QUALITY AND SAFETY

Electronic prescribing reduces prescribing error in public hospitals

Ramzi Shawahna, Nisar-Ur Rahman, Mahmood Ahmad, Marcel Debray, Marjo Yliperttula and Xavier

Decleves

Aims and objectives. To examine the incidence of prescribing errors in a main public hospital in Pakistan and to assess the

impact of introducing electronic prescribing system on the reduction of their incidence.

Background. Medication errors are persistent in today’s healthcare system. The impact of electronic prescribing on reducing

errors has not been tested in developing world.

Design. Prospective review of medication and discharge medication charts before and after the introduction of an electronic

inpatient record and prescribing system.

Methods. Inpatient records (n = 3300) and 1100 discharge medication sheets were reviewed for prescribing errors before and

after the installation of electronic prescribing system in 11 wards.

Results. Medications (13,328 and 14,064) were prescribed for inpatients, among which 3008 and 1147 prescribing errors were

identified, giving an overall error rate of 22Æ6% and 8Æ2% throughout paper-based and electronic prescribing, respectively.

Medications (2480 and 2790) were prescribed for discharge patients, among which 418 and 123 errors were detected, giving

an overall error rate of 16Æ9% and 4Æ4% during paper-based and electronic prescribing, respectively.

Conclusion. Electronic prescribing has a significant effect on the reduction of prescribing errors.

Relevance to clinical practice. Prescribing errors are commonplace in Pakistan public hospitals. The study evaluated the impact

of introducing electronic inpatient records and electronic prescribing in the reduction of prescribing errors in a public hospital

in Pakistan.

Key words: hospitals, medication errors, nurses, nursing, Pakistan, prescribing errors

Accepted for publication: 18 December 2010

Introduction

Patient safety has become a nationwide priority. Conse-

quently, healthcare delivery has come under tighter scrutiny

during the last decade (Stone et al. 2009). Annually,

medication errors exact an astoundingly high financial and

human toll on society through direct injury to patients and

substantial increase in medical expenditure (Dean Franklin

et al. 2005). The medication process which can be described

in different stages of prescribing, transcribing, dispensing,

administration and monitoring, has proven to be error-prone

(Ghaleb et al. 2010). Clinical decision and prescription

Authors: Ramzi Shawahna, MPhil, PhD Student, Faculty of

Pharmacy and Alternative Medicine, The Islamia University of

Bahawalpur, Bahawalpur, Pakistan, Faculte de Pharmacie,

Universite Paris Descartes, Paris, France and Division of

Biopharmaceutics and Pharmacokinetics, Faculty of Pharmacy,

University of Helsinki, Helsinki, Finland; Nisar-Ur Rahman, PhD,

Associate Professor and Chairman, Department of Pharmacy, The

Islamia University of Bahawalpur; Mahmood Ahmad, PhD, Professor

and Dean Faculty of Pharmacy and Alternative Medicine, The

Islamia University of Bahawalpur, Bahawalpur, Pakistan; Marcel

Debray, Professor of Biostatistics, Universite Paris-Descartes, Faculte

des Sciences Pharmaceutiques et Biologiques, Departement Sante

Publique et Biostatistique, Paris, France; Marjo Yliperttula, PhD,

Professor, Division of Biopharmaceutics and Pharmacokinetics,

Faculty of Pharmacy, University of Helsinki, Helsinki, Finland;

Xavier Decleves, PhD, Associate Professor, Faculte de Pharmacie,

Universite Paris Descartes, Paris, France

Correspondence: Ramzi Shawahna, PhD Student, Faculty of

Pharmacy and Alternative Medicine, The Islamia University of

Bahawalpur, Bahawalpur-63100, Pakistan. Telephone: +92 62

9255243.

E-mail: ramzi_shawahna@hotmail.com

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3233

doi: 10.1111/j.1365-2702.2011.03714.x

writing are highly challenging, even in sophisticated health-

care facilities where medications are handled by trained and

credentialed healthcare professionals (Friedman et al. 2007).

In the UK’s National Health Service, more than 2Æ5 million

prescriptions are written every day, it has been estimated that

errors affect 11% of those prescriptions with a cost of £400

million per year (Fitzgerald 2009). Prescribing errors are

common cause of iatrogenic injury. The causes of prescribing

errors include slips of attention, failure in applying adequate

prescribing rules, insufficient knowledge of pathology, phys-

iology and clinical drug therapy. Furthermore, proper

prescribing requires adequate consideration of patient char-

acteristics, dose calculation, medication nomenclature and

dosage formulation (Dean et al. 2002a, Lesar 2002).

Universally, handwritten prescribing on a non-standardised

and cumbrous paper-based medication and discharge medi-

cation charts predominates in hospital settings. In the UK and

Australia there have been calls to standardising medication

charts, since the content and layout of these charts contrib-

uted to prescribing errors (Coombes et al. 2008). Similarly,

interventions like moving to electronic prescribing, clinical

decision support systems and clinical pharmacist interven-

tions were also recommended endeavours to effectively

reducing prescribing errors both in inpatient and outpatient

settings (Pollock et al. 2007, Stone et al. 2009). Electronic

prescribing systems as represented by the computerised

physician order entry (CPOE) are known to reduce prescrib-

ing errors. In the USA, only 5% of medical facilities use some

form of the CPOE (Stone et al. 2009). Recently, the Office of

National Coordinator for Health Information Technology

was created to ensure the establishment of electronic forms of

medical records by 2014 in all healthcare organisations in the

USA (DesRoches et al. 2010, Stone et al. 2009). Such

initiatives are highly welcomed elsewhere in the world.

The issue of medication errors has been extensively

explored in the developed countries; conversely, scarce data

have been reported in the developing world. In Pakistan, a

country with a complex healthcare infrastructure, little is

known about the nature of prescribing errors in public

hospitals. Recently, an orphan study reported prescribing

errors in a psychiatry ward in a public hospital (Shawahna &

Rahman 2008). We conducted the present study to investi-

gate and compare the nature and incidence of prescribing

errors in a main public hospital in Pakistan before and after

the introduction of electronic inpatient record and prescrib-

ing system. Such comparison should allow elucidating the

impact of the move to electronic inpatient record and

electronic prescribing on the incidence of prescribing errors

in public hospitals in Pakistan.

