OBJECTIVES. The objectives were (1) to identify the drugs reported most commonly during NICU care, (2) to examine how different methods of documenting drug use could influence prioritization of drugs for future research aimed at evaluating safety and efficacy, (3) to describe the demographic differences in the population samples for some specific medications, (4) to identify which reported medications are used for patient populations with >20% mortality rates, and (5) to examine how reports of drug use change over time.
METHODS. A retrospective review of a large national data set was performed.
RESULTS. The 10 medications reported most commonly for the NICU were ampicillin, gentamicin, ferrous sulfate, multivitamins, cefotaxime, caffeine citrate, furosemide, vancomycin, surfactant, and metoclopramide. Medications used for patient populations with >20% mortality rates included amphotericin, clonazepam, dobutamine, epinephrine, ethacrynic acid, insulin, lidocaine, metolazone, milrinone, inhaled nitric oxide, nitroglycerin, octreotide, pancuronium, phenytoin, sodium nitroprusside, sodium polystyrene sulfonate (Kayexalate), tris-hydroxymethylaminomethane acetate buffer, and tolazoline. Several of these drugs (eg, amphotericin B and bumetanide) were used primarily for extremely premature neonates, and this usage might explain the high mortality rates for the population of neonates treated with these medications. Other medications (clonazepam, milrinone, inhaled nitric oxide, and phenytoin) were used primarily for near-term and term infants. The explanation for the high mortality rates for these neonates is less clear and may be related primarily to the severity of illness for which the medications are used. Utilization rates for several different medications (eg, cisapride, metoclopramide, and dexamethasone) changed by >50% during the past 5 years.
CONCLUSIONS. Data reported here are the first from a large national data set on the use of different medications for neonates admitted for intensive care and should be helpful in establishing priority agendas for future drug studies in this population.
Approximately 75% of the drugs licensed since the early 1970s have been approved without pediatric drug labeling.1–6 Although not all of these drugs have pediatric indications, studies in various settings show that a considerable proportion of the drugs used for children are either outside the product license (off-label) or without a license at all (unlicensed).1–6
Usually clinical trials for US Food and Drug Administration (FDA) approval exclude children, and typically labels for new medications provide physicians with no guidance regarding a product's effectiveness, dosing, or safety among pediatric patients. The FDA reported, “For most drug classes, there is almost no information on use in patients under 2 years of age.”7 The limited available information regarding adverse drug events in the pediatric population focused on medication errors such as overdosing or accidental exposure.8,9 The potential for adverse drug reactions among children is greater than that among adults.7 Young children have immature detoxification mechanisms, which decreases their ability to process drugs and drug metabolites. In addition, drug doses must be adjusted individually for a significantly broader range of body sizes and weights.7 Labeling of drugs is important because it offers a clear concise presentation of information that has been approved by the FDA as substantial evidence of efficacy, safety, and dose for the cited indications and age groups.10
In January 2003, the first step in setting a national pediatric drug agenda was undertaken by the National Institutes of Health (NIH), in consultation with the FDA and a panel of pediatric experts. This group developed a list of drugs for which pediatric studies are needed,11 as mandated by the Best Pharmaceuticals for Children Act. This list prioritized certain drug studies that were needed for children, to ensure safety and efficacy, on the basis of (1) the frequency of medication use in the pediatric population, (2) the severity of the condition being treated, and (3) the potential for providing a health benefit in the pediatric population. The NIH, the FDA, and pediatric experts have updated this list on an annual basis (see below). The specific aims of this study were (1) to identify the drugs reported most commonly for NICU patients, (2) to examine how different methods of documenting drug utilization could influence prioritization of drugs for future research aimed at evaluating their safety and efficacy for use among neonates, (3) to evaluate demographic differences in the population samples for some specific medications, (4) to identify medications used for patients with >20% mortality rates, and (5) to evaluate how reported drug use may change over time.
