Effect of an Electronic Medication Reconciliation Application and Process Redesign on Potential Adverse Drug Events

ORIGINAL INVESTIGATION

Effect of an Electronic Medication ReconciliationApplication and Process Redesign on PotentialAdverse Drug Events

A Cluster-Randomized Trial

Jeffrey L. Schnipper, MD, MPH; Claus Hamann, MD, MS; Chima D. Ndumele, MPH;Catherine L. Liang, MPH; Marcy G. Carty, MD, MPH; Andrew S. Karson, MD, MPH;Ishir Bhan, MD; Christopher M. Coley, MD; Eric Poon, MD, MPH; Alexander Turchin, MD, MS;Stephanie A. Labonville, PharmD, BCPS; Ellen K. Diedrichsen, PharmD; Stuart Lipsitz, ScD;Carol A. Broverman, PhD; Patricia McCarthy, PA, MHA; Tejal K. Gandhi, MD, MPH

Background: Medication reconciliation at transitionsin care is a national patient safety goal, but its effects onimportant patient outcomes require further evaluation.We sought to measure the impact of an information tech-nology–based medication reconciliation intervention onmedication discrepancies with potential for harm (po-tential adverse drug events [PADEs]).

Methods: We performed a controlled trial, random-ized by medical team, on general medical inpatient unitsat 2 academic hospitals from May to June 2006. We en-rolled 322 patients admitted to 14 medical teams, forwhom a medication history could be obtained before dis-charge. The intervention was a computerized medica-tion reconciliation tool and process redesign involvingphysicians, nurses, and pharmacists. The main out-come was unintentional discrepancies between pread-mission medications and admission or discharge medi-cations that had potential for harm (PADEs).

Results: Among 160 control patients, there were 230PADEs (1.44 per patient), while among 162 interven-

tion patients there were 170 PADEs (1.05 per patient)(adjusted relative risk [ARR], 0.72; 95% confidence in-terval [CI], 0.52-0.99). A significant benefit was foundat hospital 1 (ARR, 0.60; 95% CI, 0.38-0.97) but not athospital 2 (ARR, 0.87; 95% CI, 0.57-1.32) (P=.32 for testof effect modification). Hospitals differed in the extentof integration of the medication reconciliation tool intocomputerized provider order entry applications atdischarge.

Conclusions: A computerized medication reconcilia-tion tool and process redesign were associated with a de-crease in unintentional medication discrepancies with po-tential for patient harm. Software integration issues arelikely important for successful implementation of com-puterized medication reconciliation tools.

Trial Registration: clinicaltrials.gov Identifier:NCT00296426

Arch Intern Med. 2009;169(8):771-780

E FFORTS TO IMPROVE THE

quality and safety of healthcare include attention to un-intentional medication dis-crepancies, defined as un-

explained differences among documentedregimens across different sites of care (eg,prior to admission compared with hospi-tal admitting orders).1,2 Discrepancies arehighly prevalent; up to 67% of inpatientshave at least 1 unexplained discrepancy intheir prescription medication history atadmission.3

Because medication discrepancies arean important contributor to adverse drugevents (ADEs) among hospitalized and re-cently discharged patients,4-6 The JointCommission on Accreditation of Health-

care Organizations designated medica-tion reconciliation as a “National PatientSafety Goal” in 2005.7 Medication recon-ciliation is “a process of identifying themost accurate list of all medications a pa-tient is taking . . . and using this list to pro-vide correct medications for patients any-where within the health care system.”8

Hospitals have undertaken diverse ap-proaches to comply with The Joint Com-mission’s mandate. Some studies supportmedication reconciliation as a means to re-duce medication discrepancies or ADEs9-15;most are either pre-post studies or uncon-trolled evaluations of the types of errorsintercepted by the process. Few rigorousstudies demonstrate that medication rec-onciliation efforts improve important pa-

Author Affiliations are listed atthe end of this article.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM771

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

tient outcomes or define the best ways to implement theseprocesses or identify the patients most likely to benefit.We sought to determine the effects of a redesigned pro-cess for medication reconciliation, supported by infor-mation technology (IT), on potential ADEs (PADEs). Wehypothesized that our intervention would decrease PADEsin all patients and that patients at high risk for medica-tion discrepancies would benefit most.

METHODS

STUDY DESIGN, SETTING, AND PARTICIPANTS

The Partners Medication Reconciliation Study was a cluster-randomized controlled trial conducted from May 1 through June20, 2006, at 2 large academic hospitals in Boston, Massachu-setts. The design of this study has been described.16 Eligible pa-tients were admitted to one of several general medicine teamsand floors of each hospital, according to a rotating call cycle.Each team (6 at hospital 1 and 8 at hospital 2) consisted of 1attending physician, 1 junior or senior resident, 2 to 4 interns,

and 1 or 2 medical students. Patients were enrolled if study phar-macists (generally 1 pharmacist per weekday per hospital) hadtime to obtain a medication history prior to discharge. Pa-tients admitted to 1 of 7 randomly chosen medical teams andfloors were assigned to the intervention, while patients admit-ted to the other teams and on different floors received usualcare. Thus, patients in the 2 arms were cared for by differentphysicians and nurses. Randomization was stratified bystudy hospital and assigned by the principal investigator( J.L.S.) using random number generation in Microsoft Excel(Microsoft Corp, Redmond, Washington). Patients dis-charged from a nonstudy team or floor, patients transferredbetween a control team or floor and an intervention team orfloor, and patients discharged after June 20, 2006, wereexcluded. The study was approved by the institutionalreview board of Partners HealthCare, Boston; patient con-sent was deemed unnecessary.

