In capturing and reporting the less than optimal confidence calibration of nurses and students, this study offers both information (and a methodology) for those developing high fidelity clinical simulations (particularly for assessment of critical care events). In the high fidelity clinical simulation environment, we observed a clear relationship between nurses’ subjective confidence ratings and accuracy in their risk assessments: experienced nurses were generally overconfident, while student nurses tended toward underconfidence. The difference of this measure between the two groups was statistically significant (p = 0.01). The findings showed that the subjective probability judgments of experience nurses and students were subject to systematic bias; either they overestimated or underestimated their judgmental abilities or knowledge of self judgment. Our findings replicate the more general psychological picture that suggests people (including decision makers with more experience) are often systematically overconfident with regard to judgment accuracy [6, 24, 29–34].
An appropriate level of confidence, given someone’s clinical experience, is one marker of a nurse’s competency, and clinical experience is a significant factor in building confidence in nurses’ judgment [9, 35, 36]. Our study showed that experienced nurses were significantly more confident in their judgments than students, and that nurses’ confidence increased in line with clinical experience. However, we saw no significant benefit on judgment accuracy arising from clinical experience on judgment accuracy in the high fidelity clinical simulation environment. Similar findings were also observed in other studies. For instance, the study by Oskamp  showed that experienced clinicians’ judgments were no better than those of graduate students. A further study by Corcoran  did not find better performance in the accuracy of treatment plans developed by experienced nurses compared with novice nurses. Hamers et al.  also observed a similar level of assessment performance in pain intensity between experienced and student nurses. Ericsson et al.  demonstrated that a failure to reliably isolate superior performance amongst nurses with extensive years of experience appears to be a common trend. A similar pattern was observed in doctors; the systematic review of effects of clinical experience on medical performance showed a higher risk of providing lower quality of care in doctors with more years of clinical experience .
Given that nurses experience significant amounts of audio and visual information (which is a mix of important signals and ‘noise’) in daily clinical activities, it is reasonable to hypothesise that experienced nurses are more likely to have better calibration performance than student nurses in high fidelity simulated conditions. However, our findings showed no significant difference in calibration and resolution between experienced and student nurses. Our study does not support the hypothesis that confidence calibration performance is a linear function of clinical experience, even in the less than perfect environment of the high fidelity clinical simulation.
Task difficulty and calibration
Our findings further reveal that nurses’ calibration differs with the difficulty of the judgment task they are faced with. Nurses’ calibration and resolution were generally worse on the more difficult and uncertain tasks. By varying the task difficulty, a hard-easy effect was seen: nurses are overconfident in hard judgments and underconfident in easy judgments.
Similar findings have been documented in psychological studies since the 1970s [3, 11–14]. These studies consistently conclude that the extent of miscalibration relies on the degree of ease or difficulty of tasks: overconfidence is most extreme in judges faced with tasks of greater difficulty . Lichtenstein et al.  note that the hard-easy effect seems to arise from people’s inability to appreciate the ease or difficulty of a task. Therefore, nurses’ confidence miscalibration may reflect a lack of sensitivity (and commensurate lack of subjective probability adjustment) to the difficulty of tasks.
As with confidence miscalibration, nurses’ ability to resolve information altered as a result of task difficulty: their ability to differentiate correct from incorrect judgments decreased as task difficulty increased. This is also consistent with the psychological literature [42–44] showing that resolution is often better in easier judgments. Similarly, the nurses’ discrimination abilities differed drastically between easy and difficult tasks; with discrimination fairly good on easy tasks, but deteriorating as tasks become more difficult. The strikingly different calibration curves for the two levels of task difficulty (Figure 3a and 3b) may result from nurses’ not really “knowing” their judgments, particularly in difficult cases.
Time pressure and calibration
Time pressure had no significant impact on nurses’ confidence, the percentage of correct judgments or their overall calibration. This finding runs counter to those studies [17–19], showing that decision makers’ confidence lessens under time pressure and tends to increase with the amount of time spent on tasks. One plausible explanation for this finding is that nurses may experience a “mild” state of time pressure that does not necessarily reduce their confidence. Thus, without sacrificing confidence and accuracy, nurses adapt well to this state of time pressure by accelerating information processing under time constraints. This is in line with the thesis that humans think “adaptively” in situations in which resources are limited .
Time pressure increased nurses’ confidence in easy cases and reduced nurses’ confidence in the difficult ones. Such a significant interaction revealed that time pressure had a different effect on confidence between easy and difficult judgments. This phenomenon could be partially explained by “the need for closure” effect . Need for closure refers to a need for certainty, it arises from the impact of time pressure on participants’ motivation and confidence [46–48]. Particularly, when an immediate judgment must be made within a limited time, the need for closure motivates participants to consider fewer hypotheses and be more confident in their favoured hypothesis. Thus, the raised confidence is highly correlated with the (perceived) need for closure. In contrast, without the need for closure (i.e. no time pressure), nurses would tend to seek more information in information processing with a number of competing hypotheses considered, thereby reducing their confidence in any hypothesis. These studies, however, did not differentiate the effect of need for closure on easy and difficult cases. In this study, the need for closure under time pressure significantly impacts on easy judgments in the form of increased confidence. However, the inverse effect of time pressure on confidence associated with difficult tasks suggests that it acts differently in difficult judgment situations.
Recent evidence has shown a significant interaction between the need for closure, judgmental performance and changing judgment task demands (for example, altered task difficulty) . Roets et al.  suggest that once tasks are perceived as difficult, willingness to invest effort is reduced, even though the task demands are high with an initial high level of motivation arising from the need for closure. The feeling of investing a great deal of cognitive effort in difficult tasks in a judgment process can decrease the level of confidence [49, 50]. Furthermore, others have shown that task difficulty has a significant influence on judgmental response times: response time increases as judgment difficulty increases . Thus response time is often required to be longer in difficult judgments than easy judgments. Our findings suggest that, due to minimising cognitive efforts for difficult judgments under time pressure, it is reasonable that nurses correspondingly assign lower confidence to difficult judgments that require more cognitive efforts when the response time is decreased.
A non-random sampling method to enrol nurse participants was a limiting feature of the study. Whilst deliberately sampling experienced and inexperienced nurses allowed us to investigate the mediating effect of clinical experience on confidence calibration, we could have increased the risk of non-representativeness within subgroup samples. Furthermore, the focus on judgment task of risk assessments in acute care means that the generalisation of the findings to other settings is limited. Further research is required to establish whether the patterns of confidence miscalibration observed in this study are replicated in different clinical contexts.