1. Background
Pneumonia is a common illness all around the world and is considered as a major cause of death among all age groups (1). In Iran, pneumonia is the most common cause of hospitalization in the elderly and patients with primary antibody deficiencies (2, 3). Nonetheless, only 5% and 28% of patients with acute respiratory symptoms who are referred respectively to primary care physicians (4) and emergency departments (EDs) (5) develop pneumonia. According to most clinical guidelines, the gold standard tool for diagnosing pneumonia is a chest X-ray (CXR), which can distinguish this condition from other respiratory tract infections (6). However, many clinicians might only rely on the patient’s medical history and physical examination findings to diagnose or exclude pneumonia.
Community-acquired pneumonia (CAP) is the most common type of pneumonia. Despite frequent use of X-ray in clinical settings, there are still debates regarding its necessity in patients suspected of having CAP. In this regard, Ayalon et al. investigated the role of physical examination in the diagnosis of pneumonia and demonstrated the significant benefit of CXR as an ancillary test for the diagnosis of pneumonia in ED settings (7). In addition, according to a review published in 2003, CXR was considered as the gold standard for diagnosing pneumonia. Researchers doubted that the decision to perform CXR should solely rely on experts' opinions, which are based on strategies for optimizing the balance between harms and benefits. Consequently, they recommended a combination of the patient’s medical history, physical examination, and laboratory data to estimate short-term risks and to determine the appropriate treatment (8).
2. Objectives
Although several investigators have attempted to provide prediction rules to improve the detection of CAP in outpatient settings during the past three decades, there have been contradictory conclusions in this regard (4, 9-17). Accordingly, the aim of the present study was to evaluate the predictive value of the eight most common clinical parameters and three most frequent laboratory variables in patients presenting with acute respiratory symptoms. Therefore, the subjects were divided into two large groups of 420 to determine the effect of clinical and laboratory data for X-ray-based diagnosis. The groups were as follows: the case group included those with positive CXR findings and the control group consisted of subjects with insignificant CXR findings.
3. Patients and Methods
In this case-control study, all adult patients (18 years of age or older) with acute respiratory symptoms (one or more of the following symptoms: cough, sputum, dyspnea, and chest pain) and positive findings on CXR were included as positive cases (n = 420). The subjects had been referred to outpatient clinics or EDs of medical university hospitals in Birjand, Iran, from December 2008 to December 2009. Positive findings on CXR were defined as the presence of new consolidations without an air bronchogram, pleural effusion, abscess or empyema in the setting of acute respiratory symptoms. Patients with subtle airspace opacities (equivocal findings) were also considered as positive cases. An equal number of cases (n = 420) with acute respiratory symptoms but insignificant findings on CXR were recruited as the control group; the subjects were matched in terms of age and the date of the CXR procedure. Those with findings suggestive of hospital-acquired pneumonia (positive findings on CXR within ten days of discharge), aspiration pneumonia, and those who had recently received antibiotics or steroids were excluded.
An infectious disease specialist examined all patients and the data regarding clinical symptoms and signs were recorded. Afterwards, CXRs were performed and the clinician was blinded to the radiographic results. A board-certified radiologist, unaware of the clinical findings, interpreted all CXRs. The clinical indicators included cough, sputum production, dyspnea, chest pain, fever (body temperature≥ 38℃), tachycardia (heart rate≥ 100 beats per minute), tachypnea (respiratory rate≥ 20 breaths per minute), and abnormal findings on pulmonary auscultation and examination (crackles, decreased breath sounds, dullness on percussion, and rhonchi). The erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) levels, and white blood cell (WBC) count were measured for all patients. Abnormal values of these laboratory tests were defined as follows: ESR (mm/h) higher than the males’ age divided by two or females’ age plus ten divided by two (16), CRP > 5 mg/L (47.62 mmol/L), and WBC > 10000/µL or < 4000/µL.
The sensitivity and specificity of all clinical indicators were calculated using the IBM SPSS 19 software for Windows (IBM Corp., IBM SPSS Statistics for Windows, Armonk, NY: IBM Corp.); odds ratios (OR) were also calculated for each of the indicators. Additional indicators were defined as “any abnormal findings in vital sign assessment” (positive test results = fever, tachycardia, and/or tachypnea), “any abnormal findings in vital sign assessment or physical examination” (positive test results = fever, tachycardia, tachypnea, and/or abnormal findings on pulmonary examination) and “any abnormal findings in laboratory test” (positive test results = abnormal values of ESR, CRP, or WBC). According to the analyses, we attempted to provide a prediction rule to determine which patients would benefit from CXR. The diagnostic accuracy of this rule was determined regarding the area under the receiver operating characteristic (ROC) curve.
