Validation of Methicillin-Resistant Staphylococcus aureus (MRSA) Risk Factors in Predicting MRSA Community-Acquired Pneumonia at an Academic Medical Center

Abstract

Background: The 2019 Infectious Diseases Society of America community-acquired pneumonia (CAP) guidelines recommend antimethicillin- resistant Staphylococcus aureus (MRSA) therapy in patients with CAP based on previously identified risk factors for MRSA with an emphasis on local epidemiology and institutional validation of risk. Thus, we sought to assess the ability of guideline-recognized risk factors to predict MRSA CAP at our institution. Methods: This was a single-center, retrospective cohort study from January 2016 to March 2020. Patients were included if they were >18 years old, diagnosed with CAP, and had a MRSA nasal screen and respiratory culture obtained on admission. Patients were excluded if CAP diagnosis was not met, respiratory cultures were not obtained within 48 hours of antibiotic initiation, or they had cystic fibrosis. Sensitivity, specificity, negative predictive value, positive predictive value, and likelihood ratios (LR) were calculated using Vasser Stats 2019. Pre/post-test odds and pre/post-test probabilities were calculated using Excel 2019. Results: Of 705 screened patients, 221 were included. MRSA prevalence in CAP patients at our institution was 3.6%. History of MRSA isolated from a respiratory specimen had high specificity (98%), high positive LR of 20 (95% CI 5.3–74.8), and high post-test probability of 42.8%. Receipt of IV antibiotics during hospitalization within the past 90 days had a positive LR of 1.9 (95% CI 0.74–4.84). A positive MRSA nasal screen on admission had a positive LR of 6.9 (95% CI 4.0–12.1), negative LR 0.28 (95% CI 0.08–0.93), positive post-test probability of 20.7%, and negative post-test probability of 1.04%. Conclusion: Our study utilized institutional data to validate guideline recognized risk factors for MRSA CAP specifically at our institution. Risk factors including history of MRSA isolated from a respiratory specimen, and positive post-admission MRSA nasal screen were validated as significant risk factors; receipt of IV antibiotics during hospitalization within the past 90 days was not shown to be a risk factor for MRSA CAP based on our institutional data. Validated risk factors may help providers discern which patients with CAP at our institution would benefit most from empiric MRSA treatment.

Background

Community-acquired pneumonia (CAP) caused by methicillin-resistant Staphylococcus aureus (MRSA) is uncommon but has historically been associated with more severe disease and worse outcomes including increased mortality. ¹ Despite the low prevalence of MRSA CAP, many patients receive empiric anti-MRSA therapy. ² Recent guidelines from the Infectious Diseases Society of America (IDSA) and American Thoracic Society (ATS) recommend empiric coverage of MRSA for patients with prior lower respiratory tract (LRT) isolation of MRSA, and for those with severe disease and recent hospitalization requiring intravenous (IV) antibiotics in the past 90 days. ³ For patients with recent hospitalization requiring IV antibiotics in the past 90 days with non-severe pneumonia, the guidelines recommend to withhold anti-MRSA coverage unless the results of the MRSA nasal screen are positive.

Given the increased risk of antibiotic resistance and adverse effects associated with the use of broad spectrum antibiotics in patients with CAP, the IDSA/ATS suggest that institutions should validate risk factors for MRSA pneumonia at a local level. Thus, we sought to assess the ability of the MRSA risk factors identified in the IDSA/ATS CAP guidelines to predict MRSA pneumonia in a cohort of patients with CAP at our institution.

Methods

Study Design

A single-center, retrospective cohort study was conducted at a tertiary, academic medical center. A query of the electronic medical record was generated using a search algorithm to identify patients 18 years and older using the following Boolean search terms: MRSA nasal PCR or MRSA nasal culture, and LRT culture (sputum, endotracheal aspirate, or bronchoalveolar lavage) between January 1 2016 and March 30 2020. The MRSA nasal PCR, MRSA nasal culture, and LRT culture must have been obtained within 48 hours of hospital admission. Patients from this query were further reviewed for inclusion if they had a documented admission chart note with diagnosis of CAP. A chart review was completed to identify patients that met inclusion criteria as follows. A chest x-ray on admission was required to show radiographic evidence of pneumonia as identified by academic radiologists and 2 or more of the following clinical signs and symptoms: (1) temperature less than 36.0°C or greater than 38.0°C; (2) respiratory rate greater than 20; (3) cough; (4) hypoxia as evidenced by oxygen saturation less than 90% on room air; (5) increased sputum production; and (6) a white blood cell count less than 4000/mm³ or greater than 10 000/mm. ⁴ Patients were excluded for the following reasons: (1) diagnosis criteria for pneumonia not met; (2) aspiration pneumonia or pneumonia acquired in the hospital; (3) LRT cultures were not obtained within 48 hours of hospital admission; and (4) cystic fibrosis. The study was deemed exempt by the Institutional Review Board.

