Articleonimprovingmedicationerror.pdf

Improving Medication Safety in Psychiatry – A ControlledIntervention Study of Nurse Involvement in Avoidance of

Potentially Inappropriate PrescriptionsAnn L. Soerensen1,2, Marianne Lisby3, Lars P. Nielsen4, Birgitte K. Poulsen4 and Jan Mainz5,6,7

1Danish Center for Healthcare Improvements, Faculty of Social Science, Aalborg University, Aalborg, Denmark, 2University College of NorthernDenmark, Aalborg, Denmark, 3Research Center for Emergency Medicine, Aarhus University Hospital, Aarhus, Denmark, 4Department of Clinical

Pharmacology, Aarhus University Hospital, Aarhus, Denmark, 5Department of Psychiatry, Aalborg University Hospital, Aalborg, Denmark,6Department of Clinical Medicine, Faculty of Health, Aalborg University, Aalborg, Denmark and 7Department of Community Mental Health, The

University of Haifa, Haifa, Israel

(Received 9 October 2017; Accepted 12 February 2018)

Abstract: The aim of this controlled, before-and-after study in the Department of Psychiatry in a university hospital in Denmarkwas to examine the potential effects and characteristics of nurses reviewing psychiatric patients’ medication records to identifypotentially inappropriate prescriptions (PIPs). The control group and the intervention group each consisted of two bed units cho-sen based on patients’ diagnoses and age categories. There were 396 patients (age ≥18 years) included in the study. Senior clini-cal pharmacology physicians performed medication reviews for all patients in the study; these medication reviews wereconsidered gold standard. The intervention group: nurses were given a pharmacology course after which the nurses reviewedmedication lists and subsequently conferred any identified PIPs with physicians. The control group: medication was reviewed asusual and nurses did not participate. Primary outcome measure was the potential difference in PIPs between the control groupand the intervention group, analysed in two ways: (i) difference in mean number of PIPs and (ii) difference in number of patientsexposed to ≥1 PIP, using regression analysis with an approximated difference-in-difference (DID) approach. Secondary outcomemeasure was characteristics of PIPs where physicians responded to nurse-identified PIPs. The DID between intervention groupand control group for mean number of PIPs per patient was �0.23 (�1.07 to 0.60), and for number of patients receiving ≥1 PIP,the odds ratio was 0.61 (0.25 to 1.46). Physicians changed most prescriptions in the category interaction between drugs. Nursescould not significantly reduce the prevalence of PIPs for psychiatric patients.

Potentially inappropriate prescriptions (PIPs) are common andsignificant in older patients [1,2] and present a specific chal-lenge in psychiatry [3–6]. There appears to be limited researchon PIP in general psychiatric populations that include youngerpatients. However, a Danish study estimated the prevalence ofPIPs in newly admitted psychiatric patients to affect 59% ofpatients, and 12% of the identified PIPs were assessed as poten-tially fatal [7]. There is a need to reinforce medication safety ini-tiatives in psychiatry by exploring new initiatives and engagingother staff groups such as nurses [8]. Nurses are professionalswho have received training in pharmacology and medicinesmanagement during their education. However, researchers sug-gest nurses have insufficient knowledge of the medications theyhandle [9,10] and should receive regular medication manage-ment courses [11]. Nurses are also the staff group who spendmost time at the patients’ bedside, continually observing andmonitoring the patient for both effects and side effects of medi-cations [12]. Nurses are key members of the psychiatric teamsurrounding the patient [13,14] and provide continuous vitalknowledge about patients and their medications [15].In studies on nurses’ role in identifying and mitigating medi-

cal errors, nurses often appear to be the last barricade between

patients and many medical errors [16–19]. In much of the litera-ture mentioned above, errors relating to medications are themost prevalent issue. Studies demonstrate how nurses identifydrug-related problems (DRP) and improve medicines manage-ment [20–24], but similar studies are scarce in psychiatry. OneJapanese study demonstrated improvement in Global Assess-ment of Functioning and a trend towards better acceptance ofmedication by psychiatric patients when physicians – based onnurses’ reports and suggestions – changed medications [25], anda British cluster randomized study demonstrated significantimprovement in total Positive and Negative Syndrome Scale(PANSS) score in patients with schizophrenia after a medicationmanagement training course for mental health nurses [26]. Theaim of this study was to examine the characteristics and magni-tude of PIPs as well as the potential effect of nurses’ systematicreviews of medication records on the appropriateness of medica-tions prescribed to newly admitted psychiatric patients. Ourhypothesis was that pharmacological training of nurses and ageneral increased awareness of medication safety have thepotential to reduce PIPs in adult psychiatric inpatients.

Method

Definitions. Potentially inappropriate prescription is defined asprescribing that introduces a significant risk of an adverse drug-relatedevent where there is evidence of an equally or more effective but lower-

Author for correspondence: Ann Lykkegaard Soerensen, AalborgUniversity, Department of Clinical Medicine, Danish Center forHealthcare Improvements, Fibigerstraede 11, 9220 Aalborg, Denmark(e-mail: [email protected]).

