PEDIATRICS Vol. 119 No. 1 January 2007, pp. e69-e75 (doi:10.1542/peds.2006-1388)
ARTICLE |
To What Extent Do Pediatricians Accept Computer-Based Dosing Suggestions?
a New York Presbyterian Hospital, New York, New York
b Columbia University and the Mailman School of Public Health, New York, New York
d Department of Biomedical Informatics, Columbia University, New York, New York
c Department of Public Health, Weill Medical College, Cornell University, New York, New York
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
OBJECTIVE. Pediatric medication errors occur frequently among hospitalized patients and are often related to dosing. Computerized physician order entry systems with decision support can decrease dosing errors, as well as other types of errors; however, their use in pediatrics has not been extensively studied. Our objective was to determine physician acceptance of dosing and frequency decision support elements in an inpatient pediatric computerized physician order entry system at 1 academic medical center.
PATIENTS AND METHODS. We performed a retrospective analysis of all electronic medication orders entered for pediatric inpatients at a large, urban teaching hospital between April 15, 2004, and December 31, 2004. Rates of physician acceptance of computerized physician order entry system–generated dosing and frequency suggestions were determined.
RESULTS. We analyzed 54413 orders in the computerized physician order entry system, of which 27313 orders had dosing or frequency decision support. Of the orders with decision support, approximately one third (8822) were accepted exactly by prescribers. Of the 18491 remaining orders, 8708 were changed for dose, 2466 for frequency, and 7317 for both. Among the 18491 orders that were changed, the majority 11322 deviated by a substantial amount (>50%) from the total daily dose initially suggested by the decision support feature. Overall, patient weight was missing 31.3% of the time, although patient age alone sometimes was sufficient for the computer to make a dosing suggestion.
CONCLUSIONS. Although dosing-decision support systems have the potential to improve care, more work needs to be done to determine and optimize their effectiveness. Commercial vendors of dosing knowledge bases need to deliver effective products, because most health care organizations will not have the resources to customize decision support rules.
Key Words: medication error • medication order entry systems • pediatric inpatient
Abbreviations: CPOE—computerized physician order entry
Medication errors (defined as errors in drug ordering, transcribing, dispensing, administering, or monitoring) among pediatric inpatients are a serious problem.1–6 Although they are not often life-threatening, these errors tend to occur commonly. A recent study by Kaushal et al2 demonstrated medication error rates among pediatric inpatients of 5.7 per 100 medication orders, a rate similar to that for adult patients. Others have shown medication error rates for children that are even higher.3 Of note, the Kaushal et al study showed that potentially harmful errors occur in pediatric inpatients at a rate 3 times higher than that for adults.
In pediatrics, the most common type of medication error is a dosing error at the ordering stage.1,4,5,7 One of the major reasons for the high rate of dosing errors is that most medications for children are prescribed on the basis of patient weight. Weight-based dosing requires practitioners to calculate a patient-specific dose and frequency regimen for each medication. Furthermore, complicated patients such as those in the NICU setting may have significant weight changes within a short period of time and require repeated calculations and adjustments, providing increased opportunity for dosing errors to occur.7 One particularly dangerous type of error, a 10-fold decimal point error, occurs frequently in children.8 In Toronto, Kozer reviewed 22 10-fold errors reported over an 8-month period. Seventeen errors were intercepted by hospital staff before reaching the patient. Eighteen of the errors occurred during physician prescribing, and 13 were potentially lethal or serious overdoses.9 Medication errors in pediatrics are especially problematic because children have limited physiologic reserves and may be unable to inform caregivers of untoward reactions.
