a b s t r a c t

Objectives: There are no guidelines regarding the hospitalization of female patients with acute pyelo- nephritis (APN); therefore, we performed a retrospective analysis to construct a clinical prediction model for hospital admission.

Methods: We conducted a retrospective analysis of a prospective database of women diagnosed as having APN in the emergency department between January 2006 and June 2012. Independent risk factors for admission were determined by multivariable logistic regression analysis in half of the patients in this database. The risk of admission was categorized into 5 groups. The internal and external validations were conducted using the remaining half of the patients and 192 independent patients, respectively.

Results: Independent risk factors for admission were age of 65 years or greater (odds ratio [OR], 2.62; 1 point), chill (OR, 2.40; 1 point), and the levels of segmented neutrophils greater than 90% (OR, 2.00; 1 point), serum creatinine greater than 1.5 mg/dL (OR, 2.41; 1 point), C-reactive protein greater than 10 mg/dL (OR, 2.37; 1 point), and serum albumin less than 3.3 g/dL (OR, 7.36; 2 points). The admission risk scores consisted of 5 categories, which were very low (0 points; 5.9%), low (1 point; 10.7%), intermediate (2 points; 20.7%), high (3-4 points; 51.9%), and very high (5-7 points; 82.8%) risk, showing an area under the curve of 0.770. The areas under the curve of the internal and external validation cohorts were 0.743 and 0.725, respectively.

Conclusion: This model can provide a guideline to determine the admission of women with APN in the emergency department.

(C) 2013

Introduction

acute pyelonephritis (APN) is a common infection of the upper urinary tract that is most often observed in young healthy women and has an annual incidence of 15 to 17 cases per 10 000 women [1]. In the United States, more than 250 000 cases yearly are treated by primary care physicians, and ~ 200 000 patients require hospitaliza- tion [2,3]. Acute pyelonephritis cases are classified as uncomplicated or complicated based on the clinical status. Uncomplicated pyelone- phritis is caused by a number of typical pathogens in an immuno- competent patient without conditions predisposing to the anatomical

? Financial support/conflict of interest: Nothing to declare.

* Corresponding author. Department of Emergency Medicine, Seoul National University Bundang Hospital, 166 Gumi-ro, Bundang-gu, Seongnam 463-707, Korea. Tel.: + 82 31 787 7572; fax: + 82 31 787 4055.

E-mail address: [email protected] (K. Kim).

or functional impairment of the upper urinary tract [4]. In contrast, complicated pyelonephritis can be caused by anatomical or func- tional abnormalities such as polycystic kidney disease, pregnancy, indwelling urinary catheter, or recent urinary tract instrumentation. In addition, diabetes and immunosuppression are also associated with complicated pyelonephritis and can contribute to an increased risk of treatment failure [5].

Cases of complicated pyelonephritis require hospitalization and treatment with intravenous antimicrobials; however, uncomplicated cases may also involve severe symptoms such as high fever, high white blood cell count, vomiting, dehydration, and evidence of sepsis, and also require hospitalization.

The correct decision of hospitalization for the patients with APN may be important to reduce mortality, complication, recurrence, and cost. However, there are no predictive criteria or grading systems regarding hospital admission for APN, and thus, this decision is at the discretion of the attending physician. Therefore, in the present study, we formulated a scoring system to evaluate risk factors indicating

0735-6757/$ - see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2013.03.048

hospital admission for patients with APN upon visiting our emer- gency department (ED).

Methods

Study design

We performed a retrospective analysis of a prospective database, which was created to assess the effectiveness of an institutional admission protocol for women with APN [6] and to derive a scoring system for Internal validation. Furthermore, external validation was performed retrospectively using records from an independent aca- demic hospital ED. Upon the completion of the present study, we continued to collect patient data. This retrospective analysis was approved by our institutional review board and exempted from obtaining informed consent from the patients.

Study setting and population

A retrospective database was created from records collected from a single, urban, academic hospital ED with an annual census of 75 000 patients between January 2006 and June 2012. The ED was staffed by board-certified emergency medicine attending physicians and rotat- ing residents.

