Point-of-Care Urine Gram Stain for Managing Febrile Urinary Tract Infection in Adults

Febrile urinary tract infections (fUTIs), which include pyelonephritis, prostatitis, and urosepsis, are the most common cause of sepsis. However, the treatment has become more complex because of the world-wide increase in antimicrobial resistance (AMR). The objective of this study was to clarify whether point-of-care Gram stain (PCGS) of urine contributed to fUTI diagnosis and treatment in adults. This hospital-based observational study was undertaken between January 2013 and March 2015 in Okinawa, Japan. All enrolled patients were adults who had been admitted to the Division of Infectious Diseases with suspected fUTI. The usefulness of PCGS results were compared for urinalysis (U/A) and urine cultures (U/Cs). The targeted therapy type and its effectiveness based on PCGS were analyzed, and each was investigated in two groups: the uncomplicated pyelonephritis group and the complicated pyelonephritis/prostatitis group. 21], S. saprophyticus was not detected, but instead, S. aureus and A. urinae were cultured in this study.


Results
Two hundred sixty-six patients were enrolled. The results of PCGS were closely correlated with those of U/A for pyuria and bacteriuria, and moderately correlated with the results of U/C for bacterial types. In the uncomplicated group, narrow-spectrum antimicrobials such as cefotiam were initially selected in 97.9% (47/48) of patients, and their effectiveness was 97.9% (47/48).

Conclusion
Urine PCGS led to a more precise fUTI diagnosis and prompted clinicians to select narrower-spectrum antibiotics with high effectiveness.

Background
Febrile urinary tract infections (fUTIs), which include acute pyelonephritis, prostatitis, and urosepsis, can cause sepsis, septic shock, and death [1,2]. Febrile UTI diagnosis is not straightforward [3], and the treatment has become more complicated due to the increasing antimicrobial resistance (AMR) [2,4]. Ceftriaxone, a third-generation cephalosporin, is commonly recommended as an empirical treatment for uncomplicated pyelonephritis, while carbapenems are recommended for complicated pyelonephritis because of the concern about an increase in extended-spectrum beta-lactamases (ESBLs) [5].
Gram staining in clinical microbiology assists in diagnosing fUTI, while evidence for the fUTI treatment strategy is limited [3,6]. Point-of-care Gram stain (PCGS) performed by clinicians has reduced broad-spectrum antimicrobial overuse at Okinawa Chubu Hospital (OCH) in Japan [7]. Urine PCGS has been used here in children [8]. However, the usefulness of PCGS in diagnosing and treating fUTI in adults has never been investigated, and its clinical impact remains unclear.
The objective of this study was to clarify whether urine PCGS contributed to etiologic agent estimation and targeted treatment of fUTI in adults. In addition, fUTI patients were divided into uncomplicated and complicated groups because patients with complicated pyelonephritis are more likely to have AMR [9]. For each group, we investigated whether PCGS detected pyuria and bacteriuria, estimated bacterial types, and promoted narrower-spectrum antimicrobial use and effective treatment. On the basis of our ndings, we recommend using urine PCGS to ensure a more precise fUTI diagnosis and to select narrower-spectrum antimicrobial agents.

Study design and setting
This was a single hospital-based, retrospective observational study. The study setting was OCH, Japan. Approximately 39,000 patients visit the emergency room (ER) annually and nearly 7,000 patients are hospitalized after attending the ER each year [10]. Most patients with a suspected fUTI are initially examined in the ER. Adolescent and adult patients who are diagnosed with fUTI and who need to be hospitalized are admitted to the Division of Infectious Diseases. Patients with complicated pyelonephritis requiring surgical interventions such as double-J stent insertion, suprapubic cystostomy, or nephrostomy are admitted to the Department of Urology.