In response to these concerns, several patient safety advocates have recommended the implementation of computerized physician order entry (CPOE) and dosing-decision support systems to decrease the incidence of preventable medication errors in children. The Institute of Medicine report, Crossing the Quality Chasm: A New Health System for the 21st Century, recommends replacing handwritten orders with automated systems by 2010. Similarly, a policy statement from the American Academy of Pediatrics, entitled "The Prevention of Medication Errors in the Pediatric Inpatient Setting," recommends the use of CPOE and guided decision support with pediatric-specific templates and standardized order sets as they become available.10
CPOE is a technology that can improve the process of ordering and the subsequent management of prescriptions.11–13 CPOE provides an automated environment for prescribing, with assured legibility and standardization, improved access to patient information, assured completeness of drug ordering, and increased reliability of communication with nursing and pharmacy. When combined with clinical decision support, CPOE can provide drug-drug and drug-allergy interaction information and generate weight-based dosing and frequency suggestions, among other features.14–16 Dosing-decision support can be implemented in a variety of ways. For example, the computer can present a suggested dose or list of doses,17,18 or the computer can present an alert if a certain dosing range is exceeded.19
To date, automated dosing-decision support tools have not been extensively evaluated in pediatrics. The primary goal of this study was to better understand the performance characteristics of a dosing-decision support feature in an inpatient pediatric CPOE system at 1 academic medical center. Specifically, we wanted to determine the rates of physician agreement with suggestions for dose and frequency that were generated by a dosing-decision support algorithm designed by a team of hospital-based pediatricians, pharmacists, and information technology staff. We also sought to determine the magnitude of disagreement when clinicians changed these suggestions and entered different dose and frequency specifications. Finally, we sought to identify whether there were ways to improve the dosing-decision support system. Approval from the local institutional review board was obtained for this study.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Study Site
We studied orders entered into the CPOE system at a general urban teaching hospital with a socioeconomically diverse patient population. This hospital treats both adult and pediatric patients but has a separate pediatric service. There are 30 general pediatric beds, 20 pediatric ICU beds, and 50 NICU beds. Physicians use CPOE at the hospital to enter all patient medication orders, as well as other medical orders. The majority of orders are entered by housestaff physicians, although attending physicians enter some orders as well. The CPOE system was implemented in the late 1990s. The CPOE system was commercially available, but the institution developed a customized dosing-decision support feature for pediatric medication orders.
Development of the Dosing-Decision Support Rules
In 2002, a set of dosing- and frequency-decision support rules for patients aged 0 to 18 years was developed for the 200 most commonly ordered medications in the pediatric setting. The list of 200 medications was derived from the over 2500 medications dispensed by pharmacy hospitalwide and was chosen by the developers in hopes of covering the vast majority of pediatric medication orders. The rules were developed through a collaborative effort by a team of general pediatricians, pediatric pharmacists, and pediatric subspecialists. The pharmacists culled dosing information from multiple neonatal and general pediatric medication references. The pharmacists' initial recommendations were reviewed, edited, and validated by subspecialty experts in each discipline (eg, infectious disease specialists for antibiotics, cardiologists for antiarrhythmic agents, etc). Each rule covered (a) a particular medication form, (b) an age range, and (c) a weight range. On the basis of these parameters, dose and frequency suggestions were generated. The suggested doses could have units of strength (eg, milligrams), strength per unit weight (eg, milligrams per kilogram), or strength per unit body surface area (eg, milligram per meter squared). Each rule assumed the most common indication for the medication. A total of 743 rules were developed for these 200 medications. For example, for injectable tobramycin in a patient between 0 and 8 days of age, who weighs between 2 and 10 kg, the computer would suggest a dose of 2.5 mg/kg every 12 hours.
In addition to dose and frequency suggestions, drug information related to prescribing such as dosing information for alternate indications, dosing in renal dysfunction, and other important conditions was compiled. This information could be displayed at the time of ordering either interruptively in a pop-up window when the amount of text was extensive, or noninterruptively on the usual ordering screen if the amount of text to be shown was <150 characters. For example, when ordering acyclovir, the algorithm calculates a dose and frequency suggestion on the basis of the assumption that the child has normal immune status; drug-specific dosing information for immunocompromised children is presented in a pop-up box. The information was specific to the medication being ordered but was not customized to the individual patient. Like the dosing and frequency decision support, this additional information was compiled only for the 200 most commonly ordered medications. If the medication being ordered was not 1 of 200 most common medications, the computer offered no dosing assistance.
Even for the 200 most common medications, rules were not written exhaustively for all combinations of age, weight, and drug form; the rules were written to cover the most likely clinical uses of specific medication forms. For example, no rule was written for acetaminophen elixir in a patient under 6 months of age, regardless of weight.
Medication Ordering
To enter a medication order, prescribers first pick the medication from an automated predefined list. This list of medications includes the form, eg, acetaminophen tablet is an item on the list, acetaminophen elixir is another. For medications with clinical decision-support rules, the computer calculates a suggested dose and frequency of administration for that patient on the basis of the dosing rules described above.