Data Collection and Definitions

All data were collected by a single investigator trained in data collection using a standardized data collection form. A second investigator re-abstracted a sample of charts, blinded to the information obtained by the first investigator. If discrepancies were identified, the first investigator reviewed inconclusive data points for accuracy. Discrepancies that were found included weight and age differences. These discrepancies were rectified and re-collected for each patient to include the weight and age of each patient at the time of admission. Collected data included demographics; pertinent culture results, vital signs, and laboratory results; history of hospitalization and IV antibiotics within the last 90 days; and history of a positive MRSA nasal screen or positive MRSA LRT culture within the last year. LRT MRSA infection was defined as a documented LRT culture which grew MRSA. Respiratory MRSA colonization was defined as a positive MRSA nasal screen, performed either via polymerase chain reaction or culture.

Statistical Analyses

Collected data were analyzed using Excel 2019 (Microsoft, Inc., Redmond, WA) and presented using descriptive statistics, including mean and standard deviation (SD) and number and percentage. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (+LR) and negative likelihood ratio (−LR) were calculated to estimate the degree to which the following risk factors predict MRSA CAP: (1) history of previous LRT MRSA infection or nasal colonization within the past year; (2) history of hospitalization requiring IV antibiotics within the last 90 days; and (3) post-admission MRSA nasal screen. The efficient-score method, corrected for continuity, was used to calculate 95% confidence intervals (CIs) for sensitivity, specificity, PPV, NPV, +LR, and −LR. These calculations were performed using an internet-based statistical software package (Vassar Stats [www.vassarstats.net]; accessed 12 May 2020).

MRSA CAP pre- and post-test probability was estimated if a risk factor had 95% CI for a +LR or −LR not containing one. Pre-test probability was the prevalence of MRSA CAP within our cohort. Post-test probability was estimated using the following procedure: (1) pre-test odds were calculated using the following equation: pre-test odds = prevalence/(1 − prevalence); (2) post-test odds positive was calculated using the following equation: post-test odds = + LR × pre-test odds; (3) post-test odds negative was calculated using the following equation: post-odds negative = −LR × pre-test odds; (4) post-test probability positive was calculated using the following equation: post-test probability positive = post-test odds positive/(post-test odds positive + 1); and (5) post-test probability negative was calculated using the following equation: post-test probability negative = post-test odds negative/(post-test odds negative + 1). These calculations were performed using Excel 2019 (Microsoft, Inc., Redmond, WA).

Results

In total, 705 patients were screened for inclusion and 221 were included after exclusion criteria were applied (Figure 1). Table 1 shows demographics, clinical characteristics, and microbiologic information. The majority of patients were male (59.3%) and had a mean age and weight of 62.1 years (SD 16.9) and 81.6 kg (SD 29.6), respectively. MRSA nasal culture was the most common MRSA nasal screen method performed, and expectorated sputum was the most common specimen type collected. MRSA was isolated from LRT cultures in 3.6% (8/221) of included patients.

Figure 1.

Inclusion and exclusion criteria.

LRT = lower respiratory tract.

Table 1.

Demographics, Clinical Characteristics, and Microbiologic Information (n = 221).

