Summary of main results
Our review found two trials that investigated the effect of chest radiographs in lower respiratory tract infections (LRTIs); one in adults (Bushyhead 1983) and one in children (Swingler 1998). Severely ill patients were excluded from both randomised controlled trials (RCTs). Both trials came to the conclusion that making chest radiograph results available to clinicians resulted in similar outcomes for patients with acute LRTIs compared to when no chest radiograph results were available.
For cases in adults, Bushyhead 1983 states that the differences in outcomes between the group with chest radiographs and the group without were not statistically significant. This suggests that chest radiographs did not result in significant changes in management plans or differences in patient outcomes. However, chest radiographs appear to be of benefit in the subgroup of the participant population with an infiltrate on their radiograph. In this subgroup of patients whose chest radiographs showed infiltrative abnormalities, the "use of the chest radiograph was associated with [a] reduction in the length of illness, duration of cough, and duration of sputum production (P < 0.05)". This effect was not evident in patients whose chest radiographs showed non-infiltrate radiographic abnormalities. It was interesting to note that although patients with non-infiltrate radiographic abnormalities often had additional return visits, diagnostic tests and changes in treatment, these changes were rarely effective, i.e. rarely led to improved patient outcomes. Bushyhead 1983 reported that "chest radiographs were not ordered efficiently by physicians". It is also interesting to note that only 54% of patients with infiltrates on chest radiographs would have had chest radiographs ordered by the treating physician under normal practice conditions. None of the other 46% of patients would have been suspected to have pneumonia despite definite radiographic infiltrates. Nonetheless, this finding has limited practical implications as, to date, the simplest and most accessible way of knowing whether patients will have pulmonary infiltrates is to perform a chest radiograph.
For cases in children, Swingler 1998 states that "there are no clinically identifiable subgroups of children within the WHO case definition of pneumonia who are likely to benefit from a chest radiograph." Swingler 1998 summarised that "the most favourable 95% CI for the estimate of benefit of a chest radiograph is the prevention of 3 days of relatively trivial symptoms, while the least favourable is the cause of an additional 2 days of symptoms". It was thought-provoking to note that at 28 days, every 11 chest radiographs performed in children would result in one more antibiotic script (95% CI -5.8 to 61.7). Additionally, every 42 chest radiographs performed at 28 days would lead to one additional hospital admission (95% CI -18 to 127). Swingler 1998 concluded that "the use of chest radiographs did not reduce time to recovery or subsequent health-facility use in children over two months with ambulatory acute lower-respiratory [tract] infection" and hence recommended that "chest radiographs should not be routinely done in this group of patients" (Figure 4).
According to Swingler 1998, the effect of chest radiographs was independent of both the severity of the respiratory tract infection and the clinician's clinical experience. The patient's "age, weight for age, duration of symptoms before presentation, respiratory rate, or the clinicians' perception of the need for a radiograph" did not influence the effect of the chest radiograph. This effect is also not influenced by the clinicians' qualifications or experience, i.e. whether they were recently qualified doctors with no previous paediatric outpatient experience or had a postgraduate paediatric qualification.
With regard to hospitalisation rates, Swingler 1998 reported higher hospitalisation rates for patients randomised to chest radiographs but this result was not statistically significant (Figure 4). Unfortunately, Bushyhead 1983 did not analyze hospitalisation rates for patients whose radiographs were provided versus those for whom radiographs were not provided, which is one of the objectives of our review. Instead Bushyhead 1983 reported hospitalisation rates in the group of patients whose radiographs were provided to their physicians and compared whether the wishes of the physician to view the radiograph or not had an impact on the patients' hospitalisation rates.
Both RCTs set strict exclusion criteria that excluded patients with suspected severe disease, either based on variations in patients' vital signs, other clinical signs of severe disease such as localised wheeze or when the clinicians' assessment deemed chest radiographs mandatory. Strict exclusion criteria are important but may limit the clinical practicability of the results of the trials as both study populations were well-filtered sample populations which may not reflect those presenting in clinical practice.
There were no data available from the included studies to assess the impact on mortality, complications of infection and adverse effects from chest radiographs. Swingler 1998 briefly comments on the potential drawbacks of ordering chest radiographs. However, no data regarding this were collected in either of the included studies. Listed disadvantages of chest radiographs in Swingler 1998 include the "exposure of ionising radiation, cost (especially if travel to another facility is necessary), the time and space used waiting for the radiograph and the need to be seen again by a clinician". Theoretical long-term complications of radiation exposure from chest radiographs, such as risk of malignancy later in life, were not discussed. Statistics show that the radiation exposure from the use of chest radiographs is extremely low; a chest radiograph in two views is associated with an effective dose of ionising radiation of 0.06 to 0.25 mSV (Diederich 2000). This is in comparison to the average background radiation dose of around 2.4 mSV per year (WNA 2011). This means that at its most, one chest radiograph would be equivalent to approximately one month of background radiation, i.e. there is essentially a negligible risk of malignancy in the long term with a single chest radiograph.
Another feature of chest radiographs that is worth mentioning is the discovery of incidental findings, which may be advantageous or disadvantageous. For example, in Bushyhead 1983 there were 17 intrathoracic masses detected, six of which were absent on follow-up radiographs and the remainder proving to be acute infiltrate, pericardial cyst, hiatus hernia and granulomas. However, one mass proved to be lung cancer in a 74-year old man. Chest radiographs may at times find non-specific nodules which more likely than not are benign but nonetheless require follow-up (Gould 2007). This may mean further investigations such as biopsies to rule out malignancy - these additional tests each have their own associated morbidity. In cases of malignancy found on chest radiographs requested for clinically suspected pneumonia, the malignancy may be the primary cause of the pneumonia (e.g. the tumour causes bronchial obstruction and pneumonia may be the complication of the tumour).
Overall completeness and applicability of evidence
Both of the included studies provided some relevant evidence for the objectives of this review. However, our review included objectives which were not objectives of the included trials and, therefore, there were no results regarding this. One of our primary outcomes was to assess mortality and this was not assessed by either of the studies. Our other primary outcome was to investigate the time to resolution of clinical signs and symptoms. However, given that the reported data are incomplete, results could not be pooled together in a meta-analysis. In addition, our secondary outcome of assessing hospitalisation rates was only reported by Swingler 1998. Other secondary outcomes, such as complications of infection and adverse effects from chest radiographs, were not assessed by either study. This limits the amount of evidence we could analyze. Furthermore, as the participant population in Bushyhead 1983 were adults whilst the participants in Swingler 1998 were children, extrapolation of these results to come to a general conclusion may be inappropriate and of uncertain validity.
Quality of the evidence
Despite the large number of participants (2024), given that the reported data were incomplete, there is limited evidence to formulate robust conclusions regarding the objectives of our review. We assessed both RCTs as having a high risk of bias with regards to selection bias and likely imprecision of results and, hence, we downgraded them to 'low' based on the GRADE working group grades of evidence (Higgins 2011). Although the results of the two studies could not be pooled into a meta-analysis, both come to the same conclusion regarding the use of chest radiographs in acute LRTIs, except in the subset of patients with infiltrates on their radiograph.
Potential biases in the review process
Identification of all relevant studies
We used a broad search strategy to ensure that we could ascertain all relevant trials which met our criteria. It is possible that we may have missed studies. However, contacting experts in the field did not yield any other relevant references and therefore we think it is unlikely that studies were missed.
Introduction of bias
We independently appraised the studies for inclusion as well as for risk of bias and data extraction in order to minimise the risk of bias in the review process.
Agreements and disagreements with other studies or reviews
We are not aware of any other RCTs or systematic reviews available to compare results with.