All consecutive APN female patients 15 years or older were eligible for enrollment. Acute pyelonephritis was diagnosed by (1) body temperature (BT) higher than 38?C, (2) pyuria, and (3) costovertebral angle tenderness. Pyuria was defined as the presence of more than 10 leukocytes per high-power field (HPF) in a centrifuged urine speci- men. If the patient had received any antipyretics within 4 hours before visiting the ED, a BT higher than 38?C was not included in the diag- nostic criteria. The inclusion of the patients of this study was made by our own review, not by the diagnosis of the physicians caring for the patients. The exclusion criteria of this analysis were obstructive APN, patient transfer to another hospital, and death in the ED.

Records between January 2009 and December 2009 were collected for external validation from a second urban, academic hospital ED with an annual census of 50 000 patients, where the severity of patients is higher than that in the institution used in the derivation of the model. Eligible patients for the external validation cohort were those with a discharge diagnosis of APN. After reviewing all medical records of the eligible patients, we excluded those using the same criteria as described above. This cohort was validated in a previous study regarding the prediction of bacteremia in APN cases [7].

Study protocol

Since 2006, we have used an APN management protocol to reduce the number of admissions into our hospital from our ED [6]. According to the protocol, patients with APN were intravenously hydrated and administered antiemetic agents. Antipyretic agents (intravenous ketorolac or oral acetaminophen) were administered if BT was higher than 38?C, and Intravenous antibiotics were dispensed as needed after sample culturing. After 6 hours of observation, each case was reeval- uated in the ED to determine the final disposition. Discharge was determined if the patient met all the following criteria: (1) capable of receiving Oral medications, (2) BT lower than 39?C, and (3) Shock Index less than 1 (heart rate [HR]/systolic blood pressure [SBP]).

Measures

We constructed a prospectively collected database that included the following parameters: age; initial symptoms including fever, chill, vomiting, and dysuria; and presence of underling diseases such as diabetes, hypertension, Chronic liver disease, chronic kidney disease, stroke, and cancer [6,7]. The initial vital signs and laboratory

results were collected, which included SBP, diastolic blood pres- sure, HR, respiration rate (RR), BT, WBC, urine WBC, and the levels of segmented neutrophils, hemoglobin, platelets, blood urea nitrogen, serum creatinine, C-reactive protein (CRP), serum albumin, and urine nitrogen.

The primary outcome of this analysis was mandatory admission to a ward directly after the initial ED visit or subsequent admission after a revisit within 7 days via the ED or outpatient department.

Statistical analysis

Continuous variables are presented as the mean with SDs and were compared using the Student t test. Binomial variable analysis results are presented as the frequency of occurrence and were compared with the ?² test. Continuous data were categorized using clinical or statistical considerations, in which the variables (age >= 65 years, RR N 20 cycles/min, BT N 39?C, WBC b 4000 or N 12 000 cells/ mm³, segmented neutrophil level N 90%, hemoglobin level b 12 mg/dL, platelet count b 150 000 cells/mm³, serum creatinine level N 1.5 mg/ dL, CRP level N 10 mg/dL, serum Albumin level b 3.3 g/dL, and urine WBC N 50/HPF) were dichotomized by clinical relevancy or reference range, as described elsewhere [6-8]. Systolic blood pressure less than 90 mm Hg and HR more than 120 beats/min were dichotomized by statistical significance with an optimal cutoff point.

Univariate analysis was performed using dichotomized variables for admission, and multivariable logistic regression was conducted using the candidate variables with a significance of P b .05. A multi- variable logistic regression with the best-fit model was determined using Bayesian information criterion (BIC). When using maximal likelihood estimation, the likelihood can be increased by adding variables. Thus, BIC was used as a criterion for model selection by introducing a penalty term to the number of parameters in the model. When comparing fitted models, the smaller the BIC, the better the fit [9]. The prediction rules for admission were developed from the logistic regression models using a regression coefficient-based scoring method [10,11]. Integer values were assigned after dividing each ? coefficient by 1.0 and rounding the number up. The overall risk score was the sum of each variable amount. The risk of hospital admission was stratified as very low, low, intermediate, high, and very high. The discrimination of the model was quantified using an area under the receiver operating characteristic (ROC) curve, and the calibration was assessed with stratification tables. All analyses were performed using STATA version 11.0 statistical software (Stata Corp, College Station, TX).

