Pneumococcal Urinary Antigen Testing in United States Hospitals: A Missed Opportunity for Antimicrobial Stewardship

Abstract

Background.

The Infectious Diseases Society of America recommends pneumococcal urinary antigen testing (UAT) when identifying pneumococcal infection would allow for antibiotic de-escalation. However, the frequencies of UAT and subsequent antibiotic de-escalation are unknown.

Methods.

We conducted a retrospective cohort study of adult patients admitted with community-acquired or healthcare-associated pneumonia to 170 US hospitals in the Premier database from 2010 to 2015, to describe variation in UAT use, associations of UAT results with antibiotic de-escalation, and associations of de-escalation with outcomes.

Results.

Among 159 894 eligible admissions, 24 757 (15.5%) included UAT performed (18.4% of intensive care unit [ICU] and 15.3% of non-ICU patients). Among hospitals with ≥100 eligible patients, UAT proportions ranged from 0% to 69%. Compared to patients with negative UAT, 7.2% with positive UAT more often had a positive Streptococcus pneumoniae culture (25.4% vs 1.9%, P < .001) and less often had resistant bacteria (5.2% vs 6.8%, P < .05). Of patients initially treated with broad-spectrum antibiotics, most were still receiving broad-spectrum therapy 3 days later, but UAT-positive patients more often had coverage narrowed (38.4% vs 17.0% UAT-negative and 14.6% untested patients, P < .001). Hospital rate of UAT was strongly correlated with de-escalation following a positive test. Only 3 patients de-escalated after a positive UAT result were subsequently admitted to ICU.

Conclusions.

UAT is not ordered routinely in pneumonia, even in ICU. A positive UAT result was associated with less frequent resistant organisms, but usually did not lead to antibiotic de-escalation. Increasing UAT and narrowing therapy after a positive UAT result are opportunities for improved antimicrobial stewardship.

Although pneumococcal vaccination in the United States has been recommended for adults aged >65 years since 1984 and for children since 2001, Streptococcus pneumoniae remains the most common bacterial cause of community-acquired pneumonia (CAP) [1] and can be identified by blood cultures, sputum cultures, and urinary antigen testing (UAT). The 2007 Infectious Diseases Society of America (IDSA)/American Thoracic Society (ATS) Consensus Guidelines on the Management of Community-Acquired Pneumonia in Adult (IDSA guidelines) recommend diagnostic testing for etiology of CAP when knowledge of the specific pathogen would alter management decisions [2]. Thus, UAT is recommended when broad-spectrum empirical antibiotics are recommended: that is, in cases of severe CAP, failure of outpatient antibiotics, leukopenia, alcohol abuse, severe chronic liver disease, asplenia, and pleural effusion. In those cases, identification of S. pneumoniae should allow for rapid de-escalation of antibiotic therapy.

Theoretically, S. pneumoniae could be identified through cultures. However, <30% of pneumococcal pneumonia cases have positive blood cultures [3], and the yield from sputum cultures is <50% [4]. UAT has several advantages over cultures; it is rapid and easy to perform, allowing for prompt antimicrobial adjustment, and it is unaffected by prior antibiotic administration. UAT has a higher sensitivity than cultures (74% in adults) with specificity >90% [3, 5].

The IDSA guidelines do not generally recommend UAT outside the intensive care unit (ICU), although it is recommended “whenever the result is likely to change antibiotic management” [2]. Since the guidelines stipulate that non-ICU patients should receive empiric therapy with agents that offer appropriate coverage for S. pneumoniae, there would presumably be no role for UAT outside the ICU. Despite the guidelines, however, a growing percentage of patients with CAP are treated with empiric agents having activity against methicillin-resistant Staphylococcus aureus (MRSA) and resistant gram-negative organisms [6]. In these cases, UAT would be guideline-concordant and represents an opportunity for antibiotic stewardship. The 2005 ATS/IDSA guideline regarding healthcare-associated pneumonia (HCAP) did not address UAT, but UAT could potentially be useful for HCAP patients, too, as broad-spectrum empirical antibiotic therapy is recommended for them [7], and the most recent guideline recognized that multidrug-resistant pathogens are less common in HCAP than previously thought [8].

