Background

Pharmacotherapy is an important but under-evaluated element of critical care medicine.[1–3] Medications are employed to treat primary disease, to facilitate initiation of mechanical organ support (most commonly respiratory, cardio-circulatory, renal, or hepatic), and to enable the continued use of these therapies. Medications are also used to counteract unwanted or excessive effects of other medications. In contrast to descriptions of more singular intensive care unit (ICU) therapies, such as mechanical ventilation and extracorporeal membrane oxygenator therapy, for which large-scale epidemiologic studies have been performed,[4–7] descriptions of the extent of pharmacotherapy in critical illness are limited.[8–12]

Critically ill children receive more medications than other hospitalized children,[13–16] but there is little information quantifying drug orders written, numbers of different medications ordered, and drug administration. In order to better understand the nature of pharmacotherapy in pediatric critical illness, we prospectively evaluated pharmacotherapy in a cohort of critically ill children admitted to a university-affiliated pediatric ICU.

Materials and Methods

A prospective, observational study was performed. We quantified the extent of pharmacotherapy in pediatric critical illness by evaluating the numbers and composition of drug orders, the number and types of different medications ordered, and the number of drug administrations. Consecutive eligible patients were prospectively identified during daily review by an investigator who was not involved in the clinical care of the selected patient population.

Eligible patients were admitted to the 16-bed pediatric ICU (PICU) of a 300-bed university-affiliated pediatric hospital. This was a closed PICU, with approximately 1100 patient admissions per annum. Drug orders were hand-written onto the orders sheet, and carbon-copies of medication orders could be detached as required to facilitate remote dispensing by pharmacy staff. Verbal ordering was permitted, with appropriate indication (urgency), and required timely signature. Medication administration was recorded on a paper medication administration record.

Eligible patients were admitted to the PICU for >24 hours and had not been studied previously. We included only the initial, index admission in order to reduce over-representation of re-admitted patients, and to simplify the analysis by avoiding repeated measurements from the same individuals. Short-stay patients (elective surgical, diabetic ketoacidosis, and drug ingestion) were excluded as these patients were in the ICU for 2 incomplete ICU days. We excluded these patients to avoid producing an under-estimate of the number of drug orders and administrations per full ICU day. Our focus was on the sicker patients who received mechanical ventilation and other ICU therapies.[17]

The primary outcomes were drug orders written, different medications ordered, and medication administrations. A drug order was defined as documentation on the orders sheet, ordering either the commencement or discontinuation of a medication, or modifying an existing drug order. For the purposes of this study, medications were defined as including all drugs and any electrolytes ordered for delivery by infusion. We excluded orders for maintenance fluids, electrolytes added to maintenance fluids, total parenteral and enteral nutrition, blood component transfusions, unfractionated heparin (1 U/mL) used to maintain the patency of intravascular lines, and sodium chloride (0.9%) used for tracheal suctioning. Medication administrations were defined as the administration of an intermittent dose of an eligible medication that was documented in the medication administration record.

Data Collection

Data were abstracted from electronic and handwritten medical records. Data describing organ function and ICU therapies were abstracted from an electronic documentation system (Eclipsys, Atlanta, GA, USA). Drug orders, drug administration, and other clinical data were abstracted from handwritten sources. The patient’s age, pre-existing morbidities, admission diagnosis, and the number of regular medications being taken before hospital admission were documented at the time of admission.

Each patient’s ICU admission was divided into 24-hour periods ending at midnight. Within each period we abstracted drug orders, use of mechanical ventilation, continuous renal replacement therapy, and use of inotrope or vasopressor therapy. For each drug order we documented the unique medication, its therapeutic class (sedative/analgesia, antimicrobial, bronchodilator, corticosteroid, diuretic-electrolyte, muscle relaxant), route of administration, and the frequency of administration (intermittent bolus or continuous infusion).

The Pediatric Risk of Mortality III (PRISM III) score[18] and the Pediatric Logistic Organ Dysfunction (PELOD) score[19] were calculated as measures of severity of illness. The PRISM III score was calculated on admission, while the PELOD score was calculated for the entire ICU stay. The number of organ systems with dysfunction were determined using the PELOD criteria.[19]

At PICU discharge, the number of times each medication was administered during the PICU admission was abstracted from the drug administration sheets and PICU survival was recorded. Data were collected for the duration of PICU admission, or until 3 weeks after discharge of the last patient enrolled.