Demographics
Age (years), mean (SD)*	62.1 (16.9)
Weight (kg), mean (SD)	81.6 (29.6)
Male gender, n (%)	131 (59.3)
Clinical characteristics
Temperature (C), mean (SD)	36.8 (2.5)
Respiratory rate (breathes per minute), mean (SD)	22 (5.8)
White blood cell count (/mm3), mean (SD)	14.9 (15.4)
Oxygen saturation (%), mean (SD)*	92.2 (9.6)
On room air, n (%)	141 (63.8)
Increased sputum production, n (%)	155 (70.1)
Cough, n (%)	192 (86.9)
Microbiologic information
MRSA screen methodology
Nasal swab – culture, n (%)	164 (74.2)
Nasal swab – PCR, n (%)	57 (25.8)
LRT culture positive for MRSA, n (%)	8 (3.6)
Sputum culture, n (%)	7 (87.5)
Bronchoalveolar lavage, n (%)	1 (12.5)
LRT culture negative for MRSA, n (%)	213 (96.4)
Sputum culture, n (%)	192 (90.1)
Bronchoalveolar lavage, n (%)	5 (2.3)
Endotracheal aspirate, n (%)	16 (7.5)

LRT = lower respiratory tract; MRSA = methicillin-resistant Staphylococcus aureus; PCR = polymerase chain reaction.

^*One patient did not have a documented age and one did not have documented oxygen saturation.

Table 2 shows sensitivity, specificity, NPV, PPV, −LR, and +LR. All risk factors had poor to moderate sensitivity and moderate to good specificity. The strongest predictors for MRSA in CAP, based on +LR, were previous history of LRT MRSA infection or colonization within the past year, and a positive post-admission MRSA nasal screen. Recent hospitalization with IV antibiotics appeared to be a poor positive predictor, given its weak +LR (1.9) and a 95% CI containing one. All risk factors had poor to moderate PPV, but were difficult to interpret, as PPV is dependent on disease prevalence, and our prevalence rate for MRSA in CAP was low. Negative post-admission MRSA nasal screening appeared to be a moderate to strong predictor for absence of MRSA in CAP, given its −LR (0.28) and NPV (99%). The following risk factors demonstrated a high NPV, but a poor −LR with a 95% CI containing one: (1) no prior history of previous LRT MRSA infection or colonization within the past year (NPV 98%, −LR 0.64); and (2) no prior history of previous hospitalization with receipt of IV antibiotics within the past 90 days (NPV 97%, −LR 0.78).

Table 2.

Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, Positive Likelihood Ratio, and Negative Likelihood Ratio.

	Risk factor prior to CAP admission		Risk factor following CAP admission
	History of MRSA lower respiratory tract infection or colonization past year	Hospitalization and intravenous antibiotics past 90 days	Post-admission MRSA nasal screen
True positive, n	3	3	6
False positive, n	4	42	23
True negative, n	209	171	190
False negative, n	5	5	2
Sensitivity (95% CI)	38% (10%-74%)	38% (10%-74%)	75% (36%-96%)
Specificity (95% CI)	98% (95%-99%)	80% (74%-85%)	89% (84%-93%)
Positive predictive value (95% CI)	43% (12%-80%)	7% (2%-19%)	21% (9%-40%)
Negative predictive value (95% CI)	98% (94%-99%)	97% (93%-99%)	99% (96%-99%)
Positive likelihood ratio (95% CI)	20 (5.3-74.8)	1.9 (0.74-4.84)	6.9 (4.0-12.1)
Negative likelihood ratio (95% CI)	0.64 (0.37-1.1)	0.78 (0.45-1.33)	0.28 (0.08-0.93)

CAP = community acquired pneumonia; MRSA = methicillin-resistant staphylococcus aureus.

Table 3 shows the pre- and post-test probability for MRSA in CAP. Pre-test probability was 3.6%, indicating that 3.6% of the patients included in this study had a confirmed MRSA community acquired pneumonia. Post-test probability for MRSA in CAP increased to 42.9% if patients had a previous history of LRT MRSA infection or colonization within the past year. Post-test probability increased to 20.7% in patients with a positive post-admission MRSA nasal screen. The increase in post-test probability in these scenarios indicated a higher likelihood of true disease if patients had the listed risk factor upon presentation. Post-test probability decreased to 1% in patients with a negative post-admission MRSA nasal screen. This indicated a lower likelihood of true disease if MRSA nasal screen was negative. Table 4 briefly describes the equations and definitions for NPV, PPV, +LR, and -LR.

Table 3.

Pre- and Post-Probability for MRSA in CAP.

	Risk factor prior to CAP admission	Risk factor following CAP admission
	History of MRSA lower respiratory tract infection or colonization past year	Post-admission MRSA nasal screen
Pre-test probability	3.6%	3.6%
Pre-test odds	0.0376	0.0376
Post-test odds positive	0.75	0.2609
Post-test odds negative		0.0105
Post-test robability positive	42.9%	20.7%
Post-test probability negative		1.0%

MRSA = methicillin-resistant staphylococcus aureus; CAP = community acquired pneumonia.

