Point-of-care urinary pneumococcal antigen test in the emergency department for community acquired pneumonia
- 1Department of Microbiology and Infectious Diseases, Concord Repatriation General Hospital, Sydney, Australia
- 2Emergency Department, Concord Repatriation General Hospital, Sydney, Australia
- Dr R Paoloni, Emergency Department, Concord Repatriation General Hospital, Concord, Sydney, NSW 2139, Australia;
- Accepted 28 September 2007
Background: Streptococcus pneumoniae is the most common cause of community-acquired pneumonia (CAP). Early diagnosis would allow more directed therapy and confidence in appropriate treatment for a majority of patients. The BinaxNOW pneumococcal urinary antigen (PNAG) test has been evaluated at laboratory level and is easy to perform and interpret, but its use as a point-of-care test has not been evaluated. A study was undertaken to assess whether PNAG testing can be reliably performed and interpreted by staff in an adult emergency department and whether rapid results influence initial treatment decisions.
Methods: Community-living adult patients presenting to the emergency department with clinical and radiological findings of pneumonia had PNAG testing performed on the same sample in both the emergency department and the microbiology laboratory in a blinded fashion. Accuracy and turnaround time were assessed. Diagnostic yield was compared with routine culture methods.
Results: Fifty-nine patients were enrolled of whom nine (15%) had positive PNAG tests. These included three culture-proven cases and six additional cases. There was 98% concordance between emergency department and laboratory results. Turnaround time was significantly shorter when tested in the emergency department (median 2 h 39 min vs 19 h 40 min). Antibiotic prescribing was not influenced by results in this small sample.
Conclusions: PNAG diagnosis of pneumococcal pneumonia can be accurately performed as a point-of-care test by emergency department clinical staff. Without specific efforts to achieve early urine collection, the timeframe of testing will frequently fall outside the 4-hour patient stay of a UK emergency department and may be more appropriately considered as a test for the medical admissions unit in this setting. Sensitivity is at least equal to conventional culture methods and the result is available rapidly enough to potentially influence treatment decisions, a strategy that warrants further investigation.
Streptococcus pneumoniae remains the most common pathogen in community-acquired pneumonia (CAP). Despite increasing antimicrobial resistance in many human pathogens, benzylpenicillin or amoxicillin orally continue to be regarded as agents of choice for pneumococcal pneumonia, as reflected in current British Thoracic Society (BTS) guidelines on community acquired pneumonia and Australian Therapeutic Guidelines for antibiotics.1–3 Under the BTS guidelines, the ability to confirm pneumococcal aetiology for patients requiring hospital admission would allow monotherapy rather than combination therapy to be given. This makes treatment less costly and may reduce the emergence of antibiotic resistance due to exposure to unnecessarily broad-spectrum antimicrobials.
Confirming the diagnosis of pneumococcal pneumonia traditionally relies on isolation from blood or sputum cultures which require 24–48 h. Isolation from blood is specific for the diagnosis, but bacteraemia occurs in only 10–25% of cases.4 Sputum diagnosis suffers from difficulties obtaining adequate and timely specimens and lack of sensitivity. A positive Gram stain or culture can be expected in at best 33% and 44% of cases, respectively.5
Detection of the C-polysaccharide cell wall antigen has 90–98% specificity for S pneumoniae infection in adults, with sensitivity exceeding Gram stain and culture.6–8 The BinaxNOW Streptococcus pneumoniae immunochromatographic test (ICT) (Binax Inc, Portland, Maine, USA) is a simple commercially available test performed on unprepared urine with results in 15 min. It thus lends itself to use as a point-of-care test.
We sought to determine whether the BinaxNOW ICT could be reliably performed and interpreted in the emergency department (ED) by clinical staff rather than sent to a laboratory for evaluation. We also sought to quantify advantages in relation to time to diagnosis and initiation of specific therapy gained by availability as a “bedside” test.
A prospective blinded cross-sectional analytical (diagnostic test) study was performed. Ethics approval was obtained from the Institutional Review Board (Concord Hospital Ethics Committee) with the requirement for informed consent of study subjects being waived.
Study setting and population selection
The study was undertaken in the ED at Concord Hospital, a 373-bed tertiary teaching hospital, in Sydney, Australia from May to November 2004. The ED has 25 000 presentations per year, resulting in 8000 inpatient admissions. Patients over the age of 14 with a clinical diagnosis of pneumonia and a radiological infiltrate, as determined by the treating clinician, were eligible. Exclusion criteria were neutropenia (<1.0×109/l) and discharge from hospital in the preceding 7 days. Patients already taking antibiotics were included. Consecutive patients were recruited by ED medical staff.