The system uses the patient's weight to calculate a dose per administration (eg, 120 mg per dose for an 8-kg child) and "rounds" to the closest reasonable amount. Each rule includes a precision to which to round, for example, to the nearest whole number, the nearest first decimal point, etc. In addition, there is a separate entry that specifies the maximum error tolerated for rounding. If the rounded number differs from the raw calculation by more than the maximum error tolerance, the suggestion defaults back to the "raw" calculated dose.
The dose and frequency fields in the computer ordering screen are then prefilled with the suggested dose and frequency. The prescriber can accept the suggestion or override it and enter a different dose and/or frequency. If the rule requires the patient weight to be present and the patient's weight is absent, the computer will make no dosing suggestion.
Data Collection
All medication orders for all patients admitted to the pediatric inpatient service between April 15, 2004, and December 31, 2004, were analyzed, including general medical/surgical pediatric wards, medical/surgical ICUs, NICU, labor and delivery floors, newborn nurseries, and behavioral health. Orders written in the emergency department were excluded. We obtained age, gender, race, and hospital location for all patients. For each order, the medication name and both suggested dose and frequency, as well as actual order dose and frequency, were obtained.
Analysis
We defined "acceptance" as orders that matched both CPOE-generated dose and frequency suggestions exactly. When the computer's suggestions were not accepted exactly, we determined whether there was a deviation in the actual dose, the actual frequency, or both. We determined overall acceptance rates and did subanalyses by specific medications and medication classes. We also calculated rates of acceptance by hospital unit. When the suggestion was not accepted exactly, we determined the degree of deviation by calculating the absolute value of the difference of the actual total daily dose and the suggested daily dose divided by the total daily dose (we defined total daily dose as dose times frequency in a 24-hour period). The deviation was defined as "small" if it was 
25%, "medium" if it was 26% to 49%, and "large" if it was 
50%. Finally, we identified a list of the most commonly ordered medications for which rules were not created. SAS 9.0 for Windows (SAS Institute Inc, Cary, NC) was used.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Description of Data
There were a total of 6119 admissions in our study (Table 1). The patients included 2901 (47%) females and 3218 (53%) males. There were 3810 (62%) admissions for neonates, 388 (6%) for infants, 1260 (21%) for preschoolers and school-aged children, and 661 (11%) for teenagers. The mean length of stay was 7.0 days. Of the admitted patients, 3381 (55%) were white, 1002 (16%) were Hispanic, 787 (13%) were black, 482 (8%) were Asian, and 467 (8%) were for patients of other or unknown races. Of a total of 54413 orders over the 8.5-month study period, there were 27313 orders with CPOE- generated suggestion, and 27100 orders without suggestions (50% each). There were 576 different providers who entered an average of 87.2 orders each (range: 1–946). Of the providers, 154 (27%) entered
50 orders over the study period, accounting for 83% of all orders.
|
Presence of Patient Weights
Of the 54413 medication orders, a weight for the patient was absent in 17051 (31.3%) (Table 1). For some medications (eg, albuterol), the computer could make a suggestion on the basis of age alone; weight was not required. There were 4528 medication orders for which a rule was not generated because a weight was absent. If weight had been universally present and the computer had made suggestions on these rules as well, there would have been suggestions made for 59% (27313 + 4528/54413) of the orders. At the time of the study, usual practice was to enter patient weight in advance of writing any orders; however, the system did not enforce this practice. The system has since been modified to require a weight for patients <18 years of age before a medication order can be written.