Methods

Setting

The study was conducted in a 1280-bed teaching hospital

situated in Lahore (Pakistan). The main public hospital

provided secondary and tertiary care services; however, the

tertiary services were limited to cardiac care. The hospital

operated typical Pakistani multi-ward administrative infra-

structure. Each ward was operated by a team of physicians

led by a head professor. The team included associate

professor(s), assistant professor(s), senior registrar(s) and

intern(s). The hospital admitted a patient mix of mainly

middle class and government employees’; patients were

admitted to the different wards according to their symptoms

and indications. The hospital employed 10 pharmacists who

provided classic pharmacy services including procurement of

medicines, storage and record keeping. None of those

pharmacists was involved in inpatient record auditing,

medication chart or discharge medication reviewing. In

routine practice, ward doctors handwrote medications onto

a pre-printed paper-based inpatient record. The inpatient

record contained the following information: (1) basic patient

information (name, date of birth, age, gender and address),

(2) history, (3) diagnosis, (4) treatment plan, (5) medication

chart and (6) assessment. Medications prescribed onto the

medication chart were then transcribed by the nursing staff

onto separate paper-sheets. These sheets were then used to

procure medications from the central pharmacy store. Nurs-

ing staff used the medication chart to determine drugs and

doses due and noted their administration. At the time of

patient discharge, medications to be continued were then

transcribed onto a paper-based pre-printed discharge medi-

cations sheet. Consequently, medications were then tran-

scribed onto separate sheets where they could be brought

from the central pharmacy or could be purchased from

private pharmacies.

Paper-based medication charts

We have decided to review prospectively 150 inpatient

records from each ward (n = 1650) for prescribing errors.

An inpatient case in this study refers to one particular

hospital admission. The study was conducted in the following

wards: (1) emergency, (2) ear, nose and throat (ENT), (3)

gynaecology, (4) medical 1, (5) medical 2, (6) medical 3, (7)

paediatrics, (8) pulmonary, (9) cardiology, (10) coronary care

and (11) general cardiac surgery. Inpatient records were

sampled from each ward randomly.

R Shawahna et al.

3234 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

Paper-based discharge medications sheets

Consequently, 50 discharge medications sheets were reviewed

prospectively from each ward (n = 550) for prescribing and

transcribing errors. Similarly, sheets were sampled randomly.

Prescribing errors

A definition of prescribing errors and scenarios representing

prescribing error situations in Pakistan hospitals were

described in a previous study (Shawahna & Rahman 2009).

Briefly, a practitioner-led definition of prescribing error and

scenarios representing error situations were developed with

the participation of a panel of expert judges composed of 21

physicians, 19 pharmacists, five nurses, three pharmacolo-

gists and two risk managers (the demographic details of the

panel are given elsewhere (Shawahna & Rahman 2009).

Similar definitions have been used in previous studies in UK

(Dean et al. 2000, Ghaleb et al. 2005). The categories of

prescribing errors included in this study are listed in Table 1.

Data collection

A five-day pilot study was conducted in a representative ward

to assess the practicality and validate the collection protocol.

During data collection, prescribers were blinded to maintain

usual prescribing practice. However, only the head of the

ward and the hospital administration were aware of the

collection protocol.

In each ward, entire inpatient records of the recruited

inpatients were prospectively scanned twice a day (morning

and evening), throughout their stay, using a digital camera

scanner (ORITE Technology Co., Ltd, Taipei, Taiwan) by two

researcher pharmacists. Scans were then transferred to a

computer where they were enlarged and viewed. The discharge

medications sheets of 50 patients from each ward were scanned

similarly.

The ethics of this study were approved by the board of

advance studies and research of the Islamia University of

Bahawalpur. A risk management panel composed of two

physicians and two pharmacists reviewed the scans for

prescribing errors. Each reviewer viewed the scans indepen-

dently and gave a score of 0 for ‘no error situation’ or 1 in

case of error. Disputed ratings were arbitrated by a risk

manager with pharmaceutical background. Decision on error

was based on official monographs, product specifications

given by manufacturers and decisions rendered by the panel

of expert judges. Each medication prescribed was reviewed

against all error categories. Thus, one medication prescribed

could be associated with more than one error. All medica-

tions prescribed during the entire study, including regular and

once only medications, were considered in the analysis. This

included medications prescribed on day 0 (the day of

inpatient admission). In the consecutive days, only newly

prescribed drugs were included in the analysis. Thus, each

medication prescribed was counted only once. Error analysis

of medications prescribed was carried out in the same day

after each scanning, the time lag between scanning and

analysis was between one and three hours. Errors identified in

the morning scanning were communicated immediately to the

ward head or senior during the duty hours and error

reporting was initiated. Errors identified after the evening

scanning were discussed in the next day either with the head

of the ward or senior before the morning dose; otherwise, a

serious error was reported immediately and corrective action

was pursued.

Severity of error

Severity of error was classified according to a classification by

the National Coordinating Council for Medication Error

Reporting and Prevention as modified in a previous study

(van den Bemt et al. 2002). Similarly, reviewers classified

errors into six categories: A1, A2, B, C, D and E. Disputed

ratings were arbitrated by the risk manager. Seniority of

prescribers was defined by experience in years as either junior

(< 4) or senior (‡4).

Seminar and newsletters

Doctors, nurses and pharmacists were invited to series of

seminars on medication errors (including prescribing errors).

Attendants were educated on error scenarios and their

consequences. Throughout the study, newsletters concerning

such errors and safe prescribing practice circulated the

different wards of the hospital. Newsletters contained error

examples and recommendations for safe prescribing.

