Research

Educating junior doctors and pharmacists to reduce

discharge prescribing of opioids for surgical patients:
a cluster randomised controlled trial
Ria E Hopkins1,2 , Thuy Bui1, Alex H Konstantatos1, Carolyn Arnold1,3, Dianna J Magliano4,5, Danny Liew5, Michael J Dooley 1,2

Abstract
The known: Opioid medications can cause harm and
dependence, and inappropriate prescribing for patients after Objectives: To evaluate whether educating junior doctors and
surgery or trauma, without planning for de-­escalation, exposes hospital pharmacists about analgesic prescribing improved
them to the risk of chronic use. discharge prescribing of opioids for opioid-­naïve patients after
surgical admissions.
The new: Delivery of a brief education module by an analgesic
stewardship pharmacist to junior clinicians and pharmacists was Design: Cluster randomised controlled trial, undertaken during the
followed by significantly reduced opioid prescribing for surgical first half of 2019.
patients at discharge, including prescribing of slow release Setting: The Alfred Hospital, a major Melbourne teaching hospital
formulations associated with greater risk. with 13 surgical units.
The implications: Educating junior clinicians and pharmacists Participants: Opioid-­naïve patients discharged from surgical units
about appropriate analgesia prescribing for surgical and trauma after a stay of at least 24 hours.
patients is an effective tool for reducing the prescribing of slow Intervention: Surgical units were randomised to the intervention or
release opioids. control arms. Interns, residents, and clinical pharmacists assigned to
intervention arm units attended education sessions, presented by
the hospital analgesic stewardship pharmacist, about appropriate
analgesic prescribing for patients in hospital surgical units.
Three-­quarters of surgical patients report moderate to severe
Main outcome measures: The patients prescribed slow release
pain after their procedures, and opioid medications are fre-
opioids on discharge from hospital during the baseline (1 February –
quently prescribed as first line treatment.1 Many patients con-
30 April 2018) and post-­intervention periods (17 February – 30 April
tinue to use opioids after leaving hospital; 49–92% of surgical 2019).
patients in the United States and Canada are prescribed opioids
Results: During the baseline period, 1369 intervention unit and
on discharge.1,2 Opioids can have adverse effects, including cen-
1014 control unit admissions were included in our analysis; during
tral nervous system depression and opioid-­induced ventilatory the evaluation period, 973 intervention unit and 706 control unit
impairment, and tolerance and hyperalgesia develop with sus- episodes were included. After adjusting for age, length of stay, pain
tained use.3 Excessive use can be fatal; the annual number of score, acute pain service involvement, and use of immediate release
opioid-­related deaths in the US increased 345% between 2001 opioids prior to admission, patients in the intervention group were
and 2016, to an estimated 47 600 in 2017.4,5 In Australia, deaths prescribed slow release opioids at discharge less frequently than
caused by oxycodone, morphine and codeine increased 102% patients in the control group (adjusted odds ratio [aOR], 0.52; 95%
during 2006–2017, and deaths involving fentanyl, pethidine and CI, 0.35–0.77) and were more frequently discharged without any
tramadol increased 1000%.6 prescribed opioids following the intervention (aOR, 1.69; 95% CI,
1.24–2.30). Providing de-­escalation plans was more frequent for
Acute opioid therapy can lead to chronic use.7,8 It has been re- intervention than control group patients prescribed slow release
ported that patients prescribed opioids on discharge from sur- opioids on discharge post-­intervention (OR, 2.36; 95% CI, 1.25–4.45).
gical care are 44% more likely to be taking opioids one year Conclusions: Specific education for clinicians and pharmacists
later than those discharged without opioids.9 The likelihood about appropriate analgesic prescribing for surgical patients is
of long term opioid use after minor and major operations is effective in reducing prescribing of opioids at discharge.
similar.2,10 Trial registration: Australian New Zealand Clinical Trials Registry,
ACTRN12618000876291 (prospective).
In the United States, the Centers for Disease Control and
Prevention recommend not prescribing slow release opioids
for people with acute pain, prescribing the lowest adequate
opioid dose for patients with chronic pain and minimising the Inadequate knowledge of appropriate analgesia is a problem for
quantities supplied on discharge, and providing patients and prescribers, particularly junior medical officers.3,15,16 Education
general practitioners with documented post-­discharge plans.11 may optimise opioid prescribing, but most studies have evaluated
MJA 213 (9) ▪ 2 November 2020