4. Results
During the 12-month study period, 420 patients with acute respiratory symptoms and positive findings on CXR were eligible for participating in the study. An equal number of individuals were selected as the control group with similar demographic characteristics as the case group (Table 1). The analyzed data are presented in Table 2. The clinical symptoms and signs were compared, and as listed in Table 2, there was a significant difference between the two groups regarding clinical indicators including fever, tachycardia and tachypnea (abnormal vital signs). Similarly, the difference between the two groups in terms of pulmonary auscultation according to physical examination was statistically significant. With regards to the aims of this study, the sensitivity and specificity (95% confidence interval) of the statistically significant variables were calculated (Table 2). The clinical indicator with the greatest sensitivity for pneumonia (diagnosed by CXR) was cough, with a sensitivity of 89.5%.
The presence of “any abnormal findings in vital sign assessment”, “any abnormal findings in vital sign assessment or physical examination”, and “any abnormal findings in laboratory test” was studied, and these indicators were compared according to the recorded data (Table 2). All three of these new “combined indicators” exhibited higher prevalence in patients with CAP (positive findings on CXR in the setting of acute respiratory symptoms) in comparison with the controls. However, only for one parameter (any abnormal findings in vital sign assessment or physical examination), the sensitivity (94.0%), specificity (57.4%), and OR (23.1) were optimal for being used as a screening parameter. This combined prediction rule had an area under the ROC of 0.751, which suggested that it was a relatively effective rule with 75% accuracy and 94% sensitivity. Table 2 lists the sensitivity and specificity of other clinical, laboratory, and “combined” indicators. Elevated CRP alone had a specificity of 94.6%. The sensitivity of elevated ESR was 28.3%, whereas its specificity was 86.9%.
| Variables | Case Group (No. = 420) | Control Group (No. = 420) |
|---|---|---|
| Mean Age, y | 60 | 63 |
| Age Range, y | 18-87 | 19-86 |
| Gender | ||
| Male | 220 (52.4) | 227 (54.0) |
| Female | 200 (47.6) | 193 (46.0) |
| Underlying Lung Disease c | 50 (11.9) | 28 (6.7) |
a Data are presented as No. (%) or Mean.
bNone of these variables showed a significant difference between case and control groups, as indicated by the independent-samples t test and Chi square test (P > 0.05).
cUnderlying lung disease is defined as chronic obstructive disease, interstitial lung disease, or cancer.
Table 2. Analysis of Clinical and Laboratory Indicators in Patients with Acute Respiratory Symptoms a
| Indicators | Cases (n = 420) b | Controls (n = 420) b | Sensitivity c (95% CI) | Specificity c (95% CI) | Odds Ratio (95% CI) |
|---|---|---|---|---|---|
| Cough | 376 (89.5) | 272 (64.8) | 89.5 (86.7-91.5) | 35.5 (30.4-40.8) | 4.6 (2.9-6.7) |
| Sputum | 354 (84.3) | 168 (40.0) | 84.3 (80.6-87.0) | 60.0 (55.4-64.7) | 8.0 (5.6-10.4) |
| Dyspnea | 323 (76.9) | 180 (42.9) | 76.9 (72.6-80.4) | 57.1 (60.5-70.2) | 4.4 (2.7-6.5) |
| Chest pain | 315 (75.0) | 201 (47.9) | 75.0 (71.1-79.0) | 52.1 (46.3-57.3) | 3.5 (2.4-5.1) |
| Fever d | 285 (67.9) | 68 (16.2) | 67.9 (62.9-73.2) | 83.8 (80.1-86.5) | 7.0 (4.8-10.2) |
| Tachycardia d | 231 (55.0) | 63 (15.0) | 55.0 (49.6-59.4) | 85.0 (81.0-88.4) | 6.9 (4.7-10.1) |
| Tachypnead | 252 (60.0) | 44 (10.5) | 60.0 (55.3-64.9) | 89.5 (86.5-91.5) | 12.8 (7.6-17.4) |
| Abnormal Auscultation Findings e | 370 (88.1) | 133 (31.7) | 88.1 (85.1-90.3) | 68.3 (63.3-73.6) | 16.0 (11.2-20.8) |
| Elevated ESR | 119 (28.3) | 55 (13.1) | 28.3 (22.3-33.6) | 86.9 (82.8-90.3) | 7.2 (5.1-10.3) |
| Elevated CRP | 168 (40.0) | 22 (5.2) | 40.0 (34.7-45.4) | 94.6 (91.8-96.7) | 12.7 (7.5-17.3) |
| Abnormal WBC | 204 (48.6) | 94 (22.4) | 48.6 (42.7-54.0) | 77.6 (73.4-81.1) | 3.3 (2.2-4.9) |
| Any Abnormal Findings in VS | 361 (86.0) | 118 (28.1) | 86.0 (81.9-89.4) | 77.9 (73.7 81.5) | 9.5 (7.2-12.8) |
| Any Abnormal Findings in VS or PE | 395 (94.0) | 179 (42.6) | 94.0 (91.1-96.3) | 57.4 (60.8-70.5) | 23.1 (12.7-43.1) |
| Any Abnormal Findings in LAB | 254 (60.5) | 110 (26.2) | 60.5 (55.2-65.5) | 73.8 (69.8-77.7) | 6.8 (4.5-10.0) |
aAbbreviations: CI, confidence interval; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; LAB, laboratory tests (ESR, CRP, or WBC); PE, physical examination (pulmonary auscultation); and VS, vital sign (body temperature, heart rate, and respiratory rate).
