Patient safety and analysis of errors

Medication errors (ME) may potentially be a threat to patient safety. They are among the most common treatment errors.(1–3) According to the literature, up to every fifth medication is wrong, while every fourteenth imposes a serious risk to the patient.(1–5) Many of these ME could be prevented once their causes were known and the underlying processes adjusted. This article gives an overview on the techniques of error analysis, how to draw conclusions from these analyses and how to prevent ME in the future.

Errors are a natural element of any work.(3,4) The Swiss Cheese model, propounded by Dante Orlandella and James T Reason of the University of Manchester, demonstrates the relation between a hazard and its underlying causes.(6) It illustrates how a sequence of ‘holes’ in a process eventually leads to a hazard. Only when all barriers between the respective cause and the possible damage fail is a patient in actual danger. The majority of errors, however, can be prevented by either removing the primary cause or by imposing effective barriers that obstruct the path leading to the error. Patient safety can be increased by using smart and efficient error-prevention strategies; these include errors that are constantly analysed and the results of such an analysis are utilised to draw the appropriate conclusions. Changing the process that ultimately endangered the patient can then eliminate the causes of such errors. The entire staff involved in the patient’s care should perform both the analysis and the respective prevention strategy.

There are a great number of different methods to analyse errors. Below, we will restrict ourselves to some techniques that turned out to be particularly useful in our own work in the hospital. However, in order to understand the analysis, we first have to focus on the data sources that may be utilised for this process. In essence a differentiation between two primary sources can be made. Either single case reports or a complex set consisting of many collated reports may be used. Systems based on a large pool of data are usually preferable. In addition, this approach allows the possibility of drawing conclusions about the underlying causes and to gain a deeper knowledge on the different types of errors. Single case reports offer a better insight into the particular processes and the respective causes of the error. For an assumption on the actual risk to patient safety, an expert familiar with the respective case has to be involved.

It should be mandatory to summon an interdisciplinary team of experts that performs the initial analysis. This team should be structured hierarchically and have all the resources necessary to apply the measures that are deduced from the analysis. The first step is deciding whether a detailed analysis of an error report has to be performed at all. A risk analysis only makes sense when the events related to the error has a potentially severe outcome, when a process is very complex or bears hidden risks. Also, when an accumulation of similar events had taken place in the same environment, such an analysis may be considered.

Root cause analyses are performed to deduct the cause of isolated events. They are applied preferentially on a singular process or a system with a low degree of complexity. The analysis should be based on the following questions:

A chronological sequence of the chain of events is of particular importance when this method is employed. The root cause analysis (RCA) can be visualised with a flow chart or alternatively with a timeline that includes all causes that contributed to the event. The RCA is particularly useful when a process consists of only few steps. With complex events, RCAs tend to get confusing. The more detailed the initial process, the easier it is to find the reasons contributing to the error. Based on the chronology of the events, each step in the process and the respective responsible person should be defined. This approach simplifies the search for solutions and opens the way to reassign responsibilities. Figure 1 depicts a RCA that evaluates the risk of double prescription of proton pump inhibitors.

The London protocol (LP) sets the base for the analysis of unwanted clinical events. It employs a structured and interdisciplinary approach that goes beyond the RCA. For the LP it is essential that the analysis is supplemented by an assessment of the most probable causes. The identification thereof should play an important role in error-prevention strategies. It is apparent that errors are hardly caused by a single person but rather by the system itself. In this context, all departmental or hierarchical structures and the decisions made by the individuals responsible have to be taken into account.

A comprehensive analysis will study each step in this process. The first step consists always in the identification of errors or actions that might eventually lead to harm or have the potential to do so. An example of how to apply the LP is the Ishikawa diagram (ID).

The ID is a useful technique to isolate factors that give way to errors. It can be recognised easily by its unique ‘fishbone structure’. The ID is also suitable to visualise complex processes. Nevertheless, it should only be employed to study a single, isolated event. In addition, the ID can be recommended to point out prevention strategies, once a cause for an error has been identified. The very first step is to define the basic structure of the ID and to assign labels to the scales of the diagram. The structure may be designed individually and categories such as ‘Man’, ‘Machine’, ‘Material’ and ‘Methods’ can be used in this process. In some cases the factors already described in the LP, such as ‘Organisation’, ‘Institution’, ‘Patient’, ‘Individual’, ‘Team’, ‘Task’ and ‘Work Environment’, may be applied instead.

