Computerized Decision Support to Reduce Potentially Inappropriate Prescribing to Older Emergency Department Patients: A Randomized, Controlled Trial

Kevin M. Terrell, DO, MS, wz Anthony J. Perkins, MS,w Paul R. Dexter, MD,z Siu L. Hui, PhD,wz Christopher M. Callahan, MD,wz and Douglas K. Miller, MDwz

OBJECTIVES: To evaluate the effectiveness of computerassisted decision support in reducing potentially inappropriate prescribing to older adults. DESIGN: Randomized, controlled trial. SETTING: An academic emergency department (ED) in Indianapolis, Indiana, where computerized physican order entry was used to write all medication prescriptions. PARTICIPANTS: Sixty-three emergency physicians were randomized to the intervention (32 physicians) or control (31 physicians) group. INTERVENTION: Decision support that advised against use of nine potentially inappropriate medications and recommended safer substitute therapies. MEASUREMENTS: The primary outcome was the proportion of ED visits by seniors that resulted in one or more prescriptions for an inappropriate medication. The main secondary outcomes were the proportions of medications prescribed that were inappropriate and intervention physicians reasons for rejecting the decision support. RESULTS: The average age of the patients was 74, two-thirds were female, and just over half were African American. Decision support was provided 114 times to intervention physicians, who accepted 49 (43%) of the recommendations. Intervention physicians prescribed one or more inappropriate medications during 2.6% of ED visits by seniors, compared with 3.9% of visits managed by control physicians (P 5 .02; odds ratio 5 0.55, 95% condence interval 5 0.340.89). The proportion of all prescribed medications that were inappropriate signicantly decreased
From the Department of Emergency Medicine, wIndiana University Center for Aging Research, zRegenstrief Institute, Inc., and School of Medicine, Indiana University, Indianapolis, Indiana. This research was presented at the 2008 Annual Scientic Meeting of the American Geriatrics Society on May 2, 2008, in Washington, DC, and the 2008 Annual Meeting of the Society for Academic Emergency Medicine on May 29, 2008, in Washington, DC. Address correspondence to Kevin Terrell, Regenstrief Institute, Inc., 410 West 10th Street, Suite 2000, Indianapolis, IN 46202-3012. E-mail: kterrell@regenstrief.org DOI: 10.1111/j.1532-5415.2009.02352.x

from 5.4% to 3.4%. The most common reason for rejecting decision support was that the patient had no prior problems with the medication. CONCLUSION: Computerized physican order entry with decision support signicantly reduced prescribing of potentially inappropriate medications for seniors. This approach might be used in other efforts to improve ED care.Trial Registration: Clinical trials.gov Identier: NCT00297869. J Am Geriatr Soc 57:13881394, 2009.

Key words: emergency service, hospital; geriatrics; aged; decision support systems, clinical; randomized controlled trials

mergency departments (EDs) are major healthcare providers for older adults. In 2005, patients aged 65 and older made 16.7 million visits to EDs (approximately 48 visits per 100 persons per year), an increase of 26% over 10 years.1,2 Adverse drug events are a frequent cause of delirium in older adults3 and lead to more than 10% of ED visits by older adults.4 Seniors are more than twice as likely as younger adults to be treated in an ED for an adverse drug event.5 A number of methods have been developed to evaluate medication appropriateness for elderly adults.6 The Beers criteria7 are commonly used and are well supported by the medical literature.8,9 The central piece of the Beers criteria is a list of 21 medications that are potentially inappropriate for use by seniors. It has been specically recommended that potentially inappropriate medications be avoided when caring for seniors in EDs.10 Use of potentially inappropriate medications by seniors is associated with lower health-related quality of life11,12 and higher health services utilization.13,14 Prior research indicates that 6% to 7% of older adults who are discharged from an ED are provided one or more

JAGS 57:13881394, 2009 r 2009, Copyright the Authors Journal compilation r 2009, The American Geriatrics Society

