Gender-Dependent Differences in Outcome After the Treatment of Infection in Hospitalized Patients FREE

Disparity in diagnosis, treatment, and outcome between men and women among many disease processes has garnered increasing interest over the past decade, and gender-dependent differences in outcome and quality of care have been acknowledged by the American Medical Association's Council on Ethical and Judicial Affairs and others as requiring prudent investigation.^1- 2 Part of the impetus for examination of gender differences over a broad range of diagnoses stems from reports of increased mortality in women with ischemic cardiac disease.^3- 7 Studies of other cohorts have demonstrated mixed results regarding gender inequality in outcome. In initial reports of patients who were mechanically ventilated, for example, mortality was not affected by gender,⁸ while a more recent study found being female to be independently associated with hospital mortality in a similar population.⁹ However, men with cirrhosis undergoing surgery appear to have a higher complication rate than women with cirrhosis.¹⁰

Among hospitalized and critically ill patients, infections remain a leading cause of morbidity and mortality, prompting efforts to identify risk factors for their development and subsequent outcome. The effect of gender under these circumstances remains unclear. In animal studies, female mice tolerate polymicrobial sepsis better than males,¹¹ and survival was improved in males after either testosterone receptor blockade¹² or dihydroepiandrosterone¹³ (an inactive metabolite of testosterone) under similar conditions. Clinically, male gender is an independent risk factor for the development of nosocomial bloodstream infection,¹⁴ and has been associated with in-hospital mortality in septic surgical patients.¹⁵ On the other hand, female gender is found to be an independent predictor of mortality in patients with Enterococcus bloodstream infections.¹⁶ Women, in addition, may have a higher mortality among patients with necrotizing soft tissue infection¹⁷ and a higher rate of multisystem organ dysfunction in postoperative cardiac patients developing nosocomial infections.¹⁸ Among specific sites of infection, a retrospective study of 62 patients with nosocomial pneumonia demonstrated no difference in crude mortality for men vs women,¹⁹ and a separate study of surgical patients with pulmonary infiltrates showed no difference in mortality after multivariate analysis,²⁰ both despite the reported association between gender and death during mechanical ventilation.⁹ A large cohort of hospitalized patients older than 65 years demonstrated no gender differences in outcomes for multiple diagnoses including pneumonia.²¹

These conflicting reports prompted our analysis of surgical patients to identify gender-related differences in infection-associated outcome, with an initial hypothesis that among a large number of hospitalized patients, overall mortality would not be associated with gender after controlling for other relevant variables.

The study was approved by the University of Virginia Human Investigation Committee and conducted at the University of Virginia Health Sciences Center Hospital from December 1996 through January 1999. Given the nature of this observational study, the Human Investigation Committee guidelines did not require acquisition of informed consent. All patients in the adult general surgery, trauma, and transplantation units were evaluated prospectively by the investigators every other day. Infections were initially diagnosed by the ward staff, but were entered into the study by the investigators only if they met the study definitions. Patients were identified by every other day chart review, review of daily microbiologic and laboratory data, antibiotic use, and attending physician interview. All patients with infections meeting criteria during the study period were included.

Pulmonary infection was diagnosed when a predominant organism was isolated from appropriately obtained cultures in the setting of purulent sputum production and a new or changing infiltrate on chest x-ray with systemic evidence of infection. Bloodstream infections were diagnosed by isolation of organisms from blood cultures from any site, with the exception of Staphylococcus epidermidis or other coagulase-negative staphylococci, which required isolation from 2 sites. Criteria for urinary tract infections included isolation of more than 10⁵ organisms per milliliter of urine or more than 10⁴ organisms per milliliter of urine and dysuria. Criteria for catheter-related infection included isolation of 15 or more colony-forming units from catheter tips by semiquantitative roll plate technique in the setting of suspected infection (systemic symptoms or localized purulence). Cellulitis, peritoneal infections, and surgical site infections were diagnosed clinically, frequently without obtaining cultures. Treatment generally consisted of the initiation of nonprophylactic antibiotic therapy, but also included open or percutaneous drainage of infection without antibiotics (eg, surgical site infection). Nosocomial infections were those occurring in patients admitted without evidence of infection at the diagnosed site. Episodes of infection occurring more than 72 hours apart in the same patient were considered separately and individually for analysis.

