https://orcid.org/0000-0002-7317-5157
Abstract
To calculate a risk-adjusted mortality ratio (RAMR) for bloodstream infections (BSIs) using all-patient refined diagnosis-related groups (APR-DRGs) and compare it with the crude mortality rate (CMR).
Retrospective observational study of prevalent BSI at our institution from January 2019 to December 2022. In-hospital mortality was adjusted with a binary logistic regression model adjusting for sex, age, admission type and mortality risk for the hospitalization episode according to the four severity levels of APR DRGs. The RAMR was calculated as the ratio of observed to expected in-hospital mortality, and the CMR was calculated as the proportion of deaths among all bacteraemia episodes.
Of 2939 BSIs, 2541 were included: Escherichia coli (n = 1310), Klebsiella pneumoniae (n = 428), Pseudomonas aeruginosa (n = 209), Staphylococcus aureus (n = 498) and candidaemia (n = 96). A total of 436 (17.2%) patients died during hospitalization and 279 died within the first 14 days after the onset of BSI. Throughout the period, all BSI cases had a mortality rate above the expected adjusted mortality (RAMR value greater than 1), except for Escherichia coli (1.03; 95% CI 0.86–1.21). The highest overall RAMR values were observed for P. aeruginosa, Candida and S. aureus with 2.06 (95% CI 1.57–2.62), 1.99 (95% CI 1.3–2.81) and 1.8 (95% CI 1.47–2.16), respectively. The temporal evolution of CMR may differ from RAMR, especially in E. coli, where it was reversed.
RAMR showed higher than expected mortality for all BSIs studied except E. coli and provides complementary to and more clinically comprehensive information than CMR, the currently recommended antibiotic stewardship programme mortality indicator.
Introduction
Antibiotic stewardship programmes (ASPs) are initiatives designed to combat antimicrobial resistance and improve clinical outcomes for patients. Among the metrics used to assess the effectiveness of these programmes, one of the most widely used is the crude mortality rate (CMR) at 14 or 30 days for primary causes of bloodstream infections (BSIs) as an infectious syndrome reference.1,2 However, CMR does not allow for contextual interpretation of outcomes within the healthcare process and has significant limitations in intercentre comparisons and, longitudinally, within the same centre.
The widespread adoption of electronic health records (EHRs) has facilitated the acquisition of administrative databases, such as the Minimum Basic Data Set, mandated for National Health System hospitals in Spain, and the widespread utilization of patient classification systems, such as All-Patient Refined Diagnosis-Related Groups v35 (APR-DRGs), for clinical management and health outcomes research.3,4
The present study aimed to use the standardized clinical information available in our hospital to calculate the BSI risk-adjusted mortality ratio (RAMR) and compare it with the CMR, a currently recommended clinical outcome indicator for ASPs.
Methods
A retrospective observational study was conducted on all patients with the most prevalent bacteraemia (Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa or Staphylococcus aureus) and candidaemias registered between 1 January 2019 and 31 December 2022 in a public tertiary healthcare centre of high complexity. This centre has 1200 beds and provides specialized healthcare to an area with more than 450 000 inhabitants. The study excluded outpatient episodes, non-admitted emergencies, and inpatients assigned to a non-specific APR-DRG.
EHRs served as the data source, and APR-DRGs were derived from the Minimum Basic Data Set (MBDS), which includes clinical data, primarily diagnoses and procedures, coded using the International Classification of Diseases 10th Revision (ICD-10), and demographic information of hospitalized patients.4
In-hospital mortality was risk-adjusted using a binary logistic regression, considering all patients admitted during the study period and the following dependent variables: gender, age, type of admission (emergency, scheduled or following ambulatory care) and the APR-DRG’s risk of mortality (ROM) subclass for the hospitalization episode.5 In-hospital CMR was calculated as the proportion of deaths among all patients with bacteraemia or candidaemia, and RAMR was calculated as the ratio of observed to expected in-hospital mortality obtained from individual predictions of the logistic regression model.6 Discrimination was assessed using the area under the receiver operating characteristic curve (AUROC), and 95% CIs for the RAMR were obtained using the Byar approximation to the exact Poisson test.7 Calculations were performed with Stata v12.
