Computer screen suggesting that the physician change intravenous (IV) medication to the oral (PO) route. QD indicates every day; Q24H, every 24 hours; Q8H, every 8 hours; and PC, after meals.
Change in average defined daily doses (DDDs) 4 months before vs 4 months after the intervention. The change is given per admission for patients who began taking the intravenous (IV) version of the 4 medications for which DDDs could be calculated: fluconazole, levofloxacin, metronidazole, and ranitidine. Changes in average oral (PO) and IV DDDs are shown (P= .002 for both, Wilcoxon rank sum test), along with hospitalwide trends in length of stay (LOS), case-mix index (CMI), and total drug expenditures for comparison.
Change in total intravenous milligrams used at the hospital for each of the 5 study medications 4 months before vs 4 months after the intervention. The average total milligrams per month in the before and after periods are listed in Table 1.
Change in average defined daily doses (DDDs) of intravenous (IV) and oral (PO) medication used per admission: average in 4-month periods before and after implementation of the intervention. The change is given for patients who began taking the IV version of the 4 medications for which DDDs could be calculated: fluconazole, levofloxacin, metronidazole, and ranitidine.
Fischer MA, Solomon DH, Teich JM, Avorn J. Conversion From Intravenous to Oral MedicationsAssessment of a Computerized Intervention for Hospitalized Patients. Arch Intern Med. 2003;163(21):2585-2589. doi:10.1001/archinte.163.21.2585
Many hospitalized patients continue to receive intravenous medications longer than necessary. Earlier conversion from the intravenous to the oral route could increase patient safety and comfort, reduce costs, and facilitate earlier discharge from the hospital without compromising clinical care. We examined the effect of a computer-based intervention to prompt physicians to switch appropriate patients from intravenous to oral medications.
This study was performed at Brigham and Women's Hospital, an academic tertiary care hospital at which all medications are ordered online. We targeted 5 medications with equal oral and intravenous bioavailability: fluconazole, levofloxacin, metronidazole, ranitidine, and amiodarone. We used the hospital's computerized order entry system to prompt physicians to convert appropriate intravenous medications to the oral route. We measured the total use of the targeted medications via each route in the 4 months before and after the implementation of the intervention. We also measured the rate at which physicians responded to the intervention when prompted.
The average intravenous defined daily dose declined by 11.1% (P = .002) from the preintervention to the postintervention period, while the average oral defined daily dose increased by 3.7% (P = .002). Length of stay, case-mix index, and total drug use at the hospital increased during the study period. The average total monthly use of the intravenous preparation of all of the targeted medications declined in the 4 months after the intervention began, compared with the 4 months before. In 35.6% of 1045 orders for which a prompt was generated, the physician either made a conversion from the intravenous to the oral version or canceled the order altogether.
Computer-generated reminders can produce a substantial reduction in excessive use of targeted intravenous medications. As online prescribing becomes more common, this approach can be used to reduce excess use of intravenous medications, with potential benefits in patient comfort, safety, and cost.
SEVERAL MEDICATIONS commonly used in hospitalized patients are equally bioavailable intravenously and orally.1- 5 Patients often begin to receive a medication intravenously when acutely ill or postoperatively, but are not switched to the oral medication when stable and taking oral medications or eating an oral diet. Oral medications have several advantages, including a lower cost and less nursing time for medication administration. Eliminating the requirement for intravenous catheters can lead to increased patient comfort and safety. Prior research6- 9 in the setting of community-acquired pneumonia has demonstrated that early conversion from intravenous to oral medications can also shorten the duration of hospitalization.
Physicians vary widely in their criteria for deciding when patients are ready for oral medications.10 There are not set protocols for this conversion, and physicians may delay initiation of oral medications because of uncertainty about the best time to make this change.11 In addition, the conversion decision is often simply not attended to in increasingly busy inpatient settings. Person-intensive interventions to encourage conversion have been attempted,9,12,13 but these can be costly and time-consuming.
We tested whether an automated computer-generated intervention prompting physicians to switch appropriate patients from intravenous to oral medications would accelerate such regimen changes and decrease unnecessary use of the intravenous forms of the targeted medications.