Using a deidentified data set from which several other observations have been published,12–16 we assembled a retrospective cohort of neonates for this study. Clinical data on these neonates were recorded during their hospitalization in the NICU. Admission, discharge, and daily progress notes were generated with a proprietary computer-assisted tool, and the data were stored in an electronic database. Pediatrix Medical Group provides intensive care services in 220 hospitals in 32 states and Puerto Rico. The data in the electronic database are used for medical record documentation, billing, and quality improvement projects. Clinicians providing care to patients interact with the patients' data on a daily basis, to generate progress notes and billing. Each day's notes are stored with medications used and diagnoses made. The local data are consolidated within the Pediatrix Medical Group data warehouse, deidentified, made compliant with Health Insurance Portability and Accountability Act of 1996 regulations, and configured into tables that can be joined and queried for statistical analyses. Certification of the deidentified data and approval for use of the data reported here were provided by the Wichita Medical Research and Education Foundation (Wichita, KS) institutional review board. Research is performed only with the deidentified database.
Data on estimated gestational age represented the best estimates based on both obstetrical data and neonatal examination findings. Designation of race was based on the options contained in the database, which were white, black, Hispanic, Native American, and Asian.
Our analytical approach to these data was descriptive in nature. Our goal was to present a picture of how NICU medication use is reported. Specific database tables within the data warehouse used for this analysis were “patients,” “admissions,” “medications,” “diagnoses,” and “cultures.” All reports of medication use collected within the medications table were reviewed. For comparative analyses, we counted a patient only once for each medication. Differences in the demographic characteristics of patient populations treated with a specific medication were analyzed by comparing the specific medication population sample with the overall population sample in bivariate analyses. Continuous variables (estimated gestational age and birth weight) were evaluated with 2-tailed t tests. Categorical variables (eg, race and gender) were evaluated with a 2-tailed χ2 test. Nonparametric data were assessed with Kruskal-Wallis analysis of variance. Changes in medication use over time were assessed with χ2 analysis. We also used the linear trend test and the Cochran-Armitage trend test to evaluate time-related changes.
Reporting Methods for Estimating Medication Utilization
We report counts of medication use in 3 different ways (Fig 1). (1) Frequency was defined as the number of times a unique medication name was reported in the medications table (ie, the raw count). (2) Courses was defined as the number of times a unique medication name was reported for a single patient with a specific start date. For example, if a patient had 2 reports of ampicillin and both events occurred on the same date, then this likely represented a duplicate report of ampicillin or 1 of the events represented a single dose (such as a loading dose) and not a full course of therapy. In this instance, ampicillin would be counted for only 1 course. If a patient had 2 reports of ampicillin with different start dates, then both reports of ampicillin would be counted for that patient, that is, the patient had 2 courses of ampicillin during the hospitalization. By examining the number of courses, we could estimate medications that were given more than once for a single patient. Examples of medications reported more than once are listed below. (3) Exposure was defined as the number of unique medication names reported for each patient. In our exposure count, we counted a patient only once for each medication. Multiple courses of the same medication for a specific patient were excluded in this count.
We report medication usage rates as (1) courses per exposure, which is an estimate of the number of times patients were exposed to >1 course of a specific medication, and (2) population-based exposure rate, which is a calculated value based on exposure and normalized to the total study cohort. The population-based exposure rate is expressed as the number of medication exposures per 1000 NICU-discharged patients and is calculated as (1000 × exposure)/n, where n is the total number of patients discharged from the NICU.
There were a total of 253651 discharges during the study period. There were 45192 discharged patients (18%) for whom there were no reports of any medication being used and 208459 patients (82%) for whom ≥1 medication was reported. Characteristics of the total patient population reported in our database (study cohort) are shown in Table 1.
The frequency of medication reporting was found to be 1425992 records of 409 unique medications covering the time period between January 1996 and April 2005. This represented 1359216 medication courses and 1130983 medication exposures.
The most commonly reported medications are shown in Table 2, and the rarely reported medications are shown in Table 3. A complete listing of all drugs reported to our data set, along with basic demographic information on the population for which the drugs are used, is available on request.
Although there is value in clarifying the method of reporting medication exposure and use, the order of the most commonly reported medications was not significantly altered by the counting method (frequency, courses, or exposures) or by indexing the number of reports to the number of total discharges (Tables 2 and 3).
Most medications were reported only once for a specific patient (1 course of treatment per exposed patient) (Tables 2 and 3). There were, however, several predictable exceptions, including the following drugs: caffeine citrate, furosemide, vancomycin, and dexamethasone. These medications were often reported twice for the same patient, which suggests that treatment courses of these medications were often repeated for the same patient. The number of medications whose frequency was <100 equaled 188, or 46% of the 409 different medications (examples are listed in Table 3).