Performance of UAT is variable [9–11]. Moreover, few studies have investigated its utility in clinical practice, and none have employed a national sample. To determine adherence to UAT recommendations and whether UAT results impact prescribing behavior, we examined trends in the use of UAT and associations of UAT results with antimicrobial de-escalation in a national hospital discharge database. We also assessed whether de-escalating broad-spectrum antibiotics in response to a positive UAT result was associated with harm as measured by in-hospital deaths and ICU transfers.

METHODS

Patient Sample

We included patients with pneumonia from hospitals that contributed data to the Premier Health Care Database from 2010 to 2015 [12] and that also report microbiological data through the Safety Surveillor infection tracking tool. The Premier Health Care Database is widely used for research and has been well-described [13]. It includes demographic information, discharge International Classification of Diseases, Ninth Revision diagnosis codes, and a date-stamped record of each item billed during hospitalization. Safety Surveillor includes the results of all microbiological tests, including blood and respiratory cultures and UAT. Hospitalizations of patients aged ≥18 years with either a principal diagnosis code for pneumonia with a present-on-admission flag, or with principal diagnosis code for respiratory failure, acute respiratory distress syndrome, respiratory arrest, sepsis, or influenza paired with a secondary diagnosis code for pneumonia (present on admission), and for whom at least 1 test for pneumococcus (blood or respiratory culture or UAT) was performed by hospital day 2 were included. Patients with hospital-acquired pneumonia or ventilator-associated pneumonia were excluded. HCAP was defined as having any of the following: admitted from a skilled nursing facility, previous admission within 6 months, having a diagnosis for end-stage renal disease or hemodialysis by hospital day 2, or receiving immunosuppressant medications [7]. The remaining patients were considered to have CAP. Antibiotic de-escalation was defined as narrowing therapy to a single agent with activity against S. pneumoniae. For example, after beginning treatment with vancomycin and piperacillin-tazobactam, if vancomycin was stopped or piperacillin-tazobactam changed to a cephalosporin or quinolone, it would be considered de-escalation. The de-escalation analysis considered only patients who received an antipseudomonal other than quinolones or anti-MRSA therapy on the day UAT was performed. Patients with questionable pneumonia (because all antibiotics were stopped within 3 days) or positive cultures for other bacteria were excluded from the analysis of de-escalation, because cultures indicating other pathogens could impact the de-escalation decision. We included only 1 randomly selected hospitalization for each patient. This study was approved by the Cleveland Clinic Institutional Review Board.

Statistical Methods

First, we summarized and compared baseline characteristics of patients with and without UAT using frequencies, proportions, and Pearson’s χ² tests for categorical data, and means, standard deviations, and Student t tests for continuous data. We also used a histogram and a line chart to describe, respectively, variation in UAT rates across hospitals and over time, a bar chart to show association of CAP resistance with UAT result, and a line chart to show the joint association of de-escalation with UAT result and hospital UAT rate.

Next, we assessed whether de-escalation would have been appropriate based on microbiological outcomes of blood and respiratory cultures, including sputum cultures, obtained in the first 2 hospital days, and determined, stratified by UAT result, the percentages of cultures that grew S. pneumoniae, other organisms, and organisms resistant to recommended CAP therapy (either ceftriaxone and azithromycin or a quinolone), based on sensitivity testing.

We then conducted a patient-level analysis of the association between UAT result and de-escalation of antibiotics. We limited this analysis to patients who received an antipseudomonal drug or a drug with activity against MRSA on the day of UAT. De-escalation frequency among patients who did not undergo UAT was also included in this analysis for comparison, by randomly assigning a pseudo-UAT date based on the observed distribution of UAT dates (1% in the emergency department, 57% on hospital day 1, and 42% on hospital day 2) and including patients who received such drugs on the day assigned. We compared the proportions de-escalated within 3 days between the subgroups with and without positive UAT, as well as those with no UAT, treating the hospitals as clusters and using the second-order, null variance–based Rao-Scott analogue to the Pearson χ² statistic for hypothesis testing [14].