Nevertheless, prescribing patterns in the US are highly vari- physician satisfaction with education or knowledge acquisition
able,12 and a 2017 review found that only 6–59% of opioids sup- rather than prescribing patterns.17–19 Studies examining prescrib-
plied after surgery were used by patients, suggesting that they ing have focused on specific patient groups (eg, outpatients or
are overprescribed.13 A review of post-­operative prescribing for patients with chronic pain), have evaluated limited selections of
more than 18 000 surgical patients found that 45% of the 6548 opioids, and have often not distinguished between slow and im-
who had not required opioids during the preceding 24 hours mediate release formulations or between opioid-­naïve and opioid-­
were prescribed opioids at discharge.14 tolerant patients, limiting the generalisability of their findings.20–24

1
Alfred Health, Melbourne, VIC. 2 Centre for Medicine Use and Safety, Monash University, Melbourne, VIC. 3 Central Clinical School, Monash University, Melbourne, VIC. 4 Baker IDI Heart and
Diabetes Institute, Melbourne, VIC. 5 School of Public Health and Preventive Medicine, Monash University, Melbourne, VIC. ria.hopkins@monash.edu ▪ doi: 10.5694/mja2.50812 ▪ See 417
Editorial (Schug).
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
In our cluster randomised controlled trial, we exam-
ined whether educating junior doctors and pharma- 1 Timeline of study for evaluating the effect on prescribing of specific
cists about analgesic prescribing improved discharge education for junior medical officers and pharmacists about appropriate
prescribing of opioids by reducing the prescribing of analgesia
slow and immediate release opioids, the daily dose
prescribed, and the quantity of opioids supplied to
opioid-­naïve surgical patients.

Methods

Our pragmatic cluster randomised controlled trial

was undertaken at the Alfred Hospital, a major
teaching hospital in Melbourne. About 1000 patients
are discharged home or to residential aged care from
its 13 surgical units each month. Surgical units were
allocated to the intervention or control arm by simple
computer-­supported randomisation; two units were
combined during randomisation because some staff * Two units were combined for because some staff members worked in both units. ◆

members worked in both units. The study was pro-

spectively registered with the Australian New Zealand Clinical prescribing by individual clinicians within units, surgical units
Trials Registry (ACTRN12618000876291, 23 May 2018). Major were treated as clusters.
institutional changes (electronic record roll-­out, building reno-
We analysed discharge opioid prescribing data for adult pa-
vations) caused a four-­month delay in implementing and evalu-
tients (aged 18 years or more) discharged home or to residen-
ating the intervention, as well as a two-­week shortening of the
tial care after admission to surgical units for at least 24 hours,
evaluation period, but the study otherwise adhered to the regis-
identified from discharge coding. Exclusion criteria included
tered protocol. Prescribing records at the hospital changed from
pre-­admission use of immediate or slow release opioids (except
paper to electronic charts and discharge prescriptions during
pro re nata immediate release opioid use), opioid agonist therapy,
the period of the study.
transfer to another hospital or rehabilitation facility, and missing
discharge documentation.
Intervention and control groups
Other information collected included age, sex, substance use
In the intervention arm, surgical interns, residents, and
disorder, intravenous drug use, immediate release opioid use
clinical pharmacists received single 30-­ m inute face-­
to-­ face
prior to surgical admission, hospital length of stay, surgical
group education sessions delivered by the hospital analgesic
procedure, elective or emergency procedure, intensive care
stewardship pharmacist. The learning objectives included
unit (ICU) admission, acute pain service or palliative care
understanding pain assessment, the importance of multi-
team involvement, and final Verbal Numerical Rating Scale
modal analgesia, analgesia selection, and discharge and de-­
(VNRS; scores range from 0, no pain, to 10, worst pain imagin-
escalation plans, including communication with patients and
able) score before discharge.
general practitioners. Staff members were invited to attend, at
their convenience, one of the ten sessions offered during 1–14
February 2019. Five additional sessions were offered during Primary and secondary outcomes
17–24 April 2019 for pharmacists and interns newly rostered to The primary outcome was the difference between the propor-
the intervention units. Attendance was encouraged by senior tions of patients prescribed slow release opioids on discharge
unit staff but was not compulsory. All eligible clinicians re- in 2018 and 2019 for each of the intervention and control arms
ceived the presentation slides by email; no subsequent assess- (Box 1). An audit in March 2018 found that 47 of 197 opioid-­
ment was undertaken. naïve surgical patients (24%) had been prescribed slow release
In the control arm, routine continuing education was pro- opioids on discharge. To detect a 5 percentage point change in
vided to junior medical officers and pharmacy staff, but not prescribing in the intervention group, and allowing a 3 per-
the intervention-­specific session. All staff members had access centage point change in the control group (90% power; α =
to institutional resources and hospital-­ approved guidelines, 0.05 [two-­sided]), we calculated that 853 admissions per group
including the hospital guideline on pharmacological and non-­ were required.
pharmacological management of acute pain. Participating clini- Secondary outcomes were the prescribing of immediate re-
MJA 213 (9) ▪ 2 November 2020

