The review concluded that there was clear evidence that physically abused children attending accident and emergency (A&E) departments were missed. There was no evidence that any test was highly predictive of physical abuse in children attending A&E. The limited quality of the included studies and wide range of reported sensitivity and false-positive rates mean that these conclusions should be interpreted cautiously.

This abstract addresses only the research question that met DARE criteria.

To assess the accuracy of screening tests for physical abuse in children attending accident and emergency (A&E) departments.

MEDLINE and the publications catalogue of National Society for the Prevention of Cruelty to Children (NSPCC) were searched for published studies from 1975 to September/October 2005. PubMed, Web of Science, Department of Health database, British Nursing Index and Cumulative Index to Nursing and Allied Health (CINAHL) were searched up to January 2006. Search terms were reported. Reference lists of relevant publications were searched. Paediatrics, Child Abuse and Neglect and Child Abuse Review journals were handsearched in October 2006. Study authors were contacted for additional studies.

Studies that evaluated the performance of any type of clinical assessment for the detection of physical abuse in injured children attending hospital were eligible for inclusion. Eligible studies had to report the results of any type of reference standard for confirming or excluding physical abuse. Studies in developing countries were excluded.

The included studies evaluated various types of clinical assessments, including standard care (clinical screening assessment based on the initial history and examination by a triage nurse and/or A&E doctor) and the addition of a checklist and full undressed examination of the child. The reference standard varied between included studies, including the risk score for abuse based on retrospective examination of records or inadequate documentation to exclude abuse, confirmation of abuse by social services or police and confirmation or exclusion of abuse by the child protection team. Where reported, included children were less than one to nine years old. The included children had different conditions, such as a fracture in A&E or an orthopaedic clinic, admitted with a head injury and those who had an injury or poisoning in A&E.

Two reviewers assessed studies for inclusion.

The quality of included studies was assessed using the QUADAS tool to assess representative patient spectrum, acceptable reference standard, delay between tests, partial and differential verification bias, incorporation bias, blinded interpretation of the index test and reference standard, clinical review bias, reporting of uninterpretable results and explanation of withdrawals from the study. Items were rated as yes, unclear or no.

The authors did not state how many reviewers performed validity assessment.

Data were extracted on the numbers of true positive, false negative, false positive and true negative results, together with the corresponding sensitivity and false positive rates.

The authors did not state how many reviewers performed data extraction.

The studies were combined in a narrative synthesis supported by accompanying data tables.

Three studies that reported both sensitivity and false-positive rates were included in the review (n=2,949). Major methodological flaws for all studies included a lack of independence and blinding between clinical assessment and reference standard, and use of a different reference standard in children referred and those not referred. Full details of quality assessment results were reported in the appendix.

Two studies reported that clinical screening assessment based on the initial history and examination by a triage nurse and/or A&E doctor had sensitivity of 36% to 67% and false positive rates of 1% to 41%. One study reported that clinical screening assessment based on a checklist and full undressed examination of the child had sensitivity of 89% and specificity of 1%.

For studies that reported only sensitivity of clinical assessment, sensitivity ranged from 60% to 100% (four studies).

For studies reported only the false-positive rates of clinical assessments, false-positive rates ranged from 3.2% to 8.8% (four studies) when applying a standard threshold for confirming abuse and from 21.1% to 67.5% (two studies) when applying a high threshold for confirming abuse.

There was clear evidence that physically abused children attending A&E were missed, but the performance of the clinical screening assessment was poorly quantified. There was no evidence that any test was highly predictive of physical abuse in children attending A&E.

The review addressed a clearly stated research question, defined by appropriate inclusion criteria. Relevant databases were searched. Efforts were made to find both published and unpublished studies, which minimised the risk of publication bias. It was unclear whether language restrictions were applied in the search, which made it difficult to assess the risk of language bias. Steps were undertaken to minimise the reviewer errors and bias in study selection. It was unclear whether the processes of data extraction and validity assessment were performed in duplicate to minimise the reviewer errors and bias. Appropriate criteria were used to assess study quality, but a formal validity assessment was only performed in the review of studies that reported both sensitivity and false-positive rates. Given the diversity of included studies, a narrative synthesis was appropriately employed.

A degree of caution is required in interpreting the authors' conclusions, given the limited quality of the included studies and the wide range of reported sensitivity and false-positive rates.

Practice: The authors stated that standardised recording of the diagnosis or suspicion of abuse or neglect and whether or not children are referred to social services should improve potential for the detection of children at risk of abuse.

Research: The authors stated that well-designed large-scale studies were required to assess the effectiveness of assessments used in A&E for the identification of abused children. Future research is required to investigate the role and effectiveness of the community liaison nurse. Future studies should evaluate the feasibility and effectiveness of the following new tests: direct questioning of school-age children about injuries; assessment of bruising on the head and face; timing of attendance at A&E; assessment of information from the cumulative record of health care use; and assessment of the information from agencies outside health.

Health Technology Assessment Programme.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.