Skip to main content

Some NLM-NCBI services and products are experiencing heavy traffic, which may affect performance and availability. We apologize for the inconvenience and appreciate your patience. For assistance, please contact our Help Desk at info@ncbi.nlm.nih.gov.

HHS Author Manuscripts logoLink to HHS Author Manuscripts
. Author manuscript; available in PMC: 2021 Feb 26.
Published in final edited form as: Clin Infect Dis. 2018 Mar 19;66(7):1004–1012. doi: 10.1093/cid/cix947

Risk of Subsequent Sepsis within 90 Days of a Previous Hospital Stay by Type of Antibiotic Exposure

James Baggs 1, Johm A Jernigan 1, Alison Laufer Halpin 1, Lauren H Epstein 1, Kelly M Hatfield 1, L Clifford McDonald 1
PMCID: PMC7909479  NIHMSID: NIHMS1019524  PMID: 29136126

Abstract

Background:

We examined the risk of sepsis within 90 days after discharge from a previous hospital stay by type of antibiotic received during the previous stay.

Methods:

We retrospectively identified a cohort of hospitalized patients from the Truven Health MarketScan Hospital Drug Database. We examined the association between the use of certain antibiotics, determined a priori, during the initial hospital stay and risk of post-discharge sepsis controlling for potential confounding factors in a multivariable logistic regression model. Our primary exposure was receipt of antibiotics more strongly associated with clinically important microbiome disruption. Our primary outcome was a hospital stay within 90 days of the index stay that included an ICD-9-CM discharge diagnosis of severe sepsis (995.92) or septic shock (785.52).

Results:

Among 516 hospitals, we randomly selected a single stay for eligible patients. Of those, 0.17% developed severe sepsis/septic shock within 90 days after discharge. The risk of sepsis associated with exposure to our high risk antibiotics was 65% higher compared to those without antibiotic exposure.

Conclusions:

Our study identified an increased risk of sepsis within 90 days of discharge among patients with exposure to high risk or increased quantities of antibiotics during hospitalization. Given a significant proportion of inpatient antimicrobial use may be unnecessary, this study builds on previous evidence suggesting that increased stewardship efforts in hospitals may not only prevent antimicrobial resistance, CDI and other adverse effects, but also reduce unwanted outcomes potentially related to disruption of the microbiota, including sepsis.

Keywords: sepsis, septic shock, anti-bacterial agents, administrative data, health-care associated infections

Background:

Sepsis is a life-threatening clinical syndrome characterized by acute organ dysfunction resulting from infection and a major contributor to excess morbidity, mortality, and healthcare costs.[1] Nearly one-quarter of sepsis cases have suspected gastrointestinal or an unknown source of infection.[24] In addition, there is a long-recognized role for the middle and lower gastrointestinal tract microbiota in the regulation of the immune response, specifically in sepsis.[57] Emerging evidence shows major disruptive forces such as antibiotics can lead to shifts in the microbiota that have greater pathogenic potential[8, 9], possibly leading to bacterial translocation[10, 11], a dysregulated immune response[5], or both.

Antibiotics are essential treatments for many hospitalized patients. While over half of hospitalized patients receive an antibiotic, [12, 13] an estimated 30–50% of antibiotic use in hospitals is inappropriate.[13, 14] Widespread use of antibiotics not only leads to selection for drug resistance and increases risk for Clostridium difficile infection (CDI), but also may increase a patient’s risk for later development of sepsis.[15] Prescott, et al. observed an increase in sepsis after hospital discharge for patients with either an infection-related hospitalization or hospitalization with CDI, which they suggested may be due to a distortion of the microbiota at least partially by antibiotics.[16] Understanding the association between antibiotic administration and accurately estimating the potential effect size of antibiotics in precipitating sepsis is important.

Our objective was to examine, among a cohort of US hospitalized patients, the risk of sepsis within 90 days after discharge according to receipt during a previous hospitalization of antibiotics categorized a priori on the basis of their propensity to disrupt the microbiome in a clinically important way.

Methods:

Data Sources:

Adult hospital discharge and drug use data was obtained from the Truven Health MarketScan® Hospital Drug Database (HDD), which contains individual billing records for all patients from approximately 500 hospitals. The use of this database to estimate US antimicrobial usage has been described previously and been shown to be representative of acute care hospitals in the US [12, 13, 17]. Since the information required to follow individuals longitudinally changed from 2010 to 2011, we included hospital admissions for all patients discharged during two time periods, January 1, 2007 through September 30, 2010 and January 1, 2011 through September 30, 2014. Similar to a previous study [12], for each hospitalization, we identified patient demographic and clinical information from the discharge billing data and antibiotic doses administered from the drug utilization data. Further, we categorized antibiotic doses into fourteen classes: aminoglycoside, 1st/2nd generation cephalosporin, 3rd/4th generation cephalosporin, lincosamide, fluoroquinolone, macrolide, vancomycin, sulfa, beta-lactam/beta-lactamase inhibitor combinations, carbapenem, penicillin, tetracycline, metronidazole, and miscellaneous. We excluded drugs with non-oral, non-parental routes of administration.

Study Settings and Patients:

Patients 18 years of age or older with an inpatient stay were included. For patients with multiple hospital stays within the study period, one stay was randomly selected to be the index stay for each patient. Patients with previously documented sepsis, sepsis documented during the index stay, who died during the index stay, or died in the hospital within the 90 days following a non-sepsis outcome were excluded. Further we excluded childbirth-related inpatient stays (ICD-9-CM codes: V30-V39).