Data and Analysis

The median and interquartile ranges (IQR) of the number of orders, medications, and administrations were calculated and the data are presented graphically. The number of drug orders (in total), drug orders by therapeutic class, different medications ordered, and drug administrations were compared between groups with and without pre-hospital admission medication use by reason for PICU admission (acute respiratory illness vs other), use of mechanical ventilation, use of inotropic or vasopressor medications, and use of haemodialysis, using the Wilcoxon rank sum test. The number of drug orders, different medications ordered, and drug administrations were compared with the number of dysfunctional organ systems using analysis of variance (ANOVA). We tabulated the types of drug orders across the ICU admission (day of admission [day 0]; first full day in ICU [day 1]; last day of admission [day of discharge], excluding patients for whom this was also day 1; and all other days in the ICU), and tabulated the therapeutic class against the use of three different ICU therapies (mechanical ventilation, inotropes or vasopressor medications, and dialysis) and for children with multiple organ dysfunction.

Multiple regression analyses were then used to identify factors independently associated with the number of drug orders written, the number of different medications ordered, and the number of drug administrations throughout the PICU admission. A stepwise backwards approach was used where the least significant variable was removed from the regression model until only variables significant at the p = 0.05 level remained. The initial model for each regression consisted of nine variables: the reason for admission (respiratory vs other reason), patient age, regular medication used before ICU admission (yes/no), PRISM III score, PELOD score, the number of days intubated, the number of days receiving haemodialysis, the number of days receiving inotropes while in the ICU, and the ICU length of stay. The r2 (correlation coefficient) was reported as a measure of the proportion of variability explained by the variables remaining in the model.

Data were entered into a custom-made Microsoft® Access (Microsoft Corporation, Redmond, CA, USA) database. The completeness of data entry was checked by comparing the number of ICU days of data entered into the database with a manual count of the number of days abstracted into the paper data collection form, and by independent validation of all data from 10% of the patients. Analyses were performed using SAS version 9.2 (SAS, Cary, NC, USA). Prospective approval for the study was obtained from the authors’ institution’s Research Ethics Board. The need for patient consent was waived.

Results

There were 362 patients admitted to the PICU during the 20-week period ending 10 June 2005. After excluding 9 patients by diagnosis, 206 post-surgical admissions of <24 hours, 17 re-admitted patients, and 30 missed patients, we evaluated 100 emergency admissions who were admitted for a total of 851 patient-days (table I). The median age of patients studied was 40 months (IQR 9–82). The median PRISM III score was 5 (IQR 2–9). Fifty one (51%) patients were admitted to the ICU with an acute respiratory illness and 50 (50%) patients were treated at home with regular medications prior to hospital admission (median of three medications [IQR 2–4]).

Table I
figure Tab1

Patient characteristics and number of drugs, drug orders, and drug administrations during intensive care unit (ICU) stay for 100 children admitted to a quaternary pediatric ICU for >24 hours

Full size image

The median PELOD score was 12 (IQR 10–22) and the median daily PELOD score was 2 (IQR 1–11). There were 544 (64%) ventilator days, 34 (4%) haemodialysis days, and 83 (10%) days on which inotrope or vasopressor therapy was used in 78, 4, and 20 patients, respectively (table II).

Table II
figure Tab2

Comparison by patient type of drug orders written, different medications ordered, and drug administrations during intensive care unit (ICU) stay for 100 children admitted to a quaternary pediatric ICU for >24 hours

Full size image

Ninety-five (95%) patients survived to PICU discharge. There were four non-survivors, and one patient who had been admitted to the ICU for 62 days at the completion of data collection. The median PICU length of stay for all 100 patients was 3.9 days (IQR 2.3–8.1). Four patients who were each admitted for more than 28 days accounted for a total of 215 patient-days.