0002-8614/09/$15.00

JAGS

AUGUST 2009VOL. 57, NO. 8

DECISION SUPPORT FOR INAPPROPRIATE PRESCRIBING

1389

prescriptions for a potentially inappropriate medication, as dened according to the Beers criteria.11,15 These estimates are similar to those reported in ofce-based care settings.1618 From 1997 through 2001, there was no evidence of a downward trend in the frequency of inappropriate prescribing.19 In response to the sustained high frequency of inappropriate prescribing, a recently published research agenda for geriatric emergency medicine recommended interventional trials of methods to reduce prescription of potentially inappropriate medications, such as . . . computer-assisted decision support in their list of highest-priority research in emergency medicine.20 Computer-assisted decision support has been found to signicantly improve physician performance and clinical practice in approximately two-thirds of trials.21,22 Decision support that is linked to computerized physician order entry has been shown to improve quality of care in the inpatient environment.2325 and the outpatient clinic setting,26,27 although to the knowledge of the authors of the current article, physician order entry with decision support has not been tested in the ED setting, and it is unclear whether decision support to improve prescribing safety would be effective in the chaotic environment that characterizes EDs, where the pace is rapid and patients typically have high illness acuity and are not generally known by the providers. A randomized, controlled trial was conducted to examine the effect of decision support to decrease the prescription of potentially inappropriate medications to older adults discharged from the ED and to identify the various reasons why providers reject decision support. It was hypothesized that the computerized decision support would reduce prescribing of potentially inappropriate medications.

Randomization and Intervention

Using a computerized random number generator, a biostatistician randomized physicians in blocks of two, stratied according to stage of training (faculty vs resident status and according to year of residency training) into the experimental or usual care (control) group. As new physicians commenced practice in the ED, they were randomized in the same manner. Physicians randomized during residency training who later joined the faculty during this 30-month study (n 5 4) remained in the study groups to which they were initially assigned. Investigators were blinded to physician assignments, except the biostatisticians, who required this information to analyze the results. Subjects were blinded to the study hypothesis. The intervention was computer-assisted decision support designed to reduce prescribing of medications that are potentially inappropriate for older adults. The intervention period was January 12, 2005, to July 7, 2007. Known characteristics of effective decision support, such as tting the decision support into the users workow and offering alternatives to the targeted potentially inappropriate medications, were implemented.30 An expert panel of two doctors of pharmacy, two physician information technology experts, three geriatricians, and three emergency physicians participated in the design of the intervention. First, the expert panel identied the medications on the Beers list to target in this investigation. The panel was provided Wishard ED prescribing data from the preceding year. The expert panel chose to target nine high-use and high-impact potentially inappropriate medications (listed in the left column of the supplemental table in Appendix S1, which is available in the on-line manuscript). These medications represented 80% of potentially inappropriate medications that had been prescribed to seniors in the Wishard ED in the previous year. The expert panel then identied safer substitute therapies for each targeted inappropriate medication. They limited the recommended medications to those on the formulary at Wishard. The right column of the supplemental table lists the therapies that were suggested in place of each inappropriate medication. For most, the recommendations varied according to the indication for prescribing the medication. Some medications had specic doses or durations of therapy recommended. The doses or durations of therapy for other medications were left to the preexisting default prescribing menu within the computerized physician order entry system. Decision support was provided only when a physician in the intervention group attempted to prescribe a targeted inappropriate medication for a patient aged 65 and older who was being discharged from the ED. An example of the decision support is shown in Figure 1. It displayed when the intervention physician ordered promethazine. The prescriber had the option to order a recommended alternative therapy or to reject the recommendations by selecting promethazine from the menu. If the physician chose to continue prescribing promethazine, then a second menu displayed to query the most important reason for rejecting the decision supports recommendation. The choices were This is a medication rell; The patient has had no problems with the medication in the past; I lowered the medications

METHODS Design Overview This was a randomized, controlled trial with physicians as the unit of randomization. The institutional review board of Indiana University/Purdue University approved the study. The requirement for written informed consent was waived. Setting and Participants The Wishard Memorial Hospital is a tax-supported, 450bed, university-afliated, urban, public hospital located on the Indiana University Medical Center campus. The Wishard Hospital ED is a Level I Trauma Center with approximately 100,000 annual visits. During routine patient care, physicians electronically write discharge orders, including medication prescriptions, using a computerized physician order entry system.28,29 Subjects included emergency medicine faculty and resident physicians. The three emergency medicine faculty physicians who participated in the conduct of this study were excluded. Postgraduate year-1 (PGY-1) resident physicians were also excluded, although it had been planned to include them. Upon nal reconciliation of data, it was detected that a computer malfunction had inadvertently prevented PGY-1 resident physicians randomized to the intervention arm from receiving decision support.