Before pneumococcal antigen (PNAG) testing, medical staff recorded the length of illness and previous treatment, Pneumonia Severity Index and drug allergies.9 They were asked to nominate their choice of antibiotic therapy before the test was performed. No direction in relation to choice of antibiotic therapy was given, but Australian Therapeutic Guidelines were available in written and electronic form.
Urine was collected and the ICT was performed in the ED by nursing staff who had received training in performing the test. Briefly, a swab moistened with neat patient urine is inserted into the ICT card and drops of reagent applied; the reaction is left for 15 min. The appearance of a precipitant line indicates a positive result. Each card has an internal control. Each kit of 20 cards had both positive and negative controls run. The nurse then informed the treating doctor of the result.
Medical staff were informed that positive results were likely to indicate pneumococcal infection reliably but that negative tests did not rule out the diagnosis. The treating clinician recorded the time of the test result, antibiotic regimen prescribed and whether the patient was admitted or discharged. All medication choices and other diagnostic tests were at the discretion of treating doctors in the ED, but may have been influenced by the admitting physician (predominantly thoracic or aged-care physicians). These groups were also educated in relation to test characteristics.
The urine sample was forwarded to the microbiology laboratory where laboratory scientists blinded to the ED result repeated the test. Each urine specimen was also tested for Legionella pneumophila type 1 antigen (BinaxNOW Legionella pneumophila ICT; Binax Inc). The microbiology laboratory is staffed 10 h daily but assays were not performed on weekend days.
Measurements and data analysis
ED and microbiology laboratory test results were recorded, as were the results of blood and/or sputum cultures. The primary analysis was performance of the test (ED performed testing) against the gold standard (laboratory performed testing). Given the simplicity of the test (both in terms of performance and interpretation), and that the ED nursing staff who were to perform the test had already been performing and interpreting urine pregnancy tests based on a very similar platform, it was estimated that their sensitivity and specificity against the gold standard (laboratory use of the same test) would be approximately 95%. To achieve 95% confidence intervals for both sensitivity and specificity whose lower limit was 85%, the calculated sample size was 100 patients (estimated prevalence of 20%).
Time-to-test intervals were calculated starting from the later of the following two times: (1) time seen by ED medical officer or (2) time of performance of chest radiography, as both clinical and radiographic information was necessary to fulfil inclusion criteria. The time delays to laboratory testing were also analysed in the subgroup of patients who presented during business hours. Non-parametric tests of association (Fisher exact test, Wilcoxon two-sample test) were used owing to the small dataset and likely non-normal distribution of variables. Data were analysed using SAS statistical software (SAS Institute, Cary, North Carolina, USA).
Characteristics of study subjects
Fifty-nine patients satisfied the study criteria over the 6-month period, 33 of whom (56%) were male. The median patient age was 79 years. The mean duration of symptoms before presentation was 4 days, and 15 patients (26%) had received prior antibiotics. The median length of stay was 6 days.
The mean Pneumonia Severity Index was 92 (median 85.5). Seventeen patients (29%) were classified as class I or II, 18 (31%) as class III, 15 (25%) as class IV and 9 (15%) as class V. Only 6 recruited patients were discharged for outpatient therapy.
Adequate sputum specimens were obtained from 19 patients (32%) with a predominant or pure growth of a respiratory pathogen obtained from 7 (12%): Haemophilus influenzae in 4, S pneumoniae in 2 and both H influenzae + S pneumoniae in 1 patient. In addition, bronchoalveolar lavage yielded Mycobacterium tuberculosis in 1 patient. Blood cultures were collected from 41 patients and yielded a pathogen in 3 cases: Streptococcus milleri in 1 patient with a dental abscess and aspiration pneumonia, Streptococcus pyogenes in 1, and Group B streptococcus in a patient with concomitant cellulitis. No cases of pneumococcal pneumonia were bacteraemic. Serology for respiratory pathogens was requested in 12 patients and yielded one diagnosis of Mycoplasma pneumoniae and one of influenza A. Legionella pneumophila type 1 urinary antigen testing was negative in all patients.
PNAG testing was performed for all patients in the ED. A delay in urine collection meant that nine patients had left the ED before the test was performed, with a potential impact on antibiotic selection. Nine patients (15%) had a positive test. All three cases with a later laboratory culture of S pneumoniae were detected. On review, the remaining six patients had a clinical presentation consistent with pneumococcal pneumonia but no other test confirmed the diagnosis. Only one of the nine patients had received antibiotics prior to urine collection. With the combination of blood culture, serology, sputum and PNAG testing, a microbiological cause for admission was identified in 19 patients (32%), 6 of which were diagnosed by PNAG only. There was no significant difference between patients with a positive or negative PNAG in Pneumonia Severity Index, duration of illness prior to presentation, length of hospital stay or time on antibiotics (table 1).