Case de nitions and data collection
Febrile UTI was considered if a patient had symptoms of systemic in ammation (fever, chills, and malaise) and bladder symptoms (urinary frequency, urgency, and dysuria), which was supported by urinalysis (U/A) results that showed pyuria or bacteriuria (or both) and urine culture (U/C) results that showed substantial concentrations of a uropathogen [1]. Febrile UTI patients were divided into the following two groups: the uncomplicated group, which included patients with uncomplicated pyelonephritis; and the complicated group, which included patients with complicated pyelonephritis and prostatitis. Uncomplicated pyelonephritis was de ned as acute pyelonephritis in women who did not have any underlying diseases [11]. Complicated pyelonephritis was de ned as acute pyelonephritis in patients who had underlying diseases such as neurogenic bladder, indwelling bladder catheter, diabetes mellitus, or prostate hypertrophy, or patients whose kidney had urolithiasis, abscess, or emphysematous pyelonephritis/cystitis. Unde ned cases were categorized into the complicated group [11]. Prostatitis was de ned as fUTI without ank pain, with prostate tenderness, or a signi cant increase in prostate speci c antigen (PSA) or an imaging diagnosis [9,12].
In this study, all patient information was collected from medical charts between January 2013 and March 2015. The inclusion criteria were all patients aged 15 years or older who were admitted to the Division of Infectious Diseases and nally diagnosed with fUTI. The exclusion criteria were as follows: 1) uncomplicated cystitis; 2) previous antibiotics exposure within 48 hours; 3) U/C not submitted; 4) not diagnosed with a fUTI after admission; 5) patients with co-infection other than fUTI; and 6) transferred to another hospital from the ER.

Urinalysis and urine culture
Clean-voided midstream urine or urine collected with a catheter was used for PCGS in the ER, and the residue specimen was then delivered to the laboratory for U/A and U/C. U/A was performed after centrifugation without staining by a technician and using an automated urine analyzer (Aution Hybrid, Arkray, Kyoto, Japan). The pyuria result was considered to be positive when ve or more leukocytes per a 400× magni cation eld (high power eld: HPF) were observed, and the number of leukocytes was quanti ed. The bacteriuria result was considered to be positive if an average of one or more bacteria was identi ed in a HPF of view, and the number of bacteria was semi-quanti ed as follows: 1+, an average of one or more in one visual eld; 2+, many in one visual eld; and 3+, a large number in one visual eld.
U/C was performed by a technician using uncentrifuged urine. The number of bacteria was quanti ed using quantitative medium (Caldip Plus, Merck, Darmstadt, Germany). If the urine sample amount was insu cient, this quanti cation was omitted, but semi-quanti cation was con rmed using culture plates. If Gram-positive rods such as Bacillus or Corynebacterium, coagulase-negative staphylococci other than Staphylococcus lugdunensis or Staphylococcus saprophyticus, or other skin or vaginal normal ora grew in the culture, it was considered to be contaminated and non-pathogenic.
Blood culture At least two sets of blood cultures, which included aerobic and anaerobic bottles with Bactec Plus resin medium (Becton, Dickinson and Company, Franklin Lakes, NJ, USA), were taken from the upper or lower limbs but not from the femoral vessels.
All bottles were incubated for at least 5 days using the Bactec 9240 system, which is an automated blood culture system [13]. Coagulase-negative staphylococci, Bacillus, Propionibacterium, Micrococcus, Clostridium, and α-streptococci were considered to be potential skin contaminants [13]. With the exception of α-streptococci, if any of these grew in only one set of blood cultures, it was considered to be a contaminant.