Exposures:

Antibiotic exposures were identified from the selected index hospital stay and stratified into three groups of a priori risk based on published epidemiologic strength of association with CDI, which was considered a marker for intestinal microbiota disruption with demonstrated clinical importance. [18, 19] Because the intrinsic activity of an antibiotic against C. difficile may reduce this association by suppressing C. difficile while the patient is receiving the antibiotic, oral vancomycin was moved to a higher category of risk than would be suggested by its association with CDI, reflecting recent data demonstrating its profound microbiota-disruption potential.[20] High risk exposures included receipt of 3rd/4th generation cephalosporins, fluoroquinolones, lincosamides, beta-lactam/beta-lactamase inhibitor combinations, oral vancomycin, and carbapenems. [21, 22] Low risk exposures included receipt of 1st/2nd generation cephalosporins, macrolide, tetracycline, metronidazole, and sulfa without receipt of a high-risk antibiotic. Control exposures included receipt of an aminoglycoside, penicillin or intravenous vancomycin (antibiotics that minimally disrupt GI flora), without receipt of intermediate- or high-risk antibiotics. Finally, we compared the risk of sepsis in exposed patients to patients without exposure to any antibiotic, our reference group.

Outcome:

Our primary outcome (severe sepsis) was a hospital stay within 90 days of the index stay that included an ICD-9-CM discharge diagnosis of severe sepsis (995.92) or septic shock (785.52), identified in any position on the hospital discharge bill. We evaluated a secondary outcome (sepsis), using a published definition for hospital administrative data, [23] which requires ICD-9-CM codes for both infection and acute organ dysfunction within the same hospitalization or a sepsis specific diagnosis.[23] This definition was previously validated against chart review with high specificity and sensitivity.[24] For one data source within the HDD, admission dates are masked; therefore, instead of within 90 days, stays within the two months following the discharge month were identified.

Statistical Analysis:

Univariate comparisons of exposure and outcome groups were conducted using a chi-square test for categorical variables. To evaluate the risk of sepsis by exposure group, we conducted a multivariable logistic regression model comparing the odds of sepsis for those with high- and low-risk antibiotic exposures to control antibiotic exposures and to those without any antibiotic agent exposures. In addition, we evaluated the dose-response relationship in multivariable logistic models, which included either total days of antibiotic therapy or the number of antibiotic classes the patient received during the index stay as dose-response variables. All models included patient demographic and clinical characteristics from their index stay including sex, age category, length of stay (LOS), primary payer, previous hospitalization, co-morbidity score[25], certain chronic conditions as determined through ICD-9-CM codes (Table 1), diagnosis-related group (DRG) type, admission from the emergency room, critical care admission, index stay month and year, and hospital characteristics (bed size, urban/rural location, teaching status, census division). In addition, we conducted a similar analysis that used any readmission within 90 days as the outcome rather than either of the sepsis outcomes.

Table 1:

Demographic and Clinical Characteristics of Patients by Anti-bacterial Risk Group

Characteristic No anti-
bacterial
 % High Risk Anti-
bacterial
Low Risk
Anti-bacterial
 % Control Anti-
bacterial
%
All 5443331 42.8 3571964 28.1 3037618 23.9 653610 5.1
 
Severe Sepsis/Septic Shock1 6220 0.1 11510 0.3 2863 0.1 517 0.1
Sepsis2 30114 0.6 46509 1.3 12884 0.4 2381 0.4
 
Sex3
Male 2200860 40.4 1513291 42.4 1079404 35.5 125154 19.1
Female 3242471 59.6 2058673 57.6 1958214 64.5 528456 80.9
 
Age3
18-45 2124939 39.0 962866 27.0 1150774 37.9 436453 66.8
45-55 804304 14.8 540402 15.1 441473 14.5 44241 6.8
55-65 792920 14.6 585951 16.4 493663 16.3 54316 8.3
65-75 708661 13.0 577380 16.2 485401 16.0 55991 8.6
75-85 654360 12.0 558486 15.6 343580 11.3 45671 7.0
85+ 358147 6.6 346879 9.7 122727 4.0 16938 2.6
 
Length of Stay3
1-3 3962918 72.8 1680801 47.1 2010098 66.2 513705 78.6
4-6 980109 18.0 1032770 28.9 697238 23.0 93555 14.3
7-10 317700 5.8 489645 13.7 221196 7.3 30099 4.6
11+ 182604 3.4 368748 10.3 109086 3.6 16251 2.5
 
Days of Therapy3
0 DOT 5443331 100 0 0 0
1-2 DOT 0 818936 22.9 1999401 65.8 490392 75.0
3-6 DOT 0 1421841 39.8 891584 29.4 141357 21.6
7-13 DOT 0 914789 25.6 123828 4.1 18606 2.9
14+ DOT 0 416398 11.7 22805 0.8 3255 0.5
 
Co-morbidity score3
Missing 1646636 30.3 795458 22.3 1203561 39.6 397237 60.8
0 629472 11.6 423018 11.8 338376 11.1 38112 5.8
1 1142711 21.0 757460 21.2 533573 17.6 92480 14.2
2 561470 10.3 433523 12.1 200985 6.6 30276 4.6
3 307751 5.7 280385 7.9 108209 3.6 17547 2.7
4 177527 3.3 180496 5.1 62073 2.0 10377 1.6
5+ 309425 5.7 338113 9.4 121260 4.0 17126 2.6
NA 668339 12.3 363511 10.2 469581 15.5 50455 7.7
 
Critical Care Days3
0 days 5143787 94.5 3234679 90.6 2804869 92.3 626966 95.9
1-4 days 283559 5.2 248025 6.9 207910 6.8 23888 3.7
5-8 days 13039 0.2 50361 1.4 19604 0.7 2133 0.3
9+ days 2946 0.1 38899 1.1 5235 0.2 623 0.1
 
Previous Visits in last 90 Days3
0 5043450 92.7 3251389 91.0 2889772 95.1 621829 95.1
1 335442 6.2 268041 7.5 130638 4.3 27486 4.2
2+ 64439 1.2 52534 1.5 17208 0.6 4295 0.7
 
