Empirical Anti-MRSA vs Standard Antibiotic Therapy and Risk of 30-Day Mortality in Patients Hospitalized for Pneumonia | Infectious Diseases | JAMA Internal Medicine | JAMA Network
[Skip to Navigation]
Figure 1.  Study Population
Study Population

MRSA indicates methicillin-resistant Staphylococcus aureus; VA, Veterans Affairs.

aStandard antibiotics were defined as either a β-lactam plus macrolide or tetracycline, or a respiratory fluoroquinolone (moxifloxacin or levofloxacin). β-Lactams included nonpseudomonals (ampicillin, amoxicillin, ampicillin-sulbactam, amoxicillin-clavulanate, cefuroxime, cefotaxime, ceftriaxone, ceftizoxime, cefixime, cefpodoxime, ceftibuten, cefdinir, or ertapenem) or antipseudomonals (piperacillin-tazobactam, ticarcillin-clavulanate, ceftazidime, cefepime, meropenem, doripenem, or imipenem).

Figure 2.  Relative Distribution of Propensity Scores for Treatment With Anti–Methicillin-Resistant Staphylococcus aureus (MRSA) Therapy
Relative Distribution of Propensity Scores for Treatment With Anti–Methicillin-Resistant Staphylococcus aureus (MRSA) Therapy

Conditional density curves demonstrating relative distributions of propensity scores for treatment with anti-MRSA therapy with and without standard antibiotics.

Figure 3.  Patient Characteristics Before and After Inverse Probability of Treatment Weighting
Patient Characteristics Before and After Inverse Probability of Treatment Weighting

Balance plot of patient characteristics for primary analysis before and after inverse probability of treatment weighting using 41 patient characteristics. Dots represent the maximum of pairwise absolute standardized mean difference between 3 treatment groups for each of the covariates. For each variable, lines connect dots for the same variable with and without weighting. BUN indicates blood urea nitrogen.

