Objective. Adverse drug reactions (ADRs) occur frequently in children. However, the exact incidence of ADRs is unknown. Therefore, we studied ADRs in 1 ward and assessed whether a general approach, eg, by a computerized monitoring system, to detect ADRs in children is feasible and likely to yield a higher rate of early detected ADRs. The aim was to assess the usefulness of a computerized monitoring system before implementing costly adaptations.
Methods. An 8-month prospective study was conducted at a 10-bed pediatric isolation ward of the University Hospital. Charts were reviewed once weekly by a pharmacoepidemiological team. Clinical signs as well as laboratory changes were documented and assessed. Algorithms were used to assess the probability and severity of each detected event.
Results. All 214 patients admitted were enrolled in the study. A total of 68 ADRs were detected in 46 of 214 patients by the pharmacoepidemiological team. Thirty-four ADRs (50%) were detected by the staff physician, and 27 (40%) were detected primarily by analyzing laboratory parameters. Antibiotics-associated ADRs (50%) predominated, followed by glucocorticoids (16%), tuberculostatic (4%), and immunosuppressive agents (4%). In 5 cases, an ADR was responsible for the prolongation of hospital stay, and in 4 children, the ADR was responsible for hospitalization.
Conclusions. The detection rate of ADRs would almost be doubled by a computerized monitoring system analyzing laboratory data. Implementation of a computer monitor system that automatically generates laboratory signals may help to identify ADRs in children, and to reduce morbidity and hospital stay, as well as costs.
Adverse drug reactions (ADRs) may cause a variety of symptoms and changes in laboratory values. The incidence of ADRs in hospitalized patients ranges from 1.5% to 35%,1–4 depending on the different definitions of ADRs, the methods used, and the vigor with which ADRs are sought, as well as the number of drugs administered simultaneously leading to drug interactions.5,6 Fatal ADR may occur in 0.32% of hospitalized adult patients.7 In adults, hospital admission as a result of ADR range between 1% and 8.8%.4,8,9
Data on the efficacy, tolerability of the drugs, and particularly information about ADRs in children are often lacking, in part because drug administration authorities and pharmaceutical industry have ignored routine drug evaluation in pediatric patients. In a systematic review and meta-analysis of 17 prospective studies, the incidence of ADRs in children ranged between 4,37% and 16,78% among the studies with severe ADRs occurring in 7% to 20% of ADR positive cases.10 Another issue is that many drugs used for the treatment of children are either not licensed for the use in children (unlicensed) or are prescribed outside the terms of the product license (off-label).11,12 Safety data on these drugs is needed.
Therefore, we systematically studied ADRs on a ward for infectious disease to assess whether the general approach of detecting ADRs in children by a computer-based monitoring system is feasible and clinically relevant. Before adapting the computerized monitoring system developed for adults for the use in a pediatric clinic, we simulated this system on the study ward. One goal is to provide the physicians with a practical tool for the daily rounds and thus increase drug safety. Furthermore, we hope that collecting data with the computerized monitoring system and subsequent analysis will enable us to supply specific safety information regarding the use of certain drugs in children.
All patients admitted to the 10-bed pediatric isolation ward of a University Hospital were studied prospectively during an 8-month period. The charts of all 214 patients were reviewed weekly by a pharmacological team consisting of a clinical pharmacologist, a pharmacist, and a pediatrician. In addition, the staff physicians and nurses were asked to report all events. The patients were admitted to the ward either for infectious diseases or immunodeficiency. We documented each of the 214 patients age, weight, the reason for admission, all known diagnoses, as well as all administered drugs with dosage information. In addition to clinical data, laboratory findings were integrated in the documentation system. For each irregularity suspected of being an ADR, the probability was assessed. The study was approved by the institutional ethic review board.
ADR definitions vary in the literature. We used the definition of the World Health Organization (WHO) in our study. The WHO defines an ADR as “an effect which is noxious and unintended, and which occurs at doses used in man for prophylaxis, diagnosis and therapy.” An adapted Naranjo13 algorithm score was used to assess the probability of each event, and the severity of the suspected ADR was scored using a special severity score.14 In addition, events were classified as predictable or unpredictable. Predictable ADRs may either be avoidable or tolerated, implying such events as toxicity, drug interactions, and secondary effects. Unpredictable and usually unavoidable ADR include idiosyncratic or allergic reactions as well as intolerance. Furthermore, the mechanism of the ADR was documented. If the duration of hospitalization was prolonged because of an ADR it was documented separately.
The 214 patients received a total of 1032 drugs, eg, 4.8 drugs per patient. The age of patients ranged between 1 month and 35 years with a median of 8,6 years. They were hospitalized for 1 to 187 days with a median of 6,9 days.
Altogether, 68 ADRs were detected in 46 patients, corresponding to 21.5% of the documented patients. Eight patients developed 2 ADRs during their hospital stay, 4 patients 3 ADRs, and 2 patients 4 ADRs. The detection of 27 ADRs (40%) was based solely on laboratory findings, and 34 ADRs (50%) were detected by the staff physicians. The overlap was only 3 ADRs between these 2 groups, thus staff physicians aided by the computer monitoring system could have detected up to 58 (85%) of 68 ADRs based on clinical signs, eg, flush, rash, diarrhea, and abnormal laboratory findings.
Seven ADRs (10%) were classified as severe with a severity score of 8 points or greater. Of these 7 ADRs, 6 (88%) were detected by the staff physician. A significant elevation of liver enzymes (γ-glutamyltransferase, alkaline phosphatase) secondary to treatment with fosfomycin in a child with cystic fibrosis was missed. Of the other ADRs, 28 (41%) were classified as moderate and 33 (49%) as mild.