Drugs Used in Premature Neonates
The profile of the patient population for each of the commonly used drugs was also explored. In order of decreasing utilization, the drugs used primarily for premature neonates with a median estimated gestational age of <32 weeks included caffeine citrate, surfactant, vancomycin, furosemide, metoclopramide, dopamine, nystatin, and aminophylline. In contrast, ampicillin, gentamicin, cefotaxime, phenobarbital, morphine, and vitamin K1 (phytonadione) were used primarily for mature neonates.
Drugs Used in Neonates Who Experienced More Than 20% Mortality
The overall mortality rate in the study population (n = 253651 discharges) was 2.7%. For drugs reported to have been used for >50 patients, we queried the data set to identify medications used for patients who experienced >20% mortality rates (Table 4). Several of these drugs (eg, amphotericin B, amphotericin B liposome, and bumetanide) were used primarily for extremely premature neonates, whereas others (clonazepam, milrinone, nitric oxide, and phenytoin) were used primarily for near-term and term infants.
Changes in Medication Use Over Time
We also found significant changes over time in the use of commonly reported medications (Fig 2). In Fig 2A, the decrease in the use of cisapride preceded its withdrawal from the market by the manufacturer because of reports of cardiac rhythm disturbances. A concurrent increase in the use of metoclopramide and ranitidine is also seen. In Fig 2B, changes in the use of various surfactant preparations seemed to coincide with the introduction of new surfactant products at various times during the past 8 years. Table 5 lists other commonly reported medications (frequency of >1000) for which reports changed by >50% in the time period of 1999 to 2004. Understanding the source and impact of these trends will be critical for the management of prescribing habits of physicians and could be a valuable source of postmarketing surveillance for neonatal medications.
Using a large national data set, we have described the medications reported most commonly for neonates admitted for intensive care. We think that the large sample size of our data set allows us to provide a more-comprehensive description of both commonly and rarely reported medications.
As the FDA and NIH set study priorities for the use of drugs for children, we think that these data will be essential in helping to focus future research activities. We applaud the efforts of these governmental organizations, and we wrote this article in part because of questions we received from and presentations we made to the FDA and the NIH in the past 2 years. Several lists of drugs for which pediatric studies are needed have been published. The lists include azithromycin, baclofen, bumetanide, dobutamine, dopamine, furosemide, heparin, lithium, lorazepam, rifampin, sodium nitroprusside, and spironolactone11; ampicillin/sulbactam, diazoxide, isoflurane, lindane, meropenem, metoclopramide, piperacillin/tazobactam, and promethazine17; ampicillin, ketamine, vincristine, dactinomycin, and metolazone18; and ivermectin, hydrocortisone valerate ointment, hydrochlorothiazide, ethambutol, griseofulvin, methadone, and hydroxychloroquine.19
Although there is some overlap between our drug list and the drugs listed above,11,17–19 there are several commonly used “neonatal drugs” that are not on the government lists. For example, drugs used commonly to treat apnea (caffeine, aminophylline, and theophylline) and drugs used commonly to treat sepsis (vancomycin and cefotaxime) are not included in the FDA/NIH lists. It must be noted, however, that the efforts of the FDA and NIH are directed at evaluating drugs in need of study for all children and not just neonates. The difference between our list and that of the FDA/NIH highlights the importance of evaluating drug utilization for specific pediatric subpopulations (eg, neonates, infants, adolescents, children with cancer, or children with pain).
Inpatient databases such as ours allow statistical analyses of drug use in relation to risk factors, indications, and outcomes, a feature generally not available in outpatient databases. Exploratory analyses can provide information about drugs and outcomes, which should be helpful for planning clinical trials. In addition, data analysis can provide guidance about potential adverse events,20,21 as we described previously for the associations between patient risk factors and risk of necrotizing enterocolitis16 and between prenatal corticosteroid therapy and birth weight and head circumference.22 We also reported a comparison of the effects of 2 surfactants on neonatal death, intraventricular hemorrhage, and necrotizing enterocolitis.14 These types of analyses provide important postmarketing surveillance of medications used in the NICU.