Chronic Conditions3
Metastatic Disease 97138 1.8 122258 3.4 60398 2.0 4643 0.7
Congestive Heart Failure 525770 9.7 435818 12.2 171177 5.6 31166 4.8
Dementia 98348 1.8 95488 2.7 24452 0.8 3303 0.5
Renal Failure 363606 6.7 275064 7.7 124797 4.1 23734 3.6
Weight Loss 60871 1.1 136950 3.8 24924 0.8 3951 0.6
Hemiplegia 68670 1.3 56366 1.6 19101 0.6 3445 0.5
Alcohol 110518 2.0 42127 1.2 18763 0.6 2434 0.4
Any Tumor 74126 1.4 91766 2.6 61029 2.0 4772 0.7
Arrhythmia 717807 13.2 489748 13.7 282338 9.3 44430 6.8
Pulmonary Disease 673730 12.4 829731 23.2 379966 12.5 56521 8.7
Coagulopathy 125395 2.3 80564 2.3 61966 2.0 10438 1.6
Complicated Diabetes 124829 2.3 117080 3.3 46541 1.5 9283 1.4
Anemia 493951 9.1 466309 13.1 283187 9.3 55380 8.5
Electrolytes 793661 14.6 900042 25.2 255241 8.4 36739 5.6
Liver Disease 114774 2.1 107250 3.0 36770 1.2 5577 0.9
Peripheral Vascular Disorder 211656 3.9 184949 5.2 132250 4.4 16758 2.6
Psychosis 585436 10.8 157763 4.4 81167 2.7 14681 2.3
Pulmonary Circulatory Disorders 95414 1.8 82525 2.3 31191 1.0 5239 0.8
HIV/AIDS 10992 0.2 24722 0.7 9170 0.3 880 0.1
Hypertension 2079013 38.2 1462200 40.9 1122538 37.0 129930 19.9
Obesity 424144 7.8 355440 10.0 316854 10.4 41423 6.3
Hyperlipidemia 1152995 21.2 642845 18.0 476396 15.7 60169 9.2
Uncomplicated Diabetes 762020 14.0 558242 15.6 376413 12.4 49085 7.5
Ischemic Heart Disease 96773 1.8 50710 1.4 38531 1.3 7209 1.1
Atrial Fib 474208 8.7 33160 9.3 177456 5.8 28288 4.3
Ventricular Fib 7819 0.1 5482 0.2 3957 0.1 854 0.1
 
DRG Type3
Med 4312907 79.2 2234137 62.6 644113 21.2 445452 68.2
Surgical 778517 14.3 1153118 32.3 2258226 74.3 188693 28.9
Other 351907 6.5 184709 5.2 135279 4.5 19465 3.0
 
Emergency Room3 2337488 42.9 1769377 49.5 532129 17.5 85763 13.1
 
Primary Payer3
Medicare 1677801 30.8 1437289 40.2 888556 29.3 119557 18.3
Medicaid 761141 14.0 356802 10.0 348551 11.5 160066 24.5
Blue Cross 757561 13.9 455286 12.8 550557 18.1 107452 16.4
Other Insurance 1104014 20.3 586823 16.4 590583 19.4 128696 19.7
HMO 649407 11.9 416726 11.7 426392 14.0 94544 14.5
Other 493407 9.1 319038 8.9 232979 7.7 43295 6.6
 
*

Footnote: Table does not include 39576 (0.3%) whose antibiotic exposure did not meet the criteria for the pre-defined exposure groups.

1 -

Severe sepsis/septic shock defined as a hospital stay within 90 days of the index stay that included an ICD-9-CM discharge diagnosis of severe sepsis (995.92) or septic shock (785.52), identified in any position on the hospital discharge bill.

2 -

Secondary outcome, sepsis used a published definition for hospital administrative data, “Angus definition”, which requires ICD-9-CM codes for both infection and acute organ dysfunction within the same hospitalization or a sepsis specific diagnosis.[23]

3 –

Characteristic based on information in the index stay.

As both facility- and patient-level data in the HDD are non-identifiable, it was determined this work did not constitute research involving human subjects. All data were analyzed using SAS version 9.3 (SAS Institute Inc., Cary, NC).

Results:

Among 516 hospitals, we identified 14,120,553 randomly selected index stays among adults. Of those, 1,205,226 (8.5%) either experienced sepsis during or prior to the index stay, and 305,428 (2.2%) died during the index stay or within 90 days of discharge; these patients were excluded. Of the remaining 12,746,135 index stays, there were 21,247 (0.17%) who had severe sepsis or septic shock identified within 90 days of their index stay using our primary outcome while 92,467 (0.7%) had sepsis identified within 90 days using our secondary outcome, Table 1.

Severe sepsis cases within 90 days of an index stay had a mean LOS of 13.1 days during their sepsis stay, 38% died during their sepsis hospitalization, and unspecified septicemia (038.9) was the most common primary diagnosis code listed in that stay, Table 2. Pneumonia was the most common primary ICD-9-CM diagnosis code listed for the index stay.

Table 2:

Demographic and Clinical Characteristics of Patients with Severe Sepsis/Septic Shock within 90 Days Post-Discharge

Characteristic Post Discharge
Severe
Sepsis/Septic
Shock1
%
All 21247
 
Characteristics of the Severe Sepsis/Septic Shock Stay
Mean Length of Stay (days) 13.1
 
Died 8019 37.7
 
Index Stay Characteristics
Sex
Male 10810 50.9
Female 10437 49.1
 
Age
18-45 1408 6.6
45-55 2135 10.1
55-65 3642 17.1
65-75 4818 22.7
75-85 5668 26.7
85+ 3567 16.8
 
Index Length of Stay
1-3 6357 29.9
4-6 6398 30.1
7-10 4392 20.7
11+ 4100 19.3
Mean (days) 7.4
 
Days of Therapy
0 DOT 6220 29.3
1-2 DOT 3135 14.8
3-6 DOT 4618 21.7
7-13 DOT 4044 19.0
14+ DOT 3230 15.2
 