Table 1.  Patient Characteristics and Outcomes by Treatment Group
Patient Characteristics and Outcomes by Treatment Group
Table 2.  Adjusted Risk Ratios for 30-Day Mortality Among Primary and Subgroup Inverse Probability–Weighted Analyses
Adjusted Risk Ratios for 30-Day Mortality Among Primary and Subgroup Inverse Probability–Weighted Analyses
1.
Centers for Disease Control and Prevention, National Center for Health Statistics. About underlying cause of death, 1999-2017. https://wonder.cdc.gov/ucd-icd10.html. Accessed October 23, 2017.
2.
Jain  S, Self  WH, Wunderink  RG; CDC EPIC Study Team.  Community-acquired pneumonia requiring hospitalization.  N Engl J Med. 2015;373(24):2382. doi:10.1056/nejmc1511751PubMedGoogle Scholar
3.
Mandell  LA, Wunderink  RG, Anzueto  A,  et al; Infectious Diseases Society of America; American Thoracic Society.  Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults.  Clin Infect Dis. 2007;44(suppl 2):S27-S72. doi:10.1086/511159PubMedGoogle ScholarCrossref
4.
Wunderink  RG.  Community-acquired pneumonia versus healthcare-associated pneumonia: the returning pendulum.  Am J Respir Crit Care Med. 2013;188(8):896-898. doi:10.1164/rccm.201308-1536EDPubMedGoogle ScholarCrossref
5.
Berger  A, Edelsberg  J, Oster  G, Huang  X, Weber  DJ.  Patterns of initial antibiotic therapy for community-acquired pneumonia in U.S. hospitals, 2000 to 2009.  Am J Med Sci. 2014;347(5):347-356. doi:10.1097/MAJ.0b013e318294833fPubMedGoogle ScholarCrossref
6.
Jones  BE, Jones  MM, Huttner  B,  et al.  Trends in antibiotic use and nosocomial pathogens in hospitalized veterans with pneumonia at 128 medical centers, 2006-2010.  Clin Infect Dis. 2015;61(9):1403-1410. doi:10.1093/cid/civ629PubMedGoogle ScholarCrossref
7.
Aliberti  S, Cilloniz  C, Chalmers  JD,  et al.  Multidrug-resistant pathogens in hospitalised patients coming from the community with pneumonia: a European perspective.  Thorax. 2013;68(11):997-999. doi:10.1136/thoraxjnl-2013-203384PubMedGoogle ScholarCrossref
8.
Shorr  AF, Zilberberg  MD, Micek  ST, Kollef  MH.  Predictors of hospital mortality among septic ICU patients with Acinetobacter spp. bacteremia: a cohort study.  BMC Infect Dis. 2014;14:572. doi:10.1186/s12879-014-0572-6PubMedGoogle ScholarCrossref
9.
Shindo  Y, Ito  R, Kobayashi  D,  et al.  Risk factors for drug-resistant pathogens in community-acquired and healthcare-associated pneumonia.  Am J Respir Crit Care Med. 2013;188(8):985-995. doi:10.1164/rccm.201301-0079OCPubMedGoogle ScholarCrossref
10.
Webb  BJ, Jones  B, Dean  NC.  Empiric antibiotic selection and risk prediction of drug-resistant pathogens in community-onset pneumonia.  Curr Opin Infect Dis. 2016;29(2):167-177. doi:10.1097/QCO.0000000000000254PubMedGoogle ScholarCrossref
11.
Jones  M, Huttner  B, Leecaster  M,  et al.  Does universal active MRSA surveillance influence anti-MRSA antibiotic use? a retrospective analysis of the treatment of patients admitted with suspicion of infection at Veterans Affairs Medical Centers between 2005 and 2010.  J Antimicrob Chemother. 2014;69(12):3401-3408. doi:10.1093/jac/dku299PubMedGoogle ScholarCrossref
12.
Dangerfield  B, Chung  A, Webb  B, Seville  MT.  Predictive value of methicillin-resistant Staphylococcus aureus (MRSA) nasal swab PCR assay for MRSA pneumonia.  Antimicrob Agents Chemother. 2014;58(2):859-864. doi:10.1128/AAC.01805-13PubMedGoogle ScholarCrossref
13.
Paonessa  JR, Shah  RD, Pickens  CI,  et al.  Rapid detection of methicillin-resistant Staphylococcus aureus in BAL: a pilot randomized controlled trial.  Chest. 2019;155(5):999-1007. doi:10.1016/j.chest.2019.02.007PubMedGoogle ScholarCrossref
14.
Metlay  JP, Waterer  GW, Long  AC,  et al.  Diagnosis and treatment of adults with community-acquired pneumonia: an official clinical practice guideline of the American Thoracic Society and Infectious Diseases Society of America.  Am J Respir Crit Care Med. 2019;200(7):e45-e67. doi:10.1164/rccm.201908-1581STPubMedGoogle ScholarCrossref
15.
Attridge  RT, Frei  CR, Restrepo  MI,  et al.  Guideline-concordant therapy and outcomes in healthcare-associated pneumonia.  Eur Respir J. 2011;38(4):878-887. doi:10.1183/09031936.00141110PubMedGoogle ScholarCrossref
16.
Webb  BJ, Sorensen  J, Jephson  A, Mecham  I, Dean  NC.  Broad-spectrum antibiotic use and poor outcomes in community-onset pneumonia: a cohort study.  Eur Respir J. 2019;54(1):1900057. doi:10.1183/13993003.00057-2019PubMedGoogle Scholar
17.
Jones  BE, Brown  KA, Jones  MM,  et al.  Variation in empiric coverage versus detection of methicillin-resistant Staphylococcus aureus and Pseudomonas aeruginosa in hospitalizations for community-onset pneumonia across 128 US Veterans Affairs Medical Centers.  Infect Control Hosp Epidemiol. 2017;38(8):937-944. doi:10.1017/ice.2017.98PubMedGoogle ScholarCrossref
18.
Jones  BE, Haroldsen  C, Madaras-Kelly  K,  et al.  In data we trust? comparison of electronic versus manual abstraction of antimicrobial prescribing quality metrics for hospitalized veterans with pneumonia.  Med Care. 2018;56(7):626-633. doi:10.1097/MLR.0000000000000916PubMedGoogle ScholarCrossref
19.
Lindenauer  PK, Lagu  T, Shieh  MS, Pekow  PS, Rothberg  MB.  Association of diagnostic coding with trends in hospitalizations and mortality of patients with pneumonia, 2003-2009.  JAMA. 2012;307(13):1405-1413. doi:10.1001/jama.2012.384PubMedGoogle ScholarCrossref
20.
Fine  MJ, Auble  TE, Yealy  DM,  et al.  A prediction rule to identify low-risk patients with community-acquired pneumonia.  N Engl J Med. 1997;336(4):243-250. doi:10.1056/NEJM199701233360402PubMedGoogle ScholarCrossref
21.
Rubin  DB.  The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials.  Stat Med. 2007;26(1):20-36. doi:10.1002/sim.2739PubMedGoogle ScholarCrossref
22.
Hernán  MA, Alonso  A, Logan  R,  et al.  Observational studies analyzed like randomized experiments: an application to postmenopausal hormone therapy and coronary heart disease.  Epidemiology. 2008;19(6):766-779. doi:10.1097/EDE.0b013e3181875e61PubMedGoogle ScholarCrossref
23.
McCaffrey  DF, Griffin  BA, Almirall  D, Slaughter  ME, Ramchand  R, Burgette  LF.  A tutorial on propensity score estimation for multiple treatments using generalized boosted models.  Stat Med. 2013;32(19):3388-3414. doi:10.1002/sim.5753PubMedGoogle ScholarCrossref