The primary outcome variable was death prior to discharge. Secondary outcome variables included length of stay after initiation of therapy for infection, time from initiation of treatment to defervescence (maximum temperature <38°C for 24 consecutive hours), and time to reduction of white blood cell (WBC) count to 11 × 10⁹/L or less in patients with initial significant fever (oral, rectal, or core temperature ≥38.5°C) or leukocytosis (WBC ≥15 × 10⁹/L). Intake variables included gender, age, WBC count, maximum temperature, the Acute Physiology and Chronic Health Evaluation II (APACHE II) score,²² the Acute Physiology score,²² and time from hospital admission to initiation of treatment for the infection. The WBC count, maximum temperature, APACHE II score, and Acute Physiology scores were all the maximal value within 24 hours of initiation of treatment for infection. Other parameters recorded included infection site, culture data, antibiotic regimen, total duration of treatment, and the presence of significant comorbidities including diabetes mellitus (either type), chronic renal insufficiency defined as a documented serum creatinine level of 176.8 µmol/L (2.0 mg/dL) or more prior to admission, mechanical ventilation, coexisting malignancy, requirement for hemodialysis, steroid or other immunosuppressive therapy, administration of nonautologous blood transfusion during hospitalization but prior to diagnosis of infection, and presence of a solid organ transplantation. Mechanical ventilation was defined as the requirement for ventilatory support prior to initiation of treatment for infection. Routine mechanical ventilation in the immediate perioperative period was excluded. Significant steroid treatment included both chronic low-dose and acute high-dose therapy, as defined by APACHE II.²²

All univariate comparisons were unpaired and all tests of significance were 2-tailed. For univariate analysis, continuous variables were compared by t test with equal or unequal variance as determined by F test analysis of each parameter. Categorical data were compared using χ² or Fisher exact test, depending on sample size. All values are expressed as mean (SE) (continuous variables) or as a percentage of the group from which they were derived (categorical variables). On univariate analysis, P≤.05 was considered significant. An initial comparison of demographic data, severity of illness, frequency of comorbidities, and outcome variables between men and women with infection was followed by a specific comparison of gender differences for site-specific infections such as pneumonia. Logistic regression was subsequently performed to identify independent risk factors for mortality associated with all infections and pneumonia. Since the duration of hospitalization for nosocomial infections prior to infection was not a normally distributed variable, we categorized episodes as occurring 0 to 7 days or more than 7 days after admission. Backward stepwise logistic regression analysis was used to estimate the odds ratio of inpatient mortality (dependent variable) and the presence of comorbidities or potential prognostic factors (independent variables). The odds ratio was estimated using the final logistic regression model as exponential [β-coefficient] and the 95% confidence intervals for the odds ratio were calculated. An initial Pearson correlation coefficient was determined for all continuous variables to screen for highly correlated parameters. Since APACHE II and Acute Physiology scores were highly correlated, only the APACHE II score was included in the model. After establishing the logistic regression model and identifying factors significantly associated with mortality, gender was added to the models to identify its role as an independent predictor of mortality. Statistical analysis was performed using SAS software (SAS Institute Inc, Cary, NC) and the assistance of the University of Virginia Division of Statistics.

Among 1470 total infectious episodes occurring in 892 patients, 782 (53.2%) were in men, and 688 (46.8%) were in women. A univariate comparison of age, severity of illness, comorbidities, and outcome variables for all infections is included in Table 1. Although there was a trend toward increased crude mortality for women, men had a longer length of stay despite similar severity of illness. Women were slightly older and were more likely to have had diabetes prior to infection. Men were more likely to have been transfused or mechanically ventilated and had a longer interval from admission to initiation of treatment.

A univariate analysis of factors associated with death from all infections is shown in Table 2. Overall mortality for men was 11.1%; for women, 14.2%. Since female gender showed a nonsignificant association with mortality (P = .07), logistic regression was performed (Table 3). Death was independently associated with APACHE II score, increasing age, previous blood transfusion, hospitalization more than 7 days prior to treatment for infection, and a lower maximum temperature at time of diagnosis, but not gender.

