The flow diagram is presented in Figure 1. A total of 35 diagnostic studies were identified, assessed and interpreted. The characteristics of these studies are presented in Table 3. 20 studies were performed in an emergency department, four studies in a traumatology setting and three other studies in a radiology department. The patients in the studies by Mallee et al. [54, 58, 60] were derived from one prospective study; therefore the setting was the same for each study: patients were initially seen by the emergency physicians and in follow-up by the orthopaedic department and/or trauma surgery department, depending on who was on call. In five studies the setting was not specified. To our knowledge, all first authors of those five studies were working in a hospital care setting, so we assume all to have been done in hospital care. History taking, physical examination and imaging as index tests were investigated in 0%, 20% (7/35) [43, 44, 46, 47, 67 71, 74] and 86% (30/35) [47–73, 75–77] of the studies, respectively.
Quality Assessment
There was considerable underreporting of important quality domains in 23 of the 35 studies (see Table 4). In 13 of the 35 studies [43, 44, 51, 53, 56, 57, 59, 61, 62, 70, 71, 73, 74], patient selection was not well documented. Furthermore, the risk of bias was predominantly due to the absence of a proper description of the index test (9/35) [43, 47, 49, 51, 55, 57, 61, 65, 71] or the reference standard (13/35) [43, 46, 49, 51, 52, 55, 57, 65, 67, 69, 72, 74, 75]. Twelve of the studies (34%) demonstrated no limitations when risk of bias was assessed, according to QUADAS–2 [45, 48, 50, 54, 58, 60, 63, 64, 66, 68, 76, 77]. Eight showed incorporation bias [46, 48–52, 68, 76].
Diagnosing Carpal Fractures in Hospital Care
Table 5 presents the accuracy of the diagnostic tests of all the carpal fractures. One study described the diagnostic accuracy of history taking by asking for patients’ pain score, using the Visual Analogue Scale method with anatomical snuffbox tenderness [74]. Physical examination [43, 44, 46, 47] and combined physical and radiological examination [45] for diagnosing scaphoid fractures showed Se, Sp, accuracy, PPV and NPV ranging from 15–100%, 13–98%, 55–73%, 14–73% and 75–100%, respectively.
Repeated physical examination with radiological examination after 38 days [45]for diagnosing other carpal bone fractures showed a Se of 100% with the exception of the triquetrum (75%).
Radiographs used as an index test for diagnosing scaphoid fractures showed Se, Sp, accuracy, PPV and NPV ranging from 25–87%, 50–100%, 48–88%, 14–100% and 49–94%, respectively. For diagnosing scaphoid fractures, Magnetic Resonance Imaging (MRI) as an imaging modality showed Se, Sp, accuracy, PPV and NPV ranging from 67–100%, 89–100%, 85–100%, 54–100% and 93–100%, respectively. Multi Detector Computed Tomography (MDCT) showed Se, Sp, accuracy, PPV and NPV ranging from 33–100%, 85–100%, 79–100%, 28–100% and 86–100%, respectively. Bone Scintigraphy (BS) as an index test for diagnosing scaphoid fractures showed Se, Sp, accuracy, PPV and NPV ranging from 78–100%, 87–97%, 86–97%, 62–78% and 90–100%, respectively. For diagnosing scaphoid fractures, Ultrasonography (US) as an imaging modality showed Se, Sp, accuracy, PPV and NPV ranging from 78–100%, 34–100%, 49–100%, 30–100% and 75–100%, respectively.
Diagnosing Phalangeal and Metacarpal Fractures in Hospital Care
Table 5 also presents the accuracy of the diagnostic tests for metacarpal and/or phalangeal fractures, as described in six studies [67, 69–73]. Physical examination [71] for diagnosing phalangeal and metacarpal fractures showed Se, Sp, accuracy, PPV and NPV ranging from 26–55%, 13–89%, 45–76%, 41–77% and 63–75%, respectively. Imaging for metacarpal and finger fractures showed Se, Sp, accuracy, PPV and NPV ranging from 73–100%, 78–100%, 70–100%, 79–100% and 70–100%, respectively. The reported diagnostic accuracy measures of phalangeal and metacarpal fractures were characterized by markedly heterogeneous results among the eligible studies.
Combined Diagnostic Accuracy of the Studies with No Limitations and No Incorporation Bias
Table 6 shows combined diagnostic accuracy measures of the studies that had no limitations and no incorporation bias. A wide range of results were found for the specificity, accuracy and NPV of MRI, US, CT and BS. The sensitivity of BS and US showed similar, acceptable results. US and MRI are imaging tools that have similar PPV, but with large confidence intervals.