INTERVENTION

The intervention consisted of an IT application designed to fa-cilitate medication reconciliation, integrated into the inter-

Hospital stage Nurse Physician Pharmacist

Admission

Duringhospitalization

Discharge

Confirms accuracy of PAML during admission assessment, identifies discrepancies, notifies admitting physician

Takes medication history: collects sources of preadmission medication data, interviews patient/family, creates and documents PAML

Chooses and documents planned action on admission for each medication in PAML

Writes admission orders

Resolves discrepancies with nurse and pharmacist, updates PAML and inpatient orders as necessary

Reviews PAML, planned actions on admission, and admission orders, confirms reconciliation, identifies important discrepancies, notifies admitting physician

Gathers additional sources of PAML information, helps resolve uncertainties or discrepancies in PAML, notifies responsible physician

Gathers additional sources of PAML information, resolves uncertainties or discrepancies in PAML with help of nurse and/or pharmacist

Gathers additional sources of PAML information, helps resolve uncertainties or discrepancies in PAML, notifies responsible physician

Updates PAML and inpatient orders as necessary

Reviews PAML, current medications, and discharge orders, confirms reconciliation

Reviews discharge medications with patient/caregiver

Identifies discrepancies, notifies discharging physician

Reviews PAML and current medications, creates discharge orders, documents reconciliation

Resolves any remaining discrepancies with nurse, updates discharge orders as necessary, documents discharge medications and changes from prior to admission

Figure 1. Redesigned medication reconciliation workflow. Adapted from Poon et al17 with permission. PAML indicates preadmission medication list.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM772

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

nally developed computerized provider order entry (CPOE) sys-tems at the 2 hospitals, and process redesign involvingphysicians, nurses, and pharmacists.

IT APPLICATION: THE PREADMISSIONMEDICATION LIST BUILDER

The medication reconciliation application, the PreadmissionMedication List (PAML) Builder, has been previously de-scribed.17 It is a Web-based application that promotes the cre-ation of a preadmission medication list from several electronicsources (including 2 ambulatory electronic medical record sys-tems used at Partners HealthCare and discharge orders fromthe 2 study hospitals), documents a planned action on admis-sion for each PAML medication (eg, continue on admission,discontinue), facilitates review of a completed PAML and ad-mission medications by a second clinician, and facilitates rec-onciliation of the PAML with current inpatient medications whendischarge orders are written. Features of the PAML Builder ap-plication include the following:

v Creation of the preadmission medication list1. Displays “medications from electronic sources,” com-

prising the active medications from the 2 ambulatoryelectronic medical records used at the 2 study hospi-tals and the medications ordered at the most recent dis-charge from the study hospitals.

2. Allows the admitting clinician to move selected medi-cations from electronic sources into the PAML, withor without changes in dose or frequency; to add newmedications (ie, not available in electronic sources)based on the admission medication history; and to up-date the PAML during the hospitalization as more ac-curate information becomes available.

v Reconciliation of medications at admission1. Requires one of the following planned actions on ad-

mission for every medication in the PAML as prepa-ration for writing admission orders: continue as writ-ten, discontinue, continue at different dose/frequency/route, or substitute with a different medication.

2. Facilitates confirmation of reconciliation by a clinicalpharmacist.

v Reconciliation of medications at discharge1. Displays the PAML and current inpatient medication

orders at the time discharge orders are being written.2. Requires confirmation that reconciliation of the PAML

with discharge medications has taken place.

At study time, the PAML Builder was not fully integrated withthe CPOE system at either study hospital, so PAML medicationscould not automatically become CPOE admission orders. At dis-charge, hospital 1 displayed current inpatient medications along-side preadmission medications with the ability to order medica-tions from either list. At hospital 2, these 2 lists existed in separatedisplay windows, and preadmission medications needed to be or-dered separately to be resumed at discharge.

PROCESS REDESIGN

To implement medication reconciliation successfully, roles andworkflows of residents, staff physicians, nurses, and pharma-cists at the 2 hospitals were redesigned (Figure 1). Physi-cians were given primary responsibility for taking preadmis-sion medication histories and referring to this list when orderingmedications in the hospital and at discharge. Pharmacists wereresponsible for confirming the reconciliation process at admis-sion, while nurses were responsible for confirming reconcili-ation at discharge. Most notably, we replaced redundant medi-

cation history taking performed independently and withoutcommunication with interdisciplinary collaboration to obtainthe most accurate medication history possible; we also used pur-poseful cross-checking to increase compliance (Figure 1).

Training in process redesign and in the use of the PAML Builderwas similar at the 2 hospitals and generally consisted of 30 min-utes to 2 hours of small-group training sessions, computer-based training, and e-mail announcements, supplemented by edu-cational handouts, online materials, and IT staff support. Timingof the trial differed in relation to hospitalwide rollout of medica-tion reconciliation at the hospitals. At hospital 1, this study wasconducted during the first phase of a phased hospitalwide roll-out, so that clinicians were exposed to a wide range of materialspublicizing the importance of medication reconciliation (eg, aninformation booth stationed in the hospital’s main hallway, groupand individual e-mails to clinicians, and presentations to hospi-tal leadership). At hospital 2, the study was conducted prior toan all-at-once hospitalwide rollout, with consequently less pub-licity about medication reconciliation.

Patients in the usual care study arm received the preinterven-tion standard of care. Residents documented medication histo-ries in admission notes; pharmacists reviewed medication or-ders for appropriateness. At discharge, physicians wrote dischargeorders (without facilitated access to preadmission medication his-tories); nurses educated patients about their medications.

OUTCOMES

The main study outcome was the number of unintentional medi-cation discrepancies with potential for causing harm (PADEs)per patient. Defined as “incidents with potential for injury re-lated to a drug,”18 PADEs have been used as a medication safetymeasure in numerous studies,18-22 and reductions in PADEs dueto interventions tend to track with reductions in actual ADEs.23

Our process for outcome assessment has been previouslydescribed.16 A “gold standard” preadmission medication his-tory was taken of all study patients by 1 of 2 study pharma-cists at each hospital, following a strict protocol but not blindedto intervention status. The resulting preadmission medicationlist was compared with the medication history taken by the medi-cal team, with all admission medication orders, and with alldischarge orders. Discrepancies between the gold standard pread-mission medication history and admission or discharge orderswere identified, and intentional reasons for changes were soughtfrom the medical record. If necessary and when possible, phar-macists communicated directly with the medical team after dis-charge orders were written to verify intent. Medication dis-crepancies that were not clearly intentional were recorded.

Recorded discrepancies were shown by the study pharmacistto rotating adjudication teams of 2 physicians (from a pool of 6)blinded to intervention status. Each medication discrepancy andthe patient discharge summary were reviewed. Additional elec-tronic patient information such as ordered medications and testresults were reviewed as needed. Using an expert-derived classi-fication scheme,24 the 2 physicians recorded whether each medi-cation discrepancy was intentional and, if unintentional, the time(admission vs discharge) and type of discrepancy (eg, omission,change in dose). An error in preadmission medication history wasrecorded as a “history error” (eg, not including aspirin on thepreadmission medication list, thus explaining why it is not or-dered at discharge). Conversely, an error of reconciling the medi-cation history with medication orders was recorded as a “recon-ciliationerror” (eg, aspirin therapynot restartedatdischargedespitebeing on the preadmission medication list and clinically indi-cated at discharge). Independently, the 2 reviewers judged thepotential for harm for each unintentional discrepancy and its po-tential severity, as in previous studies.18 All disagreements were

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM773

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

resolved by discussion and by a third adjudicator if necessary(needed in 86 of 4700 adjudicated medication discrepancies[1.8%]).