Basic & Clinical Pharmacology & Toxicology, 2018, 123, 174–181 Doi: 10.1111/bcpt.12989

risk alternative therapy available for the same condition [2,27].Additionally, PIP includes the use of drug combinations with knowndrug–drug interactions, drug–disease interactions, overdosing, use ofdrugs for a longer time than clinically indicated, as well as the omissionof prescribing drugs that are clinically indicated [2,27]. Studies onelderly patients provide this understanding of PIP but do not include thescenario of omission of therapeutic drug monitoring (TDM). TDM is aquantification of drug concentrations in blood plasma or serum to titratethe dose of individual patients. This quantification helps to obtain a drugconcentration associated with the highest possible probability ofresponse and tolerability with an acceptably low risk of toxicity [28].TDM is central to the prescribing of, for example, tricyclicantidepressants and certain antipsychotic- as well as mood-stabilizingdrugs [28]. In this study, omission of TDM was included and extendedto encompass electrocardiography and other general paraclinical datarequired for appropriate prescribing, for example, electrolytes in bloodplasma. Categories and elaboration on definitions of PIPs are given inTable S1.

Setting and study design. This controlled, interventional before-and-after study took place in the Aalborg University Hospital of Psychiatryand enrolled patients from November 2014 until June 2015. Thebaseline measurements were performed for a 2-month period inNovember 2014 and December 2014, and the intervention wasperformed for a 6-month period from January to June 2015. Thepsychiatric hospital has 249 beds across 14 bed units. Two bed unitswith 18 beds each were selected as the intervention group, and two bedunits with 12 and 22 beds were selected as the control group. Theintervention group and the control group each contained one bed unitspecializing in psychotic disorders and one bed unit specializing inaffective disorders. The bed units were purposely selected and matchedon patients’ diagnoses and age categories to prevent, if possible, otherfactors influencing outcome.

Ethical considerations. The Regional Ethical Committee for MedicalResearch in the North Denmark Region assessed the study and reportedthat ethical approval was not required according to Danish national lawon medical research. The Danish Health and Medicines Authority, TheDanish Data Protection Agency and hospital management approved thestudy. Furthermore, the study was registered with Clinicaltrials.gov,NCT02052505, 29 January 2014. Patients’ informed consent was notrequired according to Danish regulations as all measures to maintainpatients’ anonymity were taken. Senior clinical pharmacologyphysicians (SCPPs) performed medication reviews during the studywhich were only used descriptively and not in patients’ treatment.However, the SCPPs were ethically obliged to intervene in cases wherepatients’ safety was assessed acutely endangered. This happened onthree separate occasions.

Participants. The study included all patients admitted to the four bedunits, either from the psychiatric emergency department or through thepatient’s general practitioner, between November 2014 and June 2015.Exclusion criteria were as follows: terminally ill patients with ananticipated short life expectancy; an expected length of stay of lessthan 48 hr; patients transferred from another psychiatric unit orprevious inclusion in the study; and patients whom nurses failed toinclude and consequently whose medication lists were not reviewed.

Intervention. Senior clinical pharmacology physicians performedmedication reviews in both intervention and control bed units duringthe baseline and the intervention period according to the method usedand described in a previous study [7]. It was not possible to blind theSCPPs to time period nor allocation of group. A graphical overviewof the medication review procedure is provided as Figure S1. Themedication reviews served as a gold standard against which to

compare the intervention period and as a benchmark for the nurses’observations. After the baseline period, the intervention consisted of a5-day course for the participating nurses from the intervention bedunits followed by the same nurses carrying out reviews of medicationlists for newly admitted patients. The 5-day course covered generalpharmacology, psychopharmacology and treatment principles for someof the major mental disorders, principles for medication reviews,exercises in identifying PIPs and how to systematically documentrelevant observations. Lecturers were SCPPs (authors LPN and BKP),psychiatrists, physicians, a pharmacist and the course leader (authorALS). After the course, 22 registered nurses, across the twointerventional bed units, would review patients’ electronic medicationrecords to identify PIPs. The review was carried out within 1–2 daysafter admission and after the physician reviewed and released themedications to be administered. The nurses used the electronicmedication records, the electronic patient records, paraclinical data,their clinical knowledge in general and their knowledge of theindividual patient and consulted with nursing colleagues.The nurses recorded their observations and supplemented these

records with a short text explaining the rationale for assessing the pre-scription(s) as a PIP. The nurses also recorded the physicians’ subse-quent responses to the reported observations.

Data collection. Data included demographic details, clinical status,medications and a medication review performed by the SCPP. Allpatient records and medication records were electronic. The primaryoutcome was the potential difference in PIPs between the controlgroup and the intervention group following the nurses’ observations inthe intervention group. Secondary outcome included prevalence andtypes of PIPs where physicians responded to nurses’ observations andchanged medications.