Next, we assessed the association between hospital rate of UAT and de-escalation following a positive result. We hypothesized that higher rates of UAT would indicate greater belief in and willingness to act on the results, and thus that positive UAT would be more strongly linked to de-escalation in hospitals where UAT was used more frequently. We calculated the fraction of patients at each hospital undergoing UAT, restricting to hospitals with ≥100 eligible patients. Hospitals were then categorized into 7 subgroups by 10% increments in utilization fraction, and the proportions of patients de-escalated were plotted for patients in each subgroup with and without positive UAT. We examined the interaction of antigen test positivity and hospital UAT utilization fraction in a mixed logistic analysis of covariance model, with random additive hospital effects and a 5-knot restricted cubic spline in the hospital’s proportion of patients receiving UAT.

Last, we assessed if de-escalation following UAT was associated with subsequent ICU transfer or in-hospital death among patients who were de-escalated by day 3 after a positive UAT result to those de-escalated with a negative UAT result and to those with no UAT, using mixed logistic regression models with random hospital effects, adjusting for demographics, comorbidities, and indicators of disease severity. We compared UAT-positive to UAT-negative de-escalation, because UAT-positive patients were more likely to be de-escalated to narrower-spectrum drugs effective against pneumococcus, and if this were dangerous, we would expect them to have worse outcomes. All analyses were performed using SAS software, version 9.4 (SAS Institute, Cary, North Carolina).

RESULTS

Of the 159 894 patients with pneumonia included for analysis, 25 932 (16.2%) had UAT (18.4% of ICU patients and 15.3% of non-ICU patients). A flowchart for inclusion appears in the Supplementary Data. UAT was performed on 15.5% of patients initially receiving antipseudomonal (other than a fluoroquinolone) or anti-MRSA therapy, and 17.2% of patients receiving guideline-recommended CAP therapy. Rates of UAT across hospitals ranged from 0% to 69% and were highly skewed; approximately 1 in 3 hospitals reported no UAT, and another third performed UAT for <10% of patients (Figure 1). However, overall utilization of UAT rose steadily and doubled during the 5-year study period (Figure 2). Characteristics of patients with and without UAT appear in Table 1. Because of the very large sample, even small differences are highly statistically significant (P < .001). Patients with UAT were younger, less likely to have aspiration pneumonia (6.3% vs 10.2% without UAT), and more likely to have sepsis (37.6% vs 32.5% without UAT) or to be admitted to ICU (34.2% vs 29.3% without UAT). The rate of positive tests was 7.2% and did not vary by hospital testing rate, implying that hospitals that ordered fewer tests did not test more selectively. It also did not vary over time. Positive tests were more common inside than outside the ICU (8.9% vs 6.4%, respectively, P < .001). Patients with CAP and HCAP were similarly likely to test positive (7.4% vs 6.7%, respectively, P = .06).

Figure 1.

Hospital rates of pneumococcal urinary antigen testing (UAT). Distribution of hospitals across binned proportions of UAT utilization among patients admitted with pneumonia and undergoing blood or respiratory (including sputum) culture or UAT within 2 days of hospitalization. Hospitals reporting <100 such patients, reporting on 2.8% of such patients, are excluded. Data include 159 353 patients from 164 hospitals, including hospitals with ≥100 patients. Sixty-five (39.6%) hospitals did not have any UAT ordered.

Figure 2.

Time trends in urinary antigen testing (UAT) ordering and results. Proportion (percentages) of patients undergoing UAT and fraction of positive UAT results among patients admitted with pneumonia and undergoing blood or respiratory (including sputum) culture or UAT within 2 days of hospitalization.

Table 1.