Results
A total of 2939 BSIs with growth of the studied microorganisms were identified, of which 377 (12.83%) were excluded for the following reasons: 316 were Emergency Department (ED) cases without subsequent admission (10.75%), 33 were outpatient episodes (1.12%), 26 could not be matched with the MBDS (0.9%) and 2 were included in unspecified APR-DRGs. Of the 316 BSIs excluded because there was no subsequent hospital admission, 226 (71.5%) were due to E. coli, 39 (12.3%) to S. aureus, 33 (10.4%) to K. pneumoniae, 12 (3.8%) to P. aeruginosa and 6 (1.9%) to candidaemia (Table S1, available as Supplementary data at JAC Online). Regarding their outcome, most of them (82.6%) were discharged home and didn’t require further ED or hospital admission (Table S2). There were 21 episodes with polymicrobial BSI (0.7%), which were not included in the calculation of the overall CMR and RAMR but were included in specific calculations based on the type of microorganism. Finally, a study population of 2541 (86.46%) BSIs was obtained.
During the study period, a global incidence of 7.29 BSIs per 1000 discharges was observed for E. coli, 2.38 for K. pneumoniae, 1.16 for P. aeruginosa, 2.77 for S. aureus and 0.53 for Candida. A total of 436 patients died during hospitalization. Of these, 279 (64.0%) died within the first 14 days after the onset of the BSI, 117 (26.8%) after more than 14 days and 40 (9.2%) after more than 30 days. The highest CMRs were observed for P. aeruginosa and Candida, with values of 28.3% and 27.08%, respectively, while the lowest CMRs were observed for E. coli and K. pneumoniae, with values of 10.15% and 14.95%, respectively (Table 1).
CMRs and RAMRs by microorganism, 2019 to 2022
| CMR (%) | In-hospital RAMR (95% CI) | |
|---|---|---|
| E. coli (n = 1310) | 10.15 | 1.03 (0.86–1.21) |
| E. coli CTX-R (n = 239) | 10.04 | 0.91 (0.58–1.29) |
| K. pneumoniae (n = 428) | 14.95 | 1.33 (1.02–1.67) |
| K. pneumoniae CTX-R (n = 77) | 23.38 | 1.87 (1.11–2.8) |
| K. pneumoniae ETP-R (n = 86) | 24.42 | 1.62 (1–2.37) |
| P. aeruginosa (n = 209) | 28.23 | 2.06 (1.57–2.62) |
| P. aeruginosa MEM-R (n = 42) | 42.86 | 2.9 (1.72–4.34) |
| S. aureus (n = 498) | 20.28 | 1.8 (1.47–2.16) |
| S. aureus CLX-R (n = 93) | 23.66 | 1.94 (1.22–2.81) |
| Candidaemiaa (n = 96) | 27.08 | 1.99 (1.3–2.81) |
| Total (n = 2541) | 15.07 | 1.39 (1.26–1.54) |
| CMR (%) | In-hospital RAMR (95% CI) | |
|---|---|---|
| E. coli (n = 1310) | 10.15 | 1.03 (0.86–1.21) |
| E. coli CTX-R (n = 239) | 10.04 | 0.91 (0.58–1.29) |
| K. pneumoniae (n = 428) | 14.95 | 1.33 (1.02–1.67) |
| K. pneumoniae CTX-R (n = 77) | 23.38 | 1.87 (1.11–2.8) |
| K. pneumoniae ETP-R (n = 86) | 24.42 | 1.62 (1–2.37) |
| P. aeruginosa (n = 209) | 28.23 | 2.06 (1.57–2.62) |
| P. aeruginosa MEM-R (n = 42) | 42.86 | 2.9 (1.72–4.34) |
| S. aureus (n = 498) | 20.28 | 1.8 (1.47–2.16) |
| S. aureus CLX-R (n = 93) | 23.66 | 1.94 (1.22–2.81) |
| Candidaemiaa (n = 96) | 27.08 | 1.99 (1.3–2.81) |
| Total (n = 2541) | 15.07 | 1.39 (1.26–1.54) |
CLX-R, cloxacillin resistant; CTX-R, cefotaxime resistant; ETP-R, ertapenem resistant; MEM-R, meropenem resistant.
aData for candidaemias with reduced susceptibility to azoles have not been analysed due to a frequency of less than 30 episodes per year.
