ID indicates infectious diseases.
eTable. Time-Dependent Hazard Ratios for Associations Between ID Consultation and Outcomes
eFigure 1. Diagram of Semicompeting Risks Considered for the Analysis
eFigure 2. Time-Dependent Hazard Ratio of ID Consultation for Composite Outcome Over Time
eFigure 3. Kaplan-Meier Survival Plots for Composite Outcome by Place of Acquisition
eFigure 4. Kaplan-Meier Survival Plots for Composite Outcome by Methicillin Susceptibility
eFigure 5. Kaplan-Meier Survival Plots for Composite Outcome by Methicillin Susceptibility and Place of Acquisition
eFigure 6. Kaplan-Meier Survival Plots for Composite Outcome by Case Volume at Each Facility
eFigure 7. Kaplan-Meier Survival Plots for Composite Outcome by ID Consultation Utilization Rate at Each Facility
eFigure 8. Kaplan-Meier Survival Plots for Composite Outcome by Year of S aureus Bacteremia Onset
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Goto M, Jones MP, Schweizer ML, et al. Association of Infectious Diseases Consultation With Long-term Postdischarge Outcomes Among Patients With Staphylococcus aureus Bacteremia. JAMA Netw Open. 2020;3(2):e1921048. doi:10.1001/jamanetworkopen.2019.21048
Is infectious diseases consultation associated with improved long-term outcomes for patients with Staphylococcus aureus bacteremia?
In this large, multicenter cohort study of 31 002 patients with S aureus bacteremia, infectious diseases consultation during the index hospital stay was associated with reduced risk of all-cause mortality and recurrence of bacteremia for 5 years after discharge.
Investigations of the association of infectious diseases consultations with outcomes among patients with S aureus bacteremia should include long-term follow-up.
Staphylococcus aureus bacteremia (SAB) is common and associated with poor long-term outcomes. Previous studies have demonstrated an association between infectious diseases (ID) consultation and improved short-term (ie, within 90 days) outcomes for patients with SAB, but associations with long-term outcomes are unknown.
To investigate the association of ID consultation with long-term (ie, 5 years) postdischarge outcomes among patients with SAB.
Design, Setting, and Participants
This cohort study included all patients (N = 31 002) with a first episode of SAB who were discharged alive from 116 acute care units of the nationwide Veterans Health Administration where ID consultation was offered. Data were collected from January 2003 to December 2014, with follow-up through September 30, 2018. Data analysis was conducted from February to December 2019.
Infectious diseases consultation during the index hospital stay.
Main Outcomes and Measures
The primary outcome was time to development of a composite event of all-cause mortality or recurrence of SAB within 5 years of discharge. As secondary outcomes, SAB recurrence and all-cause mortality with and without recurrence were analyzed while accounting for semicompeting risks.
The cohort included 31 002 patients (30 265 [97.6%] men; median [interquartile range] age at SAB onset, 64.0 [57.0-75.0] years). Among 31 002 patients, there were 18 794 (60.6%) deaths, 4772 (15.4%) SAB recurrences, and 20 414 (65.8%) composite events during 5 years of follow-up; 12 773 deaths (68.0%) and 2268 recurrences (47.5%) occurred more than 90 days after discharge. Approximately half of patients (15 360 [49.5%]) received ID consultation during the index hospital stay; ID consultation was associated with prolonged improvement in the composite outcome (adjusted hazard ratio at 5 years, 0.71; 95% CI, 0.68-0.74; P < .001). Infectious diseases consultation was also associated with improved outcomes when all-cause mortality without recurrence and SAB recurrence were analyzed separately (all-cause mortality without recurrence: adjusted hazard ratio at 5 years, 0.77; 95% CI, 0.74-0.81; P < .001; SAB recurrence: adjusted hazard ratio at 5 years, 0.68; 95% CI, 0.64-0.72; P < .001).
Conclusions and Relevance
Having an ID consultation during the index hospital stay among patients with SAB was associated with improved postdischarge outcomes for at least 5 years, suggesting that contributions of ID specialists to management during acute infection may have a substantial influence on long-term outcomes. Further investigations of the association of ID consultation with outcomes after S aureus should include long-term follow-up.
