Effect and associated factors of a clinical pharmacy model in the incidence of medication errors in the hospital Pablo Tobón Uribe eacpharmodel study: stepped wedge randomized controlled Trial (NCT03338725)

Background The World Health Organization considers medication errors to be an issue that requires attention at all levels of care, to reduce the severe and preventable harm related to drug therapy. Different standards for clinical pharmaceutical practices have been proposed by various organizations across the world, where the pharmacist, as part of the multidisciplinary health team, can help improve patient safety. Objective To assess the impact of the introduction of a clinical pharmacy practice model on medication error in patients of a university hospital. Setting The study was conducted in a tertiary care hospital, Medellín, Colombia. Methods A randomized, controlled cluster-wedge staggered trial with a duration of 14 months was conducted to compare the clinical pharmacy practice model with the usual care process in the hospital. Five hospital health care units were included, which were initially assigned to the control group, and after an observation period of 2 months, they were randomly assigned to the intervention group. The trial protocol was registered in ClinicalTrials.gov (identifier NCT03338725). Main outcome measure The incidence of medication errors in hospitalized patients was the main outcome measure. Results The incidence of medication error was 13.3% and 22.8% for the intervention group and control group, respectively. The probability of presenting a medication error was 48% lower when the patient was in the intervention group (RR 0.52; 95% CI: 0.34–0.79). The probability of presenting a medication error over time was 44% lower in the intervention group (p = 0.0005); meanwhile, the resolution of a medication error over time was 70% higher in the intervention group (p = 0. 0029). Conclusion The clinical pharmacy practice model, made up of strategies focused on reducing medication errors, significantly reduces medication errors in patients during hospitalization compared with usual practice. This work assessed the effect of a clinical pharmacy model on the incidence of medication errors and demonstrated its effectiveness in reducing these errors in hospitalized patients. Trial registration ClinicalTrials.gov, NCT03338725. Registered on 9 November 2017. First patient randomized on February 2, 2018.


Impacts on Practice
• Clinical pharmacy models could be considered a complex intervention with a high number of subprocesses and players that can intervene in the outcomes, increasing intergroup variability, these programs must be evaluated with all their subprocesses to determine their effectiveness.

3
• A clinical pharmacy practice model with the active role of pharmacists in the evaluation and monitoring of the medication-use process is a strategy to prevent medication-related problems that can result in patient harm, and if the medication error occurs, the time for its correction is less.

Background
In the hospital setting, approximately 4% of inpatients are at risk or harm caused by health care errors or adverse events, and 14% of fatal incidents are caused by such errors or events, which result in between 44,000 and 98,000 deaths each [1,2]. Moreover, approximately 37% of adverse events that occur during the health care process are related to medication errors (ME). A medication error is defined as "any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is under the control of the health care professional, patient, or consumer." These errors may be associated with any phase of the drug delivery process, from prescription to drug administration [3]. ME are considered preventable and generate costs for the health system. According to a report by the World Health Organization (WHO), in the United States, ME cause at least one daily death and damage to approximately 1.3 million people a year, generating an associated global cost of US $42 billion a year that is, almost 1% of global health spending [4][5][6][7].
Knowing the importance of ME and their outcomes has led to patient safety being a global concern and on the agendas of health systems around the world and of the main international organizations that promote the health of patients. The population encourages the implementation of strategies focused on mitigating and preventing these errors. The WHO in October 2004 launched the World Alliance for Patient Safety, calling on hospital and government institutions to carry out a series of actions in favor of the safety of hospitalized patients, again in 2017 it launched a global initiative to reduce the risk of medication-related errors by half in 5 years [8].
Different strategies have been proposed as a global initiative to reduce ME, including the implementation of clinical pharmacy services. Direct patient care by clinical pharmacists within multidisciplinary teams is recognized as one of the best pharmacy practice models because it minimizes ME, care costs, morbidity and mortality rates, and time spent in the hospital and improves pharmacotherapy results [9,10].
In mid-2016, a tertiary university hospital located in Medellín (Colombia), systematized a clinical pharmacy practice model (CPPM) [11]. This model is actively applied in the hospital; however, the effect on patient safety is unknown. The study aim of the EACPharModel was to assess the impact of the introduction of a clinical pharmacy practice model (CPPM) on ME in patients of a university hospital.

Methods
This study was a 14-month, randomized, controlled, prospective, single-center, stepped wedge clinical trial performed to assess the effect of a clinical practice pharmacy model (CPPM) on the incidence of ME. The study began in February 2018 and ended in March 2020. Detailed methods of the EACPHARMODEL trial have been published previously in the protocol paper [12], and the trial has been registered at ClinicalTrials.gov (identifier NCT03338725).

