A-D indicates Anderson-Darling goodness-of-fit statistic; PR, probabillity.
Positive vs negative net fluid balance (I/O) at hospital discharge, followed up for up to 30 days. The shaded areas indicate 95% Hall-Wellner confidence bands; crosses indicate censoring.
eTable 1. Chronic Kidney Disease and Heart Failure Diagnosis Codes
eTable 2. Diuretics List
eTable 3. Association of Net Fluid Balance (I/O) at Discharge With Different Transformations for 30-Day All-Cause Readmission
eTable 4. Association of Total IV Fluid Volume Administration at Discharge and Readmission (Including Readmission by Category)
eTable 5. Association of Net Fluid Balance (I/O) at Discharge With Readmission in Patients Who Received Early Adequate Fluid Resuscitation for Sepsis
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.
Yoo MS, Zhu S, Lu Y, et al. Association of Positive Fluid Balance at Discharge After Sepsis Management With 30-Day Readmission. JAMA Netw Open. 2021;4(6):e216105. doi:10.1001/jamanetworkopen.2021.6105
For non–critically ill patients hospitalized with sepsis, is there an association between positive fluid balance at the time of discharge and 30-day readmission?
In this cohort study that included 57 032 adults hospitalized with sepsis, no association was found between net fluid balance at the time of discharge and 30-day readmission. Findings may be limited by incomplete capture of intake and output net fluid balance and residual confounding.
No association was found between positive fluid balance and readmission in this large observational study of non–critically ill patients with sepsis, but further investigation is needed.
Although early fluid administration has been shown to lower sepsis mortality, positive fluid balance has been associated with adverse outcomes. Little is known about associations in non–intensive care unit settings, with growing concern about readmission from excess fluid accumulation in patients with sepsis.
To evaluate whether positive fluid balance among non–critically ill patients with sepsis was associated with increased readmission risk, including readmission for heart failure.
Design, Setting, and Participants
This multicenter retrospective cohort study was conducted between January 1, 2012, and December 31, 2017, among 57 032 non–critically ill adults hospitalized for sepsis at 21 hospitals across Northern California. Kaiser Permanente Northern California is an integrated health care system with a community-based population of more than 4.4 million members. Statistical analysis was performed from January 1 to December 31, 2019.
Intake and output net fluid balance (I/O) measured daily and cumulatively at discharge (positive vs negative).
Main Outcomes and Measures
The primary outcome was 30-day readmission. The secondary outcomes were readmission stratified by category and mortality after living discharge.
The cohort included 57 032 patients who were hospitalized for sepsis (28 779 women [50.5%]; mean [SD] age, 73.7 [15.5] years). Compared with patients with positive I/O (40 940 [71.8%]), those with negative I/O (16 092 [28.2%]) were older, with increased comorbidity, acute illness severity, preexisting heart failure or chronic kidney disease, diuretic use, and decreased fluid administration volume. During 30-day follow-up, 8719 patients (15.3%) were readmitted and 3639 patients (6.4%) died. There was no difference in readmission between patients with positive vs negative I/O (HR, 1.00; 95% CI, 0.95-1.05). No association was detected between readmission and I/O using continuous, splined, and quadratic function transformations. Positive I/O was associated with decreased heart failure–related readmission (HR, 0.80 [95% CI, 0.71-0.91]) and increased 30-day mortality (HR, 1.23 [95% CI, 1.15-1.31]).
Conclusions and Relevance
In this large observational study of non–critically ill patients hospitalized with sepsis, there was no association between positive fluid balance at the time of discharge and readmission. However, these findings may have been limited by variable recording and documentation of fluid intake and output; additional studies are needed to examine the association of fluid status with outcomes in patients with sepsis to reduce readmission risk.