Staphylococcus aureus bacteremia (SAB) is common and frequently associated with poor outcomes, with 30-day case-fatality rates ranging from 15% to 30% in recent studies.1-4 In addition, SAB is associated with a high frequency of recurrence after the completion of therapy, especially in the absence of definitive source control.5-11 Previous studies have demonstrated an association between receipt of infectious diseases (ID) consultation and improved short-term outcomes among patients with SAB, including reduced mortality and SAB recurrence within 90 days.10-16 Possible mechanisms explaining the improved short-term outcomes among patients receiving ID consultation include early identification and management of sources of infection, documentation of bacteremia clearance, appropriate choice and duration of antibiotic therapy, and close follow-up and monitoring of therapy.2,14,17
It is plausible that ID consultation in the early phase of SAB care (ie, during the initial hospitalization) could lead to long-term improvements in patient outcomes after discharge, far beyond 90 days. Staphylococcus aureus bacteremia is associated with excess mortality that lasts long after the initial illness.18-22 A retrospective matched cohort study in the Danish national health care system21 described higher attributable mortality rates among patients with SAB compared with age- and gender-matched controls that lasted at least 5 years after the initial SAB hospitalization. Higher long-term mortality rates among patients with SAB may be associated with both the presence of underlying comorbidities associated with risk of SAB and long-term consequences of the infection, including endocarditis, other organ injuries, and late bacteremia recurrence. Staphylococcus aureus bacteremia may recur several years after the initial infection if primary sources are not adequately controlled or infected medical devices are left in place.2,23,24 Improved management of SAB through ID consultation during the acute illness could potentially reduce late SAB recurrence and long-term sequelae, including death. If true, this would expand our understanding of the benefits of ID consultations for patients with SAB. However, prior studies have not examined outcomes after 90 days among patients with and without ID consultation.
The Veterans Health Administration (VHA) system provides an opportunity to study the association of ID consultation with long-term outcomes for patients with SAB. Administrative data from the VHA include information on the date of death for all enrolled patients regardless of the place of death, allowing for complete ascertainment of long-term, all-cause mortality following discharge among patients hospitalized with SAB. In addition, microbiology data within the VHA system can be aggregated across care episodes and hospitals for individual patients, allowing determination of SAB recurrence following discharge if the recurrence is identified within the VHA system. We examined national VHA data and aimed to determine the associations between ID consultations during initial hospitalizations for SAB and long-term outcomes for up to 5 years after discharge, including all-cause mortality and SAB recurrence. We hypothesized that ID consultation during the initial hospitalization would be associated with improved long-term, risk-adjusted outcomes among patients with SAB.
This was a retrospective cohort study of all patients hospitalized with SAB who were discharged alive from acute care settings in 116 VHA hospitals between January 2003 and December 2014. We excluded patients discharged from hospitals with no ID consultation available, which was determined if there were no consultations at that hospital throughout the study period. The institutional review board at the University of Iowa and the Research and Development Committee at the Iowa City Veterans Affairs Health Care System approved this study and granted a waiver for informed consent for this retrospective cohort study with minimal risk. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
We obtained data through the Veterans Affairs Informatics and Computing Infrastructure, which includes data extracted from the VHA integrated electronic medical record. Criteria for patient inclusion in the cohort were as follows: (1) a positive blood culture for S aureus during an inpatient stay in a medical or surgical acute care ward where ID consultation was offered and (2) survival to discharge from acute care wards. If a patient had more than 1 episode of SAB during the study period, we included only the first episode in the analysis.
