Abnormal shock index exposure and clinical outcomes among critically ill patients: A retrospective cohort analysis

Journal Pre-proof Abnormal shock index exposure and clinical outcomes among critically ill patients: A retrospective cohort analysis

Kamal Maheshwari, Brian H. Nathanson, Sibyl H. Munson, Seungyoung Hwang, Halit O. Yapici, Mitali Stevens, Carlos Ruiz, Charles F. Hunley PII:

S0883-9441(19)31738-1

DOI:

https://doi.org/10.1016/j.jcrc.2020.01.024

Reference:

YJCRC 53482

To appear in:

Journal of Critical Care

Please cite this article as: K. Maheshwari, B.H. Nathanson, S.H. Munson, et al., Abnormal shock index exposure and clinical outcomes among critically ill patients: A retrospective cohort analysis, Journal of Critical Care(2020), https://doi.org/10.1016/j.jcrc.2020.01.024

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier.

Journal Pre-proof

Abnormal Shock Index exposure and clinical outcomes among critically ill patients: a retrospective cohort analysis Kamal Maheshwari1,6, MD, MPH; Brian H. Nathanson2, PhD; Sibyl H. Munson3, PhD; Seungyoung Hwang3, MS, MSE; Halit O. Yapici3, MD, MPH, MBA; Mitali Stevens4, PharmD, BCPS; Carlos Ruiz5, MD; Charles F. Hunley5, MD 1

2

of

Department of Outcomes Research, Center for Perioperative Intelligence, Anesthesiology Institute, Cleveland Clinic, Cleveland, OH, USA OptiStatim, LLC, PO Box 60844, Longmeadow, MA, 01116, USA; [email protected]

ro

3

-p

Department of Health Economics and Outcomes Research, Boston Strategic Partners, Inc., 4 Wellington St., Suite 3, Boston, MA, 02118, USA; [email protected], [email protected], [email protected] 4

re

Edwards Lifesciences, One Edwards Way, Irvine, CA, 92614, USA; [email protected] 5

lP

Department of Critical Care Medicine, Orlando Regional Medical Center, 86 W Underwood Suite 101, Orlando, FL, 32806, USA; [email protected], [email protected], 6

Corresponding Author Contact Details:

ur

1

na

Department of General Anesthesiology, Anesthesiology Institute, Cleveland Clinic, 9500 Euclid Avenue, E‑ 31, Cleveland, OH 44195, USA; [email protected]

Jo

Kamal Maheshwari, MD, MPH Department of General Anesthesiology, Cleveland Clinic Main Campus, Mail Code E3, 9500 Euclid Avenue, Cleveland, OH 44195 Phone: 216.445.4311 Fax: 216.444.9247 Email: [email protected] Declaration of Interest: Dr. Maheshwari and Dr. Hunley work as consultants for Edwards Lifesciences. Dr. Munson works as a consultant and Mr. Hwang and Dr. Yapici as employees, for Boston Strategic Partners, Inc. which received funds from Edwards Lifesciences to perform the research. Dr. Nathanson is an employee of OptiStatim, LLC, which received consulting fees from Boston Strategic Partners, Inc. The remaining authors have disclosed that they do not have any declarations. Funding: Funding for this research was provided by Edwards Lifesciences. 1

Journal Pre-proof

Abstract Purpose: To assess the predictive value of a single abnormal shock index (SI ≥0.9; heart rate/systolic blood pressure [SBP]) value for mortality, and association between cumulative abnormal SI exposure and mortality/morbidity. Materials and Methods: Cohort comprised of adult patients with an intensive care unit (ICU)

of

stay ≥24-hours (years 2010-2018). SI ≥0.9 exposure was evaluated via cumulative minutes or

-p

acute kidney injury (AKI), and myocardial injury.

ro

time-weighted average; SBP ≤100-mmHg was analyzed. Outcomes were in-hospital mortality,

re

Results: 18,197 patients from 82 hospitals were analyzed. Any single SI ≥0.9 within the ICU

lP

predicted mortality with 90.8% sensitivity and 36.8% specificity. Every 0.1-unit increase in maximum-SI during the first 24-hours increased the odds of mortality by 4.8% [95%CI; 2.6-

na

7.0%; p<0.001]. Every 4-hours exposure to SI ≥0.9 increased the odds of death by 5.8% [95%CI;

ur

4.6-7.0%; p<0.001], AKI by 4.3% [95%CI; 3.7-4.9%; p<0.001] and myocardial injury by 2.1% [95%CI; 1.2-3.1%; p<0.001]. ≥2-hours exposure to SBP ≤100-mmHg was significantly

Jo

associated with mortality.