Data analysis. The primary outcome was analysed in two ways:firstly, an estimation of potential difference in mean number of PIPs,and secondly, an estimation of potential difference in patientsreceiving ≥1 PIP using linear regression and logistic regression withan approximated difference-in-difference (DID) approach, respectively.The DID method permits an interpretation of data similar to arandomized controlled trial as differences that are constant over timeare taken into account [29]. An illustration of the DID method hasbeen provided as Figure S2. Regarding the first analysis of theprimary outcome, mean number of PIPs was considered the dependentvariable and time (before/after), group (control/intervention bed units)and interaction between time and the group as independent variablesin a linear regression. The estimated interaction variable is a DIDbetween intervention and control bed units. Regarding the secondanalysis of the primary outcome: patients receiving ≥1 PIP (presenceof PIP was coded: 1 and for absent: 0) were considered the dependentvariable and time, group and interaction between time and the groupas independent variables in a logistic regression. The DID approach asdescribed above was applied; however, due to the nonlinear nature ofthe logistic regression, this cannot be interpreted as a genuine DID.The estimated odds ratio for the interaction variable is an estimatedDID between intervention and control bed units. A statisticalsignificance level of 0.05 was applied for both main outcomes.Secondary outcome included prevalence and types of PIPs where

physicians responded to nurses’ observations and changed medications.Characteristics of the study population were provided using descrip-

tive statistics. Baseline measurements by the SCPPs were descriptiveand consisted of prevalence, categories and potential severity of PIPsfor all four bed units. The presentation of prevalence and proportionsof PIPs were in absolute numbers and percentages. Sample size calcu-lations preceded the intervention. Based on a previous study investi-gating PIP in a psychiatric population, the mean number of PIPs perpatient was assumed to be 1.8, the standard deviation 2.19 [7], and a

IMPROVING MEDICATION SAFETY IN PSYCHIATRY 175

reduction by 50% considered clinically relevant. To detect a reductionfrom 1.8 to 0.9 in mean number of PIPs per patient, at a two-sided0.05 significance level, a sample size of 94 patients per group duringthe intervention period was needed to ensure 80% power. Also basedon a previous study [7], and the composition of diagnoses representedin the included bed units, it was assumed that 50% of patients wouldbe affected by at least one PIP at baseline [7], and a reduction to 30%of patients affected by at least one PIP would be clinically relevant.To detect a difference of 20 percentage points in the proportion of

patients receiving at least one PIP, at a two-sided 0.05 significancelevel, a sample size of 94 patients per group during the interventionperiod was needed to ensure 80% power. It was eventually decided toset the sample size at 120 patients to account for potential loss ofpatients. Data analysis was performed in Stata/IC 14.0 (Stata Corp,College Station, TX, USA).

Results

Characteristics of patients.Inclusion and exclusion of patients, as illustrated in fig. 1,resulted in 396 patients in the study during baseline and inter-vention. Furthermore, there was no observable differencebetween the two groups regarding patient characteristics

(table 1). Moreover, during the intervention, there were no dif-ferences between patients reviewed by the nurses (n = 121)and included in the analysis and patients whom the nursesfailed to include (n = 15); these differences had been evalu-ated based on the following variables: age (p = 0.43), gender(p = 0.21), primary psychiatric disorder (p = 0.75), comorbidi-ties (p = 0.89), alcohol or substance abuse (p = 0.93) andmedication characteristics (p = 0.47).

Potentially inappropriate prescriptions.The distribution of PIPs was consistent on intervention andcontrol wards for the baseline period and the subsequent fol-low-up period indicating no other external factors influencingthe prevalence of PIP in the control wards.Over the 8-month data collection period, 396 patients and

2625 prescriptions, differentiated as 1894 regular prescriptionsand 758 pro re nata (p.r.n. (medication administered ‘asneeded’)) prescriptions, were reviewed by the SCPPs. Allmedication reviews by the SCPPs yielded a prevalence ofpatients with at least one PIP (n = 262) of (262/396) 66%. In

Fig. 1. Flow chart illustrating the inclusion and exclusion of patients in the study (n = 396).

176 ANN L. SOERENSEN ET AL.

total, 761 PIPs were identified by the SCPPs. The leading cat-egories of PIPs were interactions between drugs (232/761(30%)), omission of indication for treatment (119/761 (16%)),drug dosage too high (96/761 (13%)) and interaction between

drug and disease (84/761 (11%)). The primary outcome is dis-played in table 2 and demonstrates the scenario of potentialimprovement if all PIPs identified by SCPPs and nurses hadreceived relevant alterations.

Table 1.Characteristics of the entire study population (n = 396).

Patient characteristicsTotal

(n = 396)Intervention

bed unit A (n = 85)Intervention bedunit B (n = 94)

Control bedunit A (n = 56)

Control bedunit B (n = 161)

Age distributionMedian (years), IQR (years) 43, 30–56 42, 32–53 44, 26–61 36, 26–47 46, 32–58

Age category (years), n (%)<30 108 (27) 24 (28) 28 (30) 21 (38) 35 (22)30–44 112 (28) 27 (32) 22 (23) 19 (34) 44 (27)45–59 94 (24) 23 (27) 20 (21) 8 (14) 43 (27)≥60 82 (21) 11 (13) 24 (26) 8 (14) 39 (24)

Males, n (%) 169 (43) 43 (51) 35 (37) 30 (54) 61 (38)Primary psychiatric disorders, n (%)Schizophrenia and other psychotic disorders 121 (31) 59 (69) 3 (3) 45 (80) 14 (9)Affective disorders 175 (44) 7 (8) 64 (68) 4 (7) 100 (62)Other 100 (25) 19 (23) 27 (29) 7 (13) 47 (29)