cians were not informed about the nature of the study; they were lease opioids on discharge, the proportion of patients dis-
told that the education module would be provided in stages to charged without opioid prescription, prescribed daily dose of
staff in all surgical units. slow release opioid (as oral morphine equivalent, calculated
with the Faculty of Pain Medicine Opioid Calculator: www.
Evaluation of prescribing opioi​dcalc ​u lator.com.au), quantity of opioid supplied on
discharge (dose units), documented slow release opioid de-­
Discharge prescribing was evaluated during the 10-­week pe-
escalation plan, and non-­opioid adjuvant prescribing. Dose
riod 17 February – 30 April 2019 and compared with prescrib-
units were tablets, capsules, and transdermal patches; for liq-
ing during a three-­month baseline period (1 February – 30 April
uids, dose units were calculated from the prescribed dose and
2018), to control for differences between prescribing in the surgi-
bottle size.
cal units assigned to the intervention and control arms (Box 1).
418 Data were collected retrospectively from discharge summaries Outcome assessors were not blinded with respect to whether
and electronic medical records. As it was impractical to evaluate data were from intervention or control group participants.
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Statistical analysis During 1 February – 30 April 2018, 2383 of 2685 admissions were
included in our analysis (intervention units, 1369; control units,
We summarised patient characteristics as descriptive statistics. 1014); during 17 February – 30 April 2019, 1679 of 1916 episodes
We assessed differences between groups in univariate analyses were included (intervention units, 973; control units, 706) (Box 2).
(Fisher exact or Mann–Whitney U tests). The normality of residu-
als for continuous outcomes was examined in Shapiro–Wilks For the included admissions during 1 February 2018 – 30 April
tests and Q–Q plots to determine appropriate statistical analysis. 2018, differences between the intervention and control units
Variables for which moderately statistically significant differ- were moderately statistically significant (P ≤ 0.10) for age, sex,
ences between groups (P ≤ 0.10) were evident in the 2018 base- length of stay, surgical procedure, elective procedure, ICU ad-
line data were included in stepwise multivariable models. Pain mission, substance use disorder, pain score and acute pain ser-
score and pre-­admission immediate release opioid use were in- vice referral (Box 3); these variables were therefore included in
cluded as variables in all models as we expected they would be our multivariate models, as was pre-­admission immediate re-
associated with discharge prescribing. Associations between the lease opioid use.
intervention and prescribing were assessed in mixed regression
models (logistic and negative binomial), with interaction terms Primary outcome
for study allocation and time period; odds ratios (ORs) and inci-
The proportion of discharged patients prescribed slow release
dent rate ratios (IRRs) are reported with 95% confidence intervals
opioids during the 2019 evaluation period was lower than dur-
(CIs). Surgical units were included as random effects in all mod-
ing the 2018 baseline period in both study arms: by 15.0 percent-
els to account for clustering. Intra-­class correlation was calculated
age points in the intervention group (2018, 395 of 1369 [28.8%];
for the primary outcome to assess variability among participants
2019, 134 of 973 [13.8%]; P < 0.001) and 6.4 percentage points in
within clusters. All analyses were performed in Stata/IC 15.0.
the control group (2018, 231 of 1014 [22.8%]; 116 of 706 [16.4%];
P = 0.001) (Box 4).
Ethics approval
The investigation was approved by the Human Research Ethics
Secondary outcomes
Committees of Alfred Health (reference, 226/18) and Monash
University (reference, 13859); individual consent by clinicians Patients in the intervention units were prescribed slow re-
and patients was not required. lease opioids at discharge significantly less frequently than
patients in the control units following the intervention, both
Results before (OR, 0.61; 95% CI, 0.43–0.88) and after adjusting for age,
length of stay, pain score, acute pain service involvement, and
In the intervention arm, all invited clinical pharmacists attended use of immediate release opioids pro re nata prior to admis-
education sessions (eight in February, six in April); four of eight sion (adjusted OR, 0.52; 95% CI, 0.35–0.77) (Box 5). The intra-­
interns and three of eight invited residents attended sessions in class correlation was 0.25, suggesting moderate within-­c luster
February, and four of eight invited interns in April. variability.

