Antibiotics, Vol. 11, Pages 1732: Epidemiology and Economic Outcomes Associated with Timely versus Delayed Receipt of Appropriate Antibiotic Therapy among US Patients Hospitalized for Native Septic Arthritis: A Retrospective Cohort Study

1. Introduction Annual incidence of septic arthritis (SA) in the United States (US) and Western Europe is between two and ten cases per 100,000 persons, and is substantially higher among patients with rheumatoid arthritis—70 cases per 100,000 persons [1]. SA infections, which are most commonly caused by Staphylococcus aureus, are drivers of patient mortality, morbidity, and high healthcare costs [2,3]. In one observational study, over 80% of patients who presented to the emergency department (ED) for SA were hospitalized, with an average length of stay (LOS) of seven days and total annual charges of $30.6 to $36.9 million for ED care and $601 to $759 million for hospital care (USD) [3]. Among infectious diseases, delayed initiation of appropriate therapy is associated with increased risk of mortality, longer durations of antibiotic therapy and/or LOS in hospital, higher treatment costs, a decreased likelihood of being discharged home (vs. rehabilitation facilities or hospice), and long-term disability [2,4,5,6,7,8,9,10,11]. Microbiologic testing (including susceptibilities) for a suspected bacterial infection typically requires two to three days to complete [12]. Initial antibiotic therapy is therefore often empiric and based upon the infection and likely implicated pathogen(s). Early knowledge of causative pathogens thus reduces time to selecting appropriate therapy and potentially the magnitude of exposure to/duration of antibiotics [13].

To the best of our knowledge, the impact of delayed appropriate antibiotic therapy in SA is less well understood than in other infections. Accordingly, we utilized a large US hospital encounter database to examine the demographics, characteristics of admitting hospitals, clinical characteristics, infection details, treatment patterns, and outcomes among patients hospitalized for SA who received timely versus delayed antibiotic therapy.

2. Materials and MethodsWe used data from the Premier Healthcare Database (“Premier”) that spanned the period 1 January 2017 to 31 December 2019 (“study period”). Premier contains information on approximately 121 million inpatient visits and 897 million outpatient visits from >231 million patients seen at over 700 acute care hospitals, ambulatory surgery centers, and clinics [14]. The database includes patient demographics, hospital/facility characteristics, and data on admitting and discharge diagnoses (in International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] format); in addition it provides day-of-stay data on procedures (in International Classification of Diseases, Tenth Revision Procedure Coding System [ICD-10-PCS] format) and medications. Premier also includes financial data (defined as cost to institutions to render care). Laboratory test data—including specimen ID, test name, test day of service and time, and test results—are available for a subset of 370 hospitals [14]. Premier data are fully de-identified and compliant with the Health Insurance Portability and Accountability Act (HIPAA) of 1996 [14]. We identified all patients with at least one admission during the study period with a principal discharge diagnosis code of pyogenic (septic) arthritis (ICD-10-CM: M00.0XX, M00.1XX, M00.2XX, M00.8XX, M00.9) or direct infection of an unspecified joint in infectious and parasitic diseases classified elsewhere (M01X0) (see Supplementary Material for coding details). Principal diagnosis was used as it indicates the condition most responsible for admission. Among these patients, we focused on those who within two days of admission had: (1) ≥1 cultures from a relevant site (synovial fluid, blood, and related sites) for which microbiology data were available, and (2) receipt of ≥1 antibiotics (parenteral or oral preparations). For patients with multiple “qualifying” admissions, we selected the earliest admission (i.e., each patient contributed one admission to the analysis). Patients who died on the day of admission, and those transferred from other hospitals or discharged on the day of admission, were excluded. Consistent with prior studies [6,7,8,9,10,11], we excluded patients with evidence of other infection as it could not be ascertained from the database which treatment(s) were given for which infection; we also excluded those with evidence of conditions related to pregnancy or childbirth (additional factors not captured by the database may come into treatment considerations for such patients). For all remaining patients, the date of presentation to hospital was defined as the index date. Patients were followed from the index date until the earliest of either death, end of the study period, or 180 days post discharge (Figure S1).