Comorbidities1, n (%)

Cardiac disease2 96 (24) 16 (19) 21 (22) 12 (21) 47 (29)Diabetes mellitus 2 38 (10) 11 (13) 5 (5) 7 (13) 15 (9)Allergies 15 (4) 3 (4) 2 (2) 3 (5) 7 (4)COPD 29 (7) 5 (6) 6 (6) 6 (11) 12 (7)Other 97 (24) 14 (16) 23 (24) 10 (18) 50 (31)No comorbidities 273 (69) 64 (75) 67 (71) 41 (73) 101 (63)

Patients with alcohol and/or substance abuse, n (%) 118 (30) 28 (33) 21 (22) 18 (32) 51 (32)Medication characteristicsPatients prescribed regular medications, n (%) 358 (90) 75 (88) 84 (89) 55 (98) 144 (89)Median number of prescriptions(excluding p.r.n. medications), IQR

4, 2–7 4, 2–7 3, 2–6 5, 3–8 4, 2–7

Median number of prescriptions(including p.r.n. medications), IQR

5, 3–9 5, 2–9 5, 3–8 7, 4–11 5, 3–9

0 medications, n (%) 15 (4) 5 (6) 5 (5) 0 5 (3)1–5 medications, n (%) 185 (47) 41 (48) 50 (49) 22 (40) 76 (47)6–9 medications, n (%) 107 (27) 20 (24) 27 (26) 17 (30) 46 (29)≥10 medications, n (%) 89 (23) 19 (22) 20 (20) 17 (30) 34 (21)

IQR, interquartile range; COPD: chronic obstructive pulmonary disorder; p.r.n: pro re nata (medication administered ‘as needed’).1The levels of prevalence and percentages do not round up correctly because each patient could have had more than one diagnosis.2Cardiac disease includes patients with coronary artery disease, arrhythmias, congestive heart failure and subsequent conditions thereof.

Table 2.Potential improvements in number of potentially inappropriate prescriptions (PIPs) had all PIPs common for SCPPs and nurses been altered relevantly.

Baseline period Intervention period Difference, 95% confidence interval p-Value

Mean number of PIP per patient (�SD)Interventional bed units 1.69 � 1.79 1.55 � 2.00 0.14 (�0.47 to 0.76) 0.30Control bed units 1.84 � 1.99 1.92 � 2.34 �0.09 (�0.72 to 0.54) 0.92Difference-in-difference �0.23 (�1.07 to 0.60) 0.59

Baseline period Intervention period OR, 95% confidence interval p-Value

Number of patients prescribed≥1 PIP (%)Interventional bed units 38 (65.5) 65 (53.7) 0.61 (0.32–1.17) 0.14Control bed units 48 (65.8) 95 (66.0) 1.01 (0.56–1.83) 0.97Difference-in-difference 0.61 (0.25–1.46) 0.26

SD, standard deviation.Differences in means were compared using a Wilcoxon rank sum test, and difference-in-difference was estimated with a linear regression model. Oddsratios for the intervention and control bed units comparing before-and-after was estimated using logistic regression, and difference-in-difference wasestimated by the OR for the coefficient for interaction between groups (intervention bed unit/control bed unit) and time (before/after) using logisticregression.

IMPROVING MEDICATION SAFETY IN PSYCHIATRY 177

The DID between control and intervention bed units inmean number of PIPs per patient was statistically non-signifi-cant. The proportion of patients receiving ≥1 PIP in the inter-vention bed units was potentially reduced by 11.8 percentagepoints, compared to almost no variation observed in the con-trol bed units. However, this result was statistically insignifi-cant. Table 3 provides both an overview of types of PIPsidentified by SCPPs and nurses as well as physicians’responses to nurses’ observations in the intervention wardsduring the intervention. Categories containing five or fewerPIPs are not included in table 3. The excluded categories wereas follows: inappropriate dosing time, inappropriate dosageform, inappropriate route of administration and inappropriateduration of treatment.

Nurses’ identifications of PIP.In the six months nurses recorded observations of PIPs theyreviewed 121 patients with a total of 756 prescriptions. The 756prescriptions were distributed on 548 regular prescriptions and208 p.r.n. prescriptions (table 3). The nurses identified (62/121)51% of patients as having one or more PIP. The overlapbetween nurses and SCPPs, in the sample of patients reviewedby nurses, consisted of 38 PIPs equalling (38/224) 17% of theSCPPs’ identifications. When only including PIPs overlappingwith SCPPs’ assessments, the nurses identified two PIPs per100 prescriptions.

Secondary outcome.Nurses identified and presented physicians with PIPs in 11 cate-gories and physicians altered or wrote prescriptions in 10 of the11 categories, distributed across 25 patients. Physicians alteredor wrote most prescriptions in the category interaction betweendrugs, followed by omission of indication for treatment, as seenin table 3. Only in the category interaction between drugs didphysicians alter prescriptions assessed to be potentially harmful.During the intervention, the proportion of PIPs altered or writtenby physicians in response to nurses’ observations was 47/137(34%). In contrast, only (8/47) 17% of the PIPs identified by thenurses and leading to an intervention by the physicians werealso PIPs identified and assessed for severity by SCPPs.