The site for this intervention was Brigham and Women's Hospital, an academic tertiary care medical center where physicians order and renew all medications on a computer-based system.14 Once the physician enters the name of the medication and the route by which it will be administered, doses, dosing schedules, and other instructions regarding the order are entered via keyboard or by selecting options from menus that appear on the screen. All orders require an electronic signature. After being entered by the physician, medication orders are valid for 1 or more days, depending on the type of medication. At the end of that period, if no action is taken, the order expires and is automatically discontinued. When an order is nearing expiration, a reminder is sent to the physician caring for the patient (or the covering physician), who is asked to renew, cancel, or change the order.
We identified 5 target medications with nearly identical oral and intravenous bioavailability.1- 5 Three were antimicrobial agents: levofloxacin, fluconazole, and metronidazole; the other 2 were an antiarrhythmic agent (amiodarone) and the hospital's main histamine antagonist (ranitidine). A sixth medication (ketorolac tromethamine) could be replaced by an oral medicine with similar efficacy (ibuprofen).15
The Brigham and Women's Hospital computerized order entry system has previously been used for interventions to improve physician ordering practices.16- 20 The intervention described herein was implemented as part of the daily order renewal process. An algorithm is programmed to review patient orders every evening, checking for those approaching the predetermined expiration date. Additional programming was written so that during this process, the system notes if a patient is taking one of the medications targeted for intravenous to oral conversion. If a patient is receiving such a targeted medication and has active orders for scheduled oral medications (ie, not as needed) or an order for an oral diet, the algorithm identifies the intravenous order as potentially convertible. Such orders are then added to the list of immediate medication renewals for that patient.
A reminder that renewals are due appears next to a given patient's name on the physician's computer monitor. When the physician responds to the reminder, a message appears noting that the patient is receiving oral medications or an oral diet and, thus, may be an appropriate candidate for conversion of the targeted intravenous medication to an oral route (Figure 1). The physician can then choose from 3 options: (1) to accept the change to the oral route, (2) to reject the change and renew the order as originally written, or (3) to cancel the medication altogether.
To evaluate the impact of the initiation of this program, we extracted data on medication use from the hospital's pharmacy computing system, which includes information on dose, route of administration, starting date, and total amount of all medications dispensed. Data were obtained on every dose dispensed in the oral and intravenous forms for the 5 targeted medications for which an identical preparation was available orally for the 4 months before and after the intervention was implemented. Data on the ketorolac to ibuprofen conversion were analyzed separately. After the data were aggregated, patient medical record numbers and other identifiers were removed to protect confidentiality. Approval for the medical record review was obtained from the Brigham and Women's Hospital Institutional Review Board.
We examined several measures to determine the effect of the intervention on the duration of intravenous therapy, analyzing use of each drug for patients who began receiving the intravenous version of the medication. The total milligrams used for each medication by each route were compared for the 4 months before and after the intervention start date. To obtain a more accurate assessment of the duration of intravenous therapy, we next calculated the average number of defined daily doses (DDDs) of each medication used intravenously per patient. The number of milligrams administered to each patient was divided by the most commonly used daily dose of the medication; this was done separately for the oral and intravenous forms. Because amiodarone is often given as a bolus followed by a continuous infusion, it was not possible to calculate the DDD for this agent. Ketorolac, often used on an intermittent or as-needed basis, was also excluded from the DDD calculation. We then compared the average DDD administered by both routes before and after the intervention. The Wilcoxon rank sum test was used to test for statistical significance of the change in average DDD for the combination of all 4 medications.
A sample of physician responses to the suggested conversion was studied during 3 nonconsecutive 2-week periods in the 6 months after the intervention was implemented. In this sample, the intervention was activated for 1045 separate orders for intravenous medications; we compared the actions selected (conversion to the oral route, cancellation of a medication, or renewal of an order without change) across medications.
Use of the 4 medications for which DDD could be calculated, as measured by average intravenous DDD, decreased by 11.1% (P = .002), as shown in Figure 2. The average oral DDD increased by 3.7% (P = .002) in the same period. In contrast, the average hospital length of stay and case-mix index increased slightly, while the total drug expenditure grew by 12.0%. Figure 3 includes amiodarone as well, and shows decreases in the total milligrams used hospitalwide for all 5 intravenous medications. Table 1 shows the total milligrams used by both routes. Use of oral levofloxacin and metronidazole increased, while use of the oral versions of the other medications decreased. Total milligram use of ketorolac decreased by 0.5%.