2 Screening and evaluation of admissions for inclusion in our analysis

MJA 213 (9) ▪ 2 November 2020

419
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3 Characteristics of participants admitted to surgical units, 1 February 2018 – 30 April 2018, by study allocation
Control
Intervention surgical units surgical units P

Number of patients 1369 1014

Age (years), median (IQR) 51 (34–66) 55 (36–70) 0.001

Sex (men) 759 (55%) 681 (67%) < 0.001

Length of stay (days), median (IQR) 2.8 (1.6–5.2) 2.2 (1.3–5.0) < 0.001

Surgical procedure 823 (60%) 843 (83%) < 0.001

Surgical procedures that were elective 427 [52%] 470 [56%] 0.10

Intensive care unit admission 92 (7%) 15 (1%) < 0.001

Substance use disorder 81 (6%) 32 (3%) 0.004

Intravenous drug use 10 (0.7%) 5 (0.5%) 0.28

Immediate release opioid use pro re nata prior to admission 109 (8%) 96 (9%) 0.21

Verbal numerical rating scale, mean (SD) 1.29 (2.02) 1.15 (1.94) 0.14

Acute pain service referral 111 (8%) 46 (4%) < 0.001

IRQ = interquartile range; SD = standard deviation. ◆

4 Opioid prescribing on discharge of surgical patients during the baseline (1 February 2018 – 30 April 2018) and post-­intervention
periods (17 February 2019 – 30 April 2019), by study allocation
Intervention group Control group

2018 2019 P 2018 2019 P

Number of patients 1369 973 1014 706

Slow release opioid 395 (28.8%) 134 (13.8%) < 0.001 231 (22.8%) 116 (16.4%) 0.001

Oral morphine equivalent 30 (15–30) 15 (15–30) < 0.001 30 (15–30) 30 (15–30) 0.06
(mg), median (IQR)*

Dose units,† median (IQR)* 10 (8–14) 10 (6–14) < 0.001 10 (10–14) 10 (6–14) 0.002

Immediate release opioid 749 (54.7%) 530 (54.5%) 0.93 538 (53.1%) 436 (61.8%) < 0.001
† ‡
Dose units, median (IQR) 10 (5–10) 10 (5–10) 0.32 10 (10–15) 10 (6–10) 0.014

Total opioid quantity, dose 5 (0–12) 4 (0–10) 0.001 5 (0–15) 6 (0–10) 0.40
units,† median (IQR)

No opioid 546 (39.9%) 437 (44.9%) 0.015 459 (45.2%) 267 (37.8%) 0.002

IQR = interquartile range. * Data for patients prescribed slow release opioids. † Tablets, capsules, transdermal patches; for liquids, dose units were calculated from the prescribed dose and
bottle size. ‡ Data for patients prescribed immediate release opioids.◆

For the intervention surgical units, the proportion of patients dis- prescribed opioids following the intervention (aOR, 1.69; 95% CI,
charged without prescribed opioids was significantly greater in 1.24–2.30) (Box 6).
2019 than in 2018 (44.9% v 39.9%; P = 0.015); the proportions pre-
Among the 876 patients prescribed slow release opioids on dis-
scribed immediate release opioids were similar (54.5% v 54.7%);
charge, the odds of being discharged with a documented de-­
and the median prescribed daily dose of slow release opioid
MJA 213 (9) ▪ 2 November 2020