Discussion

This is the first study to investigate nurses’ skills in identifyingPIPs in a psychiatric department of a university hospital by com-paring with SCPP’s medication reviews. The most importantfinding was a non-significant reduction in the mean number ofPIPs per patient and a non-significant reduction in the propor-tion of patients receiving ≥1 PIP. The study did not have enoughpower to demonstrate a statistically significant reduction inPIPs, as the nurses’ identification of PIPs was less extensivethan that, which was considered clinically relevant for the study.

Implications.This is only the second study of PIPs in general psychiatricpatients and the only study dealing with nurses’ role in

identifying and drawing attention to PIPs. In this study, nursesidentified PIPs but were not able to facilitate a statistically sig-nificant reduction in PIPs whether it was measured as meannumber of PIPs per patient or proportion of patients prescribedat least one PIP. Unfortunately, we have not been able to iden-tify comparable studies; thus, the results are being discussedin general.The prevalence of patients having at least one PIP was

found to be 66%, which is higher than the 59% previouslyreported [7]. In comparison, researchers reported prevalencesbetween 21% and 79% in an extensive systematic review ofcriteria used to identify PIPs in the elderly [30]. However,such comparisons must be interpreted with caution as psychi-atric patients may be vulnerable to certain practices of inap-propriate prescribing [31–32] and may not be directlycomparable to the elderly. Recently, in one of the first studiesof the prevalence of potentially inappropriate medications inolder patients admitted to psychiatric hospitals, a prevalenceof 47% and 79% using Beers criteria [33] and screening toolof older person’s prescriptions (STOPP)/START criteria [34],respectively, was found. These levels of prevalence supportour finding of 66%, although the methodologies differ becausewe used an implicit approach (medication reviews) and owingto the fact that both Beers and STOPP/START consist ofexplicit criteria for prescriptions to be avoided.One important finding was nurses identifying 17% of all

PIPs identified by SCPPs and that the majority of PIPs forboth SCPPs and nurses were in the category interactionbetween drugs. In general, there was only a small overlapbetween PIPs identified by nurses and SCPPs. SCPPs will per-form medication reviews based on an extensive understandingof pharmacological and medical properties. In comparison,nurses review medications using their clinical knowledge ofthe patient, the often limited pharmacology training theyreceived during their education, occasional self-studies and, inthis study, an intensive 5-day training course. A Swedishstudy on nurses identifying DRP demonstrated that, after aneducational intervention, 22% of DRPs identified by nurseswere potential drug interactions [35] as opposed to 45% of allnurse-identified PIPs in this present study. However, theSwedish study used a tool based on symptoms that are morelikely to identify adverse drug reactions and our study focusedon PIP, which primarily relies on prescriptions. Omission ofindication for treatment was the nurses’ second largest cate-gory of PIPs and shared only a few PIPs identified by theSCPPs. This might be explained by the fact that nurses pri-marily use nursing documentation rather than the patient’smedical record, possibly leading to not fully understanding theindications given for each drug.The nurses in the present study observed and presented physi-

cians with PIPs of differing potential clinical severity, frompotentially non-significant to potentially fatal. However, of theprescriptions altered or written by the physicians, only (8/47)17% were common to SCPPs and nurses, although physiciansaltered or wrote prescriptions for little more than a third of allPIPs presented to them by the nurses. It is possible that SCPPsand physicians view appropriateness of prescribing somewhat

178 ANN L. SOERENSEN ET AL.

Table 3.Types of potentially inappropriate prescriptions (PIPs) and assessments of severity in the intervention wards.

Types of PIP and assessment ofpotentially clinical consequences1

PIPs identifiedby clinical pharmacologists

in sample of patients reviewedby nursesn (%)2

PIPs identified bynurses after theintervention

Identical identificationof PIPs by clinical

pharmacologists and nursesn (%)

Total number ofchanges in patients’prescriptions made by

physicians following nurses’observations

Omission of indicationfor treatment3

32 (14) 20 4 (11) 9

Potentially non-significant 6 (19) 2 1Potentially significant 16 (50) 0 0Potentially serious 9 (28) 2 0Potentially fatal 1 (3) N/A 0 0

Drug dosage too low 11 (5) 8 1 (3) 5Potentially non-significant 0 0 0Potentially significant 9 (82) 1 0Potentially serious 2 (18) 0 0Potentially fatal 0 N/A 0 0

Drug dosage too high 35 (16) 13 6 (16) 4Potentially non-significant 4 (11) 1 1Potentially significant 14 (40) 3 2Potentially serious 15 (43) 2 0Potentially fatal 2 (6) N/A 0 0

Interaction between drugs 64 (29) 61 24 (63) 12Potentially non-significant 1 (2) 0 0Potentially significant 12 (19) 6 0Potentially serious 42 (66) 13 3Potentially fatal 9 (14) N/A 5 0

