Etiology and outcomes of veterans with spinal cord injury and disorders hospitalized with community-acquired pneumonia

262

Etiology and Outcomes of Veterans With Spinal Cord Injury and Disorders Hospitalized With Community-Acquired Pneumonia Heidi T. Chang, MPH, Charlesnika T. Evans, MPH, Frances M. Weaver, PhD, Stephen P. Burns, MD, Jorge P. Parada, MD, MPH ABSTRACT. Chang HT, Evans CT, Weaver FM, Burns SP, Parada JP. Etiology and outcomes of veterans with spinal cord injury and disorders hospitalized with community-acquired pneumonia. Arch Phys Med Rehabil 2005;86:262-7. Objective: To determine whether documentation of a causative organism for community-acquired pneumonia (CAP) is associated with outcomes, including mortality and length of stay (LOS), in hospitalized veterans with spinal cord injuries and disorders (SCI&D). Design: Retrospective cohort study. Setting: Patients with SCI&D admitted with CAP to any Veterans Affairs medical center between September 1998 and October 2000. Participants: Hospital administrative data on 260 patients with SCI&D and a CAP diagnosis. Interventions: Not applicable. Main Outcomes Measures: All-cause, 30-day mortality and hospital LOS. Results: An organism was documented by International Classification of Diseases, 9th Revision, discharge codes in 24% of cases. Streptococcus pneumoniae and Pseudomonas aeruginosa accounted for 32% and 21%, respectively, of the identified bacterial pathogens. The overall mortality rate was 8.5%. No significant association was found between etiologic diagnosis of CAP and 30-day mortality. Lower mortality was associated with treatment at a designated SCI center (relative risk⫽.35; confidence interval, .12–.99). Pathogen-based CAP diagnosis was significantly associated with longer LOS (adjusted r2⫽.023, P⫽.024). Conclusions: There was no association between etiologic diagnosis of CAP and 30-day mortality among people with SCI&D. Documentation of CAP etiology was associated with the variance in LOS. Pneumococcal vaccination and antibiotic therapy with antipseudomonal activity may be particularly

From the Stritch School of Medicine, Loyola University, Maywood, IL (Chang); Spinal Cord Injury Quality Enhancement Research Initiative, Midwest Center for Health Services and Policy Research, Department of Veterans Affairs, Edward Hines Jr VA Hospital, Hines, IL (Chang, Evans, Weaver, Parada); Institute for Health Services Research and Policy Studies, Northwestern University, Chicago, IL (Weaver); Department of Rehabilitation Medicine, University of Washington, Seattle, WA (Burns); and Spinal Cord Injury Service, VA Puget Sound Health Care System, Seattle, WA (Burns); and Division of Infectious Diseases, Department of Medicine, Loyola University, Maywood, IL (Parada). Presented as a poster to the American Academy of Physical Medicine and Rehabilitation, October 9 –12, 2003, Chicago, IL. Supported by the Spinal Cord Injury Quality Enhancement Research Initiative, Health Services Research and Development Service, US Department of Veterans Affairs (grant no. SCI 98-001). No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the author(s) or on any organization with which the author(s) is/are associated. Chang was a student at the School of Epidemiology and Public Heath, Yale University, New Haven, CT, during the research project. Reprint requests to Frances M. Weaver, PhD, Midwest Center for Health Services & Policy Research, Dept of Veterans Affairs, Hines VA Hospital, 5th & Roosevelt Rd, PO Box 5000 (151H), Hines, IL 60141, e-mail: [email protected]. 0003-9993/05/8602-8554$30.00/0 doi:10.1016/j.apmr.2004.02.024

Arch Phys Med Rehabil Vol 86, February 2005

prudent in these patients given the high frequency of these pathogens among SCI&D patients with CAP. Key Words: Community-acquired infections; Disease management; Pneumonia; Rehabilitation; Spinal cord injuries. © 2005 by American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation IFTY YEARS AGO, renal failure and other urinary tract F complications were the leading causes of mortality among persons with spinal cord injury (SCI). Currently, however, 1-3

