Priorities in the Prevention Strategies for Medication Error Using the Analytical Hierarchy Process Method

As medication error is inherently “preventable”, we should try to minimize errors to improve patient safety and quality of care. The aim of this study was to prioritize strategies to prevent medication errors using the analytic hierarchy process (AHP) method. The hierarchy structure consisted of three stages: goal of the decision, decision criteria, and alternatives. Ten experts of patient safety research or clinical pharmacology compared each pair of criteria and alternatives and assigned a nine-point numerical scale. We used the eigenvector method to aggregate the pairwise comparisons obtained from experts and to estimate the weights of each criterion and alternative. Among the decision criteria, system improvement in reporting was the most preferred criterion, followed by cultural improvement and system improvement in the counterplan. The preferred alternative was a counterplan by healthcare institutions, followed by a change from a blame culture to safety culture and the building of a reporting system. A sensitivity analysis indicated that priorities were generally robust in the methods used for calculating the integrated matrices. We have suggested the priority of preventive strategies against medication errors using the AHP method. The prioritization of preventive strategies could help policymakers understand current needs and therefore develop evidence-based policies on patient safety.


Introduction
According to the definition of the National Coordination Council for Medication Error Reporting and Prevention, a medication error is defined as "a preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the health care professional, patient, or consumer" [1]. Some medication errors result in adverse drug events (ADEs), called "preventable ADEs". Some medication errors do not result in adverse events but still have the potential to cause adverse events; these are called "potential ADEs" [2].
The Institute of Medicine reported that more than 1.5 million preventable ADEs occurred annually in the United States [3]. In a previous study, there were 7.3 preventable events (including both preventable ADEs and potential ADEs) per 100 hospital admissions, and the incidence of potential ADEs was three times that of preventable ADEs [4]. In Korea, of all drug-related damages reported to the Korea Consumer Agency increased between 2010 and 2014, 9.2% of them resulted from medication error [5]. Medication errors result in a considerable economic burden, which was estimated to be GBP 98.5 million per year in England [6].
Because medication errors are "preventable", the errors can be minimized by effort, leading to improvements in safety and quality of care [3,7]. However, medication errors cannot be prevented merely by an individual's effort. In fact, a majority of medication errors result from fundamental causes such as human factors, defects in the system, and inadequate healthcare products [8]. Owing to their diverse causes, there are numerous

Materials and Methods
To prioritize strategies to prevent or reduce medication errors, we used an AHP method that consisted of three steps: decomposition of the structure, comparison of judgements, and hierarchical composition of priorities [15]. This study was approved by the Institutional Review Board of Pusan National University (PNU IRB/2020_58_HR).

Decomposition of the Structure
In the hierarchical structuring step, we first investigated previous literature and the systems in various countries (i.e., Australia, Canada, New Zealand, United Kingdom, United States of America) to prevent medication errors . Three independent reviewers categorized the elements in a hierarchical manner and reached a consensus by discussion. The fourth reviewer independently assessed the structure to ensure the validity of the structure. The hierarchy structure consists of three stages. The first stage of the structure was the goal of the decision, and the second stage was the decision criteria. Finally, the third stage consisted of alternatives within each criterion [41]. The concept of medication error used in the hierarchy structure was based on previous studies on the definitions and types of medication errors in various countries [42,43].

Comparison of Judgements
Ten experts in South Korea, selected as respondents, compared each pair of criteria and alternatives. Because both clinical and institutional perspectives were needed to set effective strategies for preventing medication errors, we recruited five experts who had experience in patient safety research and medication errors, and five experts who had experience in clinical pharmacotherapy. The characteristics of the experts are summarized in Table 1. Each respondent assigned a nine-point numerical scale (1 = equal importance, 3 = moderate importance of one over another, 5 = essential or strong importance, 7 = very strong importance, 9 = extreme importance, and 2, 4, 6, and 8 for intermediate values between the two adjacent judgements) to compare each pair of criteria and alternatives [44,45]. Each respondent evaluated each pair of criteria first and then made a comparative assessment among alternatives within each criterion [46].