Figure 4 shows the changes in average DDD per admission from the 4 months before the intervention to the 4 months after the intervention for each medication. The average intravenous DDD of all 4 drugs shown in the figure decreased, ranging from fluconazole, with a decrease of 22.8%, to metronidazole, with a decrease of 9.2%. Ranitidine was the most commonly used drug, and the decrease in average intravenous DDD per admission was 11.1%. There was a smaller decrease in the average oral DDD of ranitidine. Oral fluconazole had a decrease in the average DDD per admission (−11.6%), and levofloxacin and metronidazole had large increases in the average oral DDD per admission.
There were 1045 unique orders for which the intervention was activated. In 372 of these cases (35.6%), physicians either changed the route of the drug from intravenous to oral or canceled the medication altogether; 21.6% of the intravenous medications were replaced with the oral version and 14.0% were canceled (Table 2). Frequently, the conversion suggestion appeared more than once before the order was changed or canceled. Often, after an order was canceled, the physician entered a new order for the oral version of the drug. Amiodarone orders were canceled 23.1% of the time, and amiodarone was not replaced at all in this sample of orders. Ketorolac was often canceled and occasionally replaced. Among the other medications, orders for ranitidine were changed or canceled most often (42.5%), followed by those for levofloxacin (30.3%), metronidazole (27.0%), and fluconazole (18.2%).
Automated reminders can be an important tool to promote quality care, and may reduce use of some medications without compromising outcomes. The intervention described herein identified medications that could be converted from the intravenous to the oral route by using data on other orders written for a given patient, and then provided the ordering physician with a recommendation about conversion to the oral route. We found reductions in the use of all of the targeted medications, although not all of the changes were statistically significant. When the intervention was in use, 35.6% of intravenous regimens identified to physicians were either converted to the oral route or discontinued.
The clinical impact of early conversion from intravenous to oral medications has been examined in previous research, most often in the setting of antibiotic use. Multiple observational studies7,21,22 have demonstrated that early switching of patients with community-acquired pneumonia to oral antibiotics produced a low clinical failure rate and a reduced length of stay. Siegel and colleagues8 randomized patients to a predetermined duration of intravenous antibiotics at admission for pneumonia and found that patients converted to oral antibiotics earlier had shorter lengths of stay without a compromise in clinical outcomes. With prior studies demonstrating the clinical efficacy of early conversion as context, our findings present an advance in the methods for increasing conversion to oral medications.
Other groups have encouraged early conversion of intravenous medications to the oral route using various techniques. Zamin and colleagues23 described a teaching hospital–based educational intervention that included distribution of written materials and oral presentations to house officers, medical students, and pharmacists. They found decreased use of intravenous antibiotics in the period after the educational intervention. The intervention described herein has the advantage of appearing automatically at the point of medication renewal, so that it does not rely on physicians to remember a previous educational message. In addition, such an automated intervention can continue to appear indefinitely and does not demand the ongoing commitment of time and effort required for educational sessions.
Ehrenkranz and coworkers9 used a nurse interventionist to suggest conversion to oral antibiotics for patients in 3 nonteaching hospitals and showed decreased use of intravenous antibiotics, decreased length of stay, and no worsening of clinical outcomes. Stirling and colleagues13 described a direct intervention by clinical pharmacists that reduced the use of intravenous ofloxacin. Przybylski et al12 also used clinical pharmacists to suggest oral replacements for intravenous antibiotics and found decreased drug use in the patients in whom the medications were converted. These 3 studies demonstrated conversion rates between 75% and 83%; however, each one focused on a single medication or drug class and used interventions that required time for medical record review to identify patients and discussion between educator and clinician. The studies included fewer than 10 patients per week, whereas the intervention described herein was presented to clinicians about 25 times per day. By automating the intervention, we were able to address several clinical areas (infectious diseases, gastroenterology, and cardiology) and reach many more patients while using much less person-time than prior interventions. Because it is embedded in the order renewal process, this intervention appears to physicians at a time when they are already thinking about the patient's medication orders.