The exposure variable of interest was receipt of ID consultation during the index hospital stay, defined as a completed consultation order or documentation from ID providers in the electronic medical record between the time the first positive blood culture was drawn and the time the patient was discharged from acute care. Specific details of findings and recommendations from ID consultation notes were not available. The primary outcome was time to a composite event of all-cause mortality or recurrence of SAB after discharge from acute care units, whichever occurred first. We limited analyses of death or SAB recurrence to the 5-year period after discharge because prior studies indicated that elevated attributable mortality among patients with SAB lasted at least this long.21 Recurrence of SAB was defined as any positive blood culture for S aureus after discharge from acute care units, obtained in an outpatient or inpatient setting in the VHA.24 We excluded cultures with discordant methicillin susceptibility results compared with the isolate in the index hospital stay (ie, methicillin-susceptible S aureus [MSSA] bacteremia following initial methicillin-resistant S aureus [MRSA] bacteremia), given that these presumably reflected recurrence because of reinfection and we were primarily interested in recurrence that could reflect relapse. Dates of death were obtained from the VHA Vital Status File, which combines mortality information from multiple VHA and non-VHA sources and traces patients even if they leave the VHA system. The Vital Status File has excellent agreement with the National Death Index. Hospice or palliative care enrollment status at the time of discharge was not available for analysis.
For each patient, we defined the time of bacteremia onset as the time the first positive blood culture was obtained. We classified places of acquisition as community acquired (CA), health care–associated community onset (HCA CO), or hospital onset (HO) per previously published definitions.25,26 The episode was considered HO bacteremia when the patient had been in the hospital for 48 hours or longer at the time of bacteremia onset. Among episodes that did not meet the criteria for HO bacteremia, HCA CO bacteremia was defined as follows: (1) admission to an acute care facility within 90 days before the bacteremia episode, (2) residence in a nursing home or rehabilitation facility, (3) receipt of renal replacement therapy, or (4) receipt of wound care or specialized nursing care in an outpatient setting or at home in the 30 days before the onset of bacteremia. If the patient did not meet any of these criteria, the episode was classified as CA bacteremia.
We also generated a list of risk-adjustment variables at the time of initial hospitalization, including patient age and sex, indicator variables for 29 comorbid conditions, methicillin susceptibility of the S aureus isolate, and body mass index (calculated as weight in kilograms divided by height in meters squared). Comorbid conditions were assessed using diagnosis codes based on the methods of Elixhauser et al27 and the algorithm proposed by Quan et al.28 Body mass index was categorized as underweight (<18.5), normal (18.5-24.9), overweight (25.0-29.9), and obese (≥30.0) according to World Health Organization classification.29
We first compared characteristics of patients with and without ID consultation during the index hospital stay using Fisher exact tests or χ2 tests for categorical measures and 2-sample t tests or Mann-Whitney U tests for continuous measures. To assess unadjusted associations between ID consultation during the index hospital stay and time to the primary composite outcome, we first used stratified Kaplan-Meier plots with log-rank tests, then univariable Cox regression models including only ID consultation as an explanatory variable. Patients who did not experience death or SAB recurrence by September 30, 2018, were censored at that point. Because the proportional hazards assumption was not met for patients with and without ID consultation, we included an interaction term between ID consultation and the logarithm of time (ie, log10[days after discharge + 1]) in the model and estimated time-dependent hazard ratios (HRs) at specific points following discharge.30-32 Briefly, these time-dependent HRs estimated the instantaneous hazards of outcome events at specific follow-up points among patients who had not experienced the event before that point and remained in the risk pool, stratified by receipt of ID consultation.
We then fitted multivariable Cox regression models, adjusting for patient demographic characteristics, body mass index, comorbidities, place of infection acquisition, and methicillin susceptibility. Variables were considered candidates for inclusion in the risk-adjustment model if there were bivariable associations with the primary outcome event at P < .10. We selected final variables for the model using a backward elimination strategy. The proportional hazard assumption was verified for all included explanatory variables through inspection of scaled Schoenfeld residual plots. We assessed overall model appropriateness by inspecting multiple residual plots, including Martingale residuals, deviance residuals, and score residuals. A hospital-specific random intercept was also included to account for clustering of patients within hospitals. We estimated time-dependent adjusted HRs for the association of ID consultation with event-free survival as described earlier.