Electronic medication charts

Inpatient record was coded on spreadsheet software package

(EXCEL, Microsoft). The following sections of the previous

paper-based record (1) patient basic information, (2) history,

(3) diagnosis, (4) treatment plan and (5) assessment, were

retained. The section of medication chart was modified to

contain a space for indispensable information to be consulted

at the time of medication prescription. Information concern-

ing medication history, allergy, renal and hepatic function

was included in the medication chart section. Crossable

boxes indicating instructions on dose, frequency and route of

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3235

Table 1 Prescribing errors identified during the two phases of the study

Error

category Example

Paper-based profile Electronic profile

p

Number of

errors

(percentage

of all errors)

Total

(percentage)

Number of

errors

(percentage

of all errors)

Total

(percentage)

(1) Dosing error A dosing error was

considered when

one or a combina

tion of these

scenarios occurred:

1179 (39Æ2%) 583 (50Æ8%) < 0Æ001

(a) The dose or

frequency is

sub-therapeutic

or toxic

(especially

when the drug

has a narrow

therapeutic

range).

A patient was

prescribed half

tablet of

Theophylline

50 mg q12h.

322 (10Æ7%) 150 (13Æ1%)

(b) The dose/regi-

men was not

altered after

the steady-state

serum

concentration

was believed to

be achieved

when such an

objective was

stipulated

while

prescribing the

drug or regimen

A patient was

prescribed with

Amoxicillin 1 g

injection q12h

followed by

500 mg tablets

q12h

419 (13Æ9%) 162 (14Æ1%)

(c) The dose/

regimen was

inappropriate

for the patient’s

renal function

according to

the Pakistan

National

Formulary,

summary of

product

characteristics,

or reference

sources

A doctor prescribed

Ketorolac 30 mg

q6h for a patient on

dialysis

342 (11Æ4%) 201 (17Æ5%)

(d) The dose written

in ‘milligrams’

when

‘micrograms’

were intended

A doctor prescribed

Digoxin in 250 mg

instead of 250 lg

96 (3Æ2%) 70 (6Æ1%)

R Shawahna et al.

3236 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

Table 1 (Continued)

Error

category Example

Paper-based profile Electronic profile

p

Number of

errors

(percentage

of all errors)

Total

(percentage)

Number of

errors

(percentage

of all errors)

Total

(percentage)

(2) The maximum

dose was not

specified when

the medication

was prescribed

as ‘S.O.S’

(Latin ‘si opus

sit’, meaning:

when needed)

A patient was

prescribed Injection

Voltaren 75 mg

(Dicolfenac

Sodium) as S.O.S

without indicating

the maximum dose

274 (9Æ1%) 11 (1%) < 0Æ001

(3) The name of the

medication was

misspelled,

leading to

confusion

(writing

illegibly was

not considered

a prescribing

error, even when

hardly

readable)

A doctor misspelled

tablet Zantac

(ranitidine) with Z

looked like X

without indicating

the dose, the order

could be confused

with Xanax

(Alprazolam)

281 (9Æ3%) 14 (1Æ2%) < 0Æ001

(4) Ambiguous

medication

order

The medication

order was consid-

ered ambiguous

when one or a

combination of

these scenarios

occurred:

667 (22Æ2%) 231 (20Æ1%) < 0Æ001

(a) The order was

completely

unclear or leads

to confusion

A patient was pre-

scribed with Syrup

Lomogel, a name

which never ex-

isted, when asked

he said it was

Somogel (Ligno-

caine)

155 (5Æ2%) 21 (1Æ8%)

(b) The order was

not rewritten in

full when a

change was

made, thus fail-

ing to provide a

requisite neat

and clean order

A doctor crossed a

medication

prescribed then

ticked it with � as

still valid

75 (2Æ5%) 14 (1Æ2%)

(c) The route of

drug adminis-

tration was not

stated when

more than one

route was

applicable

A doctor recom-

mended nebulisa-

tion with

Salbutamol, nei-

ther the duration

nor the frequency

were specified

152 (5Æ1%) 82 (7Æ1%)

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3237

Table 1 (Continued)

Error

category Example

Paper-based profile Electronic profile

p

Number of

errors

(percentage

of all errors)

Total

(percentage)

Number of

errors

(percentage

of all errors)

Total

(percentage)

(d) The medication

was prescribed

using a non-

standard abbre-

viation and/or

nomenclature

A doctor prescribed

‘Diclo’ 75 mg a

short for Dicloran

75 mg (Diclofenac

Sodium) ‘Diclo’

could be misunder-

stood with Diclo-P

(Diclofenac

Potassium)

285 (9Æ5%) 114 (9Æ9%)

(5) Dosage form

error

A dosage form was

considered when

one or a

combination of

these scenarios

occurred:

205 (6Æ8%) 92 (8%) <0Æ01

(a) The medication

was prescribed

in a dosage

form that was

not available

commercially

A doctor prescribed

capsule Risek

40 mg (Omepra-

zole), Risek is

available in two

dosage forms, one

is capsule 20 mg

and the other is

infusion 40 mg

93 (3Æ1%) 20 (1Æ7%)

(b) The medication

was prescribed

in a dosage

form that

couldn’t be

administered

to the patient

in his/her

clinical

situation

An unconscious

patient was

prescribed with

tablet Lasix 40 mg

(Furosemide)

112 (3Æ7%) 72 (6Æ3%)

(6) A medication

wasn’t

prescribed

when the

patient’s clinial

condition

required a

medication

A patient had severe

cough attacks with

no cough remedy

prescribed

57 (1Æ9%) 62 (5Æ4%) ns

(7) The medication

prescribed was

contraindicated

for the clinical

condition of the

patient

A doctor prescribed

tablet Tegral

200 mg (Carba-

mazepine) for

anaemic patient

141 (4Æ7%) 73 (6Æ4%) <0Æ05

R Shawahna et al.