Weekly meetings were conducted to ensure consistency be-tween sites and among study pharmacists and physician adju-dicators. Reliability of the gold standard preadmission medi-cation histories and physician adjudication of outcomes havebeen shown to be moderate to high.16 Prespecified secondaryoutcomes, based on hospital administrative data, included emer-gency department visits and hospital readmissions within 30days of discharge.

STATISTICAL ANALYSIS

Patient characteristics and study results were calculated using pro-portions, means with standard deviations, and medians with in-terquartile ranges. Poisson log-linear regression was used to de-termine associations between the number of PADEs per patientand study arm. To adjust for potential confounding, a weightedpropensity score approach was used in which the data for eachpatient were weighted by the inverse of the probability of beingin the treatment arm given each of the potential confounding co-variates.25 Covariates, chosen a priori on clinical grounds, in-cluded patient age, number of outpatient visits within the previ-ousyear, inpatient admissions to indexhospitalwithin thepreviousmonth, number of preadmission medications, number of high-risk preadmission medications, admission source, primary carephysician (PCP) from within the hospital network, whether thePCP was the discharging physician, family or caregiver as a sourceof preadmission medication information, level of training of ad-mitting physician, and patients’ understanding of their medica-tions (subjectively categorized by the study pharmacist as low,medium, or high). High-risk medication classes, based on theirrisk for causing PADEs in the control group when prescribed, in-

cluded gout medications, muscle relaxants, hyperlipidemic medi-cations, antidepressants, and respiratory medications.16 Health-care utilization outcomes were compared using logistic regressionwith general estimating equations clustered by admittingphysician.

Subgroup analyses explored possible effect modification; pre-specified subgroups included study hospital, PADE risk score,number of preadmission medications, transfers from outsideinstitutions, level of training of admitting physician, and pa-tient understanding of medications. The PADE risk score, de-rived from the control group, was calculated by assigning 2points for patients younger than 85 years and 1 point each forlow or medium patient understanding of medications, having16 or more preadmission medications, 4 or more high-risk pread-mission medications (as defined in the previous paragraph),13 or more outpatient visits in the previous year, and having afamily member or caregiver as a source of preadmission medi-cation information.16 Interaction terms (ie, subgroup�studyarm) were entered into secondary models to evaluate the sta-tistical significance of any effect modification.

Generalized estimating equations, using a robust covariancematrix, were applied to adjust for clustering of results by the ad-mitting physician. Model fit for the propensity score model of theprimary outcome was assessed based on aggregates of residu-als26 using the ASSESS statement in SAS statistical software (SASInstitute Inc, Cary, North Carolina), with a P value computed basedon 10 000 simulated paths (P=.60, suggesting good model fit).Analyses were intention to treat. P� .05 (2 sided) was consid-ered significant. Analyses were implemented with SAS statisticalsoftware, version 9.1 (SAS Institute Inc).

Our target sample comprised 460 patients, which was es-timated to provide a 90% power to detect a 60% relative de-crease (based on studies of paper-based medication reconcili-ation9,10) in the presence of any PADE (from 27%5,6,11 to 11%

801 Patients on 14 medical teams assessed for eligibility 479 Excluded

443 Met exclusion criteria397 Discharged prior to completion of

“gold standard” medication history10 Discharged from nonstudy floor12 Transferred between study arms1 Died prior to completion of

medication history3 Discharged after end of study

37 Did not meet inclusion criteria37 Never admitted to study team or

floor0 Declined to participate9 Other reasons

322 Patients on 14 medical teams randomized by team

162 Patients on 7 medical teams allocated to intervention

162 Received intervention162 Received process change

education160 Used PAML Builder75 Completed PAML within

24 hours0 Did not receive intervention

0 Lost to follow-up0 Discontinued intervention

162 Analyzed0 Excluded from analysis

160 Patients on 7 medical teams allocated to usual care

160 Received usual care

0 Lost to follow-up

160 Analyzed0 Excluded from analysis

Figure 2. Trial flow diagram. PAML indicates preadmission medication list.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM774

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

of patients), assuming an � level of .05, 5 patients per admit-ting physician, and an intraclass correlation coefficient by phy-sician of 0.10 (we did not adjust for additional correlation atthe team level). The study was not powered to detect a differ-ence in health care utilization.

RESULTS

DESCRIPTION OF STUDY SAMPLE

Of 801 patients originally identified as potentially eli-gible for the study, 322 were enrolled and were cared forby each of the 14 randomized medical teams and by 117different admitting physicians. The most common rea-son for patient exclusion was lack of time of the studypharmacist to obtain a medication history before dis-charge. Compared with enrolled patients, unenrolled pa-tients had shorter lengths of stay (a table of the charac-teristics of the study sample and the excluded subjectsis available from the corresponding author).

Of 162 patients assigned to the intervention arm, thePAML Builder application was used in 160 patients (99%).

A PAML was “completed” (planned actions on admis-sion assigned for all medications and the list signed offas “ready for review” by a pharmacist) within 24 hoursof admission for 75 patients (46%), including 42 of 84patients (50%) at hospital 1 and 33 of 78 patients (42%)at hospital 2 (P=.35 for comparison). A PAML was com-pleted prior to discharge in 121 patients (75%). The pri-mary outcome could be evaluated in all 322 patients.Figure 2 shows the study flow. Patient characteristicsin the 2 study arms were similar, except for a signifi-cantly lower proportion of PCPs within the PartnersHealthCare network and a higher proportion of non-medicine interns in the intervention group (Table 1).Some differences were also noted by study hospital (a tableof the site differences in baseline patient characteristicsis available from the corresponding author).