24.
Ridgeway  G, McCaffrey  DF, Morral  AR, Burgette  LF, Griffin  BA.  Toolkit for Weighting and Analysis of Nonequivalent Groups: A Tutorial for the R TWANG Package. Santa Monica, CA: RAND Corporation; 2014.
25.
Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706. doi:10.1093/aje/kwh090PubMedGoogle ScholarCrossref
26.
Jones  M, Jernigan  JA, Evans  ME, Roselle  GA, Hatfield  KM, Samore  MH.  Vital signs: trends in Staphylococcus aureus infections in Veterans Affairs Medical Centers—United States, 2005-2017.  MMWR Morb Mortal Wkly Rep. 2019;68(9):220-224. doi:10.15585/mmwr.mm6809e2PubMedGoogle ScholarCrossref
27.
Terza  JV.  Two-stage residual inclusion estimation in health services research and health economics.  Health Serv Res. 2018;53(3):1890-1899. doi:10.1111/1475-6773.12714PubMedGoogle ScholarCrossref
28.
Pizer  SD.  Falsification testing of instrumental variables methods for comparative effectiveness research.  Health Serv Res. 2016;51(2):790-811. doi:10.1111/1475-6773.12355PubMedGoogle ScholarCrossref
29.
Swanson  SA, Hernán  MA.  Commentary: how to report instrumental variable analyses (suggestions welcome).  Epidemiology. 2013;24(3):370-374. doi:10.1097/EDE.0b013e31828d0590PubMedGoogle ScholarCrossref
30.
Attridge  RT, Frei  CR, Pugh  MJ,  et al.  Health care–associated pneumonia in the intensive care unit: guideline-concordant antibiotics and outcomes.  J Crit Care. 2016;36:265-271. doi:10.1016/j.jcrc.2016.08.004PubMedGoogle ScholarCrossref
31.
Cano  EL, Haque  NZ, Welch  VL,  et al; Improving Medicine through Pathway Assessment of Critical Therapy of Hospital-Acquired Pneumonia (IMPACT-HAP) Study Group.  Incidence of nephrotoxicity and association with vancomycin use in intensive care unit patients with pneumonia: retrospective analysis of the IMPACT-HAP Database.  Clin Ther. 2012;34(1):149-157. doi:10.1016/j.clinthera.2011.12.013PubMedGoogle ScholarCrossref
32.
Jeffres  MN, Isakow  W, Doherty  JA, Micek  ST, Kollef  MH.  A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care–associated methicillin-resistant Staphylococcus aureus pneumonia.  Clin Ther. 2007;29(6):1107-1115. doi:10.1016/j.clinthera.2007.06.014PubMedGoogle ScholarCrossref
33.
Lodise  TP, Rosenkranz  SL, Finnemeyer  M,  et al; Antibacterial Resistance Leadership Group.  The emperor’s new clothes: prospective observational evaluation of the association between initial vancomycin exposure and failure rates among adult hospitalized patients with MRSA bloodstream infections (PROVIDE)  [published online June 3, 2019].  Clin Infect Dis. doi:10.1093/cid/ciz460PubMedGoogle Scholar
34.
Ewig  S, Schlochtermeier  M, Göke  N, Niederman  MS.  Applying sputum as a diagnostic tool in pneumonia: limited yield, minimal impact on treatment decisions.  Chest. 2002;121(5):1486-1492. doi:10.1378/chest.121.5.1486PubMedGoogle ScholarCrossref
35.
Madaras-Kelly  K, Jones  M, Remington  R,  et al.  Antimicrobial de-escalation of treatment for healthcare-associated pneumonia within the Veterans Healthcare Administration.  J Antimicrob Chemother. 2016;71(2):539-546. doi:10.1093/jac/dkv338PubMedGoogle ScholarCrossref
36.
Almirall  D, Nahum-Shani  I, Sherwood  NE, Murphy  SA.  Introduction to SMART designs for the development of adaptive interventions: with application to weight loss research.  Transl Behav Med. 2014;4(3):260-274. doi:10.1007/s13142-014-0265-0PubMedGoogle ScholarCrossref
37.
Chakraborty  B, Murphy  SA.  Dynamic treatment regimes.  Annu Rev Stat Appl. 2014;1:447-464. doi:10.1146/annurev-statistics-022513-115553PubMedGoogle ScholarCrossref
38.
Jones  BE, South  BR, Shao  Y,  et al.  Development and validation of a natural language processing tool to identify patients treated for pneumonia across VA emergency departments.  Appl Clin Inform. 2018;9(1):122-128. doi:10.1055/s-0038-1626725PubMedGoogle ScholarCrossref
39.
Kaplan  V, Angus  DC, Griffin  MF, Clermont  G, Scott Watson  R, Linde-Zwirble  WT.  Hospitalized community-acquired pneumonia in the elderly: age- and sex-related patterns of care and outcome in the United States.  Am J Respir Crit Care Med. 2002;165(6):766-772. doi:10.1164/ajrccm.165.6.2103038PubMedGoogle ScholarCrossref
40.
Marik  PE.  Obituary: pulmonary artery catheter 1970 to 2013.  Ann Intensive Care. 2013;3(1):38. doi:10.1186/2110-5820-3-38PubMedGoogle ScholarCrossref
41.
Poole  D, Bertolini  G, Garattini  S.  Withdrawal of “Xigris” from the market: old and new lessons.  J Epidemiol Community Health. 2012;66(7):571-572. doi:10.1136/jech-2012-200977PubMedGoogle ScholarCrossref
42.
Rowan  KM, Angus  DC, Bailey  M,  et al; PRISM Investigators.  Early, goal-directed therapy for septic shock—a patient-level meta-analysis.  N Engl J Med. 2017;376(23):2223-2234. doi:10.1056/NEJMoa1701380PubMedGoogle ScholarCrossref
43.
Wachter  RM, Flanders  SA, Fee  C, Pronovost  PJ.  Public reporting of antibiotic timing in patients with pneumonia: lessons from a flawed performance measure.  Ann Intern Med. 2008;149(1):29-32. doi:10.7326/0003-4819-149-1-200807010-00007PubMedGoogle ScholarCrossref
44.
Qi  C, Hountras  P, Pickens  CO,  et al.  Detection of respiratory pathogens in clinical samples using metagenomic shotgun sequencing.  J Med Microbiol. 2019;68(7):996-1002. doi:10.1099/jmm.0.000968PubMedGoogle ScholarCrossref
45.
Blaschke  AJ, Hersh  AL, Beekmann  SE, Ince  D, Polgreen  PM, Hanson  KE.  Unmet diagnostic needs in infectious disease.  Diagn Microbiol Infect Dis. 2015;81(1):57-59. doi:10.1016/j.diagmicrobio.2014.10.005PubMedGoogle ScholarCrossref
46.
Messacar  K, Parker  SK, Todd  JK, Dominguez  SR.  Implementation of rapid molecular infectious disease diagnostics: the role of diagnostic and antimicrobial stewardship.  J Clin Microbiol. 2017;55(3):715-723. doi:10.1128/JCM.02264-16PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    1 Comment for this article
    EXPAND ALL
    RE: Empirical Anti-MRSA vs Standard Antibiotic Therapy and Risk of 30-Day Mortality in Patients Hospitalized for Pneumonia
    Rajiv Kumar, MBBS, MD. | Faculty, Dept. of Pharmacology, Government Medical College & Hospital, Chandigarh, 160030. India.
    The authors rightly mentioned that empirical anti-MRSA antibiotic therapy was not supported in previous studies and this study also confirms the same.