3238 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

administration were added. Similar changes were made to the

discharge medications sheet. A comprehensive standard

medication information database was constructed using

standard sources (Anderson et al. 2002, Joint Formulary

Committee 2007). The database provided proprietary and

non-proprietary (brand and generic) names, dosage forms,

available strengths, adverse drug reactions, contraindications,

drug–drug, drug–food and drug–disease interactions for all

medications on the hospital’s formulary and commonly

prescribed medications. The electronic inpatient record was

interfaced to the database. The system did not provide

interactive alerts (pop-ups) neither decision support such as

drug interaction or allergy checks. All relevant information

was listed and physicians needed to consult them at the time

of prescribing.

The new electronic inpatient record was sent to a collective

of 20 doctors, 20 nurses and 10 pharmacists to elicit their

opinions and suggestions on the design. Suggestions rendered

by the collective were considered and the electronic inpatient

record was modified accordingly, later, the final design was

sent for approval to a total of 200 doctors, 200 nurses and

100 pharmacists, employed by the hospital or similar settings

in Lahore. After approval, doctors were given a one-day

training workshop on the new electronic inpatient record and

discharge medications sheet. Doctors were provided hand-

held computers and were asked to prescribe medications onto

the medication chart in the electronic inpatient records and

similarly for discharge medications (onto the electronic

discharge medications sheets).

Consequently, identical sample size (150 inpatient cases

and 50 discharge medications sheets) from the same wards

were recruited to investigate prescribing errors. Electronic

inpatient records were printed and reviewed for prescribing

errors by the same panel. Prints were treated like the scans in

the paper-based inpatient records analysis. The time lag

between scanning and analysis was shorter than that of scans

0Æ5–1Æ5 hours. Similarly, errors identified were communicated

to the head of the ward or senior and corrective action was

pursued.

Analysis

The null hypothesis to be tested was that the electronic

inpatient record and discharge medications sheet interfaced

to the medication information database would not have any

effect on the incidence of prescribing errors. The numerator

of this study was the number of errors identified and the

denominator was the number of medications prescribed.

Error rates in different wards were compared after normal-

ising the absolute error number to the number of medications

prescribed for each inpatient in the each ward.

Prescribing errors data were treated with Graphpad Prism

4Æ0 (GraphPad Software Inc., San Diego, CA, USA). Statis-

tical significance (p < 0Æ05) was tested with Kruskal–Wallis

test (95% CI) for non-normally distributed data. Paper-based

and electronic prescribing data were compared with Mann–

Whitney U-test. Nominal data were compared using the v2 or

Fisher’s exact test and odds ratios (OR) with (95% CIs).

Inter-rater agreement was determined by Fleiss’ generalised

kappa, using a free Excel spreadsheet-based program, the

program is available at http://www.ccit.bcm.tmc.edu/jking/

homepage/genkappa.doc.

Table 1 (Continued)

Error

category Example

Paper-based profile Electronic profile

p

Number of

errors

(percentage

of all errors)

Total

(percentage)

Number of

errors

(percentage

of all errors)

Total

(percentage)

(8) The medication

prescribed

interacted with

another

concomitant

medication

A doctor prescribed

Voltaren 75 mg

(Dicofenac Sodium)

for a patient on

Warfarin

99 (3Æ3%) 43 (3Æ7%) <0Æ05

(9) The medication

prescribed was

not clinically

indicated for

the patient

A doctor prescribed

tablet Convul

200 mg (Carba-

mazepine) for a

patient with no

indication

105 (3Æ5%) 38 (3Æ3%) <0Æ01

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3239

Results

Prescribing errors

During the paper-based prescribing phase, there were 13,328

medications prescribed with a median of 5 and a range of

(2–12). Similarly, there were 14,064 medication prescribed

onto the electronic inpatient records reviewed with a median

of 5 and a range of (2–11). Table 1 shows prescribing errors

identified during the two phases of the study. Table 2 shows

error rates calculated for each ward during the paper-based

and electronic prescribing. Throughout the paper-based

prescribing phase, 3008 errors were identified giving an

overall error rate of 22Æ6%, while error rates varied between

different wards ranging from 40Æ8% to 8%. During electronic

prescribing, there were 1147 identified prescribing errors,

giving an overall error rate of 8Æ2%, similarly, error rates

varied between different wards ranging from 19Æ4–5Æ1%. The

difference in both rates (paper-based and electronic) was

statistically significant (p < 0Æ01). The inter-rater reliability

among the risk management panel was good (j = 64, CI:

60Æ2–67Æ8). Throughout the study, there were 380 (2Æ9%)

omissions of prescriber’s signature these were not considered

as prescribing errors.

Dosing errors dominated prescribing errors types in both

phases of the study. The incidence of dosing errors was

witnessed in all wards at variable rates. The difference in

rates was statistically significant in emergency, paediatrics,

cardiology and general cardiac surgery wards (p < 0Æ05).

Similarly, ambiguous medication orders occurred also at

variable rates in the different wards investigated. For exam-

ple, 9Æ8% of medications ordered in the pulmonary ward

were ambiguous. However, the ambiguity of orders was

reduced in electronic prescribing phase (p < 0Æ01). During

the paper-based prescribing, 9Æ3% of medications ordered

were misspelled, however, the occurrence of misspelling

errors was reduced to 1Æ2% after the installation of electronic

prescribing (p < 0Æ01). Similarly, 9Æ3% of medications were

prescribed as S.O.S (si opus sit) or (when needed) without

indicating the maximal daily dose, such errors were reduced

to 1% in the electronic prescribing phase (p < 0Æ01). The

rest of error types were also reduced, while the reduction was

barely significant.

Table 3 shows the severity of errors identified during the

both phases. During paper-based prescribing, 46% of errors

were in the act of writing; while the rest 54% were errors in

the clinical decision. Electronic prescribing significantly

reduced the error rate to 12Æ2% errors in the act of writing

and 87Æ8% errors in the clinical decision (p < 0Æ01). During

the paper-based prescribing 62Æ9% of errors were rated

minor errors (i.e. without clinical consequences: A1 through

B) against 39Æ7% in the electronic prescribing phase. A type

A1 prescribing error had a likelihood of eight times to occur

in the paper-based context than in the electronic one, while a

type E error had a 0Æ4 times more likely to occur in the paper-

based context.