Timely appropriate antibiotic therapy was defined as the receipt of antibiotic(s) within two days of admission (i.e., date of initial presentation plus two days) that collectively provided coverage for identified causative pathogens. Delayed therapy was defined as the receipt of appropriate therapy any day thereafter while hospitalized. All antibiotics administered within two days of admission were classified as initial therapy.

2.1. Study MeasuresPatient and hospital characteristics were assessed based on the available information from admission and during the 180 days prior (“pre-index”). Patient demographics included patient age, gender, race/ethnicity, and payer type; clinical characteristics included comorbidities (and the Charlson Comorbidity Index [CCI]), prior antibiotic use, prior hospitalizations and ED visits (all-cause and infection-related, with the latter defined as any encounter resulting in an infection diagnosis), and resource intensity [4,8,9,10,11]. Hospitals were classified based on geographic region, rural/urban setting, and teaching status. Organisms and corresponding susceptibility status were identified based on relevant cultures drawn within two days of admission. The proportion of patients with at least one ESKAPE organism (i.e., Enterococcus faecium, S. aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp.) was ascertained [15]. Antibiotic agents and treatment regimens received by patients during the index admission were also tabulated.Antibiotic treatment patterns (e.g., escalation, de-escalation, unchanged) were evaluated among patients with ≥5 consecutive days of antibiotic treatment during admission. Escalation was defined as any increase in the quantity of antibiotics or spectrum score, or any change from oral to parenteral therapy (vs. the previous day); de-escalation, as any decrease in the quantity of antibiotics or spectrum score, or any change from parenteral to oral therapy. In cases where criteria for both escalation and de-escalation were met, a hierarchical approach was applied, as follows: (1) a change in route (oral vs. parenteral); (2) the number of unique antibiotics administered daily; (3) a reduction in spectrum score [16]. Only patients with ≥2 days of antibiotic therapy, LOS ≥ 3 days following the initial day of antibiotic exposure in hospital, and no evidence of death within five days following antibiotic initiation, were included in analyses of escalation/de-escalation [16]. Per Moehring et al. [16], escalation/de-escalation analyses were conducted from treatment initiation to day five of therapy or discharge (whichever occurred first), and to the last day of therapy/discharge, respectively.

Utilization included the duration of in-hospital antibiotic therapy (i.e., number of days of admission during which ≥1 antibiotics were administered), total antibiotic exposure (i.e., cumulative total days of antibiotic exposure during admission), and LOS. LOS was measured from day one (hospital admission) to the date of discharge. Costs included cost of antibiotics, other pharmacotherapies, room and board, medical care, and other costs of care; total in-hospital costs were the sum of these “component” costs. In all instances, these represented costs to the hospital to render care, and were presented in US dollar values for the year during the study period in which they were incurred.

2.2. Statistical Analysis

For categorical variables, frequencies and percentages were reported; for continuous variables, means, medians, and ranges were reported. The statistical significance of differences between treatment groups was ascertained using Student’s t-tests for continuous measures, and chi-square tests for categorical measures.

Inverse probability treatment weighting (IPTW) was utilized to balance timely and delayed groups. Weights were estimated using the average treatment effect on the treated (ATT) methodology, in which the “treated” group (delayed appropriate therapy) were assigned a weight equal to one; corresponding weights for the “control” group (timely appropriate therapy) were estimated using propensity scores based on logistic regression models [17]. These scores were defined based on the outcome of interest (i.e., receipt of timely appropriate therapy) and were bound by zero and one. Demographic and clinical characteristics incorporated into this model included age group, gender, race/ethnicity, payer type, total number of hospital beds, teaching hospital status, obesity, CCI score, prior antibiotic use, markers of immunosuppression/general frailty, resource intensity, prior all-cause hospitalizations, and prior all-cause ED visits. Standardized differences were used to assess the degree to which IPTW methodology resulted in clinical equipoise between the groups (measures where standardized difference

IPTW-adjusted utilization and cost outcomes were estimated using generalized linear models and a gamma distribution with a log link. As the presence of respiratory diseases remained unbalanced between the two groups following weighting, it was included in the models. To increase precision, we also included several covariates that were used in the IPTW process. Results from these analyses were presented as adjusted means, 95% confidence intervals (CI), and p-values.