respiratory complications, particularly pneumonia, are the most frequent cause of morbidity and mortality, both in the first year after injury and in the long-term among people with SCI.3-9 Patients with SCI and disorders (SCI&D) are at increased risk for pneumonia because of impairment or weakening of abdominal, intercostal, and diaphragmatic muscles necessary for effective coughing and mucous clearance.6,7,10 A US study,2 which examined over 9000 people who sustained a traumatic SCI between 1973 and 1983, determined that pneumonia was the leading cause of death for each age group and during all time periods. In that study, the overall standardized mortality ratio (SMR) for pneumonia and influenza was 37.1 (95% confidence interval [CI], 31.0 – 43.2) compared with the general population. Patients with complete tetraplegia have been shown to have a higher SMR for respiratory complications of all classifications of SCI&D than the general population.2 Aging and older age at time of injury also increase risk. Persons who have an SCI at 61 years or older are 46 times more likely to die from respiratory-related causes than are persons injured at age 30 years or younger.3 Strategies for the diagnosis and treatment of communityacquired pneumonia (CAP) remain controversial. The utility of the Gram stain and a culture of expectorated sputum to establish microbiologic diagnosis and guide antimicrobial therapy in cases of CAP has been debated for decades. Factors contributing to this debate are the difficulty in procuring an adequate sputum sample and the variable sensitivity and specificity of the test. The American Thoracic Society10 (ATS) and the Infectious Disease Society of America11 (IDSA) have published conflicting guidelines on the routine use of the Gram stain and culture for patients hospitalized with CAP. The ATS’s newly revised evidence-based guideline for the management of CAP acknowledges that organism-directed antimicrobial therapy would be ideal; however, the ATS asserts that the status of current diagnostic tests forces the medical community to rely on empiric antibiotic therapy in most cases of CAP.10 This guideline indicates that Gram stains and sputum cultures should not be routinely conducted for patients hospitalized with CAP, but rather should be reserved for patients suspected of harboring a drug-resistant pathogen or an organism not covered by usual empiric therapy.10 The IDSA’s guideline differs from that of the ATS in that the IDSA recommends that all patients hospitalized with CAP have Gram stain and sputum culture tests.11 Because delaying the

PNEUMONIA MANAGEMENT IN SCI, Chang

initiation of antimicrobial therapy is associated with negative patient outcomes,10,11 the IDSA recommends that empiric therapy be prescribed until results of the Gram stain and sputum culture are available.11 The IDSA has various reasons for endorsing the establishment of an etiologic diagnosis via microbiologic tests, but those potentially most important in the SCI&D population are to provide optimal pathogen-directed treatment, to prevent injudicious use of antibiotics, and to avoid selection for resistant bacteria.11 Studies12-15 that have been conducted in the general population to investigate the value of microbiologic testing in adults hospitalized with CAP have shown no benefit in patient outcomes associated with establishment of CAP etiology. In a study that examined the value of initial microbiologic testing in patients hospitalized with CAP, Sanyal et al15 found that pathogen identification had little influence on treatment decisions and patient outcomes. Sanyal concluded that, although the results of the study were in accord with other studies, study conclusions could not be extrapolated to certain categories of patients who have a greater likelihood of harboring resistant pathogens.15 Persons with SCI&D may be a group in which results obtained from studies conducted in the general population cannot be extrapolated. Those with SCI&D may be at increased risk of harboring resistant pathogens. The incidence of infection, most commonly of the respiratory tract, urinary tract, and/or infected pressure ulcers, is higher among people with SCI.16 Patients with SCI&D are treated for an average of 2.4 urinary tract infections (UTIs) per year.17 Because the most significant risk factor for infection with drug-resistant pathogens is previous treatment with antimicrobial agents,18 persons with SCI&D may have higher rates of colonization and infection with nonsusceptible strains. The value of establishing a microbiologic diagnosis and its effect on outcomes for SCI&D persons hospitalized with CAP have not been documented. Pathogendirected antibiotic therapy may be the best management strategy in this population. The primary goal of our study was to assess whether the presence of an etiologic diagnosis of CAP was related to mortality and length of stay (LOS) in persons with SCI&D. METHODS To discern whether documentation of the causative organism of CAP improves medical and utilization outcomes in hospitalized patients with SCI&D, we conducted a retrospective cohort study. Patients who met the Veterans Affairs (VA) Allocation Resource Center’s diagnostic criteria for SCI&D were eligible for this study. From this population, we identified all persons with an outpatient diagnosis of pneumonia, an admission on the same day to any VA medical center, and a discharge diagnosis of pneumonia for the period between September 1998 and October 2000 (fiscal years 1999 and 2000). These criteria distinguished CAP from nosocomial pneumonia. Only hospitalized patients were examined; however, the criteria did not distinguish between patients who were admitted from long-term care facilities and those who lived in the community. Exclusion criteria were a discharge from the hospital within 10 days of index CAP hospitalization, a diagnosis indicating a nonbacterial pneumonia, and concurrent multiple sclerosis (MS). Patients with MS were excluded because the natural history of MS that leads to paralysis differs from the sequelae after SCI&D. Data sources included 2 VA administrative databases—the National Patient Care Database (NPCD) and the Spinal Cord Dysfunction registry (SCD-R). The NPCD contains a comprehensive set of data on each episode of care in the VA health