Interaction between drugand disease

14 (6) 13 1 (3) 6

Potentially non-significant 0 0 0Potentially significant 4 (29) 0 0Potentially serious 7 (50) 0 0Potentially fatal 3 (21) N/A 1 0

Duplicate drug 2 (<1) 0 0 0Potentially non-significant 0Potentially significant 1 (50)Potentially serious 1 (50)Potentially fatal 0 N/A N/A N/A

Inappropriate dosing interval 20 (9) 1 1 (3) 1Potentially non-significant 3 (15) 1 0Potentially significant 16 (80) 0 0Potentially serious 1 (5) 0 0Potentially fatal 0 N/A 0 0

Omission of a potentiallyuseful medication

9 (4) 1 0 0

Potentially non-significant 2 (22)Potentially significant 4 (44)Potentially serious 3 (33)Potentially fatal 0 N/A N/A N/A

Omission of therapeuticdrug monitoring

16 (7) 3 1 (3) 3

Potentially non-significant 0 0 0Potentially significant 8 (50) 1 1Potentially serious 7 (44) 0 0Potentially fatal 1 (6) N/A 0 0

Other 17 (8) 7 0 4Potentially non-significant 2 (18) 0 0Potentially significant 11 (65) 0 0Potentially serious 4 (24) 0 0Potentially fatal 0 N/A 0 0

1Only categories containing five or more PIPs are displayed.2The number of patients in the sample reviewed by nurses was 121.3The categories contain the levels of prevalence of PIPs identified by nurses that resulted in an alteration or writing of a prescription by the physician.

IMPROVING MEDICATION SAFETY IN PSYCHIATRY 179

differently, as one would have expected physicians to alter ahigher proportion of the PIPs common to SCPPs and nurses.This is in line with a review suggesting that acceptance rate byphysicians is a measure for perceived clinical relevance [36].Other reasons for reluctance to alter prescriptions might be lackof knowledge, a conservative unwillingness to adopt new guide-lines and procedures or fear of making a mistake that will harmthe patient. The suggested reasons remain speculative as theissue has not been subjected to much research. Physicians’acceptance rate to changing medications, for instance accordingto pharmacists’ suggestions, is moderate [36]. High acceptancerates by physicians to changing medications must in all likeli-hood reinforce the possible effect of input from other healthcareprofessionals. Therefore, interventions seeking to unveil theeffects of studies depending on physicians’ partaking need toreport on the magnitude of the collaboration between profes-sionals. It is not known to what extent physicians change pre-scriptions directly because of nurses’ observations.In medication error research, it is important to detect all errors

regardless of significance, as errors of minor significance indi-cate flaws in the medication process that might result in harmunder other circumstances [37]. When acknowledging that allerrors are important, it is interesting that nurses are capable ofidentifying not only PIPs with the potential for harm but alsoPIPs considered to be of minor importance. Nurses’ achieve-ments in medication-related activities such as evaluating pre-scriptions, ensuring patient adherence to medication andobserving effects and side effects of medication improve witheducational level [11]. For example, a nurse with a bachelor’sdegree is associated with more frequent information-seekingbehaviour on pharmacotherapy than one with a diploma [11]. InDenmark, all nurses who have graduated after 2008 hold bache-lor degrees [38]. However, it is not possible to investigatewhether this played a role in the results of the present study.Research indicates that nurses’ participation in courses seekingto improve medication competencies improve their skills inpharmacovigilance [22] along with their self-perceived ability toundertake pharmacovigilant activities [39]. Future researchshould investigate the type of pharmacological training nursesshould receive during their education and the skills and compe-tencies that can be reasonably expected from a qualified nurseto improve medication safety for psychiatric patients.

Strengths and limitations.The significant strengths of this study were the training ofnurses in identifying and documenting observations related toPIP in a consistent and systematic way, the pursued validationof the nurses’ observations by comparing with medicationreviews by highly skilled SCPPs and documenting the extent towhich physicians adopt nurses’ observations. The use of a‘DID’ approach reduces the potential bias in the before-and-afterdesign, heightens the internal validity and thus increases credi-bility. However, the study also has some limitations. Firstly,generalization must be made with caution as the study onlyinvolved selected wards and thus represents an incomplete pic-ture of prescribing patterns. Also, the nurses failed to review

12% of the eligible patients in the intervention period, whichmight introduce selection bias. There were, however, no obviousdifferences in age, gender, primary psychiatric disorder, comor-bidities, alcohol or substance abuse, or medication characteris-tics between the missing patients and the patients included in theanalysis. Secondly, it was not feasible to practice blinding of theSCPPs or randomization in the study. Blinding was not possibledue to the nature of the intervention and because randomizationon an individual level would have increased the risk of educa-tional bias between nurses working in the same bed unit.Thirdly, patient outcome was not evaluated. Fourthly, intrarater-and interrater reliability was not assessed for the data collectors.Additionally, the choice of SCPPs’ medication reviews as thegold standard is somewhat problematic as it leads to questionsconcerning the considerable proportion of PIPs identified by thenurses and responded to by the physicians, which were not sug-gested as PIPs by the SCPPs. However, this perspective wasbeyond the scope of the present study.