PCGS and antimicrobial selection
PCGS was performed and interpreted by physicians in the ER soon after urine samples were obtained. Uncentrifuged urine specimens were placed on glass slides, xed with a ame or hot air blower, and Gram-stained using Barmii M (Muto Pure Chemicals, Tokyo, Japan). We used a four-step of dyeing procedure including crystal violet, 2% iodine sodium hydroxide, acetone ethyl alcohol, and 0.1% fuchsine. The slides were examined in the ER by in-house staff members including trained resident physicians (postgraduate years 1 and 2) to identify each etiologic agent and select an appropriate targeted antimicrobial therapy [7,14]. Physicians examined the slides using conventional light microscopy at 100× magni cation and then at 1000× magni cation under immersion oil. If leukocytes were visible, the patient was considered to have pyuria, and the number of leukocytes was semi-quanti ed as follows: an average of <1 in one per visual eld (1000×) was ±; an average of 1 to 9 was 1+; 10 to 99 was 2+; and >99 was 3+. Similarly, if bacteria were present, the patient was considered to have bacteriuria, and the number of bacteria was semi-quanti ed as follows: an average of <1 per visual eld (1000×) was ±; an average of 1 to 9 was 1+; 10 to 99 was 2+; and >99 was 3+. All ndings were con rmed by senior residents (postgraduate years 3 to 5) and attending physicians.
Annual cumulative antimicrobial susceptibility test data were reported and distributed to medical staff by the bacteriology laboratory. The organisms that were estimated using PCGS led physicians to select antibiotics along with the cumulative antibiogram. For example, 1415 cultures were positive for Escherichia coli in 2015, and of which sensitivity to ampicillin was 62%, ampicillin/sulbactam 69%, cefotiam (a second-generation cephalosporin, alternative to cefuroxime in Japan) 82%, cefmetazole (a cephamycin, alternative to cefoxitin in Japan) 99%, cefotaxime 84%, tobramycin 94%, and cipro oxacin 79%.
In urine PCGS, medium-or large-sized Gram-negative rods suggested E. coli, Proteus mirabilis, or Klebsiella pneumoniae. In these cases, cefotiam 1 g every 6 to 8 hours (q6-8h) intravenously (iv) was mostly administered. A third-generation cephalosporin (cefotaxime 1 g q6-8h iv or ceftriaxone 1-2 g q12-24h iv) was selected when Citrobacter, Enterobacter, or other Enterobacteriaceae was suspected based on previous culture ndings. If an ESBL-producing bacterial strain was suspected based on a living place or previous culture results, cefmetazole 1 g q6-8h iv was selected. If a patient's condition was unstable due to septic shock, a carbapenem (meropenem 1 g q8h iv or imipenem/cilastatin 0.5 g q6h iv) was used. Small-sized Gramnegative rods suggested P. aeruginosa, and ceftazidime 1 g q6-8h iv, aztreonam 1 g q6-8h iv, tobramycin 120-240 mg q24h iv, or a carbapenem (meropenem or imipenem/cilastatin) was chosen. Gram-positive cocci in chains suggested Enterococcus or Streptococcus, and mainly ampicillin 1 g q6-8h iv or rarely vancomycin 1 g q12h was selected. Gram-positive cocci in clusters suggested Staphylococcus or Aerococcus, and cefazolin 1 g q6-8h iv or rarely vancomycin was chosen. If Gram-negative rods and Gram-positive cocci in chains were observed, ampicillin/sulbactam 1.5 g q6-8h iv was chosen to cover both Enterobacteriaceae and Enterococcus. When more than two types of bacteria were con rmed, a polymicrobial infection including AMR was suspected, and a cephalosporin, a cephamycin, or a carbapenem was selected. These polymicrobial infection cases were excluded from the analysis (Table 3). If Gram-positive rods, small amounts of Gram-positive cocci, or mixed organisms were observed, they were considered to be non-pathogenic.
Penicillins and rst-or second-generation cephalosporins were de ned as narrow-spectrum antibiotics; fourth-generation cephalosporin, carbapenems, and vancomycin as broad-spectrum antibiotics; and all other antibiotics as intermediate-spectrum antibiotics [7].

Outcome measures
The primary outcome was the effectiveness of targeted therapies such as initial antimicrobial selection based on PCGS according to urine and/or blood culture susceptibility test results in the uncomplicated and complicated groups. Secondary outcomes were pyuria and bacteriuria detection by PCGS compared to U/A, and bacterial type estimation by PCGS compared to U/C.

Statistical analysis
To ensure an adequate sample size to calculate the effectiveness of the initial antimicrobial choice, we hypothesized that there would be a difference between the uncomplicated and complicated groups. We assumed that the rate of effectiveness in the uncomplicated group would be more than 0.9 because the patients were young and the rate of AMR was still low. In the complicated group, however, we expected that the effectiveness could be lower than 0.9 because they were older, and repeated antimicrobial exposure increases AMR colonization. Therefore, assuming that the rate of effectiveness would be 0.95 in the uncomplicated group and 0.8 in the complicated group, and assuming that the uncomplicated group-to-complicated group ratio was 1:4, 80% power, and a two-sided alpha level of 0.05, we calculated that 260 patients would be required.