EFFECT OF THE INTERVENTION ON PADEs

Among the 162 patients assigned to the intervention, therewere 170 unintentional medication discrepancies withpotential for patient harm (1.05 PADEs per patient) vs

Table 1. Baseline Characteristics of Study Population

Characteristic

Total Population Hospital 1 Hospital 2

Intervention(n=162)

Control(n=160)

Intervention(n=84)

Control(n=86)

Intervention(n=78)

Control(n=74)

Age, y�50 35 (22) 35 (22) 17 (20) 15 (17) 18 (23) 20 (27)50-59 27 (17) 25 (15) 15 (18) 18 (21) 12 (15) 7 (9)60-74 49 (30) 35 (22) 26 (31) 16 (19) 23 (30) 19 (26)75-84 34 (21) 48 (30) 17 (20) 29 (34) 17 (22) 19 (26)�85 17 (10) 17 (11) 9 (11) 8 (9) 8 (10) 9 (12)

Female 84 (52) 92 (57) 37 (44) 46 (54) 47 (60) 46 (62)Median income by zip code, $

�39 000 37 (23) 31 (19) 19 (23) 12 (14) 18 (23) 19 (26)39 001-47 000 40 (25) 43 (27) 21 (25) 25 (29) 19 (24) 18 (24)47 001-63 000 48 (29) 36 (23) 23 (27) 19 (22) 25 (32) 17 (23)�63 000 37 (23) 50 (31) 21 (25) 30 (35) 16 (21) 20 (27)

InsurancePrivate 50 (31) 43 (27) 19 (23) 25 (29) 31 (40) 18 (24)Medicare only 4 (3) 7 (5) 3 (4) 2 (2) 1 (1) 5 (7)Medicare with secondary 76 (47) 79 (49) 45 (54) 44 (51) 31 (40) 35 (47)Free care/Medicaid 25 (15) 26 (16) 15 (17) 13 (16) 10 (13) 13 (18)Other/self-pay 7 (4) 5 (3) 2 (2) 2 (2) 5 (6) 3 (4)Medicare under age 65 y 18 (11) 11 (7) 13 (15) 8 (9) 5 (6) 3 (4)PCP within Partners

HealthCare networka,b81 (50) 98 (61) 48 (57) 53 (62) 33 (42) 45 (61)

Preadmission sourceEmergency department 106 (65) 96 (60) 57 (68) 49 (57) 49 (63) 47 (63)Transfer from other service 17 (10) 15 (9) 0 (0) 0 (0) 17 (22) 15 (20)Transfer from outside

institution16 (10) 23 (14) 6 (7) 14 (16) 10 (13) 9 (12)

Scheduled from home 9 (6) 11 (7) 8 (10) 9 (11) 1 (1) 2 (3)Day procedure 14 (9) 15 (9) 13 (15) 14 (16) 1 (1) 1 (1)

Admission DRG weight,c median(IQR)

0.99 (0.76-1.27) 1.03 (0.83-1.28) 0.99 (0.77-1.24) 1.02 (0.83-1.22) 1.01 (0.73-1.26) 1.03 (0.82-1.44)

Charlson Comorbidity Scored

0-1 54 (34) 46 (29) 21 (26) 21 (26) 33 (43) 25 (34)2-3 49 (31) 53 (34) 28 (35) 24 (29) 21 (27) 29 (39)4-5 25 (16) 21 (13) 13 (16) 10 (12) 12 (15) 11 (15)�6 31 (19) 36 (23) 19 (23) 27 (33) 12 (15) 9 (12)

(continued)

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM775

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

230 (1.44 PADEs per patient) among the 160 patientsassigned to usual care, an adjusted 28% relative risk re-duction (Table 2). Ninety-eight PADEs were consid-ered serious, ie, to have potential to cause serious harmsuch as rehospitalization or persistent alteration in healthfunction, including 43 PADEs in the intervention arm(0.27 per patient) and 55 PADEs in those assigned to usualcare (0.34 per patient). The intervention was associatedwith a significant reduction in PADEs at discharge butnot at admission (Table 2). Table 3 gives examples ofdifferent types of PADEs identified during the study.

No significant differences were found in health care uti-lization. The rate of hospital readmission or emergency de-partment visit within 30 days was 20% in the interventionarm and 24% in the usual care arm (clustered odds ratio,0.76; 95% confidence interval [CI], 0.43-1.35).

In subgroup analyses, we found effect modification byPADE risk score and a suggestion of an effect modifica-tion by hospital. The effect of the intervention was greaterin the 167 patients with a PADE risk score of 4 or higher(adjusted and clustered relative risk, 0.62; 95% CI, 0.41-0.93) than in the 155 patients with a risk score of 3 or

lower (adjusted and clustered relative risk, 1.09; 95% CI,0.49-2.44) (P value for interaction, .02). The interven-tion was associated with a significant reduction in PADEsat hospital 1 but not at hospital 2 (P value for interac-tion, .32) (Table 4). Differences between hospitals werequalitatively similar for the adjusted effect of the inter-vention on PADEs due to history errors, those due to rec-onciliation errors, PADEs at admission, and those at dis-charge (Table 4). No other subgroup differences werefound in the effect of the intervention.

COMMENT

In this 2-hospital cluster-randomized controlled trial, wefound that a medication reconciliation intervention con-sisting of novel IT and process redesign involving phy-sicians, nurses, and pharmacists was associated with a28% relative risk reduction in unintentional medicationdiscrepancies with potential for harm, a type of PADE.The absolute risk reduction between the 2 arms was 0.39PADE per patient or a number needed to treat 2.6 pa-

Table 1. Baseline Characteristics of Study Population (continued)

Characteristic

Total Population Hospital 1 Hospital 2

Intervention(n=162)

Control(n=160)

Intervention(n=84)

Control(n=86)

Intervention(n=78)

Control(n=74)

Total No. of preadmission medicationsb,e

Quartile 1 (1-8) 45 (28) 42 (26) 23 (27) 27 (31) 22 (28) 15 (21)Quartile 2 (9-12) 54 (33) 38 (24) 28 (33) 20 (23) 26 (33) 18 (24)Quartile 3 (13-17) 37 (23) 38 (24) 19 (23) 18 (22) 18 (23) 20 (27)Quartile 4 (18-33) 26 (16) 42 (26) 14 (17) 21 (24) 12 (16) 21 (28)

No. of high-risk preadmission medicationsf

Quartile 1 (0-1) 37 (23) 43 (27) 15 (18) 27 (31) 22 (28) 16 (21)Quartile 2 (2-3) 42 (26) 34 (21) 25 (30) 15 (17) 17 (22) 19 (26)Quartile 3 (4-5) 44 (27) 42 (26) 22 (26) 22 (26) 22 (28) 20 (27)Quartile 4 (�6) 39 (24) 41 (26) 22 (26) 22 (26) 17 (22) 19 (26)