Table 4 shows the pharmacological classification of med-

ications prescribed throughout the study. During the paper-

based prescribing phase, out of the 13,328 medications

prescribed, 3050 (28Æ88%) were antibiotics. Antibiotic class

had the highest error percentage (39% of total errors).

Likewise, in the electronic prescribing, antibiotics were the

commonest prescribed class.

Discharge medication sheet

During paper-based prescribing, 2480 medications were

prescribed with a median of 4 and a range of (2–9), while

2790 medications were prescribed during the electronic

prescribing phase with a median of 5 and a range of 2–11.

A total of 418 errors were identified in the paper-based

prescribing giving an overall error rate of 16Æ9%, the number

of errors was reduced to 123 errors giving an overall error

rate of 4Æ4% during electronic prescribing. Table 5 shows

prescribing errors detected in discharge medications sheets

during the two phases of the study. Omission errors prevailed

during the paper-based prescribing phase, followed by

misspelling errors. However, the omission errors remained

the highest in the electronic prescribing followed by dosing

errors. Omission, misspelling and ambiguous order errors

were significantly reduced in electronic prescribing phase

(p < 0Æ05).

Table 2 Error rates calculated during the paper-based and electronic

prescribing phases

Ward

Error rate (%) (95% CI)

pPaper-based Electronic

Medical

emergency

40Æ8 (30Æ2–49) 19Æ4 (10Æ8–23Æ9) <0Æ001

Paediatrics 34 (27Æ3–38Æ4) 6 (4–11) <0Æ001

Pulmonary 31Æ3 (24–36Æ2) 8Æ9 (6Æ2–12Æ4) <0Æ001

ENT 27 (23Æ7–33Æ4) 7Æ8 (6Æ2–17) <0Æ01

Gynaecology 26Æ6 (22–32Æ1) 7 (4–11Æ2) <0Æ01

Medical 1 19Æ3 (11–29Æ3) 8Æ3 (7Æ2–8Æ9) <0Æ05

Medical 2 19Æ6 (13Æ2–24Æ5) 6Æ3 (5Æ2–7Æ1) <0Æ05

Medical 3 25 (17–29Æ7) 6 (2–8) <0Æ01

Cardiology 19Æ8 (13Æ9–27Æ6) 6 (5Æ5–10Æ6) <0Æ05

General cardiac

surgery

9Æ6 (5Æ2–13Æ4) 5Æ3 (4Æ2–9Æ3) ns

Coronary care 8 (4–12) 5Æ1 (3Æ8–10) ns

R Shawahna et al.

3240 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

Table

3Sev

erit

yra

ting

of

pre

scri

bin

ger

rors

iden

tified

thro

ugh

out

the

study

Sev

erit

y

rati

ng

Des

crip

tion

Exam

ple

from

this

study

Paper

-base

dpro

file

Ele

ctro

nic

pro

file

pO

dds

rati

o

(95%

CI)

Num

ber

of

erro

rs

(n)

Per

centa

ge

(%)

Num

ber

of

erro

rs

(n)

Per

centa

ge

(%)

Pro

ble

mord

ers:

Err

ors

inth

eact

of

wri

ting

A1

Apre

scri

bin

ger

ror

has

bee

nm

ade,

but

the

erro

ris

so

min

or

that

the

med

icati

on

ord

er

cannot

be

mis

under

stood

Adoct

or

pre

-

scri

bed

capsu

le

Ris

ek40

mg

(Om

epra

zole

),

Ris

ekis

avail

able

ntw

odosa

ge

form

s,one

is

capsu

le20

mg

nd

the

oth

eris

infu

sion

40

mg

650

21Æ6

80

6Æ9

7<

0Æ0

19Æ0

(7–11Æ3

)

A2

Apre

scri

bin

grr

or

has

bee

nm

ade,

but

the

nurs

eca

nnot

adm

inis

ter

the

med

icati

on

wit

hout

havin

gto

gath

er

addit

ional

info

rmat

ion

An

unco

nsc

ious

pati

ent

was

pre

scri

bed

wit

h

table

tL

asi

x40

mg

(Furo

sem

ide)

733

24Æ4

60

5Æ2

3<

0Æ0

113Æ6

(10Æ4

–17Æ7

)

Err

ors

inth

e

clin

ical

dec

isio

n

BA

pre

scri

bin

ger

ror

has

bee

nm

ade

but

adm

inis

trat

ion

to

the

pati

ent

wil

l

have

no

clin

ical

conse

quen

ces

Adoct

or

pre

scri

bed

‘Dic

lo’

75

mg

a

short

for

Dic

lora

n

75

mg

(Dic

lofe

nac

So

diu

m)

‘Dic

lo’

could

be

mis

un

der

stood

wit

h

Dic

lo-P

(Dic

lofe

nac

Pota

ssiu

m)

510

17Æ0

315

27Æ4

6<

0Æ0

11Æ7

(1Æ5

–2)

CA

pre

scri

bin

ger

ror

has

bee

nm

ade,

that

could

pote

nti

all

y

resu

ltin

the

nee

d

for

an

incr

ease

d

freq

uen

cyof

pati

ent

monit

ori

ng

Apati

ent

was

pre

scri

bed

half

table

tof

Theo

phyllin

e

50m

gq12h

470

15Æ6

307

26Æ7

7<

0Æ0

11Æ6

(1Æ4

–1Æ9

)

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3241

Error tendency

In both paper-based and electronic phases of the study, junior

doctors tended to make more errors than their seniors. A junior

doctor had likelihood to make a prescribing error 58Æ7% more

than a senior doctor (OR 1Æ58, 95% CI: 1Æ46–1Æ71) in the

paper-based phase, however, the likelihood decreased to

46Æ3% (OR 1Æ46, 95% CI: 1Æ4–1Æ51) in the electronic phase.