Drug Orders

There were 4419 drug orders written for 241 different drugs (figure 1). Drug orders comprised 3557 (80.4%) new orders, 107 (2.4%) modification orders, and 755 (17.1%) discontinuation orders. The routes of administration ordered were intravenous 3017 (68.3%), enteral 802 (18.1%), respiratory tract 142 (3.2%), topical 30 (0.7%), subcutaneous 23 (0.5%), intrathecal 7 (0.2%), and 2 (0.1%) intramuscular. No route of administration was specified in 659 (14.9%) orders. The 3017 intravenous medication orders were for intermittent doses in 1915 (63.5%) orders and for continuous infusions in 1102 (36.5%) orders. Each patient received a median of 29.5 (IQR 16.5–48.5) drug orders during their PICU admission, and a median of four orders (IQR 1–8) per PICU day. Nine patients received 80 or more drug orders during their PICU stay (figure 1).

Fig. 1
figure 1

Number of drug orders written and number of different medications ordered for critically ill children admitted to a university-affiliated intensive care unit (ICU). Data are from 100 children who were admitted to a pediatric medical-surgical ICU for a total of 851 ICU days.

Full size image

Unadjusted analyses demonstrated that more drug orders were written for children who received regular medication at home before hospital admission; for children who, during their ICU admission, received inotropes or vasopressor medications, ventilation, or dialysis; and for children with more organ dysfunction (table II). Each additional day of ICU stay was associated with 3.26 additional drug orders (p < 0.0001). Children with greater PRISM III scores at ICU admission had more orders written for them during their ICU stay (p = 0.003; r2 = 0.09). Multiple regression analysis demonstrated that four factors were independently associated with increased drug orders: increasing PELOD score (p = 0.027), increased duration of dialysis (p = 0.002), increased duration of inotropic or vasopressor medication therapy (p = 0.0002), and greater length of ICU stay (p < 0.0001). These four factors explained 76% of the variability in the number of orders (r2 = 0.76). Each day of inotrope therapy was independently associated with 30 additional orders.

Ordered Drugs

There were 241 unique medications ordered, with a median of 14 (IQR 9–18.5) per ICU admission (figure 1). The most frequently ordered drugs were morphine (457 orders, 10.6%), furosemide (frusemide) [337 orders, 7.8%], potassium (237 orders, 5.5%), lorazepam (226 orders, 5.2%), and albuterol (salbutamol) [158 orders, 3.7%]. The 21 most frequently ordered drugs were ordered 2684 (60.5%) times (figure 2). Forty-eight (20%) drugs were ordered on one occasion only. The most commonly ordered therapeutic categories were sedative/analgesic medications (29%), antimicrobials (14%), diuretics (9%), electrolytes (8%), bronchodilators (7%), muscle relaxants (5%), corticosteroids (4%), and vasopressor medications (4%) [table III]. The majority of new orders occurred on the first day of ICU admission (table IV).

Fig. 2
figure 2

The 21 most frequently ordered medications for 100 children admitted to a pediatric medical-surgical intensive care unit (ICU) for a total of 851 ICU days. The bars colored black represent drugs not classified in the key.

Full size image
Table III
figure Tab3

Number of drug orders by therapeutic category and clinical condition during intensive care unit (ICU) stay for 100 children admitted to a quaternary pediatric ICU for >24 hours

Full size image
Table IV
figure Tab4

Types of drug orders (new, modification, and discontinuation orders) per day during the course of intensive care unit (ICU) stay for 100 children admitted to a quaternary pediatric ICU for >24 hoursa

Full size image

Unadjusted analyses demonstrated that more different medications were ordered for children who had received regular medications at home prior to hospital admission; for children who, during their ICU admission, received inotropes or vasopressor medications, ventilation, or dialysis; and for children with more organ dysfunction (table II). Each additional day of ICU stay was associated with 0.37 additional drugs ordered (p < 0.0001). Children with higher PRISM III scores at PICU admission also received a greater number of ordered drugs (p < 0.0001; r2 = 0.14). Multiple regression analysis demonstrated five factors that were independently associated with an increased number of medications ordered; treatment with regular medications prior to hospital admission (p = 0.01), increasing PELOD score (p < 0.0001), increased duration of dialysis (p = 0.006), increased duration of inotropic or vasopressor medications (p = 0.001), and increased duration of intubation (p < 0.0001). These five factors explained 72% of the variability in the number of drugs ordered (r2 = 0.72).