The microbiological results, Pneumonia Severity Index and antibiotic prescriptions for the nine patients with a positive PNAG are summarised in table 2.
Blinding of laboratory staff to the ED test result was successfully achieved. There was 98% concordance between emergency and microbiology department results, the single exception being a “weak positive” PNAG on ED interpretation but negative on retesting in the laboratory. Using the prevalence and diagnostic test performance data obtained yielded a sensitivity of 100% (95% confidence interval (CI) 70.1% to 100%) and a specificity of 98% (95% CI 89.5% to 99.7%). The day 3 urine specimen from this patient also tested negative but sputum culture yielded both S pneumoniae and H influenzae.
Trial eligibility, based on the results of clinical examination and chest radiography, was established a mean of 1 h 47 min after patient arrival in the ED (median 1 h 43 min). The ED test result was available in a median time of 2 h 39 min after confirmation of trial eligibility. The mean time to the ED test result was 4 h 20 min as patient dehydration at time of arrival caused lengthy delays to urine collection in some patients. This compared with the mean time of 26 h 16 min and median time of 19 h 49 min from confirmation of trial eligibility to laboratory result availability. ED results were available within 1 h of confirmation of trial eligibility in 27% of cases and within 4 h in 65%. Twenty-five tests (42%) were performed outside routine laboratory working hours (08:00–17:30 daily). For patients who presented during laboratory hours and had their urine tested on the day of presentation, the median delay between the ED and laboratory test results was 2 h 20 min (mean 3 h).
Repeat urine tests were performed on day 3 of admission where possible. Of patients with an initial positive test, five remained positive, two became negative, and two had no day 3 specimen collected. Of 27 patients retested at day 3 after an initial negative PNAG, all remained negative.
Of 59 ED tests, six results remained unavailable at the time of antibiotic prescription, usually due to difficulty in obtaining a urine specimen. In 42% of the remaining cases the ED doctor recorded that they found the test “useful”. Despite this, there was no trend to either a narrower spectrum of antibiotic use (benzylpenicillin, ampicillin or amoxycillin) in patients with a positive PNAG or prescription of broader spectrum antibiotics in those with a negative test. In fact, in those with a positive PNAG, 5/9 patients (55%) had a change in antibiotics compared with only 25% of those with a negative PNAG (p = 0.07). Of those patients who had a change in antibiotics, this change broadened antibiotic coverage in 80% with a positive PNAG and 64% with a negative PNAG (p = 0.51).
Our study in a small consecutive ED patient cohort shows that the new generation of ICT-based S pneumoniae diagnostic kits can be reliably performed and interpreted at the bedside by clinical staff in EDs. Routine use enhances detection of the most common cause of CAP when compared with standard microbiological techniques.
Point-of-care testing in the ED has been shown to be valuable in the diagnosis and early initiation of antiviral medication in paediatric influenza.10 Bonner and colleagues demonstrated a higher rate of specific and appropriate influenza treatment as well as a reduction in the number and cost of further investigations. Their study used trained laboratory staff seconded to the ED to collect and test nasopharyngeal washings.
The BinaxNOW test uses a urine specimen which requires no additional skills for collection or preparation. Performance of the test involves a single reagent application and a 15 min wait during which other tasks can be performed. The ease of performance and similarity to urinary βHCG testing, with which the ED staff were familiar, allowed reliable testing by nursing staff after a single training session. This allowed test results to be available within 4 h, regardless of time of day, while simultaneously increasing the diagnostic yield in patients with pneumonia from 22% to 32%. Although the ICT is less likely than culture-based tests to be affected by prior antibiotic treatment, performing the test early may increase the diagnostic yield. We noted that, of seven patients who had re-collection of urine after both ED and laboratory confirmed positive tests, two became negative by day 3 of admission. This extends observations by Gutierrez et al7 and van der Eerden et al11 that antibiotic treatment can reduce test sensitivity, while contrasting with other findings that PNAG persists as long as 6 weeks in some patients.12 It seems likely that PNAG sensitivity is affected by treatment, but to a lesser extent than culture-based diagnosis.