Critical Care Days
0 days 18589 87.5
1-4 days 1892 8.9
5-8 days 449 2.1
9+ days 317 1.5
 
Ten Most Frequent Primary Diagnosis Codes
486 – Pneumonia, organism unspecified 690 3.3
428.0 – Congestive heart failure, unspecified 460 2.2
038.9 – Unspecified septicemia 455 2.1
599.0 – Urinary tract infection, site not specified 427 2.0
491.21 – Obstructive chronic bronchitis with acute exacerbation 416 2.0
584.9 – Acute kidney failure, unspecified 398 1.9
507.0 – Pneumonitis due to inhalation of food or vomitus 349 1.6
434.91 – Cerebral artery occlusion, unspecified with cerebral infarction 322 1.5
410.71 – Subendocardial infarction, initial episode of care 306 1.4
518.81 – Acute respiratory failure 304 1.4
1 -

Severe sepsis/septic shock defined as a hospital stay within 90 days of the index stay that included an ICD-9-CM discharge diagnosis of severe sepsis (995.92) or septic shock (785.52), identified in any position on the hospital discharge bill.

For those with an infection or CDI diagnosis in the index stay, the unadjusted proportion of patients with subsequent severe sepsis was higher compared to those without infection or CDI diagnosis, (0.3% vs 0.13%, p<0.0001) and (1.0% vs 0.16%, p<0.0001), respectively. Among patients with exposure to a high risk antibiotic agent during the index stay, the proportion of patients with severe sepsis post-discharge was 0.3%, p<0.0001, compared to just 0.1% of patients without any antibiotic exposures. Exposure to low risk or control antibiotic agents was not associated with an increased risk of sepsis compared to patients not exposed to any antibiotics in the unadjusted analysis, Table 1.

In the multivariable logistic model, exposure to a high risk antibiotic was associated with a higher risk of severe sepsis within 90 days of discharge compared to our referent group, OR=1.65 95% CI:1.59–1.70. Exposure to low risk and control antibiotic agents were not as strongly associated with severe sepsis (OR=1.07, 95% CI: 1.02–1.13, OR=1.22, 95% CI:1.12–1.34 respectively), Table 3. Further, both the number of unique antibiotics classes and total days of antibacterial therapy demonstrated significant dose-response association with post-discharge severe sepsis. Patients exposed to four or more antibiotic classes or those with 14 or more days of antibiotic therapy had over twice the risk of severe sepsis (OR=2.23, 95% CI:2.12–2.36, OR=2.17, 95% CI:2.06–2.29, respectively), compared to those without antibiotic exposure. Similar results were found for our secondary outcome, Table 3. In contrast, when using any readmission within 90 days, the association between a high risk antibiotic and readmission was close to one (OR=1.03, 95% CI:1.03–1.04), Table 3.

Table 3:

Adjusted Odds Ratio Describing the Association between Defined Exposures and Severe Sepsis and Septic Shock within 90 Days of Hospital Discharge in a Cohort of US Hospitals*

Primary Outcome:
Severe Sepsis/Septic
Shock1
Secondary Outcome:
Sepsis2
OR Lower
CI
Upper
CI
OR Lower
CI
Upper
CI
High Risk Anti-bacterial agents3 1.65 1.59 1.70 1.49 1.47 1.52
Low Risk Anti-bacterial agents4 1.07 1.02 1.13 1.04 1.02 1.06
Control Anti-bacterial agents5 1.22 1.12 1.34 1.20 1.15 1.25
No Exposure to Anti-bacterial agents Ref Ref
 
# Antibiotic Classes Exposed to during Stay
4+ 2.23 2.12 2.36 1.92 1.86 1.97
3 1.80 1.72 1.89 1.57 1.53 1.61
2 1.49 1.43 1.56 1.36 1.34 1.39
1 1.30 1.25 1.35 1.26 1.24 1.28
0 Ref Ref
 
# Days of Anti-bacterial Therapy
14+ 2.17 2.06 2.29 1.89 1.84 1.94
7-13 1.68 1.61 1.75 1.52 1.49 1.55
3-6 1.41 1.36 1.47 1.34 1.32 1.37
1-2 1.23 1.18 1.29 1.16 1.13 1.18
0 Ref Ref
 
For Patients Receiving High Risk Anti-bacterial Agents
# Antibiotic Classes Exposed to during Stay
4+ 1.53 1.44 1.63 1.36 1.32 1.40
3 1.27 1.20 1.34 1.14 1.11 1.17
2 1.08 1.03 1.14 1.02 1.00 1.05
1 Ref Ref
 
# Days of Anti-bacterial Therapy
14+ 1.61 1.49 1.74 1.47 1.41 1.52
7-13 1.28 1.19 1.37 1.20 1.16 1.24
3-6 1.15 1.08 1.23 1.10 1.07 1.13
1-2 Ref Ref
 
For Patients Receiving Low Risk or Control Anti-bacterial Agents
# Antibiotic Classes Exposed to during Stay
4+ 1.20 0.83 1.74 1.72 1.48 2.01
3 1.13 0.96 1.33 1.08 1.00 1.17
2 1.03 0.95 1.12 0.99 0.95 1.03
1 Ref     Ref
 
# Days of Anti-bacterial Therapy
14+ 1.21 0.98 1.50 1.43 1.29 1.59
7-13 1.11 0.99 1.26 1.20 1.14 1.28
3-6 0.92 0.85 0.99 1.02 0.98 1.06
1-2 Ref Ref
 
Patients with Primary Infectious Diagnosis Code
High Risk Anti-bacterial agents 1.53 1.43 1.64 1.41 1.37 1.46
Low Risk Anti-bacterial agents 1.00 0.91 1.11 1.04 0.99 1.09
Control Anti-bacterial agents 1.07 0.92 1.26 1.06 0.98 1.15
No Exposure to Anti-bacterial agents Ref Ref
 