    Although the Infectious Diseases Society of America and the American Thoracic Society have guidelines for the management of pneumonia / community-acquired pneumonia,  it is all about careful, judicious and rational use of antibiotics.

    The decision to use empirical anti-MRSA or standard antibiotics  depends on clinical judgment, and it varies from patient to patient.
     
    It is essential for physicians and other clinicians to promote respiratory hygiene measures and cough etiquette
    to reduce transmission of respiratory infections in patients and for good clinical recovery.

    Regards,

    Dr.Rajiv Kumar, Dr.Sangeeta Bhanwra. Faculty Dept. of Pharmacology, Government Medical College & Hospital, Chandigarh, 160030. India.

    DRrajiv.08@gmail.com
    CONFLICT OF INTEREST: None Reported
    READ MORE
    Original Investigation
    February 17, 2020

    Empirical Anti-MRSA vs Standard Antibiotic Therapy and Risk of 30-Day Mortality in Patients Hospitalized for Pneumonia

    Author Affiliations
    • 1Division of Pulmonary and Critical Care, Veterans Affairs Salt Lake City Health Care System, Salt Lake City, Utah
    • 2University of Utah, Salt Lake City
    • 3Division of Epidemiology, University of Utah, Salt Lake City
    • 4Division of Epidemiology, Veterans Affairs Salt Lake City Health Care System, Salt Lake City, Utah
    • 5Division of Internal Medicine, University of Utah, Salt Lake City
    • 6Department of Health Economics and Epidemiology, Veterans Affairs Salt Lake City Health Care System, Salt Lake City, Utah
    • 7Department of Internal Medicine, Veterans Affairs Salt Lake City Health Care System, Salt Lake City, Utah
    • 8Division of Infectious Disease, Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, California
    JAMA Intern Med. 2020;180(4):552-560. doi:10.1001/jamainternmed.2019.7495
    Key Points

    Question  What is the association of empirical anti–methicillin-resistant Staphylococcus aureus therapy with 30-day mortality for patients hospitalized with pneumonia?

    Findings  This national cohort study of 88 605 hospitalizations for pneumonia that used detailed clinical data to emulate a clinical trial did not find a mortality benefit of empirical anti–methicillin-resistant S aureus therapy vs standard antibiotics for any group of patients examined, even those with risk factors for methicillin-resistant S aureus.

    Meaning  This study contributes to a growing body of evidence suggesting that empirical anti–methicillin-resistant S aureus therapy using existing risk approaches may not be beneficial to most patients hospitalized with pneumonia.

    Abstract

    Importance  Use of empirical broad-spectrum antibiotics for pneumonia has increased owing to concern for resistant organisms, including methicillin-resistant Staphylococcus aureus (MRSA). The association of empirical anti-MRSA therapy with outcomes among patients with pneumonia is unknown, even for high-risk patients.

    Objective  To compare 30-day mortality among patients hospitalized for pneumonia receiving empirical anti-MRSA therapy vs standard empirical antibiotic regimens.

    Design, Setting, and Participants  Retrospective multicenter cohort study was conducted of all hospitalizations in which patients received either anti-MRSA or standard therapy for community-onset pneumonia in the Veterans Health Administration health care system from January 1, 2008, to December 31, 2013. Subgroups of patients analyzed were those with initial intensive care unit admission, MRSA risk factors, positive results of a MRSA surveillance test, and positive results of a MRSA admission culture. Primary analysis was an inverse probability of treatment–weighted propensity score analysis using generalized estimating equation regression; secondary analyses included an instrumental variable analysis. Statistical analysis was conducted from June 14 to November 20, 2019.

    Exposures  Empirical anti-MRSA therapy plus standard pneumonia therapy vs standard therapy alone within the first day of hospitalization.

    Main Outcomes and Measures  Risk of 30-day all-cause mortality after adjustment for patient comorbidities, vital signs, and laboratory results. Secondary outcomes included the development of kidney injury and secondary infections with Clostridioides difficile, vancomycin-resistant Enterococcus species, or gram-negative bacilli.

    Results  Among 88 605 hospitalized patients (86 851 men; median age, 70 years [interquartile range, 62-81 years]), empirical anti-MRSA therapy was administered to 33 632 (38%); 8929 patients (10%) died within 30 days. Compared with standard therapy alone, in weighted propensity score analysis, empirical anti-MRSA therapy plus standard therapy was significantly associated with an increased adjusted risk of death (adjusted risk ratio [aRR], 1.4 [95% CI, 1.3-1.5]), kidney injury (aRR, 1.4 [95% CI, 1.3-1.5]), and secondary C difficile infections (aRR, 1.6 [95% CI, 1.3-1.9]), vancomycin-resistant Enterococcus spp infections (aRR, 1.6 [95% CI, 1.0-2.3]), and secondary gram-negative rod infections (aRR, 1.5 [95% CI, 1.2-1.8]). Similar associations between anti-MRSA therapy use and 30-day mortality were found by instrumental variable analysis (aRR, 1.6 [95% CI, 1.4-1.9]) and among patients admitted to the intensive care unit (aRR, 1.3 [95% CI, 1.2-1.5]), those with a high risk for MRSA (aRR, 1.2 [95% CI, 1.1-1.4]), and those with MRSA detected on surveillance testing (aRR, 1.6 [95% CI, 1.3-1.9]). No significant favorable association was found between empirical anti-MRSA therapy and death among patients with MRSA detected on culture (aRR, 1.1 [95% CI, 0.8-1.4]).

