Study design, setting, and selection of the study population

We conducted a prospective study involving recruited consecutive cases of acute and chronic poisoning exposure in children and adults who presented between January 1, 2001 and December 31, 2002 to the EDs of the only two medical centers in the Tainan area. This area is in the southwest region of Taiwan, includes Tainan City and the Tainan County, and had a population of about 1,850,000 during the study period. The study proposal was reviewed and approved by the Research Grant Review Committee of National Science Council, Taiwan, R.O.C., and the Research Grant Review Committee of the National Cheng Kung University.

Data collection and processing

All consecutive cases of poisoning exposure, including cases with adverse drug effects of prescribed medication were enrolled within 24 hours of arrival at the ED, but cases diagnosed as food poisoning were excluded. We also enrolled patients who died on arrival. Poisoning was defined as exposure to drugs or any environmental substance resulting in an ED visit, even if the patient was considered symptom-free by the ED staff. Acute exposure was defined as a single exposure, continuous exposure lasting less than 8 hours, or repeated exposures over a period no longer than one week [2]. Chronic exposure was defined as continuous exposure lasting for 8 hours or more, or repeated exposures over a period longer than one week [1, 2]. A poisoning report form was developed for data collection, and periodically evaluated and refined on the basis of preliminary data. The information collected included demographic data; exposure agents, time, types, routes and reasons; history of suicide attempts, psychiatric diseases, substance abuse or major systemic disease; presenting symptoms and signs; details of psychiatric consultation; methods of management; clinical observations; ED and final outcomes; and laboratory data. Poisoning report forms were completed for all enrolled patients except those diagnosed with food poisoning. Report form data were initially obtained through direct interview of patients or patient caregivers by the ED staff and medical assistants. The reports were completed by trained assistants who followed the patients during hospitalization. Those who were discharged with stable condition were routinely followed by the psychiatric or toxicological outpatient clinic for at least once, and therefore we could obtain information on each patient's condition after discharge. For those who were discharged with unstable condition, were transferred to other hospitals, or could not return to the outpatient clinics, follow-up phone calls were made to the patients or their friends or relatives to obtain information.

According the practice of the two participating EDs, all poisoning cases receive psychiatric consultation or social worker consultation, and therefore, the suicide intents were assessed by professional staff. For those who were unable to be assessed in the ED, the consultation would be performed after their conditions were stabilized. On most of the critically ill and die-on-arrival patients, we managed to obtain information from their friends, relatives, emergency paramedical technicians, and policemen.

Outcome measures

All enrolled patients were tracked throughout the hospitalization period and during follow-up to document final outcomes. These were categorized as survival or poisoning-related fatality, the latter being defined as death during hospitalization or death after discharge that could be attributable to the poisoning episode.

Primary data analysis

The potential poisoning-related fatality predictors evaluated in this study included demographic characteristics (age and gender), exposure types (acute or chronic), nature of the exposure agents, the number of exposure agents (multiple or single), suicidal intent as the cause of poisoning, and concomitant use of alcohol. Vital signs (initially checked on the patient's arrival at the ED) including consciousness level, body temperature (BT), respiration pattern and rate, blood pressure and heart rate were also assessed. Past medical history of attempted suicide, psychiatric disease (e.g., schizophrenia, major depression, sleeping disorder), and other major or chronic diseases (cancers and diseases for which long-term medication is needed) was obtained from the patient or medical records. Laboratory data also recorded for reference included blood-cell counts, liver and renal function, blood gas analyses and x-rays.

We analyzed the data using SPSS statistical software (11.5.0; SPSS Inc., Chicago, IL). The exploratory data analyses checked the distribution of values and presented the results as the mean and standard deviation (SD) for numerical data, and a proportion (%) for nominal data. Differences between survivors and fatalities were tested using the chi-square test or Fisher's exact test for categorical variables, and t test for continuous variables. A p value < 0.05 was considered statistically significant, and all the statistical tests were two-tailed. We calculated the odds ratios (ORs) to assess the relationship between potential predictors and outcomes. Multiple logistic regression analyses were applied to adjust for confounders, to obtain adjusted estimates of OR and the 95% confidence interval (CI) for the independent effects of the potential predictors, and to identify significant independent predictors of outcomes.

Characteristics of the poisoning cases

From January 1, 2001 to December 31, 2002, a total of 1512 individuals presented with poisoning to the ED of the two medical centers, corresponding to an average of 4.2 per 1000 ED visits. The estimated incidence of poisoning events was around 0.41 poisoning exposures per 1000 population served in the Tainan area annually. Acute poisoning accounted for 98.0% of the enrolled poisoning cases, and involved more women (828) than men (684). Most cases involved people between 19 and 50 years of age (66.8%), and particularly those 19–30 years (29.4%). The mean age was 38.8 years (SD = 18.8, median = 35). Of 1548 poisoning agents implicated during the study period, drugs dominated in all age groups (49.9%), followed by pesticides (14.5%) (Figure 1). Overall, 66.1% of the poisoning exposures involved suicidal intent. Concomitant use of alcohol was recorded for 62.4% of the enrolled patients, and 17.3% of them used alcohol in the episode (Table 1).