Drug Administrations

Patients received a total of 11 911 administrations of intermittent medication during their ICU admission (figure 3), and a mean of 14 medication administrations per day. The most commonly administered classes of medications were antimicrobials (2392 [20%]), sedatives (1500 [12.6%]), analgesics (1119 [11.9%]), corticosteroids (731 [6.1%]), and diuretics (678 [5.7%]) [figure 4].

Fig. 3
figure 3

Frequency histogram of the number of drug administrations during intensive care unit (ICU) stay. Data are from 100 children who were admitted to a pediatric medical-surgical ICU for a total of 851 ICU days.

Full size image
Fig. 4
figure 4

Frequency histogram of the 27 most frequently administered drugs (by intermittent administration) to 100 children who were admitted to a pediatric medical-surgical ICU for a total of 851 ICU days. The bars colored black represent drugs not classified in the key.

Full size image

Each patient received a median of 58 (28–129) administrations (figure 3). Unadjusted analyses demonstrated that higher numbers of drugs were administered to children who received regular medication before hospital admission; to children who, during their ICU admission, received mechanical ventilation or dialysis; and to children who had more organ dysfunction (table II). Each additional day of ICU stay was associated with 11 additional drug administrations (p < 0.0001). The number of drug administrations was not related to the PRISM III score (p = 0.09; r2 = 0.03). Multiple regression analysis found that increasing PELOD score (p = 0.006) and increasing length of ICU stay were associated with increasing frequency of administrations (p < 0.0001). The total number of doses administered increased by 11 for each ICU day and by 2.3 for each PELOD score point. These two factors explained 70% of the variability in the number of drug administrations (r2 = 0.70).

Discussion

We performed a prospective observational study of pharmacotherapy in critically ill children admitted to a quaternary PICU in a university-affiliated pediatric hospital. We found that drug ordering and drug administration were common activities. The 100 patients studied had 4419 drug orders written for 241 different drugs, and received 11 911 doses of medication, with a mean of 14 administrations per day.

Increasing numbers of drug orders written, different medications ordered, and drug administrations were all associated with increasing PELOD score. Increasing numbers of drug orders and ordered drugs were independently associated with duration of dialysis and circulatory support in the ICU. The number of drug orders and drug administrations were independently associated with ICU length of stay (the total number of doses administered increased by 11 for each ICU day) whereas the number of drugs ordered was independently associated with duration of mechanical ventilation.

Our findings reflect the practice of critical care medicine and the characteristics of the patients studied, and have implications for the evaluation of medicines and medication safety. First, the types of drugs and drug classes most frequently used reflect the general therapeutic objectives of the PICU.[1,20–25] Sedative and analgesic mediations were common therapies. Morphine and lorazepam were the most commonly ordered and administered medications (by bolus administration), respectively. Fentanyl was also frequently ordered and administered (figures 2 and 4). The frequent use of sedatives, analgesia, and muscle relaxant medications likely reflect the use of mechanical ventilation on nearly two-thirds of days studied, and in nearly 80% of patients. Diuretics and electrolytes were commonly ordered and administered, reflecting the management of fluid overload and secondary electrolyte depletion. Medications to prevent or treat gastrointestinal bleeding and to promote gastrointestinal motility were also commonly used. Anticoagulation was uncommonly prescribed, in contrast with practice in critically ill adults.[26]

Second, the medications used reflected the therapeutic needs of the patient groups studied. Our population included children with acute respiratory illness, conditions treated with regular medications before hospital admission, and complex conditions, including malignancy, bone marrow transplantation, and solid organ transplantation. We found that children who were receiving regular medications at home before hospital admission received more medications while in the PICU. Children receiving more medications at baseline are likely to be more complex and have greater risk of critical illness than children receiving fewer medications.[27] Moreover, continuation of regular medications while in the PICU may increase the number of drug orders, different medications ordered, and drug administrations. The frequent use of antimicrobial therapy most likely reflects the frequency of acute infectious illness and immunocompromised patients. The use of corticosteroids and immunosuppressant medications (tacrolimus) was consistent with the admission of 14 patients who had either an oncological diagnosis or who had received a bone marrow or solid organ transplant.