263

care system, including inpatient care data and outpatient care data. Variables used for this study included admission and discharge dates; diagnostic information, using International Classification of Diseases, 9th Revision (ICD-9) codes; LOS; discharge type, which identifies whether a patient died in the hospital; and demographic variables. CAP was defined as any ICD-9 code for pneumonia in the NPCD. Subjects whose discharge diagnoses reflected a nonbacterial pneumonia were excluded from the study. Only 2 subjects were excluded because of the presence of a viral pneumonia diagnosis. Patients with a diagnosis code of 486.0 (pneumonia, organism unspecified), 485.0 (bronchopneumonia, organism unspecified), and 482.9 (bacterial pneumonia, unspecified) were defined as having no pathogen identified. Patients with all other codes for bacterial pneumonia, 481 to 482.89, were classified as having an established pneumonia etiology. The SCD-R was used to obtain information on patient level and completeness of injury and duration of injury. The primary outcome—30-day, all-cause mortality—was defined as death within 30 days of hospital admission. Death was restricted to those who died within 30 days of hospitalization, to better relate mortality to the initial diagnosis and treatment of patients and to distinguish it from complications that may have occurred during hospitalization. The secondary outcome, LOS, was restricted to patients who did not die during hospitalization. Covariates included patients’ neurologic level of injury, completeness of injury, and age at injury. Neurologic impairment was classified as tetraplegic or paraplegic and complete or incomplete.19 When data concerning level and/or completeness of injury were not contained in the SCD-R, data in the NCPD were used as a proxy. Information concerning age, gender, race, marital status, month of diagnosis, treatment at a designated SCI center, respiratory complications during hospitalization, and comorbidities were collected through the NPCD. Admissions from November through February were considered flu season admissions, and admissions between March and October were considered nonflu season admissions. Respiratory complications of interest included aspiration, respiratory failure, bronchitis, atelectasis, bacteremia, and septicemia. Occurrence of these complications was ascertained via the diagnosis fields in the discharge record from the NPCD. The Deyo-Charlson comorbidity index,20 which has been used in the VA population,21 was used to rank patients’ level of comorbidity. Comorbidities were measured by ICD-9-CM codes in the NPCD in fiscal years 1999 through 2000. The Deyo-Charlson score ranges from 0 to 3, with 0 indicating the absence of any one of 18 disease conditions and 3 indicating the presence of 3 or more of the conditions.20 Admission into an intensive care unit (ICU) on the same day as hospital admission was used to control for severity of illness. This information was obtained from the bed section file of the NPCD. Descriptive statistics were calculated for all the demographic, illness, and outcome variables. Independent predictors of LOS were identified through Pearson correlations, t tests, and analysis of covariance statistical analyses. To assess independent predictors of 30-day mortality, the chi-square and Fisher exact tests were performed. Relative risks, with 95% CIs, were calculated. Confounders were identified in the bivariate analysis, using a P value of .25. The only confounder identified was age; therefore, an age-adjusted regression model was used to determine the extent to which etiologic diagnosis of CAP affects the hospital LOS. Age-adjusted logistic regression was performed to determine whether etiologic diagnosis Arch Phys Med Rehabil Vol 86, February 2005

264

PNEUMONIA MANAGEMENT IN SCI, Chang Table 1: Characteristics of Veterans With SCI&D With and Without Established CAP Etiology

Characteristics

Mean age (y) Mean LOS (d) Percentage who died within 30d Percentage male Percentage white Percentage admitted at SCI&D center Percentage admitted to ICU Percentage admitted during flu season Tetraplegia* Complete lesions† Percentage with complications Comorbidities %0 %1 %2 % 3⫹

Total Sample (frequency) (N⫽260)

Known Etiology (frequency) (n⫽62)

Unknown Etiology (frequency) (n⫽198)

P Value

62.5⫾13.5 13.5⫾18.9 8.5 (22) 98.5 (256) 76.2 (198) 37.3 (97) 10.4 (27) 43.5 (113) 141 (59) 61 (56) 21.2 (55)

58.3⫾13.7 16.5⫾20.1 8.1 (5) 96.8 (61) 74.2 (46) 40.3 (25) 12.9 (8) 38.7 (24) 42 (30) 15 (25) 24.2 (15)

