A total of 1038 patients were eligible during the four-year study period, but 41 patients were excluded due to missing data or wrong registration. Subsequently, 997 acute poisoning patients were enrolled (Figure 1).
Differences in demographic characteristics and poison agents between groups
Table 1 summarizes the demographic characteristics and the results of univariate analysis for fatalities and survivors among the enrolled patients. The 70 fatal cases (6.7%) showed male predominance (72.9 vs. 48.8%, p < 0.01), lower body temperature (36.1 ± 1.2 vs. 36.4 ± 0.8°C, p = 0.03), and tachycardia (101.6 ± 29.3 vs. 92.2 ± 23.4 bpm, p = 0.01) compared with the surviving group. No significant difference in the mean age, triage respiratory rate, triage blood pressure, suicide attempts, psychiatric medical histories, and length of hospital stay were found between the groups. We identified the patient with mixed agent poisoning and took them into statistical analysis and be one of the variable. The fatal group has less mixed agent poisoning (n = 3, 4.2%) compared with the surviving group (n = 106, 11.4%) but there is no significant difference (p = 0.073). The three fatal cases were intoxicated by paraquat and amphetamine, organophosphate and benzodiazepine, and organophosphate and caustic agents.
Different poison agents among the fatal and surviving groups are shown in Table 2. In order to present the character of individual poison agents, we excluded the mixed poison agents, including three patients in fatal group and 32 patients in surviving group. The most common lethal agents were paraquat (N = 31, 46.3%), caustic agents (N = 7, 10.4%), digoxin (N = 6, 9.0%), and organophosphate (N = 4, 5.9%). The lethal agent associated with high odds ratio of in-hospital mortality was paraquat (OR 22.5, 95% CI 12.4-40.7), followed by carbamate (OR 13.7, 95% CI 1.9-99.1), amphetamine (OR 6.9, 95% CI 1.2-38.1), and digoxin (OR 4.8, 95% CI 1.8-12.5). Significant difference in the paraquat, carbamate, digoxin, and hypnotics were found between the fatal and surviving groups.
The association between ED triage vital signs and poison-related in-hospital mortality
The odds ratios (OR) of in-hospital mortality for SBP, BT, HR, and RR revealed J-shaped relationships (Figure 2). Patients with an SBP of more than 190 mmHg or less than 100 mmHg had a greater than two-fold increase in the OR for in-hospital mortality, respectively. Initial BT of less than 34°C or over 38°C showed seven- and two-fold increased OR for in-hospital mortality, respectively. A triage HR of below 50 bpm or above 120 bpm was associated with increase in OR for in-hospital mortality, respectively. RR >28 or <12 per minute was associated with higher odds of in-hospital mortality (RR <12, OR = 27.2; RR >28, OR = 7). The patients with extremely abnormal vital signs had the greatest risk of in-hospital mortality. Therefore, further analysis was performed to find out the proper cut-off values to predict the in-hospital mortality.
By constructing a receiver operating characteristic (ROC) curve, we plotted the true-positive rate (sensitivity) against the false-positive rate (1-specificity) at each point (Figure 3). The optimum cut-off points using triage vital signs to predict in-hospital mortality were BT <36 or >37°C, SBP <100 or >150 mmHg, HR <35 or >120 bpm, RR <16 or >20 per minute (Figure 3A).
After the univariate analysis, logistic regression analysis was performed (Table 3). ED triage vital signs exceeding cut-off values independently predicted in-hospital mortality after adjusting for variables (BT <36 or >37°C, OR 2.8, 95%CI 1.5 – 5.3, p < 0.01; SBP <100 or >150 mmHg, OR 2.5, 95%CI 1.4 – 4.7, p < 0.01; HR <35 or >120 bpm, OR 3.1, 95%CI 1.5 – 6.6, p < 0.01; RR <16 or >20 per minute, OR 1.4, 95%CI 0.7 – 2.9, p = 0.38).
The impact of paraquat in the study
Among the 70 fatal cases, 32 patients (45.7%) were intoxicated by paraquat. Nearly half (n = 32/68, 47%) of paraquat poisoning cases was fatal and has high odds ratio of mortality. Identifying the paraquat itself had a higher predictive value than the vital signs. To diminish or realize the impact of paraqaut in this study, we excluded the paraquat poisoned patients and re-conducted the statistic analysis. In Table 1, the mean age (p = 0.01), male gender (p < 0.01), mean body temperature (p < 0.04), mean heart rate (P < 0.01), mean respiratory rate (p < 0.01), mean diastolic blood pressure (p < 0.01), and length of hospital stay (p < 0.01) between the fatal and surviving groups revealed statistically significance after excluding the patients with paraquat intoxication. We constructed the ROC curve (Figure 3B) and found that the cut-off values to predict in-hospital mortality are nearly the same with total poisoning population. Logistic regression analysis was performed and the ED triage vital signs exceeding cut-off values independently predicted in-hospital mortality after adjusting for variables (BT <36 or >37°C, OR 3.2, 95%CI 1.4 – 7.1, p < 0.01; SBP <100 or >150 mmHg, OR 2.2, 95%CI 1.0 – 4.5, p = 0.04; HR <35 or >120 bpm, OR 2.7, 95%CI 1.2 – 6.0, p = 0.01; RR <16 or >20 per minute, OR 2.4, 95%CI 1.0 – 5.1, p = 0.03).