escalation plan were greater for the intervention (2018, 215 of 395
(for patients prescribed slow release opioids) and median total
[54%]; 2019, 93 of 134 [69%]) than the control group (2018, 163 of
opioid quantity supplied on discharge (all patients) were each
231 [70%]; 2019, 77 of 116 [66%]) following the intervention (OR,
lower. For the control surgical units, the proportion of patients
2.36; 95% CI, 1.25–4.45). No significant interaction effect was ob-
prescribed immediate release opioids was greater in 2019 than
served between time and study allocation, and the odds of being
in 2018 (61.8% v 53.1%; P < 0.001) and that of patients discharged
prescribed any non-­opioid adjuvant medication (any adjuvant:
without a prescribed opioid smaller (37.8% v 45.2%; P = 0.002); the
OR, 0.94; 95% CI, 0.69–1.28) (Box 7).
median prescribed daily dose of slow release opioid was higher
and the median opioid quantity supplied on discharge smaller
in 2019 than in 2018 (Box 4). After adjusting for age, length of Discussion
stay, pain score, acute pain service involvement, and use of im-
420 mediate release opioids prior to admission, patients in the in- In our cluster randomised, controlled trial, the odds of surgical
tervention group were more frequently discharged without any inpatients being prescribed opioid medications at discharge were
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
influenced prescribing.25 Multivariate anal-
5 Prescribing of slow release opioids on discharge of surgical patients: univariate ysis identified that the odds of slow release
and multivariate analyses of 1720 control unit admissions and 2342 intervention opioid prescribing for patients from the inter-
unit admissions vention units were lower than for those from
Adjusted odds ratio* control units, and that the odds of patients in
Variable Odds ratio (95% CI) (95% CI) intervention group units not being prescribed
Intervention 1.09 (0.29–4.04) 1.01 (0.29–3.54) any opioid medications on discharge were sig-
nificantly greater than for patients in control
Time 0.52 (0.39–0.69) 0.40 (0.30–0.55) units.
Intervention × time 0.61 (0.43–0.88) 0.52 (0.35–0.77)
The control and intervention groups differed
Age, per year 0.99 (0.99–1.00) 0.99 (0.98–0.99) during the baseline period with regard to cer-
Length of stay, per day 1.03 (1.02–1.04) 1.02 (1.01–1.03)
tain features, probably reflecting differences
in casemix. Characteristics such as acute pain
Verbal numerical pain score, per 1.17 (1.13–1.21) 1.23 (1.18–1.28) service referral significantly influenced opioid
point
prescribing patterns; as pain management for
Acute pain service 6.39 (4.79–8.52) 7.17 (5.22–9.86) patients referred to acute pain services is more
Immediate release opioid use pro re 1.54 (1.17–2.03) 1.41 (1.05–1.91) difficult, these patients are more likely to need
nata prior to admission opioid analgesia at discharge. However, the
odds of opioid prescribing were also lower for
* Adjusted for age, length of stay, pain score, acute pain service involvement, and use of immediate release opioids
pro re nata prior to admission. Sex, surgical procedure, elective v emergency procedure, intensive care unit admis-
the intervention group in multivariate analy-
sion, and substance use disorder, were removed in a backwards stepwise procedure (P > 0.10). ◆ ses after adjusting for acute pain service input.

The impact of education on slow release opi-

lower in 2019 than in 2018 in both the intervention and control oid prescribing at discharge has not previously been evaluated.
arms of the study. The decline in the proportion of patients pre- A study of inpatient opioid use found that the proportion of
scribed slow release opioids at discharge was greater for surgical patients receiving slow release opioids declined from 35% to
units in which junior clinicians had received specific education 16% following an education intervention, and that of patients
about appropriate analgesia prescribing than in control surgical discharged without opioids increased from 9% to 13%.21 Other
units. studies have reported that prescribed daily opioid doses declined
Increased awareness of the harms associated with opioids and by one-­half following educational interventions.20,22 The reduc-
institutional strategies may have generally reduced opioid pre- tions in our study were less pronounced, but the median baseline
scribing at the Alfred Hospital. The hospital did not introduce dose was lower (30 mg daily oral morphine equivalent) than in
specific policy changes during the study period, but the publica- previous studies (150 mg,20 90 mg21). Similarly, the reductions in
tion in March 2018 by the Australian and New Zealand College quantities supplied were not as pronounced as in previous stud-
of Anaesthetists of their Position statement on the use of slow re- ies,20,22,24 but the baseline quantities were also lower in our study;
lease opioid preparations in the treatment of acute pain may have this is unsurprising, as opioids are now prescribed in smaller