Orders With Suggestions
Among the 27313 orders with suggestions, 8822 (32.3%) were accepted exactly (for dose and frequency) by users (Table 2). Of the remaining 18491 orders that were changed, 8708 (47.1%) were changed for dose only, 2466 (13.3%) were changed for frequency only, and 7317 (39.6%) were changed for both dose and frequency. When an order was changed for either parameter, the ordered total daily dose ordered was less than the suggested total daily dose 9687 times (52.4%) and was greater than the suggestion 8308 times (44.9%); the total daily dose was equivalent in 496 orders (2.7%). The difference between the actual and suggested total daily dose was >50% of the suggested daily dose in 11322 (61.2%). We did not find any orders where the ordered dose differed from the suggested dose by a factor of 10 or more.
|
Acceptance by Category and by Unit
We then analyzed the frequency with which suggestions were made and accepted by medication category. Within each category the most common drug was also determined (Table 3). The class of medications with the most suggestions was antibiotics, yet only 21% of these suggestions were accepted by providers. Acceptance was highest for antipyretics (41%) and lowest for diuretics (3%). It should also be noted that almost half (43%) of the suggestions for the "other" category were accepted, attributable primarily to the overwhelming number of accepted orders for Vitamin K (96%).
|
We found a great deal of variation in acceptance among the various pediatric unit locations (Table 4). Acceptance was extremely high (96.1%) in the newborn nursery, because 3350 of 3494 orders for healthy newborns were vitamin for K. Acceptance was lowest (10.8%) in the NICU, with general pediatric floors at 23.6%.
|
Medications Without Dosing Rules
Finally, to understand whether the medication rule base was deficient, ie, whether there were medications that were being ordered that should have had rules but did not, we compiled a list of the most commonly ordered medications for which suggestions were not made. Of the 27100 orders for which no suggestions were made, the most commonly ordered drugs were erythromycin ointment (3791 [14%]), insulin lispro (480 [2%]), tuberculin (423 [2%]), Maalox plus (384 [1%]), phenobarbital (379 [1%]), quetiapine (260 [1%]), and apiprazole (257 [1%]).
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Medication ordering is one of the most complex aspects of medical care, requiring physicians to simultaneously integrate a thorough understanding of available drugs, disease processes, and patient-specific information in the context of a particular clinical circumstance. Medication ordering in pediatrics is further complicated by the need to take into account the patient's weight. It is not surprising, therefore, that in the pediatric inpatient setting, dosing errors are common and serious.
Information technology, in general, has the potential to improve patient safety, and a CPOE system with embedded dosing-decision support features is an appealing way to try to decrease the incidence of medication errors and improve patient care.17,20 In this analysis, we studied an application that suggested a specific dose and frequency on the basis of patient age and weight using a set of rules developed by an interdisciplinary team from the hospital. Overall, in this study physicians accepted the computer's dose and frequency suggestions less than one third of the time.
The acceptance rates in this study are low; however, it is difficult to know what the "ideal" acceptance rate should be, because there are many situations where deviations from a CPOE- generated suggestion would be considered medically appropriate. For example, dosing in the context of impaired renal function or titrating a drug to a particular physiologic effect may appropriately cause the physician to disagree with the computer's dosing suggestions and change the dose and/or frequency of an order. Also, because opinions differ on dosing appropriateness, the physicians' education may have been in conflict with the rules that were embedded in the computer. Dosing for different or less common indications may also lead to disagreement with the computer's suggestion. In our system, the specific dosing suggestion assumed the most common indication, and information boxes were displayed that gave general dosing guidance for other indications; the specific indication the order was based on was not available for analysis. In the future, more sophisticated computer ordering applications should capture the indication for the order so that dosing logic can be more context-specific. The developers of the system we studied originally had considered including dosing logic that would take the patient's renal function into account, but ultimately rejected the idea because it would have significantly increased the number of rules that would have been required.
Medical knowledge, including knowledge about medication dosing and frequency, is evolving. New medications are added to the formulary and new knowledge about optimal dosing strategies is learned. Organizations that implement computer-based dosing-decision support must have a strategy to stay abreast of evolving knowledge and periodically incorporate it into the system over time. In our analysis of orders without suggestions, we found medications that had been added recently to the formulary for which dosing rules had not yet been written.
In our analysis, we found circumstances where the computer might have made a dosing suggestion but did not. One reason for this was that the dosing rules were age-range specific and were written to cover the medication formulations that are used most frequently in certain age groups. For example, for acetaminophen elixir, a rule was written to cover patients between 6 months and 13 years of age, but there was no rule to cover patients <6 months. Although orders for acetaminophen elixir in patients <6 months of age were uncommon, they did occur and no computer suggestions were made. Also, the computer did not generate a suggestion when the patient's weight was missing. To make pediatric dosing-decision support systems as useful as possible, developers should consider how weight will be entered, as well as the completeness of the dosing-decision support rules.