1390

TERRELL ET AL.

AUGUST 2009VOL. 57, NO. 8

JAGS

Figure 1. Computer-assisted decision support for promethazine.

dose; No good substitute exists for the medication; The patient insists on the medication; and Other (please specify: __________). Physicians in the control group did not receive the decision support, but the computer system tracked their prescribing.

Outcomes Data were obtained from the electronic medical record system.29 The a prioridetermined primary outcome was the proportion of ED visits by older adults that resulted in one or more prescriptions for a targeted inappropriate medication. As a secondary outcome, the proportion of all prescribed medications that were potentially inappropriate was examined. Other outcomes included the number of times that each potentially inappropriate medication was initially prescribed, the proportion of times that intervention physicians chose an offered alternative therapy separately for each medication, intervention physicians reported primary reasons for rejecting the decision support, and patient and physician factors associated with decisions to prescribe inappropriate medications for physicians in the control and intervention groups. Statistical Analysis Logistic models with mixed effects were used to compare the binary outcomes between the intervention and control groups to accommodate the correlated data. To test the effect of the intervention on the proportion of visits with an inappropriate medication, mixed effects models were used that included the group assignment as a xed effect and physicians as random effects to account for within-physician correlation. For the proportion of prescription orders written for a potentially inappropriate medication, two random effects were included in the model: one for physicians to account for within-physician correlation and one for patients nested within physicians to account for within-

patient correlation, because many patients had multiple medication orders. All ED visits (rst and subsequent visits) by older adults during the study period were included. Two sensitivity analyses were performed: one adjusted for provider and patient characteristics associated with the prescription of potentially inappropriate medications, and the other included PGY-1 resident physicians and the patients who received their care. To examine whether there was an effect that would decrease the number of attempts of inappropriate prescribing (i.e., contamination in the control group or a learning effect in the intervention group) over the course of the study, the proportion of ED visits with such attempts (out of all eligible patient visits) was estimated, and a time trend was tested for within each randomized group and their interaction. All analyses were performed using SAS version 9.1 (SAS Institute, Inc., Cary, NC).

RESULTS Sample Characteristics Figure 2 describes the physician subjects and the older adults who received their care in the ED. Pertinent characteristics of the 63 physician subjects are provided in Table 1. There were no important differences in the characteristics of intervention and control physicians. During the study, older adults made 7,458 visits to physician subjects. The 2,271 (30%) visits that resulted in hospital admission and the 25 (0.3%) visits resulting in death in the ED were excluded. The remaining 5,162 (69%) visits that led to an ED discharge were included in the study. The patients relevant demographic information is provided in Table 1. Main Outcomes The main outcomes are displayed in Table 2. The decision support signicantly reduced the proportion of ED discharges that resulted in a potentially inappropriate prescription

JAGS

AUGUST 2009VOL. 57, NO. 8

DECISION SUPPORT FOR INAPPROPRIATE PRESCRIBING

1391

Figure 2. Study ow diagram.This does not include postgraduate year 1 resident physicians, who were excluded from the study. ED 5 emergency department.