Design
Stepped wedge randomized trial designs involve sequential roll-out of an intervention to participants over several periods. At the end of the study, all clusters had received the intervention, and the order in which participants received the intervention was determined at random. The clusters were hospital units.
Each hospital ward began with a control period (Baseline S0) and changed to an intervention period following randomization. The study design consisted of six consecutive 60-day periods ( Fig. 1). At the end of each 60-day period, another hospital ward changed to an intervention period until all units were in the intervention period during the final 60-day period. The randomization carried out by the trial defined the point at which each hospital unit changed from the control to the intervention period. There was a pharmacist who was responsible for the recruitment of potential patients to the hospitalization units. A cluster was defined as a hospital ward, 2 months were considered a step, and the inclusion of participants within the cluster was dynamic.

Setting and study population
The study was conducted with hospitalized patients prescribed five or more drugs attended in HPTU, a tertiary care university institution. The hospital has 452 beds; however, for the duration of this study, only five medical hospitalization units were used.
The recruited patients were evaluated for compliance with the following inclusion criteria: age 18 years or older, hospitalization in the HPTU for a minimum of 24 h, and inclusion of at least five drugs in their pharmacological therapy.
Each cluster was assigned to the intervention group (IG) or the control group (CG) through a computer-generated randomization sequence, using Microsoft Excel® (version 2010, Microsoft® Corporation, Redmond Washington).

Data collection and blinding
Data were collected from February 2018 to January 2019. The study had a total length of 14 months, the recruitment period was 12 months, and patients were evaluated for 2 months, starting from the date of their recruitment. Once the enrolment period ended, the final 2 months were only used to evaluate the latest patients. The data obtained in this study were registered in an electronic database. Data regarding medication history, interviews with the pharmacist, health status, plans of action, and data related to the primary outcome were registered.
A specially trained clinical pharmacist and an internist were the staff that reviewed each patient's clinical history at baseline (t0) and 2 (t1), 4 (t2), 6 (t3), 8 (t4), and 10 months post-baseline (t5). The staff were an external evaluation group who evaluated if a ME was presented and whether it was resolved. This evaluation was performed at the time of discharge from the hospital. Additionally, the evaluators were blinded to patient inclusion in the clinical pharmacy model.

Sample size
An overall incidence of ME of 15% was taken as a reference value. The sample size was calculated accepting an alpha error of 0.05, a beta error of 0.84, five clusters, six steps, two months for each step, 24 participants per cluster, and a coefficient of variation (k) of 0.15. The sample size Fig. 1 Design of the EACPhar-Model study. This figure shows the design and timeframe of the EACPharModel study. The outcomes to be measured during each evaluation stage are the incidence of medication error, time to error, and time to error recovery needed to detect a difference equal to or greater than 10% was 720 patients [13].

Outcomes
The study outcomes were the change in the incidence of ME upon application of the CPPM; the identification, quantification, and classification of ME; estimation of the probability that a subject remains without ME; and measurement of the time until ME were resolved.

Intervention group
For patients who were under the CPPM of the hospital, pharmacists performed the following activities: (1) Participation in medical rounds at the time of drug prescribing, reviewing and discussing doses, interactions, duplications, allergies.; (2) patient review, which consists of reviewing patients' medical records and providing verbal or written follow-up concerning the clinical condition (involves repeated monitoring, with review of all medication orders and documentation of pharmacist's interventions, paraclinics review, efficacy of therapy); (3) identification and management of Adverse Drug Reactions (ADR), which consists of detecting potential ADRs, providing and documenting appropriate follow-up until the ADR has resolved, and reporting ADRs to the national pharmacovigilance program; (4) pharmacological counselling to patients and/or relatives during the hospital stay or after discharge, which involves providing information and education on the proper use of medications, their conservation and individualized schedule for taking them. The patients to be educated are those who have started anticoagulant treatment, all patients with kidney or liver transplantation (5) pharmacotherapy review, in which pharmacists conducted an appropriateness review of medical orders and determine, for example, correct dosage, frequency, route of administration, administration, length of therapy, indicated drugs, contraindicated drugs, lack of treatment, drug-drug interactions, drug-food interactions, therapeutic duplicity, and allergies. DEPICT 2 was used to specify the pharmacist's intervention [14].
Every day in the morning, the pharmacist was provided a list of patients that they must evaluate as specified in activities 2, 3, 5, and the others when necessary or when they had patients who required it. The model can be seen in Fig. 2.