Additional univariate analyses were performed to identify potential gender-related differences in subcategories of patients felt to be at increased risk of death. Among 291 critically ill patients with APACHE II scores of 20 or more, 142 (48.8%) were men and 149 (51.2%) were women, with no difference in either mortality (men, 34.5% vs women, 38.3%; P = .51) or length of stay expressed as mean (SE) (men, 34 [3] days vs women, 30 [2] days; P = .32). Of 363 infections diagnosed in the intensive care unit, 210 (57.9%) occurred in men and 153 (42.1%) in women. There was no difference in mortality (men, 30.0% vs women, 36.6%; P = .19), although the hospital stay was longer for men (men, 40 [3] days vs women, 30 [2] days; P = .003). Among 526 (53.2%) men and 462 (46.8%) women with nosocomial infections, there was no difference in mortality (men, 15.6% vs women, 19.0%; P = .19), although the overall length of stay was higher for men (men, 23 [1] days vs women, 19 [1] days; P = .02). Mortality for the 163 men and 107 women ventilated at the time of diagnosis was similar (men, 30% vs women, 36%; P = .30).

Table 4 gives the number of infections by site and gender and the associated mortality. Pneumonia and bloodstream infections were more common in men, while urinary tract infections were more common in women. Mortality was significantly higher for women with pneumonia or soft tissue infections. Infected sites identified concurrently with the 267 episodes of bacteremia or fungemia included lung, 24%; catheter, 16%; urinary tract, 15%; peritoneum, 11%; surgical site, 5%; and other, 6%. Another 25% of bloodstream infections occurred without any identifiable primary source. Men accounted for 159 (59.6%) and women 108 (40.4%) of these patients, without a difference in overall mortality (men, 18.2% vs women, 20.4%; P = .66) or length of stay (men, 28 [3] days vs women, 27 [3] days; P = .93). There were no statistically significant differences in outcomes following urinary tract infections.

Univariate comparisons of 205 men and 121 women with pneumonia are given in Table 5. Of all pneumonia cases, 94% were hospital-acquired with no difference between men and women. The previously noted increase in mortality among females occurred despite similar severity of illness, age, time from admission to initiation of treatment, and overall duration of treatment. Length of stay was longer for men, but was not attributable to early deaths among women, since the time to discharge after intervention for men was similar for survivors and nonsurvivors (28 [2] days vs 28 [5] days), as it was for female survivors and nonsurvivors (23 [2] days vs 21 [3] days). The mortality for ventilator-associated pneumonia was 23.2% for men and 40.8% for women (P = .03) while the mortality for nonventilator-associated pneumonia was 13.3% and 28.6% for men and women, respectively (P = .01). Among all causative organisms in pneumonia, gram-positive organisms accounted for 25% and 20% (P = .26), gram-negative organisms for 43% and 44% (P = .81), Staphylococcus aureus for 16% and 14% (P = .50), and Pseudomonas aeruginosa for 9% and 13% (P = .09) in men and women, respectively. Men with pneumonia received more penicillin derivatives as a proportion of all antibiotics used (38.7% vs 16.7%; P = .02), although there were no differences in the proportional use of any other antibiotic class (eg, cephalosporins, aminoglycosides, fluoroquinolones, vancomycin, or carbapenems), and mortality for pneumonia treated with penicillins was similar to that in the group treated with other agents (data not shown). Time from fever (≥38.5°C) to initiation of therapy among febrile men and women was similar (men, 20 [1] hours vs women, 23 [3] hours; P = .29).

The univariate comparison of variables potentially related to pneumonia-associated mortality is shown in Table 6. Using logistic regression analysis (Table 7), increasing APACHE II score, preexisting malignancy, diabetes mellitus, time from admission to diagnosis of infection of more than 7 days, organ transplantation, increasing age, and female gender were independent risk factors for pneumonia-associated mortality.