Third, organ dysfunction was common and had therapeutic implications. Half the patients had more than two dysfunctional organ systems.[28] We found that these patients received more drug orders, more unique medications, and more drug administrations than patients with two or fewer dysfunctional organ systems (table II). Organ dysfunction may modify drug elimination, alter expected responses to standard doses, and increase the risk of toxicity and other adverse drug events. Alterations in pharmacokinetic and pharmacodynamic parameters in critical illness are likely for many drugs.[29,30] Our data show that in patients with increasing organ dysfunction and those who received inotropes, the number of orders increased more rapidly than the number of drugs (table II). This supports the notion that more ‘fine tuning’ of pharmacotherapy occurred in sicker patients, underscores the importance of drug-specific pharmacokinetic and pharmacodynamic knowledge in critical illness, and raises questions about the contribution of pharmacotherapy to organ dysfunction in critical illness.

Fourth, the volume of drug orders and drug administrations was substantial. This represents a considerable workload for frontline care providers and provides opportunity for error, drug incompatibility, and drug interactions in patients already vulnerable to adverse outcomes associated with their critical illness.[31,32] Our data illustrate a substantial potential for adverse drug events. Efforts to improve patient safety through the use of clinical pharmacists,[33] computerized medication management systems,[34] and drug preparation[35] and administration mechanisms[36] may be feasible and warrant clinical and economic evaluation.[37] To date, the only intervention shown to reduce harmful medication errors in ICU patients is the addition of clinical pharmacists to multidisciplinary rounds.[33,38]

Finally, the number of different medications ordered was substantial and was more than that observed in many other pediatric populations. The number of different medications ordered was similar to that in a recent study of Spanish neonates,[39] but considerably higher than those reported in studies conducted in pediatric outpatient populations.[15,16] These data underscore earlier calls for more comprehensive regulatory standards and postmarketing studies in children.[3,40]

Limitations

There are several limitations to this study. First, our results may not be representative of other, non-quaternary PICUs. The complexity and acuity of the population studied were high; more than 75% of patients received mechanical ventilation, half had more than two dysfunctional organ systems, and we excluded short-stay patients. Other ICUs caring for patients of lower acuity may have lower rates of drug utilization.

Second, the population represented was admitted for >24 hours. Further studies would be required to describe pharmacotherapy in the pediatric ICU population admitted for <24 hours. Third, our rates of medication administration included only intermittent doses. We did not include administration of continuous infusions in our analysis of drug administrations. We also assumed that all doses administered were documented. Failure to document administration of a drug would also under-estimate the frequency of administrations. Fourth, this study was conducted over winter and spring months. Consequently, our results may not be representative of all seasons. Longer duration studies will be required if seasonal phenomena are to be described. Fifth, we did not quantify the appropriateness of drug therapy. Given the complexities of describing the rationale for a given drug order and the potential for direct observation to alter prescribing practice, we elected not to evaluate this aspect of care.

Finally, we did not assess if the therapeutic needs of children studied were met by the therapy provided. We described the provision of pharmacotherapy to facilitate organ system support (mechanical ventilation, dialysis, circulation), and fluid and electrolyte management. We did not assess the specific needs of the children studied in order to identify areas of unmet therapeutic need. This will require a different study design.

Conclusions

We quantified pharmacotherapy in pediatric critical illness and found that critically ill children received a substantial number of drug orders for a diverse range of medications, and that sicker patients received more drugs and had more drug orders. The most commonly employed medications were those that support the establishment and maintenance of invasive therapies, provide sedation and analgesia while receiving these therapies, treat infectious diseases, and prevent ICU-related complications such as gastrointestinal bleeding, fluid overload, and electrolyte depletion. Our results underscore the relevance of pharmacotherapy to the care of critically ill children, and the potential for improved pharmacokinetic knowledge to improve the outcomes of critical illness.