Acute Pyelonephritis in Adults: Prediction of Mortality and Failure of Treatment

Background  To formulate a classification tool for early recognition of patients admitted with acute pyelonephritis (AP) who are at high risk for failure of treatment or for death.

Methods  A retrospective chart review of 225 patients (102 men) admitted with AP. We considered 13 potential risk factors in a multivariate analysis.

Results  Recent hospitalization, previous use of antibiotics, and immunosuppression were found to be independent correlates of the prevalence of resistant pathogens in both sexes. Additional predictors included nephrolithiasis in women and a history of recurrent AP in men. Prolonged hospitalization should be expected for a man with diabetes and long-term catheterization who is older than 65 years or for a woman of any age with the same characteristics, when the initial treatment was changed according to the results of urine culture. For mortality prediction, we derived an integer-based scoring system with 6 points for shock, 4 for bedridden status, 4 for age greater than 65 years, and 3 for previous antibiotic treatment for men and 6 points for shock, 4 for bedridden status, 4 for age greater than 65 years, and 3 for immunosuppression for women. Among patients with at least 11 points, the risk for in-hospital death was 100% for men and 91% for women.

Conclusions  Simple variables available at presentation can be used for risk stratification of patients with AP. The additional identification of certain risk factors by means of a carefully obtained history could contribute to early recognition of patients infected by resistant bacteria and optimize the selection of antimicrobial agents.

ACUTE INFECTION of the upper urinary tract (acute pyelonephritis [AP]) occurs in at least 250 000 adults per year in the United States,¹ representing the most severe form of urinary tract infection (UTI).^2,3 It often requires hospitalization and prolonged therapy, and when accompanied by bacteremia, AP has a mortality rate of 10% to 20%.^4,5 Efforts to provide emergency department physicians with simple tools for the prediction of antimicrobial resistance and risk for death are justified by the increasing prevalence of resistance of various uropathogens to combined trimethoprim and sulfamethoxazole, β-lactams, and fluoroquinolones (FQs)⁶ and the finding in different clinical studies that almost half of the patients who were initially empirically treated with antibiotics received inappropriate treatment.⁷ Although numerous studies have focused on antimicrobial resistance of bacteria that cause UTIs in general,^8-11 data on the ability to predict mortality and failure of treatment due to resistant pathogens in patients with AP are lacking. Moreover, despite the fact that AP in women has been extensively studied, scant information regarding AP in men exists.^12,13

The aims of the present study were to formulate a classification tool for early recognition of men and women admitted with AP who are at high risk for death, and to identify independent predictors of failure of treatment and of prolonged hospitalization.

We conducted a retrospective medical chart review of all admissions to the Third Department of Medicine, Athens University, Sotiria General Hospital, Athens, Greece, that were due to AP from January 1, 1997, through December 31, 2001. Criteria for diagnosis of AP were the isolation of the same pathogen from urine and blood cultures or the simultaneous presence of (1) fever, defined as an axillary temperature of 38°C or greater; (2) pyuria, defined as the presence of 10 or more leukocytes per high-power field in the centrifuged specimen; and (3) positive urine culture (>10⁵ colony-forming units [cfu]/mL).¹⁴ In all cases, a positive urine culture was necessary for inclusion in the study. Blood samples were obtained for cultures from all subjects. Factors that defined complicated APs included 1 or more of the following: (1) male sex; (2) age greater than 65 years; (3) the presence of an indwelling catheter or the use of intermittent catheterization; (4) greater than 100 mL of residual urine retained after voiding; (5) obstructive uropathy; (6) vesicoureteral reflux or other urologic abnormalities; (7) azotemia due to intrinsic renal disease; and (8) renal transplantation.¹⁴

Demographic data, medical history, clinical findings, biochemical variables, ultrasonography (US) of the urinary system, record of resistance of microbes to antibiotics, and treatment with antibiotics were recorded. In addition, we assessed the presence of each of the following 13 variables considered to be potential risk factors for death or failure of treatment: age greater than 65 years, change in initial treatment, recent (in the previous month) hospitalization for any reason, prostatic hypertrophy (confirmed by US and defined as a prostate volume >25 mL),¹⁵ recent (in the previous month) use of antibiotics, hydronephrosis (demonstrated on admission by US), nephrolithiasis (confirmed by US), bedridden status, septic shock at presentation, history of recurrent APs (defined as ≥3 episodes of AP documented by means of culture in the past 12 months), diabetes mellitus, immunosuppression, and long-term indwelling catheterization. Antimicrobial agents were administered empirically until the results of urine or blood cultures were known. The duration of hospital stay and outcome (discharge or death) were also recorded.