We also examined time to SAB recurrence and mortality as secondary outcomes, taking semicompeting risks of death and recurrence into account (eFigure 1 in the Supplement).33 We first examined stratified cumulative incidence function plots for recurrence and mortality without SAB recurrence, using the Gray test for equality to test differences between patients with and without ID consultation.34 We then used subdistribution hazard Cox regression models with similar interaction terms between time and ID consultation to estimate time-dependent HRs for recurrence and death without recurrence, while accounting for competing risks (ie, death and SAB recurrence) simultaneously (h1 and h2 in eFigure 1 in the Supplement).35 We also fitted separate Cox regression models to estimate time from recurrence to death among those patients who experienced recurrence (h3 in eFigure 1 in the Supplement) with similar interaction terms between logarithm of time from discharge to death and ID consultation, while accounting for logarithm of time from discharge to recurrence as a covariate.
Finally, we conducted subgroup analyses to determine whether associations between ID consultation and long-term outcomes varied according to methicillin-susceptibility of isolates, place of acquisition, facility-level case volume, and hospital-level frequency of ID consultation. All P values were 2-sided, with P < .05 defined as the threshold of statistical significance. All analyses were conducted using SAS version 9.4 (SAS Institute).
There were 36 868 patients admitted to 124 VHA hospitals with a first episode of SAB during the study period. The inpatient mortality rate was 15.6% (5760 of 36 868 patients), and 31 108 patients (84.4%) were discharged alive from acute care units and included in this analysis. After excluding 106 patients in 8 hospitals with no ID consultation service available, we included 31 002 patients from 116 hospitals in the final analyses (30 265 [97.6%] men; median [interquartile range (IQR)] age at SAB onset, 64.0 [57.0-75.0] years). The characteristics of included patients are summarized in Table 1. Approximately half of patients (15 360 [49.5%]) received an ID consultation during the index hospital stay, and 15 642 patients (50.5%) did not. At the hospital level, the total case volume ranged from 22 to 1115 cases (median [IQR], 250 [105-369] cases), and the proportion of patients who received ID consultation ranged from 4.4% to 92.8% (median [IQR], 45.2% [31.2%-58.9%]). Compared with patients without an ID consultation, patients with a consultation were less likely to be older than 75 years (3564 [23.2%] vs 4434 [28.3%]; P < .001) and to have HO bacteremia (3974 [25.4%] vs 2964 [19.3%]; P < .001). Compared with patients without an ID consultation, those with a consultation were more likely to have obesity (4795 [30.7%] vs 5099 [33.2%]; P < .001) and to have a longer median (IQR) stay in acute care wards (8 [4-13] days vs 13 [8-20] days; P < .001). Although these differences were statistically significant in this large data set, the magnitudes of these differences were generally small. We identified 18 794 (60.6%) deaths, 4772 (15.4%) SAB recurrences, and 20 414 (65.8%) composite events (death or SAB recurrence) during the 5-year follow up period, and 10 588 patients (34.2%) had no outcome event. A total of 12 773 deaths (68.0%) and 2268 SAB recurrences (47.5%) occurred more than 90 days after discharge; 8290 deaths (44.1%) and 1174 SAB recurrences (24.6%) occurred more than 1 year after discharge. A relatively small proportion (ie, 1151 [3.7%]) were censored on October 1, 2018, with less than 5 years of follow-up. Table 2 summarizes the Cox regression risk-adjustment model.
Infectious diseases consultation during the index hospital stay was associated with a lower probability of experiencing the composite event in unadjusted survival analysis (P < .001 by log-rank test) (Figure 1). In a univariable Cox regression model using time-dependent coefficients, the hazards of experiencing the composite event were lower among patients who had an ID consultation compared with patients without an ID consultation at all time points following discharge (Table 3; eFigure 2 in the Supplement). The instantaneous HR at day 0 (ie, at time of discharge) was 0.48 (95% CI, 0.44-0.52; P < .001). It decreased over the 5-year period after discharge but maintained statistical significance at all time periods (eg, HR at 5 years, 0.64; 95% CI, 0.62-0.66; P < .001) (Table 3). Results were similar in a multivariable Cox regression model that adjusted for all covariates in Table 2 (adjusted HR at day 0, 0.54; 95% CI, 0.50-0.58; P < .001; adjusted HR at 5 years, 0.71; 95% CI, 0.68-0.74; P < .001) (Table 3).