# Antibiotic Classes Exposed to during Stay
4+ 2.06 1.90 2.24 1.79 1.71 1.86
3 1.65 1.52 1.79 1.49 1.43 1.55
2 1.36 1.26 1.47 1.33 1.27 1.37
1 1.20 1.11 1.30 1.17 1.13 1.22
0 Ref Ref
 
# Days of Anti-bacterial Therapy
14+ 1.95 1.79 2.12 1.76 1.68 1.83
7-13 1.56 1.45 1.68 1.44 1.39 1.50
3-6 1.35 1.25 1.46 1.30 1.25 1.35
1-2 1.06 0.96 1.17 1.04 0.99 1.09
0 Ref Ref
 
Model using any readmission within 90 days as outcome instead of sepsis6
High Risk Anti-bacterial agents 1.03 1.03 1.04
Low Risk Anti-bacterial agents 0.88 0.88 0.89
Control Anti-bacterial agents 0.90 0.89 0.92
No Exposure to Anti-bacterial agents Ref
*

Multivariable Logistic Model adjusted for sex, age, primary payer, previous hospitalizations within 90 days, length of stay, co-morbidity score, surgical or medical DRG, emergency room visit, critical care stays during the index visit, month and year of the index visit, hospital bed size, hospital urban/rural location, hospital teaching status, hospital census division, and various chronic conditions based on ICD-9-CM discharge codes including: metastatic disease, congestive heart failure, dementia, renal failure, weight loss, hemiplegia, alcohol, any tumor, arrhythmia, pulmonary disease, coagulopathy, complicated diabetes, anemia, electrolytes, liver disease, peripheral vascular disorder, psychosis, pulmonary circulatory disorders, HIV/AIDS, hypertension, obesity, hyperlipidemia, uncomplicated diabetes, ischemic heart disease, atrial fib, and ventricular fib.

1 -

Severe sepsis/septic shock defined as a hospital stay within 90 days of the index stay that included an ICD-9-CM discharge diagnosis of severe sepsis (995.92) or septic shock (785.52), identified in any position on the hospital discharge bill.

2 -

Secondary outcome, sepsis used a published definition for hospital administrative data, “Angus definition”, which requires ICD-9-CM codes for both infection and acute organ dysfunction within the same hospitalization or a sepsis specific diagnosis.[23]

3 -

High risk anti-bacterial exposures included any receipt of 3rd/4th generation cephalosporins, fluoroquinolones, lincosamides, beta-lactam/beta-lactamase inhibitor combinations, oral vancomycin, and carbapenems.

4 -

Low risk anti-bacterial exposures included receipt of 1st/2nd generation cephalosporins, macrolide, tetracycline, metronidazole, and sulfa without receipt of a high risk antibiotic.

5 -

Control anti-bacterial exposures included any receipt of an aminoglycoside, penicillin or intravenous vancomycin (antibiotics that minimally disrupt GI flora) without receipt of intermediate or high-risk antibiotics.

6 -

In addition, we conducted a similar model with the same exposures that used any readmission within 90 days as the outcome rather than either of the sepsis outcomes.

Since most patients exposed to four or more different classes of antibiotics were also in the high risk antibiotic group, we further evaluated the dose-response within the high risk group alone. We also limited the analysis to those with an infection-related primary discharge code during the index stay. Dose responses were observed when our analysis was limited to one of these groups, Table 3.

Discussion:

We found a significant association between antibiotic exposure in the hospital and severe sepsis and septic shock either as the cause of or occurring during a subsequent hospitalization within 90 days of discharge. Exposure to antibiotics such as 3rd/4th generation cephalosporin, lincosamide, fluoroquinolone, beta-lactam/beta-lactamase inhibitor combinations, oral vancomycin, and carbapenem were associated with an increased risk of sepsis. Furthermore, significant dose-response effects were observed for the number of antibiotic classes a patient received during the index hospitalization as well as the total days of therapy. In contrast, the risk of post-discharge sepsis for exposure to low risk antibiotics was diminished.

Our findings support, but do not prove, the hypothesis that microbiota disruption is associated with an increased risk of severe sepsis and septic shock within 90 days of discharge from a hospital stay. Prescott, et al. previously demonstrated that the rate of sepsis 90 days post-hospitalization was 3-fold greater than other observation periods.[16] They also found that hospital events, such as infection or CDI further increased this rate.[16] Presumably these events, infection and CDI, would disrupt the patient’s microbiota in part due to anti-bacterial agents. Our study further supports this hypothesis by showing that increased antibiotic exposure, or exposure to specific anti-bacterial agents more likely to disrupt the microbiota are associated with an increased risk in severe sepsis in the 90 days following hospital discharge. We were able to study a large population of over 500 hospitals over a seven year period. Unlike the study by Prescott, et al, we were able to include hospital pharmacy data, which was previously shown to be consistent with other estimates of hospital antibiotic usage and a representative sample of hospitals in the US.[12]

In addition, we determined a priori the antibiotic exposure categories based upon their epidemiologic association with clinically important microbiome disruption (i.e., CDI risk). While the types of antibiotic-mediated disruptions that predispose to sepsis may ultimately be determined to be different from those that predispose to CDI, hypothesis-driven a priori analyses based upon a theoretical framework may lessen the risk for unmeasured bias or spurious associations based upon chance alone. Our study only identified a significantly large association between sepsis and those antibiotics most likely to disrupt the patient’s microbiota [1820] while low risk and control antibiotics showed much smaller increases in the risk of sepsis. In addition, both dose response variables showed significant trends with increasing amounts of antibiotics, further supporting our hypothesis that disruption of the patient’s microbiota leads to an increased risk of post-discharge sepsis. Furthermore, we were able to control for a number of demographic and clinical characteristics including certain chronic conditions likely associated with antibiotic use and hospital readmission in our multivariable models. In sensitivity analyses, we found similar estimates to those by Prescott comparing infection-related or CDI-related hospitalizations to non-infection-related hospitalizations without our antimicrobial exposures [26]. We also eliminated patients with an ICD-9-CM code for CDI either in their index visit or during the post-discharge sepsis visit and found consistent results with our primary model, suggesting that our association was not confounded by the well described relationship between antibiotics and CDI. However, additional epidemiologic and biologic studies may further explore this hypothesis.

