Hospitalization Type and Subsequent Severe Sepsis

Abstract

Rationale: Hospitalization is associated with microbiome perturbation (dysbiosis), and this perturbation is more severe in patients treated with antimicrobials.

Objectives: To evaluate whether hospitalizations known to be associated with periods of microbiome perturbation are associated with increased risk of severe sepsis after hospital discharge.

Methods: We studied participants in the U.S. Health and Retirement Study with linked Medicare claims (1998-2010). We measured whether three hospitalization types associated with increasing severity of probable dysbiosis (non-infection-related hospitalization, infection-related hospitalization, and hospitalization with Clostridium difficile infection [CDI]) were associated with increasing risk for severe sepsis in the 90 days after hospital discharge. We used two study designs: the first was a longitudinal design with between-person comparisons and the second was a self-controlled case series design using within-person comparison.

Measurements and main results: We identified 43,095 hospitalizations among 10,996 Health and Retirement Study-Medicare participants. In the 90 days following non-infection-related hospitalization, infection-related hospitalization, and hospitalization with CDI, adjusted probabilities of subsequent admission for severe sepsis were 4.1% (95% confidence interval [CI], 3.8-4.4%), 7.1% (95% CI, 6.6-7.6%), and 10.7% (95% CI, 7.7-13.8%), respectively. The incidence rate ratio (IRR) of severe sepsis was 3.3-fold greater during the 90 days after hospitalizations than during other observation periods. The IRR was 30% greater after an infection-related hospitalization versus a non-infection-related hospitalization. The IRR was 70% greater after a hospitalization with CDI than an infection-related hospitalization without CDI.

Conclusions: There is a strong dose-response relationship between events known to result in dysbiosis and subsequent severe sepsis hospitalization that is not present for rehospitalization for nonsepsis diagnoses.

Keywords: dysbiosis; humans; microbiota; patient readmission; self-controlled case series.

Figure 1.

Conceptual diagram of the self-controlled case series analysis. (A) Hypothetical timeline for a patient with three exposures and one severe sepsis hospitalization. (B) Hypothetical shifts in microbiome health associated with the patient’s clinical history. Microbial diversity is in constant flux, with periods of disruption (dysbiosis) corresponding with hospitalization. (C) Classification of baseline and higher risk periods used to calculate the incidence risk ratios for severe sepsis following each of the three exposures. The baseline risk of sepsis increases over time as patients age, which is accounted for in the model. CDI = Clostridium difficile infection.

Figure 2.

The probability of severe sepsis readmission correlates with probable microbiome disruption during index hospitalization. This figure depicts adjusted probabilities of 90-day hospital readmission for severe sepsis (red) and nonsepsis diagnoses (blue) following live discharge from hospitalization without infection, hospitalization with non–Clostridium difficile infection, and hospitalization with C. difficile infection. Probabilities are adjusted for age, sex, race, ethnicity, education, total wealth, limitations of activities and instrumental activities of daily living, and 31 Elixhauser comorbidities. Adjusted probabilities of rehospitalization for nonsepsis diagnoses are not different following the three exposures, whereas rehospitalization for severe sepsis is incrementally greater after exposures with increasing probable microbiome disruption.