Discussion

Our present study reports the incidence of prescribing errors

in a busy hospital in Pakistan comparing error rates of paper-

based and electronic prescribing. Prescribing errors resulting

in patient harm (iatrogenic injury) are common in today’s

healthcare system. The dilemma is multifaceted in nature.

Prescribing errors in the US and UK’s healthcare systems have

been extensively studied with fewer reports described the

situation in European and Australian context (Dean FranklinTable

3(C

onti

nued

)

Seve

rity

rati

ng

Des

crip

tion

Exam

ple

from

this

study

Paper

-base

dpro

file

Ele

ctro

nic

pro

file

p

Odds

rati

o

(95%

CI)

Num

ber

of

erro

rs

(n)

Per

centa

ge

(%)

Num

ber

of

erro

rs

(n)

Per

centa

ge

(%)

DA

pre

scri

bin

g

erro

rhas

bee

n

made,

that

could

pote

nti

all

y

resu

ltin

dam

age

toth

epati

ent

Adoct

or

pre

scri

bed

table

tT

egra

l

200

mg

(Car

bam

aze

pin

e)

for

anae

mic

pati

ent

415

13Æ8

215

18Æ7

4<

0Æ0

12Æ1

(1Æ8

–2Æ4

)

EA

pre

scri

bin

g

erro

rhas

bee

n

made,

that

could

pote

nti

all

y

resu

ltin

the

dea

thof

the

pati

ent

Adoct

or

pre

scri

bed

Dig

oxin

in

250

mg

inst

ead

of

250

lg

230

7Æ6

170

14Æ8

2<

0Æ0

51Æ4

(1Æ2

–1Æ8

)

Table 4 Pharmacological classification of medications prescribed

during the study

Pharmacological class

Absolute number

Paper-based Electronic

Antibiotics 3050 3500

Analgesics 2800 3100

Diuretics 2500 1540

Sedatives-hypnotics 1300 1214

Antiasthmatics 700 862

Antihypertensives 624 953

Beta-blockers 490 450

Hormones and steroids 430 250

Histamine antagonists 308 732

Antiepileptics 241 412

Anticoagulants 230 370

Antipsychotics 200 180

Nitrates 150 210

Others 305 291

Table 5 Prescribing errors detected in the discharge medications

sheets during the study

Paper-based Electronic

p

Absolute

number

of errors

Error

rate

(95% CI)

Absolute

number

of errors

Error

rate

(95% CI)

Omission 152 6Æ1 (5–10Æ2) 50 1Æ8 (0Æ9–2Æ6) <0Æ01

Misspelling 103 4Æ2 (3Æ6–8) 11 0Æ4 (0Æ2–1Æ1) <0Æ01

Ambiguous

order

54 2Æ2 (1Æ2–3Æ9) 12 0Æ4 (0Æ1–0Æ9) <0Æ05

Dosing error 71 2Æ9 (2Æ1–5Æ3) 40 1Æ4 (0Æ8–2Æ6) ns

Dosage form 38 1Æ5 (0Æ7–2Æ9) 10 0Æ4 (0Æ1–0Æ8) ns

R Shawahna et al.

3242 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

et al. 2005, Lewis et al. 2009). The structure of healthcare

system in Pakistan is different from those in the US and

Europe. Therefore, it was necessary to investigate the nature

of prescribing errors in such environments. Error reporting

system is severely underdeveloped in Pakistan public hospi-

tals and errors often go unnoticed.

Our study is the first extensive report from Pakistan

hospitals. The hospital chosen for this study operated typical

Pakistani administrative and operative system. The multi-ward

specialties and cases included in this study make it possible to

generalise the results to similar institutions in Pakistan. The

overall prescribing error rate reported in our study (Table 2)

before intervention (during paper-based prescribing) (22Æ6%)

was higher than rates reported in other studies in UK and the

USA. The error rate ranged between (0Æ4–15Æ4%) in the US and

(7Æ4–18Æ7%) in the UK (Dean Franklin et al. 2005, Lewis et al.

2009). Recently, a study conducted in UK showed an error rate

of 13Æ2% in paediatric settings (Ghaleb et al. 2010). Our results

could be essentially different since error rate depends on

definition, methodology, setting investigated and depth of

investigation, furthermore, it has been proven observer depen-

dent (Dean Franklin et al. 2005, Lewis et al. 2009). Prescribing

errors identified in our study concerned 66% of the inpatients

during the paper-based prescribing phase against 42% during

electronic prescribing. Such result was consistent with studies

conducted abroad. Studies showed that prescribing errors

occurred in 40Æ5–100% of inpatient cases studied, likewise,

consistent results were observed across the wards included in

those studies (Kaushal et al. 2001). The highest error rate

(40Æ8%) occurred in the Medical emergency ward (Table 2).

Potts et al. (2004) reported a comparative rate (39Æ1%) in a

study conducted at a paediatric intensive care unit (ICU). In our

study setting, the dynamicity of the emergency ward and

diversity of cases admitted could also further complicate the

situation (Dean et al. 2002a). The variability in error rate

across different wards could be attributed to differential

experience, since the demographics of prescribers and their

background were largely different across wards. Consistent

with our findings, former research revealed that the majority of

prescribing errors are related to dose (Table 1). In a study

conducted by Dean et al., 54% of the identified errors were

related to dose (Dean et al. 2002b). Clarity of medication

orders concerned 5% of the errors identified in this study.

Failing in communicating clear orders often led to dangerous

consequences, such as depriving the patient from the benefits of

the right medication and subjecting the patient to the dangers

of unnecessary medication. In a communication, the American

Hospital Association and collaborating associations strongly

discouraged the use of non-standard nomenclature, abbrevia-

tion and symbols. Such usage frequently results in the misin-

terpretation of the intent of the order (Pollock et al. 2007).

Prescribing errors dominated in the most prescribed medica-

tion class (antibiotics) (Table 4). Our findings were consistent

with previous research (Thomsen et al. 2007, Lewis et al.