    Conclusions and Relevance  This study suggests that empirical anti-MRSA therapy was not associated with reduced mortality for any group of patients hospitalized for pneumonia. These results contribute to a growing body of evidence that questions the value of empirical use of anti-MRSA therapy using existing risk approaches.

    Introduction

    Pneumonia is the leading cause of death from infection in the United States,1 and timely empirical antibiotic therapy against the most likely pathogens is a cornerstone of care. However, causative pathogens are rarely identified,2 leaving uncertainty in the choice of empirical antibiotic therapy. For patients hospitalized for community-onset pneumonia, this uncertainty has been magnified by the emergence of resistant organisms, chiefly methicillin-resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa. The concept of health care–associated pneumonia, intended to assist clinicians in risk assessment,3 likely contributed to the overuse of broad-spectrum antibiotics.4,5 Consequently, although fewer than 5% of hospitalized patients have resistant organisms detected, more than one-third receive broad-spectrum antibiotics.6

    Efforts to enhance clinical prediction of resistant organisms have resulted in several new promising results.7-10 For patients at risk of MRSA infection, the use of molecular diagnostic testing with the nasal polymerase chain reaction (PCR) surveillance test has also been proposed as a potential tool to stratify patients.11-13 However, no single strategy has proved to substantially enhance decision accuracy,7 and new clinical practice guidelines emphasize the need for validation of these approaches.14 Some studies have suggested that empirical broad-spectrum therapy may be harmful.15,16 It is thus unclear which patients benefit sufficiently from empirical treatment with broad-spectrum agents to warrant such therapy.

    The primary objective of this study was to evaluate 30-day risk of death in patients hospitalized for pneumonia who were receiving empirical anti-MRSA therapy plus guideline-concordant (ie, standard) antibiotics compared with those receiving standard therapy alone among groups of patients who may warrant therapy. The Veterans Health Administration comprises a large integrated health care system with a shared electronic health record and clinical data repository for more than 5 million veterans at 140 medical centers. Previous studies have reported substantial variation in the decision to use empirical anti-MRSA therapy across and within facilities17 and year6 that was unexplained by differences in patient characteristics. We leveraged this variation to examine risks of 30-day mortality among all patients as well as subgroups that may expect greater benefit from an empirical anti-MRSA strategy, including those initially admitted to the intensive care unit (ICU), those with a history of MRSA infection or colonization or other clinical risk factors, those with MRSA detected on results of nasal PCR, and those with MRSA detected by culture within 2 days of admission.

    Methods
    Study Design, Setting, and Participation

    We conducted a retrospective cohort study of all hospitalizations for community-onset pneumonia in the Veterans Affairs (VA) health care system from January 1, 2008, to December 31, 2013 (Figure 1), using an existing data set that contained extensively validated clinical data18 and demonstrated sufficient variation in treatment17 to allow for comparative effectiveness research. We identified hospitalizations in acute inpatient wards with a principal International Classification of Diseases, Ninth Revision (ICD-9) code for pneumonia or secondary ICD-9 code for pneumonia with a principal ICD-9 code for sepsis and respiratory failure, a case-finding approach that has been found to be resilient to variation in diagnostic coding.17,19 Patients were excluded if they were not administered an antimicrobial within the first calendar day of hospitalization, were hospitalized with pneumonia in the previous month, or were transferred from other acute care facilities. The study was reviewed and approved, and waivers of consent were granted on the basis of infeasibility and minimal risk of harm to participants, by the University of Utah Institutional Review Board and the Research and Development Committee of the VA Salt Lake City Health Care System.

    Study Data and Measurements

    The primary exposure of interest was treatment with anti-MRSA therapy (vancomycin hydrochloride or linezolid) plus guideline-recommended standard antibiotics (β-lactam and macrolide or tetracycline hydrochloride, or fluoroquinolone)14,20 vs standard therapy alone. Because many patients received anti-MRSA therapy without standard antibiotics, we also evaluated this strategy as an additional treatment group. To conduct an analysis from observational data similar to an intention-to-treat clinical trial,21,22 we classified all patients according to the treatment they received on the first calendar day of hospitalization. Medication administration was captured using the VA standardized barcode medication administration data, which was previously validated against manual medical record review to accurately represent empirical therapy.18

    The primary outcome of interest was all-cause mortality within 30 days of hospitalization. Death data were obtained from the VA Vital Status file. Patient demographics, clinical risk factors for resistant organisms, and features associated with illness severity were extracted and curated for each hospitalization using a previously validated approach.18 Comorbidities included age, sex, renal disease, liver disease, congestive heart failure, cerebrovascular disease, neoplastic disease, immunocompromise (including HIV, solid organ transplant, neutropenia, and immunosuppressive therapy), residence at a nursing home, wound care, number of hospitalization days in the past 60 days, antibiotic therapy in the past 60 days, history of MRSA colonization or infection in the past year, and receipt of tube feeding. Acute illness severity features included vital signs and laboratory test results routinely collected from patients with pneumonia. We extracted the first laboratory test results (blood urea nitrogen, creatinine, glucose, potassium, sodium, white blood cell count, hematocrit, platelet count, albumin, bilirubin, arterial pH, arterial PaO2, lactate, and troponin) between 12 hours before and 24 hours after admission. Because the patient’s initial trajectory, best reflected by changing vital signs, can influence both the initial treatment decision and final outcome, we extracted the worst vital signs (minimum and maximum heart rate, minimum systolic blood pressure, maximum respiratory rate, and minimum oxygen saturation) within 12 hours prior to the time of treatment assignment (antibiotic administration). Missing vital signs, which occurred in less than 5% of all cases, were treated as missing at random and imputed with median values. All missing laboratory test values were imputed with normal values; we also created indicator variables when laboratory test values were missing that had more than 5% prevalence of missingness in the population (albumin, bilirubin, lactate, arterial pH, and troponin). We extracted MRSA culture and PCR surveillance data occurring within the first 2 calendar days and year prior to the admission. We defined a clinically relevant positive MRSA culture as one occurring within the first 2 calendar days from the day of admission from blood or a respiratory source (sputum, endotracheal aspirate, bronchoalveolar lavage, or pleura), as previously described.17

    Statistical Analysis

    Statistical analysis was conducted from June 14 to November 20, 2019. Our primary analysis was an inverse probability of treatment propensity score–weighted analysis that estimated the mean treatment effect for the entire population of the 3 defined empirical treatments—standard therapy, anti-MRSA therapy plus standard therapy, and anti-MRSA therapy without standard therapy—on 30-day mortality after controlling for patient characteristics potentially associated with both the propensity of treatment and the risk of 30-day mortality. To implement this approach, we first computed propensity scores for the 2 anti-MRSA treatments based on 41 patient characteristics as covariates, including all extracted comorbidities, vital signs, and laboratory test values mentioned above. The propensity scores were estimated by applying generalized boosted machine-learning models as described by McCaffrey et al23 to minimize the maximum standardized mean differences in the covariates between the 3 treatment groups using the twang package in the R statistical computing environment.24 The distributions of propensity scores in the 3 treatment groups were visually inspected for the degree of common support between patients prior to weighting (Figure 2). The balance of characteristics between treatment groups after weighting was considered adequate if standardized differences were less than 0.2 (Figure 3). We then fit an inverse propensity score–weighted regression with generalized estimating equation that accounted for clustering of patients within facility using independent working covariance matrices under modified Poisson regression models25 to estimate population-average adjusted risk ratios (aRRs) of 30-day mortality for the 2 anti-MRSA treatment groups compared with standard therapy as the control group. We hereafter use weighted propensity score analysis to refer to this full sequence of analyses.