Conclusions: A single SI reading ≥0.9 is a poor predictor of mortality; cumulative exposure is associated with greater risk of mortality/morbidity. Cumulative exposure to SBP ≤100-mmHg is clinically comparable in predicting mortality. Dynamic interactions between hemodynamic variables need further evaluation among critically ill patients. Keywords: Shock, hypotension, hemodynamic monitoring, mortality, acute kidney injury, myocardial ischemia

2

Journal Pre-proof

Introduction* Over 5.7 million patients are admitted to US intensive care units (ICUs) annually, resulting in roughly $82 billion in healthcare costs.[1–4] At least one-third of critically ill patients experience shock, which is either a cause or a consequence of inadequate organ perfusion.[5,6] The severity and duration of hypotension, the main component of shock, is a

of

major risk factor for mortality, acute kidney injury (AKI) and myocardial injury.[7,8] Clinicians widely use blood pressure-based hemodynamic management to ensure adequate perfusion to

ro

vital organs. For example, titrating therapy to a mean arterial pressure (MAP) >65 mmHg is

-p

recommended by Surviving Sepsis Campaign guidelines.[9] The role of other related variables

re

such as the Shock Index [SI] in clinical decision making is unclear.[10,11]

lP

The Shock Index (SI), a ratio of HR to SBP, was first introduced in 1967 as a simple and effective way to measure the degree of hypovolemia in shock states.[12] SI is primarily used in

na

the Emergency Department (ED) to assess disease severity, identify high-risk patients, and

ur

predict mortality.[13,14] Some clinicians use abnormal SI as a trigger to initiate goal-directed

Jo

therapy to optimize perfusion and fluid management.[15,16] However, single SI readings have low specificity when predicting poor outcomes,[16,17] and the association between cumulative exposure to abnormal SI and clinical outcomes is unknown. Thus, the objectives of this study were to determine: (I) the utility of a single abnormal SI in predicting mortality within the general ICU population; (II) the association between cumulative exposure of SI ≥0.9 with mortality and morbidity; and (III) the association between cumulative SBP ≤100-mmHg exposure and mortality, and its comparison with cumulative SI ≥0.9 exposure. *

Abbreviations: Shock Index (SI); Intensive Care Unit (ICU); systolic blood pressure (SBP), acute kidney injury (AKI); heart rate (HR); Emergency Department (ED); length-of-stay (LOS); serum creatinine (SCr); time-weighted average of shock index (TWA-SI); Acute Physiology and Chronic Health Evaluation (APACHE); generalized linear models (GLM); area under the receiver operating characteristics (AUROC); positive predictive value (PPV); negative predictive value (NPV); Bayesian Information Criterion (BIC); odds ratio (OR)

3

Journal Pre-proof

Materials and Methods Study Design and Population We conducted a retrospective analysis using the Cerner Health Facts® electronic health records database (Kansas City, MO) which includes clinical and administrative data from US hospitals. A predefined statistical analysis plan was submitted and determined exempt from

of

review in advance by the Western Institutional Review Board (Puyallup, WA).

ro

The study included patients ≥18 years-of-age whose records included HR and SBP

-p

measurements, and who were admitted and discharged from the ICU between January 1, 2010,

re

and June 30, 2018. The first hospital visit containing an ICU admission was considered for

lP

patients with multiple qualifying hospitalizations.

Patients were excluded when they lacked ≥6 months of database history before hospital

na

admission; had ICU length-of-stay (LOS) of <24 hours, or hospital stay of >180 days; had

ur

admission to a neurology ICU; lacked a baseline serum creatinine (SCr) measurement within 6 months before and at least one measurement during the ICU stay; had a history of AKI or

Jo

myocardial injury within 6 months before hospital admission (Supplemental Table 1); received dialysis within 6 months before hospital admission through the first 6-hours within the ICU (time from which outcomes were analyzed); had AKI or myocardial injury during index visit prior to ICU admission (assessed by SCr and troponin); received extracorporeal membrane oxygenation between hospital admission and ICU admission +6-hours; had a pregnancy-related admission, spinal trauma, brain injury or 3rd degree burns over ≥20% of body surface (Supplemental Table 2); <5 valid SI readings per each 24-hours (ICU admission to discharge, or day 7, whichever is shorter); had >1 4-hour gap between SI readings; or were missing age, sex, diagnosis codes, or 4

Journal Pre-proof medication records. The exclusion criteria were based on a previous study [16], and were made more conservative via reducing the number and duration of allowed gaps between measurements). We defined admission time as the time of the first laboratory test or medication order from the ICU. Discharge times were used when available; when missing, order location was used

of

to estimate discharge time.[8]

ro

Prognostic analyses included the following subgroups: patients with sepsis diagnosis

re

-p

(Supplemental Table 3) and patients who came from the operating room.

lP

Exposure

Abnormal SI and SBP exposure were determined from ICU admission through the first of

na

ICU discharge or 7 days, whichever was sooner. We defined cumulative exposure to abnormal

ur

SI in two ways: (I) Cumulative time measured in minutes during which SI was >0.9; (II) Timeweighted average of SI (TWA-SI) >SI threshold of 0.9. Sensitivity analyses were performed

Jo

using the SI >0.7 threshold for both cumulative minutes and TWA-SI exposures. The abnormal SI thresholds for the primary and sensitivity analyses (≥0.9 and ≥0.7) were based on published literature.[18] SI readings, calculated as HR/SBP from HR and SBP values within the database, were excluded if SBP ≤20-mmHg or ≥300-mmHg (0.12% readings) or HR <30-bpm or >200bpm (0.03%). A median of 12.7 SI readings ([25th, 75th]: 9.3, 17.0) were available per patient per ICU day. For abnormal SBP exposure, cumulative time was measured in minutes during which SBP was <100-mmHg.