Urine specimens, obtained by means of the clean-catch midstream technique, in-and-out urethral catheterization, or preexisting long-term indwelling catheterization, were incubated by means of calibrated loops on blood-agar plates with 5% sheep blood. The cultures were incubated for 18 to 24 hours before quantitative reading. All bacterial counts with at least 10⁵ cfu/mL in pure culture or a yield of at least 2 species were considered to be significant bacteriuria. For blood cultures, 10 mL of venous blood was obtained from all patients and inoculated into aerobic and anaerobic bottles (Organon Teknika Corp, Durham, NC). At least 2 sets of blood cultures were obtained from each patient. Susceptibility tests were performed using the disk agar diffusion method and interpreted as described by the National Committee for Clinical Laboratory Standards.¹⁶

Statistical analysis was performed using SPSS software.¹⁷ Acute pyelonephritis due to bacteria resistant to trimethoprim-sulfamethoxazole, to combined ampicillin sodium and sulbactam sodium or combined amoxicillin sodium and clavulanate potassium, or to FQs (norfloxacin, ofloxacin, and ciprofloxacin hydrochloride), prolonged (≥10 days) hospitalization, and death due to AP were used as end points in the analysis.

We evaluated univariate correlations using the χ² or the Fisher exact test as appropriate for categorical data and the t test for continuous variables. All tests of significance were 2-tailed, and a P value of less than .05 was considered to be significant. Variables with a significant univariate correlation with mortality and prolonged hospitalization were considered in a logistic regression analysis using a backward-selection algorithm to identify the respective independent correlates. Furthermore, the coefficients of the regression model, expressed as multiples of a common denominator and rounded to the nearest interval, allowed for the transformation of the multivariate model into an integer-based scoring system, which assigned a weight (number of points) to each predictor and summed the weights of the predictors that were present for a subject.

Clinical and laboratory characteristics of patients according to sex are illustrated in Table 1. Overall median age was 67 years (range, 42-96 years). Women were older than men (median age, 70 and 61 years, respectively [range, 42-96 and 51-92 years, respectively]; P<.001). Women had a decreased level of consciousness on admission in a greater proportion (35.8%) compared with men (12.7%; P<.001).

The results of urine cultures are presented in Table 2. Escherichia coli accounted for 56.4% of overall cases, followed by enterococci (10.7%), Staphylococcus species (8.0%), Proteus mirabilis (6.7%), Enterobacter species (5.3%), and Pseudomonas aeruginosa (5.3%). Bacteriuria was accompanied by bacteremia (with the same pathogen) in 60 men (58.8%) and 75 women (60.9%). Uncomplicated APs accounted for only 5.7% of women (3.1% of the total study population).

Amoxicillin-clavulanate or ampicillin-sulbactam was used as the initial empirical treatment in 54% of cases, whereas second-generation cephalosporins were administered in 11%, combined ticarcillin sodium and clavulanate potassium in 10%, FQs in 6%, trimethoprim-sulfamethoxazole in 6%, third-generation cephalosporins in 3%, and the combination of a β-lactam plus an aminoglycoside in 10% of all subjects. The rate of resistance of E coli to ampicillin was 27%; to trimethoprim-sulfamethoxazole, 11%; and to at least 2 drugs, 10%. Enterococci were resistant to ampicillin in 7%, ciprofloxacin in 11%, vancomycin hydrochloride in 2%, and teicoplanin in 1% of cases, whereas P aeruginosa was resistant to combined piperacillin sodium and tazobactam sodium in 17%, ticarcillin in 19%, ciprofloxacin in 40%, netilmicin sulfate in 23%, ceftazidime in 30%, and imipenem in 8% of subjects. Overall susceptibility of bacteria isolated from urine culture to amoxicillin-clavulanate or ampicillin-sulbactam was 86%; to trimethoprim-sulfamethoxazole, 67%; to FQs, 91.5%; and to netilmicin, 80%. Overall resistance of bacteria to all of these antibiotics was 4%. Multiresistant pathogens, ie, pathogens resistant to amoxicillin-clavulanate or ampicillin-sulbactam, trimethoprim-sulfamethoxazole, and FQs, included 2 strains of Acinetobacter species, 3 strains of P aeruginosa, and 4 strains of Klebsiella species. Among the 9 patients who were infected by multiresistant strains, 4 (44%) had recently used antibiotics (P = .02 compared with recent users of antibiotics who had uropathogens susceptible to at least 1 of the above 4 agents).