CMRs and RAMRs by microorganism, 2019 to 2022
| CMR (%) | In-hospital RAMR (95% CI) | |
|---|---|---|
| E. coli (n = 1310) | 10.15 | 1.03 (0.86–1.21) |
| E. coli CTX-R (n = 239) | 10.04 | 0.91 (0.58–1.29) |
| K. pneumoniae (n = 428) | 14.95 | 1.33 (1.02–1.67) |
| K. pneumoniae CTX-R (n = 77) | 23.38 | 1.87 (1.11–2.8) |
| K. pneumoniae ETP-R (n = 86) | 24.42 | 1.62 (1–2.37) |
| P. aeruginosa (n = 209) | 28.23 | 2.06 (1.57–2.62) |
| P. aeruginosa MEM-R (n = 42) | 42.86 | 2.9 (1.72–4.34) |
| S. aureus (n = 498) | 20.28 | 1.8 (1.47–2.16) |
| S. aureus CLX-R (n = 93) | 23.66 | 1.94 (1.22–2.81) |
| Candidaemiaa (n = 96) | 27.08 | 1.99 (1.3–2.81) |
| Total (n = 2541) | 15.07 | 1.39 (1.26–1.54) |
| CMR (%) | In-hospital RAMR (95% CI) | |
|---|---|---|
| E. coli (n = 1310) | 10.15 | 1.03 (0.86–1.21) |
| E. coli CTX-R (n = 239) | 10.04 | 0.91 (0.58–1.29) |
| K. pneumoniae (n = 428) | 14.95 | 1.33 (1.02–1.67) |
| K. pneumoniae CTX-R (n = 77) | 23.38 | 1.87 (1.11–2.8) |
| K. pneumoniae ETP-R (n = 86) | 24.42 | 1.62 (1–2.37) |
| P. aeruginosa (n = 209) | 28.23 | 2.06 (1.57–2.62) |
| P. aeruginosa MEM-R (n = 42) | 42.86 | 2.9 (1.72–4.34) |
| S. aureus (n = 498) | 20.28 | 1.8 (1.47–2.16) |
| S. aureus CLX-R (n = 93) | 23.66 | 1.94 (1.22–2.81) |
| Candidaemiaa (n = 96) | 27.08 | 1.99 (1.3–2.81) |
| Total (n = 2541) | 15.07 | 1.39 (1.26–1.54) |
CLX-R, cloxacillin resistant; CTX-R, cefotaxime resistant; ETP-R, ertapenem resistant; MEM-R, meropenem resistant.
aData for candidaemias with reduced susceptibility to azoles have not been analysed due to a frequency of less than 30 episodes per year.
Discrimination of the in-hospital mortality adjustment model showed an AUROC of 0.88 (Figure S1). Female gender (OR = 0.88; 95% CI 0.84–0.91), scheduled admission (OR = 0.43; 95% CI 0.39–0.46) or admission after ambulatory care (OR = 0.32; 95% CI 0.26–0.40) were protective factors (Table S4).
Throughout the period, all BSIs had a mortality rate above the expected adjusted mortality, as indicated by an RAMR value greater than 1, except for E. coli (1.03; 95% CI 0.86–1.21). The highest values were observed for P. aeruginosa, Candida and S. aureus, with values of 2.06 (95% CI 1.57–2.62), 1.99 (95% CI 1.3–2.81) and 1.8 (95% CI 1.47–2.16) in 2020, respectively. In the analysis of antibiotic-resistant subgroups, RAMRs were generally higher than for their sensitive counterparts, with statistically significant differences in P. aeruginosa resistant to meropenem (2.9; 95% CI 1.72–4.34) and K. pneumoniae resistant to ceftriaxone (1.87; 95% CI 1.11–2.8).