Medical student involvement in patient care 18 (11) 12 (8) 8 (10) 6 (7) 8 (10) 6 (8)Experience level of admitting physiciana,b,g

Medicine intern 124 (77) 104 (65) 70 (84) 64 (75) 54 (69) 40 (55)Medicine resident 13 (8) 32 (20) 11 (13) 8 (9) 2 (3) 24 (32)Attending/fellow physician 4 (2) 8 (5) 1 (1) 5 (6) 3 (4) 3 (4)ED/Ob-Gyn intern 18 (11) 1 (1) 0 (0) 0 (0) 18 (23) 1 (1)Unknown 3 (2) 15 (9) 2 (2) 9 (10) 1 (1) 6 (8)

Patient understanding of preadmission medicationsh

High 51 (32) 55 (35) 36 (43) 35 (41) 15 (20) 20 (27)Medium 79 (50) 71 (44) 32 (39) 35 (18) 47 (18) 36 (49)Low 29 (18) 33 (21) 15 (18) 16 (41) 14 (62) 17 (24)

Previous hospitalization in the last 31 d 33 (20) 30 (19) 17 (20) 20 (23) 16 (21) 10 (14)No. of outpatient visits in the last 12 mo

0 40 (25) 36 (22) 15 (18) 14 (16) 25 (32) 22 (30)1-4 52 (32) 37 (23) 28 (33) 21 (24) 24 (31) 16 (22)5-11 40 (25) 41 (26) 24 (29) 24 (28) 16 (20) 17 (23)�11 30 (18) 46 (29) 17 (20) 27 (32) 13 (17) 19 (25)

Abbreviations: DRG, diagnosis-related group; ED, emergency department; IQR, interquartile range; Ob-Gyn, obstetrics/gynecology; PCP, primary care physician.aP� .05 for comparison between intervention and control groups (total population).bP� .05 for comparison between intervention and control groups at hospital 2.cBased on Medicare 2006 weights (version 23).27

dBased on administrative billing codes.28

eExcluding “as needed” medications and topical agents.f In 5 medication classes most likely to cause potential adverse drug events when prescribed: gout medications, muscle relaxants, hyperlipidemic agents,

antidepressants, and respiratory medications.16

gP� .05 for comparison between intervention and control groups at hospital 1.hBased on pharmacist assessment.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM776

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

tients to prevent 1 PADE. The intervention was more suc-cessful in patients at high risk for medication discrep-ancies and possibly more successful at hospital 1 than athospital 2. To our knowledge, this is the first random-ized controlled trial of an IT-based medication recon-ciliation intervention.

We believe our intervention was successful becauseit combined effective process redesign with IT. The newreconciliation process encouraged interdisciplinary com-munication and cross-checks. The PAML Builderapplication facilitated accurate medication historiesby presenting several sources of available medication in-formation, and it displayed the PAML with current in-patient medications during the discharge orderingprocess.

However, our intervention was far from perfect ineliminating potentially harmful medication discrepan-cies (1.05 PADEs per patient in the intervention arm).Possible reasons include incomplete and inaccurate elec-tronic sources of ambulatory medications, lack of pa-tient and caregiver knowledge of preadmission medica-tion regimens, lack of clinician adherence with thereconciliation process, and software usability issues (suchas ease of adding new medications to the PAML and lackof integration with the admission ordering process).

Our intervention was more successful in patients athigh risk for medication discrepancies, based on a riskscore derived from the control group. If prospectively vali-dated in other populations, this score could prove use-ful in prioritizing those most in need of intensive recon-ciliation efforts, ie, above the minimum Joint Commissionstandard.7

The intervention reduced PADEs at discharge but notat admission. This result was likely because of the fol-lowing circumstances: (1) PADEs in general were morecommon at discharge than at admission16; (2) most er-rors of reconciliation occurred at discharge, and the PAMLBuilder may have been particularly effective at reducingthese kinds of errors; and (3) delays in completing a PAMLwould also have attenuated the effectiveness of the in-tervention at admission but not at discharge.

A significant reduction in PADEs with the interven-tion at hospital 1 but not at hospital 2 could have beendue to chance, since the study was not powered to de-tect a difference in PADEs at each hospital individually.Alternatively, part of the answer may be the phased hos-pitalwide rollout of the intervention at hospital 1, whichled to more publicity about medication reconciliation,possibly resulting in greater compliance with the new pro-cesses. Hospital differences related to reducing “medi-cation history errors” may be due in part to greater in-volvement of nurses at admission at hospital 1. Based oninformal, unblinded interviews with clinical personnelconducted after study completion, nurses at hospital 1more consistently confirmed PAML accuracy at admis-sion compared with hospital 2. Hospital differences inreducing “reconciliation errors” were likely due to dif-ferences in discharge medication computer screens de-veloped at the 2 hospitals—a hypothesis we generated apriori. At hospital 1, it was easier to view inpatient andpreadmission medications simultaneously and to ordermedications from either list at discharge.

Regarding other studies of medication reconcilia-tion, 2 recent pre-post studies of paper-based interven-tions evaluated effects on actual ADEs based on retro-spective review of a random selection of medical records,demonstrating a 43% to 84% relative risk reduction.9,10

A recent randomized controlled trial of a pharmacist-run “medication reconciliation and seamless care ser-vice” found that 56.3% of 119 control patients had a drugtherapy inconsistency or omission at discharge on ret-rospective medical record review compared with 3.6%of intervention patients (based on an independent re-view of 17% of intervention patients’ medical records bya second pharmacist).29 Mostly descriptive studies of IT-based medication reconciliation tools are also begin-ning to be published,30-34 and commercial vendors arestarting to provide medication reconciliation modules intheir applications.30,35

We believe IT-based medication reconciliation inter-ventions have several advantages over paper-based so-lutions, including the ability to use existing electronic

Table 2. Incidence and Relative Rates of Potential Adverse Drug Events Due to Unintentional Medication Discrepancies

Outcome

PADEs, No. (per Patient)in the Control Arm

(n=160)

PADEs, No. (per Patient)in the Intervention Arm

(n=162)Unadjusted RR

(95% CI)Adjusted and Clustered RR

(95% CI)a,b

All PADEs 230 (1.44) 170 (1.05) 0.74 (0.60-0.89) 0.72 (0.52-0.99)PADEs by type of errorc