Sepsis is a leading cause of hospital death and readmission in the United States.1-6 Prior work has suggested that early intravenous fluid administration during sepsis hospitalization is associated with lower mortality, even among patients considered to be at risk for fluid overload, such as those with chronic kidney disease (CKD) or heart failure.7 Thus, adequate fluid administration during the early phase of sepsis care is recommended in current guidelines.8
At the same time, a growing body of evidence suggests that excess fluid accumulation is harmful for hospitalized patients, with many concerned about positive fluid balance in sepsis being associated with adverse in-hospital and postsepsis sequelae.9-21 Positive fluid balance in critically ill patients has been associated with increased mortality,9-14 and fluid-restrictive strategies have resulted in improved outcomes for this population.15-18 However, positive fluid balance during hospitalization for sepsis is widely recognized as a surrogate measure of illness severity and thus may be less amenable to practice change.19-21
To our knowledge, few studies address fluid practice in patients with sepsis outside of the intensive care unit (ICU). Although a previous study showed that nearly 80% of patients with sepsis do not require ICU level of care,22 little evidence is available to guide practice. In addition, many are concerned about the potential that aggressive fluid resuscitation among patients with sepsis is associated with subsequent rehospitalization.23,24 For example, heart failure is a leading diagnosis among patients with sepsis who are readmitted,25 yet the association of interventions such as diuretics (shown to be beneficial in patients with acute decompensated heart failure26,27) with net fluid balance and readmission risk in patients with sepsis remains poorly understood. To our knowledge, no studies have evaluated the association between fluid balance and readmission, yet it is plausible that conservative fluid management may be associated with considerable postdischarge benefit.
We therefore undertook this study to evaluate whether positive fluid balance among non–critically ill patients with sepsis was associated with increased readmission risk, including readmission for heart failure. We sought to test the association between net fluid balance at discharge and 30-day readmission and additionally to evaluate if readmission characteristics differ by net fluid balance during the preceding hospitalization for sepsis.
We conducted a multicenter retrospective cohort study at Kaiser Permanente Northern California (KPNC) of adults hospitalized with sepsis from January 1, 2012, through December 31, 2017. Kaiser Permanente Northern California is an integrated health care system with a diverse, community-based population of more than 4.4 million members. This study was approved by the KPNC institutional review board with a waiver of informed consent because data were deidentified. All variables of interest were collected from KPNC electronic databases based on existing methods that have been validated and used in prior studies of KPNC inpatient data.1,5-7,28 This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
We identified patients admitted through the emergency department who met Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) criteria29 within 6 hours of presentation. Patients were excluded if they were younger than 18 years at the time of admission, spent any time in the ICU, died during hospitalization, were transferred from or discharged to another acute care facility, had less than 1 year of KPNC membership prior to admission or loss of membership less than 30 days after discharge (unless due to death), had length of stay greater than 7 days, initiated comfort care during hospitalization, or were diagnosed with end-stage renal disease, acute liver failure, or hypoalbuminemia within 1 year prior to hospitalization. An index visit was assigned as the first qualifying hospitalization.
Our exposure of interest was cumulative intake and output net fluid balance (I/O) at discharge, calculated from daily flowsheets. Daily flowsheet data obtained from the electronic health record (EHR) measured patients’ total intake and output at every 24-hour interval from midnight until midnight the following day. Net fluid balance was characterized both continuously and dichotomously as positive (>0) vs negative (≤0) at discharge. Electronic health record I/O data were reviewed by clinicians for accuracy of fluids received and documented in a random sample of 120 patient records and were found to be concordant, including confirmation that fluid electronically recorded as being administered in the emergency department was captured in the EHR flowsheets. However, as this study was retrospective, we were unable to confirm the accuracy of actual fluid administration and output with electronic documentation, as no non-EHR I/O record exists. We evaluated net fluid balance distribution at discharge, excluding patients with extreme outlier I/O values, defined as those whose values were less than 0.1% (<−19.0 L) or greater than 99.9% (>20.8 L) at the distribution percentiles. Finally, we compared total fluid intake with total intravenous fluid intake at discharge.