All analyses were conducted using version 9.4 of SAS (Cary, NC, USA).

4. Discussion

Among severe infections, timely initiation of appropriate antibiotic treatment is critical, as delays are associated with greater risk of morbidity and mortality, longer LOS, increased antibiotic exposure, and higher costs to provide care. Since the causative pathogen is typically unknown at presentation, initial therapy is often empiric, based on patient characteristics and knowledge of the implicated pathogen(s). Antibiotic stewardship further constrains clinicians in balancing comprehensive coverage against concerns of fostering resistance. Our findings indicate that receipt of timely appropriate antibiotic therapy for SA is associated with reduced exposure to antibiotics, shorter LOS, and an 18% reduction in costs to hospitals to render care.

Our findings are consistent with prior research indicating that the receipt of delayed appropriate antibiotic therapy among patients hospitalized with serious bacterial infections is associated with worse outcomes (all relative to the receipt of timely appropriate therapy) [4,15]. A high proportion of patients in our study sample received timely appropriate therapy, while similar studies in other disease areas suggest that delays in appropriate therapy occur more frequently (10–25%) [8,9,10,11,12,13,18]. This could be due to the high prevalence of S. aureus within our sample, which can be addressed with several antibiotics (including vancomycin, which also addresses MRSA, and was the most commonly used antibiotic agent in our study). Consistent with similar studies on SA, S. aureus was the most commonly implicated pathogen among our sample [2]. We also observed high-risk Gram-negative pathogens, including P. aeruginosa, Escherichia coli, Enterobacter cloacae complex, and yeasts (Candida albicans, Candida parapsilosis). These pathogens were more likely to be identified among patients in whom appropriate therapy was delayed, suggesting that clinicians may be less likely to provide appropriate coverage empirically for relatively atypical pathogens and instead require the results of microbiologic testing. This is consistent with other studies that note that these pathogens are associated with the receipt of delayed therapy [4,19]. Lodise and colleagues have noted that early knowledge of pathogens often results in a greater likelihood of timely and appropriate antibiotics [4]. If providers at hospitals within our sample utilized rapid microbiology testing for the 38.5% of patients in whom an ESKAPE pathogen was detected and appropriate therapy delayed, it may have resulted in a reduction of 14 days in hospital, 11 days of antibiotic therapy, and $35,310 (i.e., $3531 per patient) in total in-hospital costs. While traditional diagnostic methodologies such as synovial fluid culture typically require ≥48 h, molecular PCR panels can achieve results in a few hours, with accuracy reported equal or superior to traditional cultures; one study reported the sensitivity of synovial fluid and PCR to be 52% and 60%, respectively [20]. PCR has been found superior to culture in the identification of difficult-to-detect bacteria such as Cutibacterium spp. [20]. Use of rapid diagnostic technology provides results within a single provider shift, which allows for quicker and more efficient intervention with better follow-up such as minor dose and regimen adjustments (vs. waiting days for culture results to make adjustments). While the decision to purchase such technology is not always an easy one given resource-limited settings, such tools can often be used across infection types, thereby increasing not only their utility to clinicians in their selection process of antibiotic therapy and their detection of antimicrobial resistance markers but presumably also the cost-effectiveness/return on investment associated with such devices [21]. Without the availability of a rapid diagnostic tool, risk prediction tools also may help inform selection of empiric therapy (although further study is needed to determine the degree to which these tools accurately predict infection with various pathogens) [22].This study classified treatment as timely if at least one agent within the initial regimen was appropriate for the identified pathogen(s), although this does not necessarily indicate that treatment was optimal; therapy regimens included extended-spectrum coverage (e.g., ceftriaxone and vancomycin for Gram-positive bacteria) or coverage of pathogens not identified (piperacillin/tazobactam and vancomycin for the treatment of Gram-negative bacteria). It is reasonable to assume that rapid pathogen identification afforded