Infectious diseases consultation during the index hospital stay was also associated with lower cumulative incidence of SAB recurrence and all-cause mortality without SAB recurrence (P < .001 by Gray test for both outcomes) (Figure 2). In a univariable subdistribution hazard Cox regression model, time-dependent HRs for associations of ID consultation with SAB recurrence were significant at all points (HR at day 0, 0.35; 95% CI, 0.31-0.40; P < .001; HR at 5 years, 0.68; 95% CI, 0.63-0.72; P < .001) (Table 3). Results were similar in a multivariable model adjusting for patient characteristics during the initial admission (ie, adjusted for all covariates in Table 2), with a persistent and significant association of ID consultation during the index hospital stay with lower cumulative incidence of SAB recurrence (adjusted HR at day 0, 0.35; 95% CI, 0.31-0.40; P < .001; adjusted HR at 5 years, 0.68; 95% CI, 0.64-0.72; P < .001) (Table 3). Similarly, the univariable subdistribution hazard Cox regression model showed that time-dependent HRs for associations between ID consultation and all-cause mortality without SAB recurrence were persistently significant over time (HR at day 0, 0.56; 95% CI, 0.51-0.62; P < .001; HR at 5 years, 0.71; 95% CI, 0.68-0.74; P < .001) (Table 3). Results were similar in a multivariable model adjusted for all covariates in Table 2 (adjusted HR at day 0, 0.61; 95% CI, 0.56-0.68; P < .001; adjusted HR at 5 years, 0.77; 95% CI, 0.74-0.81; P < .001) (Table 3). There was no significant association of ID consultation during the index hospital stay with all-cause mortality after the recurrence of SAB (Table 3). Results were similar when patients with metastatic cancer were excluded (data not shown).
In a subgroup analysis by place of acquisition, ID consultation was most strongly associated with improved outcomes among patients with CA SAB, followed by those with HCA CO and HO SAB (CA: adjusted HR at 5 years, 0.69; 95% CI, 0.64-0.76; P < .001; HCA CO: adjusted HR at 5 years, 0.70; 95% CI, 0.67-0.74; P < .001; HO: adjusted HR at 5 years, 0.76; 95% CI, 0.71-0.82; P < .001) (Table 3; eFigure 3 in the Supplement). Subgroup analysis by methicillin susceptibility showed that ID consultation was more strongly associated with long-term outcomes among patients with MSSA compared with patients with MRSA (MSSA: adjusted HR at 5 years, 0.68; 95% CI, 0.64-0.72; P < .001; MRSA: adjusted HR at 5 years, 0.74; 95% CI, 0.70-0.77; P < .001) (Table 3; eFigure 4 in the Supplement). Similar associations were observed with stratification by both place of acquisition and methicillin susceptibility (eTable 1 and eFigure 5 in the Supplement). In subgroup analyses stratified by case volume, ID consultation utilization rate at each facility, and year of bacteremia onset, associations between ID consultation and improved outcomes remained in all categories (eTable 1, eFigure 6, eFigure 7, and eFigure 8 in the Supplement).
In this large postdischarge cohort of patients with SAB, we observed that ID consultation during the index hospital stay was persistently associated with lower risk of death and SAB recurrence for at least 5 years. Cox regression models with time-dependent coefficients showed the associations were strongest at discharge and waned over time but remained significant and substantial throughout the follow-up period. This indicates that ID consultation during the index hospital stay was associated with long-term outcomes among patients with SAB and suggests that prior studies underestimated the benefits of ID consultation by only assessing short-term outcomes.