63.9⫾13.2 12.6⫾18.4 8.6 (17) 99.0 (196) 76.4 (152) 36.4 (72) 9.6 (19) 44.9 (89) 99 (70) 46 (75) 20.2 (40)

.006 .16 .90 .22 .68 .57 .46 .39 .11 .59 .50

20.0 (52) 21.1 (60) 21.9 (57) 35.0 (91)

16.1 (10) 32.1 (23) 21.0 (13) 25.8 (16)

21.2 (42) 18.7 (37) 22.2 (44) 37.9 (75)

Reference .03 .65 .81

NOTE. Values are mean ⫾ standard deviation (SD) or % (n). *Data missing for 22 patients. † Data available for 109 subjects.

affected 30-day mortality. All tests were 2-sided, using an ␣ of .05 as the significance level. RESULTS Two hundred sixty patients satisfied the eligibility criteria (table 1). Subjects had a mean age of 62.5 years; most were men (98.5%) and white (76%). Thirty-seven percent were admitted to an SCI center, and 10% were admitted to an ICU on the same day as hospitalization. Forty-three percent were admitted during flu season. A microbiologic discharge diagnosis was indicated on 24% of all discharge diagnostic reports. Patients with a documented established etiology were significantly younger than patients without an established etiology (mean ages, 58y and 64y, respectively; P⫽.006). There were no other significant differences in patient characteristics between those with and without an etiologic diagnosis. The mean LOS was 13.5 days, and almost 9% died within 30 days of admission. There was no significant association between these outcomes and the establishment of an etiologic diagnosis. The most commonly identified pathogen was Streptococcus pneumoniae (32%), followed by Pseudomonas aeruginosa (21%) (table 2). Fifty-five patients (21%) experienced 75 respiratory-related complications. Complications included bronchitis (n⫽22), respiratory failure (n⫽20), atelectasis (n⫽13), septicemia (n⫽7), bacteremia (n⫽6), and aspiration (n⫽7). Twenty-two subjects (8.5%) died within 30 days of hospital admission (table 3). The mean time to death for these subjects ⫾ standard deviation (SD) was 11.1⫾9.7 days. Subjects who died within 30 days of hospitalization (mean age, 71.1y) were significantly older than subjects who survived (mean age, 61.7y) (P⫽.001). Mortality was significantly lower in patients treated at SCI centers (3.1%) than in patients treated at non-SCI centers (11.7%) (P⫽.02). Pathogen-related CAP documentation was not associated with 30-day mortality (unadjusted relative risk [RR]⫽.95; 95% CI, 0.43–2.12). Furthermore, after adjusting for age, pathogen identification was not associated with 30-day mortality (adjusted RR⫽.98; 95% CI, 0.86 –1.10). The mean hospital LOS was 13.5⫾18.9 days (median, 8d), excluding subjects who died during hospitalization (table 4). Arch Phys Med Rehabil Vol 86, February 2005

Bivariate analysis revealed that nonwhite race, treatment at an SCI center, admission during flu season, respiratory complications during hospitalization, and direct admission into an ICU were all significantly associated with increased length of hospitalization (all P⬍.05). Older age was marginally associated with increased LOS (r2⫽.12, P⫽.07). However, pathogen identification was not independently associated with LOS. Age was the only variable identified as a confounder in the bivariate analysis (P⬍.25). An age-adjusted linear model of the natural log of the LOS (because of the skewedness of LOS) was used to account for confounding. Age and LOS had an adjusted r2 value of .023 (P⫽.24), which explains only 2.3% of the variance of LOS. Pathogen identification was a significant predictor in LOS (␤⫽.02, P⫽.03), as was age (␤⫽.009, P⫽.04). When the model was expanded to include variables associated with LOS—namely race, admission to an ICU on hospitalization, treatment at an SCI center, respiratory complications, and admission during flu season—the adjusted r2 was 203. This explains 20.3% of the variance in LOS. In the expanded model of LOS, pathogen identification was also a significant predictor (␤⫽.27, P⫽.03). Pathogen identification was associated with longer LOS.

Table 2: Discharge Diagnoses of Patients With Established Microbiologic Etiology of CAP Pathogen

Frequency

%

Streptococcus pneumoniae Pseudomonas Haemophilus influenza Klebsiella Streptococcus group A Streptococcus group B Streptococcus, other Escherichia coli Other organism Total

20 13 6 5 3 3 2 1 9 62

32.3 21.0 9.7 8.1 4.8 4.8 3.2 1.6 14.5 100.0

NOTE. Twenty-four percent of sample.