Antibiotic-mediated gut microbiota disruptions may increase the risk of sepsis via any one or a combination of three broad pathways. The first of these is loss of direct inhibition and competitive nutrient utilization, leading to loss of colonization resistance against more virulent and potentially pathogenic microbiota members.[9] Another pathway emphasizes the loss of immune regulatory dampening functions of the gut microbiota itself, whereby, at least theoretically, antibiotic effects on the gut microbiota may contribute to a more pronounced septic response from even a non-gut-related site of primary infection.[5] A third area is loss of integrity of the gut mucosal barrier function, largely due to loss of short chain fatty acids normally produced by a healthy microbiota that serve as the main nutrient source for large intestinal enterocytes.[27]

Direct adverse drug events, such as allergic reactions and toxicities like tendon rupture or renal toxicity, as well as the microbiota-mediated effects of antibiotic-associated diarrhea and especially CDI, are long-recognized forms of patient harm resulting from antibiotics.[13, 21, 28] While exact mechanisms remain under investigation, there is now a small but increasing body of human observational evidence, and animal data suggesting broader detrimental effects on patient outcomes rooted in microbiota disruptions that result from, among other environmental insults,[29] antibiotic use.[8, 11] Taur et al. showed that, even after controlling for confounders, 3-year mortality in bone marrow transplant recipients was associated with gut microbiota diversity at engraftment.[30] Mai et al. found that antibiotic-mediated changes in microbiota composition, especially the loss of potentially protective members and ‘bloom’ of proteobacteria, leading up to onset, were associated with late-onset sepsis in human neonates.[31] In adult patients, the population evaluated here, poorer outcomes in patients with the systemic inflammatory response (SIRS) are associated with greater microbiota disruption. [32]

One hope from our findings is that future innovations focused on restoring or protecting the lower intestinal microbiota from antibiotic-mediated disruption might become a possible approach for preventing sepsis.[33] Recent studies have established fecal microbiota transplantation (FMT) as a front-line therapy for multiply recurrent C. difficile infection.[34] Despite at least two case reports of FMT apparently used successfully to treat sepsis [35, 36], this remains highly experimental and carries unknown risk. Although animal data suggest that a defined probiotic consortia could be developed to restore the barrier function of the gut and thereby possibly prevent antibiotic-mediated sepsis on that account [37], there are examples where probiotics administered in the throes of severe illness, specifically acute pancreatitis, have increased mortality.[38] Recently, a large, randomized, double-blind, placebo-controlled trial of an oral synbiotic preparation given to infants in rural India, observed a 40% reduction in sepsis outcomes.[39] Protecting the lower intestinal microbiota from antibiotic-mediated disruption may be another strategy available soon. Though still under development, methods to inactivate antibiotics that reach the lower intestine via either enzymatic deactivation, (e.g., an orally administered beta-lactamase[40]), or by binding with an absorbent [41], appear promising.

However, another currently available prevention strategy is improved antibiotic stewardship. Although early antibiotic administration is critical for the management of sepsis, [4244] there are many other conditions for which antibiotics are unnecessary and yet often prescribed, thereby needlessly increasing patients’ risk for complications including future sepsis;[13] for example, treatment of asymptomatic bacteruria or positive cultures from non-sterile body sites where colonization is likely. In addition, recent studies suggest that certain common, serious infections may not need to be treated with broad spectrum or as many agents [45] or for as long duration as previously thought.[46]

This study has several limitations. First, administrative data such as the HDD are not collected for research purposes, and misclassification in the pharmacy, clinical, and facility data including the use of ICD-9-CM diagnostic codes can lead to bias. However, this bias is likely non-differential and would typically bias the results towards null values. Also, this type of pharmacy charge data was previously validated in small samples with excellent agreement.[47, 48] In addition, our outcome was based on ICD-9-CM diagnostic codes, but this definition of sepsis was previously validated. [24] Although we controlled for several demographic and clinical characteristics in the multivariable analysis, residual confounding from unknown factors could affect our findings, particularly the presence of underlying conditions or characteristics that increase antibiotic use in the index hospitalization and the risk of subsequent infection. However, in an analysis restricted to patients with no discharge diagnosis codes indicating an infection during the index hospitalization, our findings were similar, suggesting that an underlying predisposition to infection is less likely to confound our observed association. Further, when we included any readmission within 90 days as our outcome instead of sepsis, we observed that the odds ratio for our high risk antibiotic group was reduced to nearly one, providing additional support for our hypothesis, rather than underlying disease, explaining the association. In addition, we could only include post-discharge cases of sepsis in which patients returned to the same hospital, as patients in the HDD cannot be followed longitudinally across different hospitals. As such, our estimate of the proportion of sepsis cases following hospitalization was smaller than the previous study and death outside the same hospital was not detectable.[26] Finally, our study did not include any exposure data from health care encounters outside of the hospital or antibiotics prescribed at discharge.

In conclusion, our study observed a significant increase in severe sepsis and septic shock within 90 days of discharge for patients exposed to antibiotics in the hospital likely to disrupt the patient’s microbiota. Given that a significant proportion of inpatient antimicrobial use may be unnecessary [14, 49], this study builds on a growing evidence base suggesting that increased stewardship efforts in hospitals may not only prevent antimicrobial resistance, CDI and other adverse effects, but also reduce other unwanted outcomes potentially related to disruption of the microbiota, including sepsis.

Key Points.

Among a retrospective cohort of patients, the risk of sepsis was 65% higher for patients exposed to antibiotics more likely to disrupt the gut microbiota compared to those without any antibiotic exposure.