6 Prescribing of opioids on discharge of surgical patients: univariate and multivariate analyses of 1720 control unit admissions and
2342 intervention unit admissions
Mixed regression models (logistic and negative binomial):
interaction effect (intervention × time)

Univariate analysis: Multivariate analysis:

Prescribing at discharge: categorical outcomes odds ratio (95% CI) adjusted odds ratio† (95% CI)

Immediate release opioid prescribed 0.69 (0.53–0.89) 0.74 (0.54–1.00)

Excluding patients who used immediate release opioids pro re nata 0.68 (0.52–0.90) 0.70 (0.51–0.96)
prior to admission

Both slow and immediate release opioids prescribed 0.69 (0.49–0.97) 0.67 (0.45–1.00)

Excluding patients who used immediate release opioid pro re nata prior 0.63 (0.44–0.90) 0.64 (0.42–0.97)
MJA 213 (9) ▪ 2 November 2020

to admission

No opioid prescribed 1.67 (1.29–2.16) 1.69 (1.24–2.30)

Prescribing at discharge: continuous outcomes: Incident rate ratio* (95% CI) Adjusted incident rate ratio† (95% CI)

Slow release oral morphine equivalent (mg) 0.57 (0.33–0.96) 0.58 (0.35–0.98)
‡
Slow release quantity (dose units) 0.50 (0.32–0.79) 0.52 (0.33–0.81)

Immediate release quantity (dose units)‡ 0.93 (0.74–1.17) 0.91 (0.73–1.14)

‡
Total opioid quantity (dose units) 0.79 (0.63–0.99) 0.78 (0.62–0.97)

CI = confidence interval. * An incident rate ratio of 0.57, for example, indicates that the median amount prescribed to patients from intervention units was 57% of that for patients discharged
from control units. † Adjusted for age, length of stay, pain score, acute pain service referral, and immediate release opioid use pro re nata prior to admission. ‡ Tablets, capsules, transdermal
patches; for liquids, dose units were calculated from the prescribed dose and bottle size. ◆ 421
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
detect meaningful differences between the control and post-­
7 Prescribing of non-­opioid adjuvant medications on discharge intervention periods in prescribing despite the lower number of
of surgical patients: univariate analysis of 1720 control unit episodes during the 2019 evaluation period than during the 2018
admissions and 2342 intervention unit admissions baseline period.
Interaction effect
(intervention × time): Cluster randomisation allowed comparisons of intervention
Non-­opioid adjuvant odds ratio (95% CI) and control arms within the baseline and post-­intervention
periods, controlling for changes in practice between the two
Any non-­opioid adjuvant 0.94 (0.69–1.28)
periods. We mitigated the problem of differences in baseline
Gabapentinoid 0.78 (0.47–1.28) variables between the two groups by undertaking multivari-
Non-­s teroidal anti-­inflammatory drug 0.93 (0.70–1.24)
ate analyses.

Paracetamol 0.90 (0.67–1.21) Education was delivered by an analgesic stewardship pharma-