Results

In this prospective cohort, a total of 1717 cases were screened, with 100 being excluded, including 32 cases of obstructive APN, 66 hospital transfers, and 2 patients who died in the ED. Of the 1617 eligible cases, 809 (50%) were randomly assigned to the derivation set and 808 (50%) to the internal validation set. A total of 385 (23.8%) patients were hospitalized. In the external validation set, 198 APN cases were enrolled, with 6 excluded because of obstructive APN. Finally, 192 cases were included for the analysis. The baseline demo- graphics and clinical characteristics of the derivation and validation cohorts are described in Table 1.

Univariate analysis was performed to predict hospital admission in the derivation cohort. As shown in Table 2, age of 65 years or greater, fever, chill, vomiting, diabetes, hypertension, chronic kidney disease, cancer, SBP less than 90 mm Hg, HR more than 120 beats/min, WBC less than 4000 or more than 12 000/mm³, segmented neutrophil level higher than 90%, hemoglobin level lower than 12 mg/dL, platelet count less than 150 000/mm³, serum creatinine level higher than 1.5 mg/dL, CRP level higher than 10 mg/dL, Serum albumin level lower

Table 1

Baseline characteristics of all cohort groups

Derivation (n = 809)

Internal validation (n = 808)

External validation (n = 192)

P

Derivation vs

Derivation vs

				internal validation	external validation
Admissions, n (%)	191 (23.6)	194 (24.0)	67 (34.9)	.850	.001
Age (y)	50.9 +- 18.8	51.5 +- 19.2	56.0 +- 18.5	.562	.001
Initial symptoms, n (%)
Fever	666 (82.3)	652 (80.7)	164 (85.4)	.398	.306
Chill	556 (68.7)	561 (69.4)	124 (64.6)	.759	.269
Vomiting	183 (22.6)	180 (22.3)	54 (28.1)	.869	.107
Dysuria	238 (29.4)	224 (27.7)	54 (28.1)	.450	.723
comorbid diseases, n (%)
Diabetes	136 (16.8)	120 (14.9)	44 (22.9)	.280	.059
Hypertension	210 (26.0)	213 (26.4)	71 (37.0)	.854	.002
Chronic liver disease	12 (1.5)	16 (2.0)	5 (2.6)	.444	.280
Chronic kidney disease	21 (2.6)	16 (2.0)	10 (5.2)	.408	.060
Stroke	22 (4.1)	31 (3.8)	10 (5.2)	.803	.488
Cancer	53 (6.6)	37 (4.6)	27 (14.1)	.084	.001
Initial vital signs
SBP (mm Hg)	125.9 +- 21.8	127.2 +- 22.7	126.3 +- 22.4	.218	.820
DBP (mm Hg)	69.1 +- 13.1	69.9 +- 13.7	72.5 +- 14.1	.221	.003
HR (beats/min)	98.6 +- 18.9	99.0 +- 18.6	100.5 +- 18.5	.658	.216
RR (breaths/min)	19.8 +- 2.4	19.8 +- 2.1	20.7 +- 5.9	.869	.001
BTs (?C)	37.9 +- 1.2	38.0 +- 1.2	37.8 +- 1.3	.574	.153
Laboratory findings WBC (x103/mm3)	11.8 +- 5.0	11.8 +- 4.8	11.7 +- 5.0	.807	.890
Segmented neutrophils (%)	81.4 +- 10.1	82.0 +- 10.1	83.0 +- 10.3	.289	.063
Hemoglobin (mg/dL)	12.5 +- 1.4	12.5 +- 1.4	12.0 +- 1.8	.338	b.001
Platelet (x103/mm3)	225.6 +- 82.5	227.1 +- 86.5	226.5 +- 104.4	.738	.909
BUN (mg/dL)	15.7 +- 9.9	15.7 +- 10.2	20.0 +- 17.3	.991	.001
Serum creatinine (mg/dL)	1.0 +- 0.5	1.0 +- 0.6	1.3 +- 1.1	.855	b.001
CRP (mg/dL)a	9.7 +- 8.9	9.9 +- 8.9	11.2 +- 9.7	.638	.061
Serum albumin (g/dL)	4.0 +- 0.5	4.0 +- 0.5	3.6 +- 0.6	.630	b.001
Urine WBC N 50 (/HPF), n (%)	303 (37.5)	313 (38.7)	99 (51.6)	.589	b.001
Positive urine nitrite, n (%)	334 (41.3)	357 (44.2)	80 (41.9)	.239	.880

Data were presented as means with SDs or number with percentage. DBP, diastolic blood pressure; BUN, blood urea nitrogen.