Conclusion

In conclusion, pharmacologically trained psychiatric nurses didnot yield a statistically significant reduction in PIPs in a psychi-atric hospital setting. However, the study and the interventionwere relatively small, and the results ought to spur questionsabout other ways nurses might affect medication safety andmedication appropriateness. PIPs were frequent throughout thestudy in both the intervention and control group. The leadingcategory of PIPs identified by nurses and SCPPs was ‘interac-tion between drugs’ with a small overlap. Additionally, theresults demonstrated that operationalizing nurses’ observationsto achieve any useful improvements in medication safety relieson the collaboration and actions of physicians. More researchneeds to be focused on the effect of registered nurses’ pharma-cological knowledge on adverse patient outcome in psychiatry,pharmacological requirements of nurses, as well as the medica-tion-related collaboration between nurses and physicians toinform future efforts aimed at improving the quality of prescrib-ing in psychiatry.

AcknowledgementThe authors thank Morten Berg for statistical advice and

comments on the manuscript.

Competing InterestsNone declared.

References

1 Hill-Taylor B, Walsh KA, Stewart S, Hayden J, Byrne S, SketrisIS. Effectiveness of the STOPP/START (screening tool of olderpersons’ potentially inappropriate prescriptions/screening tool toalert doctors to the right treatment) criteria: systematic review andmeta-analysis of randomized controlled studies. J Clin Pharm Ther2016;41:158–69.

2 O’Conner M, Gallagher P, O’Mahony D. Inappropriate prescribing:criteria, detection, and prevention. Drugs Aging 2012;29:437–53.

3 Lang PO, Vogt-Ferrier N, Hasso Y, Le Saint L, Dram�e M, ZekryD et al. Interdisciplinary geriatric and psychiatric care reduces

180 ANN L. SOERENSEN ET AL.

potentially inappropriate prescribing in the hospital: interventionalstudy in 150 acutely ill elderly patients with mental and somaticcomorbid conditions. J Am Med Dir Assoc 2012;13:406.e1–7.

4 Hefner G, Stieffenhofer V, Gabriel S, Palmer G, M€uller KM,R€oschke J et al. Side effects related to potentially inappropriatemedications in elderly psychiatric patients under everyday pharma-cotherapy. Eur J Clin Pharmacol 2014;71:165–72.

5 Lang PO, Hasso Y, Dram�e M, Vogt-Ferrier N, Prudent M, Gold Get al. Potentially inappropriate prescribing including under-useamongst older patients with cognitive or psychiatric co-morbidities.Age Ageing 2010;39:373–81.

6 Rongen S, Kramers C, O’Mahony D, Feuth TB, Olde Rikkert MG,Ahmed AI et al. Potentially inappropriate prescribing in olderpatients admitted to psychiatric hospital. Int J Geriatr Psychiatry2016;31:137–45.

7 Soerensen AL, Nielsen LP, Poulsen BK, Lisby M, Mainz J. Poten-tially inappropriate prescriptions in patients admitted to a psychi-atric hospital. Nord J Psychiatry 2016;70:365–73.

8 Gabe ME, Davies GA, Murphy F, Davie M, Johnstone L, JordanS. Adverse drug reactions: treatment burdens and nurse-led medi-cation monitoring. J Nurs Manag 2011;19:377–92.

9 Ndosi ME, Newell R. Nurses’ knowledge of pharmacology behinddrugs they commonly administer. J Clin Nurs 2009;18:570–80.

10 Haw C, Stubbs J, Dickens G. Medicines management: an interviewstudy of nurses at a secure psychiatric hospital. J Adv Nurs2015;71:281–94.

11 Dilles T, Vander Stichele R, Van Rompaey B, Van Bortel L, Else-viers M. Nurses’ practices in pharmacotherapy and their associa-tion with educational level. J Adv Nurs 2010;66:1072–9.

12 Hughes RG, ed. Medication administration safety. In: Patient safetyand quality: an evidence-based handbook for nurses. Agency forHealthcare Research and Quality, Rockville, MD, 2008;397–457.

13 Liberman RP, Hilty DM, Drake RE, Tsang HW. Requirements formultidisciplinary teamwork in psychiatric rehabilitation. PsychiatrServ 2001;52:1331–42.

14 Fagin L, Garelick A. The doctor–nurse relationship. Adv PsychiatrTreat 2004;10:277–86.

15 Marland GR, Sharkey V. Depot neuroleptics, schizophrenia andthe role of the nurse: is practice evidence based? A review of theliterature. J Adv Nurs 1999;30:1255–62.

16 Rogers AE, Dean GE, Hwang WT, Scott LD. Role of registerednurses in error prevention, discovery, and correction. Qual SafHealth Care 2008;17:117–21.

17 Rothschild JM, Landrigan CP, Cronin JW, Kaushal R, LockleySW, Burdick E et al. The Critical Care Safety Study: the incidenceand nature of adverse events and serious medical errors in inten-sive care. Crit Care Med 2005;33:1694–700.

18 Dykes PC, Rothschild JM, Hurley AC. Medical errors recoveredby critical care nurses. J Nurs Adm 2010;40:241–6.

19 Henneman EA, Gawlinski A, Blank FS, Henneman PL, Jordan D,McKenzie JB. Strategies used by critical care nurses to identify,interrupt, and correct medical errors. Am J Crit Care 2010;19:500–9.