cist, but could be delivered by another clinician experienced in
Other 1.03 (0.37–2.83)
peri-­operative or pain medicine. We included clinical pharma-
CI = confidence interval. ◆ cists in the intervention because of previous findings regarding
their roles in medication management.27 The Alfred Hospital
has a seven-­day unit-­based clinical pharmacy service, and the
quantities because of growing awareness of the potential harms
impact of a similar intervention at institutions with more limited
and poor functional outcomes associated with their long term
pharmacy services may be different. As doctors and pharma-
use. Nevertheless, we found that educational interventions can cists were educated together, it was not possible to separately
have a significant effect even when baseline opioid prescribing evaluate the effect of the intervention on the two groups. The
rates and quantities are low. appropriateness of prescribing for individual patients was not
We found that the odds of a plan for slow release opioid de-­ assessed.
escalation being included in discharge summaries were 2.4 Clinicians were blinded to the purposes of the study to reduce
times as high for patients from the intervention units as for observer bias. Diffusion of the educational message was possi-
those from control units following the intervention. Effective ble, particularly if presentation slides were shared; this would
communication is crucial for ensuring the safe transfer of care have improved the control group outcomes. We could not con-
after discharge, but omissions in the reporting of medication trol for the effect of clinicians working outside normal hours
changes in hospital have also been described previously.26 in units other than their home unit. Only about one-­half of the
We included all available opioid formulations in our evalu- invited doctors attended the education sessions, and we cannot
estimate the potential effect of more complete attendance.
ation of total daily oral morphine equivalents prescribed on
discharge. Reducing the prescribing of one agent may be as-
Conclusion
sociated with increases in others; in one study, for example,
tramadol and morphine prescribing increased following an Providing brief education sessions for junior clinicians and clin-
education intervention in an emergency department that re- ical pharmacists was followed by significantly reduced opioid
duced pethidine use.23 Opioid medications could be spared prescribing at discharge for opioid-­naïve surgical patients, sug-
by providing non-­opioid analgesics, but non-­opioid analgesic gesting that education is an effective evidence-­based strategy
prescribing patterns were not influenced by our intervention. for optimising opioid prescribing in acute care. However, junior
Adjuvant prescribing may have already been appropriate, or medical officers in teaching hospitals frequently rotate between
optimising opioid-­sparing strategies may need more emphasis specialties, and the value of the education module would de-
in education. pend on its regular delivery, which is time-­consuming and per-
haps unsustainable. Possible solutions include incorporating the
We included patients with substance use disorders, but ex- module into orientation programs and offering it online. Future
cluded those prescribed opioid agonist therapy. Patients treated iterations of the intervention could include post-­education as-
in ICUs or undergoing emergency surgery have been excluded sessment and personalised feedback. Longer term evaluation of
in previous studies, but we included these treatment variables outcomes is required, as is evaluation of the impact of the inter-
in our multivariate analyses. We also included all surgical units vention on prescribing appropriateness.
at the hospital, whereas earlier studies have been restricted to
Acknowledgements: We gratefully acknowledge the contribution of Agus Salim to
single specialties or procedure categories, limiting the general- the statistical analysis of our data.
isability of their results. We found that education can be deliv-
Competing interests: No relevant disclosures.
ered with benefit across a range of surgical unit types.
MJA 213 (9) ▪ 2 November 2020

Data sharing statement: De-­identified and aggregate data for this study may be
Strengths and limitations made available upon request from the corresponding author. ■
Our large sample size allows confidence in our findings. Because Received 22 November 2019, accepted 5 May 2020
of organisational changes, the post-­intervention period was re-
duced by two weeks, but the study was adequately powered to © 2020 AMPCo Pty Ltd

  1 Fujii MH, Hodges AC, Russell RL, et al. Post-­ surgery: population based cohort study. BMJ   4 Gomes T, Tadrous M, Mamdani MM, et al. The
discharge opioid prescribing and use after 2014; 348: g1251. burden of opioid-­related mortality in the United
common surgical procedure. J Am Coll Surgeons   3 Adewumi AD, Staatz CE, Hollingworth SA, et al. States. JAMA Netw Open 2018; 1: e180217.
2018; 226: 1004–1012. Prescription opioid fatalities: examining why the   5 Scholl L, Seth P, Kariisa M, et al. Drug and opioid-­
422   2 Clarke H, Soneji N, Ko DT, et al. Rates and risk healer could be the culprit. Drug Saf 2018; 41: involved overdose deaths: United States, 2013–2017.
factors for prolonged opioid use after major 1023–1033. MMRW Morb Mortal Wkly Rep 2019; 67: 1419–1427.
 Research

