Online supplemental material:
Articles that used the PSS correctly
Walter and Persson reviewed toxic exposures called into the Swedish Poisons Information Centre in 1996 [44]. The study included all patients over 10 years of age that were exposed to a chemical during a home or leisure activity. A total of 1492 cases were included, of which follow up phone calls or a review of the medical record occurred in 1033 cases (69%). While 28% of exposures resulted in medical evaluation, only 6% of cases were hospitalized. The PSS was calculated for all patients that received follow-up. A score of 0 was assigned to 12% of the cases, a score of 1 to 78%, and a score of 2 to 7%. No patients received a score of 3 or 4, and in 3% of patients, a score could not be classified. Cleaning agents were the most common chemicals involved in moderate exposures (42%). Patients exposed to corrosives were 3.6 times more likely to have a PSS of 2 compared to other exposures. Of note, one of the two authors was the lead author of the original manuscript detailing the PSS. Most patients were only mildly poisoned as evidenced by the low number requiring evaluation or admission making us wonder if the inclusion of the PSS provided any additional information. Had the authors evaluated a population with more severe poisonings, it is possible that the impact of the PSS would have been greater.
Seok et al. retrospectively reviewed charts of patients admitted to two South Korean hospitals following an acute ingestion of chloracetanilide herbicides from 2006-2010 [40]. Patients were included if they were older than 18 years of age and had intentionally ingested the herbicide, but were excluded if they ingested an additional pesticide. Specialists in clinical toxicology reviewed medical records. The PSS was calculated for all patients without modification. In addition, symptoms were also stratified into 1 of 4 categories: gastrointestinal, central nervous system, cardiovascular, and respiratory. The study included 35 patients with a mean age of 49.8 years. Most patients were asymptomatic or mildly symptomatic (76.2%) after their exposure; one patient died. Seventeen patients had a PSS of 0 (48.6%), 10 had a score of 1 (28.6%), 5 had a score of 2 (14.3%), 2 had a score of 3 (5.7%), and 1 patient had a score of 4 (2.9%). Symptomatic patients (PSS of 1-4) were older (55.7 vs. 43.6 years, p=0.013) and had a lower calculated bicarbonate (18.8 vs. 21.6 mEq/L, p=0.017) than patients with a score of 0. While used correctly instead of attempting to correlate age with the PSS, it would be clearer and likely easier to note that few patients had significant symptomatology, and those that were older, were more likely to be symptomatic.
Kuzelová et al. reviewed the records of all patients less than 18 years old admitted to a Slovakian children’s hospital with alcohol intoxication from 1996-2005 [35]. Patients that consumed multiple substances or presented after an accidental or suicidal ingestion were excluded. The review included 537 patients with a mean age of 15.1±1.67 years and a mean blood alcohol concentration (BAC) of 1.98±0.57 g/L (198 mg/dL). The number of children hospitalized for alcohol intoxication increased each year from 1996 through 2005. The mean BAC and total number of admissions was higher in 2001-2005 than from 1996-2000. At admission, the PSS was 0 in 10 patients (1.9%), 1 in 256 patients (47.7%), 2 in 250 patients (46.5%), and 3 in 21 patients (3.9%). From 1996-2000, a higher percentage of admitted patients had a PSS of 1, while from 2001-2005 a higher percentage were hospitalized with a PSS of 2. Despite the differences in PSS score between hospitalized and non-hospitalized patients, the problem of adolescent alcohol consumption was effectively communicated by the increase in BAC between the 2 time periods and the increasing number of hospitalizations without including the PSS.
Schaper et al. performed a retrospective, cross sectional case series involving exotic venomous pet envenomations reported to 4 poison centers in Northeastern Germany and Southeastern France [39]. Pets were classified into 1 of 4 groups: snakes, aquatic animals, arthropods, and miscellaneous. The review consisted of 404 cases reported over 11 years. The average patient’s age was 36 years (range 2-75) with most being male (73%). One hundred fifty-five bites were from snakes (39%), 123 from aquatic animals (30%), 109 by arthropods (27%), and 17 cases were classified as miscellaneous (4%). There were 29 patients with severe effects (7.1%; PSS 3), 55 with moderate effects (14.2%; PSS 2), and 320 with minor effects (78.7%; PSS 1). No deaths or asymptomatic patients were reported. All 29 cases classified as severe envenomations were due to snakebites. Specific symptoms for all other patients were not reported. While the inclusion of a severity score was reasonable, stating the number of patients that developed certain symptoms (e.g. coagulopathy), were intubated, and were treated with antivenom would be more useful descriptors regarding the degree of envenomation compared to what the PSS added. This is especially true regarding the PSS as opposed to other scoring systems since extreme pain, severe thrombocytopenia, and compartment syndrome are all criteria for a PSS 3. Clearly, the clinical significance of an isolated lab value is very different compared to the development of a surgical emergency.