Acknowledgments

Funding/Support: This work was supported by the Centers for Disease Control and Prevention. All authors were employees of the CDC. This manuscript underwent the CDC clearance process.

Disclaimer: The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Footnotes

Conflict of Interest Disclosures: No financial conflicts or potential conflicts of interest were reported by any of the authors.

References

  • 1.Angus DC, van der Poll T. Severe sepsis and septic shock. The New England journal of medicine 2013; 369(21): 2063. [DOI] [PubMed] [Google Scholar]
  • 2.Ranieri VM, Thompson BT, Barie PS, et al. Drotrecogin alfa (activated) in adults with septic shock. The New England journal of medicine 2012; 366(22): 2055–64. [DOI] [PubMed] [Google Scholar]
  • 3.Vincent JL, Rello J, Marshall J, et al. International study of the prevalence and outcomes of infection in intensive care units. Jama 2009; 302(21): 2323–9. [DOI] [PubMed] [Google Scholar]
  • 4.Novosad SA, Sapiano MR, Grigg C, et al. Vital Signs: Epidemiology of Sepsis: Prevalence of Health Care Factors and Opportunities for Prevention. MMWR Morbidity and mortality weekly report 2016; 65(33): 864–9. [DOI] [PubMed] [Google Scholar]
  • 5.Mittal R, Coopersmith CM. Redefining the gut as the motor of critical illness. Trends in molecular medicine 2014; 20(4): 214–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Alverdy JC, Chang EB. The re-emerging role of the intestinal microflora in critical illness and inflammation: why the gut hypothesis of sepsis syndrome will not go away. Journal of leukocyte biology 2008; 83(3): 461–6. [DOI] [PubMed] [Google Scholar]
  • 7.Alverdy JC, Krezalek MA. Collapse of the Microbiome, Emergence of the Pathobiome, and the Immunopathology of Sepsis. Crit Care Med 2017; 45(2): 337–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Modi SR, Collins JJ, Relman DA. Antibiotics and the gut microbiota. J Clin Invest 2014; 124(10): 4212–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.McKenney PT, Pamer EG. From Hype to Hope: The Gut Microbiota in Enteric Infectious Disease. Cell 2015; 163(6): 1326–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Knoop KA, McDonald KG, Kulkarni DH, Newberry RD. Antibiotics promote inflammation through the translocation of native commensal colonic bacteria. Gut 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Yu LC, Shih YA, Wu LL, et al. Enteric dysbiosis promotes antibiotic-resistant bacterial infection: systemic dissemination of resistant and commensal bacteria through epithelial transcytosis. American journal of physiology Gastrointestinal and liver physiology 2014; 307(8): G824–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Baggs J, Fridkin SK, Pollack LA, Srinivasan A, Jernigan JA. Estimating National Trends in Inpatient Antibiotic Use Among US Hospitals From 2006 to 2012. JAMA Intern Med 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Fridkin S, Baggs J, Fagan R, et al. Vital signs: improving antibiotic use among hospitalized patients. MMWR Morbidity and mortality weekly report 2014; 63(9): 194–200. [PMC free article] [PubMed] [Google Scholar]
  • 14.Dellit TH, Owens RC, McGowan JE Jr., et al. Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America guidelines for developing an institutional program to enhance antimicrobial stewardship. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 2007; 44(2): 159–77. [DOI] [PubMed] [Google Scholar]
  • 15.Ayres JS, Trinidad NJ, Vance RE. Lethal inflammasome activation by a multidrug-resistant pathobiont upon antibiotic disruption of the microbiota. Nature medicine 2012; 18(5): 799–806. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Prescott HC, Dickson RP, Rogers MA, Langa KM, Iwashyna TJ. Hospitalization Type and Subsequent Severe Sepsis. Am J Respir Crit Care Med 2015; 192(5): 581–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Yi SH, Jernigan JA, McDonald LC. Prevalence of probiotic use among inpatients: A descriptive study of 145 U.S. hospitals. Am J Infect Control 2016; 44(5): 548–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Slimings C, Riley TV. Antibiotics and hospital-acquired Clostridium difficile infection: update of systematic review and meta-analysis. The Journal of antimicrobial chemotherapy 2014; 69(4): 881–91. [DOI] [PubMed] [Google Scholar]
  • 19.Brown KA, Khanafer N, Daneman N, Fisman DN. Meta-analysis of antibiotics and the risk of community-associated Clostridium difficile infection. Antimicrob Agents Chemother 2013; 57(5): 2326–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Isaac S, Scher JU, Djukovic A, et al. Short- and long-term effects of oral vancomycin on the human intestinal microbiota. The Journal of antimicrobial chemotherapy 2017; 72(1): 128–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Lewis BB, Buffie CG, Carter RA, et al. Loss of Microbiota-Mediated Colonization Resistance to Clostridium difficile Infection With Oral Vancomycin Compared With Metronidazole. J Infect Dis 2015; 212(10): 1656–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Johnson S, Louie TJ, Gerding DN, et al. Vancomycin, metronidazole, or tolevamer for Clostridium difficile infection: results from two multinational, randomized, controlled trials. Clin Infect Dis 2014; 59(3): 345–54. [DOI] [PubMed] [Google Scholar]
  • 23.Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001; 29(7): 1303–10. [DOI] [PubMed] [Google Scholar]
  • 24.Iwashyna TJ, Odden A, Rohde J, et al. Identifying patients with severe sepsis using administrative claims: patient-level validation of the angus implementation of the international consensus conference definition of severe sepsis. Med Care 2014; 52(6): e39–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A Combined Comorbidity Score Predicted Mortality in Elderly Patients Better Than Existing Scores. Journal of clinical epidemiology 2011; 64(7): 749–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Prescott HC, Langa KM, Iwashyna TJ. Readmission diagnoses after hospitalization for severe sepsis and other acute medical conditions. JAMA 2015; 313(10): 1055–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Yamada T, Shimizu K, Ogura H, et al. Rapid and Sustained Long-Term Decrease of Fecal Short-Chain Fatty Acids in Critically Ill Patients With Systemic Inflammatory Response Syndrome. JPEN Journal of parenteral and enteral nutrition 2015; 39(5): 569–77. [DOI] [PubMed] [Google Scholar]