Control group
The usual patient care process began with a medical evaluation and the respective formulation of pharmacotherapy. Later, the pharmacy technicians verified with a spreadsheet the quantities to be dispensed, allergy detection, and therapeutic duplicities, and finally, the medication was dispensed (Fig. 3).

Statistical analysis
The statistical analyses of the full-analysis set followed the intention-to-treat principle. This dataset included all subjects in the assigned cluster who met all the inclusion criteria. Baseline and demographic characteristics were analyzed descriptively (number of valid cases, mean, standard deviation, median, interquartile range, and proportions for qualitative variables). A mixed model evaluated the primary outcome with treatment group and time as fixed effects and clustering structure as a random effect. A significance level set to alpha = 5% (two-sided) was used to compare proportions. Comparisons were conducted by using the chi-square test (or Fisher's exact test when appropriate) for categorical variables and the Mann-Whitney U test for continuous variables. Relative Risk and 95% confidence intervals (CIs) were estimated as well. Multivariable analyses were performed to explain the association of multiple variables with the factors significantly related to the primary outcome: the sociodemographic and clinical variables assessed were sex, age, social security system, scholarship, weight, height, allergies, caregiver, diagnosis of admission, hospitalization 6 months prior, number of services, previous stay in the intensive care unit, ADR, isolated patient, hospital stay, and number of medications.

Ethics approval and consent to participate
The trial was carried out in compliance with the protocol and the declaration of Helsinki, following the International Conference on Harmonization. The protocol was approved by the Institutional Review Board of the HPTU (2017.050/2017). The study characteristics are such that the data were collected from clinical records, and the proposed intervention does not entail a risk of causing biological, psychological, or social damage, as all the clusters, in the end, received the intervention. Informed consent was obtained from all study participants.

Results
During the estimated period for the inclusion of patients in the study, 765 patients were identified. After verification of the inclusion criteria, of the 765 eligible patients, 45 were excluded because they had a hospital stay of fewer than 24 h. The initial population under study of the EACPHAR-MODEL was made up of 720 patients, which were divided into five different clusters and six steps with 24 participants for each cluster-period of time; all the patients finished the study (Fig. 1). The patient population presented a mean age of 60.9 (SD = 20.4) years and a male prevalence of 50.5%. Table 1 shows the sociodemographic characteristics of the population. At the beginning of the study, no statistically significant differences were found between the CG and the IG. During follow-up, there were no losses and no statistically significant differences at the sociodemographic level (p > 0.05).

Incidence of medication errors
A total of 130 ME were detected (82: CG; 48: IG) The errors identified in the intervention group were: wrong dose prescription (31), wrong time (7), non-prescribed medication (4), therapeutic duplication (4), wrong medication (1) and inappropriate route of administration (1). Medication errors in the control group were, wrong dose prescription (44), incorrect drug (15), non-prescribed medication (10), therapeutic duplication (6), wrong time (4), incorrect route of administration (3), and the difference was statistically significant. The incidence of ME was 13.3% for IG and 22.8 for CG. The estimated RR was 0.52 (IC 95% 0.34-0.79), determining that the probability of occurrence of a ME was 48% lower when the CPPM followed up the patient. In addition to the intervention, a significant association was also found, and a protective factor was a previous stay in the ICU (RR: 0.44 IC 95% 0.24-0.74). Conversely, the increase in the ). The most frequent ME among groups were those classified as C, that is, those that reached the patient, but did not generate damage with 31% for the CG and 26% for the IG, ME classified as D or B were only presented in the CG (Table 2).

Medication errors and the measurement of time
The probability of presenting a ME over time was 44% lower in the IG (p = 0.0005); meanwhile, the probability of presenting the resolution of a ME over time was 70% higher in the IG (p = 0. 0029). The Cox regression results showed that the group variable was statistically significant (p = 0.0005). Medication quantity (p = 0.0004) and previous stay in the ICU (p = 0.0018) and, consequently, instantaneous risk (HR:0.56, 95% CI = 0.39-0.81), were lower in the IG than in the CG. For ME recovery, Cox regression showed that only the study group variable was statistically significant (p = 0.0131); consequently, instantaneous risk (HR:1.70, 95% CI = 1.12-2.61) was higher in the IG than in the CG (Fig. 4).

Differences based on in-hospital stay
The hospital stay data did not follow a normal distribution. Therefore, the Mann-Whitney U test was performed, and no statistically significant differences were found between the median length of stay (p = 0.5620).