In a separate study, Schaper et al performed a retrospective review of all poisonous snake envenomations reported to the GIZ-Nord poison center in Germany and the Centre Antipoison at Marseille in France, over 20 years with a focus on snakebites [15]. There were 671 cases of poisonous snakebites included, 21 of which involved rattlesnakes (3.1%). However after further review, it appears other snakes such as Agkistrodons were classified as rattlesnakes by the authors. All patients envenomated by rattlesnakes were male with an average age of 37.2 years old (range 20-64). Using the PSS, 8 patients (38.1%) with rattlesnake envenomations were classified as grade 1, 5 patients (23.8%) were grade 2, 8 patients (38.1%) were grade 3; no fatalities were reported. Five patients (23.8%) received antivenom therapy. Of the patients that received antivenom, four were classified as severe envenomations and one was mild. The same limitations of the PSS that applied to the previous study by Schaper et al. also apply here.
Wong et al. reviewed all intentional overdoses of antidepressant medications reported in the Victorian Emergency Minimum Dataset (VEMD) from 1998-2007 and all cases reported in the Victorian Poisons Information Centre (VPIC) from 2005-2008 [47]. VEMD contains anonymous demographic and clinical data from patients presenting to hospitals with 24-hour EDs in Victoria, Australia while VPIC receives calls related to toxic exposures from 0700 to 2100 hours daily. Data from patients reported in both datasets was included. The PSS was used to compare symptoms for all cases included from the VPIC but not from VEMD; no explanation is given for this difference. There were 1833 exposures included from the VPIC dataset. The majority of calls were related to overdoses with selective serotonin reuptake inhibitors (SSRIs) (1016; 55.4%). Only 722 patients were symptomatic at the time of the call to VPIC (39.4%). The PSS classified 118 patients as having moderate or severe symptoms (6.4%). Thirty patients had moderate or severe symptoms following ingestion of a TCA (15.9%), while a lower percentage of patients had moderate or severe symptoms following ingestion of SSRIs (n=32; 3.2%), monamineoxidase inhibitors (n=2; 7.4%), or other antidepressants (n=47; 8.6%). No deaths were reported. It arguably would have been more useful to know how many patients developed cardiotoxicity, seizures, or respiratory depression following a TCA overdose than to use the PSS. Similar important symptoms or endpoints could have been included for other antidepressants that would be better descriptors to the reader than the PSS.
Hung et al. retrospectively reviewed 60 consecutive patients admitted to an intensive care unit in Vietnam with Bungarusmulticinctusenvenomations from 2000-2003 [31]. Most patients were male (77%) with a mean age of 33.3 ± 13.5 years (range 12-67) with most bites to the hand (53%) or foot (27%). The most common initial manifestations were pharyngeal pain (n=50), ptosis (n=56), general myalgias (n=41), and dyspnea (n=52). Using the PSS, 54 patients had a score of 3 or 4 (90%) including the 4 patients (7%) who died; 3 patients had a score of 1 (5%). Scores for the remaining 3 patients were not reported. While it was reasonable to include a severity score, describing the frequency of interventions, such as 87% of patients requiring intubation, more transparently established the severity of illness in this cohort as compared to using the patient’s PSS.
Abahussan and Ball performed a retrospective review of all poisonings in patients under 12 from January 2004-December 2005 at six hospitals in Kuwait [20]. Nine hundred seventy eight patients were included. A PSS of 0 was assigned to 578 patients (59.2%), a PSS of 1 to 364 patients (37.3%), and a PSS of 2 or 3 to 34 patients (3.5%). It was not differentiated how many were in each of these two grading classes. No patients were classified as PSS 4 and no fatalities were reported. Given how the PSS was incorporated, it would have been easier to state that over 95% of patients developed minimal or no symptoms.