(3.9% vs 2.6%; P 5.02; odds ratio (OR) 5 0.55, 95% condence interval (CI) 5 0.340.89). This difference represents an absolute risk reduction of 1.3% (95% CI 5 0.4 2.3%). When analyzed as a percentage of all medications prescribed by physician subjects, the proportion of medications that were potentially inappropriate was signicantly reduced, from 5.4% to 3.4% (P 5.006; OR 5 0.59, 95% CI 5 0.410.85), with an absolute reduction of 2.0% (95% CI 5 0.73.3%). Table 3 provides the potentially inappropriate medications prescribed by study physicians as well as the effect of decision support on prescribing by intervention physicians. The medications prescribed by control physicians and initial prescribing by intervention physicians (before receiving decision support) were similar except for a substantially higher number of indomethacin prescriptions in the intervention group and a higher number of amitriptyline prescriptions in the control group. Computer-assisted decision support was provided 114 times to intervention physicians during 107 different ED visits. At seven visits, intervention physicians received decision support for two different potentially inappropriate medications. Overall, intervention physicians accepted 49 (43%) of 114 decision support recommendations. The most common reasons that intervention physicians reported for rejecting decision support recommendations were that the patient has had no problems with the medication in the past (40% of rejections), this is a medication rell (17%), no good substitute exists for the medication (16%), I lowered the medications dose (13%), and the patient insists on the medication (6%).

Sensitivity Analyses
In the rst sensitivity analysis, a mixed-effect logistic regression was performed that included patient characteristics (sex, age (in 5-year increments), and race) and physician characteristics (sex and years since rst year of residency). Female patients were signicantly more likely to receive inappropriate medications (Po.001; OR 5 2.03, 95% CI 5 1.392.96), whereas increasing patient age was signicantly associated with a lower likelihood of receiving an inappropriate medication (Po.001; OR 5 0.95, 95% CI 5 0.920.97). Increasing years since rst year of residency was associated with a lower likelihood of prescribing an inappropriate medication (Po.001; OR 5 0.94, 95% CI 5 0.910.97). Patient race and physician sex were not signicant factors. After adjusting for these items, the intervention was still signicantly associated with a lower likelihood of prescribing an inappropriate medication (P 5 .004; OR 5 0.58, 95% CI 5 0.400.84). The second sensitivity analysis was performed to account for the inadvertent deletion of PGY-1 resident physicians from the study and, consequently, the analysis. When including PGY-1, PGY-2, and PGY-3 resident physician, and faculty physician data, one or more potentially inappropriate medications were prescribed at 3.3% of ED visits in which the provider was an intervention physician, compared with 4.6% of visits by seniors who received care from control physicians (P 5 .04; OR 5 0.68, 95% CI 5 0.470.98). An analysis of only PGY-1 resident physicians found that there was no signicant difference in prescribing inappropriate medications between the inter-

1392

TERRELL ET AL.

AUGUST 2009VOL. 57, NO. 8

JAGS

Table 1. Characteristics of Emergency Department Physicians and Their Patients

Characteristic Control Intervention

Physicians, n Sex, n (%) Female Male Status, n (%) Faculty physician Resident physician through entire study Resident who became faculty Years since postgraduate year 1 of training, mean SD Patient visits, n Sex, % Female Male Age, mean SD Race, % African American Caucasian Hispanic Other

31 8 (25.8) 23 (74.2) 11 (35.5) 19 (61.3) 1 (3.2) 4.9 6.2 2,515 65.0 35.0 73.7 6.9 56 37 4 3

32 9 (28.1) 23 (71.9) 12 (37.5) 17 (53.1) 3 (9.4) 5.3 7.8 2,647 64.9 35.1 73.5 6.8 56 37 4 3

Four physicians were randomized as residents and then joined the faculty physician group. They remained in the group to which they were initially randomized. SD 5 standard deviation.

vention physicians (6.4%) and the usual care physicians (6.2%) (P 5 .87; OR 5 1.03, 95% CI 5 0.671.59). Finally, there was no signicant time trend during the study period in the proportion of ED visits with attempts at inappropriate prescribing in either study group, and there was no signicant interaction between groups. These ndings indicate that there was no signicant learning effect or contamination during the course of the study.