Discussion
This work demonstrated the ability of a clinical pharmacy model to reduce ME in hospitalized patients, generating high-quality evidence on this topic. Another stepped-wedge study measured the impact of collaborative pharmaceutical care on preventable ME rates, but their results are not yet available [15]. The use of pharmacists to reduce ME and adverse drug reactions is well established and is considered a standard practice in most countries. A randomized, clinical, multicenter study showed that a multifaceted pharmacist's intervention based on medication review, patient interview, and follow-up in patients receiving multiple medications can reduce the number of readmissions and emergency department visits [16]. Khalili et al. reported that clinical pharmacists detected 112 ME (0.13 errors/patient). Dosage, choice, use, and drug interactions were the main causes of error in medication processes, respectively [17]. Additionally, Kuo et al. identified 924 ME, of which almost half were corrected before the medications reached the patients [18].
The results of a meta-analysis, in which the effect of the intervention of the pharmacist in intensive care units (ICU) on the incidence of ME was evaluated, showed no significant beneficial effect of the intervention on general ME; however, the global analysis supported the role of pharmacists in reducing preventable ADEs and prescribing errors [19]. These results contrast with the EACPharmodel; however, the differences in the study population must be borne in mind. Basheti et al. evaluated the pharmacist's impact on the reduction of ME and adherence to treatment in outpatients with chronic diseases and showed a significant decrease in the number of ME with a mean of (1.23 ± 1.19 p < 0.001), in the IG versus (0.29 ± 1.24, p = 0.114) CG [20].
Previous stay in intensive care unit was a protective factor for ME incidence, although it is known that in ICUs, patients are most vulnerable to being exposed to ME because it brings together high-risk patients and interventions in a complex environment [21][22][23][24][25]; nevertheless, in the HPTU, work has been done on the standardization of processes and medication prescription and administration through computerized physician orders.
Polypharmacy is widespread and associated with medication-related harms, including ADR and ME [26]. Our results are consistent and reported the increase in the number of medications ordered as a risk factor (RR = 1.09), raising the probability of presenting a ME. Medication errors lead to an increase in hospitalization duration [27]; nevertheless, in the present study, the length of stay was similar in both groups, although the incidence of ME was higher in the CG. In general, ME generates an increase in hospital stay [27][28][29][30], so it would be expected that the hospital stay would be longer in the CG; however, in EACPharmodel study, no statistically significant differences were found in the length of stay between the two groups (p = 0.5620). This finding is similar to those of other studies in the hospitalization setting, in which no differences were found in the length of stay [33][34][35]. A possible explanation could be that a high percentage of errors do not reach the patient or do not cause harm (Type A error). In this sense, only four type E errors were recorded in the CG in the present study. For their part, the most frequent ME among the groups were classified as type C errors (those that reached the patient but did not cause harm: 36.2%), followed by type B errors (did not reach the patient: 59.2%). These results are similar to those reported by a study carried out in Spain, in which the incidence of ME in the processes of drug use was described, showing that type B errors were the most frequent (84.5%), followed by type C errors (14.5%) [34]. Similarly, in a pharmacovigilance study carried out in Colombian hospitals, 9062 ME were found, of which the most frequent were type A (48.06%), followed by type B (45.69%) and type C (5.95%) [35]. As the ME detection mechanism was the evaluation of medical records, it was not possible to find classification A errors, "circumstances or incidents with the capacity to cause error," since these events are not recorded in the medical record or the management software of clinical risks.
As the information collected will be based on the review of medical records, some sub-registries may be present that will affect ME detection. However, HPTU is an institution accredited by the International Joint Commission, which has made it possible to adequately standardize the process, including the quality of medical records, and this bias is therefore expected to be controlled.

Conclusion
The CPPM is made up of strategies focused on reducing ME, such as (a) pharmacovigilance; (b) medication review; (c) the trigger tools method; (d) drug reconciliation, and (d) health education, which significantly reduces the incidence of ME in patients during hospitalization. Regarding ME characterization, those classified as C, D, and E are more frequent in the CG than in the intervention group. The probability over time of presenting a ME is higher in the CG than the intervention group. Moreover, the probability over time of ME resolution is higher in the IC than in the CG.
Funding The Pharmaceutical Promotion and Prevention Group received financial support from the Committee for Development Research (CODI) and sustainability program (2018-2019), Universidad de Antioquia.

Conflicts of interest
The authors declare no conflict of interest.