0
Correspondence |

Analytical Shortfalls in Multivariate Regression AnalysisResponse FREE TO VIEW

Yu-Hsiang Hsieh, PhD; Richard E. Rothman, MD, PhD
Author and Funding Information

Affiliations: The Johns Hopkins University School of Medicine, Baltimore, MD,  School of Medicine and Pharmacology, University of Western Australia, Perth, WA, Australia,  Northwestern University Feinberg School of Medicine, Chicago, IL

Correspondence to: Yu-Hsiang Hsieh, PhD, Department of Emergency Medicine, The Johns Hopkins University School of Medicine, 5801 Smith Ave, Suite 3220, Davis Building, Baltimore, MD 21209; e-mail: yhsieh1@jhmi.edu



Chest. 2007;131(5):1613-1614. doi:10.1378/chest.07-0237
Text Size: A A A
Published online

We read with great interest the recent article in CHEST (July 2006) by Waterer et al,1investigating predictors for the time to first antibiotic dose (TFAD) and its association with mortality in patients with community-acquired pneumonia (CAP). We would like to commend the authors on their effort to attempt a prospective evaluation of the now widely touted Joint Commission on Accreditation of Healthcare Organizations (or JCAHO) recommendations to deliver therapy with antibiotics in ≤ 4 h to CAP patients. The study however, has several analytic shortfalls (in addition to those raised by Houck2) that limit its ability to interpret results, and potentially distort some of the findings and conclusions.

First, the authors’ presentation of their findings lacks critical detail. While they state that their goal was to examine clinical factors influencing TFAD, there is no description of the prevalence of each clinical characteristic among study subjects. As such, it is impossible to understand the utility (in a clinical context) of the significant independent variables (eg, altered mental state) for predicting TFAD. Although odds ratios are described in Table 1 of the article by Waterer et al,1 the relative importance of these predictors for the reduction of TFAD (ie, the population-attributable risk,3) cannot be estimated without knowing the prevalence of each clinical characteristic. An appreciation of the population-attributable risk would allow rationale prioritization of limited emergency department resources toward the goal of improved timeliness of antibiotics administration. Table 1 also fails to clarify whether the odds ratios shown were derived from univariate or multivariate analysis (our speculation is that they were derived from univariate analysis). A multivariate approach (which controls for confounders) would better identify true predictors for TFAD.

The second methodological issue that we would like to draw attention to involves the concept of “overadjustment,” which we believe occurred in this study. Overadjustment is a condition whereby statistical adjustment is inadvertently carried out for a variable that is either in the causal pathway between the exposure and the outcome (an intermediate cause), or is strongly related to either the exposure (multicollinearity) or the outcome.35 In this particular study, “altered mental state,” which is clearly a strong predictor for TFAD of > 4 h, was also a potent predictor for mortality. Meanwhile, TFAD of > 4 h was a predictor for mortality in univariate analysis but was not found to be so after multivariate analysis (see Table 3 in the article by Waterer et al1). Here, we cannot rule out the possibility that the variables altered mental status, TFAD of > 4 h, and mortality are all are on a causal pathway, with altered mental status resulting in a TFAD of > 4 h, and a TFAD of > 4 h leading to increased mortality. Overadjustment may have occurred when the authors inappropriately incorporated factors predicting TFAD, together with TFAD, in the multivariate analysis for predicting mortality. An alternative analytic approach that would help to clarify this relationship would be to perform a stratified multivariate analysis (for those with and without altered mental status) evaluating whether TFAD of > 4 h is associated with mortality.

In summary, Waterer and colleagues1 have concluded that there is a specious association between TFAD and mortality in CAP patients. Unfortunately, the limited reporting of data and the analytic methods employed by the authors do not allow the reader to make any reliable conclusions regarding the important and controversial issue.

The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

The authors have no conflicts of interest to disclose.

Waterer, GW, Kessler, LA, Wunderink, RG (2006) Delayed administration of antibiotics and atypical presentation in community-acquired pneumonia.Chest130,11-15. [PubMed] [CrossRef]
 
Houck, PM Antibiotics and pneumonia: is timing everything or just a cause of more problems?Chest2006;130,1-3. [PubMed]
 
Szklo, M, Nieto, FJ. Epidemiology: beyond the basics. 2000; Aspen Publisher. Gaithersburg, MD:.
 
Last, JM. A dictionary of epidemiology. 1995; Oxford University Press. New York, NY:.
 