Hierarchical Composition of Priorities
Among the various methods for estimating the weights from the results of the pairwise comparison matrix, the eigenvector method was used because this was the only way to estimate the consistency of the judgement of respondents in a pairwise comparison [44]. We estimated the normalized weights of each criterion and alternative using geometric mean calculated in every row of the pairwise comparison matrix. To examine the consistency of each expert's judgement, the consistency ratio (CR) was calculated by dividing the consistency index (CI) by the random index (RI). The CI represents the average of the remaining eigenvalues from dividing λ max (the maximum eigenvalues of the pairwise comparison matrix) by the size of the pairwise comparison matrix minus one. The RI was defined according to the size of the pairwise comparison matrix (n) (i.e., RI = 0 for n = 1, 2; RI = 0.58 for n = 3; RI = 0.90 for n = 4; RI = 0.58 for n = 5) [47,48]. A CR less than 0.2 was regarded as the permissible level of consistency in experts' judgements [49,50]. To aggregate each pairwise comparison from the different experts, we applied three approaches [51]. Firstly, we aggregated all the matrices of the pairwise comparisons by geometric mean and then calculated the comprehensive weight using the eigenvalue method [52]. This method, which is the most widely accepted in the AHP method, was used for base-case analysis. Secondly, we calculated the weight of each individual pairwise comparison by applying the eigenvalue method and then aggregated the weights using the geometric mean. Finally, we calculated the weight of each individual pairwise comparison using the eigenvalue method and then aggregated the weights using the arithmetic mean. The latter two methods were used for sensitivity analyses. The approach in the base-case analysis is primarily used when information or previous research on the topic is rare, and the other approaches are used to place an emphasis on the expertise of individual responders involved in decision making [53]. The Pearson correlation coefficient was calculated to assess the concordance between ranks derived from the base-case analysis and the ranks derived from the sensitivity analysis.
We calculated all the weights of alternatives by multiplying each weight of the criterion and the weights of alternative under the criterion. A criterion or alternative with a higher weight was regarded as having a higher priority. The data were analyzed using Microsoft Office Excel software (version 2010, Microsoft Corporation, Redmond, WA, USA).

Results
The final version of the hierarchy structure consisted of five criteria: cultural improvement, system improvement in reporting, system improvement in cause analysis, system improvement in counterplan, and system improvement in assessment ( Figure 1). Two to five alternatives were suggested as possible strategies to meet the corresponding criteria. The definitions of each criterion and alternative are presented in Table 2.

Results
The final version of the hierarchy structure consisted of five criteria: cultural improvement, system improvement in reporting, system improvement in cause analysis, system improvement in counterplan, and system improvement in assessment ( Figure 1). Two to five alternatives were suggested as possible strategies to meet the corresponding criteria. The definitions of each criterion and alternative are presented in Table 2.
Pairwise comparisons of criteria and alternatives were conducted by ten healthcare professionals who were deeply experienced in healthcare policy or clinical service. All comparisons between criteria (Table 3) or alternatives (Tables S1-S5) were incorporated into the integrated matrices. Some comparisons with CR larger than 0.2 were excluded from the analyses.  To induce laypeople to participate actively in safe use by providing safe use information and running campaigns for spontaneous reporting of medication errors Change from a blame culture to a safety culture To regard medication errors as a systemic problem and work together to find solutions instead of blaming an individual healthcare professional Facilitating relevant research To encourage research on safety culture or safety policy   To prepare government-level countermeasures such as the establishment of alarm systems, providing guidance for the pharmaceutical industry, dissemination of information regarding the safe use of drugs, establishment of a reimbursement system for error reporting (e.g., incentives for good reporting and legal liability for insufficient reporting), and development of a system that help institutions exchange patients information during patient transfer (e.g., medication reconciliation service)

By pharmaceutical industry
To establish industry-level countermeasures such as making patient brochure, restraint of making similar looking products, and production of pediatric-specific dosage formulation By healthcare professionals To establish professional-level countermeasures such as developing an education program/materials, regular and mandatory education for professional knowledge, introducing courses related to patient safety (e.g., patient safety law and communication skill) in College of Medical, Nursing, and Pharmacy