Caldas et al. reviewed 709 pesticide exposures reported to the toxicological information center between 2004 and 2007 [23]. The majority of cases were telephone consultations (62.8%), with the rest coming from hospital epidemiology units (37.2%). There were 51 different compounds identified as pesticides involved in these cases, generally divided into the following categories: carbamate/organophosphate, coumarin, pyrethroids, others, and ignored (cases where information was not available from either the poison center or the hospital chart). In total, 162 patients (22.8%) were a PSS of 0, 342 patients (48.2%) were a PSS of 1, 97 patients (13.7%) were a PSS of 2, 90 patients (12.7%) were a PSS of 3, and 18 patients (2.5%) were a PSS of 4. Each category of patients was also stratified by the PSS. Most patients with a PSS of 3 or 4 were exposed to either carbamates or organophosphates. However, instead of using the PSS to define the severity of illness, it would be expected to be simpler and more informative to the reader to use muscarinic signs and symptoms or bleeding in patients exposed to carbamates/organophosphates or coumarins, respectively. This would have provided more clinical insight to the patient’s condition compared to using the PSS. Some intoxications or envenomations in general may have hallmark clinical signs which signify the degree of toxicity or progression of illness much more appropriately than an overall PSS does.
Deguigne et al. reviewed all calls about metam sodium received by the Poisons and Toxicovigilance Centre at Angers University Hospital in France between 1992 and 2009 [26]. There were 102 cases included in this retrospective study, all of which were unintentional exposures. The severity of the case presentation was determined with the PSS. All scores were determined from the information obtained over the phone. While the authors do not give exact figures, they state that 99% of cases (101 patients) were “low or null” likely indicating a PSS of 0 or 1 and that there were no cases of moderate exposure according to the PSS. One death was seen in a patient who fell into an empty container that had contained metam sodium, which the authors state was “severe, sufficient to be fatal” which is confusing as severe would be a PSS of 3, while death is a PSS of 4. The authors did not state which score the patient received. Given how the authors worded the severity score outcomes, it would have been clearer to state that nearly all patients either were asymptomatic or developed minimal symptoms.
Jose et al. prospectively evaluated factors associated with childhood poisonings in South India [33]. The study took place at a tertiary care hospital between 2007-2011 and included 121 children 18 years old or younger. Hospitalizations related to envenomations were excluded. Patients with scores of > 2 were managed in the ICU. The type of poisoning, age, and PSS were compared using a chi-square test. The majority of patients were female (n=73; 60.3%). The most common toxic agents were: yellow cow dung powder (n=31, 25.6%), drugs (pharmaceuticals) (n=26, 20.5%), and organophosphates (n=18, 15%). PSS grading was classified as 0 for 24 patients, 1 for 67 patients, 2 for 17 patients, 3 for 10 patients, and 4 for 3 patients. Of patients with low PSS (0-1), 67 cases (73.6%) were suicide attempts. Of patients with high PSS (2-4), 19 cases (63.3%) were suicide attempts. The difference was not statistically significant (p=0.281). The authors did not comment on the accuracy of the PSS. While stratification with the PSS was successfully used to triage patients, a shorter, and less subjective list of signs or symptoms could more simply and rapidly triage patients to the floor verse the ICU.
Sabzghabaee et al. studied the relationship between admission blood glucose concentration and symptom severity and outcome following poisonings [38]. The authors prospectively enrolled patients admitted following a deliberate self-poisoning to a hospital in Iran. Patients with diabetes, who overdosed on xenobiotics known to cause either hyperglycemia (blood glucose > 126 md/dL) or hypoglycemia (blood glucose < 70 mg/dL), or that received dextrose or glucocorticoids prior to arrival were excluded. The PSS was scored at both the time of arrival and following the patient’s recovery. From March 2005 to May 2005, 345 patients were enrolled. The majority of patients were female (57.9%) with an average age of 27.58.6 years (range 18-65). The PSS was 0 for 169 patients (50%), 1 for 78 patients (22.6%), 2 for 58 patients (16.8%), 3 for 38 (11%), and 4 for 2 patients (0.6%). Higher PSSs were correlated with complications including respiratory failure, renal failure, liver failure and death (r2=0.371; p<0.001). Hyperglycemic patients were more likely to have a PSS of 3 or 4 than normoglycemic patients (22% and 2% compared to 4% and 0%, respectively). Normoglycemic patients were more likely to be asymptomatic (PSS 0) than hyperglycemic patients, 61% vs 39%, respectively. As opposed to correlating hyper- or hypoglycemia with PSS, it would be clearer to directly associate blood glucose concentrations with less nebulous outcomes such as respiratory failure, metabolic derangement, and death.
Turedi et al. investigated the use of ischemia-modified albumin (IMA) in patients with carbon monoxide (CO) poisoning [41]. They conducted a prospective, multicenter study in Turkey, including all patients who were admitted to the ED over a 6-month period with CO poisoning (COHb >2%). Any patient with co-existing ischemia from another process was excluded. The authors enrolled 37 patients with CO poisoning. Study patients were compared to 49 age and sex-matched healthy control patients. Study patients had a mean age of 34 years and 15 (45%) were male. Mean IMA concentrations were higher in the study group but did not correlate with COHb. No patients were given a PSS of 0 or 4; 22 patients received a score of 1 (66.7%), 10 patients a score of 2 (30.3%), and 1 patient had a score of 3 (3%). Patients with a PSS of 3 had higher CO levels than those with a PSS of 1 (p<0.05) and a PSS of 2 (p<0.03). While CO levels correlated with PSS, higher IMA concentrations were not associated with higher PSS. As the purpose of this investigation was to determine if IMA increased with CO poisoning and the clinical relevance of IMA, it is disappointing that the PSS did not correlate with IMA concentrations even though it did with CO levels.
Williams et al. prospectively studied consecutive inquiries to the UK National Poisons Information Service regarding liquid detergent capsule exposures reported by health care workers from March 2008-April 2009 [46]. If possible, each inquiry was followed up with a phone call within 4 hours. There were 647 enquiries; the product was known in 604/647 cases (93.4%). Most enquiries concerned children less than 5 (96.1%) and were unintentional (99.7%). PSS was recorded for 637/647 (98%) patients. The PSS was 0 in 382 cases (60%), 1 in 243 cases (38%), 2 in 10 cases (2%), and 3 in 2 cases (<1%). No patients died. Specifically describing the symptoms that patients experienced after capsule exposure, especially those of the 2 sickest patients (which were included), were more useful than an isolated PSS. A scoring system designed with prognostic abilities based on early findings would be more useful in this type of poisoning.
Adams et al examined the concordance of severity grading between specialists in poison information (SPI) and healthcare providers taking care of the patient [21]. Over three years in Edinburgh, United Kingdom, all accidental pesticide exposures in the National Poisons Information Service Edinburgh Unit were reviewed. Questionnaires were sent to the providers asking “how serious was the exposure considered at the time of admission/assessment?” The providers graded the case based on five categories: not at all serious, minor, moderate, major, or unknown. The providers only completed the questionnaires while the SPIs only completed the PSS. SPIs graded the cases using the PSS with the addition of the category uncertain. In order to equate the 2 scoring systems, patients with a PSS of 3 or 4 equated to major toxicity on the questionnaire. In the study time frame, 1567 questionnaires were completed with 1284 cases included in the analysis. Cases marked uncertain or missing information were excluded. SPIs grading of the severity was lower than that of the providers, with the difference being more pronounced for pediatric exposures. Providers graded 690 patients (93%) of a total 745 pediatric exposures as level 1 or 2, while SPIs graded 738 patients (99%) at the same level. Providers scored 448 patients (78%) out of 539 total adult cases as scores 1 or 2, while SPIs graded 509 patients (94%) the same level. Given all the modifications that were required, it may have been easier to have both groups grade the severity of symptoms based on the questionnaire and not use the PSS.
Studies Comparing the Poisoning Severity Score to another Severity Score or Using the Poisoning Severity Score Uniquely
Hrabetz et al. retrospectively reviewed a cohort of patients in Germany that presented following an ingestion of an organophosphate [12]. All patients admitted to the ICU over a 12 year period that required either atropine or obidoxime were included. Ten patients were excluded due to missing data or not requiring ICU admission or an antidote leaving 33 for inclusion. The authors used the PSS, GCS on admission, QT interval, APACHE-II, and SOFA scores to evaluate the severity of poisoning. The median age was 56 (range 19-84) and patients were predominately male (n=25;75%). Median scores were: PSS 3 (range 0-4), APACHE-II 19 (range 2-34), SOFA 7 (range 1-19), and GCS 9.5 (range 3-15.). Nine patients (27%) had prolonged QT intervals on arrival, defined as greater than 440 ms in men and 450 ms in women. The authors do not include the breakdown of PSS grades. Five patients (15%) died in this study and tended to be older; have worse prognostic scores on the PSS (all grade 4), APACHE-II (median 24, range 17-34), SOFA scores (median 10, range 7-19), GCS (3); have prolonged QT intervals (497 ms median; range 440-520); require prolonged mechanical ventilation; and require longer ICU admissions. Twenty patients (61%) made a full recovery, while 8 patients (24%) required neurologic rehabilitation at discharge. Patients that died had higher PSS than those that survived (PSS 4 vs PSS 3; p<0.01) and higher SOFA scores (10 vs 7; p=0.02) than those that survived. Of course by definition, only patients that died could have a score of 4 and so their score should be higher. Unfortunately, the severity scores were not compared to one another in regards to their accuracy or prognostic ability, which would have been extremely helpful in determining the clinical value of the PSS.