5

Journal Pre-proof Outcomes The primary outcome was in-hospital mortality. The secondary outcomes were AKI and myocardial injury, which were determined from ICU admission +6-hours through ICU discharge. AKI was defined as ≥Stage-1 according to the Kidney Disease Global Outcomes 2012

of

guidelines based on SCr readings (an increase of ≥0.3 mg/dl or increase to 1.5- to 2-fold from

ro

baseline [the lowest reading within 6 months prior, and closest to ICU admission] and with respect to SCr values within 48-hours).[19] Myocardial injury was defined by ≥1 elevated

-p

troponin value >0.03-ng/ml of Troponin I or T, or “Troponin” before the onset of AKI.[8,20]

re

Myocardial injury was not evaluated past the date upon which AKI was identified to avoid the

lP

detection of myocardial injury within an AKI setting.

na

Two exploratory outcomes were evaluated: ICU LOS and total hospital LOS (total number of in-patient days) for survivors. ICU LOS was calculated from the time of a first

ur

laboratory test or medication order documenting an ICU setting through a discharge entry or the

Jo

last time for a test or medication order from an ICU setting. ICU LOS was not evaluated for the TWA-SI exposure since the time spent in the ICU is considered within the TWA calculation.

Statistical Analyses Baseline patient characteristics were summarized via counts and percentages for binary or categorical variables and with means and standard deviations, or medians and interquartile ranges for continuous variables. For univariate inferences, the chi-square test or t-test was used,

6

Journal Pre-proof as appropriate. Multivariable logistic regression quantified the relationship between abnormal SI (cumulative time and TWA-SI) and the binary outcomes: mortality, AKI, and myocardial injury. The primary goal of the multivariable modeling was effect estimation of SI measures on the outcomes. Given the large sample size, we created models with a large number of clinicallyrelevant confounders rather than opting for a limited, parse model.[21,22] Covariates utilized for models are listed in Table 1 and include Acute Physiology and Chronic Health Evaluation

ro

of

(APACHE) III score to adjust for patient acuity, and Charlson comorbidities.[23] We modeled continuous outcomes with Generalized Linear Models (GLM) and assumed

-p

a Gamma distribution and logarithmic link function to account for the data being skewed and

re

strictly positive. Both the logistic regression and GLM models used robust standard errors that

lP

adjusted for potential within-cluster correlations among patients treated within the same

na

hospital.[24]

We assessed the need for restricted cubic splines by plotting deciles and ventiles of the

ur

abnormal SI exposure variable versus the mean proportion of the outcome for each exposure and

Jo

outcome. Since no non-linear trend was evident, the exposures were modeled as linear predictors. We also assessed whether certain covariates (i.e., age, gender, sepsis status) were effect modifiers with SI exposure. The area under the receiver operating characteristics (AUROC), positive predictive value (PPV) and negative predictive value (NPV) curves for SI ≥0.9 were determined for the primary outcome of mortality within the overall ICU population and both subgroups. Exposure to SBP ≤100-mmHg (cumulative time) was evaluated using the same statistical methods outlined above. We used the Bayesian Information Criterion (BIC) to compare the value of SI and SBP in

7

Journal Pre-proof predicting mortality by keeping all other covariates constant. AUROC values were also calculated to assess the model fit for primary and secondary outcomes. We plotted the regression model results as marginal probabilities of the outcome to facilitate interpretation of the results. All statistical analyses were performed using Stata/MP 15.1

Jo

ur

na

lP

re

-p

ro

of

for Windows (StataCorp, College Station, TX).

8

Journal Pre-proof

Results Study Cohort and Patient Characteristics 18,197 patients from 82 US hospitals met all inclusion criteria (Figure 1). The mean (SD) age of the patient population was 62.7 (16.4), the APACHE III score was 52 (19), and 50.5% of patients were female.

of

The unadjusted in-hospital mortality was 7.8% (n=1,418). AKI and myocardial injury

ro

were experienced by 20.7% (n=3,771) and 6.8% (n=1,237) of the patients, respectively. Mean (SD) LOS for survivors was 4.2 (5.3) days (ICU) and 9.1 (10.2) days (hospital). Cohort

re

-p

characteristics for all covariates included in the regression models appear in Table 1.

lP

Predictive Validity of Abnormal SI ≥0.9 for Mortality

na

We analyzed the ability of any ICU SI reading ≥0.9 threshold to predict patient mortality.

ur

The sensitivity was 90.8%, the specificity was 36.8%, the PPV was 10.8% and the NPV was

Jo

97.9%. A summary of AUROC values is provided in Supplemental Table 4.

Association of First and Maximum 24-hour ICU SI Readings with Mortality The association between the first recorded ICU SI reading and in-hospital mortality was not statistically significant: odds increased by 2.6% [95% CI (0.0, 5.3%); p=0.06] for every 0.1 unit increase in SI (Supplemental Table 5). However, exposure to the maximum abnormal SI recorded within the first 24-hours in the ICU significantly increased the odds of mortality; for

9

Journal Pre-proof every 0.1 unit increase in the maximum SI reading, the odds of death increased by 4.8% [95% CI (2.6, 7.0%); p<0.001] (Supplemental Table 5).