Table 4.

Calculated Abbreviations, Meaning, and Equations.

Term	Meaning	Equation
Negative Predictive Value (NPV)	The probability that a person does not have a disease or condition, given a negative result	(sum of true negative results)/(sum of true negative results + sum of false negative results) × 100
Positive Predictive Value (PPV)	The probability that subjects with a positive screening test truly have the disease	(sum of true positive results)/(sum of true positive results + sum of false positive results)
Positive likelihood ratio (+LR)	Explains how much to increase the probability of having a disease, given a positive test result	Sensitivity/(100−specificity)
Negative likelihood ratio (−LR)	Explains how much to decrease the probability of having a disease, given a negative test result	(100−sensitivity)/specificity

Discussion

This study, to our knowledge, is the first real-world validation of the 2019 IDSA/ATS recommendations for empiric anti-MRSA treatment in CAP. Empiric treatment with anti-MRSA agents is common in patients with CAP. A previous study ² demonstrated 30% of patients with CAP are initiated on empiric antibiotic therapy with an anti-MRSA agent upon hospital admission. MRSA, however, is an uncommon pathogen, and previous estimates^2,5 suggest a CAP prevalence of 2.1% to 8.2%, indicating anti-MRSA treatment is likely unnecessary in most CAP patients. For these reasons, the 2019 IDSA/ATS guidelines ³ recommend using risk factors to identify CAP patients with a higher probability for MRSA infection. In brief, the 2019 IDSA/ATS guidelines ³ recommend empiric anti-MRSA treatment for patients with a history of previous LRT MRSA infection or colonization within the past year as well as for patients with severe CAP who had a previous hospitalization with receipt of IV antibiotics within the last 90 days. However, empiric anti-MRSA treatment is not recommended for patients with non-severe CAP who have a history of recent hospitalization with receipt of IV antibiotics within the last 90 days. Thus, it is recommended to add an anti-MRSA agent only if there is a positive post-admission MRSA nasal screen. MRSA nasal screening is recommended post-admission in patients on empiric anti-MRSA treatment, as a negative result may allow for de-escalation of the anti-MRSA agent.^2,6-8

Our study has a few interesting findings. First, CAP caused by MRSA appeared to be strongly predicted by history of LRT MRSA infection or colonization within the past year. This risk factor had a high specificity, indicating few false positives. It also had a strong +LR, which produced an approximate 40% increase in the post-test probability for MRSA. PPV was only moderate in strength, but likely influenced by the low rates of MRSA infection observed. Previous studies^5,9 have shown this risk factor was associated with a 6 to 14-fold increase in the odds for MRSA pneumonia. Severe adverse reactions associated with anti-MRSA treatment such as vancomycin-associated nephrotoxicity occurs at variable rates amongst hospitalized patients (5%–43%). ¹⁰ These rates are generally lower than the estimated post-test probability found in this study, indicating the benefits of treatment would outweigh the risk in these patients.^2,10,11 Based on our findings, empiric anti-MRSA treatment appears warranted for patients treated at our institution with CAP who have had a history of LRT MRSA infection or colonization within the past year.

Second, a post-admission MRSA nasal screen appeared to predict MRSA pneumonia to some degree. A negative MRSA nasal screen had a high NPV of 99% which is similar to previously reported literature.^7,12 A moderate -LR decreased the post-test probability for MRSA to 1%. A positive MRSA nasal screen had a moderate to strong +LR, and increased the post-test probability for MRSA to 20.7%. PPV was poor, but was confounded by low MRSA rates observed at our institution. These results suggest 2 potential applications within our health-system for post-admission MRSA nasal screening. First, a negative result appeared to suggest low probability for CAP due to MRSA, which may be used to withhold or de-escalate anti-MRSA treatment in select patients. Previous data^2,6-8 has suggested a similar application. Second, a positive result appeared to suggest moderate probability for CAP due to MRSA and may be used to add anti-MRSA treatment in select patients with unclear MRSA risk. In our opinion, and in agreement with current literature,^4,12 given the low positive predictive value of a positive post-admission MRSA nasal screen, this test should not be used as the sole clinical factor to initiate anti-MRSA therapy.