Results
There were 48 and 218 patients enrolled into the uncomplicated and complicated groups, respectively. Table 1 shows the comparison between the two groups. Patient age ranged from 16 to 92 years in the uncomplicated group and from 19 to 105 years in the complicated group. The complicated group included 34 patients with prostatitis, ten with stone pyelonephritis, three with uterine prolapse, two with ileal conduit, two with renal abscess, one with emphysematous cystitis, and one with vesicoureteral re ux. The performance status was classi ed in accordance with the Eastern Cooperative Oncology Group scale [15]. Two patients died. One had been admitted to another hospital and treated with multiple antibiotics. Soon after discharge, he developed fever with shock, and was transferred to our hospital. His Foley catheter was obstructed with debris, and he was diagnosed as having obstructive pyelonephritis with hydronephrosis. Ceftriaxone, meropenem, and vancomycin with catecholamine were administered, but he died 2 days later. Both carbapenem-resistant P. aeruginosa and S. dysgalactiae subsp. dysgalactiae were cultured from the patient's blood and urine. The other patient who died was a steroid user with E. coli urosepsis. Her fUTI had resolved, but she developed cytomegalovirus gastritis and it eventually caused her death after 3 weeks. Table 2 compares pyuria and bacteriuria detection by PCGS with U/A. In the complicated group, U/A was not submitted for one patient who was not included in the analysis (N=217). As the number of white blood cells in the U/A results increased, the number of white blood cells in PCGS tended to increase in both the uncomplicated and complicated groups (each p<0.0001). Similarly, as the number of bacteria in the U/A results increased, the number of bacteria in PCGS tended to increase in both groups (each p<0.0001).   Table 2). The two patients who had negative bacteria results for both U/A and PCGS were the above-mentioned patient with water intoxication and another who had taken sulfamethoxazole/trimethoprim 3 weeks before because of cystitis, and 10 3 CFU/mL of E. coli was cultured from this patient's urine. Table 3 shows the agreement between bacterial estimation using PCGS and U/C. In the uncomplicated group, the U/C results for one patient were considered to be a skin or vaginal contaminant, and this patient was excluded from the analysis. The agreement was 90.2%, and the kappa coe cient was 0.518 (95% con dence Interval [CI] 0.318-0.547). The sensitivity to detect Enterobacteriaceae was 93.6%, and the speci city was 100% using PCGS.  In the complicated group, the U/C results of ve patients were considered to be contaminated, and these results were excluded from Table 3. The agreement was 85.8%, and the kappa coe cient was 0.608 (95% CI 0.571-0.653). The sensitivity of PCGS to detect Enterobacteriaceae was 95.4% and the speci city was 87.2%.
As mentioned above, 19 patients with PCGS results that showed more than two types of bacteria were excluded from the analysis in Table 3. Gram-negative rods such as Enterobacteriaceae, Pseudomonas, and Myroides, Gram-positive cocci such as Enterococcus and Streptococcus, and Gram-positive rods such as Corynebacterium or others were cultured from these patients. Table 4 shows the comparison between PCGS bacterial semi-quanti cation and U/C bacterial quanti cation. Contamination was suspected in one patient in the uncomplicated group and ve patients in the complicated group, and they were excluded from this table. As the number of cultured bacteria increased, the PCGS semi-quanti ed bacteria increased in both groups (each p<0.0001). There was one patient with a negative urine culture result in the complicated group. This patient was diagnosed with acute prostatitis, and K. pneumoniae was identi ed in the blood culture. After treatment, his PSA level decreased from 33 to 7 ng/mL.     Table 6 shows the nally chosen iv antibiotics after the U/C results were reported. Combination antimicrobials were selected for three patients. Narrow-spectrum antibiotics were chosen in 91.7% (44/48) of patients in the uncomplicated group and 67.8% (150/221) in the complicated group (p=0.001). Ampicillin, the narrowest-spectrum antimicrobial in this study, was nally selected in 42.4% (114/269). Broad-spectrum antibiotics were nally used in 7.8% (17/218) of patients in the complicated group only. Among 18 cefmetazole cases, ESBL-producing E. coli were cultured from 17 patients. One patient also had positive blood culture results.