By healthcare institutions
To establish industry-level countermeasures such as staff training, regular discussion on errors occurring in the institution, the establishment of computerized physician order entry, improvements to the workflow and work environment, and developing guidelines for providing patient with medication information

By patients
To establish patient-level countermeasures such as an education program on medication error and participation of patient/caregiver in patient safety committee System improvement in assessment Regularization of system assessment To regularly evaluate the system related to medication error and seek ways to improve the system Development and spread of guidelines for assessment To develop and disseminate the guidelines for assessment (e.g., design, criteria/indices, measurement, and analysis method) to acquire high-quality results Pairwise comparisons of criteria and alternatives were conducted by ten healthcare professionals who were deeply experienced in healthcare policy or clinical service. All comparisons between criteria (Table 3) or alternatives (Tables S1-S5) were incorporated into the integrated matrices. Some comparisons with CR larger than 0.2 were excluded from the analyses. Normalized weights of criteria and alternatives were calculated from the integrated matrices and are shown in Tables 3 and 4, respectively. The best preferred criterion was system improvement in reporting, and the best preferred alternative was counterplan by healthcare institutions (Table 5). We conducted sensitivity analyses to check the robustness of the priorities calculated from the AHP using various methods for calculating the integrated matrices. As mentioned in the methods section, there are two other methods to calculate weights: aggregating individual matrices using the geometric mean and aggregating individual matrices using the arithmetic mean. Of the two methods, the former (Pearson correlation coefficient = 0.974) resulted in priorities that corresponded more with the base-case analysis than the latter (Pearson correlation coefficient = 0.819) ( Table 5). Participation of laypeople 0.047 12 5 10 Development and spread of tools for cause analyses 0.044 10 12 11 Development and spread of guidelines for reporting 0.042 11 15 12 Counterplan by healthcare professionals 0.041 9 10 13 Regularization of system assessment 0.039 13 8 14 Facilitating relevant research 0.035 15 9 15 Integration of cause analyses and IT technology 0.031 † 17 17 16 Counterplan by regulatory government agency 0.031 † 14 14 17 Development and spread of guidelines for system assessment 0.031 † 16 13 18 Constitution of exclusive committee for cause analyses 0.023 18 18 19 Counterplan by patients 0.014 19 19 * The normalized weight of cultural improvement was larger than that of system improvement in counterplan. † The normalized weight of integration of cause analyses and IT technology was larger than that of counterplan by regulatory government agency. Additionally, the normalized weight of counterplan by regulatory government agency was larger than that of development and spread of guidelines for system assessment.