Finally, history of hospital admission with receipt of IV antibiotics in the past 90 days did not appear to predict MRSA in our CAP patients. We observed moderate specificity and a weak +LR whose 95% CI included one, indicating limited predictive ability. Our results, however, are not sufficient to preclude the possibility of MRSA risk, as numerous other studies^5,9 have identified recent hospital admission and/or IV antibiotics as an independent risk factor. We, therefore, recognize the possibility of risk, and believe a reasonable approach should be based on illness severity. Any risk for MRSA CAP could be detrimental in a critically ill patient. ¹³ Therefore, we believe it is reasonable to start empiric anti-MRSA treatment in patients with severe CAP who have a recent hospitalization with receipt of IV antibiotics but plan to deescalate based on post-admission MRSA nasal screening and culture results as appropriate. In non-critically ill patients, it is our opinion that the risk of MRSA CAP is not sufficient to warrant initiation of empiric anti-MRSA treatment, as the risk for severe adverse effects from anti-MRSA treatment could be higher. In patients with non-severe CAP who had a recent hospitalization and receipt of IV antibiotics, we believe a reasonable approach is to obtain a MRSA nasal screen, and only initiate empiric anti-MRSA treatment if the result is positive.

Our analysis supports using the 2019 IDSA/ATS recommendations as a framework for deciding which patients warrant anti-MRSA treatment at our institution. Given the overall low prevalence of MRSA CAP, previous data ¹³ suggesting that empiric anti-MRSA therapy did not provide a mortality benefit, and anti-MRSA agents may cause harm, anti-MRSA agents should be used selectively. The 2019 IDSA/ATS CAP guidelines ³ emphasize the increased risk of MRSA CAP for patients with a history of a MRSA LRT infection or colonization, or positive MRSA screen taken after admission. The guidelines also suggest considering the severity of illness, especially in patients with recent hospitalization and IV antibiotic use, when determining whether empiric anti-MRSA is warranted. This is a considerable deviation from the recommendations made in the 2005 IDSA/ATS hospital-acquired and ventilator-associated guidelines to use empiric anti-MRSA treatment for a subset of patients that acquired pneumonia in the community, but had nosocomial exposure, referred to as healthcare associated pneumonia (HCAP). ³ Several studies have shown that patients categorized as having HCAP did not have a higher risk of having multi-drug resistant organisms including MRSA. ³ Therefore, our assertion is that adherence to the revised IDSA/ATS CAP guidelines will decrease the empiric use of anti-MRSA therapy for CAP at our institution, while providing MRSA coverage for patients at highest risk for MRSA CAP or for those with the greatest risk for poorer outcomes based on severity of illness. Further investigations are needed to identify the impact of current IDSA/ATS recommendations on empiric anti-MRSA treatment at our institution. Additionally, we plan to look for further opportunities to decrease empiric anti-MRSA treatment in our CAP patients.

This study has a few limitations. First, there may have been gaps in the medical record such that risk factors may not have been documented. We attempted to minimize this limitation by collecting readily available information. Second, the prevalence of MRSA at our institution was relatively low, which may warrant a larger sample size in future studies. This analysis was specific to our institution and was intended to be an institutional validation. Third, factors such as institutional procedures could influence a similar analysis at other institutions, and we would encourage clinicians to perform a similar analysis before applying our results to their practice. Also, we did not collect antibiotic related information, as it was outside scope of this study. We are, therefore, uncertain of the clinical impact adherence to the current IDSA/ATS recommendations may have. Finally, we did not differentiate between severe or non-severe CAP, as again, we felt this was outside the scope of this study.

Despite the limitations of this study, it may be utilized as a template for other institutions to follow. Additionally, if institutions have similar MRSA prevalence and characteristics such as size and patient population, they may be able to directly apply this to their center.

Conclusion

We observed low MRSA prevalence in a cohort of hospitalized CAP patients. The majority of risk factors identified in the IDSA/ATS guidelines appeared to predict MRSA pneumonia in patients with CAP at our institution. Post-admission MRSA nasal screening also appeared, to some degree, to predict MRSA presence and absence. The optimal approach for empiric anti-MRSA treatment in CAP patients is unclear, but given our results, we believe it is reasonable to apply the 2019 IDSA/ATS recommendations for empiric anti-MRSA CAP treatment to our institution.