Discussion
Unlike previous studies on urine Gram staining in a laboratory [3,6,[16][17][18][19], this observational study investigated the utility of urine PCGS in estimating etiologic agents and choosing narrower-spectrum antimicrobials to treat fUTI in adults. Our results yielded three main ndings. First, PCGS and U/A were good indicators of both pyuria and bacteriuria. Second, estimated bacterial types using PCGS were moderately correlated with U/C results. Third, initial antibiotic effectiveness was quite high in the uncomplicated group although they were almost all narrow-spectrum antimicrobials, and the effectiveness was also high in the complicated group although there was a high level of AMR and low usage of broad-spectrum antimicrobials.
Pyuria and bacteriuria detection by PCGS and U/A showed similar results in both groups. U/A and PCGS both showed no falsenegative results for pyuria and bacteriuria in the uncomplicated group. In the complicated group, there was only one patient with water intoxication, and she did not have pyuria or bacteriuria (0.46%). U/A samples were centrifuged before analysis, and thus, the sensitivity for bacterial detection may be higher than using PCGS. However, U/A was unable to distinguish between Gram-positive and -negative bacteria, and it was impossible to distinguish between pathogenic Enterobacteriaceae and nonpathogenic normal ora. PCGS does not require centrifugation, and it easily identi ed Gram-positive and -negative bacteria, with a small risk of technical error in the process of a four-step dyeing procedure. Additionally, especially for fUTI in the complicated group, a clinical diagnosis is often not straightforward because the patients' symptoms and physical examination are not reliable [3]. Therefore, using both U/A and PCGS together to complement each other is strongly recommended for a precise fUTI diagnosis.
Estimating the bacterial types using PCGS was moderately correlated with U/C results in the uncomplicated group. Most Enterobacteriaceae were detected with PCGS except for three false negative results. The bacteria quantity in these false negative samples were relatively low: 10 4 , 10 5 , and 10 6 CFU/mL. However, all Enterobacteriaceae levels that were greater than 10 6 CFU/mL were detected by PCGS. PCGS did not detect the two Streptococcus results, but both of the patients grew Enterobacteriaceae together with Streptococci, and we considered the latter to be contaminants. Additionally, PCGS precisely identi ed non-pathogenic bacterial infections. Therefore, PCGS detected most pathogens unless the bacterial quantity was too low, and this led to selecting narrow-spectrum antibiotics.
In the complicated group, various types of bacteria were detected with PCGS, and identi ed with U/C. AMRs were more frequently found, and so it was not easy to choose the initial antimicrobial treatment. PCGS did not identify four cases of Enterobacteriaceae that grew in the cultures. The quantity of these bacteria was 10 3 , 10 4 , and two results of 10 7 CFU/mL; the latter two culture results suggested that technical errors had occurred during the staining process. In these two cases, PCGSs were repeated using the centrifuged residues, and leukocytes and Gram-negative rods were found in both cases. The sensitivity of estimating Pseudomonas based on PCGS was only 50% (5/10), which would be unreliable for differentiating Pseudomonas from Enterobacteriaceae. However, the sensitivity for detecting Gram-negative rods including Enterobacteriaceae and Pseudomonas was 97.8% (181/185). When only Gram-negative rods were con rmed using PCGS, an antibiotic that excluded coverage of Gram-positive bacteria was chosen. For Streptococcus and Staphylococcus, the sensitivity was 87.5% (7/8) and 66.7% (2/3), respectively. Non-pathogenic bacteria were moderately detected using PCGS (54.5%; 6/11). Compared to other reports about Gram-positive cocci in clusters [20,21], S. saprophyticus was not detected, but instead, S. aureus and A. urinae were cultured in this study.