Men with (49/102 [48.0%]) and without (53/102 [52.0%]) prostatic hypertrophy did not differ in terms of the prevalence of bacteria resistant to trimethoprim-sulfamethoxazole (21/49 [42.9%] vs 16/53 [30.2%]; P = .18), to amoxicillin-clavulanate or ampicillin-sulbactam (3/49 [6.1%] vs 1/53 [1.9%]; P = .35), or to FQs (1/49 [2.0%] vs 2/53 [3.8%]; P = .60). Moreover, no significant differences were found between these groups with regard to mortality (7/49 [14.3%] vs 5/53 [9.4%]; P = .54) and prolonged hospitalization (7/49 [14.3%] vs 5/53 [9.5%]; P = .54).

Table 3 shows the comparisons between men and women in terms of historical features, antibiotic resistance, and clinical course of the AP episode. The variables identified by univariate analysis to be associated with each of the end points are presented in Table 4 and Table 5, whereas the independent risk factors are shown in Table 6. Table 7 shows the independent predictors of prolonged hospitalization and death. Although the presence of hydronephrosis was found to be associated with prolonged hospitalization and resistance to trimethoprim-sulfamethoxazole in univariate analysis, it was not isolated as an independent risk factor for the end points in the regression model.

The coefficients of the 4 independent correlates of mortality, divided by 0.5 and rounded to the nearest interval, allowed a simpler reexpression of the final regression model, in the form of the following integer-based scoring system:

Number of Points (Men) = 6 for Septic Shock + 4 for Bedridden Status + 4 for Age >65 Years + 3 for Previous Antibiotic Treatment

and

Number of Points (Women) = 6 for Septic Shock + 4 for Bedridden Status + 4 for Age >65 Years + 3 for Immunosuppression

According to this prediction tool (Figure 1), most patients (89% of men and 87% of women) could be classified in the subgroups with low scores (<7 points), which predict a very limited risk for death (0%-2.5% for men and 0%-2.3% for women). Patients classified in the highest-score subgroup (≥11 points) had a very high risk for death (100% for men and 91% for women).

Acute pyelonephritis is the only UTI that may require hospital admission and could be fatal in patients with immunosuppression,¹⁸ elderly individuals,¹⁹ pregnant women,²⁰ and patients with diabetes.²¹ Thus, in patients admitted for AP, the early identification of subgroups of patients at high risk for in-hospital death, prolonged hospitalization, and/or failure of treatment might contribute to ameliorative treatment.

In the present study, women admitted for AP were older than men (Table 1), more frequently bedridden, and more likely to undergo long-term catheterization and to be immunosuppressed, and they more often had a history of nephrolithiasis, diabetes mellitus, and recent hospitalization (Table 3). Therefore, they had a higher incidence of infection with resistant bacteria. Because of this difference, and because of the fact that few studies on AP have analyzed data regarding men,^12,13 we decided to perform separate analyses of data regarding men and women.

Our study provides additional support for the view that, although susceptibility patterns have changed, the spectrum of microorganisms that cause community-acquired AP has remained relatively constant. Among the overall population, E coli accounted for 57% of cases, whereas the rate of its resistance to trimethoprim-sulfamethoxazole (11%) was lower than the corresponding estimated rate in Greece during the same period (17%).²² Because cost-effectiveness analyses have shown that the threshold trigger for empirical FQ use in E coli– related UTI should be a trimethoprim-sulfamethoxazole resistance of more than 20%,²³ and that the effect of the latter on clinical outcomes depends on the level of resistance in the community,²⁴ administration of trimethoprim-sulfamethoxazole might be recommended as a first-line treatment for community-acquired AP due to E coli in Greek settings. However, considering that in our study E coli was responsible for only 57% of APs, 51% of all other bacteria causing AP were resistant to trimethoprim-sulfamethoxazole, and the causative agent of AP is not usually known at the time of admission to the hospital, trimethoprim-sulfamethoxazole is not recommended as a first-line treatment for AP.

In accord with previous reports,^10,24 a history of recent hospitalization and use of antibiotics correlated independently with the prevalence of any trimethoprim-sulfamethoxazole–resistant bacteria in men and women. Thus, clinicians must consider the extent of the patient's exposure to hospital pathogens and the unique bacteriological features of each hospital when devising antibiotic strategies.

