We searched for studies estimating the economic costs attributable to antibiotic drug-resistant CA-UTIs published from January 1, 2008, to January 31, 2023, using a combination of broad-based (and wildcard) search criteria, including terms for UTI, community-acquired, ABR, and health economic cost. We searched the PubMed, Ovid MEDLINE and Embase, Cochrane Review Library, and Scopus databases using strings developed for each database (Table S2 in Multimedia Appendix 1). The bibliographies of the identified studies were also reviewed.

The study inclusion/exclusion criteria are presented in Table 1, including the Patient/Population, Intervention, Comparison, and Outcomes (PICO) eligibility. Two authors (NJZ and MW) independently screened the titles and abstracts of the records yielded from the database search and independently screened the full-text articles. The discrepancies during title/abstract screening and full-text screening were resolved by consulting the third author (EC). Any article comparing monetary or health economic costs of antibiotic-resistant versus susceptible CA-UTIs through clinical trials, observational designs (eg, cohort study, case-control study), or modeling approaches was included for full-text review.

Data were extracted from the included studies, including study identifier, authors, journal, publication year, study design, data collection period, country/region, health care setting, perspective (patient, health system [representing payer or provider], or societal), patient population, number of patients, UTI diagnosis criteria, pathogen, sensitivity profile, treatment, and outcome. We synthesized the impact of ABR on health outcomes (eg, mortality), health care system (eg, hospital LOS, medication cost), and economic system (eg, productivity), and compared these for infections caused by resistant versus susceptible pathogens. The methods to estimate the cost of illness were categorized using a top-down approach for those studies that reported total costs on a population level irrespective of the specific method used to derive these costs or a bottom-up approach for those studies that reported average costs derived from accumulating measured costs from patient samples.

A meta-analysis was performed to synthesize the reported mortality and hospital LOS using a random effect model [18]. A random effects model assumes that the true effect size of the exposure varies from study to study due to study heterogeneity. Particularly, heterogeneities in this type of analysis occurred in definitions and categories of costs across health systems, settings, and disease types; cost measurement instruments; and unit prices. Thus, a random effects model was chosen to allow aggregating cost data from different studies by circumventing this heterogeneity. In the meta-analysis of mortality, we estimated pooled odds ratios based on the crude mortality rate [19]. In the meta-analysis of LOS, we applied both the transformation-based methods (ie, estimating the sample mean and SD from the median and sample size) [20,21] and median-based methods (ie, considering study-specific median differences and data distribution) [22], considering mean and variance and median and IQR were commonly used when reporting LOS, and the distribution of LOS was heavily right-tailed (eg, not normally distributed) [23,24]. We assessed the publication bias for the mortality outcome using a funnel plot and Egger test [25,26]. No meta-analysis was performed for economic costs due to the large variation in the resource costs and the methods used to determine the cost. To compare the reported monetary costs, the outcomes were converted into 2023 US dollars by inflating the cost to 2023 original currency estimates using annual inflation rates [27], then converting this into US dollars utilizing the 2023 average exchange rates [28].

The included studies were assessed using the Integrated Quality Criteria for Review of Multiple Study Designs (ICROMS) tool [29].

A total of 380 titles and abstracts were yielded from the database search; 214 duplicates were removed, and 132 abstracts were deemed irrelevant. A full-text review was performed on 34 studies, of which, 11 studies were included. Through reference search, another 4 studies were identified and included in the final study pool. Figure 1 summarizes the screening process in a PRISMA flowchart.

The characteristics of the 15 identified studies are presented in Table 2 [12-14,30-41]. The countries that individually produced the highest number of studies were the United States (n=5, 33%) [14,33,38,39,41], followed by Spain (n=3, 20%) [13,32,40] and South Korea (n=2, 13%) [30,36]. A total of 13 (87%) studies focused on patients who were hospitalized [12-14,30,32-36,38-41], and 2 (13%) studies focused on primary care patients [31,37]. Additionally, 13 (87%) studies included adult patients of all genders [12-14,32-41], of which, 1 study included patients 65 years and older [32]. Chang et al [30] and Little et al [31] (n=2, 13%) investigated adult female patients. All hospital-based studies had UTI diagnosed via the presence of symptoms, infection biomarkers, and microbiology culture confirmation, and differentiated community-acquired cases using the 48-hour cutoff time after admission. Two (13%) studies reported hospital-acquired UTI [14,39]. The community-based study recruited patients with urinary tract symptoms (suspected UTI) or a history of dysuria and frequency [31,37]. In total, 57,251 CA-UTI cases were reported, and 47,131 UTI cases were analyzed (Table S3 in Multimedia Appendix 1).