Characteristics of Patients Admitted With Pneumonia and Undergoing Blood or Respiratory Culture or Urinary Antigen Testing Within 2 Days of Hospitalization

		UAT
	Total	Not Tested	Tested
Factor	(N = 159 894)	(n = 133 962)	(n = 25 932)
Age, y, mean	69.4	69.8	66.9
Sex, %
Female	51.0	51.1	50.4
Male	49.0	48.9	49.6
Race/ethnicity, %
White	77.1	76.8	78.5
Black	12.6	12.1	15.4
Hispanic	0.66	0.74	0.23
Other	9.6	10.4	5.9
Admission source, %
Emergency room	87.7	87.2	90.3
SNF/ICF	7.9	8.0	7.6
Clinic	4.3	4.7	2.0
Other	0.08	0.09	0.07
Discharge disposition, %
Home	44.7	43.5	51.2
Home health	13.9	13.8	14.7
Hospice	5.6	5.9	4.0
Expired	8.2	8.3	7.4
SNF	25.1	26.0	20.2
Other hospital	2.5	2.5	2.4
Insurance payer, %
Medicare	71.8	72.8	66.9
Medicaid	8.9	8.7	10.0
Managed care	10.6	10.3	12.3
Commercial indemnity	3.1	3.0	3.4
Other	5.6	5.2	7.4
Principal diagnosis, %
Pneumonia	50.8	51.0	49.7
Aspiration pneumonia	9.6	10.2	6.3
Sepsis	33.4	32.5	37.6
Respiratory failure	6.3	6.3	6.3
Dialysis, %	4.7	5.2	2.2
Immunosuppressed, %	16.1	15.8	17.7
Admission within last 6 mo, %	11.0	11.5	8.5
No. of HCAP risk factors, %
0	66.4	65.9	68.7
1	28.0	28.2	26.8
2	5.1	5.3	4.2
3	0.49	0.53	0.26
4	0.02	0.02	0.01
Intensive care unit, %	30.1	29.3	34.2
Invasive mechanical ventilation, %	12.5	12.4	13.2
Vasopressor use, %	10.1	10.1	10.4
Blood cultures within 48 h	96.9	97.7	93.1
Respiratory cultures within 48 h	33.0	31.3	41.4
Early antibiotic therapy
Anti-MRSA/antipseudomonal	55.8	56.2	53.5
Quinolones	20.1	19.9	21.1
Ceftriaxone/azithromycin	23.6	23.3	25.1
Other antibiotics	0.6	0.6	0.3
Antibiotic de-escalation within 2 d of UATa	15.3	14.6	18.6

Abbreviations: HCAP, healthcare-associated pneumonia; ICF, intermediate care facility; MRSA, methicillin-resistant Staphylococcus aureus; SNF, skilled nursing facility; UAT, urinary antigen testing.

^aPatients initially treated with anti-MRSA or antipseudomonal (other than quinolones) therapy. The reported fraction who were de-escalated within 2 days of UAT, among patients who were not tested, was approximated synthetically by randomly imputing a hospital day of testing to patients who were not tested, by drawing randomly with replacement from the distribution of hospital testing days among patients who were tested, and counting de-escalation on that day and the next 2 hospital days. The fraction in the “Total” column was then computed as the sample size–weighted average of the “Tested” and “Not tested” fractions.

Utility of UAT in Identifying S. pneumoniae and Ruling Out Other Bacteria

Patients with a positive UAT result were much more likely than patients with a negative UAT result to have S. pneumoniae isolated from blood or sputum (25.4% vs 1.9%, P < .001), and less likely to grow other organisms (8.1% vs 11.8%, P < .001). Patients with positive sputum cultures were also less likely to grow an organism resistant to CAP therapy (Figure 3). For patients with positive cultures, when UAT was positive and pneumococci were not recovered in culture, the most common nonpneumococcal organisms were S. aureus (13.9%), Escherichia coli (11.5%), Proteus mirabilis (4.9%), and Klebsiella species (4.1%).

Figure 3.

Resistance to community-acquired pneumonia (CAP) therapy. Proportions (percentages) of patients undergoing urinary antigen testing (UAT), and blood or respiratory (including sputum) culture within 2 days of hospitalization, who had organisms resistant to CAP therapy, by UAT result and source of positive culture.