2009). Electronic prescribing significantly reduced prescribing

errors in different severity classes. However, the impact of

electronic prescribing was more pronounced on minor errors

(i.e. errors without clinical consequences A1 through B)

(Table 3). In our study, the impact of electronic prescribing

brought a significant reduction in the occurrence of prescribing

errors in all wards other than the general cardiac surgery ward

and the coronary care unit (Table 2). Similar significance

reduction was observed with all types of prescribing errors

identified other than ‘not prescribing a medication when

clinically needed’ (Table 1). Previous reports recommended

the use of computerised medication order systems to reduce

prescribing errors both, occurrence and severity (Bizovi et al.

2002, Potts et al. 2004, Colpaert et al. 2006). Our study was

conducted in two-phase design (paper-based and electronic) in

the same wards by the same researchers, recruiting the same

number of patients. Such design is supposed to ensure a

consistent approach. Furthermore, the same risk management

panel rated the prescribing errors. Moreover, the panel

consisted of internal and external reviewers. Such consistent

approach strengthen the rejection of the hull hypothesis (that

there will be no effect of computerised prescribing on the

number of prescribing errors), however, the reduction of

prescribing errors was due to the computerisation of the

medication and discharge medication charts, in addition to

educating prescribers on medication errors.

The tendency of junior doctors to make more errors than

senior prescribers was evident in our study. Although the

literature is inconclusive, in a study conducted by Caruba et al.,

(2010) prescribing errors occurred at higher rates on the first

day of inpatient hospital stay, the study connected the higher

error rates with the fact that on admittance, patients were

received by junior doctors who took their medication history

(Caruba et al. 2010). Coombes et al. (2008) described an

existing culture where prescription writing is seen as a low-risk

chore undertaken by junior doctors where orders were often

incomplete, ambiguous or illegible. Consistently, junior doc-

tors performed more clerical duties than their seniors. Previous

research showed that prescribing errors could be reduced by

reporting and educating prescribers on the issue of prescribing

errors. In a study conducted by Shawet al. (2003), academic

detailing was shown to significantly affect the error rates

identified in the settings investigated (41% vs. 24%). Pharma-

cist interventions and interdisciplinary team medication rec-

onciliation could result in minimising the occurrence of

prescribing and transcription errors (Barber et al. 1997). In a

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3243

previous study, van den Bemt et al. (2002) investigated the

cost–benefit and concluded that pharmacy staff interventions

resulted in higher benefits than the costs related to time

investment. Similarly, modification of medication charts could

bring beneficial results in this domain as show in the study of

Coombes et al. (2008).

Limitations

Our study has several limitations. First, the tertiary cardiac

care services had a separate administration; however, in this

analysis we treated the data as they came from one hospital.

A multicenter study would have permitted to draw a more

solid conclusion. Second, albeit the rating by a risk-manage-

ment panel, the possibility of leaving some prescribing errors

undetected cannot be excluded, therefore, this study could be

an underestimation of prescribing errors in the study setting.

Third, throughout the study, data collectors had the oppor-

tunity to observe and interview prescribers, nursing staff and

patients. Conversely, neither error context nor situations (like

workload) leading to error were formally investigated.

Fourth, our study design had pre and postphases. Such

design has less control on bias as compared with a rando-

mised controlled trial design. Finally, the electronic inpatient

record along with the interfaced database did not provide

‘flags’ if the dose was inappropriate or a possible interaction.

A system with flags and warnings is expected to have a better

impact on the reduction of such errors.

Conclusion

Our results indicate that prescribing errors are highly

prevalent in public hospitals in Pakistan in the existing

environment. Such high prevalence serves as a call to action

to minimise these errors. We have shown that incidence of

prescribing errors can be reduced by information technology

and educating prescribers and nursing staff on medication

errors and safe prescribing. However, this fact has been

shown in previous studies conducted elsewhere. We believe

that improving the electronic system with flags and warnings

could further reduce prescribing errors. Furthermore, previ-

ous studies showed that ward-based pharmacist interventions

and clinical recommendations resulted in reduction of similar

errors (Barber et al. 1997). In the current settings, dispensary-

based pharmacists physically do not see the patient; neither

have full access to patient data. However, once an erroneous

order reaches the pharmacy, it is almost sure that the error

will reach the patient if a checking point was not installed.

Relevance to clinical practice

Prescribing errors are commonplace in Pakistan public

hospitals. The present study tested the impact of moving to

electronic patient records and prescribing helps in reducing

the incidence of such errors. In 2004, Pakistan universities

started a brand new clinically oriented PharmD program.

New graduates are expected to initiate clinical pharmacy

services in hospitals. There is a clear need to investigate the

role of the ward pharmacist in minimising such errors in

Pakistan hospitals.

Acknowledgements

The authors would like to thank the hospital superintendent

and professors of the wards involved in this study for

granting permission and cooperative behaviour.

Contributions

Study design: RS, NR; data collection and analysis: RS, NR,

MD, MA, MY, XD and manuscript preparation: RS, NR,

MD, XD.

References

Anderson P, Knoben J & Troutman W

(2002) Handbook of Clinical Drug

Data, 10th edn. McGraw-Hill Compa-

nies, Inc., Columbus, OH.

Barber ND, Batty R & Ridout DA (1997)

Predicting the rate of physician-

accepted interventions by hospital

pharmacists in the United Kingdom.

American Journal of Health System

Pharmacy 54, 397–405.

Bizovi KE, Beckley BE, McDade MC,

Adams AL, Lowe RA, Zechnich AD &

Hedges JR (2002) The effect of com-

puter-assisted prescription writing on

emergency department prescription

errors. Academic Emergency Medicine

9, 1168–1175.

Caruba T, Colombet I, Gillaizeau F, Bruni

V, Korb V, Prognon P, Begue D, Dur-

ieux P & Sabatier B (2010) Chronology

of prescribing error during the hospital

stay and prediction of pharmacist’s

alerts overriding: a prospective analysis.

BMC Health Services Research 10, 13.