^a CRP not performed in 45 cases in prospective cohorts.

than 3.3 g/dL, urine WBC more than 50/HPF, and positive urine nitrite were significantly associated with hospital admission.

As shown in Table 3, the final multivariable logistic regression model included the following variables: age of 65 years or greater, chill, and the levels of segmented neutrophils greater than 90%, serum creatinine greater than 1.5 mg/dL, CRP greater than 10 mg/dL, and serum albumin less than 3.3 g/dL. Each point was allocated according to the regression coefficient as described in Table 3. The model was applied to categorize the cohort into very low (5.9%), low (10.7%), intermediate (20.7%), high (51.9%), and very high (82.8%) risks of hospital admission. An ROC curve of stratification model was drawn, and its area under the curve (AUC) was 0.770 (95% confi- dence interval [CI], 0.730-0.809; Fig. 1A).

In the internal validation cohort, the risk groups of hospital ad- mission were categorized into very low (7.8%), low (11.6%), inter- mediate (23.8%), high (43.2%), and very high (82.5%) risk, and the AUC of the stratification model was 0.743 (95% CI, 0.701-0.784; Fig. 1B).

As shown in Table 1, the external validation cohort consisted of

192 eligible patients with a discharge diagnosis of APN, with 67 (34.9%) being admitted to our hospital. This cohort was categorized into very low (16.7%), low (13.3%), intermediate (28.3%), high (55.1%), and very high (60.7%) risk of hospital admission, and the AUC of the model was 0.725 (95% CI, 0.649-0.801; Figs. 1C and 2).

Discussion

Recent guidelines for APN treatment have recommended oral antibiotics on an outpatient basis for uncomplicated pyelonephritis cases with mild symptoms, such as low-grade fever and normal or slightly elevated peripheral leukocyte count without nausea or

vomiting, whereas patients with complicated pyelonephritis should be admitted and managed with broad-spectrum empirical antimicro- bials [1,12-14]. However, patients with severe uncomplicated pyelo- nephritis (eg, accompanied by hemodynamic instability or intolerance to oral antimicrobials) should also be hospitalized.

Hospital admission rates of 28% to 60% have been reported for APN in several case series and our previous study [8,15]. This wide range of admission rates for APN reflects the varied decision policies among physicians and among institutions. Thus, we have used an APN management protocol since 2006 to reduce the admission rates for APN in our ED, and this protocol has been effective in the management of APN [6]. The initial admission rate was significantly lower than that of the before group (15.1% vs 47.7%, P b .001). However, the ED revisit rates after initial discharge between the before (11.8%) and after (15.3%) groups were not significantly different (P = .38). Furthermore, subsequent admission after revisit was also similar in the before and after groups (8.4% vs 5.9%, P =

.42). The ultimate admission, initial admission plus subsequent ad- missions after revisit, was significantly decreased from 52.1% to 20.1% after instituting the protocol.

To the best of our knowledge, this analysis is the first to identify predictive variables for the hospitalization of patients with APN using a risk stratification model. A previous study reported that risk factors for the hospital admission of APN cases included advanced age, higher temperature, diabetes, genitourinary tract abnormalities, and vomiting; however, this study did not include a scoring system or risk stratification analysis [16]. Our model can be used to decide the admission of patients with APN in the ED and showed that very- low-risk (5.9%) and low-risk (10.7%) risk patients can be discharged based on their admission risk profiles. In contrast, high-risk (51.9%) and very-high-risk (82.8%) patients should be considered for

Table 2

Univariate predictors of hospital admission in the derivation cohort (n = 809)