Of the 15 studies, 14 (93%) reported the pathogens identified, of which, all reported GNB [12-14,30-41], 4 (29%) reported Gram-positive bacteria [32,34,36,39], 1 (7%) reported fungi [32], 3 (21%) exclusively reported UTI caused by E coli [13,30,37], 1 (7%) reported UTI caused by K pneumoniae [40]. E coli, K pneumoniae, and P aeruginosa are the most frequently identified organisms. Among the studies in specific antibiotic-pathogen combinations, 2 studies assessed carbapenem-resistant organisms, specifically GNB and Enterobacterales [14,41]. Mark et al [33] examined E coli, K pneumoniae, and Proteus mirabilis (P mirabilis) resistance to third-generation cephalosporins. Sozen et al [12] and MacVane et al [38] examined extended-spectrum β-lactamases– or inducible β-lactamases–producing GNB.

All the included studies estimated the clinical and economic outcomes of patients recruited from single or multiple health facilities. François et al [37] provided a national-level estimate of the infection incidence and costs derived from the study cohort. No study performed sensitivity analysis. The results of the quality assessment are presented in Table S5 in Multimedia Appendix 1. All studies met the minimum required score. Of the 15 studies, 6 (40%) failed to meet the minimum required criteria [13,30,37,38,40,41].

When quantifying the burden attributable to ABR, the included studies compared patient outcomes, health system outcomes, and economic costs of the CA-UTI cases caused by resistant pathogens against those caused by nonresistant pathogens. The most reported outcomes were mortality, hospital LOS, and economic costs due to antibiotic treatment (Table S4 in Multimedia Appendix 1). A health system perspective was taken by all except 1 study when estimating the costs [12-14,30-41]. François et al [37] took a societal perspective and included productivity loss due to absenteeism. When comparing the patients with resistant and nonresistant CA-UTIs, 4 studies matched case and control [13,14,38,41], 2 studies adjusted patient characteristics and other risk factors when reporting outcomes [33,35], other studies performed no matching or adjusting.

A total of 12 studies reported mortality, including in-hospital all-cause mortality [14,30,32,34,38,41], in-hospital infection-related mortality [38], 30-day all-cause mortality [13,32,40], and 90-day all-cause mortality [33] (Table S4 A in Multimedia Appendix 1). A total of 7 studies reported higher crude mortality among the patients with antibiotic-resistant UTIs [13,14,32-34,38,41], of which, 1 study demonstrated the statistical significance [41]. The pooled odds ratios of mortality outcomes for resistant UTIs are presented in Figure 2. Results presented odds ratios of resistant compared to nonresistant infections. The blue squares centered at the point estimate the effect size, with horizontal lines depicting the 95% CIs, and the sizes of the blue squares correspond to the patient group sizes. The overall effect sizes are represented by diamonds centered on their estimated values with the diamond width corresponding to the CI length. The random effects model estimated an overall odds ratio of 1.50 (95% CI 1.29-1.74), suggesting that ABR increased the overall mortality. The subgroup analysis conducted for different mortality outcomes suggested increased odds of in-hospital all-cause mortality (Figure 2). No publication bias was detected for mortality (Figure S1 in Multimedia Appendix 1).

All 13 hospital-based studies reported LOS [12-14,30,32-36,38-41], among which, 11 reported significantly higher LOSs associated with antibiotic-resistant UTIs (Table S4 B in Multimedia Appendix 1) [12-14,32-36,38,40,41]. Cardwell et al [39] reported higher LOS among patients with clinical failure due to inappropriate antibiotic therapies for resistant infections. The meta-analysis of studies reported LOS in mean and SD estimates of a pooled excess LOS of 2.45 days (95% CI 0.51‐4.39; Figure 3A). The meta-analysis of studies reported LOS in median and IQR estimates of a pooled excess LOS, ranging from the lowest value of 1.50 days (95% CI 0.71-4.00), estimated by the median of the differences of medians method, to the highest value of 2.00 days (95% CI 0.85-3.15), estimated by the linear quantile mixed models method (Figure 3A).

A total of 8 studies reported costs in monetary terms (Table S4 C in Multimedia Appendix 1) [12-14,35,37,38,40,41], including 5 that reported costs in US dollars [12,14,35,38,41] and 3 that reported costs in euros [13,37,40] (Figure 4). None of the included studies discounted the costs. Considering only 2 studies explicitly stated the year of which the costs were adjusted to [12,37], the end year of the data collection period was used to convert the reported costs into 2023 US dollars. A total of 8 studies reported direct medical costs incurred in secondary care, including emergency department costs [13,33] and outpatient parenteral antibiotic therapy costs in 1 study [13,33]. All 8 studies reported higher medical costs spent treating patients with resistant UTIs in hospitals. The highest excess cost was observed in UTIs caused by carbapenem-resistant Enterobacterales [41]. François et al [37] reported costs incurred in primary care, specifically, the costs of GP visits due to UTI symptoms. The primary care costs of single- or multidrug-resistant E coli UTIs were not significantly higher than those caused by susceptible E coli.