History errors 153 (0.96) 125 (0.77) 0.81 (0.64-1.02) 0.80 (0.55-1.15)Reconciliation errors 80 (0.50) 52 (0.32) 0.64 (0.45-0.91) 0.62 (0.29-1.34)

PADEs by time of occurrencePADEs at admission 49 (0.31) 44 (0.27) 0.89 (0.59-1.33) 0.87 (0.51-1.52)PADEs at discharge 181 (1.13) 126 (0.78) 0.69 (0.55-0.86) 0.67 (0.49-0.98)

Abbreviations: CI, confidence interval; RR, relative risk.aAdjusted for clustering by admitting physician using general estimating equations.bPropensity score–adjusted model. Propensity score based on patient age, number of outpatient visits within the previous year, inpatient admissions to the

index hospital within the previous month, number of preadmission medications, number of high-risk preadmission medications, admission source, primary carephysician from within the hospital network, whether the primary care physician was the discharging physician, family or caregiver as a source of preadmissionmedication information, level of training of the physician documenting the medication history, and patient understanding of their medications. Except for patientunderstanding of their medications, which was dichotomized as high vs medium or low, all other variables were categorized as in Table 1.

cThree PADEs in the control arm and 7 PADEs in the intervention arms were attributed to both history error and reconciliation error.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM777

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

sources of ambulatory medication information, better in-tegration into workflow in hospitals with CPOE, easiersharing of reconciliation information across providers,automatic production of documentation for dischargesummaries, comparisons of medication lists to facilitatereconciliation and patient education, provision of alertsand reminders to ensure compliance, and ability to trackcompliance to inform further process improvement.

This study has several limitations. First, because it wasconducted on general medical services at academic hos-pitals, results may not be generalizable to other settings.Patients with very short lengths of stay may have beendisproportionately discharged before enrollment, lead-ing to selection of a sicker patient population; the re-sults may not be generalizable to less sick patients or those

taking fewer medications. Second, the study measuredpotential and not actual ADEs, and while we analyzedhealth care utilization, the study was not powered to de-tect a reduction in this outcome. Third, full use of thePAML Builder was not achieved: only 46% of patients hada completed PAML within 24 hours of admission (al-though 75% were complete by discharge), thus limitingthe ability of the intervention to benefit patients. Basedon an analysis of clinician attitudes and patterns of useof the application, we attribute this nonadherence to thelack of integration of the application with CPOE at ad-mission.36 This problem has since been remedied; cur-rently 80% to 90% of PAMLs are complete within 24 hoursof admission. Fourth, we cannot exclude the possibilityof unmeasured provider characteristics confounding the

Table 3. Examples of Potential Adverse Drug Eventsa

History Timing of Error Type of Error Reason for Error Potential Severity

A patient with multiple cardiac risk factors was admitted foratypical chest pain. At home he was taking atenolol,100 mg by mouth daily. The medical team accuratelydocumented this medication in the preadmissionmedication list but did not prescribe it (or any other�-blocker) on admission or any time during thehospitalization. No medical reason could be found forwithholding it, and notes document a plan to continueoptimal medical management of possible coronarydisease. The patient was discharged on his home dose ofatenolol.

Admission Omission Reconciliation error Serious

A patient with a seizure disorder was admitted for peripheraledema and COPD exacerbation. At home the patient wastaking carbamazepine, 100 mg by mouth twice daily, butthe medical team incorrectly documented 100 mg bymouth 3 times a day in the preadmission medication listand then prescribed that frequency on admission. Noblood levels of the medication were drawn during thehospitalization.

Admission Frequency History error Serious

A patient with a history of asthma was admitted for anasthma exacerbation. At home, the patient was using analbuterol inhaler, 2 puffs every 4 hours as needed forshortness of breath. The team accurately documented themedication in the preadmission medication list. During thehospitalization the patient was given albuterol bynebulizer. At discharge, no albuterol (or any othershort-acting bronchodilator) was prescribed.

Discharge Omission Reconciliation error Serious

A patient with gastroesophageal reflux disease was admittedwith nausea and vomiting and was found to havehyponatremia. At home, the patient was takingomeprazole 20 mg by mouth daily, but the medical teamdid not document this medication in the preadmissionmedication list. During hospitalization, the patient wasprescribed esomeprazole for stress ulcer prophylaxis. Atdischarge, no proton pump inhibitor or other antirefluxmedication was prescribed.

Discharge Omission History error Significant

A patient with a history of myocardial infarction andhypertension was admitted for a gastrointestinal bleed. Athome the patient was taking lisinopril, 10 mg by mouthdaily, but the medical team incorrectly documented 25 mgby mouth daily in the preadmission medication list.During the hospitalization the patient was hypertensive,and the team appropriately ordered captopril instead oflisinopril to manage his blood pressure. At discharge, witha normal blood pressure, the team ordered what theythought was his home dose of lisinopril, 25 mg by mouthdaily.

Discharge Dose History error Serious

Abbreviation: COPD, chronic obstructive pulmonary disease.a In all cases, errors were corrected prior to finalization of the discharge orders.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM778

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

results, but the factors we adjusted for had no effects onstudy findings. Finally, hypotheses about the effects ofthe intervention’s rollout or the involvement of nursesin the medication history process at the 2 hospitals weregenerated post hoc and may be biased.

In conclusion, an interdisciplinary medication recon-ciliation intervention comprising novel IT and processredesign was associated with a significant reduction inunintentional medication discrepancies with potential forharm. Institutions should strongly consider adopting elec-tronic medication reconciliation tools as availability in-creases. Site-specific differences suggest that electronicmedication reconciliation tools should facilitate com-parisons of medication lists at transition points and useof these lists to order medications for the next care set-ting. Provider education on taking complete medica-tion histories and purposeful “independent redundan-cies”12 in the reconciliation process (eg, nurses verifyingthe accuracy of physician-produced medication histo-ries) are also likely important to the success of any medi-cation reconciliation effort. Future research should be di-rected at more rigorous evaluations of the environments,medication reconciliation interventions, and implemen-tation characteristics that best improve outcomes and atfurther development and evaluation of commercially avail-able electronic medication reconciliation tools. Ideally,multicenter studies using methods such as randomizedcontrolled trials or interrupted time series analyses shouldbe conducted using more downstream health outcomes(such as total ADEs and hospital readmissions). Morework is needed to eliminate serious medication recon-ciliation errors and make transitions in care as safe aspossible.