265

PNEUMONIA MANAGEMENT IN SCI, Chang Table 3: Bivariate Associations Between Patient Characteristics and 30-Day, All-Cause Mortality Deaths Characteristics

Pathogen identification Yes No Race White Nonwhite Direct ICU admission Yes No Paralysis† Tetraplegia Paraplegia Treatment facility SCI&D center Noncenter Season of admission Flu Nonflu Respiratory complications Yes No Comorbidities 0 1 2 3⫹

n

%

Crude RR

95% CI

P Value*

5 17

8.1 8.6

0.95 1.00

0.43–2.12 Reference

.90

19 3

9.6 4.8

1.88 1.00

0.71–4.95 Reference

.24

4 18

14.8 7.7

1.88 1.00

0.71–4.95 Reference

.21

15 7

10.6 7.2

1.47 1.00

0.62–3.48 Reference

.37

3 19

3.1 11.7

0.35 1.00

0.12–0.99 Reference

.02

10 12

8.8 8.2

1.05 1.00

0.65–1.70 Reference

.84

7 15

12.7 7.3

1.58 1.00

0.81–3.06 Reference

.20

5 5 9 3

9.6 8.3 15.8 3.3

1.00 0.87 1.64 0.34

Reference 0.27–2.83 0.59–4.58 0.09–1.37

.81 .34 .11

*For ␹2 test statistic. Data missing for 22 subjects.

†

Subgroup analyses were conducted for cases in which additional characteristics of SCI were available. There was no difference in the risk of death within 30 days between patients with paraplegia (7.2%) and patients with tetraplegia (10.6%) (␹2 test⫽.80, P⫽.37) or in the prevalence of pathogen identification (20.6% and 29.8%, respectively) (␹2 test⫽2.51, P⫽.11). Data on neurologic classification, based on the SCD-R, were available for 109 subjects. Fifty-six percent had complete lesions, and 44% had incomplete lesions. Completeness of injury was not associated with 30-day mortality (␹2 test⫽.46, P⫽.50), with etiologic diagnosis of CAP (␹2 test⫽.28, P⫽.59), or with LOS (P⫽.24). However, among subjects with tetraplegia, those with complete lesions were somewhat more likely to die within 30 days of their CAP admission than those with incomplete lesions (␹2 test⫽3.09, P⫽.08) but not more likely to have a pathogen identified (␹2 test⫽.29, P⫽.58). Completeness of injury was associated neither with mortality nor with pathogen identification among subjects with paraplegia. One hundred twenty-four subjects had information about age at onset of SCI. Those admitted to an ICU were marginally older at injury than those not admitted (48.6y vs 38.8y, P⫽0.06). In addition, those who died within 30 days were slightly older than those who survived (53.8y vs 40.8y, P⫽0.07). Age at injury was associated neither with LOS (Pearson r⫽⫺.16, P⫽.86) nor with pathogen identification (P⫽.74). DISCUSSION The results of our study are similar to other studies of CAP outcomes in the general population, with a few exceptions.11,13,15,22 The rate of etiologic diagnosis (24%) in our

sample of hospitalized veterans with SCI&D is slightly lower than that found in a review of 50 international studies of CAP, in which a microbiologic diagnosis was made in anywhere from 31% to 54% of cases.23 Possible reasons for our lower rate include a low yield of diagnostic tests (which is associated with treatment with antibiotics before sample collection), the inability of subjects to produce expectorated sputum, the failure to order or perform sputum testing, and the recording of inaccurate ICD-9 codes in the discharge diagnostic report. In addition, the mortality rate in this cohort was 8.5%, which is comparable with patients in general who are hospitalized with CAP (range, 2%–30%; average, 13.7%).24 The mortality rate for patients directly admitted into an ICU (14.8%) in this cohort was lower than the mortality rate reported in the metaanalysis for ICU patients (36.5%).24 However, unlike the metaanalysis, we did not examine patients transferred to an ICU subsequent to hospitalization in our cohort. On the other hand, mortality in this cohort was lower than that seen for pneumonia in the entire VA population. A discharge disposition of death associated with a diagnosis of pneumonia was found in approximately 20% of all cases seen in the VA hospitals between 1991 and 1996; however, CAP was not differentiated from nosocomial pneumonia in our study.25 Mortality was associated with location of care in this cohort. Patients treated at non-SCI centers were more likely to die (11.7%) than those treated at SCI centers (3.1%). Potential contributors to this finding include the presence of SCI specialists, the greater use of effective treatments for mobilizing secretions, and experienced nursing care for assisted (quad) coughing maneuvers at the SCI centers. Furthermore, physicians at SCI centers may be more knowledgeable about morArch Phys Med Rehabil Vol 86, February 2005

