Strengths and limitations of the study

This study employed mixed methods to provide a detailed account of the relationship between MW and dispensing errors detected in a final accuracy checking task. An experimental setting was useful to study the relationship between missed dispensing errors and MW under controlled conditions as shown in the cognitive systems

performance model (see figure 2 in chapter 1). This model highlights the confounding effect that the task characteristics and working environment can have on the

cognitive processes that pharmacists use to carry out a final accuracy check as well as the amount of MW experienced. The experimental setting also allowed for comparisons to be made across participants as they all carried out the same task, with the same batch of dispensed items, in the same setting which is not possible in observational studies of pharmacies and dispensing errors, or through analysis of the national reports of near-miss and unprevented dispensing errors.

Our experimental study into the relationship between MW and unprevented dispensing errors is based on previous research carried out by researchers in the United States of America (Grasha, Reilley, Schell, & Tranum, 2001; Reilley, et al., 2003; Reilley, et al., 2002; Schell & Cox-Fuenzalida, 2005; Schell, Woodruff, Corbin,

& Melton, 2005). These studies were also conducted under controlled conditions, and

| 91 used the same measures of MW, mood and mental states. However, aspects of the task design (the percentage of dispensed items that contained dispensing errors) and the participants they used in these studies (mainly undergraduate psychology

students, although one larger scale study also involved pharmacists and pharmacy technicians) limited the generalisability of some of these findings to a UK community pharmacy setting. The present study involved community pharmacists from the UK who were strategically recruited to match the national demographic characteristics of community pharmacists practising in the UK to provide the most feasibly

representative sample possible.

Our study also used a lower error rate (10% of dispensed items contained an error) compared to the studies carried out in the US (≥24% of dispensed items contained an error). The lower error rate in our study was used because such high dispensing error rates have not been observed or reported in UK community or hospital

pharmacies. This was important as the prevalence of targets (i.e. prevalence of dispensing errors) has been shown to affect performance on ecologically valid visual search tasks (e.g. airport security baggage scans: Van Wert, et al., 2009; Wolfe, et al., 2005; Wolfe, et al., 2007) and therefore this confounding variable had to be held constant at an ecologically valid level throughout the study. This presented a difficulty for this research as there were very few opportunities for pharmacists to miss or detect errors in the two accuracy checking tasks which may have restricted the ability to identify differences in performance between the 4 groups of participants and to measure the impact of the task characteristic (number memory) and environmental (distractions) manipulations. One way to provide more opportunities to measure pharmacists’ accuracy checking performance was to add in more tasks and therefore more study participation time for each participant. However this needed to be

balanced against the potential for increasing participants’ levels of study induced cognitive fatigue, which may confound error rates and MW ratings. Although most community pharmacists will work shifts longer than 2 hours which was the time it took to complete our study so likely experience these fatigue effects in real life.

Interestingly, one of the most complicated aspects involved in setting up this study was ensuring that no un-intended dispensing errors entered the task items. During the pilot study we found a couple of errors that had been made during the dispensing of the task items but had not been picked up by two community pharmacists who were asked to independently check the items before they were used in the study.

| 92 This is further illustration of how difficult it is to catch all dispensing errors on a final accuracy check.

A great strength of this study was the ability to account for the change in cognitive processes as participants move from novice by comparing the pharmacy students’

and community pharmacists’ performance on the checking tasks. However, a limitation of this study was that we were unable to also recruit pharmacy technicians to take part in our study.

In this research we also found that the working memory task used for experiment 1 did not sufficiently load our pharmacists’ or pharmacy students’ working memory or cause an increase in mental workload as expected which meant that we were unable to test directly the impact of a task which carries higher cognitive load on participants’

ability to detect errors.

The MW diary study which allowed participants to rate their MW throughout a day at work has not been used in community pharmacy practice before. The diary study showed that the overall MW scores obtained under controlled experimental conditions were comparable to those pharmacists’ reported experiencing in their

“real-life” pharmacies. However, there were differences found on the individual items of the NASA-TLX questionnaire. For example, reports of physical demand involved in the task were higher on the MW diary. This finding is not surprising, as pharmacists were asked to sit down whilst checking the dispensed items in task 1 and 2 and reflects a key limitation of the study that we did not have standing height work

benches for pharmacists to work from. Standing height benches are commonplace in community pharmacies and pharmacists will spend most of their working day on their feet, thereby increasing the physical demand of their job. We were also unable to simulate the increases in task frustration pharmacists reported in their diaries which, the semi-structured interviews identified as relating to feeling that the amount of work they were required to complete was not possible in the time available. This aspect of MW reflects the complexity of the pharmacists’ role in balancing many competing demands and tasks each day and emphasises the importance of carrying out diary studies like this to identify factors which increase MW which cannot be reliably reproduced in a laboratory setting. This also highlights the benefits of carrying out semi-structured interviews with pharmacists which allow us to identify other issues relating to MW, for example, the strategies that pharmacists use to manage their MW

| 93 which would allow them to reduce their MW during an experimental task, as well as in their day-to-day work.

Carrying out experimental work into the relationship between dispensing errors and MW is advantageous for the reasons mentioned above, however, the limitations of this approach include, the time consuming and costly nature of this work compared to survey and interview studies. This limited the sample of participants that could be recruited for the experimental phase of the research which may have restricted the power to detect significant findings, despite a-priori power calculations being carried out.

Another limitation is that the experimental setting required the deconstruction of a complex task like dispensing into discrete stages so that the specific cognitive processes involved in a final accuracy check could be measured. This limits the generalisability of these findings to other aspects of the dispensing process or other tasks pharmacists undertake which have safety critical elements. The final accuracy check itself was also designed in a way that would avoid pharmacists carrying out other checks (e.g. the legal and clinical check) of the prescription, all participants were asked to focus on the accuracy check and single item prescriptions were used to avoid pharmacists looking for drug-drug interactions. Many prescriptions that community pharmacists dispense and check will have multiple items listed, therefore this is a further limitation of this study as it does not quite reflect the complexity of the accuracy checking process that occurs in community pharmacies.

Previous research has also shown that pharmacists vary in the strategies, or more specifically the order in which they verify the label and dispensed item against the prescription (Anto et al., 2013). The checking procedure that participants utilised was not assessed in our study and therefore we cannot identify whether their checking strategy was related to their performance on the accuracy checking task.