Association of Outcomes with Abnormal SI ≥0.9 Exposure Measured by Cumulative Minutes

of

The adjusted odds ratios (OR) are summarized in Supplemental Table 6 and adjusted

ro

marginal effects are found in Table 2. Longer duration of SI ≥0.9 exposure was significantly and

-p

positively correlated with in-hospital mortality, AKI, myocardial injury, and the ICU and hospital LOS. The models had good discrimination with AUROC ≥0.80 for mortality, AKI, and

re

myocardial injury. Every 4-hour exposure increased the odds of death by 5.8% [95% CI (4.6,

lP

7.0%); p<0.001], AKI by 4.3% [95% CI (3.7, 4.9%); p<0.001] and myocardial injury by 2.1%

na

[95% CI (1.2, 3.1%); p<0.001]. Furthermore, every 4-hour increase in exposure increased the mean ICU LOS by 3.7% [95% CI (3.4, 4.0%); p<0.001] and hospital LOS by 2.3% [95% CI

ur

(2.0, 2.6%); p<0.001]. The predicted marginal probabilities of in-hospital mortality, AKI, and

Jo

myocardial injury are shown in Figure 2.

Association of Outcomes with Abnormal SI ≥0.9 Exposure Measured by TWA-SI Adjusted ORs and marginal effects are summarized in Supplemental Table 7 and Table 3, respectively. The exposure of TWA-SI ≥0.9 was significantly associated with in-hospital mortality (see Figure 2). For every 0.03 unit increase in TWA-SI, the odds of in-hospital mortality increased by 24% [95% CI (19.2, 29.0%); p<0.001] and the odds of AKI increased by 6.8% [95% CI (4.5, 9.1%); p<0.001]. The relationship between TWA-SI and hospital LOS was 10

Journal Pre-proof similar: for every 0.03 unit increase in TWA-SI, the average LOS increased by 1.9% [95%CI (0.7, 3.2%); p<0.001]. There was no significant association between exposure to TWA-SI ≥0.9 and myocardial injury: for every 0.03 unit increase in TWA-SI, the odds of myocardial injury increased by 1.3% [95% CI (0.0, 4.4%); p=0.42].

of

In-Hospital Mortality Associated with SBP ≤100-mmHg or SI ≥0.9 Exposure

ro

Exposure of ≥2-hours to SBP ≤100-mmHg was significantly associated with in-hospital mortality. Unlike SI results, the mortality did not significantly increase when exposure to SBP

-p

≤100-mmHg was less than 2-hours (Supplemental Table 8). Overall, associations with in-

re

hospital mortality were comparable for SI ≥0.9 or SBP ≤100-mmHg exposure (Figure 3). We

lP

found large improvements in model fit comparing BIC statistics after adding either SI or SBP to

ur

(Supplemental Table 9).

na

regression models incorporating all other covariates, although SI was statistically superior

Jo

Sensitivity Analysis for Abnormal SI Threshold ≥0.7 The exposures of cumulative minutes and TWA-SI for the SI threshold ≥0.7 generated consistent results with the findings of the primary analysis. A longer duration of exposure (cumulative minutes) to SI ≥0.7 was associated with increased mortality, AKI, myocardial injury, and ICU and hospital LOS. Every 4-hour exposure to SI ≥0.7 increased the odds of death by 3.4% [95% CI (2.6, 4.2%); p<0.001], AKI by 4.0% [95% CI (3.5, 4.5%); p<0.001], myocardial injury by 2.4% [95% CI (1.7, 3.0%); p<0.001]; and, for survivors, mean ICU LOS by 3.8% [95% CI (3.6, 4.1%); p<0.001] and mean hospital LOS by 2.4% [95% CI (2.2, 2.6%); 11

Journal Pre-proof p<0.001]. Exposure measured by TWA-SI for SI ≥0.7 was associated with a significant increase in rates of mortality, AKI and hospital LOS, but not significantly associated with myocardial injury. Every 0.03 unit increase in TWA-SI ≥0.7 increased the odds of death by 16.1% [95% CI (13.8, 18.5%); p<0.001], AKI by 4.2% [95% CI (2.9, 5.5%); p<0.001], myocardial injury by 0.3% [95% CI (0, 2.3%); p=0.741], and mean hospital LOS by 1.6% [95% CI (0.9, 2.4%);

Jo

ur

na

lP

re

-p

ro

of

p<0.001, survivors only].

12

Journal Pre-proof

Discussion Timely identification of poor organ perfusion is a challenge in critically ill patients. This study evaluated how SI, particularly the cumulative time of abnormal SI, is associated with mortality and morbidity, and compared abnormal SI and SBP exposures in predicting in-hospital mortality. We observed the first SI reading in the ICU does not predict mortality well; however, the

of

maximum SI reading during the first 24 hours was significantly associated with increased

ro

mortality but may be impractical to use clinically. While at least one SI ≥0.9 value had a high

-p

NPV in detecting in-hospital mortality (98%), we found low PPV and specificity, which is

re

consistent with other studies (i.e., not all patients with one or more abnormal SI ≥0.9 reading will

lP

die in the hospital).[16,17] Therefore, a single abnormal SI ≥0.9 reading has limited prognostic value for clinicians. Instead, clinicians need to examine cumulative time with abnormal SI and its

na

association with patient outcomes in the ICU setting.

ur

Our results are in contrast to the use of a single SI value as a predictor of morbidity and mortality in the emergency department, which may be explained by the difference in patient