The event, which is essentially the effect of all contributing factors, is placed on the right-hand side of each ID. When adding the main causes as labels as individual scales of the diagram, a greater degree of complexity may arise because more and more factors are identified in the process. These factors can be added as subcategories. Sometimes factors have to be assigned to more than just one single category. For example: ‘insufficient training’ can be put into the category Man as well as into Methods. After completing the diagram, the next step consists of the evaluation of all causes of the error in terms of the likelihood of being contributing factors. For this analysis, pre-existing literature and the expertise of the interdisciplinary team can be taken into account. Assigning scores for each individual cause can help in the evaluation. In order to find an optimal prevention strategy the causes with the highest score should be eliminated first.

The advantage of the ID is the comparatively low effort required, the good clarity as basis for group discussions and that it is relatively easy to learn and use. In addition, the ID helps to understand complex relationships.

Disadvantages when using the ID are that the clarity tends to get lost when the complexity of the problem increases, the impossibility of visualising interconnected relationships and insufficient validity when it comes to timing of events. A clear statement on the prevailing cause cannot be made with absolute certainty since most errors are caused by many factors simultaneously.

IDs can be created in the future using few resources since the causes for similar errors can be used for many purposes. For instance, the error ‘double prescription’ can be used for more than just one active substance (see Figures 2 and 3, labelled in red). In such a case, large subsets of IDs can be reused for other purposes. However, when errors differ by a large part only few causes can be compared (see Figures 2–4, labelled in green).

An ID and double prescription of a proton pump inhibitor

Step 1: Draw the diagram and assign the individual groups. Make sure using a concise formulation of the event or error (Figure 5).

Step 2: Add the main causes and sub-causes (branching points).

Step 3: Evaluation of all causes and identification of the most likely factor involved in or leading to the event by the team.

Errors are facilitated by multiple factors. It is apparent that the more complex a process is the more prone it is to errors. With an increase in the number of steps in a process, the risk of errors increases exponentially (Table 1).

The LP divides causes for errors into different categories:

A number of studies have demonstrated that in many different sectors (for example, aviation and medicine) similar causes for error can be detected: high workload, fatigue, insufficient skills or experience, lack of proper guidance or control, stressful environment, poor communication insufficient planning or organisation, faulty equipment or bad working materials.(10,11)

Errors can only be successfully prevented if all persons involved are able and willing to implement the recommended measures. This can be achieved by developing high sensitivity towards risks, which engenders a safety culture that can be communicated to the entire staff.

According to the Pareto principle, 80% of all errors come from only 20% of all causes. Once these main causes are identified little effort is required to prevent errors in the future; however, in order to achieve this goal the entire system has to be kept in mind. By the 1980s, Deming Edwards had demonstrated that in 94% of all errors, a problem in the process structure is involved.(12)

Generally a distinction is made between preventative and discovery measures. One can either decrease the likelihood of a respective cause to happen (that is, by optimising the process) or improve ways to discover the cause by implementing additional routines. A high level of reliability is maintained whenever a process fulfils the following conditions: sensitivity to errors, awareness towards the flow of the processes, high degree of organisation, strong interaction within the team, high flexibility and improved professional knowledge.(13,14)

An important principle in order to prevent errors is ‘Poka-yoke’ (which is a Japanese term for ‘mistake proofing’). The thinking behind the concept is that processes or products are designed so that errors are prevented a priori. A perfect example is the introduction of different types of connectors for epidural or intravenous administration of drugs.(15)

The LP emphasises in particular on the fact that optimised teamwork and communication play a major role in the prevention of errors. Hallmarks of an optimised team are:

Communication can be improved by a number of different factors:

In this article, we have described several methods that can be applied depending on the individual circumstances. Many pharmacies already use these tools as part of their quality control system and should be encouraged to discover their potential to prevent errors. The contents of the LP are useful as part of a familiarisation process of the subject. It is, however, difficult for beginners to learn the details of the LP and implement it into their daily practice. Working with the RCA is the method of choice for small teams, while the ID can best be utilised in larger and more interdisciplinary teams. Good examples for such teams are our hospital’s interdisciplinary quality circles or the conference on mortality.

Error analysis is a complex subject. The reasons for errors can be identified efficiently and prevented in the future by using tools and resources that have previously been implemented in quality management systems. Many causes of errors can be attributed to the same recurring events, which is why prevention strategies are often similar. For almost any error analysis, it is mandatory that an interdisciplinary approach is chosen in order to consider all relevant factors leading to the event. In addition, every analysis has take the entire system into account, because often factors play a role that are not directly related to the area where the error occurred.