Surgical specialties carry one of the highest risks of iatrogenic harm,1 with the majority of these adverse events deemed preventable.2 While surgeons have been reviewing the complications of practice for many years in Morbidity and Mortality (M&M) meetings, direct observation or retrospective case review are the predominant methods for reliable measurement of error in healthcare.34 Such techniques are heavily resource-dependent and are restricted to brief sampling studies with limited scope. By contrast, incident reporting is always operational, accessible to all staff and active across whole organisations. Although incident reporting can contribute to improvements and developments in safety,5 it has yet to be seen as unequivocally valuable.

It is recognised that reported incidents under-represent the true state of affairs.6^–8 Studies have investigated the reasons for incident under-reporting in obstetrics9 and across whole hospitals,10 but this has not been previously investigated in an exclusively surgical setting. The special situation of surgical complications is particularly worth considering. These are classed as adverse events by formal systems analysis1 but may not be perceived as such by staff. Better understanding of incident reporting behaviour in surgical care may help to improve the quantity and quality of reporting and the utility and accuracy of the analyses that arise.

This study considers the sequential steps in the process of surgical incident reporting. First, staff knowledge of the local reporting system is evaluated. Next, the factors that influence perception of an incident as being reportable are investigated with particular attention to the effect of outcome and the propensity to report surgical complications. Finally, a range of practical and psychological barriers that may discourage staff from reporting a recognised incident are explored.

METHODS

Study design

A questionnaire survey was carried out across the four wards responsible for elective and emergency General Surgical admissions in a UK teaching hospital between January and March 2007. A total of 203 trained staff were eligible to participate (72 doctors and 131 nurses). The questionnaire11 was anonymously completed via a web-based application.

Questionnaire design

The structure of the questionnaire was based on two previous studies910 with modifications to reflect specific issues relevant to surgery. A draft questionnaire was refined following review by three senior clinicians and the site clinical risk management officer plus piloting on five members of staff. It consisted of three sections. The first elicited basic demographics and examined whether staff understood how to complete and submit a form to the local reporting system, how recently they had completed a form and how many they had completed in the last 12 months.

The second section assessed which incident characteristics influence the likelihood of it being perceived as reportable. Staff indicated how frequently they would complete a form in a given clinical scenario on a four-point Likert scale (always, more than 50% of the time, less than 50% of the time or never). Three stem scenarios were designed based on commonly reported clinical incidents—falls, drug errors and clinical management. The scenario chosen for the latter was a group and save (G&S) incident. For each stem, three alternate outcomes were given to reflect either harm, no harm or harm prevented, as defined by the NPSA’s classification of adverse events by outcome12 (table 1). All incidents should have been reported regardless of outcome. To assess whether staff perceive surgical complications differently, this section also evaluated the reporting of five common surgical complications which, again, should all have been classed as patient safety incidents1 and reported. All questions were randomised in order by the computer.

The final section of the questionnaire investigated the practical and psychological factors that prevent reporting once an incident has been recognised. Adapted from the study by Vincent et al,9 four general areas are identified: lack of time, fear, lack of feedback and lack of understanding of what constitutes a patient safety incident. Staff were asked to rate the importance of each potential reason on a five-point Likert scale (1 = strongly agree, 5 = strongly disagree).

Statistical evaluation

The chi-square test was used for dichotomous measures. Confidence intervals were used to test for percentage differences in the likelihood of reporting. Reporting behaviour was evaluated using analysis of variance (ANOVA 3×3×2−3×level of harm; 3× incident type; doctor vs nurse), with response categories scored between 0 (never) and 3 (always).18 Bonferroni correction was applied to all 72 tests, with an alpha of 0.05 and conservatively assuming a correlation of 0.1 between measures, giving a significance level of p = 0.001.

RESULTS

Section 1: response rate and familiarity with reporting

Fifty-five (76.4%) doctors and 82 (62.6%) nurses responded, giving an overall response rate of 67.5% and a good representation of all levels of seniority (doctors comprised 33% consultant, 27% SpR/SHO, 40% F1; nurses comprised 26% sister/team leader, 74% staff nurse). One nurse and six doctors did not know of the local reporting system. Of those that were aware of the system, nurses were significantly more likely than doctors to know where to find a form (98% vs 43%, z = −4.99 p<0.001) and to know what to do with the completed form (89% vs 29%, z = −4.77, p<0.001). Figure 1 shows when respondents last completed a form and how many they have completed in the last year by profession and seniority. Doctors were significantly less likely to have ever completed a form (15% vs 65% z = 3.81, p<0.001). Most nurses (45%) had completed between two and four incident reports in the last year, while the majority of doctors (69%) had not completed any.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Time since last report submitted and number of reports made by profession and seniority.

Section 2: reporting behaviour

Figure 2 shows the level of harm, type of incident and likelihood of reporting by doctors and nurses. The level of harm (F(1.8,246) = 254.2, p<0.001), incident type (F(1.9,258) = 64.4, p<0.001) and profession (F(1,135) = 20.7, p<0.001) all significantly affected the likelihood of reporting. Staff were most likely to report an incident where harm occurred and least likely to report an incident where harm was prevented (near miss). Nurses were almost three times as likely to always report no-harm events compared with doctors. The incident type was also important. Allergic drug reactions were reported more frequently than falls and both were reported more frequently than incorrectly G&S patients (see table 2 for details).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

How often would you fill out an incident form?

Figure 3 shows the likelihood of reporting surgical complications. These types of incident were far less likely to be reported than the fall, allergy and G&S incidents, regardless of profession. Even comparing the most reported surgical complication (pressure sores) with the harm subgroup of the least reported “standard” incident (G&S) and taking the likelihood of reporting as >50 and “always,” there is a significant difference for doctors (53% G&S incidents vs 15% pressure sores, z = 4.633, p<0.001) and possibly for nurses (75% G&S vs 55% pressure sores, z = 2.737, p = 0.003) although this falls outside the assumptions of Bonferroni.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Reporting of surgical complications. DVT, deep vein thrombosis.

Section 3: perceived obstacles to reporting

Figure 4 shows how staff responded to suggested reasons for not reporting. Doctors were more likely to agree that any one reason presented a barrier to reporting than nurses. Fear was a significantly less important barrier than any of the other reasons, which were otherwise relatively uniform. The two fear parameters are significantly different from all other measures aside from each other, and the least difference is between “litigation” and “contribution to quality” items (z = −3.49, p<0.0002).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

I am unlikely to complete a clinical incident form because…

DISCUSSION

The likelihood of reporting an incident depends on three stages. First, staff must be aware of an incident reporting system and know how to use it. Second, they must be able to recognise reportable incidents. Third, they must overcome a number of psychological and practical barriers that may prevent them completing a report.

Our findings correlate with those of other studies that have found differences between doctors and nurses in the likelihood of incident reporting.91013 Doctors report less frequently because they are less likely to know how the system works, less likely to think a particular incident is worthy of reporting and more likely to be discouraged from reporting.

Both the incident type and level of harm also influenced the likelihood of reporting. An event resulting in patient harm is most likely to be reported. When harm is prevented (“near miss”), the likelihood of reporting is least. No-harm and near-miss events present a valuable learning opportunity for the identification and rectification of factors that contribute to error before harm occurs.14 However, in a system where such events are frequent,15 it may be difficult to identify them as reportable. The type of event is also influential, with falls and allergy incidents having a higher likelihood of being reported than G&S incidents. This may be because habitually reported events such as falls are immediately recognisable as reportable with consequent reinforcement of reporting behaviour. Less frequently reported events, such as an incorrectly G&S patient, may not be recognised as an incident without a degree of consideration and do not therefore automatically elicit a report. Alternatively, staff may be reluctant to report incidents where human error is more obviously to blame, such as an incorrectly G&S patient, in contrast to a situation in which an individual may not feel directly responsible, such as a fall.

The present study demonstrates that surgical complications are especially unlikely to be reported, even though they should probably be classified as adverse events.1 Clearly there is not a failure to recognise these events, since they are openly discussed and recorded at M&M meetings. M&M intends to achieve many of the same goals as incident reporting and is open to similar potential selection biases. While generally poor at achieving improvements in quality,16 it reflects a clear desire for clinicians to examine and improve their own practice. Presumably surgeons regard complications as inevitable consequences of a high-risk activity which are reviewed and learnt from at the M&M, rather than as “patient safety incidents” which need to be reported to the risk-management system. Misalignment between these two systems goes some way to explaining the under-reporting of surgical complications.

Given that surgical complications are openly presented for peer criticism at M&M meetings, their under-reporting is unlikely to be related to fear or a desire to hide mistakes. Of the perceived barriers to reporting, no unique cause was identified as being more important than any other, confirming the view that this problem is multifactorial.910 However, fear was considered to be significantly less important than other factors. This could suggest an increasing acceptance of incident reporting as a non-punitive part of everyday practice and a move away from the blame culture previously deemed a significant barrier.17 This is the first study to report this, and it may be a premature conclusion at this time. Examining whether this apparent reduction in fear is confirmed in future studies would be worth while.

We must acknowledge the limitations of a questionnaire design which cannot account for the inevitable degree of difference between what people say they do and what they actually do. Furthermore, if staff are reporting specific incident types to other systems (eg, G&S incidents to the UK SHOT reporting system)18 this could result in lower reporting rates locally, but specific questioning of a subsample of staff revealed no evidence of this. Finally, while general underlying barriers to incident reporting have been examined across the multidisciplinary spectrum in order to draw correlation with other studies, this study did not have sufficient scope to explore all the reasons why surgeons do not report complications. This would require a further targeted study of surgeons alone, perhaps benefiting from a preliminary qualitative approach to identify a range of influential factors, then conducted across several centres to gather sufficient numbers.

Incident reporting systems have helped change attitudes towards safety and error19 and the range of national projects to develop such systems should be commended. The strength of reporting systems is in characterising incidents and learning from them. However, under-reporting is commonplace, particularly among doctors, and is also influenced by incident type, level of harm and the overall view that some safety incidents may be more traditionally viewed as “complications.” Better understanding of the relative motivations and disincentives will facilitate the evolution of more robust reporting systems. In addition, integration of traditional quality and safety mechanisms such as M&M into the modern and evolving hospital safety infrastructure may facilitate the assimilation of otherwise overlooked data. If individual specialities assume ownership of incident data, then more consistent higher-quality reporting may result.1920 Improving incident reporting systems, therefore, demands progress towards more integrated safety and quality-improvement systems, many of which exist already in the minds and habits of doctors.