Jo

population and disease pathophysiology. For example, SI is used in the ED to assess hemodynamic instability quickly, and has been studied as a prognostic metric for patients with a variety of conditions such as the probability of hospital admission and increased lactate.[13,25– 34] Bourque et al. showed that standard SI is superior to nine other modified-SI versions in predicting mortality for patients with gastrointestinal bleeding.[35] The evidence for a general ICU population is limited to two small-population studies using a single SI value: Smischney et al. found a significant association between pre-intubation SI and hemodynamic derangement;[36] Trivedi et al. demonstrated that pre-intubation SI ≥0.9 could predict both post13

Journal Pre-proof intubation hypotension and mortality.[37] However, they evaluated mostly short term, intermediary outcomes like postintubation hypotension. In this comprehensive analysis of critically ill patients, single SI has limited ability to predict in-hospital mortality; thus, its use is of uncertain clinical value. Furthermore, we evaluated a novel approach to examining cumulative exposure to abnormal SI. Overall, our study demonstrates that cumulative exposure to abnormal SI is a

of

significantly better predictor of mortality and morbidity than single SI readings in the ICU. We

ro

found cumulative time spent above SI ≥0.9 is associated with worse outcomes: increased

-p

mortality, AKI (stage I and higher), and myocardial injury, and longer ICU and hospital stays.

re

We further observed consistent results using TWA-SI ≥0.9 as the exposure for in-hospital

lP

mortality and AKI but observed no significant association with myocardial injury. While sensitivity analyses using SI ≥0.7 as a threshold yielded similar results, a threshold of SI ≥0.9

na

had a stronger association with outcomes, particularly in-hospital mortality.

ur

We compared the association of either abnormal SI or one threshold measure of hypotension, SBP ≤100-mmHg, with mortality. As expected, we observed a strong association

Jo

between the cumulative time spent with the single SBP ≤100-mmHg threshold tested and mortality. It would be important to examine other SBP thresholds, but this was not the focus of the present study. Our analysis showed, however, that exposure to abnormal SBP, defined by this single absolute threshold, and SI are clinically comparable in predicting in-hospital mortality though the association of SI ≥0.9 exposure with in-hospital mortality was stronger via the BIC analysis. This is to be expected since SI incorporates both the SBP and HR information and thus can capture the decompensated stages of shock as well as cardiac contractility.

14

Journal Pre-proof Nevertheless, the superiority of SI over SBP is not clear, particularly in the ICU setting where clinical decision making is highly restricted for time and other resources. Our results show that the dynamic component of hemodynamic decomposition can be captured by cumulative exposure to abnormal SI. However, to establish a clear superiority of one measure, the interplay between SI and SBP should be investigated in critically ill patients with abnormal cumulative exposure to both of these measures among different types and stages of shock. Given our

of

findings, it appears prudent for clinicians to limit time above SI ≥0.9 and/or time below SBP

ro

≤100-mmHg and future studies should focus on appropriate strategies to achieve these goals. The

-p

correlation between cumulative abnormal SI and SBP exposures in predicting mortality suggests

re

other hemodynamic variables may also have clinical value when they are continuously

lP

monitored and kept under certain thresholds.

Unobserved patient characteristics may bias these results though we did adjust for a large

na

number of potential confounders. Untreated hypotension or tachycardia that persisted despite

ur

treatment, may indicate a worse prognosis. To address this, we adjusted for medication use and other potential comorbidity markers. However, patients in the ICU are often under aggressive

Jo

treatments which may affect HR and SBP in different directions, and our analysis may not be capturing all of these effects. Continuous-time at or above the SI or below the SBP threshold, while intuitive, is subject to two types of bias: the frequency of readings, and the total time for abnormal exposure calculation. Spurious extreme values may distort the results due to gaps between readings that can over-represent times above a given SI threshold (or below an SBP threshold); and patients with longer ICU stays may have a greater propensity for hemodynamic instability. To address this potential bias, we used TWA-SI exposure as it measures both the degree and the duration at/above an SI threshold. Additionally, while we required a minimum of

15

Journal Pre-proof 5 readings per ICU-day and allowed only one gap which exceeds 4 hours between SI/SBP readings, there is still some heterogeneity in the numbers of readings available among patients. We did not differentiate or analyze SBP readings based on the measurement device (i.e., arterial line, non-invasive BP cuff) due to a lack of this information in the database. Myocardial injury analysis results should be interpreted with caution due to lack of universal troponin testing, diverse troponin tests employed among U.S. hospitals, and troponin increases that may be

of

affected by certain medical conditions (i.e., pleural effusion) which may not be fully captured by

ro

our model. AKI was examined as combined stages I and higher; associations restricted to Stage

-p

II or Stage III AKI may differ and were not examined individually. Additional analysis is required to determine association with Stage II or III AKI. This study did not examine outcomes

re

post-discharge; therefore, SI’s association with long-term outcomes are unknown. Finally, our

lP

study examined data from 82 US hospitals and the findings may not be generalizable to patients

Jo

ur

na

treated in hospitals outside the Cerner Health Facts® system.

16

Journal Pre-proof

Conclusions A single abnormal shock index reading, which incorporates blood pressure and heart rate information, has limited predictive ability for in-hospital mortality in critically ill patients. However, prolonged exposure to abnormal shock index is significantly associated with increased mortality and morbidity. Shock index and systolic blood pressure are clinically comparable in predicting mortality (SBP ≤100-mmHg). Therefore, currently clinicians may continue to use

of

simple blood pressure-based variables to guide clinical care. The role of dynamic interaction

Jo

ur

na

lP

re

-p

ro

between hemodynamic variables in management of the critically ill needs further evaluation.

