Study design and patient population

This retrospective cohort study was conducted at Oregon Health and Science University, a 556-bed academic, quaternary-care healthcare center in Portland, Oregon, that also serves as a regional referral center. Patients age 18 years and older admitted to the Department of Family Medicine service between May 2009 and December 2011 with a primary care provider in the Department of Family Medicine were eligible for inclusion. The Department of Family Medicine includes an active inpatient service that predominantly manages patients receiving primary care at 1 of 4 family medicine outpatient clinics in the greater Portland area. The department aims to schedule follow-up visits with their inpatients within 2 weeks of discharge.

We excluded patients with a history of UTI in the 30 days before admission, identified as (1) an acute UTI diagnosis (based on International Classification of Diseases, Ninth Revision (ICD-9) codes 590.1x, 590.8x, 590.9, 595.0, 595.4, 595.89, 595.9, 597.80, 597.81, 599.0, or 996.64; (2) positive urine microbiology culture (ie, urine cultures with growth of ≥ 10,000 CFU/mL, < 3 pathogenic bacteria isolated, and taxonomy identified to the genus level; and (3) nitrofurantoin prescription in the 30 days before admission. Additionally, patients with a chronic UTI diagnosis (ICD-9 code 590.0, 590.3, 595.1, 595.2, 595.81, or 595.82) in the year preceding admission or during hospitalization were excluded. Only the first eligible admission was included for patients with repeat hospitalizations during the study period. The study was approved by the Oregon Health and Science University Institutional Review Board.

Data collection

Patient data were abstracted from the Pharmacy Research Repository (PHARR), a repository developed and maintained in collaboration with the Oregon Clinical and Translational Research Institute. PHARR includes electronic health record (EHR) data and associated databases. Abstracted data include patient characteristics and demographics as well as information related to encounters, diagnoses, and symptoms based on ICD-9 codes; laboratory tests including microbiology cultures; medication orders; and surgeries for admissions and outpatient visits to OHSU. Data were collected during patients’ hospitalization, the preceding year (medical history), and during the 30 days after discharge.

Variable definitions

We identified hospital-associated UTI using ICD codes, urinary antibiotic prescriptions (ie, nitrofurantoin, ertapenem, imipenem, amoxicillin-clavulanate, ampicillin-sulbactam, moxifloxacin, ciprofloxacin, levofloxacin, trimethoprim-sulfamethoxazole, or trimethoprim alone), and positive urine microbiology culture results. Patients with HA-HO UTI were excluded from primary analyses, but their data were quantified for descriptive purposes. Patients meeting one of the following criteria during their hospitalization and at least 48 hours after admission were defined to have a HA-HO UTI: (1) nitrofurantoin prescription, (2) acute UTI diagnosis and other non-nitrofurantoin urinary antibiotic prescription, (3) positive urine culture and other non-nitrofurantoin urinary antibiotic prescription, or (4) acute UTI diagnosis and positive urine culture.

Our primary outcome was HA-CO UTI in the 30 days following hospital discharge. Individuals without a UTI during their index hospitalization were considered to be at risk. HA-CO UTI outcome definitions were dependent on treatment setting (ie, infections treated at inpatient versus outpatient encounters) and are summarized in Table 1. For readmitted patients, HA-CO UTI was defined as patients with hospital-associated UTI criteria in the first 48 hours of that subsequent inpatient admission. Readmitted patients were censored from follow-up after the first 48 hours of that subsequent inpatient admission. For patients with outpatient visits during the follow-up period, HA-CO UTI was defined as patients’ first incident UTI within 30 days meeting any of the following criteria: (1) a nitrofurantoin prescription, (2) acute UTI diagnosis, (3) positive urine culture or positive urinalysis with any urinary antibiotic prescription, or (4) positive urine culture, positive urinalysis, and dysuria (ICD-9 code 781.1). Chart review was performed to validate our definition of HA-CO UTI. Our case definition performed with 100.0% (95% confidence interval [CI]: 96.0%–100.0%) sensitivity and 88.1% (95% CI%–92.8%) specificity compared to chart review for identifying true symptomatic UTI. Each patient was included only once in our analyses. Illustrative examples of outcome classification are provided in Fig. 1.

Fig. 1.

Illustrative examples of identification of cases of healthcare-associated, community-onset urinary tract infections. Abbreviations: UTI, urinary tract infection; HA-CO, healthcare-associated, community-onset UTI; HA-HO, healthcare-associated, hospital-onset UTI.

Table 1.

Criteria for Identifying Cases of Healthcare-Associated, Community-Onset Urinary Tract Infection by Healthcare Setting

^a Within 3 days of admission.

^b Between −1 and 3 days of visit.

Potential predictors of HA-CO UTI were identified from patient characteristics, social history, and medical history from the qualifying hospitalization and medical history from the preceding year including prior antibiotic prescriptions and comorbidities. A weighted summary score for the Elixhauser comorbidity measure was calculated for each patient’s index hospitalization.^{Reference Elixhauser, Steiner, Harris and Coffey5–Reference Thompson, Fan and Dalton7} To explore the urban versus rural place of residence as a potential predictor, patient zip code was combined with zip code–based rural–urban commuting area codes, then aggregated into urban and rural categories as recommended by the Washington–Wyoming–Alaska–Montana–Idaho Rural Health Research Center.^{Reference Hart and Cromartie8,9}

Statistical analyses

Patient characteristics were summarized with descriptive statistics. The 30-day cumulative incidence was calculated and expressed per 1,000 patients for HA-HO and for HA-CO UTI. We performed a Kaplan-Meier analysis based upon time to HA-CO UTI diagnosis following hospital discharge. Multivariable logistic regression was performed to identify potential risk factors for HA-CO UTI using patient data from the year prior to admission and during hospitalization. Variables significantly associated with the outcome (P < .05) and those with a confounding effect were retained in the model. Variables were considered confounders if their inclusion in the final model resulted in a ≥ 20% change in the odds ratio between covariates and outcome. We also tested for the presence of interaction between sex and indwelling catheterization, which had been hypothesized a priori. Adjusted odds ratios (aORs) and 95% CIs were calculated from the final model. Data management and statistical analyses were conducted in SAS statistical software version 9.2 software (SAS Institute, Cary, NC) and R version 3.1.2 software (R Foundation for Statistical Computing, Vienna, Austria).¹⁰