DISCUSSION Computer-assisted decision support for emergency physicians resulted in a statistically signicant reduction in prescribing of potentially inappropriate medications for older adults cared for in the ED. To the authors knowledge, this is the rst randomized, controlled trial of computer-assisted decision support to reduce potentially inappropriate prescribing in any clinical setting. In addition, real-time feed-

back was collected from physicians about their reasons for not following the decision support recommendations. The two most common reasons reported for rejecting the recommendations were that the patient had no previous problems with the medication and the prescription was a medication rell. These ndings demonstrate some of the cultural issues that similar intervention studies in EDs must address. In other words, some providers may be particularly reluctant to change chronic care treatment plans in situations in which the patients appear to be doing relatively well despite published literature of risk with the treatment. Specically, in the current study, the desire not to disrupt primary care practitioners regimens may have played a role. An unplanned strength of the study emanated from the exclusion of PGY-1 residents from the intervention. The analysis of only PGY-1 residents found no difference in prescribing between intervention and control physicians, suggesting that the behavior change among the intervention physicians who received the decision support was not due to unmeasured variables or a temporal trend. The ndings have important implications as the healthcare system places greater emphasis on the potential of information technology and quality improvement strategies are developed for the care of older adults. Patient-specic decision-support invoked at the time of clinical decision-making for medication use can be an effective method to improve quality of care for seniors. Two prior studies used computerized physician order entry with electronic alerts to reduce potentially inappropriate prescribing in the outpatient ofce setting. The rst reported a reduction in newly prescribed potentially inappropriate medications by 22% when medication rells were excluded.31 The second study specically targeted medication rells and found a 42% reduction in inappropriate prescribing from computerized alerts alone.32 The primary limitation of the prior studies was the use of time series study designs without concurrent control groups rather than a randomized, controlled trial. The major strengths of this investigation were its randomized, controlled trial design, its untested environment (i.e., it is the rst study to test the use of computerized decision support in an ED), the opportunity to receive realtime feedback from practitioners about their reasons for not following the decision support recommendations, and the serendipitous exclusion of PGY-1 residents. There are also several potential limitations of the current study. First, this was a single-site study targeting a small sample of academic physicians and residents who had access to a well-established health information technology infrastructure.29 The ndings may not be generalizable to other providers or

Table 2. Prescribing Outcomes of Physicians

Outcome Measure Control Intervention P-Value Odds Ratio (95% Condence Interval)z

Primary: visits with an inappropriate medication prescription, n (%) Secondary: prescriptions that were inappropriate, n (%)w

99 (3.9) 103 (5.4)

69 (2.6) 69 (3.4)

.02 .006

0.55 (0.340.89) 0.59 (0.410.85)

The analysis was calculated using a mixed-effect logistic regression with a random effect included for physician. The analysis was calculated using a mixed-effect logistic regression with random effects included for physician and patient nested within physician. z Odds of intervention physicians prescribing an inappropriate medication versus control physicians.

JAGS

AUGUST 2009VOL. 57, NO. 8

DECISION SUPPORT FOR INAPPROPRIATE PRESCRIBING

1393

Table 3. Potentially Inappropriate Medications Prescribed and the Effect of Decision Support for Each Medication
Intervention Physicians Initial Control Physicians, Prescription, n n Alternative Treatment Provided, n (%)

Inappropriate Medication

Promethazine Diphenhydramine Diazepam Propoxyphene with acetaminophen Hydroxyzine Amitriptyline Cyclobenzaprine Clonidine Indomethacin All inappropriate medications

40 15 10 9 9 8 7 4 1 103

32 22 18 8 15 1 5 3 10 114

19 (59) 8 (36) 5 (28) 2 (25) 6 (40) 0 (0) 2 (40) 2 (67) 5 (50) 49 (43)

potentially inappropriate medications and adverse events, no prior research has proven that the medications cause adverse effects. A related limitation is that the medications targeted are unsafe for most older adults, although they may be safe for some. Thus computer-assisted decision support must be viewed as a mechanism to support clinical judgment, not to replace it. In conclusion, computer-assisted decision support induced a signicant decrease in potentially inappropriate prescribing for older ED visitors. These ndings should be validated at other healthcare sites, such as inpatient settings and other EDs. Other next steps will be to conduct interviews with physicians who were in the intervention group to learn more about the episodes when they rejected the decision support and what could be done to increase acceptance of this type of intervention. Finally, another important future study will be to measure the effect of the decision support on patient-oriented outcomes.