Making the PNAG available in the ED carries the additional possibility of rapid antigen detection from cerebrospinal fluid in suspected pneumococcal meningitis, a setting in which the test has also been validated.13 14
Outcome studies in CAP suggest that starting parenteral antibiotics within 4 h of admission improves recovery.15 Our turnaround time was within this timeframe in almost two-thirds of cases and within 1 h in one-quarter of patients. In the UK, where patients are usually in the ED for no longer than 4 h, it may be appropriate to consider use of the PNAG in the medical admissions unit. No attempt was made to expedite urine collection as part of this study. If the PNAG was to be incorporated as part of an initial management guideline, turnaround times could be improved by a focus on early prompts for urine collection and/or catheterisation to obtain samples. Such initiatives might well bring the time of testing consistently within the UK ED 4-hour period. A diagnostic result available early in the admission has a greater chance of influencing treatment decisions than one available after 48 h of culture.16 17
Although we collected data on intended and actual antibiotic prescriptions, we purposefully made no attempt to direct therapy based on test results. There was a trend towards changing antibiotic prescribing in light of a positive test, with most of these changes being a broadening rather than a narrowing of antibiotic coverage. A number of patients with confirmed pneumococcal pneumonia therefore received broad-spectrum antibiotic cover. This was usually due to severe disease, in line with data suggesting dual therapy in severe CAP.18 Prescribing may also be influenced by North American guidelines (Infectious Diseases Society of America, the American Thoracic Society and Center for Disease Control) which advocate inclusion of atypical pathogen cover for all patients with pneumonia requiring hospital admission.18 However, two recent meta-analyses failed to show a benefit over β-lactam monotherapy even in subgroups subsequently shown to have Chlamydophila or Mycoplasma infection.19 20
The availability of a PNAG test result within 1–4 h of presentation and classifying 15% (33% in some studies) of patients as having pneumococcal pneumonia may improve the clinician’s level of comfort with omitting atypical cover. This would need to be combined with “change culture” strategies to affect prescribing habits. It is interesting to note that only 42% of clinicians felt that the point-of-care PNAG test result was useful. On the one hand, this figure is considerably higher than the 15% positivity rate which suggests that clinicians also found the negative test results to be useful. On the other hand, the fact that in 58% of cases the clinicians did not find the test result useful might indicate that the test will only be of use in a more highly selected patient group than was used in this study. The safety of narrow-spectrum β-lactam therapy based on a positive PNAG test result has been confirmed in populations with low and high penicillin resistance.21–25 Oosterheert et al looked at the cost-effectiveness of such targeted therapy and found no cost saving except where broad-spectrum therapy was routine empirical treatment.26 Potential benefits such as a reduction in expense and risk of investigations performed in the search for a defined pathogen and the long-term reductions in antibiotic resistance in the community are more difficult to quantify.
Our strategy tested all patients with a clinical suspicion of pneumonia and an infiltrate on the chest radiograph. A higher yield is likely if restricted to seasonal risk periods for pneumococcal pneumonia and/or those with lobar pneumonia on the radiograph. Pneumococcal disease accounts for 20–37% of pneumonia in published series. Our rate of 15% may reflect the published 73–92% sensitivity of this test. There was no reduction in the burden of invasive pneumococcal disease (isolated from sterile sites) reported to health authorities during the period of the study.27 Nevertheless, even with our simple and inclusive entry criteria, a positive diagnosis was made in 15% of patients, a considerably higher yield than with established microbiological and serological tests of similar or greater cost.
The major limitation of this study was the failure to recruit sufficient patients to achieve adequate statistical power. The small sample size resulted in wide confidence intervals for diagnostic test parameters. The time of urine collection was not recorded.
A diagnosis of pneumococcal pneumonia by PNAG detection in the ED has equivalent accuracy to—but a faster turnaround time than—laboratory-based testing. PNAG testing improves diagnostic rates in conjunction with standard microbiological testing. More timely confirmation of pneumococcal pneumonia may allow incorporation of this information into decisions regarding initial antibiotic choice. Without specific efforts to achieve early urine collection, the timeframe of testing will often fall outside the 4-hour patient stay of a UK ED and may be more appropriately considered as a test for the medical admissions unit in this setting. The observation of a non-significant trend towards broader antibiotic prescribing with a positive PNAG suggests that knowledge of, and/or compliance with, current guidelines is suboptimal. Future studies incorporating PNAG detection as part of a diagnostic and treatment algorithm should be performed to further define its clinical usefulness in the management of CAP.
The authors thank the staff of the microbiology department (particularly Seema Chitnis) and emergency department for their participation in the study. Dr John Kelly and Dr Toby Thomas undertook the pilot study in 2003.
Funding: Laboratory Diagnostics Pty Ltd (Bankstown NSW Australia) provided free pneumococcal and legionella ICT diagnostic kits as well as training for Emergency and Laboratory staff. Laboratory Diagnostics Pty Ltd did not influence the design of the study nor the collection, interpretation and presentation of data from this study.
Competing interests: None declared.