Physician Response to UAT Results

In our sample, 61 083 patients received an antipseudomonal drug other than a fluoroquinolone or anti-MRSA drug empirically, and 9960 (16.3%) of these underwent UAT. On the third day after UAT, most patients were still receiving the same drugs, but patients with a positive UAT result were more likely to have their coverage narrowed (38.4% with positive UAT vs 17.0% with negative UAT vs 14.6% with no UAT, P < .001). The median duration of broad-spectrum antibiotics for patients with positive, negative, or no UAT was 3 days, 4 days, and 5 days, respectively (P < .001). The duration of vancomycin, piperacillin-tazobactam, and carbapenems were all shorter in response to a positive UAT result (Supplementary Table 1). The rate of de-escalation following UAT positivity tended to increase with increasing hospital use (Figure 4; interaction P = .04).

Figure 4.

Proportion of patients de-escalated by urinary antigen testing (UAT) result and hospital proportion of patients undergoing UAT. The symbols connected by straight lines show the proportions de-escalated in bins of hospitals whose UAT proportions fall respectively within ranges from 0–10% to 60%–70% with intermediate intervals, each of 10% width, centered on their midpoints, for patients with positive (triangles) and negative (circles) UAT results. The dashed lines without symbols show restricted cubic spline fits from a generalized linear mixed binomial regression model with identity link function, fixed effects of UAT result, observed hospital UAT proportion and their interaction, and random hospital effects.

Patient Outcomes

In-hospital deaths were infrequent (4.4%) among the 1852 patients who underwent UAT and were de-escalated by day 3, and similar among those de-escalated after positive UAT vs negative UAT (adjusted odds ratio [aOR], 0.81; 95% confidence interval [CI], .43–1.52) or no UAT (aOR, 0.69; 95% CI, .38–1.24). ICU transfers on day 3 or later were also infrequent (3.0%), and similar for those de-escalated with positive vs negative UAT (aOR, 0.93; 95% CI, .26–3.31) or no UAT (aOR, 0.59; 95% CI, .18–1.97). Only 1 de-escalated patient with positive UAT was subsequently admitted to ICU.

DISCUSSION

In this large sample of patients with CAP at US hospitals, only 16% had rapid pneumococcal UAT performed, but there was wide variability in utilization across hospitals. A positive UAT result was associated with a lower probability of a nonpneumococcal pathogen being recovered in culture, and most of these organisms were susceptible to CAP therapy. More than half of all patients (48.0% of patients with CAP and 71.1% of those with HCAP) received empiric treatment with an antipseudomonal or anti-MRSA drug, yet <1 in 5 had these drugs de-escalated within 3 days despite negative bacterial cultures. Patients with a positive UAT result were more likely to have their empiric therapy narrowed by day 3, and physician response to a positive UAT result varied by hospital, with physicians at hospitals with the highest rates of UAT use demonstrating the highest rates of de-escalation in response to a positive test. If some UAT results were false positives, or if patients had >1 pathogen, de-escalation based on UAT could have resulted in insufficient coverage and worse outcomes. However, following de-escalation, patients with positive UAT did not appear more likely than those with negative UAT to suffer late deterioration or death. In fact, only 1 of these patients was subsequently admitted to ICU.

It is unclear why rates of UAT are so low. UAT appears most useful for identifying patients at high risk for S. pneumoniae, allowing for de-escalation of broad-spectrum empiric antibiotics. One potential explanation is the IDSA guideline recommendation to limit UAT to the ICU, instead of recommending it for all patients receiving broad-spectrum antibiotics, whether or not such antibiotics are indicated [2]. However, the difference in testing rates between ICU and non-ICU patients was only a few percent, suggesting that guidelines had minimal effect on physician ordering. Another potential explanation is cost. Physicians may not consider S. pneumoniae to be a common cause of CAP in the era of conjugated vaccine. Nevertheless, pneumococcus is still the most common bacterial cause of CAP [1]; UAT was positive in 7% of patients—higher than reported by others [15]—and the cost per test is low. Medicare reimbursement is approximately $17—similar to human immunodeficiency virus antibody testing.