20 Choo J, Hutchinson A, Bucknall T. Nurses’ role in medicationsafety. J Nurs Manag 2010;18:853–61.

21 Ulfvarson J, Mejyr S, Bergman U. Nurses are increasinglyinvolved in pharmacovigilance in Sweden. PharmacoepidemiolDrug Saf 2007;16:532–7.

22 B€ackstr€om M, Mj€orndal T, Dahlqvist R. Spontaneous reporting ofadverse drug reactions by nurses. Pharmacoepidemiol Drug Saf2002;11:647–50.

23 Ekman E, Petersson G, T�agerud S, B€ackstr€om M. Awarenessamong nurses about reporting of adverse drug reactions in Sweden.Drug Healthc Patient Saf 2012;4:61–6.

24 Jordan S, Gabe-Walters ME, Watkins A, Humphreys I, Newson L,Snelgrove Set al. Nurse-led medicines’ monitoring for patients

with dementia in care homes: a pragmatic cohort stepped wedgecluster randomized trial. PLoS One 2015;10:1–27.

25 Nakanishi M, Koyama A, Ito H, Kurita H, Higuchi T. Nurses’ col-laboration with physicians in managing medication improvespatient outcome in acute psychiatric care. Psychiatry Clin Neurosci2006;60:196–203.

26 Gray R, Wykes T, Edmonds M, Leese M, Gournay K. Effect of amedication management training package for nurses on clinicaloutcomes for patients with schizophrenia: a cluster randomizedcontrolled trial. Br J Psychiatry 2004;185:157–62.

27 Gallagher P, Barry P, O’Mahony D. Inappropriate prescribing inthe elderly. J Clin Pharm Ther 2007;32:113–21.

28 Hiemke C, Baumann P, Bergemann N, Canca A, Dietmaier O,Egberts K et al. AGNP consensus guidelines for therapeutic drugmonitoring in psychiatry: update 2011. Pharmacopsychiatry2011;44:195–235.

29 Gertler P, Martinez S, Premand P, Rawlings LB, VermeerschCMJ. Chapter 6. Impact Evaluation in Practice. The World BankPublications, Washington DC, 2011.

30 Hill-Taylor B, Sketris I, Hayden J, Byrne S, O’Sullivan D, ChristieR. Application of the STOPP/START criteria: a systematic reviewof the prevalence of potentially inappropriate prescribing in olderadults, and evidence of clinical, humanistic and economic impact.J Clin Pharm Ther 2013;38:360–72.

31 Guo JJ, Wu J, Kelton CML, Jing Y, Fan H, Keck PE et al. Expo-sure to potentially dangerous drug–drug interactions involvingantipsychotics. Psychiatr Serv 2012;63:1080–8.

32 Nomura K, Nakao M, Sato M, Yano E. Regular prescriptions forbenzodiazepines: a cross-sectional study of outpatients at a univer-sity hospital. Intern Med 2006;45:1279–83.

33 American Geriatrics Society. Beers Criteria Update Expert Panel.American Geriatrics Society updated Beers Criteria for potentiallyinappropriate medication use in older adults. J Am Geriatr Soc2012;2012:616–31.

34 Gallagher P, Ryan C, Byrne S, Kennedy J, O’Mahony D. STOPP(screening tool of older person’s prescriptions) and START(screening tool to alert doctors to right treatment). Consensus vali-dation. Int J Clin Pharmacol Ther 2008;46:72–83.

35 Bergqvist M, Ulfvarson J, Andersen Karlsson E, von Bahr C. Anurse-led intervention for identification of drug-related problems.Eur J Clin Pharmacol 2008;64:451–6.

36 Graabæk T, Kjeldsen LJ. Medication reviews by clinical pharma-cists at hospitals lead to improved patient outcomes: a systematicreview. Basic Clin Pharmacol Toxicol 2013;112:359–73.

37 Aronson JK. Medication errors: what they are, how they happen,and how to avoid them. QJM 2009;102:513–21.

38 Bekendtgørelse om uddannelsen til professionsbachelor i sygepleje.BEK nr 29 af 24/01/2008 [Internet]. https://www.retsinformation.dk/forms/R0710.aspx?id=114493 (last accessed on 21 Sep. 2017).

39 Johansson-Pajala RM, Martin L, Fastbom J, Jors€ater Blomgren K.Nurses’ self-reported medication competence in relation to theirpharmacovigilant activities in clinical practice. J Eval Clin Pract2015;21:145–52.

Supporting Information

Additional Supporting Information may be found online inthe supporting information tab for this article:Figure S1. The process of medication reviews by clinical

pharmacologistsFigure S2. The difference-in-difference approachTable S1. Categories and definitions of potentially inappro-

priate prescriptions (PIP).

IMPROVING MEDICATION SAFETY IN PSYCHIATRY 181

Copyright of Basic & Clinical Pharmacology & Toxicology is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listservwithout the copyright holder's express written permission. However, users may print,download, or email articles for individual use.