In subgroup analyses between patients with MSSA and MRSA, ID consultation was associated with higher event-free survival probability in both groups, although the association was more prominent in the MSSA group. We speculate that this is because ID consultants were more likely to use optimal antibiotics for definitive therapy for MSSA (eg, penicillinase-resistant penicillin or first-generation cephalosporin), which is associated with better outcomes5,36,37 but generally means a switch from empirical therapy with MRSA-active agents (eg, vancomycin). Previous studies38 showed that ID consultation was associated with appropriate definitive antibiotic therapy. On the other hand, it is a widespread practice to start empirical therapy with vancomycin for SAB in the United States,39,40 and vancomycin continuation can be definitive therapy if the isolate is MRSA. Infectious diseases consultation is likely to facilitate this additional step to optimize the care for MSSA and can be even more beneficial to patients with MSSA bacteremia. Similarly, when patients were stratified by place of acquisition, ID consultation had the strongest association with improved outcomes among patients with CA SAB. This is most likely because patients with CA SAB were less likely to have easily removable infection sources, such as a central venous catheter, and previous studies showed ID consultation was associated with a higher likelihood of source control,41-43 detection of endocarditis,10,13,38,44 and metastatic foci.10,13,44,45
This study also highlighted the poor long-term outcomes for patients with SAB, even with the advancement of medical therapy and increased use of evidence-based care. Nearly half of patients with SAB who were discharged alive from acute care hospitals died or experienced recurrence within 2 years of discharge. Notably, more than 15% of patients experienced SAB recurrence even though they were presumably stable enough to be discharged from acute care hospitals. Our results are consistent with other studies that evaluated outcomes of SAB during a 10-year period.21,22 For example, Yaw et al22 analyzed data from 582 patients with SAB in Australia and reported that 66% of patients died during their follow-up period of up to 15 years and that most deaths occurred during the first 5 years.
This study has limitations. As in any observational study, a limitation is the potential for unmeasured confounding. We also relied on administrative data for diagnoses of comorbidities, which may be inaccurate and allow residual confounding. In addition, we did not have detailed clinical data on sources of bacteremia, whether these sources were controlled, or all sites of infection. As discussed earlier, it is conceivable that ID consultation had differential associations with outcomes based on primary sites of infection, such as endocarditis or osteomyelitis. We were unable to identify SAB recurrences diagnosed with cultures outside the VHA system. Furthermore, most patients with SAB in VHA hospitals were men, which may limit the generalizability of these findings to female populations, although there is little biological plausibility to speculate that this association would be different between sexes.
In this study, we observed significant associations of ID consultation during the index hospital stay with outcomes among patients with SAB for up to 5 years after hospital discharge. A substantial proportion of patients with SAB had death or recurrence after being discharged alive from hospitals; notably, more than 15% of patients experienced recurrence of SAB. The associations were more prominent in subgroups of patients who were more likely to benefit from ID consultation, which suggests a causal association of ID consultation with long-term outcomes. As SAB has a poor long-term prognosis, these findings strongly support ID consultation as the standard of care for SAB where available.
Accepted for Publication: December 16, 2019.
Published: February 12, 2020. doi:10.1001/jamanetworkopen.2019.21048
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Goto M et al. JAMA Network Open.
Corresponding Author: Michihiko Goto, MD, MSCI, Division of Infectious Diseases, Department of Internal Medicine, University of Iowa Carver College of Medicine, 601 Highway 6 West, Bldg 42, Iowa City, IA 52246 (email@example.com).
Author Contributions: Dr Goto had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Goto, Schweizer, Livorsi, Perencevich, Ohl.
Acquisition, analysis, or interpretation of data: Goto, Jones, Schweizer, Richardson, Beck, Alexander, Ohl.
Drafting of the manuscript: Goto, Ohl.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Goto, Jones, Richardson, Beck.
Obtained funding: Schweizer, Perencevich.
Administrative, technical, or material support: Goto, Schweizer.
Supervision: Goto, Schweizer, Livorsi, Perencevich, Ohl.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by the Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service through the Center for Access and Delivery Research and Evaluation, grant number CIN 13-412.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views expressed herein are those of the authors and do not necessarily reflect the views of the United States Department of Veterans Affairs.
Create a personal account or sign in to: