What is important to the GP in recognizing acute appendicitis in children: a delphi study

This study had a high response rate, with almost 70% of the invited GPs agreeing to participate in, and then completing, the three Delphi rounds. This suggests that few participants were likely to have a special interest in the subject [13]. As shown in Table 1, our sample included a diverse cross-section by experience, sex, and age. However, cultural, geographical, or educational backgrounds were not used as criteria for the purposive sampling, which could limit the diversity of perspectives represented in the study. The inclusion of 22 participants was in accordance with recommendations found in the literature [13]. Furthermore, our research questions relied on the subjective opinions of experts, resulting in a long consensus list that may not reflect routine practice [13]. The modified Delhi procedure may have caused this lengthy output due to the inclusion of a comprehensive list in the first round, which we designed to avoid biasing the responses of participants. By contrast, a classic procedure without an initial list (i.e., adding features using open questions) might have yielded a different final consensus list. The consensus list can serve as starting point for further research. However, a more concise list would have benefitted clinicians, which would have required extending the Delphi procedure with one or more rounds. The requirement for reassessment in our vignette also implies uncertainty about the diagnosis, which may have resulted in the respondents hesitating to label a clinical feature as unimportant [5]. Performing a study with a different level of uncertainty could prove interesting and may yield a different list. Finally, we did not offer background information on the diagnostic value of the clinical features of appendicitis. Although this might have yielded a more substantiated consensus list, we did not want to influence the participants’ opinions in this way [30].

GPs in this study fully agreed on the importance of recording the location of any pain or tenderness, whereas a cohort study of registration data for children with acute abdominal pain revealed that only 29% recorded findings on right lower quadrant pain and that only 55% recorded tenderness [1]. These findings suggest that a discrepancy exists between what GPs consider important and what is actually implemented in practice, meaning that missing values in registration data cannot be fully explained by our results. This finding also suggests that GP select features to be assessed and recorded from a vast range of features in their cognition which is consistent with the use of illness scripts for pattern recognition.

Other researchers have used the Delphi method to reach consensus on medical reporting in primary and secondary care [31, 32]. A Delphi study on the content of medical reporting by endoscopists could not specify the diagnostic information required for such a report, [33] consistent with our findings and the intuition-based or informal diagnostic approaches used by GPs [34]. Therefore, statistical analysis of the predictive values of clinical features (e.g., in the shape of a prediction model), will benefit from standardized reporting based on a shorter and more practical list of clinical features for reporting.

Notably, participants agreed on the need to request CRP-POCT (78%), commenting that CRP is widely used when managing suspected appendicitis in children. Although the Dutch GP guideline does not support using CRP, due to a lack of evidence that it adds value above symptoms and signs of appendicitis among children in primary care, [10] a recent study of registration data has shown that a CRP test result does add value in this setting [4]. However, in the Netherlands CRP-POCT is available to GPs, whereas White Blood Counts (WBC) are available but not as point-of-care-testing, which could have had an impact the results. Consensus also existed on the need to record safety-netting advice, including the existence of prognostic uncertainty, the alarm symptoms that warrant further assessment, how to find help, and advice about the prognosis [18]. Given that GPs do not universally record safety-netting advice in their current practice, [6] this represents a prime target for improvement.

Our findings have implications for research, medicolegal reporting, decision-making, and access to medical records by out-of-hours care providers and by patients.

For research, the greater likelihood of recording positive than negative findings is incompatible with the assumption of data missing completely at random [2, 5]. Further studies could assess what imputation method (e.g., multiple imputation and zero imputation) produces least bias in datasets where negative findings are more likely to be missing than positive findings [1].

When accounting for care given in medical negligence settings, our results contradict the assumption “if it is not documented, it did not happen,” because the absence of features does not necessarily indicate that care has not been given. Given the consensus on the need to document safety-netting advice by the expert panel in this study, our results could establish a norm for this standard of care [35].

Evidently, the length of our consensus list makes it unsuitable for routine use in support of clinical decision-making in practical guidelines that cover all situations where the differential diagnosis includes appendicitis. Therefore, we suggest clinical guidelines should state the diagnostic value of the most important clinical features relevant to the differential diagnosis, rather than specifying unrealistically long lists of clinical features that should be assessed [36]. A clinical prediction rule for use in primary care would help to make the decision-making process more uniform and effective [37]. Several validated rules already exist for appendicitis in secondary care (e.g. Alvarado, AIR score), but none have been developed specifically for use in primary care [38]. Such a clinical prediction rule could be helpful when a second clinician reassesses the patient by helping to ensure that the most important features are compared. Finally, consensus on the importance of recording safety-netting and other advice may indicate that patients could benefit from signposting to this advice in their electronic health records at the end of a consultation [6].