The main discrepancies in the temporal evolution of the point estimates for CMR and RAMR were observed for E. coli, which showed an increasing CMR and a decreasing RAMR over the years, and to a lesser extent for K. pneumoniae, which showed a decreasing RAMR year by year and a CMR higher in 2020 and 2021 than in 2019 (Figure 1 and Table S3). Considering all microorganisms together during the analysed period, the CMR showed an increasing trend from 12.89% in 2019 to 17.18% in 2021 and a decrease in 2022 to 14.54%, without reaching the level of 2019. Conversely, when considering RAMR, there was an increase from 2019 (1.39) to 2021 (1.51), although the decrease in 2022 (1.19) was below the 2019 level and without a statistically significant excess mortality.
Evolution of CMRs and RAMRs between 2019 and 2022 by microorganism and overall.
Discussion
We describe, to our knowledge for the first time, a BSI mortality adjusted with international standards such as APR-DRGs and show that it is significantly higher than expected by the adjustment model in all analysed groups except E. coli. Furthermore, we show that the time course of CMR may differ from RAMR, especially in E. coli, where it was reversed. Thus, our results indicate that the use of CMR as a single outcome metric may lead to misinterpretation.
The methodology we used is increasingly becoming a global standard in other areas of healthcare. Its strengths lie in its ability to easily use large sample sizes with outcome adjustments based on standards that use structured information readily available in hospitals. We used sex and age, which have previously been used for matching in case–control studies aimed at describing mortality attributable to bacteraemia.8 In this type of study, when it comes to matching with patient clinical complexity, prior hospital stay with or without admission unit are the few matching factors usually used.8–10 However, the use of APR-DRGs in mortality adjustment has the advantage of considering clinically meaningful and financially homogeneous groups. APR-DRGs have been shown to be superior to comorbidity indices (such as Charlson) in estimating mortality rates among hospitalized patients, which in turn are independent predictors of mortality in bacteraemia.5,11,12 In light of the above, RAMR calculated by adjusting for age, sex and APR-DRG mortality risk may provide a more comprehensive understanding than CMR. In contrast, the most comprehensive estimates of deaths associated with bacterial pathogens to date uses a complex methodology that would be difficult for most ASPs or other infectious disease reports to monitor on a regular basis.13
However, our study has relevant limitations that must be considered. First, the primary use of APR-DRGs is to compare hospitals in terms of complexity and outcomes for resource allocation, and there are no previous studies that have used it to adjust for mortality in BSIs.3 Second, there may be variations in APR-DRG coding between centres, as highlighted by the discrepancies in the correlation between APR-DRG and length of stay between different hospitals.14 Furthermore, the MBDS is limited to hospitalized patients, and therefore we excluded outpatient BSIs or those treated only in the ED. In our series, this type of BSI accounted for 316 episodes (12.83% of the total), and therefore the analysis of this type of bacteraemia should be approached independently. Notably, no patient died within 30 days of ED discharge (Table S1 and S2), and most of these episodes were due to E. coli (71.5%) and did not require further ED or hospital admission (65%), suggesting a low-risk patient profile for which exclusion from the analysis may be appropriate in any case. Finally, RAMR estimates of BSI mortality may be limited if deaths occur beyond 14 or 30 days after the onset of BSI, which is the recommended time frame for assessing in-hospital mortality.1,2 This is because RAMR considers mortality at any time during hospitalization. In our cohort, 26.8% and 9.2% of deaths occurred after 14 and 30 days, respectively, from the onset of BSI, and therefore these groups would also merit a special approach or exclusion from reporting.
Our study proposes the use of RAMR as a clinical outcome indicator in ASPs, or other infectious disease mortality reports, to complement the already established CMR or other mortality adjustments without normalized data. We strongly believe that it provides a more accurate context for clinicians and healthcare managers to assess the impact of reported mortality and allows for more accurate and appropriate intercentre comparisons. In future studies, it would be desirable to compare this methodology with results from studies based on comprehensive clinical registries to refine its interpretation and assess the feasibility of generalizing its use as a reference clinical outcome indicator in the context of ASPs.
Funding
This study was carried out as part of our routine work.
Transparency declarations
The authors declare that they have no conflicts of interest related to this research. The authors affirm that this manuscript is an honest, accurate, and transparent account of the study being reported, that no important aspects of the study have been omitted, and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
Supplementary data
Figure S1 and Tables S1 to S4 are available as Supplementary data at JAC Online.