17

Journal Pre-proof

References Society of Critical Care Medicine (SCCM). Critical Care Statistics n.d. https://www.sccm.org/Communications/Critical-Care-Statistics.

[2]

Barrett ML, Smith MW, Elixhauser A, Honigman LS, Pines JM. Utilization of Intensive Care Services. Agency for Healthcare Research and Quality, Rockville, MD; 2011.

[3]

Halpern NA, Pastores SM. Critical care medicine in the United States 2000-2005: an analysis of bed numbers, occupancy rates, payer mix, and costs. Crit Care Med 2010;38:65–71. https://doi.org/10.1097/CCM.0b013e3181b090d0. Wunsch H, Angus DC, Harrison DA, Linde-Zwirble WT, Rowan KM. Comparison of medical admissions to intensive care units in the United States and United Kingdom. Am J Respir Crit Care Med 2011;183:1666–73. https://doi.org/10.1164/rccm.201012-1961OC.

[4]

of

[1]

Sakr Y, Reinhart K, Vincent J-L, Sprung CL, Moreno R, Ranieri VM, et al. Does dopamine administration in shock influence outcome? Results of the Sepsis Occurrence in Acutely Ill Patients (SOAP) Study*. Crit Care Med 2006;34:589–97. https://doi.org/10.1097/01.ccm.0000201896.45809.e3.

[6]

Vincent J-L, De Backer D. Circulatory Shock. N Engl J Med 2013;369:1726–34. https://doi.org/10.1056/nejmra1208943.

[7]

Arlati S, Brenna S, Prencipe L, Marocchi A, Casella GP, Lanzani M, et al. Myocardial necrosis in ICU patients with acute non-cardiac disease: a prospective study. Intensive Care Med 2000;26:31–7.

[8]

Maheshwari K, Nathanson BH, Munson SH, Khangulov V, Stevens M, Badani H, et al. The relationship between ICU hypotension and in-hospital mortality and morbidity in septic patients. Intensive Care Med 2018;44:857–67. https://doi.org/10.1007/s00134-0185218-5.

[9]

Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) Consensus Definitions for Sepsis and Septic Shock. JAMA 2016;315:801–10. https://doi.org/10.1001/jama.2016.0287.

Jo

ur

na

lP

re

-p

ro

[5]

[10] Smischney NJ, Seisa MO, Heise KJ, Schroeder DR, Weister TJ, Diedrich DA. Elevated Modified Shock Index Within 24 Hours of ICU Admission Is an Early Indicator of Mortality in the Critically Ill. J Intensive Care Med 2018;33:582–8. https://doi.org/10.1177/0885066616679606. [11] Koch E, Lovett S, Nghiem T, Riggs RA, Rech MA. Shock index in the emergency department: utility and limitations. Open Access Emerg Med 2019;11:179–99. https://doi.org/10.2147/oaem.S178358. [12] Allgower M, Burri C. ["Shock index"]. Dtsch Med Wochenschr 1967;92:1947–50. https://doi.org/10.1055/s-0028-1106070. [13] Berger T, Green J, Horeczko T, Hagar Y, Garg N, Suarez A, et al. Shock index and early recognition of sepsis in the emergency department: pilot study. West J Emerg Med 18

Journal Pre-proof 2013;14:168–74. https://doi.org/10.5811/westjem.2012.8.11546. [14] Montoya KF, Charry JD, Calle-Toro JS, Núñez LR, Poveda G. Shock index as a mortality predictor in patients with acute polytrauma. J Acute Dis 2015;4:202–4. https://doi.org/https://doi.org/10.1016/j.joad.2015.04.006. [15] Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, et al. Early GoalDirected Therapy in the Treatment of Severe Sepsis and Septic Shock. N Engl J Med 2001;345:1368–77. https://doi.org/10.1056/NEJMoa010307.

of

[16] Cecconi M, Hernandez G, Dunser M, Antonelli M, Baker T, Bakker J, et al. Fluid administration for acute circulatory dysfunction using basic monitoring: narrative review and expert panel recommendations from an ESICM task force. Intensive Care Med 2019;45:21–32. https://doi.org/10.1007/s00134-018-5415-2.

-p

ro

[17] Torabi M, Mirafzal A, Rastegari A, Sadeghkhani N. Association of triage time Shock Index, Modified Shock Index, and Age Shock Index with mortality in Emergency Severity Index level 2 patients. Am J Emerg Med 2016;34:63–8. https://doi.org/10.1016/j.ajem.2015.09.014.

re

[18] Rady MY, Nightingale P, Little RA, Edwards JD. Shock index: a re-evaluation in acute circulatory failure. Resuscitation 1992;23:227–34. https://doi.org/10.1016/03009572(92)90006-x.

lP

[19] Kellum JA, Lameire N, Co-Chairs WG. Kidney Disease: Improving Global Outcomes (KDIGO). KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney Int Suppl 2012;2:1–138. https://doi.org/10.1038/kisup.2012.6.

ur

na

[20] Salmasi V, Maheshwari K, Yang D, Mascha EJ, Singh A, Sessler DI, et al. Relationship between Intraoperative Hypotension, Defined by Either Reduction from Baseline or Absolute Thresholds, and Acute Kidney and Myocardial Injury after Noncardiac Surgery: A Retrospective Cohort Analysis. Anesthesiology 2017;126:47–65. https://doi.org/10.1097/aln.0000000000001432.