266

PNEUMONIA MANAGEMENT IN SCI, Chang

Table 4: Bivariate Associations Between Patient Characteristics and LOS for Patients Who Did Not Die During Hospitalization Variable

Pathogen ID Yes No Age (y)* ⬍40 40–49 50–50 60–69 70–70 80⫹ Race White Nonwhite Treatment facility Noncenter SCI&D center Direct ICU admission Yes No Season of admission Flu Nonflu Respiratory complications Yes No Comorbidities 0 1 2 3⫹

Mean LOS

P Value

14.7⫾19.8 12.3⫾17.9

.38

10.6⫾6.6 7.2⫾4.9 12.6⫾13.8 16.3⫾22.7 13.6⫾22.0 15.8⫾21.9

Reference .12 .62 .40 .64 .44

0.7⫾11.3 19.3⫾30.1

.004

9.3⫾12.5 18.3⫾23.8

.001

21.4⫾18.1 14.7⫾19.8

.025

16.8⫾25.7 9.8⫾12.5

.001

19.7⫾22.7 11.1⫾16.6

.02

13.7⫾21.8 12.2⫾14.1 11.8⫾11.4 13.3⫾21.5

Reference .67 .59 .92

NOTE. Values are mean ⫾ SD. N⫽236. *r 2⫽.12, P⫽.07.

tality from pneumonia in the SCI&D population and therefore may recommend admission more readily. Because we lacked patient-level clinical data on medical acuity at presentation, other than the need for direct ICU admission, we were unable to determine whether SCI centers tended to admit patients earlier in their illness course or with lower medical acuity, thereby lowering the overall risk of death for this study sample. Also, patients with lower medical acuity may have advocated for outpatient treatment rather than admission to a hospital that lacked a specialized SCI unit, thereby biasing the non-SCI centers to show higher mortality. Because a comparison of outcomes at different center types was not an a priori hypothesis for the study, caution should be taken with interpretation of this finding, and it should be validated in future studies. Pseudomonas aeruginosa was the second most commonly documented pathogen in this cohort of SCI&D patients, accounting for 21% of the established microbiologic etiologies. This differs from other studies of CAP in which Pseudomonas aeruginosa is an infrequent cause of CAP. A study of 103 episodes of bacteremia in 93 hospitalized patients with SCI&D found that, in 10 patients for whom pneumonia was the primary site of infection, Pseudomonas aeruginosa was detected in 2 patients.8 It is unclear why Pseudomonas aeruginosa was a common pathogen in this cohort. One possible explanation is that colonization of other body sites, such as the urethra, perineum, or rectum, occurs in most persons with SCI.26 Additionally, use of antibiotics for more than 7 days in the past Arch Phys Med Rehabil Vol 86, February 2005