No admission Admission			P
	(n = 618)	(n = 191)
Age >=65 y	151 (24.4)	95 (49.7)	b.001
Initial symptoms, n (%)
Fever	496 (80.3)	170 (89.0)	.006
Chill	404 (65.4)	152 (79.6)	b.001
Vomiting	119 (19.3)	64 (33.5)	b.001
Dysuria	178 (28.8)	60 (31.4)	.489
Comorbid diseases, n (%)
Diabetes	81 (13.1)	55 (28.8)	b.001
Hypertension	127 (20.6)	83 (43.5)	b.001
Chronic liver disease	7 (1.1)	5 (2.6)	.138
Chronic kidney disease	10 (1.6)	11 (5.8)	.002
Stroke	21 (3.4)	12 (6.3)	.078
Cancer	31 (5.0)	22 (11.5)	.002
Initial vital signs
SBP b 90 mm Hg	5 (0.8)	7 (3.7)	.010
HR N 120 (beats/min)	51 (8.3)	26 (13.6)	.027
RR N 20 (breaths/min)	84 (13.6)	33 (17.3)	.206
BTs N 39 (?C)	106 (17.2)	41 (21.5)	.177
Laboratory findings WBC b 4, N 12 (x103/mm3)	259 (41.9)	116 (60.7)	b.001
Segmented neutrophils N 90%	79 (12.8)	58 (30.4)	b.001
Hemoglobin b12 (mg/dL)	169 (27.4)	91 (47.6)	b.001
Platelet b150 (x103/mm3)	74 (12.0)	43 (22.5)	b.001
Serum creatinine N 1.5 (mg/dL)	31 (5.0)	45 (23.6)	b.001
CRP N 10 (mg/dL)a	196 (31.7)	119 (62.3)	b.001
Serum albumin b3.3 (g/dL)	15 (2.4)	38 (19.9)	b.001
Urine WBC N 50 (/HPF), n (%)	216 (35.0)	87 (45.6)	.008
Positive urine nitrite, n (%)	240 (38.8)	94 (49.2)	.011

Data were presented as number with percentage. BUN, blood urea nitrogen.

^a CRP not performed in 20 cases.

admission rather than discharge. However, those in the intermedi- ate-risk (20.7%) group should be potential candidates for discharge and outpatient management, but a decision to discharge a patient in this risk group cannot be determined simply according to presenting medical conditions, without nonmedical circumstances (eg, patient preference for admission, bed availability in ward, and ED over- crowding) also being taken into consideration. In the intermediate- risk cases, an alternative option is short-term observation in the ED depending on the policies of each hospital.

Here, we developed an APN admission risk stratification scoring system for use in the ED based on 6 variables: age of 65 years or greater, chilling sensation, segmented neutrophil level higher than 90%, serum Creatinine levels higher than 1.5 mg/dL, CRP greater than 10 mg/dL, and serum albumin levels lower than 3.3 g/dL. Although we previously reported that advanced age was not independently asso- ciated with admission for APN [8], in the present study, it was an independent factor, which may be explained by the different sam- ple size and exclusion criteria. The exclusion criteria in the previous study were severe sepsis or septic shock, an immunocompromised state (on chemotherapy or immunosuppressive medication), history of chronic renal failure, acute renal failure (serum creatinine level N 2 mg/dL), neurogenic bladder, patients with urinary catheters, and previous kidney transplantation except obstructive APN. Although a

Fig. 1. Area under the ROC curves of the risk stratification model for predicting admis- sion in the derivation (A), internal validation (B), and external validation (C) cohorts.

chilling sensation is somewhat subjective, in febrile patients, it was associated with the incidence of bacteremia in many studies, and the

degree of the chills has been reported to be important to predict

Table 3

Risk score model with best fit for predicting hospital admission

Variables	? Coefficient	OR (95% CI)	P	Score
Age >=65 y	0.96	2.62 (1.79-3.85)	b.001	1
Chill	0.88	2.40 (1.54-3.74)	b.001	1
Segmented neutrophils N 90%	0.69	2.00 (1.28-3.13)	.002	1
Serum creatinine N 1.5 mg/dL	0.88	2.41 (1.35-4.30)	.003	1
CRP N 10 mg/dL	0.86	2.37 (1.62-3.42)	b.001	1
Serum albumin b 3.3 g/dL	2.00	7.36 (3.69-14.68)	b.001	2
OR, odds ratio.

the risk of bacteremia and a higher risk of death with a positive blood culture [17-21]. A higher percentage of segmented neutrophils were also a prognostic factor for predicting bacteremia in patients with severe infections [7,22]. Thus, in this context, chills and a higher percentage of segmented neutrophils appeared to be asso- ciated with hospital admission in the present analysis.

Herein, an initial serum creatinine level higher than 1.5 mg/dL was independently correlated with admission in APN cases in the current study, and a reportedly higher initial serum creatinine was independently associated with prolonged fever in hospitalized

Fig. 2. Admission rates of patients in each risk group and cohort.

patients with APN [23], which was also the most important predictor of outcome in Emphysematous pyelonephritis [24]. Based on these results, we can assume that an elevated serum creatinine level is indicative of a poor prognosis in APN and is likely caused by poor renal function resulting in lower renal antibiotic concentrations and a lower clearance of bacterial pyrogens.