Jo

[21] Harrell Jr FE. Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis. Springer; 2015. [22] Rogers W. Regression standard errors in clustered samples. Stata Tech Bull 1994;3. [23] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–83. [24] Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics 2000;56:645–6. [25] Balhara KS, Hsieh Y-H, Hamade B, Circh R, Kelen GD, Bayram JD %J EMJ. Clinical metrics in emergency medicine: the shock index and the probability of hospital admission and inpatient mortality. Emerg Med J 2017;34:89–94. [26] Sloan EP, Koenigsberg M, Clark JM, Weir WB, Philbin N. Shock index and prediction of traumatic hemorrhagic shock 28-day mortality: data from the DCLHb resuscitation 19

Journal Pre-proof clinical trials. West J Emerg Med 2014;15:795–802. https://doi.org/10.5811/westjem.2014.7.21304. [27] Mutschler M, Nienaber U, Münzberg M, Wölfl C, Schoechl H, Paffrath T, et al. The Shock Index revisited - a fast guide to transfusion requirement? A retrospective analysis on 21,853 patients derived from the TraumaRegister DGU. Crit Care 2013;17:R172– R172. https://doi.org/10.1186/cc12851. [28] Wira CR, Francis MW, Bhat S, Ehrman R, Conner D, Siegel M. The shock index as a predictor of vasopressor use in emergency department patients with severe sepsis. West J Emerg Med 2014;15:60–6. https://doi.org/10.5811/westjem.2013.7.18472.

of

[29] Birkhahn RH, Gaeta TJ, Bei R, Bove JJ. Shock index in the first trimester of pregnancy and its relationship to ruptured ectopic pregnancy. Acad Emerg Med 2002;9:115–9.

ro

[30] Onah HE, Oguanuo TC, Mgbor SO. An evaluation of the shock index in predicting ruptured ectopic pregnancy. J Obs Gynaecol 2006;26:445–7. https://doi.org/10.1080/01443610600747314.

re

-p

[31] Huang B, Yang Y, Zhu J, Liang Y, Tan H, Yu L, et al. Usefulness of the admission shock index for predicting short-term outcomes in patients with ST-segment elevation myocardial infarction. Am J Cardiol 2014;114:1315–21. https://doi.org/10.1016/j.amjcard.2014.07.062.

lP

[32] Kilic T, Ermis H, Gulbas G, Kaya O, Aytemur ZA, Inceoglu F, et al. Prognostic role of the simplified pulmonary embolism severity index and shock index in pulmonary embolism. Pol Arch Med Wewn 2014;124:678–87.

ur

na

[33] Sohn CH, Kim WY, Kim SR, Seo DW, Ryoo SM, Lee YS, et al. An increase in initial shock index is associated with the requirement for massive transfusion in emergency department patients with primary postpartum hemorrhage. Shock 2013;40:101–5. https://doi.org/10.1097/SHK.0b013e31829b1778.

Jo

[34] Nakasone Y, Ikeda O, Yamashita Y, Kudoh K, Shigematsu Y, Harada K. Shock index correlates with extravasation on angiographs of gastrointestinal hemorrhage: a logistics regression analysis. Cardiovasc Interv Radiol 2007;30:861–5. https://doi.org/10.1007/s00270-007-9131-5. [35] Bourque JS-C, Cliche J, Chauny J, Daoust R, Paquet J, Piette É. Accuracy of the shock index and various modified shock indexes to predict early mortality in patients suffering from gastrointestinal haemorrhage. Crit Care 2013;17:P219. [36] Smischney NJ, Seisa MO, Heise KJ, Wiegand RA, Busack KD, Deangelis JL, et al. Predictors of hemodynamic derangement during intubation in the critically ill: A nested case-control study of hemodynamic management-Part II. J Crit Care 2018;44:179–84. https://doi.org/10.1016/j.jcrc.2017.10.018. [37] Trivedi S, Demirci O, Arteaga G, Kashyap R, Smischney NJ. Evaluation of preintubation shock index and modified shock index as predictors of postintubation hypotension and other short-term outcomes. J Crit Care 2015;30:861.e1-7. https://doi.org/10.1016/j.jcrc.2015.04.013. 20

Jo

ur

na

lP

re

-p

ro

of

Journal Pre-proof

21

Journal Pre-proof

Figure Legends

Patient attrition diagram. Abbreviations: AKI, Acute Kidney Injury; ECMO, Extra-corporeal Membrane Oxygenation; ICU, Intensive Care Unit; LOS, Lengthof-Stay; SCr, Serum Creatinine; SI, Shock Index

Figure 2

Predicted mortality and marginal probabilities of acute kidney injury and myocardial injury for exposure to Shock Index ≥0.9. Exposure was measured by cumulative minutes and time-weighted average of Shock Index ≥0.9. Predicted probability of outcomes from cumulative minutes with Shock Index ≥0.9 threshold are shown for mortality (panel 2a), acute kidney injury (2c), and myocardial injury (2d), in hours, and from time-weighted average of Shock Index ≥0.9 for in-hospital mortality in panel 2b.

