned patients.

Author contributions

The idea and design of the paper were conceptualized by KH and MS. Literature review and data analysis were performed by all authors. All authors revised and commented on subsequent versions of the manuscript. All authors read and approved the final version of the paper.

CRediT authorship contribution statement

Manabu Sugita: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Katsura Hayakawa: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Kazumasa Yamaguchi: Writing – review & editing, Methodology, Funding acquisition, Data curation, Conceptualization.

Ethics approval

Approval was obtained from the Institutional Review Board of Toranomon Hospital (Approval number: 2647; February 18, 2025).

Consent to participate

Consent to participate was waived because of the anonymous nature of the data.

Consent to publication

Not applicable

Funding

No external funding was received. All investigations were funded by the authors contributions.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The authors do not have permission to share data.