Our primary outcome was 30-day all-cause hospital readmission. Our secondary outcomes were 30-day readmission stratified by category (heart failure, any infection without heart failure, or other) and mortality after living discharge. Readmission categories were identified using International Classification of Diseases, Ninth Revision codes and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes from the principal discharge diagnosis, primary diagnosis, and secondary diagnoses reported for the hospitalization. Patients were followed up from index visit discharge date until death or 30 days after discharge, whichever came first. Reasons for readmission were further validated by clinician record review of a random sample of 386 patients, including 126 of 2520 patients (5.0%) from the “infection without heart failure” category and 260 of 5212 patients (5.0%) from the “other” category.
Demographic variables included age, sex, and race/ethnicity. Vital signs included weight and body mass index. Medical history included Charlson Comorbidity Index,30 heart failure, CKD, and estimated glomerular filtration rate. All vital signs and medical histories were collected within a look-back period up to 1 year prior to the index visit (diagnosis codes for heart failure and CKD are shown in eTable 1 in the Supplement). Other clinical variables collected during the index visit were diuretic use during hospitalization (eTable 2 in the Supplement); comorbid disease burden based on Comorbidity Point Score, version 2 (COPS2)31,32; acute severity of illness based on Laboratory-Based Acute Physiology Score, version 2 (LAPS2)31,32; total intravenous fluid administration; full code status (ie, full resuscitative measures would be provided in the event of a cardiac or respiratory arrest); readmission risk score (an EHR-based score calculated on the morning of discharge)33; and length of stay.
Statistical analysis was performed from January 1 to December 31, 2019. We used descriptive statistics to describe demographic and clinical characteristics of the overall cohort and patients with positive vs negative net fluid balance at discharge. We reported means (with SDs) or medians (with interquartile ranges) for continuous variables. Categorical variables were reported as numbers and proportions. Differences in patient characteristics at baseline between positive vs negative I/O were evaluated using t tests or Wilcoxon rank sum tests for continuous variables and χ2 tests for categorical variables. The correlation between total fluid input and total intravenous fluid administration was assessed using Pearson correlation.
To evaluate the association between I/O and 30-day readmission risk, we conducted bivariate and multivariable competing risk survival analysis as described by Fine and Gray.34,35 We adjusted our models for clinically important and statistically significant confounders, defined as those that were significantly associated with the outcome in bivariate analyses (P < .05). We reported the Fine-Gray hazard ratio (HR) with 95% CIs. In addition to using I/O as positive vs negative, we tested our exposure as a continuous variable and with splined and quadratic transformations. We performed these transformations to evaluate the association between I/O and readmission that could be more nuanced than using only continuous and binary I/O representations.
To test the hypothesis that negative I/O at discharge primarily reduces the likelihood of heart failure–related readmission, we performed multivariable competing risk survival analysis, with death and reasons for readmission as competing risks (ie, dependent variable = readmission due to heart failure; competing risks = death and readmission due to infection or other reasons). We adjusted for the same covariates identified in the final model that were used for the primary outcome. We then stratified our cohort by patients with or without a history of heart failure or CKD and applied all of the aforementioned analyses to the 2 subgroups.
Finally, we studied mortality at 30 days after discharge using Kaplan-Meier survival analysis to compare the survival rates between the 2 groups during follow-up. Kaplan-Meier plots and log-rank test results were presented. Cox proportional hazards regression models were used to evaluate the association between net fluid balance and 30-day mortality, adjusting for risk factors identified in the main model, with adjusted HRs and 95% CIs reported. All P values were from 2-sided tests and results were deemed statistically significant at P < .05.
We conducted 2 sensitivity analyses to account for variable EHR documentation of I/O and fluid resuscitation in our cohort using the analytic approaches described above. First, we used the total amount of intravenous fluids received during the course of hospitalization for sepsis rather than the net fluid balance as the exposure because the documentation of intravenous fluids administered is linked with direct EHR orders and likely more reliably recorded. Second, we chose to evaluate the association between net fluid balance and outcomes in a subgroup of patients who received at least 2.1 L (approximating 30 mL/kg of intravenous volume resuscitation) within the first 24 hours of hospitalization because the association between net fluid balance and outcomes would be theoretically stronger in this group who received adequate resuscitation.