13265377, 2020, 9, Downloaded from https://onlinelibrary.wiley.com/doi/10.5694/mja2.50812 by CochraneChina, Wiley Online Library on [14/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
  6 Penington Institute. Australia’s annual overdose 14 Chen EY, Marcantonio A, Tornetta P. Correlation 21 Oyler D, Bernard A, VanHoose JD, et al.
report 2019. Aug 2019. https://www.penin​gton. between 24-­hour predischarge opioid use Minimizing opioid use after acute major trauma.
org.au/wp-conte​nt/uploa​ds/Austr​alias-Annual- and amount of opioids prescribed at hospital Am J Health Syst Pharm 2018; 75: 105–110.
Overd​ose-Report-2019.pdf (viewed Oct 2019). discharge. JAMA Surg 2018; 153: e174859. 22 Stanek JJ, Renslow MA, Kalliainen LK. The effect
  7 Deyo RA, Hallvik SE, Hildebran C, et al. 15 Barth KS, Guille C, McCauley J, Brady KT. of an educational program on opioid prescription
Association between initial opioid prescribing Targeting practitioners: a review of guidelines, patterns in hand surgery: a quality improvement
patterns and subsequent long-­term use among training, and policy in pain management. Drug program. J Hand Surg 2015; 40: 341–346.
opioid-­naive patients: a statewide retrospective Alcohol Depend 2017; 173: S22–S30. 23 Kaye KI, Welch SA, Graudins LV, et al. Pethidine
cohort study. J Gen Intern Med 2017; 32: 21–27. 16 Cushman PA, Liebschutz JM, Hodgkin JG, in emergency departments: promoting
  8 Calcaterra SL, Yamashita TE, Min SJ, et al. Opioid et al. What do providers want to know about evidence-­based prescribing. Med J Aust. 2005;
prescribing at hospital discharge contributes opioid prescribing? A qualitative analysis 183: 129–133. https://www.mja.com.au/journ​
to chronic opioid use. J Gen Intern Med 2016; 31: of their questions. Subst Abus 2017; 38: al/2005/183/3/pethi​dine-emerg​ency-depar​
478–485. 222–229. tments-promo​ting-evide​nce-based-presc​ribing.
  9 Alam A, Gomes T, Zheng H, et al. Long-­ 17 Roth CS, Burgess DJ. Changing residents’ beliefs 24 Hill VM, Stucke SR, McMahon LM, et al. An
term analgesic use after low-­risk surgery: a and concerns about treating chronic noncancer educational intervention decreases opioid
retrospective cohort study. Arch Intern Med pain with opioids: evaluation of a pilot prescribing after general surgical operations.
2012; 172: 425–430. workshop. Pain Med 2008; 9: 890–902. Ann Surg 2018; 267: 468–472.
10 Brummett CM, Waljee JF, Goesling J, et al. New 18 Carter A, Spagnoletti C, Rothenberger SD, 25 Australian New Zealand College of Anaesthetists.
persistent opioid use after minor and major McNeil M. Educating faculty and resident Position statement on the use of slow-release
surgical procedures in US adults. JAMA Surg physicians on safe opioid prescribing for chronic opioid preparations in the treatment of acute
2017; 152: e170504-e. pain: impact of a brief curriculum on knowledge, pain. Mar 2018. https://www.anzca.edu.au/getat​
11 Frieden TR, Houry D. Reducing the risks of relief: confidence, and attitudes. J Gen Intern Med tachm​ent/d9e2a​7c5-0f17-42d3-bda7-c6dae​7e55c​
the CDC opioid-­prescribing guideline. N Engl J 2018; 33 (2 Suppl): S174. ed/Posit​ion-state​ment-on-the-use-of-slow-relea​
Med 2016; 374: 1501–1504. 19 Liossi C, Failo A, Schoth DE, et al. The se-opioid-prepa​ratio​ns-in-the-treat​ment-of-
effectiveness of online pain resources for health acute-pain (viewed Oct 2019).
12 Hill VM, McMahon ML, Stucke SR, Barth RJ.
Wide variation and excessive dosage of opioid professionals: a systematic review with subset 26 Belleli E, Naccarella L, Pirotta M. Communication
prescriptions for common general surgical meta-­analysis of educational intervention at the interface between hospitals and primary
procedures. Ann Surg 2017; 265: 709–714. studies. Pain 2018; 159: 631–643. care: a general practice audit of hospital
13 Feinberg AE, Chesney TR, Srikandarajah S, 20 Donaldson SR, Harding AM, Taylor SE, et al. discharge summaries. Aust Fam Physician 2013;
et al. Best Practice in Surgery Group. Opioid Evaluation of a targeted prescriber education 42: 886–890.
use after discharge in postoperative patients: intervention on emergency department 27 Nissen L. Current status of pharmacist influences
a systematic review. Ann Surg 2018; 267: discharge oxycodone prescribing. Emerg Med on prescribing of medicines. Am J Health Syst
1056–1062. Australas 2017; 29: 400–406. Pharm 2009; 66 (5 Suppl 3): S29–S34. ■

MJA 213 (9) ▪ 2 November 2020

423