Methods

To develop the procedure for the identification and categorisation of recovery opportunities, we selected 10 medication errors out of the total set of 52. This sample was representative for the complete set in terms of type of initial error and complexity. In the earlier study, the first author had composed a causal tree for each medication error.6 7 In the present study, both authors independently identified recovery opportunities in those existing causal trees. During a consensus meeting, the results for the 10 selected errors were compared and we agreed upon the identified recovery opportunities, which we subsequently independently categorised. We distinguished between planned and unplanned error recovery and between failed, missed and absent recovery opportunities. Accordingly, we distinguished six categories of recovery opportunities. Planned recovery opportunities involve organisational and technical defences and barriers that are built into the healthcare system to avoid safety-related consequences.2 9 10 Unplanned recovery opportunities are ad hoc solutions that are not formally required and supported by procedures or instructions and largely depend on the problem solving abilities of the people involved.2 Table 2 presents the coding scheme with some examples.

Subsequently, we independently identified and categorised recovery opportunities in the total set of 52 medication errors. If we could not determine which of two categories should be assigned to a particular recovery opportunity, we decided to assign both categories, which each counted for half. Consensus was achieved on all categories.

Finally, we linked the six categories for recovery opportunities to their underlying failure factors to determine the negative influences on error recovery. In the previous study, all failure factors had already been classified according to the ECM. For each recovery opportunity, we registered the related underlying failure factors. Thus, we were able to create a profile of underlying failure factors for each category of recovery opportunities.

Results

In total, 127 recovery opportunities were identified that had been absent, missed or that failed. The number of recovery opportunities per error ranged from 0 to 11; on average, 2.4 recovery opportunities were identified. In only four accidents, no recovery opportunities were identified at all.

Table 3 shows the distribution of recovery opportunities across the six categories. It should be noted that some categories show partial frequencies because in a few cases two categories were assigned to a single recovery opportunity, which each counted for half. Of the 127 recovery opportunities, 94 were planned and 33 were unplanned. Failure to detect and correct initial errors was thus more related to problems with formalised barriers than to difficulties with ad hoc problem solving. In contrast with the planned recovery opportunities, which were almost equally distributed among the three categories, most unplanned recovery opportunities were categorised as unplanned-failed. This indicates that with respect to the ad hoc problem solving cases employees frequently noticed that something was wrong, but were not able to solve it (unplanned-failed). However, it occurred less frequently that employees did not detect an error that in fact should have been detected due to professional expertise (unplanned-missed).

Negative influences on planned medication error recovery opportunities

Table 4 shows how often particular failure factors contributed to unsuccessful planned error recovery. The dominant failure factor is organisational protocols (OP). Absent, incomplete or unclear protocols prevented employees from detecting errors in drug prescriptions or performing double checks after dispensing or before administering the drug. Failure to recover from errors was also frequently because of incorrect or incomplete assessment and verification of the prescription, the drug and the patient before drug dispensing or administration (HRV). Nurses did not read drug labels, or failed to notice a difference in dose between the drug prescription and the dispensed drug. Other factors that made it impossible for employees to recover from initial errors were heavy workload due to management decisions related to staffing (OM) and an organisational culture in which compliance with safety-related procedures is low, and consequently required checks are not always carried out (OC). Regarding the main categories of failure factors, the organisational failure factors contributed the most to unsuccessful planned error recovery.

Negative influences on unplanned medication error recovery opportunities

Table 5 shows how often particular failure factors were underlying unsuccessful unplanned error recovery. No dominant failure factor was identified. Several failure factors to some extent contributed to unsuccessful unplanned error recovery. In several cases, suspicion was present. In those cases, an employee or patient was more or less aware of the initial error, but lack of verification (HRV), coordination with colleagues (HRC) or in-depth knowledge or routine (HKK) prevented them from successful error correction. For example, sometimes nurses or patients did suspect an overdose, but when the physicians were notified, the physicians held to their decision and the wrong dose was still administered. In other cases, the employees involved were not able to solve the problem because of absent, erroneous or unclear (treatment) protocols (OP). In contrast with unsuccessful planned error recovery, human failure factors contributed the most to unsuccessful unplanned error recovery.