We identified 57 032 sepsis patients admitted to a non-ICU setting who survived to discharge between January 1, 2012, and December 31, 2017. Table 1 describes the baseline characteristics of our study participants; 40 940 patients (71.8%) had a positive net fluid balance at discharge, and 16 092 patients (28.2%) had a negative net fluid balance at discharge. Overall, there were significant differences between the 2 groups. Compared with patients with positive I/O at discharge, those with negative I/O were sicker as indicated by higher acute (LAPS2) and chronic (COPS2 and Charlson Comorbidity Index) severity of illness (mean [SD] LAPS2, 93.8 [31.8] vs 90.8 [34.1]; P < .001; mean [SD] COPS2, 67.0 [46.3] vs 53.1 [43.4]; P < .001; and mean [SD] Charlson Comorbidity Index, 1.5 [1.5] vs 1.1 [1.3]; P < .001). In addition, compared with patients with positive I/O, patients with negative I/O had higher rates of preexisting heart failure (8925 [55.5%] vs 17 805 [43.5%]; P < .001) and CKD (7626 [47.4%] vs 8951 [21.9%]; P < .001), higher readmission risk score (high-risk, 1442 [9.0%] vs 2516 [6.1%]; P < .001), lower median volumes of intravenous fluid administration (1.7 L [interquartile range, 0.5-3.8 L] vs 4.6 L [interquartile range, 2.6-7.2 L]; P < .001), increased diuretic use (8783 [54.6%] vs 10 074 [24.6%]; P < .001), and longer mean (SD) length of stay (3.7 [1.7] days vs 3.4 [1.7] days; P < .001).
We evaluated net fluid balance distribution at discharge for the overall cohort. Figure 1 shows that the distribution of I/O at discharge is close to a normal distribution but with a higher peak. Pearson correlation comparing total fluid input with total intravenous fluid input at discharge indicated a strongly positive correlation (r = 0.92; P < .001).
Overall, 8719 patients (15.3%) were readmitted and 3639 patients (6.4%) died during 30-day postdischarge follow-up. For the primary outcome of readmission, we found that 2720 of the patients with negative I/O (16.9%) were readmitted compared with 5999 of the patients with positive I/O (14.7%). There was no difference in the 30-day readmission risk between patients with positive vs negative I/O at discharge after adjusting for age, sex, race, body mass index, history of CKD, history of heart failure, diuretic use, COPS2, LAPS2, readmission risk score, and length of stay (HR, 1.00 [95% CI, 0.95-1.05]) (Table 2). No association was detected between I/O at discharge and readmission using I/O as continuous, spline, and quadratic function transformations (eTable 3 in the Supplement).
For readmission by category, positive I/O at discharge compared with negative I/O was associated with lower heart failure–related readmission (HR, 0.80 [95% CI, 0.71-0.91]) and higher infection-related readmission (HR, 1.20 [95% CI, 1.09-1.33]) (Table 2). In stratified analyses accounting for a history of heart failure or CKD, positive I/O was consistently associated with a higher rate of infection-related readmission in all groups. For example, among patients with a history of heart failure or CKD, positive I/O was associated with a lower rate of heart failure–related readmission (HR, 0.68 [95% CI, 0.60-0.77]) compared with negative I/O.
For the secondary outcome of mortality, we found that 3536 patients with positive I/O (8.6%) died and 1254 patients with negative I/O (7.8%) died within 30 days of discharge. Positive I/O at discharge was associated with higher 30-day mortality compared to negative I/O (adjusted HR [aHR], 1.23 [95% CI, 1.15-1.31]). From the Kaplan-Meier plot (Figure 2), the group with negative I/O demonstrated a lower mortality rate compared with the group with positive I/O during the 30-day follow-up (P = .001).
We did not identify a statistically significant association between fluids received and 30-day all-cause readmission in our sensitivity analyses. Among patients who received larger cumulative volumes of intravenous fluids at discharge, for every 1-L increase, there was decreased risk of heart failure–related readmission (aHR, 0.93 [95% CI, 0.90-0.95]) but no association with all-cause readmission (aHR, 1.00 [95% CI, 0.99-1.00]) (eTable 4 in the Supplement). Similarly, among patients who received early adequate fluid resuscitation of at least 2.1 L within 24 hours, there was no association between net positive fluid balance at discharge and all-cause readmission (aHR, 1.02 [95% CI, 0.92-1.13]) or heart failure–related readmission (aHR, 1.18 [95% CI, 0.78-1.79]) (eTable 5 in the Supplement).