Leistikow, B Commentary: questionable premises, overadjustment, and a smoking/suicide association in younger adult men.Int J Epidemiol2003;32,1005-1006. [PubMed]
 
To the Editor:

We thank Dr. Hsieh and Dr. Rothman for their comments. The prevalence of key factors predicting a delay in receiving antibiotics were altered mental state (12.4%), absence of hypoxia (64.3%), and absence of fever (36.6%). With respect to the other statistical queries, no amount of data manipulation will convincingly prove the causal pathway. Clinicians are still forced to chose between a theory that the difference of a few hours in receiving antibiotics alters the course of a disease that has typically been present for days, or alternatively that patients who have delays in receiving antibiotics are substantially different with respect to important comorbid illnesses.

While administering antibiotics promptly is desirable, the real key to improving outcomes lies in understanding why patients who have a delay in antibiotic therapy are more likely to die. Although an enforceable time limit satisfies the need of administrators to claim they are meeting standards of care, better patient outcomes are more likely to be linked to recognizing key comorbid conditions (eg, heart failure and diabetes) and instituting appropriate therapy for them as early as possible.

Arguing that the large Medicare database studies12 controlled for the factors we identified is misleading. Confusion is known to be poorly recognized and documented in emergency departments.34 The accuracy of qualitative assessments like the presence of confusion in the Medicare studies is therefore markedly inferior to the detailed individual assessment by an experienced physician, as occurred in our study. Assessment of other diagnoses requiring detailed clinical evaluation (eg, mild chronic organ disease) may also be suspect.

We sympathize with the desire for simple quality measures to meet administrative goals. However, these targets must actually improve patient outcomes. Prospective studies of improving the delivery of antibiotics within 4 h have not documented a mortality benefit. A 4-h rule for antibiotic therapy in community-acquired pneumonia is not supported by current data.

References
Houck, PM, Bratzler, DW, Nsa, W, et al Timing of antibiotic administration and outcomes for Medicare patients hospitalized with community-acquired pneumonia.Arch Intern Med2004;164,637-644. [PubMed] [CrossRef]
 
Meehan, TP, Fine, MJ, Krumholz, HM, et al Quality of care, process, and outcomes in elderly patients with pneumonia.JAMA1997;278,2080-2084. [PubMed]
 
Hustey, FM, Meldon, SW The prevalence and documentation of impaired mental status in elderly emergency department patients.Ann Emerg Med2002;39,338-341. [PubMed]
 
Lewis, LM, Miller, DK, Morley, JE, et al Unrecognized delirium in ED geriatric patients.Am J Emerg Med1995;13,142-145. [PubMed]
 

Figures

Tables

References

Waterer, GW, Kessler, LA, Wunderink, RG (2006) Delayed administration of antibiotics and atypical presentation in community-acquired pneumonia.Chest130,11-15. [PubMed] [CrossRef]
 
Houck, PM Antibiotics and pneumonia: is timing everything or just a cause of more problems?Chest2006;130,1-3. [PubMed]
 
Szklo, M, Nieto, FJ. Epidemiology: beyond the basics. 2000; Aspen Publisher. Gaithersburg, MD:.
 
Last, JM. A dictionary of epidemiology. 1995; Oxford University Press. New York, NY:.
 
Leistikow, B Commentary: questionable premises, overadjustment, and a smoking/suicide association in younger adult men.Int J Epidemiol2003;32,1005-1006. [PubMed]
 
Houck, PM, Bratzler, DW, Nsa, W, et al Timing of antibiotic administration and outcomes for Medicare patients hospitalized with community-acquired pneumonia.Arch Intern Med2004;164,637-644. [PubMed] [CrossRef]
 
Meehan, TP, Fine, MJ, Krumholz, HM, et al Quality of care, process, and outcomes in elderly patients with pneumonia.JAMA1997;278,2080-2084. [PubMed]
 
Hustey, FM, Meldon, SW The prevalence and documentation of impaired mental status in elderly emergency department patients.Ann Emerg Med2002;39,338-341. [PubMed]
 
Lewis, LM, Miller, DK, Morley, JE, et al Unrecognized delirium in ED geriatric patients.Am J Emerg Med1995;13,142-145. [PubMed]
 
NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

CHEST Journal Articles
PubMed Articles
Guidelines
Feverish illness in children: assessment and initial management in children younger than 5 years.
National Collaborating Centre for Women's and Children's Health | 8/28/2009
Blepharitis.
American Academy of Ophthalmology | 6/5/2009
  • CHEST Journal
    Print ISSN: 0012-3692
    Online ISSN: 1931-3543