    Subgroup Analyses

    To explore associations between treatment and death for different patient groups, we applied the same weighted propensity score analysis to 4 subgroups of patients who might warrant empirical anti-MRSA treatment owing to (1) initial admission to the ICU, (2) clinical risk for MRSA detection, (3) positive initial results of MRSA PCR surveillance screening, and (4) positive results of a clinical culture for MRSA (detected on results of blood or respiratory culture within 48 hours). We defined clinical risk for MRSA as a history of MRSA infection or colonization in the past year or at least 2 of the following: previous hospitalization, nursing home residence, and previous intravenous antibiotic therapy, which was based on a previous examination of these risk factors in the VA population17 as well as a revalidation in the study cohort (eAppendix 4 in the Supplement).

    Secondary Analyses

    Even after controlling for measured patient characteristics in the weighted propensity score analysis, it is possible that residual confounding owing to unmeasured patient characteristics (ie, mental status or radiographic findings) may bias our estimates of the association of anti-MRSA therapy with mortality. In particular, physicians may have disproportionately assigned anti-MRSA therapy to patients whom they perceived to be at greater risk for infection and death; this perception may have been influenced by factors unavailable in the electronic health record. In view of this risk, we capitalized on the variation in practice patterns for use of anti-MRSA therapy across facilities and over different years that was unexplained by patient characteristics or prevalence6,17,26 to perform an instrumental variable analysis. The proportion of hospitalizations with empirical anti-MRSA therapy for each facility and year was treated as an instrument to evaluate the association of empirical anti-MRSA therapy with 30-day mortality. We implemented the instrumental variable analysis using a 2-stage residual inclusion approach.27,28 We also adjusted for patient characteristics as covariates to control for confounding at the facility and year level. Although residual confounding is also possible for instrumental variable analysis, the instrumental variable analysis should avoid the type of confounding that results from use of anti-MRSA therapy for individual patients who are perceived to be at greater risk. A full description of this analysis, including assessments of its underlying assumptions,28,29 is available in eAppendix 2, eTable 2, eFigure 6, and eAppendix 3 in the Supplement.

    Sensitivity Analysis Incorporating Antipseudomonal Antibiotics

    Because receipt of empirical anti-MRSA therapy often coincides with receipt of antipseudomonal therapy, which may also have an association with outcomes, we estimated the association of anti-MRSA therapy and antipseudomonal therapy separately with 30-day mortality by applying a weighted propensity score analysis to compare patients receiving anti-MRSA therapy alone, antipseudomonal therapy alone, and no anti-MRSA or antipseudomonal therapy as 3 treatment groups. The analysis is similar to the primary analysis except that we added receipt of standard therapy as a covariate in the outcome model. Statistical analyses were performed using SAS, version 9.2 (SAS Institute Inc), Stata, version 16.0 (StataCorp LLC), and R (R Foundation for Statistical Computing; http://cran.r-project.org) software.

    Examination of Secondary Outcomes

    The weighted propensity score analysis was applied to the following secondary events occurring between 48 hours and 30 days after hospitalization: kidney injury (defined as an increase in creatinine of 0.3 mg/dL [to convert to micromoles per liter, multiply by 88.4] or 50% from initial creatinine level), incident or recurrent Clostridioides difficile infection (detection of toxin without previous positive test results for toxin in the past 14 days), and detection of vancomycin-resistant Enterococcus spp and gram-negative rods in blood or urine cultures.

    Results

    A total of 88 605 hospitalizations for pneumonia were studied (Figure 1), with a 30-day all-cause mortality of 10% (n = 8929). Empirical anti-MRSA therapy was administered to 33 632 patients (38%). Of these, 13 528 received empirical anti-MRSA therapy plus standard antibiotics, 20 104 received empirical anti-MRSA therapy without standard antibiotics, and 54 973 received empirical standard guideline-recommended therapy alone (Table 1).20

    Patients receiving empirical anti-MRSA therapy demonstrated a greater comorbidity burden (renal disease, 29% vs 25%; congestive heart failure, 35% vs 30%; neoplastic disease, 34% vs 3%; and nursing home residents, 9% vs 3%), more risk factors for MRSA (7% vs 2% with history of MRSA infection, 36% vs 12% with previous hospitalization, and 42% vs 29% with previous antibiotics), and greater illness severity (median Pneumonia Severity Index, 124 [interquartile range, 95-156] vs 103 [interquartile range, 81-131]) as well as worse outcomes (16% vs 6% for 30-day all-cause mortality) compared with patients receiving standard therapy alone (Table 1).20 However, the distribution of propensity for treatment demonstrated sufficient overlap between the treatment groups (Figure 2A), and weighting resulted in sufficient balance in patient characteristics for all 3 pairwise comparisons (Figure 2B).

    Empirical anti-MRSA treatment was significantly associated with greater 30-day mortality compared with standard therapy alone, with a propensity score–weighted aRR of 1.4 (95% CI, 1.3-1.5) for empirical anti-MRSA treatment plus standard therapy and 1.5 (1.4-1.6) for empirical anti-MRSA treatment with nonstandard therapy (Table 2). The corresponding propensity score–weighted marginal probabilities of 30-day mortality were 11.6% for empirical anti-MRSA treatment plus standard therapy and 12.7% for empirical anti-MRSA treatment with nonstandard therapy compared with 8.6% for standard therapy alone.

    Subgroup Analyses

    Among all hospitalizations, 14 370 patients (16%) were initially admitted to the ICU, 19 045 (22%) had clinical risk factors for MRSA, 2775 (3%) had positive PCR results, and 2154 (2%) had MRSA detected by clinical culture. Sufficient common support resulted in adequate balance in patient characteristics after weighting for all subgroups (eAppendix 1; eTable 1; and eFigures 1, 2, 3, 4, and 5 in the Supplement).

    We found a significant increase in 30-day mortality associated with empirical anti-MRSA therapy plus standard therapy compared with standard therapy alone among patients admitted to the ICU (aRR, 1.3; 95% CI, 1.2-1.5), with a high clinical risk for MRSA (aRR, 1.2; 95% CI, 1.1-1.4), and with positive results of surveillance PCR (aRR, 1.6; 95% CI, 1.3-1.9) but no significant difference in risk of 30-day mortality for patients with positive results of clinical culture (aRR, 1.1; 95% CI, 0.8-1.4 [Table 2]). Similar associations were found for the group receiving anti-MRSA therapy without standard therapy (Table 2).