Discussion
In this study, we explored preventive strategies against medication errors and prioritized them using the AHP method. Among the decision criteria, system improvement in reporting, cultural improvement, and system improvement in counterplan comprise the main priorities with similar weights, whereas the system improvement in cause analyses and assessment showed lower weights. This might suggest that experts in patient safety and clinical pharmacotherapy feel the necessity of action plans rather than evaluation plans. Among the alternatives, counterplan by healthcare institutions, changing from a blame culture to a safety culture, and building of reporting systems had high weights. The results of the sensitivity analyses indicated that the priorities of decision criteria and alternatives were generally robust in aggregating approaches.
To our knowledge, this is the first study to provide evidence for priority setting in prevention strategies against medication errors using the AHP method. Although several studies have investigated strategies for the prevention of medication errors, most of them discussed the strategies in a descriptive way based on the literature review, and a few of them evaluated the priority of strategies using survey data. Matti et al. conducted a survey of staff at neonatal intensive care units in Australia and New Zealand to understand the prevention strategies utilized in clinical practice [54]. The survey found that smart infusion pumps, ward-based pharmacists, and the administration calculation test for nursing staff were the most frequently used strategies to prevent medication errors. Fortescue et al. examined the types of medication errors and strategies that might have prevented the potentially harmful errors in 1020 pediatric inpatients [55]. The study reported that the most effective interventions to prevent potentially harmful errors were ward-based pharmacists, improving communication between healthcare providers, and computerized physician order entry combined with clinical decision support system. Contrary to previous studies, our study assessed the priority of prevention strategies at a national level rather than an institutional level. However, our study demonstrated that counterplan by healthcare institutions was the most important alternative, highlighting the need for developing evidence-based strategies for healthcare institutions.
AHP is one of the most popular methods for solving complex multi-criteria decisionmaking problems because of its intuitiveness and ability to analyze both qualitative and quantitative criteria [56]. Therefore, many studies have utilized AHP to suggest evidence for decision making in healthcare problems. Hsieh et al. studied the error factors in the emergency department using a fuzzy technique for order preference by similarity to an ideal solution (TOPSIS) and AHP [57]. Fuzzy TOPSIS was utilized to rank the important error factors, and AHP was utilized to derive the criteria weights in TOPSIS. This study suggested the most important error factors, such as decision errors and crew resource management. Singh et al. explored the optimal management strategy in patients with sore throat using AHP [58]. The results indicated that the priorities among the five management strategies (i.e., no test and no treatment, rapid strep, culture, rapid strep and culture, and empiric treatment) depend on the likelihood of group A streptococcal infection and clinical judgments on the relative importance of the decision criteria. Reddy et al. introduced AHP to prioritize public health topics for guidance development [16]. The pilot test suggested that AHP could be a useful approach for prioritizing topics for guidance by structuring the decision process in a transparent way.
Decision-making on health policy often involves qualitative assessment that is inherently subject to ambiguity and implicit judgement. Multiple criteria decision analysis (MCDA) helps decision makers set priorities between the alternatives and make optimal decisions by providing a structured and explicit approach [14]. Therefore, MCDA has been applied to decision-making regarding healthcare problems at various levels, such as national, international, regional, or hospital level [59]. Considering that healthcare problems have a large impact on society, from quality of life in individuals to socioeconomic burden related to a certain disease, decisions should be made in a reasonable way. In that sense, MCDA such as AHP can be a useful tool that can help policy makers in the healthcare sector pursue evidence-based policymaking. Therefore, the number of studies using MCDA in the healthcare research field has steadily increased between 1990 and 2017; AHP (41%) was the most widely used method among the MCDA methods, followed by fuzzy logic (21%) and EVIDEM (12%) [60]. In this study, we used AHP because it enables structured decision making based on intuitive judgement of experts, reflecting their expertise and experience with patient safety. In addition, AHP has an advantage in its simplicity of application compared to other methodologies that use mathematical models. It uses a relative scale (i.e., standardized measurement unit) for pairwise comparisons, enabling efficient data accumulation and further analyses in case of conducting successive research. However, as the AHP uses linguistic expression to measure the responder's decision, it inevitably involves linguistic uncertainty [61]. To overcome this uncertainty, the fuzzy AHP that combines the AHP and fuzzy set theory can be utilized in future research.
This study has several limitations. First, the priority of the strategies to prevent medication errors could change if the survey was conducted with a large number of experts. Further research is needed to confirm the findings of this study by involving various experts in the field of patient safety. Second, although we tried to develop a hierarchical structure with mutually exclusive items, some items might be dependent on each other. Third, when pairwise comparisons between criteria or alternatives were incorporated into the integrated matrices, some comparisons had a CR larger than 0.2. When considering this, it suggests a possible inconsistency in the experts' judgement. We excluded the comparisons with CR larger than 0.2 from the integrated matrices.

Conclusions
We have suggested the priority of preventive strategies against medication errors using the AHP method, which enables structured decision-making based on expertise and experience of patient safety experts. Among the strategies, system improvement in reporting and counterplan by healthcare institutions are likely to play an important role in preventing medication errors. This priority could help policymakers understand the current needs and develop evidence-based policies on patient safety.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/healthcare10030512/s1, Table S1: Integrated matrix and normalized weights of alternatives (cultural improvement); Table S2: Integrated matrix and normalized weights of alternatives (system improvement in reporting); Table S3: Integrated matrix and normalized weights of alternatives (system improvement in cause analyses); Table S4: Integrated matrix and normalized weights of alternatives (system improvement in counterplan); Table S5: Integrated matrix and normalized weights of alternatives (system improvement in assessment).

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Pusan National University (protocol code PNU IRB/2020_58_HR; 16 July 2020).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study. Written informed consent has been obtained from the respondents to publish this paper.
Data Availability Statement: The dataset will be made available upon reasonable request.