We know that some dose-related errors are egregious. In pediatrics, 10-fold errors are common. Because we did not find any instances where dosing orders differed from suggestions by a factor of 10, the CPOE system may have helped prevent these more dangerous kinds of events by providing some kind of reasonable range as a starting point for the physician.
This study has some important limitations. Most notably, we did not measure the impact of the system on the rates of medication errors and adverse drug events, and are, therefore, not able to comment on the impact of the decision support system on error and injury reduction. During the study period, the hospital converted from a paper-based incident reporting system to a Web-based one, so rates of events obtained from the 2 systems would not be comparable. Without data describing the impact of the system on medication-related error and injury rates, the value of the system becomes a complex calculus. Certainly, when a dosing suggestion is accepted exactly, one could argue that the computer helped direct clinician decision-making. However, the provider may well have chosen those dosing parameters anyway. Furthermore, although it may seem as if a rejected suggestion is unhelpful, the computer's suggestions may have helped guide the clinician into choosing an appropriate dose or provided confirmation when their intended dose fell within an acceptable range.
We also did not survey the physician users of the system to determine whether they perceived value from the system, even if they did not accept the suggestion exactly. Although we speculated that there were medically appropriate reasons why physicians deviated from the computers' suggestions (eg, renal dysfunction, titrating to an effect, different indication, etc), we did not undertake a formal survey of physicians to document these reasons. We are unable to comment on the value of a priori medication suggestions versus posthoc critique, because ours is a suggestion-based system. The patient population was disproportionably made up of neonates and, therefore, might not be representative of pediatric inpatient populations at all institutions. Lastly, we did not determine whether the actual doses and frequencies were clinically appropriate, even if they deviated from the decision support suggestions.
The question of value is important because dosing-decision support systems are complex to build. They require a significant input of time from expert pediatricians and clinical pharmacists and other valuable and scarce resources. The system in this study took over a year to develop and involved over 2 dozen clinical experts. Furthermore, the initial effort is not the only cost. New drugs are constantly being developed and clinical knowledge regarding the optimal use of medications changes over time, thus the dosing rules must be updated and maintained. Commercially available dosing-decision support knowledge bases are available for the adult setting but must be manipulated by the local institution before they are suitable for clinical use.20 Unfortunately, for technical and other reasons, such customizations are not always transferable among provider organizations, so often each organization must repeat this work itself. Commercially available medication decision support products for the pediatric setting lag behind those available for the adult setting, probably because of their smaller market share and the increased complexity of development. In addition, most health care organizations will not have the skills and resources to develop their own rule set for computer-based dosing-decision support.20
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Although computer-based dosing-decision support has the potential to improve care in the pediatric setting, several aspects of electronic medication ordering need additional study. First, future studies of dosing-decision support should evaluate whether dosing-decision support systems can impact the rates of medication related errors and injuries. They also should evaluate whether deviations from CPOE suggestions are clinically appropriate. Second, studies should be performed that examine clinicians' perceptions of such systems, whether they are perceived to have value even if suggestions are not accepted, and the reasons for rejecting suggestions. Third, additional study needs to be performed to determine whether proactive suggestions or reactive critiques are the best way to provide dosing-decision support. Lastly, commercial vendors of dosing knowledge bases need to deliver clinically helpful products that can be used in the pediatric setting with minimal local customization.
In summary, although dosing-decision support systems have the potential to improve care, more work needs to be done to determine and optimize their effectiveness in the pediatric inpatient setting.
|
FOOTNOTES |
|---|
Accepted Sep 1, 2006.
Address correspondence to Brigid K. Killelea, MD, MPH, Center for Health Outcomes and Innovation Research at Columbia University, College of Physicians and Surgeons and the Mailman School of Public Health, 600 W 168th St, 7th Floor, New York, NY 10032. E-mail: brigidkillelea{at}yahoo.com
The authors have indicated they have no financial relationships relevant to this article to disclose.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
- Institute of Medicine, Committee on Quality of Health Care in America. Errors in health care: a leading cause of death and injury. In: Kohn L, Corrigan J, Donaldson M, eds. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 1999:22–58
- Kaushal R, Bates DW, Landrigan C, et al. Medication errors and adverse drug events in pediatric inpatients.