    Instrumental Variable Analysis, Analysis of Antipseudomonal Antibiotics, and Secondary Outcomes

    Results of secondary analyses suggested similar associations. In the instrumental variable analysis, we found a significant association between use of anti-MRSA therapy and 30-day mortality (aRR, 1.6; 95% CI, 1.4-1.9). In the weighted propensity score analysis examining separate associations of empirical antipseudomonal therapy from anti-MRSA antibiotics, we found both therapies to be separately associated with higher risk of 30-day mortality after controlling for standard therapy (anti-MRSA therapy: aRR, 1.2; 95% CI, 1.1-1.3; antipseudomonal therapy: aRR, 1.3; 95% CI, 1.2-1.4). Use of empirical anti-MRSA therapy was associated with a higher risk of kidney injury (aRR, 1.4; 95% CI, 1.3-1.5), C difficile infection (aRR, 1.6; 95% CI, 1.3-1.9), vancomycin-resistant Enterococcus spp (aRR, 1.6; 95% CI, 1.0-2.3), and secondary gram-negative rod detection (aRR, 1.5; 95% CI, 1.2-1.8).

    Discussion

    In this national observational study of patients hospitalized for pneumonia using detailed clinical data, we were unable to establish benefit of empirical anti-MRSA therapy, even when risk factors for MRSA were present or clinical severity warranted admission to the ICU. These findings, which were robust to multiple methods of analysis, contribute to a growing body of evidence that raises questions surrounding widespread empirical use of extended-spectrum antibiotics in patients with community-acquired pneumonia.15,16,30

    These findings should be interpreted carefully. Estimates of treatment effects, whether generated from randomized trials or observational studies, are population means. Individual members of a population may vary widely in outcomes from different treatments. In the patient cohort that we analyzed, it is plausible that there were individuals who would have experienced a net clinical benefit from empirical receipt of an anti-MRSA regimen and others who would have experienced net harm. Potential sources of harm from vancomycin, which accounted for 98% of the anti-MRSA therapy in our study, include renal toxic effects, allergy, and superinfection.16,31-33 In our secondary analyses, anti-MRSA therapy was associated with increased risk of kidney injury and secondary infections. The influence of the decision to treat with anti-MRSA therapy on other antibiotic choices was another pathway that likely had an association with outcomes.

    Our evaluation of patients whose admission cultures grew MRSA was not a test of whether MRSA should be treated when it is isolated. Rather, our analysis addressed only the question of whether empirical therapy against MRSA was beneficial compared with standard empirical treatment. The strategy of adding anti-MRSA therapy once results of cultures were positive was not specifically examined. However, MRSA was most commonly isolated from sputum. A recognized limitation of respiratory cultures is that they often reflect oropharyngeal colonization.34 Thus, a contributing explanation for our results is that respiratory cultures may have poor positive predictive value for MRSA pneumonia. This finding calls into question whether respiratory cultures should be used as a criterion standard for infection in pneumonia and adds urgency to the need for better diagnostic tools to more precisely identify bacterial and viral causes of pneumonia and other infections.

    Future studies should extend the work presented here to examine treatment decisions during the postempirical treatment phase, such as deescalation,3,35 as well as dosing and therapeutic drug monitoring, which were not examined in our study. Valid approaches can draw causal inferences about sequential decisions that use time-varying information, such as sequential multiple assignment randomized trials36 and observational studies of dynamic treatment regimens.37 Identifying optimal antibiotic decision-making strategies for patients—including how best to integrate information from results of cultures and molecular diagnostic tests to make subsequent decisions about antibiotics after empirical therapy—merits further research.

    Limitations

    This observational study has limitations. While our large population size, variation, and detailed clinical data allowed us to compare outcomes for patients with similar measured illness severity, and while the instrumental analysis should provide some protection against bias by unmeasured severity, residual confounding is still possible. In our secondary analysis, antipseudomonal therapy was found to be associated with 30-day mortality. This finding warrants further investigation as may be suggested by the added association of concomitant therapy. Our case-finding approach is widely used but relied on diagnosis codes assigned at the end of the hospitalization; while adequate precision has been found in this approach,38 we may have included some patients who did not initially receive a diagnosis of pneumonia. We captured all antibiotics administered in the hospital, but an estimated 2% of patients in our population received a different antibiotic in the emergency department.18 Our population is disproportionately male, and although no differences in antibiotic effects have been reported, women have different outcome patterns from men in pneumonia.39 Similarly, our population had an insufficient number of patients receiving linezolid to compare its separate effects.

    Conclusions

    Clinicians are constantly seeking innovations that might promise better outcomes for our patients. However, our eagerness to improve outcomes, particularly for critically ill patients, makes us susceptible to adopt practices that may have plausibility and promise but lack significant evidence or validation.40-43 Once adopted, these practices become norms that persist despite cautionary studies. With a mortality rate that has not substantially improved in decades, the threat of resistant organisms, and the emphasis on timely antibiotics in sepsis, it is not surprising that the strategy of early broad-spectrum antibiotics for pneumonia has become the norm. The underlying assumption of this approach is that the benefit of more potent antibiotics during the empirical phase exceeds the harms. Our study questions this assumption. We hope that newer diagnostic approaches44-46 and more evidence informing antimicrobial decisions will enhance our ability to accurately treat our patients. In the meantime, administration of empirical anti-MRSA therapy for pneumonia using current approaches should be reconsidered, even in high-risk patients.

    Back to top
    Article Information

    Accepted for Publication: December 23, 2019.

    Corresponding Author: Barbara Ellen Jones, MD, MSc, Division of Pulmonary and Critical Care, Veterans Affairs Salt Lake City Health Care System, 50 N Medical Dr, Wintrobe 701, Salt Lake City, UT 84132 (barbara.jones@hsc.utah.edu).

    Published Online: February 17, 2020. doi:10.1001/jamainternmed.2019.7495

    Author Contributions: Dr B. E. Jones had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: B. E. Jones, Ying, Stevens, Sauer, M.M. Jones, Greene, Samore.

    Acquisition, analysis, or interpretation of data: B. E. Jones, Ying, Stevens, Haroldsen, He, Nevers, Christensen, Nelson, Stoddard, Sauer, Yarbrough, Goetz, Greene, Samore.

    Drafting of the manuscript: B. E. Jones, Stevens, Nevers, Christensen, M.M. Jones.

    Critical revision of the manuscript for important intellectual content: B. E. Jones, Ying, Stevens, Haroldsen, He, Nelson, Stoddard, Sauer, Yarbrough, Goetz, Greene, Samore.

    Statistical analysis: B. E. Jones, Ying, Stevens, Nevers, Nelson, Stoddard, Sauer, M.M. Jones, Greene, Samore.

    Obtained funding: Samore.

    Administrative, technical, or material support: B. E. Jones, Stevens, Christensen, Sauer.

    Supervision: Greene.

    Conflict of Interest Disclosures: Dr B. E. Jones reported receiving grants from the Centers for Disease Control and Prevention and Veterans Affairs Health Services Research & Development during the conduct of the study. Ms Nevers reported receiving grants from the Centers for Disease Control and Prevention during the conduct of the study. Dr Sauer reported receiving grants from the Veterans Affairs during the conduct of the study. Dr Greene reported receiving personal fees from Janssen Pharmaceuticals, DURECT Corporation, and Pfizer Inc and grants from AstraZeneca and CSL outside the submitted work. No other disclosures were reported.