Discussion

Theoretical implications

As the ultimate objective of zero errors is unreachable, the current, limited focus of many error reduction methodologies on failure factors is insufficient. Besides those traditional methodologies, there is a need for methodologies that explore why errors are detected and corrected accurately and in time or why not, that is methodologies that discover successful and unsuccessful error recovery strategies.1–3 This study shows that such error recovery methodologies could use accidents as a data source in addition to near misses.

Practical implications

To gain an in-depth understanding of error recovery, hospitals should conduct two types of analysis. For near misses, the steps that lead up to successful error recovery should be identified to reveal positive influences on error recovery. For both near misses and accidents, hospitals should identify unsuccessful recovery opportunities that arose after the initial errors. The underlying failure factors represent negative influences on error recovery. Together, those two sources of information enable hospitals to enhance their resilience by reinforcing the positive influences on error recovery and reducing the negative ones. This study also shows that hospitals do not have to wait for accidents to occur to implement this novel approach, as existing case files and incidents databases can already be reused to obtain information about unsuccessful error recovery.

An additional practical advantage of considering failed, missed and absent recovery opportunities when analysing accidents relates to the fact that positively intended behaviour (albeit failed) is elucidated. Concentrating on the positive mechanisms of error detection and correction might result in larger numbers of reported incidents.11 12

Take-home lessons for enhancing medication error recovery in hospitals

For medication safety, hospitals can mainly reduce negative influences on planned error recovery by adding and improving formalised protocols that employees need to detect and correct errors, and by giving management priority to safety in terms of adequate staffing levels. Furthermore, hospitals should improve existing organisational cultures by increasing risk awareness, for instance by educating staff on safety science and by enabling voluntary and non-punitive error reporting.13 Focused training and instructions can reduce negative influences on unplanned error recovery. Hospitals should ensure that the knowledge and skills of their employees are up to date to enable them to detect and correct errors. Such training should concentrate on both standard (checking) procedures and problem solving abilities, and could (if possible) be simulation based.14 15 As these guidelines are based on data from multiple Dutch hospitals, they will probably also be applicable for other hospitals and possibly for other countries as well. However, hospitals should always verify to what extent the recommendations can be applied. For instance, adding protocols might not always be appropriate, depending on organisational culture and the way protocols are perceived and interpreted.16

Limitations and future research

A limitation of our study is that we conducted a secondary analysis of data that had already been collected in an earlier study. In the present study, we were not able to ask additional questions to the inspectors or the hospitals involved. Therefore, this study should be replicated in a setting in which it is possible to gather recent additional information. Another limitation of our approach is the potential for a hindsight bias, the tendency for people who are aware of the outcome to exaggerate the extent to which the incident could have been predicted beforehand.17 We tried to limit this bias by only using information that had been agreed upon by inspectors of the Dutch Health Care Inspectorate in the earlier study. Because our study is explorative in nature, no formal inter-rater reliability checks were conducted. Future studies should concentrate on the extent to which multiple raters agree on the categorisations. Furthermore, intervention studies should be carried out to discover best practices related to the promotion of error recovery in hospitals, like the study on error recovery strategies in the emergency department by Henneman et al.18

General conclusion

This study shows that accidents can be used as an alternative data source to near misses for the analysis and understanding of error recovery. By using both sources, hospitals can enhance their resilience by reinforcing the positive influences on error recovery and reducing the negative ones. Although this is a very important safety strategy, traditional error reduction methodologies, which concentrate on eliminating failure factors, are equally important. In other words, triangulating information is necessary to provide a complete and comprehensive picture.19 20 Hence, only by applying the two complementary safety strategies of error reduction and error recovery promotion, hospitals can significantly improve patient safety.