We realize the limitations of a retrospective review of any data set that is accumulated as part of medical record documentation. Retrospective studies are limited by incomplete data and duplicate data entry, as we described. However, validation of the observations reported here would require an exceptionally large, prospective, monitored, demographic study, which is unlikely to be performed in the foreseeable future and, if performed, would need ongoing data review as practices change. In this situation, review of data collected as part of providing care can yield important insights into clinical practice and its effects on outcomes.
To evaluate adverse events properly, it will be important to correct for differences in birth weight, estimated gestational age, degree of illness, timing of drug use, and concurrent drug use. As a first step in this complex analytic process, we described the medications used for neonates with a mortality rate of >20%. The data presented here are not meant to imply that the drugs themselves contributed to the mortality rate. Most of the medications reported in Table 4 are drugs used for critically ill neonates and may in fact reduce the mortality rate. For example, untreated fungal infections among premature neonates are associated with very high mortality rates; therefore, our finding of a mortality rate of 26% for neonates treated with amphotericin B is not unexpected. However, the mortality rate was even higher for neonates treated with the liposomal form of amphotericin B, and this deserves additional exploration and confirmation of contributing factors. Our intent in this article is not to define attributable risk but to describe the findings in a manner that is most likely to generate testable hypotheses.
We identified the drugs reported most commonly during NICU care and examined how different methods of documenting drug utilization influence prioritization of drugs for future research. Our data suggest that the method of reporting drug utilization has a minimal effect in identification of the top 10 drugs that are used for critically ill neonates. Another important observation from our data is that there are significant demographic differences in the populations of patients treated with specific drugs. These differences were most apparent for drugs used to treat the neonatal population with a very high mortality rate (>20%). Futures studies evaluating the mortality risk attributable to exposure to specific medications must evaluate the confounding effects of these demographic differences. Finally, we showed that neonatal drug utilization is a dynamic process and the reported use of drugs changes over time. Our report should enable decisions to be made by those charged with creating specific priorities about which drugs should be studied specifically for neonates.
- Accepted November 22, 2005.
- Address correspondence to Reese H. Clark, MD, Pediatrix Medical Group, 1301 Concord Terrace, Sunrise, FL 33323-2825. E-mail:
Financial Disclosures: Drs Clark and Bloom have received grants from Ross Laboratories, Forest Laboratories, and INO Therapeutics.
- ↵Wilson JT. Pragmatic assessment of medicines available for young children and pregnant or breast-feeding women. In: Morselli P, Garattini S, Sereni F, eds. Basic and Therapeutic Aspects of Perinatal Pharmacology. New York, NY: Raven Press; 1975: 411–421
- Jong GW, Vulto AG, de Hoog M, Schimmel KJ, Tibboel D, van den Anker JN. A survey of the use of off-label and unlicensed drugs in a Dutch children's hospital. Pediatrics.2001;108 :1089– 1093
- ↵Schirm E, Tobi H, de Jong-van den Berg LT. Risk factors for unlicensed and off- label drug use in children outside the hospital. Pediatrics.2003;111 :291– 295
- ↵Moore TJ, Weiss SR, Kaplan S, Blaisdell CJ. Reported adverse drug events in infants and children under 2 years of age. Pediatrics.2002;110 (5). Available at: www.pediatrics.org/cgi/content/full/110/5/e53
- ↵Fortescue EB, Kaushal R, Landrigan CP, et al. Prioritizing strategies for preventing medication errors and adverse drug events in pediatric inpatients. Pediatrics.2003;111 :722– 729
- ↵National Institutes of Health. List of drugs for which pediatric studies are needed. Fed Regist.2003;68 :2789– 2790
- Benjamin DK Jr, DeLong ER, Steinbach WJ, Cotton CM, Walsh TJ, Clark RH. Empirical therapy for neonatal candidemia in very low birth weight infants. Pediatrics.2003;112 :543– 547
- ↵National Institutes of Health. List of additional drugs for which pediatric studies are needed. Fed Regist.2003;68 :48402
- ↵National Institutes of Health. List of drugs for which pediatric studies are needed. Fed Regist.2004;69 :7243– 7244
- ↵National Institutes of Health. List of drugs for which pediatric studies are needed. Fed Regist.2005;70 :3937
- Copyright © 2006 by the American Academy of Pediatrics