Clinical status of the poisoning cases

Characteristics of the fatality cases

Predictors of fatality

Univariate analyses identified the following significant predictors for poisoning-related fatalities: age ≥ 61 years (OR 4.3, 95% CI 2.6–7.2), male gender (OR 2.5, 95% CI 1.5–4.3), attempted suicide (OR 6.3, 95% CI 2.5–15.7), history of chronic or major disease (OR 2.7, 95% CI 1.6 – 4.5), abnormal vital signs (decreased GCS, abnormal body temperature, shock status, abnormal heart rate, and insufficient respiration), and exposure agents (paraquat, organophosphates, carbamates, blasticidine-S, KCN, CO, and hypnotics/sedatives) (Tables 1, 2 and 3).

Characteristics of the poisoning cases

The frequency of poisoning-related ED visits varies around the world, ranging from 0.76 to 24 per 1000 ED visits [5, 10, 12, 34–37], and the ratio of females to males involved also varies [4, 5, 10, 12, 35, 36, 38, 39]. In our study the frequency of poisoning-related ED visits was 4.2 per 1000 (around the middle of the published range) and females dominated, as has been found in some other countries [10, 36, 38, 40]. Consistent with a recent study in southern Taiwan [20] and studies from other countries [6, 8, 10, 35, 38, 40], attempted suicide was the most common (66.1%) cause of poisoning.

Drugs (49.9%) were the dominant poisoning agent in our study. This phenomenon has been reported in other regions of Taiwan [20], in many Western countries [1, 4, 6, 37], and in some eastern Mediterranean [3] and Asian [38, 39] countries. The implementation of the national health insurance program resulted in the wide availability of prescription drugs to the residents in Taiwan, which should be one of the major factors leading to the phenomenon observed in our study. Other contributing factors might include the prescription patterns of physicians, the dissemination of news related to suicides through the mass media, and user's knowledge about the effects of drug overdose.

We also found a negative association between exposure to hypnotics/sedatives and poisoning fatality. Most hypnotics/sedatives are considered to be relatively safe, and no case in our study involved an individual taking hypnotics/sedatives and pesticides (which caused most of the fatalities; 73.0%) at the same time. We believe these are the main reasons why exposure to drugs appears to be a protective factor in our study.

Characteristics of fatality cases

In our study the poisoning-related fatality rate was 4.3%, which is very similar to that reported for Taiwan (5.7%) by the PCC [2]. This is much higher than in most Western countries [1, 4, 6, 41], probably because of the wide use of highly toxic agrochemicals in Taiwan. Although drugs accounted for most cases of poisoning in our study, pesticides were a major contributor and paraquat exposure was a strong predictor of fatality. Pesticide exposure has also been reported to be a major poisoning problem in Taiwan [2, 19, 20] and some other countries [13, 39, 42, 43]. Paraquat (24% w/v) is one of the most commonly used herbicides in Taiwan and has been the most common lethal agent of poisoning for a long time [2, 24, 44, 45]. There are various explanations for the relatively large number of pesticide poisonings in Tainan. In particular, Tainan is located in an agricultural area, making pesticides easily accessible, and appropriate storage containers are not generally used. Another factor may be a lack of knowledge amongst the general public about the toxic effects of these chemicals. In addition to the measures that have been taken by authorities, including banning some of the most toxic pesticides (e.g., endosulfan and parathion) and promoting less use of pesticides, there is a need for greater efforts aimed at reducing the number of deaths caused by pesticide poisoning.

Predictors of fatality

The diagnosis and treatment of acute poisoning are often challenging for ED staff, and some poisoning characteristics such as amount ingested [24], attempted suicide [28], pesticide exposure [28] and concomitant alcoholic consumption [46] have been found to be useful in risk stratification. However, the exposure history is sometimes hard to obtain from suicidal or critically ill patients, and information obtained at admission on illegal drug use is often incomplete or inaccurate [47]. In our study, information on the concomitant use of alcohol was only available in 62.4% of the enrolled patients. In the absence of accurate and adequate information on exposure, objective and obvious parameters for predicting fatality are valuable. Some scoring systems and biological predictors have been evaluated in previous studies for their application in risk stratification for certain herbicides. These include the Simplified Acute Physiology Score (SAPS) II [30, 31], Acute Physiology and Chronic Health Evaluation (APACHE) II [26, 30, 31, 33] and III [30, 31] scores, blood pressure, FiO₂, leukocytosis, acidosis, number of organ failures, pulmonary edema, and hyperkalemia [27, 29, 48]. Since accurate exposure history is often hard to acquire, simple and easy-to-obtain parameters that could be applied to most cases of poison exposure would be very useful in clinical settings. In this study, using multiple logistic regression analysis, we identified some easily measurable and useful parameters for predicting poisoning-related fatality, including body temperature, blood pressure, heart rate and respiration. Such objective information is routinely collected in the ED, and can help emergency physicians to quickly detect poisoned patients, predict poor outcomes, and expedite dispatch of those who need it to intensive care.

Limitations of the study

As in other similar studies, the difficulty in obtaining complete information in some cases is a limitation of our study. For example, information on alcohol consumption was provided by only 62.4% of the participants, which limited further analyses of the related issues. It should also be noted that there are limitations in generally applying some of the results of this study to other regions. Even within the same area, the pattern of poisoning may vary over time, so regularly updating epidemiology data to identify trends for specific agents and to identify risk factors is important in developing prevention strategies.