A final possibility is that physicians are not confident in the specificity of UAT. Although the specificity is about 97%, false positives do occur [3]. The current CAP guidelines do not describe UAT as a reliable stand-alone microbiologic method for identifying the etiology of CAP, nor do they advise using a positive UAT result to narrow therapy. The committee strongly recommended narrowing antibiotic therapy as best practice but offered no specifics about when to do so; they also noted “the possibility of polymicrobial CAP and the potential benefit of combination therapy for bacteremic pneumococcal pneumonia have complicated the decision” [2]. We found that 8% of patients with positive UAT grew organisms other than pneumococcus, but only 4% grew organisms that would be resistant to CAP therapy. Several smaller studies suggest good clinical outcomes following de-escalation based on UAT [11, 16, 17]. We also found that patients whose antibiotics were de-escalated following UAT had very low rates of late deterioration or death.

Overall, physicians de-escalated therapy in 38.4% of cases following positive UAT. This varied by hospital—those with the highest rates of UAT de-escalated therapy almost 50% of the time, whereas those who rarely ordered UAT de-escalated therapy in only 20% of cases. Single-system studies have reported de-escalation rates across this range. Sordé et al, at a Barcelona hospital in 2007–2008, found that 32% of positive UAT resulted in antibiotic optimization [16]. In contrast, West et al studied 7 Intermountain Healthcare hospitals from 2009 to 2012 and found that de-escalation occurred in 63% of patients [11]. Importantly, we found the rate of positivity did not vary by hospital or over time, suggesting that physicians are not good at identifying who will test positive.

Narrow-spectrum, pathogen-directed therapy has a number of potential benefits, including fewer adverse drug effects, reduced pressure for the emergence of antibiotic resistance, and less disruption of normal microbiota with associated sequelae (eg, Clostridioides difficile infection or colonization with multidrug-resistant organisms). In our study, empirical therapy with antipseudomonal or anti-MRSA drugs was common both in and out of the ICU and for CAP as well as HCAP. Use of empiric broad-spectrum antibiotics has been seen in other studies and appears to be increasing [6, 18]. Overcoming the “antibiotic inertia” for these patients is a major challenge of antimicrobial stewardship as there is often no clear microbiologic diagnosis [19, 20]. For these patients, positive UAT can guide antibiotic de-escalation and, given that UAT is easy, rapid, sensitive, and specific, we believe it should be a recommended component of the workup of pneumonia for any patient receiving broad-spectrum antibiotic coverage, whether CAP or HCAP. As the IDSA formulates its next set of CAP/HCAP guidelines, consideration should be given to broadening the population of patients who could benefit from UAT and offering more specific guidance regarding the appropriate response to positive UAT.

Our study has a number of limitations. First, its observational nature makes it subject to unmeasured confounding. In our analyses of the associations between de-escalation, UAT, and outcomes, we controlled for numerous important confounders but may have missed others. This is particularly true because our primarily administrative dataset did lack important clinical variables. At the same time, complications among patients who underwent de-escalation were rare. This study also offers a descriptive view of how UAT has been used in US hospitals.

Although we cannot explain why there is such variation in testing or in physicians’ responses to positive results, we found no differences in rates of positivity or outcomes to justify it. It is possible that some hospitals provided UAT only as a send-out test and that the results were not captured in Safety Surveillor. In that case, we may have underestimated overall use of UAT, but this underestimation would not materially impact our other findings. We excluded patients treated with quinolone monotherapy, so we cannot comment on the usefulness of UAT in that population. Last, our data span 2010–2015, when not all hospitals had easy access to UAT. We found that use increased during the study period, and this trend may have continued.

In conclusion, UAT appears to be a useful test for identifying patients with pneumonia due to S. pneumoniae. When such patients are stable and receiving broad-spectrum empirical antibiotics, a positive test appears to signal that antibiotic de-escalation is appropriate. Despite the test being inexpensive, accurate and rapid, UAT appears underused, but could be an important tool for antibiotic stewardship. Broader indications for testing featured in national guidelines might help to spread its use nationwide.