Accepted for Publication: December 6, 2008.Author Affiliations: Brigham and Women’s AcademicHospitalist Service (Drs Schnipper and Carty), Divisionof General Medicine (Drs Schnipper, Carty, Poon, Lip-sitz, and Gandhi and Mr Ndumele and Ms Liang), Cen-ter for Clinical Excellence (Drs Carty and Gandhi), andPharmacy Services (Dr Labonville), Brigham and Wom-en’s Hospital, Boston, Massachusetts; Department of Medi-cine (Dr Hamann, Karson, Bhan, and Coley), GeriatricMedicine Unit (Dr Hamann), Center for Quality and Safety(Dr Karson and Ms McCarthy), and Pharmacy Services

(Dr Diedrichsen), Massachusetts General Hospital, Boston;Partners Information Systems Clinical InformaticsResearch and Development, Boston (Drs Poon, Turchin,and Broverman); and Harvard Medical School, Boston(Drs Schnipper, Hamann, Carty, Karson, Bhan, Coley,Poon, Turchin, Lipsitz, and Gandhi).Correspondence: Jeffrey L. Schnipper, MD, MPH, Divi-sion of General Medicine, Brigham and Women’s Hos-pital, 1620 Tremont St, Boston, MA 02120-1613([email protected]).Author Contributions: Dr Schnipper had full access toall the data in the study and takes responsibility for theintegrity of the data and the accuracy of the data analy-sis. Study concept and design: Schnipper, Carty, Karson,Poon, Broverman, and Gandhi. Acquisition of data:Schnipper, Hamann, Carty, Karson, Bhan, Labonville, Die-drichsen, and Gandhi. Analysis and interpretation of data:Schnipper, Hamann, Ndumele, Liang, Coley, Poon, Tur-chin, Lipsitz, McCarthy, and Gandhi. Drafting of the manu-script: Schnipper. Critical revision of the manuscript forimportant intellectual content: Hamann, Ndumele, Li-ang, Carty, Karson, Bhan, Coley, Poon, Turchin, Labon-ville, Diedrichsen, Lipsitz, Broverman, McCarthy, andGandhi. Statistical analysis: Schnipper, Ndumele, and Lip-sitz. Obtained funding: Schnipper, Coley, Broverman, andGandhi. Administrative, technical, and material support:Liang, Karson, Coley, and McCarthy. Study supervision:Broverman.Financial Disclosure: None reported.Funding/Support: This study was funded in part by aninvestigator-initiated grant from the Harvard Risk Man-agement Foundation, including compensation for Elisa-beth Burdick, MS, Amy Bloom, MPH, and Emily Barsky,BA, as well as internal funding from Brigham and Wom-en’s Hospital (BWH), Massachusetts General Hospital,and Partners HealthCare. Dr Schnipper was supportedby a mentored clinical scientist award from the Na-tional Heart, Lung, and Blood Institute (K08 HL072806).Role of the Sponsors: The funding organizations had norole in the design and conduct of the study; collection,management, analysis, and interpretation of the data; andpreparation, review, or approval of the manuscript.Previous Presentations: Portions of this work werepresented as a poster at the Summer Meeting of theAmerican Society of Health–System Pharmacists; June25, 2007; San Francisco, California; and was presentedorally as an abstract at the Society of General InternalMedicine Annual Meeting; April 10, 2008; Pittsburgh,Pennsylvania.Additional Contributions: The following Partners In-formation Systems personnel were involved in develop-ing this intervention: Barry Blumenfeld, MD, Cheryl VanPutten, PMP, John Poikonen, BA, Eric Godlewski, BA,Linda Moroni, MBA, Michael McNamara, BA, SandraSmith, BA, Marilyn Paterno, MBI, Daniel Fuchs, BS, Ol-iver James, BS, and Greg Rath, BA. The BWH medica-tion reconciliation implementation team included ErinGraydon-Baker, MS, RRT, Christine McCormack, BA,Christina Pelletier, BA, Emily Maher, MD, Ellen Bergeron,RN, MSN, Jennifer Kuzemchak, BA, Michael Cotugno,RPh, and Andrea Giannattasio, BA). The MassachusettsGeneral Hospital medication reconciliation implemen-

Table 4. Differences in Effect of Intervention by Study Site

Outcome

Hospital 1Adjusted andClustered RR

(95% CI)

Hospital 2Adjusted andClustered RR

(95% CI)

PADEs per patient 0.60 (0.38-0.97) 0.87 (0.57-1.32)PADEs per patient due to history

errors0.66 (0.36-1.20) 0.94 (0.59-1.48)

PADEs per patient due toreconciliation errors

0.56 (0.18-1.70) 0.75 (0.36-1.58)

PADEs per patient at admission 0.57 (0.18-1.76) 1.18 (0.58-2.41)PADEs per patient at discharge 0.61 (0.35-1.04) 0.77 (0.49-1.20)

Abbreviations: CI, confidence interval; PADEs, potential adverse drug events;RR, relative risk.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM779

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from

tation team included George Baker, MD, Sally Millar, RN,and Margaret Clapp, BA. Bonnie Greenwood, Pharm D,and Trisha LaPointe, Pharm D, BCPS, assisted in phar-macist data collection; and BWH personnel John Orav,PhD, provided for biostatistical assistance, Elisabeth Bur-dick, MS, assisted in statistical programming, Amy Bloom,MPH, assisted in project management, and Emily Bar-sky, BA, Eric Tomasini, BA, and Emily Dattwyler, BA, pro-vided research assistance. Jaylyn Olivo and Jeanne Zim-merman at BWH provided editorial assistance.

REFERENCES

1. Coleman EA, Smith JD, Raha D, Min SJ. Posthospital medication discrepancies:prevalence and contributing factors. Arch Intern Med. 2005;165(16):1842-1847.

2. Smith JD, Coleman EA, Min SJ. A new tool for identifying discrepancies in post-acute medications for community-dwelling older adults. Am J Geriatr Pharmacother.2004;2(2):141-147.

3. Tam VC, Knowles SR, Cornish PL, Fine N, Marchesano R, Etchells EE. Fre-quency, type and clinical importance of medication history errors at admissionto hospital: a systematic review. CMAJ. 2005;173(5):510-515.

4. Joint Commission on Accreditation of Healthcare Organizations. Using medica-tion reconciliation to prevent errors. Jt Comm J Qual Patient Saf. 2006;32(4):230-232.