Number of times that intervention physicians attempted to prescribe the

medication before receiving decision support. There were four visits to intervention physicians with an inappropriate prescription where decision support was not provided: one each for promethazine, diazepam, propoxyphene with acetaminophen, and amitriptyline.

other healthcare settings, although a survey conducted more than 6 years ago found that 35% of academic EDs had computerized prescribing systems,33 and the proportion is likely to increase. The results thus encourage the inclusion of decision-support capabilities in computerized order entry systems. Second, a 1.3% absolute difference in the rate of prescribing potentially inappropriate medications was found, which could be adjudged as a poor return on the investment in decision support, although once such a system is put in place, it costs essentially nothing to apply it to all prescriptions written in the ED. In addition, a 1.3% absolute difference becomes clinically important when applied to a high-volume ED. Third, because emergency physicians work side by side in the ED, there was potential for contamination (even though no evidence of contamination was identied in the analyses). Such contamination, if present, would tend to bias against nding an effect of the intervention. Similar to other studies of inappropriate medication use, this study was also limited in that it did not collect patient-oriented outcomes, such as subsequent ED visits for adverse drug events. The ndings of this study must also be viewed in the context of the debate in the literature about the true risk of many of these medications in older adults. Also, some of the suggested alternative treatments had limited evidence demonstrating that they were safer and at least as effective as the potentially inappropriate medications that they were to replace. To address this uncertainty at the outset of the study, a group of pharmacists and physicians with different backgrounds was assembled to employ their expert opinions when no guidance was available in the medical literature. Although available literature supports use of the Beers criteria because of the consistent associations between use of

ACKNOWLEDGMENTS We thank William M. Tierney, MD, of the Regenstrief Institute, Inc., Indianapolis, and the Department of Medicine, Indiana University School of Medicine, Indianapolis, for his review of this manuscript and his valuable suggestions. Conict of Interest: None of the authors have any potential nancial conicts of interest related to this research. This research was supported by Dr. Terrells Dennis W. Jahnigen Career Development Award, which is funded by the American Geriatrics Society, the John A. Hartford Foundation, and Atlantic Philanthropies Inc. Author Contributions: All authors participated in the studys concept and design, analysis and interpretation of data, and preparation of the manuscript. Sponsors Role: The sponsor did not play any role in the design, methods, data collection, analysis, or preparation of the manuscript.

REFERENCES
1. McCaig LF, Stussman BJ. National Hospital Ambulatory Medical Care Survey: 1996 emergency department summary. Adv Data 1997;293:120. 2. Nawar EW, Niska RW, Xu J. National Hospital Ambulatory Medical Care Survey: 2005 emergency department summary. Adv Data 2007;386: 132. 3. Gray SL, Lai KV, Larson EB. Drug-induced cognition disorders in the elderly: Incidence, prevention and management. Drug Saf 1999;21:101122. 4. Hohl CM, Dankoff J, Colacone A et al. Polypharmacy, adverse drug-related events, and potential adverse drug interactions in elderly patients presenting to an emergency department. Ann Emerg Med 2001;38:666671. 5. Budnitz DS, Pollock DA, Weidenbach KN et al. National surveillance of emergency department visits for outpatient adverse drug events. JAMA 2006;296:18581866. 6. Shelton PS, Fritsch MA, Scott MA. Assessing medication appropriateness in the elderly: A review of available measures. Drugs Aging 2000;16: 437450. 7. Fick DM, Cooper JW, Wade WE et al. Updating the Beers criteria for potentially inappropriate medication use in older adults: Results of a US consensus panel of experts. Arch Intern Med 2003;163:27162724. 8. Jano E, Aparasu RR. Healthcare outcomes associated with Beers criteria: A systematic review. Ann Pharmacother 2007;41:438447. 9. Chutka DS, Takahashi PY, Hoel RW. Inappropriate medications for elderly patients. Mayo Clin Proc 2004;79:122139. 10. Hustey FM. Beers criteria and the ED: An adequate standard for inappropriate prescribing? Am J Emerg Med 2008;26:695696.