b Data are presented as No. (%).
c Data are presented as (%).
dBody temperature ≥ 38℃; heart rate ≥ 100 beats per minute; respiratory rate ≥ 20 breaths per minute.
ePresence of crackles, decreased breath sounds, dullness on percussion, or rhonchi.
5. Discussion
The results of the present study revealed that vital signs and physical examinations are reliable parameters for predicting chest radiograph findings in outpatient settings; thus, it is possible to determine which patients would benefit from CXR. Performing CXR, only in patients with acute respiratory symptoms and abnormal findings in vital signs assessment and physical examinations, enabled us to correctly identify 94% of patients with positive findings on CXR (positive cases) and avoided unnecessary CXRs in 57% of the patients with insignificant findings on CXR (controls); this indicates 94% sensitivity and 57% specificity of the prediction rule.
The suggested rule indicates that performing CXRs would be unnecessary in patients with acute respiratory symptoms (cough, sputum, dyspnea or chest pain), who present normal findings in vital signs assessment (body temperature, heart rate, and respiratory rate) and physical examination (lung auscultation). Our findings are consistent with those of Gennis et al. (11) and O’Brien et al. (12), who provided similar guidelines and prediction rules for ordering CXR, using clinical criteria for the detection of pneumonia in adults. However, our results differed from the findings of Singal et al. (10) and Butcher et al. (18), who demonstrated the inefficiency of clinical findings as screening parameters, and Graffelman et al. (13), who concluded that models based only on clinical findings do not reliably predict the presence of pneumonia.
According to the guidelines of the American Thoracic Society (ATS) for the management of CAP (2007), a CXR is required for the routine evaluation of patients who are likely to have pneumonia, in order to establish a diagnosis and help differentiate CAP from other common causes of cough and fever such as acute bronchitis (19). In Iran, there are no national guidelines for CAP management. Most of the clinicians in Iran (and in other countries such as the United States) (10) do not follow the aforementioned guidelines and mainly rely on their personal experiences. They often do not order a CXR for patients with suspected CAP, unless the vital signs and/or physical examination findings are abnormal. Our findings supported these practices by providing objective data.
The British Thoracic Society (BTS) guidelines for the management of CAP in adults (updated in 2009) provide recommendations, which are more similar to the present prediction rule than to the ATS guidelines. According to BTS guidelines, all patients who are admitted to hospitals with signs and symptoms suggestive of CAP should undergo a CXR as soon as possible to confirm or refute the diagnosis. However, it is stated that in non-emergent outpatient settings, it is not necessary to perform CXR in patients with signs and symptoms suggestive of CAP, unless the following conditions are present:
- The diagnosis is doubted and a CXR can help with the differential diagnosis and the management of the acute illness.
- The progress following the treatment for suspected CAP is not satisfactory at follow-up.
- The patient is at risk of underlying lung pathologies such as lung cancer (6).
By implementing this prediction rule in our study, it was shown than CXR findings were insignificant in 57% of the patients with acute respiratory symptoms although the findings of vital signs or physical examinations were normal. The implementation of this prediction rule can significantly decrease the costs associated with these unnecessary radiographs. In fact, its application can save time, increase clinicians’ efficiency, and decrease radiation exposure to the population.
Approximately 6% of the cases would be dismissed by following this prediction rule; for instance, patients with normal findings in vital signs assessment or physical examinations who have positive results for CXR. Although this might be acceptable, it is only reasonable to implement the prediction rule for patients with reliable follow-ups and a low likelihood of morbidity if CAP is not initially detected.
There are some limitations and shortcomings in our study. The first limitation was that performing radiographs for the study population was determined by the clinicians’ judgment (selection bias); thus, our results would best reflect the current medical practices, which normally start by a visit from general practitioners. The second limitation was interpreting equivocal CXR findings as positive; however, this classification minimized the false negative results. Finally, the inevitable delay between the onset of symptoms and their appearance on CXR was not explored in this study and the insignificant findings on CXR could be partly related to this issue. The prediction rule of our study must be independently validated and until then, the results should be interpreted with caution.
Our findings suggested that it might not be necessary to perform CXR in patients with acute respiratory symptoms unless the vital signs and/or physical examination findings are abnormal. However, since the sensitivity of this rule is 94% for predicting CAP, a CXR is required for patients with unreliable follow-ups or moderate to high likelihood of morbidity if CAP is not initially detected.