Our study suggested that at least 10 4 CFU/mL or more is required to make a diagnosis in the uncomplicated group, and 10 3 CFU/mL was required for the complicated group. These results were similar to the recommendations of the Infectious Diseases Society of America (IDSA) guidelines in 1992, which recommended that 10 4 CFU/mL or more is required to meet the diagnostic criteria of uncomplicated pyelonephritis, and 10 3 or more is required for a complicated UTI in men [11]. In 2018, another report recommended that 10 4 or more CFU/mL of a uropathogen is an acceptable diagnostic criterion for pyelonephritis in adults [1]. However, as mentioned above, we had one case of acute prostatitis that had a negative U/C result. His initial U/A detected only ve to nine leukocytes per HPF, and the day after starting treatment, the leukocytes increased to 30 to 49 per HPF. Therefore, the initial U/C result may not be su cient because of a small amount of pathogen even for a rare case of fUTI in men.
Initial targeted therapy effectiveness based on PCGS was very high (97.9%) in the uncomplicated group. In the IDSA guidelines, intravenous uoroquinolone is recommended for therapy in patients with pyelonephritis who require hospitalization [22]. However, if the prevalence of uoroquinolone resistance exceeds 10%, intermediate spectrum antimicrobials such as ceftriaxone or aminoglycoside are recommended for initial treatment [22]. In Japan, unfortunately, the prevalence of uoroquinolone-resistant E. coli was already 38.0% in 2015 [23]. Therefore, second-generation cephalosporins such as cefotiam, not uoroquinolone, were mostly selected at OCH, and they were su cient to treat patients in the uncomplicated group. Ceftriaxone, a third-generation cephalosporin, is more convenient than cefotiam because it is usually administered once daily. However, "it is better to use narrower-spectrum antibiotics" is one of our principles for antimicrobial selection that prefers cefotiam to ceftriaxone. The only pathogen for which cefotiam was not effective was ESBL-producing E. coli, and even ceftriaxone did not cover it. If the number of ESBL cases continues to increase in the future, cefmetazole or cefoxitin, which are potentially effective against ESBL, will be required even for uncomplicated pyelonephritis.
In the complicated group, the initial antibiotic effectiveness was lower than in the uncomplicated group because of the greater prevalence of AMR. Narrow-spectrum antimicrobials were selected for 76.9% of patients, and its effectiveness was 83.5%, which was mainly in uenced due to ESBLs. PCGS reduced broad-spectrum antimicrobial use to 7.7%, but one patient died due to P. aeruginosa that was resistant to ceftazidime, aztreonam, imipenem/cilastatin, tobramycin, and cipro oxacin. In Japan, national surveillance reported that the prevalence of carbapenem-resistant Enterobacteriaceae was only 0.1-0.2%, but that of carbapenem-resistant P. aeruginosa was 13.1-18.8% in 2015 [23]. If the prevalence of AMR does not decrease, empirical combination therapy that covers various resistant patterns will be required more often in accordance with the local AMR prevalence. PCGS can potentially identify the bacterial type even in the complicated group, and it can reduce broad-spectrum antibiotic overuse, which will prevent an increase in AMR.
Ampicillin was the most frequently used nal antibiotic in both groups. The most common fUTI pathogen was still ampicillinsensitive E. coli. The overall proportion of AMR was 12.4% (33/266) in this study, which maintained low use of empirical broadspectrum antibiotics (6.4%; 17/266). This research has some limitations. First, this was a retrospective chart review study, and some information was missing. Second, our strategy is not directly applicable to other hospitals where clinicians are not familiar with Gram staining. PCGS is not a special technique, but it still requires a certain measure of training and experience. Third, our targeted therapies based on PCGS cannot be directly applied to other regions because the prevalence of AMR may be different between regions. Fourth, PCGS results were not compared with Gram staining results from the microbiology laboratory because the data were not recorded, and thus, this difference was not investigated. Fifth, PCGS quality control was not evaluated in this study, although the utility of PCGS has been con rmed in several other studies from OCH [7,8,14,24].

Conclusions
Using PCGS with U/A led to a more precise fUTI diagnosis and encouraged clinicians to select narrower-spectrum antibiotics that showed high effectiveness. PCGS could potentially reduce the increasing AMR, and it is advisable to include PCGS in all fUTI antimicrobial stewardship programs in adults.