    References
    1.
    Centers for Disease Control and Prevention, National Center for Health Statistics. About underlying cause of death, 1999-2017. https://wonder.cdc.gov/ucd-icd10.html. Accessed October 23, 2017.
    2.
    Jain  S, Self  WH, Wunderink  RG; CDC EPIC Study Team.  Community-acquired pneumonia requiring hospitalization.  N Engl J Med. 2015;373(24):2382. doi:10.1056/nejmc1511751PubMedGoogle Scholar
    3.
    Mandell  LA, Wunderink  RG, Anzueto  A,  et al; Infectious Diseases Society of America; American Thoracic Society.  Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults.  Clin Infect Dis. 2007;44(suppl 2):S27-S72. doi:10.1086/511159PubMedGoogle ScholarCrossref
    4.
    Wunderink  RG.  Community-acquired pneumonia versus healthcare-associated pneumonia: the returning pendulum.  Am J Respir Crit Care Med. 2013;188(8):896-898. doi:10.1164/rccm.201308-1536EDPubMedGoogle ScholarCrossref
    5.
    Berger  A, Edelsberg  J, Oster  G, Huang  X, Weber  DJ.  Patterns of initial antibiotic therapy for community-acquired pneumonia in U.S. hospitals, 2000 to 2009.  Am J Med Sci. 2014;347(5):347-356. doi:10.1097/MAJ.0b013e318294833fPubMedGoogle ScholarCrossref
    6.
    Jones  BE, Jones  MM, Huttner  B,  et al.  Trends in antibiotic use and nosocomial pathogens in hospitalized veterans with pneumonia at 128 medical centers, 2006-2010.  Clin Infect Dis. 2015;61(9):1403-1410. doi:10.1093/cid/civ629PubMedGoogle ScholarCrossref
    7.
    Aliberti  S, Cilloniz  C, Chalmers  JD,  et al.  Multidrug-resistant pathogens in hospitalised patients coming from the community with pneumonia: a European perspective.  Thorax. 2013;68(11):997-999. doi:10.1136/thoraxjnl-2013-203384PubMedGoogle ScholarCrossref
    8.
    Shorr  AF, Zilberberg  MD, Micek  ST, Kollef  MH.  Predictors of hospital mortality among septic ICU patients with Acinetobacter spp. bacteremia: a cohort study.  BMC Infect Dis. 2014;14:572. doi:10.1186/s12879-014-0572-6PubMedGoogle ScholarCrossref
    9.
    Shindo  Y, Ito  R, Kobayashi  D,  et al.  Risk factors for drug-resistant pathogens in community-acquired and healthcare-associated pneumonia.  Am J Respir Crit Care Med. 2013;188(8):985-995. doi:10.1164/rccm.201301-0079OCPubMedGoogle ScholarCrossref
    10.
    Webb  BJ, Jones  B, Dean  NC.  Empiric antibiotic selection and risk prediction of drug-resistant pathogens in community-onset pneumonia.  Curr Opin Infect Dis. 2016;29(2):167-177. doi:10.1097/QCO.0000000000000254PubMedGoogle ScholarCrossref
    11.
    Jones  M, Huttner  B, Leecaster  M,  et al.  Does universal active MRSA surveillance influence anti-MRSA antibiotic use? a retrospective analysis of the treatment of patients admitted with suspicion of infection at Veterans Affairs Medical Centers between 2005 and 2010.  J Antimicrob Chemother. 2014;69(12):3401-3408. doi:10.1093/jac/dku299PubMedGoogle ScholarCrossref
    12.
    Dangerfield  B, Chung  A, Webb  B, Seville  MT.  Predictive value of methicillin-resistant Staphylococcus aureus (MRSA) nasal swab PCR assay for MRSA pneumonia.  Antimicrob Agents Chemother. 2014;58(2):859-864. doi:10.1128/AAC.01805-13PubMedGoogle ScholarCrossref
    13.
    Paonessa  JR, Shah  RD, Pickens  CI,  et al.  Rapid detection of methicillin-resistant Staphylococcus aureus in BAL: a pilot randomized controlled trial.  Chest. 2019;155(5):999-1007. doi:10.1016/j.chest.2019.02.007PubMedGoogle ScholarCrossref
    14.
    Metlay  JP, Waterer  GW, Long  AC,  et al.  Diagnosis and treatment of adults with community-acquired pneumonia: an official clinical practice guideline of the American Thoracic Society and Infectious Diseases Society of America.  Am J Respir Crit Care Med. 2019;200(7):e45-e67. doi:10.1164/rccm.201908-1581STPubMedGoogle ScholarCrossref
    15.
    Attridge  RT, Frei  CR, Restrepo  MI,  et al.  Guideline-concordant therapy and outcomes in healthcare-associated pneumonia.  Eur Respir J. 2011;38(4):878-887. doi:10.1183/09031936.00141110PubMedGoogle ScholarCrossref
    16.
    Webb  BJ, Sorensen  J, Jephson  A, Mecham  I, Dean  NC.  Broad-spectrum antibiotic use and poor outcomes in community-onset pneumonia: a cohort study.  Eur Respir J. 2019;54(1):1900057. doi:10.1183/13993003.00057-2019PubMedGoogle Scholar
    17.
    Jones  BE, Brown  KA, Jones  MM,  et al.  Variation in empiric coverage versus detection of methicillin-resistant Staphylococcus aureus and Pseudomonas aeruginosa in hospitalizations for community-onset pneumonia across 128 US Veterans Affairs Medical Centers.  Infect Control Hosp Epidemiol. 2017;38(8):937-944. doi:10.1017/ice.2017.98PubMedGoogle ScholarCrossref
    18.
    Jones  BE, Haroldsen  C, Madaras-Kelly  K,  et al.  In data we trust? comparison of electronic versus manual abstraction of antimicrobial prescribing quality metrics for hospitalized veterans with pneumonia.  Med Care. 2018;56(7):626-633. doi:10.1097/MLR.0000000000000916PubMedGoogle ScholarCrossref
    19.
    Lindenauer  PK, Lagu  T, Shieh  MS, Pekow  PS, Rothberg  MB.  Association of diagnostic coding with trends in hospitalizations and mortality of patients with pneumonia, 2003-2009.  JAMA. 2012;307(13):1405-1413. doi:10.1001/jama.2012.384PubMedGoogle ScholarCrossref
    20.
    Fine  MJ, Auble  TE, Yealy  DM,  et al.  A prediction rule to identify low-risk patients with community-acquired pneumonia.  N Engl J Med. 1997;336(4):243-250. doi:10.1056/NEJM199701233360402PubMedGoogle ScholarCrossref
    21.
    Rubin  DB.  The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials.  Stat Med. 2007;26(1):20-36. doi:10.1002/sim.2739PubMedGoogle ScholarCrossref
    22.
    Hernán  MA, Alonso  A, Logan  R,  et al.  Observational studies analyzed like randomized experiments: an application to postmenopausal hormone therapy and coronary heart disease.  Epidemiology. 2008;19(6):766-779. doi:10.1097/EDE.0b013e3181875e61PubMedGoogle ScholarCrossref
    23.
    McCaffrey  DF, Griffin  BA, Almirall  D, Slaughter  ME, Ramchand  R, Burgette  LF.  A tutorial on propensity score estimation for multiple treatments using generalized boosted models.  Stat Med. 2013;32(19):3388-3414. doi:10.1002/sim.5753PubMedGoogle ScholarCrossref
    24.
    Ridgeway  G, McCaffrey  DF, Morral  AR, Burgette  LF, Griffin  BA.  Toolkit for Weighting and Analysis of Nonequivalent Groups: A Tutorial for the R TWANG Package. Santa Monica, CA: RAND Corporation; 2014.
    25.
    Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706. doi:10.1093/aje/kwh090PubMedGoogle ScholarCrossref
    26.
    Jones  M, Jernigan  JA, Evans  ME, Roselle  GA, Hatfield  KM, Samore  MH.  Vital signs: trends in Staphylococcus aureus infections in Veterans Affairs Medical Centers—United States, 2005-2017.  MMWR Morb Mortal Wkly Rep. 2019;68(9):220-224. doi:10.15585/mmwr.mm6809e2PubMedGoogle ScholarCrossref
    27.
    Terza  JV.  Two-stage residual inclusion estimation in health services research and health economics.  Health Serv Res. 2018;53(3):1890-1899. doi:10.1111/1475-6773.12714PubMedGoogle ScholarCrossref
    28.
    Pizer  SD.  Falsification testing of instrumental variables methods for comparative effectiveness research.  Health Serv Res. 2016;51(2):790-811. doi:10.1111/1475-6773.12355PubMedGoogle ScholarCrossref
    29.
    Swanson  SA, Hernán  MA.  Commentary: how to report instrumental variable analyses (suggestions welcome).  Epidemiology. 2013;24(3):370-374. doi:10.1097/EDE.0b013e31828d0590PubMedGoogle ScholarCrossref
    30.
    Attridge  RT, Frei  CR, Pugh  MJ,  et al.  Health care–associated pneumonia in the intensive care unit: guideline-concordant antibiotics and outcomes.  J Crit Care. 2016;36:265-271. doi:10.1016/j.jcrc.2016.08.004PubMedGoogle ScholarCrossref
    31.
    Cano  EL, Haque  NZ, Welch  VL,  et al; Improving Medicine through Pathway Assessment of Critical Therapy of Hospital-Acquired Pneumonia (IMPACT-HAP) Study Group.  Incidence of nephrotoxicity and association with vancomycin use in intensive care unit patients with pneumonia: retrospective analysis of the IMPACT-HAP Database.  Clin Ther. 2012;34(1):149-157. doi:10.1016/j.clinthera.2011.12.013PubMedGoogle ScholarCrossref
    32.
    Jeffres  MN, Isakow  W, Doherty  JA, Micek  ST, Kollef  MH.  A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care–associated methicillin-resistant Staphylococcus aureus pneumonia.  Clin Ther. 2007;29(6):1107-1115. doi:10.1016/j.clinthera.2007.06.014PubMedGoogle ScholarCrossref
    33.
    Lodise  TP, Rosenkranz  SL, Finnemeyer  M,  et al; Antibacterial Resistance Leadership Group.  The emperor’s new clothes: prospective observational evaluation of the association between initial vancomycin exposure and failure rates among adult hospitalized patients with MRSA bloodstream infections (PROVIDE)  [published online June 3, 2019].  Clin Infect Dis. doi:10.1093/cid/ciz460PubMedGoogle Scholar
    34.
    Ewig  S, Schlochtermeier  M, Göke  N, Niederman  MS.  Applying sputum as a diagnostic tool in pneumonia: limited yield, minimal impact on treatment decisions.  Chest. 2002;121(5):1486-1492. doi:10.1378/chest.121.5.1486PubMedGoogle ScholarCrossref
    35.
    Madaras-Kelly  K, Jones  M, Remington  R,  et al.  Antimicrobial de-escalation of treatment for healthcare-associated pneumonia within the Veterans Healthcare Administration.  J Antimicrob Chemother. 2016;71(2):539-546. doi:10.1093/jac/dkv338PubMedGoogle ScholarCrossref
    36.
    Almirall  D, Nahum-Shani  I, Sherwood  NE, Murphy  SA.  Introduction to SMART designs for the development of adaptive interventions: with application to weight loss research.  Transl Behav Med. 2014;4(3):260-274. doi:10.1007/s13142-014-0265-0PubMedGoogle ScholarCrossref
    37.
    Chakraborty  B, Murphy  SA.  Dynamic treatment regimes.  Annu Rev Stat Appl. 2014;1:447-464. doi:10.1146/annurev-statistics-022513-115553PubMedGoogle ScholarCrossref
    38.
    Jones  BE, South  BR, Shao  Y,  et al.  Development and validation of a natural language processing tool to identify patients treated for pneumonia across VA emergency departments.  Appl Clin Inform. 2018;9(1):122-128. doi:10.1055/s-0038-1626725PubMedGoogle ScholarCrossref
    39.
    Kaplan  V, Angus  DC, Griffin  MF, Clermont  G, Scott Watson  R, Linde-Zwirble  WT.  Hospitalized community-acquired pneumonia in the elderly: age- and sex-related patterns of care and outcome in the United States.  Am J Respir Crit Care Med. 2002;165(6):766-772. doi:10.1164/ajrccm.165.6.2103038PubMedGoogle ScholarCrossref
    40.
    Marik  PE.  Obituary: pulmonary artery catheter 1970 to 2013.  Ann Intensive Care. 2013;3(1):38. doi:10.1186/2110-5820-3-38PubMedGoogle ScholarCrossref
    41.
    Poole  D, Bertolini  G, Garattini  S.  Withdrawal of “Xigris” from the market: old and new lessons.  J Epidemiol Community Health. 2012;66(7):571-572. doi:10.1136/jech-2012-200977PubMedGoogle ScholarCrossref
    42.
    Rowan  KM, Angus  DC, Bailey  M,  et al; PRISM Investigators.  Early, goal-directed therapy for septic shock—a patient-level meta-analysis.  N Engl J Med. 2017;376(23):2223-2234. doi:10.1056/NEJMoa1701380PubMedGoogle ScholarCrossref
    43.
    Wachter  RM, Flanders  SA, Fee  C, Pronovost  PJ.  Public reporting of antibiotic timing in patients with pneumonia: lessons from a flawed performance measure.  Ann Intern Med. 2008;149(1):29-32. doi:10.7326/0003-4819-149-1-200807010-00007PubMedGoogle ScholarCrossref
    44.
    Qi  C, Hountras  P, Pickens  CO,  et al.  Detection of respiratory pathogens in clinical samples using metagenomic shotgun sequencing.  J Med Microbiol. 2019;68(7):996-1002. doi:10.1099/jmm.0.000968PubMedGoogle ScholarCrossref
    45.
    Blaschke  AJ, Hersh  AL, Beekmann  SE, Ince  D, Polgreen  PM, Hanson  KE.  Unmet diagnostic needs in infectious disease.  Diagn Microbiol Infect Dis. 2015;81(1):57-59. doi:10.1016/j.diagmicrobio.2014.10.005PubMedGoogle ScholarCrossref
    46.
    Messacar  K, Parker  SK, Todd  JK, Dominguez  SR.  Implementation of rapid molecular infectious disease diagnostics: the role of diagnostic and antimicrobial stewardship.  J Clin Microbiol. 2017;55(3):715-723. doi:10.1128/JCM.02264-16PubMedGoogle ScholarCrossref
    ×