Colpaert K, Claus B, Somers A, Vandewo-

ude K, Robays H & Decruyenaere J

(2006) Impact of computerized physi-

cian order entry on medication pre-

scription errors in the intensive care

unit: a controlled cross-sectional trial.

Critical Care 10, R21.

Coombes ID, Stowasser DA, Coombes JA &

Mitchell C (2008) Why do interns make

prescribing errors? A qualitative study

The Medical Journal of Australia 188,

89–94.

R Shawahna et al.

3244 � 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245

Dean B, Barber N & Schachter M (2000)

What is a prescribing error? Quality in

Health Care 9, 232–237.

Dean B, Schachter M, Vincent C & Barber

N (2002a) Causes of prescribing errors

in hospital inpatients: a prospective

study. Lancet 359, 1373–1378.

Dean B, Schachter M, Vincent C & Barber

N (2002b) Prescribing errors in hospital

inpatients: their incidence and clinical

significance. Quality and Safety in

Health Care 11, 340–344.

Dean Franklin B, Vincent C, Schachter M &

Barber N (2005) The incidence of pre-

scribing errors in hospital inpatients: an

overview of the research methods. Drug

Safety 28, 891–900.

DesRoches CM, Campbell EG, Vogeli C,

Zheng J, Rao SR, Shields AE, Donelan

K, Rosenbaum S, Bristol SJ & Jha AK

(2010) Electronic health records’ lim-

ited successes suggest more targeted

uses. Health Affairs 29, 639–646.

Fitzgerald RJ (2009) Medication errors: the

importance of an accurate drug history.

British Journal of Clinical Pharmacol-

ogy 67, 671–675.

Friedman AL, Geoghegan SR, Sowers NM,

Kulkarni S & Formica RN Jr (2007)

Medication errors in the outpatient

setting: classification and root cause

analysis. Archives of Surgery 142, 278–

283. discussion 284.

Ghaleb MA, Barber N, Dean Franklin B &

Wong IC (2005) What constitutes a

prescribing error in paediatrics? Qual-

ity and Safety in Health Care 14, 352–

357.

Ghaleb MA, Barber N, Franklin BD &

Wong IC (2010) The incidence and

nature of prescribing and medication

administration errors in paediatric in-

patients. Archives of Disease in Child-

hood 95, 113–118.

Joint Formulary Committee (2007) British

National Formulary, 54th edn. British

Medical Association and Royal Phar-

maceutical Society of Great Britain,

London.

Kaushal R, Bates DW, Landrigan C, McK-

enna KJ, Clapp MD, Federico F &

Goldmann DA (2001) Medication errors

and adverse drug events in pediatric

inpatients. JAMA 285, 2114–2120.

Lesar TS (2002) Prescribing errors involving

medication dosage forms. Journal of

General Internal Medicine 17, 579–587.

Lewis PJ, Dornan T, Taylor D, Tully MP,

Wass V & Ashcroft DM (2009) Preva-

lence, incidence and nature of pre-

scribing errors in hospital inpatients: a

systematic review. Drug Safety 32,

379–389.

Pollock M, Bazaldua OV & Dobbie AE

(2007) Appropriate prescribing of med-

ications: an eight-step approach. Amer-

ican Family Physician 75, 231–236.

Potts AL, Barr FE, Gregory DF, Wright L &

Patel NR (2004) Computerized physi-

cian order entry and medication errors

in a pediatric critical care unit. Pediat-

rics 113, 59–63.

Shaw J, Harris P, Keogh G, Graudins L,

Perks E & Thomas PS (2003) Error

reduction: academic detailing as a

method to reduce incorrect prescrip-

tions. European Journal of Clinical

Pharmacology 59, 697–699.

Shawahna R & Rahman N-U (2008) Pre-

scribing errors in psychiatry depart-

ment: an audit from a hospital in

Lahore. Journal of Pakistan Psychiatric

Society 5, 31–33.

Shawahna R & Rahman N-U (2009)

Defining Prescribing Errors in General

Practice for the Hospitals in Pakistan.

Pakistan Armed Forces Medical Journal

2, 103–109.

Stone WM, Smith BE, Shaft JD, Nelson RD

& Money SR (2009) Impact of a com-

puterized physician order-entry system.

Journal of American College of Sur-

geons 208, 960–967. discussion 967–

969.

Thomsen LA, Winterstein AG, Sonderg-

aard B, Haugbolle LS & Melander A

(2007) Systematic review of the inci-

dence and characteristics of prevent-

able adverse drug events in

ambulatory care. Annals of Pharma-

cotherpy 41, 1411–1426.

van den Bemt PM, Postma MJ, van Roon

EN, Chow MC, Fijn R & Brouwers JR

(2002) Cost–benefit analysis of the

detection of prescribing errors by hos-

pital pharmacy staff. Drug Safety 25,

135–143.

The Journal of Clinical Nursing (JCN) is an international, peer reviewed journal that aims to promote a high standard of

clinically related scholarship which supports the practice and discipline of nursing.

For further information and full author guidelines, please visit JCN on the Wiley Online Library website: http://

wileyonlinelibrary.com/journal/jocn

Reasons to submit your paper to JCN:High-impact forum: one of the world’s most cited nursing journals and with an impact factor of 1Æ228 – ranked 23 of 85

within Thomson Reuters Journal Citation Report (Social Science – Nursing) in 2009.

One of the most read nursing journals in the world: over 1 million articles downloaded online per year and accessible in over

7000 libraries worldwide (including over 4000 in developing countries with free or low cost access).

Fast and easy online submission: online submission at http://mc.manuscriptcentral.com/jcnur.

Early View: rapid online publication (with doi for referencing) for accepted articles in final form, and fully citable.

Positive publishing experience: rapid double-blind peer review with constructive feedback.

Online Open: the option to make your article freely and openly accessible to non-subscribers upon publication in Wiley

Online Library, as well as the option to deposit the article in your preferred archive.

Quality and safety Electronic prescribing reduces prescribing error in public hospitals

� 2011 Blackwell Publishing Ltd, Journal of Clinical Nursing, 20, 3233–3245 3245