Limited evidence exists regarding the association of fluid status with readmission and other outcomes in sepsis survivors, particularly among patients in the medical ward. To our knowledge, this is the first study that addresses a common question in the management of patients with sepsis: does net positive fluid balance during hospitalization, a known adverse prognostic factor, increase a patient’s readmission risk? In this large study of non–critically ill patients with sepsis, we did not identify an association between positive fluid balance at discharge and 30-day readmission.
The answer to this question could have significant implications for enhancing recovery after sepsis. Numerous studies highlight the long-term adverse sequelae of sepsis, including increased readmission risk for infection and cardiovascular disease and exacerbation of chronic conditions.24,25 Thus, identifying actionable steps for mitigating the deleterious consequences of sepsis, including rehospitalization and mortality, are needed. Given the focus on fluid administration in the early part of sepsis care, we sought to identify whether there was an association between overall fluid balance at discharge and subsequent hospitalization. We focused on relatively stable patients treated in hospital wards rather than those with critical illness for whom fluid balance more likely reflects the severity of organ dysfunction and hemodynamic perturbations. We hypothesized that patients with positive fluid balance would have increased readmission risk for heart failure and that, if this were true, future studies examining targeted diuresis might present an avenue for improving postsepsis outcomes.
Our study has some strengths, including our large multicenter sample, use of validated algorithms to identify patients with sepsis, rigorous predefined exclusion criteria, and analysis incorporating granular data drawn from a comprehensive EHR system. We used existing algorithms to identify a cohort of contemporary patients who met Sepsis-3 criteria within 6 hours of presentation. We took careful steps to exclude patients who were critically ill, receiving comfort care, or died during hospitalization. We also excluded patients whose fluid balance might be more challenging to quantify (eg, patients with end-stage renal disease or dialysis requirement who had oliguria or anuria) or whose fluid administration would differ clinically (eg, patients with hypoalbuminemia contributing to edema formation and patients with acute liver failure with low systemic vascular resistance and intravascular depletion). We also excluded patients whose hospitalization was longer than 7 days, as a previous study showed that the mean length of stay for sepsis was 4 days.7 Finally, we investigated readmission characteristics and stratified patients based on the presence of underlying heart failure or CKD in our primary, secondary, and sensitivity analyses. Our study findings, drawn from multiple community-based hospitals throughout Northern California, are generalizable to the larger community and establish equipoise for clinical trials.
Our study also has some limitations. Despite using detailed EHR data with excellent longitudinal follow-up, we found no association between net fluid balance and readmission. Our findings were very likely to have been impacted by difficulties in ascertaining true, reliable measures of fluid intake and output during the course of a hospitalization, which is a key limitation of our study. Accurate measurement of fluid balance in clinical practice, even in a monitored hospital setting, is problematic for several reasons: difficulty measuring strict oral fluid intake volume, errors of documentation and measurement for urine and stool output, and increased gaps in documentation for healthier, ambulatory patients without fluid restriction orders who do not require urine collection instruments. Thus, it remains likely that electronically documented I/O data fail to represent the full extent of daily fluid intake and output. Although we interpret our findings in light of these challenges, our study helps to establish the need for sepsis fluid management studies that carefully incorporate prospectively collected measures of fluid balance without depending only on EHR documentation.