According to the adapted Naranjo algorithm score, 3 (4%) ADRs classified as “possible ADRs” with a probability score of 1 to 4, and 53 (78%) events were defined as probable with a probability score of 5 to 8. The remaining 12 ADRs (18%) were classified as very probable to definite ADRs (Fig 1).
A patient with tuberculosis meningitis suffering from extreme vertigo secondary to long-term streptomycin therapy despite clinical improvement of the tuberculosis can serve as an example for the latter, as well as a patient with cystic fibrosis with an anaphylactic reaction during a repeated meropenem infusion.
With the exception of 1 moderate ADR, all ADRs classified as “possible” were judged to be mild. No severe ADRs occurred in the group of “possible” events. Of the 53 “probable” ADRs, 51% were categorized as mild, 41% as moderate, and 8% (4 ADRs) were regarded to be severe. In the group of “very probable to definite” ADRs, the proportion of severe events increased to 25%; 33% of the events were classified to be of mild and 42% of moderate severity.
Of the 68 ADRs, 16 (24%) ADRs were judged to be preventable, 20 (29%) to be unavoidable, and the majority, 32 events (47%), to be tolerable. In 5 cases, ADRs were responsible for the prolongation of the hospital stay, and 4 patients were hospitalized because of an ADR significantly increasing health care expenditures in all cases.
Regarding the underlying pathomechanisms of the ADRs, immunologic mechanisms induced 14 ADRs, 27 ADRs were based on secondary effects, 20 ADRs on toxic effects, and 2 ADRs on drug interactions. Five cases could not be classified (Fig 2).
Not surprisingly for a ward with infectious disease or immunocompromised patients, most ADRs were antibiotic-associated (50%; Fig 3, Table 1). Antibiotic-associated ADRs predominated, followed by glucocorticoids, tuberculostatic, and immunosuppressive agents (Fig 3).
This study illustrates the magnitude of the problem of ADRs in this selected pediatric patient population with the finding that ADRs occurred in 21,5% of the patients. The incidences ranged from 15% to 27%,15 including 6% life-threatening ADRs15 in older studies in the United States and Canada. However, we believed that the incidence of ADRs detected in our study has been relatively high as more recent prospective studies on ADRs or adverse drug events (ADEs) in pediatric patients report incidences between 4,37% and 16,78%.10,12,16,17
The incidence of ADRs detected on this pediatric ward was influenced by several factors. First, the number of ADRs detected was influenced by the prolonged hospital stay, the classes of drugs used, and the polypharmacy. Second, chart review, in addition to questioning of the medical staff, was used as the main method of detecting ADRs by the pharmacoepidemiological team. This method has been used as the gold standard as it detects a higher percentage of ADRs than all other methods.
Finally, the definition of ADR has influenced the type and percentage of events detected. In contrast to the definition of ADE as “an injury resulting from medical intervention related to a drug,”1 we explicitly chose to use the WHO definition for ADR as it was our aim to study drug safety issues during the regular and appropriate use of drugs in children. Despite these differences in definitions between ADR and ADE, the unequivocal classification of events is not always easy especially in pediatrics because of the variances of dosages recommendations or lack of information on age-appropriate dosing.
We verified with our study that a large percentage of ADRs could be detected by a computer monitoring system. A total of 27 of 68 ADRs could have been identified by a computer monitoring system, of these only 3 ADRs were recognized by the attending physician and contributed to the 34 ADRs detected by the staff. Thus, the use of the computer monitoring system could result in a dual effect. It would significantly increase the percentage of correctly identified ADRs as well as stimulate the staff physicians to be aware of the occurrence of ADRs.
However, not all ADRs detected by this approach are of immediate clinical significance. As an example many physicians may not regard eosinophilia or leucocytosis as a relevant ADR, although they fulfill the WHO criteria. And they may be useful in the prevention of more severe ADRs as eosinophilia may precede allergic reactions and leucocytosis may simulate infectious or immunologic diseases.
In 4 (8.7%) of 46 patients, the ADR was the cause of hospital admission. Similar figures were reported by Easton and coworkers.18
As we could demonstrate the feasibility of using a computer monitoring system with automatically generated laboratory signals for pediatric patients, we are now implementing an ADR computer monitoring system. The algorithms for the detection of ADRs were developed on the basis of our findings from this preliminary project. The algorithms induce the generation of signals whenever laboratory findings vary unproportionally in time and value. In a second step, signals are evaluated on their appropriateness of identifying ADRs. Additional refinement of the algorithms is warranted to increase the sensitivity and specificity of the generated signals.
The increase in awareness for ADRs and, thus, the early detection of ADRs with the help of a computer monitoring system may improve the patients quality of life and decrease health care expenditures. Similar efforts are underway in adult patient populations.1,3,19–23 It is our goal to have an impact on drug safety in children and contribute to pharmaco-vigilance.
The support of this work by the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie (BMBF) under project number 01EC98030 is greatly acknowledged.
- Received August 14, 2001.
- Accepted February 26, 2002.
- Reprint requests to (W.R.) Klinik für Kinder und Jugendliche, Friedrich-Alexander Universität Erlangen-Nürnberg, Loschgestr. 15, 91054 Erlangen. E-mail:
Dr Weiss and Dr Krebs contributed equally to this study.
- Copyright © 2002 by the American Academy of Pediatrics