month and malnutrition are both known risk factors for Pseudomonas aeruginosa infection in the general population.27 The finding of Pseudomonas in 1 in 5 isolates suggests that initial empiric treatment in this population should include a reliable antipseudomonal antibiotic. The mean LOS in this study is somewhat longer than what has been reported in studies of CAP conducted in the general VA population22 and in the general North American population.28,29 The longer LOS in this study may be attributed to the fact that people with SCI&D are also at increased risk for nonrespiratory-related complications, such as pressure ulcers and UTIs. In addition, LOS at SCI centers may have been longer for some people, to allow providers to address other SCI-related health care issues, such as equipment prescription, patient education, and annual medical evaluations. A strength of our study is the use of a population-based sample. An estimated 22% of all persons with SCI&D in the United States are veterans.30,31 The average age at time of injury for veterans is 32 years, and the mean age at time of injury in the general population with SCI&D from the Model Spinal Cord Injury Systems of care is 31 years.31 Women were underrepresented in this study, reflecting the small proportion of female veterans with SCI&D. Ninety-eight percent of veterans with SCI&D are men, whereas, in the general population, 80% with SCI&D are men.31 Our study was based on a retrospective review of administrative data. As such, there were several limitations. Clinical information regarding patient characteristics (eg, smoking status), type and number of tests conducted, and treatments provided was unavailable. Furthermore, diagnosis was based on what was recorded in the administrative data and was not confirmed based on laboratory test results. The ICD-9 codes in the NPCD have been validated in several disease areas, including in CAP.32 However, ICD-9 coding for pneumonia may vary across facilities and coders; hence, one coder may indicate pneumococcal pneumonia, whereas another could indicate pneumonia, even though a laboratory report indicates that a culture was positive for pneumococcus. We were unable to assess whether this was a problem with our data. Two patients with known viral pneumonia were excluded from our analysis because they did not have CAP as defined in our study. We are left without certainty regarding if, or how many of, the remaining 76% of patients without an established etiologic diagnosis may have also had a viral pneumonia. Although this is a potential limitation of our study, in many ways it reflects the same diagnostic and therapeutic dilemmas that clinicians face when confronted with patients presenting with CAP. Our data sources did not support the ability to distinguish between those admitted from long-term-care facilities and those admitted directly from the community, nor the ability to use the American Spinal Injury Association classification. Patients’ completeness of injury, obtained from the SCD-R, was not available for more than half the cohort, and crude categories of paraplegia and tetraplegia were distinguished using ICD-9 codes from the NPCD. Use of the Pneumonia Severity Index, a validated prediction rule that stratifies adult inpatients with CAP into 5 risk classes with respect to the risk of death within 30 days,33 would have been an asset to our study. However, a medical chart review was not possible in our investigation. Another limitation was lack of pharmacy data regarding treatment. Most studies that have investigated use of initial empiric antimicrobial therapy in accordance with recommendations in patients hospitalized with CAP have found that adherence to treatment guidelines can improve mortality and

PNEUMONIA MANAGEMENT IN SCI, Chang

other outcomes. However, patients with SCI&D may need empiric coverage, which differs from that of the general population, given that they are more likely to be infected with Pseudomonas, an uncommon and highly resistant infection. CONCLUSIONS Documentation of CAP etiology is not a good predictor of mortality or LOS in patients with SCI&D. Pseudomonas aeruginosa was a commonly identified pathogen in this cohort. Given the large incidence of Pseudomonas isolates, antibiotics with antipseudomonal activity may be indicated as initial empiric therapy for CAP in patients with SCI&D. Pneumococcus also remains a frequent isolate in this study. This highlights an opportunity for disease prevention and the need for pneumococcal immunization in this patient population. Mortality among patients treated at SCI centers was lower compared with patients treated at non-SCI centers. The reason for this difference is unclear. Further research that examines the relationship between treatment and patient outcome is warranted, to determine best practice among patients with SCI&D. Acknowledgment: We thank Scott Miskevics for programming expertise. References 1. Harktopp A, Bronnum-Hansen H, Seidenschnur AM, BieringSorenson F. Survival and cause of death after traumatic spinal cord injury: a long-term epidemiological survey from Denmark. Spinal Cord 1997;35:76-85. 2. DeVivo MJ, Black KJ, Stover SL. Causes of death during the first 12 years after spinal cord injury. Arch Phys Med Rehabil 1993; 74:248-54. 3. Frankel HL, Coll JR, Charlifue SW, et al. Long-term survival in spinal cord injury: a fifty year investigation. Spinal Cord 1998; 38:266-74. 4. DeVivo MJ, Stover LS. Long term survival and causes of death. In: Stover SL, DeLisa JA, Whiteneck GG, editors. Spinal cord injury: clinical outcomes from the model systems. Gaithersburg: Aspen; 1995. p 289-316. 5. DeVivo MJ, Krause JS, Lammertse DP. Recent trends in mortality and cause of death in people with spinal cord injury. Arch Phys Med Rehabil 1999;80:1411-9. 6. Fishburn MJ, Marino RJ, Ditunno JF. Atelectasis and pneumonia in acute spinal cord injury. Arch Phys Med Rehabil 1990;71:197200. 7. Lucke KT. Pulmonary management following acute SCI. J Neurosci Nurs 1998;30(2):91-104. 8. Montgomerie JZ, Chan E, Gillmore DS, Canawati HN, Sapico FL. Low mortality among patients with spinal cord injury and bacteremia. Rev Infect Dis 1991;13:867-71. 9. Mansel JK, Norman JR. Respiratory complications and management of spinal cord injuries. Chest 1990;97:1446-52. 10. Niederman MS, Mandell LA, Anzueto A, et al. Guidelines for the management of adults with community-acquired pneumonia. Diagnosis, assessment of severity, antimicrobial therapy, and prevention. Am J Respir Crit Care Med 2001;163:1730-54. 11. Bartlett JG, Dowell SF, Mandell LA, File TM Jr, Musher DM, Fine MJ. Practice guidelines for the management of communityacquired pneumonia in adults. Clin Infect Dis 2000;31:347-82. 12. Levy M, Dromer F, Brion N, Leturdu F, Carbon C. Communityacquired pneumonia: importance of initial noninvasive bacteriologic and radiographic investigations. Chest 1988;93:43-8. 13. May HM, Harrison TS, Harrison BD. A criterion based audit of community-acquired pneumonia. Respir Med 1994;88:693-6. 14. Woodhead MA, Arrowsmith J, Chamberlain-Weber R, Woodings S, Williams I. The value of routine microbial investigation in community-acquired pneumonia. Respir Med 1991;85:313-7.