The CRP level is a prognostic factor of the severity of many in- fectious diseases and is associated with hospital mortality in patients with community-acquired pneumonia. Some studies have shown that when CRP levels were considered with the pneumonia severity index, the predictive capabilities of 30-day mortality improved [25,26].

The strongest association with hospital admission was a low serum albumin level. As it is commonly measured in hospitalized patients, the serum albumin concentration may be an important prognostic factor for infectious diseases. In community-acquired pneumonia, the serum albumin concentration was related to mor- tality [27,28], and a low concentration was identified as the most reliable risk factor in community-acquired severe sepsis and septic shock [22]. Furthermore, initial hypoalbuminemia might result from nutritional deficiencies, infectious processes, or underlying dis- eases; therefore, the mortality of infectious diseases might be a direct result of these underlying conditions. However, the exact mechan- isms of hypoalbuminemia remain unknown, and thus, further re- search is warranted.

Here, a risk stratification model of hospital admission in APN cases was initially developed and validated. In the derivation cohort, the AUC was 0.770 (95% CI, 0.730-0.809), and the AUC of the stratification models was 0.743 (95% CI, 0.701-0.784) and 0.725 (95% CI, 0.649-0.801) in the internal and external validation cohorts, respectively, indicating a good overall discriminating power of the model. Furthermore, the Predictive ability of the risk stratification model was also reasonable in all cohorts. Considering the perfor- mance of this risk stratification model for APN in the internal and external validations, we believe that it could be used as a tool to decide hospital admission from the general ED.

There are several limitations to this study. First, the determination of patient admission was at the discrimination of the attending phy- sicians and nonmedical patient conditions, as described above. How- ever, in the cohorts used in the derivation and internal validation analyses, an admission protocol as described herein was in place, although following it was not mandatory. Because the risk stratifi- cation model was confirmed with another data set originating from a hospital without an admission protocol, this model might be gen- erally useful. Second, we arbitrarily conducted the dichotomization of continuous predictor variables because dichotomization has the advantage that a binary variable can lead to the easier interpretation of results than a continuous variable. Third, the external validation was performed retrospectively, and potential bias remained. Thus, a

prospective study should be considered in the future. Fourth, the admission rate and some other variables showed significant dif- ferences between the baseline characteristics of the prospective and external validation cohorts. However, in terms of generalizability, this difference could be representative of a strength of the proposed model. Fifth, patients who were discharged from our hospital may have subsequently visited another ED within 7 days. However, our hospital is the largest and only referral hospital in our region, and one of our discharge plans for patients with APN is outpatient department follow-up. As a result, it is unlikely that patients visited another ED after discharge. Finally, in the external validation of the risk stratification model, the risk rates were 16.7% and 13.3% for the very-low-risk and low-risk groups, respectively, indicating a minor discrepancy within this model that might have originated from the different admission rates between the 2 data sets (23.8% vs 34.9%). However, in terms of clinical application, both patient groups could be considered for hospital discharge; thus, in this aspect, these dis- crepancies could be only minor concerns.

We believe that this risk stratification model of APN could be used as a tool to decide hospital admission from the ED. Further- more, this model can be applicable in the general clinical ED setting.

Conclusion

In conclusion, the significant predictors of hospital admission in patients with APN include 65 years or greater, chills, and the levels of segmented neutrophils greater than 90%, serum creatinine greater than 1.5 mg/dL, CRP greater than 10 mg/dL, and serum albumin less than 3.3 g/dL. This simple risk stratification model provides estimates of admission risk that may guide Clinical decisions. Future studies are needed for prospective and other population settings.

