The COMBACTE-MAGNET WP5 RESCUING study is an international, multi-centre, retrospective, observational cohort study conducted between January 2015 and August 2016. It included both patients diagnosed with cUTI as the primary reason for hospitalisation and those who developed cUTI after being hospitalised for another reason between 1st January 2013 and 31st December 2014 [12, 13]. The STROBE guidelines for reporting observational studies were followed . In the present study, we compared patients who had CA-UTI with those who had cUTI of other sources.
Setting and patients
The study was conducted at 20 hospitals in Bulgaria, Greece, Hungary, Israel, Italy, Romania, Spain and Turkey. Patient selection was done by searching the appropriate Clinical Modification codes at discharge based on versions 9 or 10 of the International Classification of Diseases. The sample size was calculated to detect an absolute difference of 10% in the treatment failure rate between infection due to MDR and other pathogens (power = 0.83, α = 0.05). We planned to include 50 to 60 patients per hospital to reach a total estimated sample size of 1000 [15, 16].
The inclusion criteria were based on the Food and Drug Administration guidance for cUTI , requiring patients to have a UTI plus at least one of the following:
1. At least one from among the following: an indwelling urinary catheter; urinary retention (at least 100 mL of residual urine after voiding); neurogenic bladder; obstructive uropathy (e.g., nephrolithiasis or fibrosis); renal impairment caused by intrinsic renal disease (estimated glomerular filtration rate < 60 mL/min); and renal transplantation; urinary tract modification (ileal loop or pouch).
2. At least one from among the following signs or symptoms: Chills or rigors associated with fever or hypothermia (temperature > 38 °C or < 36 °C); flank or pelvic pain, dysuria, urinary frequency or urgency; and costovertebral angle tenderness on physical examination.
3. A urine culture with at least 105 colony-forming units (CFU)/mL of a uropathogen (no more than two species); or at least one blood culture growing possible uropathogens (no more than two species) with no other evident site of infection.
We excluded patients who were younger than 18 years, diagnosed with prostatitis (based on Food and Drug Administration guidance), diagnosed with pyelonephritis with normal urinary tract, had polymicrobial infections that included Candida spp. or more than two bacterial species, or who had cUTI with Candida spp. as the sole uropathogen.
Data from eligible patients were collected from January 2015 to August 2016. For all patients, a standardised data set was collected retrospectively from electronic hospital records and input in a web-based electronic case report form (eCRF) with controlled access. The data set included details of demographic characteristics, comorbidities, infection acquisition site, signs and symptoms, laboratory and microbiology test results, and discharge and outcome details, including death. The follow-up period was limited to two months after hospital discharge. To ensure data quality, study sites were monitored and audited.
Acquisition of cUTI in a medical care facility was considered hospital-acquired if it started 48 h or more after hospital admission. Acquisition of cUTI was considered healthcare-associated if it was detected at hospital admission or within the first 48 h of hospitalisation, and met any of the following criteria: the patient had received intravenous therapy, wound care or specialist nursing care at home in the previous 30 days; hospital or haemodialysis ward attendance or intravenous chemotherapy administration in the previous 30 days; hospitalisation for at least 2 of the previous 90 days; residence in a long-term care facility; underwent an invasive urinary procedure in the previous 30 days; or had a long-term indwelling urethral catheter.
We then defined cUTI as either CA-UTI or cUTI of other source (other-cUTI). The CA-UTI group comprised those with UTI related to indwelling urinary catheterisation, including long-term, short-term or intermittent catheterisation. The other-cUTI group included those with all other causes of cUTI, including the following: UTI related to anatomical urinary tract modification (including any urinary diversion procedure, nephrostomy, stent or renal transplant); UTI related to obstructive uropathy (including any obstruction intrinsic or extrinsic to the urinary tract, such as lithiasis, tumour, ureteral herniation or prostate hyperplasia); UTI related to events that do not fall under any other category (such as neurogenic bladder).
MDR was defined in accordance with the international expert proposals of Magiorakos et al., as non-susceptibility to at least one agent in three or more antimicrobial categories (extended-spectrum penicillins, carbapenems, cephalosporins, aminoglycosides and fluoroquinolones) . Extensive drug-resistance (XDR) was defined as non-susceptibility to at least one agent in all but two or fewer antimicrobial categories.
Steroid therapy was defined as the administration of a dose of at least 10 mg of prednisolone or an equivalent dose of another steroid for 30 days or more prior to the diagnosis of cUTI.
Length of hospital stay was considered in all cases from the day of diagnosis of cUTI to the day of discharge or mortality.
The primary endpoint was 30-day mortality. The secondary endpoints were length of hospital stay, symptom improvement at 7 days of treatment, symptom recurrence at 30 days from diagnosis, and readmission rate at 60 days from discharge. We also aimed to identify the factors influencing 30-day mortality among patients with CA-UTI.
Demographic, clinical and outcome data for patients in the CA-UTI and other-cUTI groups were described using appropriate statistics according to the nature and distribution of the variable. The statistical analyses were performed using version 3.5.0 of R for Windows. Statistical significance was set at a probability level of < 0.05.
The crude and adjusted association between the presence of CA-UTI and the 30-day mortality was analysed by logistic modelling with mixed-effects that took into account the variability between centres. The patient demographics and variables associated with 30-day mortality by unadjusted analysis (Charlson index, having haematological malignancy, basal functional status, place of cUTI acquisition and reason for admission) were used for adjustment. Although these adjustment variables were initially chosen on clinical grounds, they were required to modify the coefficient of the main variable (CA-UTI) by more than 10 to remain in the model. The impact of the assessment centre was evaluated by the intra-class correlation (ICC), which measures how much of the overall variation in the outcome is explained simply by clustering. ICC ranges from 0 to 1; a value close to 1 indicates that patients within centers are more similar than patients between centers, and a value close to 0 indicates that patients between centers are similar.
Residues were validated graphically, and the conditions of application of the models were tested. The odds ratios (ORs) and 95% confidence intervals (CIs) were calculated when appropriate.