5. Cornish PL, Knowles SR, Marchesano R, et al. Unintended medication discrep-ancies at the time of hospital admission. Arch Intern Med. 2005;165(4):424-429.

6. Schnipper JL, Kirwin JL, Cotugno MC, et al. Role of pharmacist counseling inpreventing adverse drug events after hospitalization. Arch Intern Med. 2006;166(5):565-571.

7. Joint Commission on Accreditation of Healthcare Organizations. 2006 Criticalaccess hospital and hospital national patient safety goals. http://www.jointcommission.org/PatientSafety/NationalPatientSafetyGoals/06_npsg_cah.htm. Accessed April 18, 2008.

8. Institute for Healthcare Improvement. Medication Reconciliation Review. http://www.ihi.org/IHI/Topics/PatientSafety/MedicationSystems/Tools/Medication+Reconciliation+Review.htm. Accessed April 18, 2008.

9. Bartick M, Baron D. Medication reconciliation at Cambridge Health Alliance: ex-periences of a 3-campus health system in Massachusetts. Am J Med Qual. 2006;21(5):304-306.

10. Boockvar KS, Carlson LaCorte H, Giambanco V, Fridman B, Siu A. Medicationreconciliation for reducing drug-discrepancy adverse events. Am J GeriatrPharmacother. 2006;4(3):236-243.

11. Gleason KM, Groszek JM, Sullivan C, Rooney D, Barnard C, Noskin GA. Recon-ciliation of discrepancies in medication histories and admission orders of newlyhospitalized patients. Am J Health Syst Pharm. 2004;61(16):1689-1695.

12. Pronovost P, Weast B, Schwarz M, et al. Medication reconciliation: a practicaltool to reduce the risk of medication errors. J Crit Care. 2003;18(4):201-205.

13. Rozich JD, Rezar RK. Medication safety: one organization’s approach to thechallenge. J Clin Outcomes Manag. 2001;8(10):27-34.

14. Varkey P, Cunningham J, O’Meara J, Bonacci R, Desai N, Sheeler R. Multidisci-plinary approach to inpatient medication reconciliation in an academic setting.Am J Health Syst Pharm. 2007;64(8):850-854.

15. Vira T, Colquhoun M, Etchells E. Reconcilable differences: correcting medica-

tion errors at hospital admission and discharge. Qual Saf Health Care. 2006;15(2):122-126.

16. Pippins JR, Gandhi TK, Hamann C, et al. Classifying and predicting errors of in-patient medication reconciliation. J Gen Intern Med. 2008;23(9):1414-1422.

17. Poon EG, Blumenfeld B, Hamann C, et al. Design and implementation of an ap-plication and associated services to support interdisciplinary medication recon-ciliation efforts at an integrated healthcare delivery network. J Am Med InformAssoc. 2006;13(6):581-592.

18. Bates DW, Cullen DJ, Laird N, et al; ADE Prevention Study Group. Incidence ofadverse drug events and potential adverse drug events: implications for prevention.JAMA. 1995;274(1):29-34.

19. Gurwitz JH, Field TS, Radford MJ, et al. The safety of warfarin therapy in the nurs-ing home setting. Am J Med. 2007;120(6):539-544.

20. Wang JK, Herzog NS, Kaushal R, Park C, Mochizuki C, Weingarten SR. Preven-tion of pediatric medication errors by hospital pharmacists and the potential ben-efit of computerized physician order entry. Pediatrics. 2007;119(1):e77-e85.

21. Kopp BJ, Erstad BL, Allen ME, Theodorou AA, Priestley G. Medication errors andadverse drug events in an intensive care unit: direct observation approach fordetection. Crit Care Med. 2006;34(2):415-425.

22. Bates DW. Frequency, consequences and prevention of adverse drug events.J Qual Clin Pract. 1999;19(1):13-17.

23. Ammenwerth E, Schnell-Inderst P, Machan C, Siebert U. The effect of electronicprescribing on medication errors and adverse drug events: a systematic review.J Am Med Inform Assoc. 2008;15(5):585-600.

24. Etchells E. Reconcilable differences: enhancing medication safety at times oftransition. Paper presented at: University of Calgary Forth Annual Quality Im-provement Forum; January 27, 2006; Calgary, Alberta, Canada.

25. Lunceford JK, Davidian M. Stratification and weighting via the propensity scorein estimation of causal treatment effects: a comparative study. Stat Med. 2004;23(19):2937-2960.

26. Lin DY, Wei LJ, Ying Z. Model-checking techniques based on cumulative residuals.Biometrics. 2002;58(1):1-12.

27. Centers for Medicare & Medicaid Services. DRG Relative Weights 2006. http://www.cms.hhs.gov. Accessed January 20, 2009.

28. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use withICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613-619.

29. Nickerson A, MacKinnon NJ, Roberts N, Saulnier L. Drug-therapy problems, in-consistencies and omissions identified during a medication reconciliation andseamless care service. Healthc Q. 2005;8(special issue):65-72.

30. Anderson HJ. Medication reconciliation: what role will IT play? Health Data Manag.2007;15(7):44-48.

31. Bails D, Clayton K, Roy K, Cantor MN. Implementing online medication recon-ciliation at a large academic medical center. Jt Comm J Qual Patient Saf. 2008;34(9):499-508.

32. Groeschen HM. Electronic system improves medication reconciliation rates. AmJ Health Syst Pharm. 2007;64(18):1894.

33. Lawrence D. Bringing the outside in. Hospital and vendor collaborate on devel-opment of a medication reconciliation IT solution. Healthc Inform. 2007;24(10):10-13.

34. Yu F, Leising S, Turner S. Development approach to an enterprise-wide medi-cation reconciliation tool in a free-standing pediatric hospital with commercialbest-of-breed systems. Paper presented at: American Medical Informatics As-sociation Annual Symposium; November 12, 2007; Chicago, IL.

35. The Advisory Board Co. Electronic Medication Reconciliation: Practices for Stream-lining Information Transfer. Washington, DC: Advisory Board Co; 2007.

36. Turchin A, Hamann C, Schnipper JL, et al. Evaluation of an inpatient computer-ized medication reconciliation system. J Am Med Inform Assoc. 2008;15(4):449-452.

(REPRINTED) ARCH INTERN MED/ VOL 169 (NO. 8), APR 27, 2009 WWW.ARCHINTERNMED.COM780

©2009 American Medical Association. All rights reserved. at University of Washington Libraries, on July 21, 2011 www.archinternmed.comDownloaded from