Another key limitation is that net fluid balance is only a partial indicator of intravascular volume status. In other words, even if we had completely accurate I/O measures, they could not definitively establish whether a patient was intravascularly volume depleted or intravascularly volume overloaded. The assessment of intravascular fluid balance is one of the most challenging determinations even when experienced clinicians are at the bedside with a variety of diagnostic tools.36 Thus, even a study in which clinician assessments were used for all patients would incompletely address the issue of fluid balance. In addition, the use of tools to estimate intravascular volume (central venous pressure monitoring, noninvasive cardiac output monitoring, and point-of-care ultrasonographic evaluation) remains infrequent outside of the ICU, thereby limiting the ability to compare net fluid balance against measurable intravascular assessments. Our study was designed to identify whether an association exists and to provide critical preliminary data that could inform future interventional or prospective studies. In fact, large clinical trials evaluating different fluid volumes, types, and infusion rates are currently ongoing to investigate the question of the effect of fluid status on outcomes in patients with early septic shock.37-39 More prospective data are needed in non–critically ill patients who comprise the majority of sepsis hospitalizations.
Additional limitations include potentially limited data on patients transferred out of KPNC, incomplete data for capturing outpatient interventions carried out after discharge that may have been associated with outcomes, and absence of intravenous fluid type in our analysis. The SALT-ED (Saline Against Lactated Ringer’s or Plasma-Lyte in the Emergency Department) trial demonstrated improved renal outcomes in non–critically ill patients administered balanced crystalloids compared with normal saline.40 We are aware of these results but chose to highlight even more basic questions about fluid balance during sepsis hospitalization rather than fluid type in our study.
We found that patients with negative fluid balance at discharge were significantly different from those with positive balance: they were older, with higher comorbid disease burden and increased acute illness severity. They received far less intravenous fluid during hospitalization (median, 1.7 L for negative balance vs 4.6 L for positive balance) and were more likely to receive diuretics (54.6% vs 24.6%). These differences raise concerns about residual confounding impacting our findings. Despite analyzing the subgroup of patients thought to be at increased risk for volume overload (ie, those with heart failure or CKD) and attempting to mitigate differences through statistical analysis in this retrospective study, we did not identify an association between net fluid balance at discharge and readmission. Other causal inference or prospective approaches are warranted.
We found that patients with sepsis with positive fluid balance at discharge, with or without heart failure or CKD, had a higher adjusted risk of infection-related readmission. The reasons for this association are unclear. One possibility is that patients in the positive fluid balance group—who had fewer comorbidities and lower acute illness severity than the negative fluid balance group—were more likely to be discharged earlier (indeed, length of stay was shorter), resulting in inadequate management of infection during the index hospitalization.
Net positive fluid balance in non–critically ill patients hospitalized with sepsis was not associated with increased readmission, although our study may have been limited by incomplete capture of I/O and residual confounding. Further studies are needed to examine the association of fluid status with outcomes in patients with sepsis to reduce the risk of readmission.
Accepted for Publication: February 24, 2021.
Published: June 4, 2021. doi:10.1001/jamanetworkopen.2021.6105
Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License. © 2021 Yoo MS et al. JAMA Network Open.
Corresponding Author: Michael S. Yoo, MD, Department of Hospital Medicine, Kaiser Permanente Santa Rosa Medical Center, 401 Bicentennial Way, Santa Rosa, CA 95403 (firstname.lastname@example.org).
Author Contributions: Drs Yoo and Liu had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Yoo, Zhu, Hammer, Nemazie, Ananias, Reed, Cheung, Liu.
Acquisition, analysis, or interpretation of data: Yoo, Zhu, Lu, Greene, Hammer, Iberti, Nemazie, McCarthy, O’Malley, Young, Reed, Martinez, Cheung, Liu.
Drafting of the manuscript: Yoo, Liu.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Yoo, Zhu, Liu.
Obtained funding: Yoo.
Administrative, technical, or material support: Yoo, Zhu, Lu, Greene, Iberti, Reed, Liu.
Supervision: Yoo, Liu.
Conflict of Interest Disclosures: Dr Liu reported receiving a grant from the National Institutes of Health/National Institute of General Medical Sciences during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was funded by a grant from the Northern California Kaiser Permanente Community Benefit Program. Dr Liu was also funded by grant R35GM128672 from the National Institutes of Health/National Institute of General Medical Sciences.
Role of the Funder/Sponsor: The funding source 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.
Create a personal account or sign in to: