METHODS

Study protocol

A panel of five physicians (A, B, C, D, E), with more than 10 years of experience in emergency medicine, was appointed to give an expert opinion on appropriateness of patient management, after reviewing the records of 142 patients who attended the ED of Forl, Italy, for suspected intentional drug overdose. All patients had been submitted to comprehensive drug screening. The experts blindly received two datasheets for each case. One datasheet omitted the results of the comprehensive drug screening (see below), while the second, which included drug screening results, was slightly modified in terms of minor data concerning history, to make the paired datasheets unidentifiable as pertaining to the same patient. The sequence of cases was also randomly determined to exclude any identification.

The experts were asked to complete a questionnaire regarding the appropriateness of patients' disposal (discharge after three to six hours of observation, general wards admission, HDU/ICU observation).⁷ Before revision, experts were also informed on the criteria for evaluating the appropriateness of the different steps of care, aiming at a consensus based on universally accepted indication for main variables considered.¹ To ensure independence, all experts were selected outside the area of Forl. The local ethics committee of Morgagni Hospital approved the protocol.

Database

One author (AF) reviewed the clinical records of all consecutive adult subjects, who attended the ED of the community hospital of Forl (50 000 visits per year) for suspected intentional drug overdose from December 1996 to March 2000, and submitted to a comprehensive drug screening (table 1).

Abstracted information included personal data, poisoning to admission time, type of suspected drug(s), comorbidity, signs and/or symptoms of clinical presentation, observation period, and clinical course. Cases with Glasgow Coma Score (GCS) <9 at time of disposal, and cases in whom lifesaving support was mandatory independently of laboratory results were excluded, as were patients with acute concomitant illnesses, suspected overdose by drugs of abuse or alcohol intoxication, unintentional drug poisoning, traumatic injuries, and chemical poisoning. Finally, records with insufficient data reporting and patients voluntarily discharged from ED were also excluded. According to these criteria, the clinical records of 142 patients were selected (table 1). These patients account for about 30% of all cases admitted for suspected drug poisoning in our institution, after exclusion of chemicals and carbon monoxide intoxication. In 82 cases (57.7%) the diagnosis was acute overdose on the basis of plasma or urine concentrations of drugs or their metabolites in amounts sufficient to explain the presenting symptoms. In 55 cases (38.7%) a single drug was responsible for acute intoxication, in the remaining 27 (19.0%) two or more drugs were found. The drugs more commonly involved were benzodiazepines (BDZ, 58 cases), neuroleptics (17 cases), other hypnotics (4 cases), tricyclic antidepressants (TCA, 13 cases), serotonine specific reuptake inhibitors (SSRI, 8 cases), non-tricyclic antidepressant (12 cases), anticonvulsants (12 cases), digoxin (2 cases), drugs of abuse (8 cases), analgesics (4 cases).

In 43 cases (30.3%), one or more drugs were detected, at trace levels. The most common drugs in this group were BDZ in 30 cases, neuroleptics (7 cases), TCA (5 cases), anticonvulsants (10 cases), antiparkinsonians (4 cases). In 17 patients (12.0%) no drug was detected at urinary Remedi screening (see below). Blood alcohol concentration >100 mg/dl was present in 34 cases (27 with drug overdose).

Gastrointestinal decontamination procedures were started in 59 cases soon after arrival at the emergency department, and this information was reported in the data sheet.

Study design

The drug screening had been performed in each case by means of an integrated system, combining a multi-column HPLC drug profiling system (Remedi HS Bio-Rad, Hercules, CA) and a fluorescence polarisation immunoassay (FPIA) with a TDx analyser (Abbott, Abbott Park, IL). In sequence, the protocol included the Remedi urine drug screening in all cases and, in case of drug detection, a confirmatory test with either Remedi in plasma or a quantitative FPIA, whenever available.^8–10 FPIA was carried out in plasma for drugs and in urine samples for drugs of misuse (opioids, cannabis, amphetamines, cocaine).

A combined analysis with Remedi in urine samples and plasma was performed in 37% of cases, whereas an additional quantitative confirmatory test with a single FPIA in plasma was necessary in 35% after the urinary Remedi screening.

The current library of Remedi HS multidrug profiling system permits the determination of over 900 common pharmaceutical agents and their metabolites, with a detection limit of 200 ng/ml.⁹ This policy guarantees a final result in a turnaround time of 20–60 minutes. The system does not measure lithium, sulphonylureas, and most chemical agents.

The FPIA system permits the quantitative determination of plasma BDZ, digoxin, barbiturates, phenytoin, methotrexate, primidone, salicylates, paracetamol, theophylline, TCA, valproate, ethosuximide and the urinary detection of the commonest drugs of misuse (opioids, cannabis, cocaine, amphetamines). Blood alcohol concentration was measured by ALC, Dade Behring, Newark, NJ.

Data analysis

Median or mean values, standard deviations (SD), and frequencies were used to describe data distribution. For each outcome variable, the intraobserver agreement (reproducibility) between paired datasheets with/without the results of the drug screening was calculated using the chance corrected agreement index (κ).¹¹ After this, the interobserver agreement was also calculated using the same technique. A κ value of 0 indicates agreement no better than chance, and a value of 1 indicates perfect agreement. The degree of agreement was graded as follows: κ>0.75=excellent agreement, κ between 0.40 and 0.75=fair to good agreement, and κ<0.40=poor agreement.¹² Student's t test using standard error of κ estimates was used to determine the level of statistical significance of comparisons among κ values. Differences were considered significant when the two tailed p value was <0.05. Statistical analyses were performed running the SPSS/PC+ statistical package on a personal computer.¹³

RESULTS

Intraobserver reproducibility

The availability of drug screening significantly changed the physicians' opinion regarding patients' disposal. The overall reproducibility of the five experts when considering the same record with/without drug screening information was poor (0.233), as were the individual κ values for discharge after short observation (0.249), admission to general wards (0.185), and admission to HDU/ICU (0.290). Expert A and E had a significant non-homogeneous approach for disposal, expert A having a larger reproducibility on HDU/ICU admission, and expert E having a lower reproducibility on HDU/ICU admission and discharge (table 2).

The analysis of the total 1420 records compiled by the five experts showed that the availability of drug screening would have saved a total of 135 of 621 hospital admissions (−21.7%; p<0.001), by increasing the number of patients discharged after a brief observation from 89 to 224 (p<0.001). In particular, the screening would have reduced the number of HDU/ICU admissions from 184 to 86 (−53.3%; p<0.001), and the total number of general ward admissions from 437 to 400 (−8.5%; p=0.136) (table 3). About 20% of patients were moved from HDU/ICU to general wards, and from general wards to discharge by all experts (fig 1).

Interobserver agreement

In the absence of drug screening, a very poor agreement was observed among physicians for disposal (κ=0.238). The results of drug screening significantly (p<0.001) improved the interobserver agreement to the fair to good range (κ=0.461) The agreement also increased when admission to an intensive care unit, to a general ward, and discharge from hospital were separately analysed (table 4).

DISCUSSION

The results of this study suggest that comprehensive drug screening affects clinical decision making for patients attending the ED for suspected intentional drug overdose in terms of use of hospital resources.

The usefulness of a multidrug analysis in the management of patients with acute drug poisoning has not been resolved. Several authors claim that comprehensive drug screenings are useful for the treatment of their patients, influencing their management, but most studies do not provide quantitative evidence to support this conclusion.^{2,8,10,14–16} Earlier studies have suggested that unexpected laboratory findings scarcely influence management or outcome of patients with drug overdose.^5,6,17 Similarly, laboratory evidence of a clinically unsuspected drug does not seem to serve as a useful marker for severity of illness and does not, in itself, identify patients requiring closer observation or having poorer prognosis.¹⁸ One possible reason for benefits of laboratory results may be that they reassure the physician. The implicit psychological benefits, which are not easily quantified, seem to be inversely correlated with physician's experience and confidence.¹⁹ On the other hand, other studies concluded that laboratory results have only a limited role in the management of the acutely poisoned patient, the clinical course remaining the leading criterion.^3,4

How can a screening test demonstrate its usefulness? We have a reliable comprehensive drug identification system, testing over 900 drugs in blood and urine in our unit.⁹ In the absence of any unequivocal gold standard, we only simulated in a table top comparison the clinical position of an emergency physician when dealing with a patient with suspected drug overdose, and evaluated if the screening influenced decision making, independently of clinical data.

The decision on patients' disposal may be critical in borderline cases. The clinical course of patients with suspected drug overdose can be predicted during the first few hours of observation, sufficient to identify low risk patients.¹⁹ However, without laboratory results, physicians are prone to overestimate the likelihood of late deterioration after drug overdose, and may be reluctant to take risks whenever a minimum of uncertainty exists, leading to over-treatment and inappropriate disposal. This is perceived as a less serious error than not admitting a patient who could benefit by continuing hospital care. However, the shortage of hospital resources implies that any attempt to reduce inappropriate clinical care must be pursued.

Our simulated study demonstrates that this drug screening may be relevant for the medical management of patients with suspected drug overdose, improving the physicians' agreement on the use of resources. Our experts were well trained in emergency medicine, but not specifically trained in medical toxicology. Therefore, our data may be extrapolated to the majority of emergency departments, where physicians have only a basic knowledge of clinical toxicology.

Our protocol has obvious limitations. All relevant information was included in paired data sheets submitted to external experts, but visiting a patient is different from revising a clinical record. Physicians do not always behave the way they say they would have behaved, but in paired analysis this is not expected to bias the final results, unless relevant data were omitted. It is also conceivable that the non-homogeneous experience of experts with our comprehensive drug screening may have led them to overestimate its significance. Finally, a discussion between physicians and analysts is frequently needed to correctly interpret the results of drug screening. In the hands of non-expert physicians, comprehensive screening even detecting minimum amounts of drugs might lead to an overestimate of the results, as the detection of the presence of drugs does not necessarily imply causation. Apparently, in the present series the potential bias for “false positive” results ended up in reducing the use of resources. This implies that the busy physician is more ready to accept the “true negative” data than the “false positive” results.

We now need to know whether this result may be reproduced in the clinical setting, improving the final outcome of patients admitted to ED for suspected intentional drug overdose, and whether this policy is cost effective. The cost of each test varies between 120 and 150 depending on the need for and number of confirmatory tests. This compares favourably with the potential reduction hospital admission, but the appropriateness of this policy on final outcome needs to be demonstrated. Our results are the basis for prospective, controlled studies on indications, limits, usefulness, and cost effectiveness of the comprehensive drug screening.