267

15. Sanyal S, Smith PR, Saha AC, Gupta S, Berkowitz L, Homel P. Initial microbiologic studies did not affect outcome in adults hospitalized with community-acquired pneumonia. Am J Respir Crit Care Med 1999;160:346-8. 16. Montgomerie JZ. Infection in patients with spinal cord injuries. Clin Infect Dis 1997;25:1285-90; quiz 1291-2. 17. Esclarin De Ruz A, Garcia Leoni E, Herruzo Cabrera R. Epidemiology and risk factors for urinary tract infection in patients with spinal cord injury. J Urol 2000;164:1285-9. 18. Doern GV, Heilmann KP, Huynh HK, Rhomberg PR, Coffman SL, Brueggemann AB. Antimicrobial resistance among clinical isolates of Streptococcus pneumoniae in the United States during 1999-2000, including a comparison of resistance rates since 19941995. Antimicrob Agents Chemother 2001;45:1721-9. 19. American Spinal Injury Association and International Medical Society of Paraplegia. International standards for neurological and functional classification of spinal cord injury, revised 1996. Chicago: ASIA; 1996. 20. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:613-9. 21. Weaver F, Hynes D, Hopkinson W, et al. Preoperative risks and outcomes of hip and knee arthroplasty in the Veterans Health Administration. J Arthroplasty 2003;18:693-707. 22. Lentino JR, Krasnicka B. Association between initial empirical therapy and decreased length of stay among veteran patients hospitalized with community-acquired pneumonia. Antimicrob Agents Chemother 2002;19:61-6. 23. Finch R. Community-acquired pneumonia: the evolving challenge. Clin Microbiol Infect 2001;7(Suppl 3):30-8. 24. Fine MJ, Smith MA, Carson CA, et al. Prognosis and outcomes of patients with community-acquired pneumonia: a meta-analysis. JAMA 1996;275:134-41. 25. Roselle GA, Danko LH, Kralovic SM, Simbartl LA, Hilley J, Tryhus P. A six-year epidemiological review of pneumonia in the Department of Veterans Affairs facilities. Mil Med 1999;164: 293-7. 26. Gilmore DS, Bruce SK, Jimenez EM, Schick DG, Morrow JW, Montgomerie JZ. Pseudomonas aeruginosa colonization in patients with spinal cord injuries. J Clin Microbiol 1982;16:856-60. 27. Niederman MS. Guidelines for the management of communityacquired pneumonia: current recommendations and antibiotic selection issues. Med Clin North Am 2001;85:1493-509. 28. McCormick D, Fine MJ, Coley CM, et al. Variation in length of hospital stay in patients with community-acquired pneumonia: are shorter stays associated with worse medical outcomes? Am J Med 1999;107:5-12. 29. Dedier J, Singer DE, Chang Y, Moore M, Atlas SJ. Processes of care, illness severity, and outcomes in the management of community-acquired pneumonia at academic hospitals. Arch Intern Med 2001;161:2099-104. 30. Lasfaragues JE, Custis D, Morrone F, Carswell J, Nguyen T. A model for estimating spinal cord injury prevalence in the United States. Paraplegia 1995;33:62-8. 31. Weaver FM, Hammond MC, Guihan M, Hendricks RD. Department of Veterans Affairs Quality Enhancement Research Initiative for spinal cord injury. Med Care 2000;38:182-91. 32. Guevara RE, Butler JC, Marston BJ, Plouffe JF, File TM Jr, Breiman RF. Accuracy of ICD- 9-CM codes in detecting community-acquired pneumococcal pneumonia for incidence and vaccine efficacy studies. Am J Epidemiol 1999;149:282-9. 33. 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;336:243-50. Arch Phys Med Rehabil Vol 86, February 2005