by rapid testing may reduce the use of overly broad antibiotic coverage (especially among monomicrobial infections), although clinicians are hesitant to discontinue double-coverage therapy in septic infections [23,24]. Among our sample, patients who received timely appropriate therapy were more likely to experience de-escalation—and less likely to experience escalation—compared to patients who received delayed appropriate therapy. Other studies have noted the key role that antimicrobial de-escalation (ADE) plays in stewardship and better clinical outcomes for patients with severe infection, including a decreased risk of mortality and lower rates of ICU admission [25,26,27]. Use of a rapid PCR panel may increase the speed at which ADE can be considered (where relevant). Even among patients in whom initial therapy is appropriate, earlier knowledge of causative pathogen(s) may allow for earlier initiation of a more suitable dosing regimen. Regardless of the benefits of ADE, providers may be hesitant to de-escalate therapy for fear that the infection may not improve (or even worsen). Despite the identification of MRSA among only one in eight patients, vancomycin was used in over three-quarters of all patients. Excessive empiric use of vancomycin contributes to resistance, with minimum inhibitory concentration (MIC) “creep” widely reported in the literature [28,29,30,31]. The ability for rapid pathogen identification may preclude the prescription of irrelevant antibiotics, thereby removing the need to consider de-escalation in some cases.This study has several limitations. Capture of patient characteristics (e.g., overall levels of prior antibiotic exposure, comorbidities, infection history) within the database is limited to care rendered within hospitals that contribute to the database, resulting in the potential for misclassifying the timing of initial therapy and appropriateness as well as unmeasured confounding. This also led to an inability to characterize infections prior to presentation at hospital, including antibiotic use or culture findings prior to or after admission, limiting our ability to assess the association between the receipt of timely appropriate therapy and the overall duration of antibiotic therapy. On a related matter, outcomes also were restricted to those available in healthcare encounter data; data such as clinical efficacy evaluations from chart notes were unavailable and therefore could not be examined. Further study is needed to better understand how the timeliness of appropriate therapy impacts those evaluations. Premier lacks information on hospital policies that may influence prescription and treatment patterns. The degree to which these policies informed prescribing patterns observed in this study is unknown. Definitions of delayed appropriate therapy vary across existing literature from >24 h to >72 h following culture collection [32]. Premier antibiotic data are day-stamped but not time-stamped, which limited our ability to precisely capture the timing of the antibiotic receipt, resulting in possible misclassification of timely therapy. Without the availability of time-stamps, we used days since admission to classify timeliness, rather than assessing outcomes by actual testing turnaround time. While one systematic review reported that outcomes were similar across studies that defined delayed appropriate therapy as >48 h, >72 h, and at time of culture and susceptibility reporting (all relative to time of culture collection) [32], to the degree such misclassification occurred, it may at minimum have diluted the magnitude of delayed appropriate therapy. There also is no explicit linkage in the Premier database between the results of laboratory tests and specific diagnoses. Although we required a principal diagnosis of septic arthritis for inclusion in the study (which presumably was the reason that best explained the admission) along with a positive culture from a relevant site within two days of admission, and required review of all organisms and site among patients included in analyses by two members of the study team with substantial experience in SA, it is possible that some included culture draws may have been misclassified as relating to SA. However, without access to patients’ medical records, the degree to which misclassification occurred is unknowable. Finally, this study was limited to the microbiologic testing data available within the database; 85% of patients hospitalized for SA during the study period were excluded due to incomplete microbiologic data. This, and the knowledge that ours was a convenience sample, may limit the generalizability of findings.

留言 (0)

沒有登入
gif