  • 28.Shehab N, Patel PR, Srinivasan A, Budnitz DS. Emergency department visits for antibiotic-associated adverse events. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 2008; 47(6): 735–43. [DOI] [PubMed] [Google Scholar]
  • 29.Bajaj JS, Cox IJ, Betrapally NS, et al. Systems biology analysis of omeprazole therapy in cirrhosis demonstrates significant shifts in gut microbiota composition and function. American journal of physiology Gastrointestinal and liver physiology 2014; 307(10): G951–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Taur Y, Jenq RR, Perales MA, et al. The effects of intestinal tract bacterial diversity on mortality following allogeneic hematopoietic stem cell transplantation. Blood 2014; 124(7): 1174–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Mai V, Torrazza RM, Ukhanova M, et al. Distortions in development of intestinal microbiota associated with late onset sepsis in preterm infants. PloS one 2013; 8(1): e52876. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Shimizu K, Ogura H, Hamasaki T, et al. Altered gut flora are associated with septic complications and death in critically ill patients with systemic inflammatory response syndrome. Digestive diseases and sciences 2011; 56(4): 1171–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Tosh PK, McDonald LC. Infection control in the multidrug-resistant era: tending the human microbiome. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 2012; 54(5): 707–13. [DOI] [PubMed] [Google Scholar]
  • 34.Drekonja D, Reich J, Gezahegn S, et al. Fecal Microbiota Transplantation for Clostridium difficile Infection: A Systematic Review. Annals of internal medicine 2015; 162(9): 630–8. [DOI] [PubMed] [Google Scholar]
  • 35.Li Q, Wang C, Tang C, et al. Successful treatment of severe sepsis and diarrhea after vagotomy utilizing fecal microbiota transplantation: a case report. Critical care 2015; 19: 37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Li Q, Wang C, Tang C, et al. Therapeutic modulation and reestablishment of the intestinal microbiota with fecal microbiota transplantation resolves sepsis and diarrhea in a patient. The American journal of gastroenterology 2014; 109(11): 1832–4. [DOI] [PubMed] [Google Scholar]
  • 37.Li M, Liang P, Li Z, et al. Fecal microbiota transplantation and bacterial consortium transplantation have comparable effects on the re-establishment of mucosal barrier function in mice with intestinal dysbiosis. Frontiers in microbiology 2015; 6: 692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Besselink MG, van Santvoort HC, Renooij W, et al. Intestinal barrier dysfunction in a randomized trial of a specific probiotic composition in acute pancreatitis. Annals of surgery 2009; 250(5): 712–9. [DOI] [PubMed] [Google Scholar]
  • 39.Panigrahi P, Parida S, Nanda NC, et al. A randomized synbiotic trial to prevent sepsis among infants in rural India. Nature 2017; 548(7668): 407–12. [DOI] [PubMed] [Google Scholar]
  • 40.Kokai-Kun JF, Bristol JA, Setser J, Schlosser M. Nonclinical Safety Assessment of SYN-004: An Oral beta-lactamase for the Protection of the Gut Microbiome From Disruption by Biliary-Excreted, Intravenously Administered Antibiotics. International journal of toxicology 2015. [DOI] [PubMed] [Google Scholar]
  • 41.de Gunzburg J, Ducher A, Modess C, et al. Targeted adsorption of molecules in the colon with the novel adsorbent-based medicinal product, DAV132: A proof of concept study in healthy subjects. Journal of clinical pharmacology 2015; 55(1): 10–6. [DOI] [PubMed] [Google Scholar]
  • 42.Dellinger RP, Levy MM, Rhodes A, et al. Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012. Critical care medicine 2013; 41(2): 580–637. [DOI] [PubMed] [Google Scholar]
  • 43.Puskarich MA, Trzeciak S, Shapiro NI, et al. Association between timing of antibiotic administration and mortality from septic shock in patients treated with a quantitative resuscitation protocol. Critical care medicine 2011; 39(9): 2066–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Sterling SA, Miller WR, Pryor J, Puskarich MA, Jones AE. The Impact of Timing of Antibiotics on Outcomes in Severe Sepsis and Septic Shock: A Systematic Review and Meta-Analysis. Critical care medicine 2015; 43(9): 1907–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Postma DF, van Werkhoven CH, van Elden LJ, et al. Antibiotic treatment strategies for community-acquired pneumonia in adults. The New England journal of medicine 2015; 372(14): 1312–23. [DOI] [PubMed] [Google Scholar]
  • 46.Sawyer RG, Claridge JA, Nathens AB, et al. Trial of short-course antimicrobial therapy for intraabdominal infection. The New England journal of medicine 2015; 372(21): 1996–2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Pakyz AL, MacDougall C, Oinonen M, Polk RE. Trends in Antibacterial Use in US Academic Health Centers 2002 to 2006. Archives of Internal Medicine 2008; 168(20): 2254–60. [DOI] [PubMed] [Google Scholar]
  • 48.Beus JM, Ross R, Metjian TA, Gerber JS. The Performance of Administrative Data for Measurement of Antibiotic Use Varies Across Antibiotics. In: IDWeek. San Diego, CA, 2015. [Google Scholar]
  • 49.Hecker MT, Aron DC, Patel NP, Lehmann MK, Donskey CJ. Unnecessary use of antimicrobials in hospitalized patients: current patterns of misuse with an emphasis on the antianaerobic spectrum of activity. Arch Intern Med 2003; 163(8): 972–8. [DOI] [PubMed] [Google Scholar]

RESOURCES