1394

TERRELL ET AL.

AUGUST 2009VOL. 57, NO. 8

JAGS

11. Chin MH, Wang LC, Jin L et al. Appropriateness of medication selection for older persons in an urban academic emergency department. Acad Emerg Med 1999;6:12321242. 12. Fu AZ, Liu GG, Christensen DB. Inappropriate medication use and health outcomes in the elderly. J Am Geriatr Soc 2004;52:19341939. 13. Fick DM, Waller JL, Maclean JR et al. Potentially inappropriate medication use in a Medicare managed care population: Association with higher costs and utilization. J Manag Care Pharm 2001;7:407413. 14. Lau DT, Kasper JD, Potter DE et al. Hospitalization and death associated with potentially inappropriate medication prescriptions among elderly nursing home residents. Arch Intern Med 2005;165:6874. 15. Hustey FM, Wallis N, Miller J. Inappropriate prescribing in an older ED population. Am J Emerg Med 2007;25:804807. 16. Aparasu RR, Fliginger SE. Inappropriate medication prescribing for the elderly by ofce-based physicians. Ann Pharmacother 1997;31:823829. 17. Aparasu RR, Sitzman SJ. Inappropriate prescribing for elderly outpatients. Am J Health Syst Pharm 1999;56:433439. 18. Goulding MR. Inappropriate medication prescribing for elderly ambulatory care patients. Arch Intern Med 2004;164:305312. 19. Liu GG, Christensen DB. The continuing challenge of inappropriate prescribing in the elderly: An update of the evidence. J Am Pharm Assoc 2002;42: 847857. 20. Wilber ST, Gerson LW. A research agenda for geriatric emergency medicine. Acad Emerg Med 2003;10:251260. 21. Kawamoto K, Houlihan CA, Balas EA et al. Improving clinical practice using clinical decision support systems: A systematic review of trials to identify features critical to success. BMJ 2005;330:765768. 22. Garg AX, Adhikari NKJ, McDonald H et al. Effects of computerized clinical decision support systems on practitioner performance and patient outcomes: A systematic review. JAMA 2005;293:12231238. 23. Dexter PR, Perkins S, Overhage JM et al. A computerized reminder system to increase the use of preventive care for hospitalized patients. N Engl J Med 2001;345:965970. 24. Dexter PR, Perkins SM, Maharry KS et al. Inpatient computer-based standing orders vs physician reminders to increase inuenza and pneumococcal vaccination rates: A randomized trial. JAMA 2004;292:23662371. 25. Overhage JM, Tierney WM, Zhou XH et al. A randomized trial of corollary orders to prevent errors of omission. J Am Med Inform Assoc 1997;4: 364375. 26. 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;6:313321.

27. Sequist TD, Gandhi TK, Karson AS et al. A randomized trial of electronic clinical reminders to improve quality of care for diabetes and coronary artery disease. J Am Med Inform Assoc 2005;12:431437. 28. McDonald CJ, Tierney WM. The Medical GopherFa microcomputer system to help nd, organize and decide about patient data. West J Med 1986;145: 823829. 29. McDonald CJ, Overhage JM, Tierney WM et al. The Regenstrief medical record system: A quarter century experience. Int J Med Inform 1999;54:225 253. 30. Bates DW, Kuperman GJ, Wang S et al. Ten commandments for effective clinical decision support: Making the practice of evidence-based medicine a reality. J Am Med Inform Assoc 2003;10:523530. 31. Smith DH, Perrin N, Feldstein A et al. The impact of prescribing safety alerts for elderly persons in an electronic medical record: An interrupted time series evaluation. Arch Intern Med 2006;166:10981104. 32. Zillich AJ, Shay K, Hyduke B et al. Quality improvement toward decreasing high-risk medications for older veteran outpatients. J Am Geriatr Soc 2008;56:12991305. 33. Pallin D, Lahman M, Baumlin K. Information technology in emergency medicine residency-afliated emergency departments. Acad Emerg Med 2003; 10:848852.

SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article:
Appendix S1. Potentially Inappropriate Medications and Recommended Substitute Therapies. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.