JAMA. 2001;285
:2114
–2120
[Abstract/Free Full Text] - Marino BL, Reinhardt K, Eichelberger WJ, Steingard R. Prevalence of errors in a pediatric hospital medication system: implications for error proofing. Outcomes Manag Nurs Pract. 2000;4 :129 –135[Medline]
- Crowley E, Williams R, Cousins D. Medication errors in children: a descriptive summary of medication error reports submitted to the United States Pharmacopeia. Curr Ther Res Clin Exp. 2001;26 :627 –640[CrossRef]
- Vincer MJ, Murray JM, Yuill A, Allen AC, Evans JR, Stinson DA. Drug errors and incidents in a neonatal intensive care unit: a quality assurance activity.
Am J Dis Child. 1989;143
:737
–740
[Abstract/Free Full Text] - Potts AL, Barr FE, Gregory DF, Wright L, Patel NR. Computerized physician order entry and medication errors in a pediatric critical care unit.
Pediatrics. 2004;113
:59
–63
[Abstract/Free Full Text] - Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of adverse drug events. ADE Prevention Study Group.
JAMA. 1995;274
:35
–43
[Abstract/Free Full Text] - Koren G, Barzilay Z, Greenwald M. Tenfold errors in administration of drug doses: a neglected iatrogenic disease in pediatrics.
Pediatrics. 1986;77
:848
–849
[Abstract/Free Full Text] - Kozer E, Scolnik D, Keays T, Shi K, Luk T, Koren G. Large errors in the dosing of medications for children.
N Engl J Med. 2002;346
:1175
–1176
[Free Full Text] - Bates DW, Leape LL, Cullen DJ, et al. Effect of computerized physician order entry and a team intervention on prevention of serious medication errors.
JAMA. 1998;280
:1311
–1316
[Abstract/Free Full Text] - Bates DW, Teich JM, Lee J, et al. The impact of computerized physician order entry on medication error prevention.
J Am Med Inform Assoc. 1999;6
:313
–321
[Abstract/Free Full Text] - American Academy of Pediatrics, Committee on Drugs and Committee on Hospital Care. Prevention of medication errors in the pediatric inpatient setting.
Pediatrics. 1998;102
:428
–430
[Abstract/Free Full Text] - Teich JM, Merchia PR, Schmiz JL, Kuperman GJ, Spurr CD, Bates DW. Effects of computerized physician order entry on prescribing practices.
Arch Intern Med. 2000;160
:2741
–2747
[Abstract/Free Full Text] - Hsieh TC, Kuperman GJ, Jaggi T, et al. Characteristics and consequences of drug allergy alert overrides in a computerized physician order entry system. J Am Med Inform Assoc. 2004;11 :482 –491[Medline]
- Overhage JM, Tierney WM, Zhou XH, McDonald CJ. A randomized trial of "corollary orders" to prevent errors of omission.
J Am Med Inform Assoc. 1997;4
:364
–375
[Abstract/Free Full Text] - Kuperman GJ, Teich JM, Gandhi TK, Bates DW. Patient safety and computerized medication ordering at Brigham and Women's Hospital. Jt Comm J Qual Improv. 2001;27509 –521:
- Chertow GM, Lee J, Kuperman GJ, et al. Guided medication dosing for inpatients with renal insufficiency.
JAMA. 2001;286
:2839
–2844
[Abstract/Free Full Text] - Peterson JF, Kuperman GJ, Shek C, Patel M, Avorn J, Bates DW. Guided prescription of psychotropic medications for geriatric inpatients.
Arch Intern Med. 2005;165
:802
–807
[Abstract/Free Full Text] - Reichley RM, Seaton TL, Resetar E, et al. Implementing a commercial rule base as a medication order safety net.
J Am Med Inform Assoc. 2005;12
:383
–389
[Abstract/Free Full Text] - Kuperman GJ, Reichley RM, Bailey TC. Using commercial knowledge bases for clinical decision support: opportunities, hurdles, and recommendations.
J Am Med Inform Assoc. 2006;13
:369
–371
[Free Full Text]
PEDIATRICS (ISSN 1098-4275). ©2007 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Kemp Do pediatricians accept computer-based dosing suggestions? AAP News, March 1, 2007; 28(3): 2 - 2. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
