The Impact of a Diagnostic Strategy for Acute Appendicitis in Children With Acute Abdominal Pain in Primary Care

BACKGROUND About 10% of pediatric general practitioner (GP) consultations are for acute abdominal pain of which about 5% have acute appendicitis (AA). Delaying a diagnosis of AA and subsequent appendectomy increases the short and long-term morbidity. AA in an early stage is difficult to distinguish from other (self-limiting) causes of acute abdominal pain (e.g. urinary tract infection, constipation and gastroenteritis), resulting in missing 19% of children with AA at first presentation in primary care and 70% non-AA cases among referrals, which has a negative impact on the child and parents, such as anxiety and psychological distress.

As urgent illnesses other than AA are very rare in children with acute abdominal pain, the yield of referrals in terms of detecting other conditions than AA that need urgent specialist care is low. An evidence based diagnostic strategy for AA referral could help the GP in the diagnostic process, thereby reducing non-AA referrals without missing children with AA.

OBJECTIVE AND HYPOTHESIS The objective of this study is to evaluate the impact of a diagnostic strategy for AA, consisting of an externally validated cPR based on seven signs and symptoms, selectively followed by a CRP-POCT in the medium-risk group, on referral efficiency in children with acute abdominal pain in primary care, as compared with usual care. The hypothesis is that the diagnostic strategy will decrease the proportion of non-AA referrals, without delaying the diagnosis of AA.

STUDY DESIGN A pragmatic cluster RCT will be conducted with 1:1 permuted-block randomization of general practices to the intervention or control group using randomly varying block size. Stratification will be based on the GP practice size (greater or smaller than 5000 patients). Follow-up is 30 days for the primary outcome (efficiency) and 30 days and 3 months for secondary outcomes (7). Alongside the trial, a process evaluation will be performed according to the RE-AIM framework (Reach, Effectiveness, Adoption, Implementation, Maintenance).

STUDY POPULATION Inclusion criteria are: children, 4 to 18 years, presenting with acute abdominal pain (onset ≤ 7 days) in primary care.

Exclusion criteria are: a history of appendectomy, pregnancy, traumatic cause of abdominal pain.

INTERVENTION The intervention in this study is a diagnostic strategy for AA referral using an externally validated cPR selectively followed by a CRP POCT in the medium risk group. By using the diagnostic strategy, GPs will evaluate a set of signs and symptoms on a structural basis and get guidance for indication and interpretation of CRP POCT, for which current practice is very heterogeneous. The cPR includes seven signs and symptoms of AA: sex (2 points), duration of symptoms (< 24 hrs= 1 point, 24-48 hrs= 2 points), nausea or vomiting (2 points), elevated temperature (≥ 37.3 °C) (1 point), tenderness of the right lower quadrant (5 points), peritoneal irritation (4 points), and abnormal bowel sounds (1 point). For children in the low-risk group (score 0-7; mean probability AA 0.5%) safety netting advice should be provided. In the high-risk group (score ≥14; mean probability AA 41.0%), the GP should refer the patient to the emergency department. In the medium-risk group (score 8-13; mean probability AA 7.5%), GPs should perform CRP POCT according to Dutch laboratory quality standards. CRP values <10mg/l (mean probability AA 1.3%) imply safety netting advice and for CRP >50 mg/l a referral is needed (mean probability AA 41.8%). In case of a CRP value between 10 and 50 mg/l (mean probability AA 16.8%), the GP will discuss the following two management options with the patient: 1) a reconsultation later the same day or the next day; or 2) immediate referral. The cPR will be made available electronically for GPs, by a link integrated in the GP registry.

USUAL CARE Care provided as usual is according to the recommendations of the Dutch College of GPs guideline 'abdominal pain in children', which states that AA could be suspected when a child presents with pain and tenderness in the right lower quadrant and fever. Children with peritonitis signs and abnormal bowel sounds need to be referred. In case of doubt, GPs should perform safety netting and/or plan reassessment. CRP testing is not recommended, because lack of evidence for added diagnostic value in primary care. The GPs in the control group receive a leaflet with a summary of the NHG guideline 'abdominal pain in children'.

SAMPLE SIZE CALCULATION Results from previous studies show that the cut-off for the high-risk group based on the cPR had a specificity of 97% and that CRP value of ≥50 mg/l in children in the medium risk group had a specificity of 94%. As the weighted mean of these is a specificity (efficiency) of at least 95%, an improvement of efficiency from 88% (current care) to 95% is expected. In order to detect this difference, with 80% power, using a two-sided 5% significance level, a cluster size of 3, an intraclass correlation of 0.01, and expecting 10% loss to follow-up, a randomized controlled trial would need to include 566 children without AA (283 per group). Additionally, all children assessed for AA, including those with AA, will be included until the target sample of children without AA for the primary outcome analysis is reached. Assuming two to six eligible children per practice over two years, we aim to recruit at least 170 general practices with a mean patient population (85 per arm).

STATISTICAL ANALYSIS The primary outcome will be evaluated on an intention-to-treat basis. Children will be analyzed in the intervention or control group based on the GP practice in which they are registered, irrespective of the actual care received. In addition, a per protocol analysis will be done. In the intervention group according to the per protocol analysis, only children who received the intended diagnostic strategy and CRP-POCT testing in the prescribed way will be included. In the control group, we will only include children who did not undergo CRP POCT testing. Separately in both randomized groups, characteristics of children will be compared in the intention-to-treat and per-protocol analysis. A multilevel logistic regression model will be used to analyze the primary outcome, accounting for clustering of observations within GP practices and adjusting for the stratification variable (practice size) and, if necessary, for baseline differences. Exploratory analyses will assess subgroups by age and sex.

For secondary outcomes, multilevel logistic (dichotomous variables) or linear regression (continuous variables) will be used, as appropriate. In addition, adverse health outcomes (e.g. complications in AA cases and non-AA cases, missed other emergency diseases) will be described without statistical evaluation.

COST-EFFECTIVENESS ANALYSIS Cost-effectiveness will be calculated from societal and healthcare perspective with a time horizon of 30 days. Secondary analyses will address a 3-month time horizon. Efficiency will be the primary effect parameter. Cost-utility will be calculated based on the EQ-5D-(Y)-3L. Bootstrap re-sampling will be performed on the costs, and on costs and effect pairs (cost-effectiveness and cost-utility analyses) in order to calculate confidence intervals. Furthermore, both cost-effectiveness and cost-utility planes will be plotted, along with acceptability curves.