In initial studies at Brigham and Women's Hospital, the same algorithm as used in the present study was used, but instead of presenting the suggestions directly to the ordering physicians, lists of potential conversions were provided to pharmacists, who then suggested potential conversions to physicians by telephone.24 More than 30% of physicians contacted switched medication route or indicated that they were willing to switch in response to these suggestions,24 similar to the results of our analysis. The difference between these rates and those in the preceding paragraph suggests that an automated algorithm to select medications for potential conversion may generate more recommendations that are rejected by physicians. A subsequent analysis of the data from the pharmacist intervention indicated that physicians often rejected the conversion for patients in certain subgroups, including patients in intensive care unit settings, patients with mucositis who were undergoing chemotherapy, and patients on the first or second postoperative day. The analysis suggested that more complex rules may be needed for identifying potential conversions with high specificity. Because all interventions and responses in the present study are captured on the computer, it is possible to refine the algorithm further and it is practical to use more complex algorithms if they are validated. Once optimized, the computerized intervention has the advantage of requiring much less ongoing labor to perform a suggestion and of allowing the physician to make the change more quickly and easily using the computer.
Physician responses to the intervention differed by medication. For amiodarone, more than 20% of orders were canceled, but none were directly converted to the oral route. The dosing schedule of amiodarone, with intravenous loading protocols, may mean that the automated intravenous to oral conversion prompt does not provide physicians with the dose recommendations they need for this drug. In the case of ketorolac, many orders were canceled. Ketorolac is often written as an as-needed order, and because prior research25 has emphasized the importance of short courses of ketorolac, physicians are likely to be receptive to discontinuing the medication altogether. Thus, the computerized intervention may have an additional impact by promoting physicians to review the use of the medication altogether.
Among the other 4 medications, physicians were willing to replace or cancel ranitidine more often than the 3 antimicrobial agents. This may be consistent with a belief that infections should be treated with intravenous medications even when oral administration is possible, although all of the medications targeted herein have equivalent oral bioavailability. Among the 3 antimicrobial agents, fluconazole, an antifungal agent, was changed or canceled least frequently, followed by metronidazole, often used for anaerobic infections. This may represent a physician preference to treat fungal and anaerobic infections with longer courses of intravenous therapy than infections being treated with levofloxacin, or it might be because of the greater morbidity of patients for whom drugs like fluconazole are prescribed, causing physicians to stay with the intravenous form if they believe that it is stronger. It is difficult to ascertain whether factors not captured by the computer system were contraindications to the use of oral medications or whether physicians rejected the conversion prompt because they believe that the intravenous form of the medication is more effective.
The decreases in overall use came during a period when there were slight increases in the average length of stay and the case-mix index and when overall drug expenditure at the hospital was increasing. For levofloxacin and metronidazole, the decreases in intravenous use (total use and DDD per admission) were matched by increased oral use. This probably represents an earlier shift to oral antibiotics, with patients still receiving a similar overall course of those 2 medications. For ranitidine, on the other hand, use by both routes declined. This may be because ranitidine is often started empirically for stress-ulcer prophylaxis, for which there is not a defined course of therapy, so that for many patients, when the intravenous route was discontinued, it was not automatically replaced with an oral prescription.
These findings indicate that computer-generated reminders produced a substantial reduction in use of targeted intravenous medications. As online prescribing becomes more common, this approach can be used to shorten the duration of use of intravenous medications, with potential benefits in patient comfort, safety, and cost.
Corresponding author: Michael A. Fischer, MD, MS, Division of Pharmacoepidemiology and Pharmacoeconomics, Departments of Medicine and Emergency Medicine, Brigham and Women's Hospital, 1620 Tremont St, Suite 3030, Boston, MA 02120 (e-mail: email@example.com).
Accepted for publication December 20, 2002.
This study was supported by training grant AG-00158 on epidemiology and aging from the National Institute on Aging, Bethesda, Md (Dr Fischer).
We gratefully acknowledge the assistance of Minalkumar Patel, MD, MPH; Joshua Benner, PharmD, ScD; and Alison LaTourette; as well as all of the members of the Information Systems and Pharmacy Departments who helped to implement this program.