In contrast to the results of several studies yielding resistance rates as high as 54% in European settings,^25-27 susceptibility of all bacteria to amoxicillin-clavulanate or to ampicillin-sulbactam was sufficiently high in our population, and immunosuppression was the only independent predictor of the prevalence of bacteria resistant to these drugs in both sexes. The susceptibility pattern in the present study corroborates recent reports by the World Health Organization antimicrobial resistance monitoring²⁸ and by the first international multicenter epidemiological survey of antimicrobial susceptibility of uropathogens,²⁹ according to which no firm evidence of increased resistance to these drugs exists so far.

Overall resistance rate to FQs in the population studied (8.5%) was comparable to the rate observed in our community (6.5%) at the beginning of the present study (1997)^30,31 and the rate found in other European settings during the same preriod (6%).³² In line with previous reports,^33,34 the present study yielded a history of recurrent APs for men and a history of nephrolithiasis for women as independent determinants of the prevalence of FQ-resistant bacteria. After considering cost-effectiveness and the risk for emergence of resistance, there appears to be little justification for using FQs routinely to treat APs, unless other agents are not well tolerated or are unlikely to be effective.

A history of recurrent UTIs, the only independent predictor of multidrug resistance in men in our study, constitutes a well-established risk factor for the prevalence of multiresistant pathogens, especially in elderly patients with underlying diseases.³⁵ Nevertheless, multivariate analysis isolated only nephrolithiasis as independently associated with multidrug resistance in women, which confirmed the results of previous studies that have shown women with renal lithiasis to have APs more frequently than men.^36,37

In line with recent reports,^21,38 the regression model showed that prolonged hospitalization for AP should be expected for a man with diabetes who is undergoing long-term catheterization and is older than 65 years, or for a woman of any age with both additional characteristics, whose initial empirical treatment was changed according to the results of urine culture. Regarding mortality, the multivariate analysis identified 5 factors that have been extensively investigated and found to convey a high risk for death in previous studies. Thus, age greater than 65 years,^19,39,40 bedridden status,^41,42 and the presence of septic shock at presentation⁴³ were the common independent predictors of death for both sexes, whereas additional multivariate correlates were recent use of antibiotics in men^44-47 and immunosuppression in women.⁴⁸

The mortality prediction model identified 3 groups of patients. The largest (90% of subjects) consisted of men and women with a score of less than 7 points who had a very limited risk for in-hospital death (<2.5%), whereas the smallest (8%) consisted of patients with a score of at least 11 points who were at high risk for death due to their disease (>90%). The ambiguous zone of patients with a score of 7 to 10 points included only 3 men (2.9%) and 5 women (4.1%) with corresponding risks for death of 67% and 75%, respectively. Thus, the cutoffs for achieving sufficiently high accuracy should be less than 7 or at least 11 points for the prediction of a favorable outcome or death, respectively. Despite significant differences between men and women regarding the mean age and the severity of the illness, risk factors for mortality, prolonged hospitalization, and prevalence of resistant bacteria were quite similar between the 2 sexes. Nevertheless, since the proposed model has been evaluated only in the patients from whom it was derived, its performance requires prospective validation in independent, external populations before it is routinely applied.

Taking into account the rates of resistance in our study, patients with AP due to E coli could be treated with trimethoprim-sulfamethoxazole, whereas when the pathogen is not known, empirical treatment with trimethoprim-sulfamethoxazole is not recommended. Amoxicillin-clavulanate or ampicillin-sulbactam might be recommended as first-line treatment for APs acquired in communities with susceptibility patterns similar to those of our country. Fluoroquinolones should not be administered empirically as first-line treatment for AP, except in cases in which the risk factors for prevalence of bacteria resistant to other antibiotics are present and in men with hypertrophy of the prostate, for whom the penetration of FQs is superior compared with that of other agents.

The proposed prediction model requires the application of simple information available at the emergency department and transformed to an easily remembered score. If validated, the model could be used to confirm the necessity of hospitalization for seriously ill patients who require intravenous treatment. Therefore, the model may contribute to a more objective risk stratification of patients presenting with AP, thus optimizing the selection of antimicrobial agents, ensuring prompt management of high-risk individuals, and directing those at low risk to less intensive and less costly levels of care.

Corresponding author and reprints: Angelos V. Pefanis, MD, Third Department of Medicine, Athens University, Medical School, Sotiria General Hospital, 152 Mesogion Ave, Athens 11527, Greece (e-mail: gpp@hol.gr).

Accepted for publication August 16, 2002.