References

1. Czaja CA, Scholes D, Hooton TM, et al. Population-based epidemiologic analysis of acute pyelonephritis. Clin Infect Dis 2007;45:273-80.
2. Hooton TM, Stamm WE. Diagnosis and treatment of uncomplicated urinary tract infection. Infect Dis Clin North Am 1997;11:551-81.
3. Foxman B, Klemstine KL, Brown PD. Acute pyelonephritis in US hospitals in 1997: hospitalization and in-hospital mortality. Ann Epidemiol 2003;13:144-50.
4. Bergeron MG. Treatment of pyelonephritis in adults. - Abstract - UK PubMed Central. medical clinics of North America; 1995.
5. Bent S, Nallamothu BK, Simel DL, et al. Does this woman have an acute uncomplicated urinary tract infection? JAMA 2002;287:2701-10.
6. Kim K, Lee CC, Rhee JE, et al. The effects of an institutional care map on the admission rates and Medical costs in women with acute pyelonephritis. Acad Emerg Med 2008;15:319-23.
7. Kim KS, Kim K, Jo YH, et al. A simple model to predict bacteremia in women with acute pyelonephritis. J Infect 2011;63:124-30.
8. Rhee JE, Kim K, Lee CC, et al. The lack of association between age and diabetes and hospitalization in women with acute pyelonephritis. JEM 2011;41:29-34.
9. Schwarz G. Estimating the dimension of a model. Ann Stat 1978;6:461-4.
10. Tu JV, Naylor CD. clinical prediction rules. J Clin Epidemiol 1997;50:743-4.
11. Moons KGM, Harrell FE, Steyerberg EW. Should scoring rules be based on odds ratios or regression coefficients? J Clin Epidemiol 2002;55:1054-5.
12. Warren JW, Abrutyn E, Hebel JR, et al. Guidelines for Antimicrobial treatment of uncomplicated acute bacterial cystitis and acute pyelonephritis in women. Infectious Diseases Society of America (IDSA). Clin Infect Dis 1999;29:745-58.
13. Nicolle LE. Uncomplicated urinary tract infection in adults including uncompli- cated pyelonephritis. Urol Clin North Am 2008;35:1-12.
14. Gupta K, Hooton TM, Naber KG, et al. International clinical practice guidelines for the treatment of acute uncomplicated cystitis and pyelonephritis in women: a 2010 update by the Infectious Diseases Society of America and the European Society for Microbiology and Infectious Diseases. Clin Infect Dis 2011;52: e103-20.
15. LE N. Management of acute uncomplicated pyelonephritis. In: Bergan T, editor. Urinary Tract Infections. Infectiology; 1997.
16. Pinson AG, Philbrick JT, Lindbeck GH, et al. ED management of acute pyelo- nephritis in women: a cohort study. Am J Emerg Med 1994;12:271-8.
17. Leibovici L, Greenshtain S, Cohen O, et al. Bacteremia in febrile patients: a clinical model for diagnosis. Arch Intern Med 1991;151:1801-6.
18. Van Dissel JT, Schijf V, Vogtlander N, et al. Implications of chills. Lancet 1998; 352:374.
19. Stryjewski ME, Kanafani ZA, Chu VH, et al. Staphylococcus aureus bacteremia among patients with health care-associated fever. Am J Med 2009;122:281-2.
20. Tokuda Y, Miyasato H, Stein GH, et al. The degree of chills for risk of bacteremia in acute febrile illness. Am J Med 2005;118:1417.
21. Bryan CS. Clinical implications of positive blood cultures. Clin Microbiol Rev 1989;2:329-53.
22. Artero A, Zaragoza R, Camarena JJ, et al. Prognostic factors of mortality in patients with community-acquired bloodstream infection with severe sepsis and septic shock. J Crit Care 2010;25:276-81.
23. Behr MA, Drummond R, Libman MD, et al. Fever duration in hospitalized acute pyelonephritis patients. AJM 1996;101:277-80.
24. Wan YL, Lo SK, Bullard MJ, et al. Predictors of outcome in emphysematous pyelonephritis. J Urol 1998;159:369-73.
25. Menendez R, Martinez R, Reyes S, et al. Biomarkers improve Mortality prediction by prognostic scales in community-acquired pneumonia. Thorax 2009;64:587-91.
26. Chalmers JD, Singanayagam A, Hill AT. C-reactive protein is an independent predictor of severity in community-acquired pneumonia. Am J Med 2008;121: 219-25.
27. Lim WS, van der Eerden MM, Laing R, et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax 2003;58:377-82.
28. Charles PGP, Wolfe R, Whitby M, et al. SMART-COP: a tool for predicting the need for intensive respiratory or vasopressor support in community-acquired pneu- monia. Clin Infect Dis 2008;47:375-84.