Previous clinical and experimental studies examining gender-related differences in infectious complications and mortality have provided inconsistent results. Our study of a large cohort of hospitalized surgical patients shows that gender is not a strong independent predictor of outcome when examining all infections. The in-hospital mortality for women with pneumonia, however, was almost twice that of men, and was still significantly increased after controlling for multiple other comorbid factors. Interestingly, there was also an increased mortality for women with soft tissue infection as previously described,¹⁷ although the sample size was inadequate to perform meaningful logistic regression analysis.

Despite sparse data in the recent literature regarding the specific relationship between gender and outcome in hospital-acquired pneumonia, some studies provide for an indirect comparison of these findings. Kollef et al⁹ have recently shown that female gender independently predicts mortality in a multivariate analysis of a large cohort of patients who were mechanically ventilated. Although we did not see increased mortality during our study in female patients who were mechanically ventilated, their study differed in that less than 5% of their patients had a diagnosis of pneumonia. Rello et al¹⁹ found no difference in pneumonia-associated mortality between men and women, although that study was smaller (62 patients) and had a higher overall mortality (60%), suggesting either a more rigorous definition of pneumonia or later diagnosis. Singh et al²⁰ also found that gender was not predictive of mortality among a cohort of 129 surgical patients with pulmonary infiltrates, although only 14% of patients were women and pneumonia accounted for only 30% of the pulmonary infiltrates.

Many of the concerns regarding gender disparity in ischemic cardiac disease originated from reported delays in the time from presentation to definitive treatment and decreases in the use of hospital resources for women.^4- 5 Our study demonstrated similar time intervals from admission to initiation of treatment among male and female patients with pneumonia, as well as time from fever to initiation of treatment. Although these are indirect measures of the physician's timeliness of administration of care, they do support a relatively similar approach to male and female patients with infection. The comparable efficiency of initiation of treatment, in addition to similar length of treatment and antibiotic use between genders, minimizes the likelihood that women received a worse process of care. Consistent with our findings, Pearson et al²¹ reported comparable objective measures of quality of care between men and women initially hospitalized with pneumonia.

One additional issue regarding the disparate outcome with pneumonia is the potential for intrinsic gender differences in the pathogenesis of disease and in the host's response to infection. Clinical series examining gender disparity in the host response to infection are scarce, although recent studies in the transplantation literature have alluded to hormonal and/or immunological diversity that may affect outcome in critically ill patients.^23- 24 Experimentally, survival from polymicrobial sepsis has been found to be highly gender-dependent, and presumably hormone-dependent in rodents.^11- 13 Among patients in the current study with pneumonia, women did have a higher WBC count and slightly lower temperature at time of diagnosis, perhaps suggesting a small gender-dependent difference in the physiologic response to infection. This finding, however, may alternatively represent a difference in disease progression in women related to a delay in diagnosis. In terms of response to treatment, the time interval from initiation of treatment to normalization of temperature and WBC count was similar between men and women. Making a large difference in phenotypic response between men and women remains unlikely. One additional etiologic factor is related to respiratory tract colonization in critically ill hospitalized patients. Although oropharyngeal colonization was not assessed in this study, gender differences in skin colonization leading to subsequent catheter-related infections have been reported.²⁵

We found that despite similar severity of illness and length of treatment and a slightly younger mean age, infected men were hospitalized longer following initiation of treatment for infection. A similar increase in length of stay was also observed in male patients developing pneumonia, a trend noted for both survivors and nonsurvivors. Compared with the mortality data, these seemingly contradictory findings may be related to a large subset of predominantly male trauma patients requiring prolonged inpatient management and rehabilitation without significant increases in the acute severity of illness scores. Because the duration of therapy was similar between genders, the increased length of stay most likely is attributable to post-treatment hospitalization, again implying causation more related to underlying diseases than the largely nosocomial infections studied here.

Identification of alterable factors that can significantly affect outcomes among patients with infections, particularly hospital-acquired pneumonia, remains a difficult task. The unique association between female gender and pneumonia-associated mortality suggests the need for further investigation of characteristics of presentation of disease, administration of care, pathogenesis of infection, and host response to address potential gender differences in infection-associated outcome. A detailed understanding of any such disparity between men and women may provide valuable information for improving interventions and outcomes in all infected patients.