Figure 3

Adjusted probability of in-hospital mortality for exposure to Shock Index ≥0.9 and systolic blood pressure ≤100-mmHg. Exposures were measured by cumulative minutes. 95% confidence intervals are shown as brackets surrounding each adjusted probability value. Abbreviations: SBP, systolic blood pressure

Jo

ur

na

lP

re

-p

ro

of

Figure 1

22

Journal Pre-proof Supplemental Digital Content Supplemental Table 1. Medical codes for acute kidney injury and myocardial injury diagnoses Supplemental Table 2. Medical codes to define cohort exclusion criteria Supplemental Table 3. Sepsis Codes

Jo

ur

na

lP

re

-p

ro

of

Supplemental Table 4. Results of predictive statistics for in-hospital mortality Supplemental Table 5. In-hospital mortality multivariable analysis results for the first ICU SI reading and the maximum SI reading within the first 24 hours in the ICU Supplemental Table 6: Adjusted results of all outcomes for exposure cumulative minutes SI ≥0.9 Supplemental Table 7. Multivariable adjusted results for SI ≥0.9 for TWI-SI exposure Supplemental Table 8. Adjusted odds ratios with the outcome of in-hospital mortality Supplemental Table 9. Bayesian Information Criterion (BIC) comparisons

23

Journal Pre-proof

Author Contributions Kamal Maheswari: Conceptualization, Methodology, Writing-Review and Editing. Brian Nathanson: Methodology, Software, Validation, Formal analysis, Investigation, WritingOriginal Draft and Writing –Review and Editing, Visualization. Sibyl Munson: Methodology, Writing-Original Draft and Writing –Review and Editing, Supervision,

of

Project Management. Seungyoung Hwang: Methodology, Software, Investigation, Data Curation, Writing-Review and Editing. Halit Yapici: Writing-Original Draft and

ro

Writing –Review and Editing. Mitali Stevens: Conceptualization, Writing-Review and

-p

Editing, Supervision, Funding acquisition. Carlos Ruiz: Writing-Review and Editing.

Jo

ur

na

lP

re

Charles Hunley: Conceptualization, Methodology, Writing-Review and Editing.

24

Journal Pre-proof Table 1. Comprehensive list of potentially confounding variables for SI threshold groups SI <0.9 and SI ≥0.9.

Mortality n (%)

Mortality n (%)

African American Caucasian Other Unknown

Agea,*

Mean ± SD

Admission typea

Emergency

No N= 6,173

P Value

Yes N= 1,287

No N= 10,606

P Value

67 (51.1%)

3,226 (52.3%)

0.80

611 (47.5%)

5101 (48.1 %)

0.67

19 (14.5%) 97 (74.0%) 13 (9.9%) 2 (1.5%) 73.9 ± 12.3

752 (12.2%) 4,828 (78.2%) 511 (8.3%) 82 (1.3%) 63.9 ± 15.7

147 (11.4%) 981 (76.2%) 138 (10.7%) 21 (1.6%) 68.3 ± 14.8

1262 (11.9 %) 8247 (77.8%) 931 (8.8%) 166 (1.6%) 61.2 ± 16.8

114 (87.0%)

3860 (62.5%)

1039 (80.7%)

7026 (66.2%)

126 (9.8%) 74 (5.7%) 48 (3.7%)

2817 (26.6%) 539 (5.1%) 224 (2.1%)

6 (4.6%) 3 (2.3%) 8 (6.1%)

lP

Elective Urgent

na

Unknown

of

Racea

Yes N= 131

0.73

ro

Male

1881 (30.5%) 272 (4.4%) 160 (2.6%)

<0.001

<0.001

35 (0.6%)

8 (0.6%)

53 (0.5%)

10 (7.6%)

323 (5.2%)

74 (5.7%)

487 (4.6%)

16 (12.2%)

567 (9.2%)

122 (9.5%)

1023 (9.6%)

20 (15.3%)

835 (13.5%)

205 (15.9%)

1484 (14.0%)

2014

24 (18.3%)

1188 (19.2%)

252 (19.6%)

1934 (18.2%)

2015

19 (14.5%)

977 (15.8%)

191 (14.8%)

1706 (16.1%)

2016

18 (13.7%)

838 (13.6%)

190 (14.8%)

1701 (16.0%)

2017

12 (9.2%)

991 (16.1%)

157 (12.2%)

1541 (14.5%)

11 (8.4%) 20 (15.3%)

419 (6.8%) 1083 (17.5%)

88 (6.8%) 169 (13.1%)

677 (6.4%) 1966 (18.5%)

2013

2018 Midwest

Jo

2012

ur

1 (0.8%)

2010 2011

Census Region (Hospital)a

SI ≥0.9

-p

Gendera

Admission yeara

SI <0.9

Variable type

re

Variable name

25

0.48

0.83

0.14

<0.001

<0.001

0.06

<0.001

Journal Pre-proof Variable name

SI ≥0.9

SI <0.9

Variable type

Mortality n (%) No N= 6,173 2465 (39.9%)

South

32 (24.4%)

1614 (26.1%)

332 (25.8%)

2970 (28.0%)

West

23 (17.6%)

1011 (16.4%)

312 (24.2%)

2320 (21.9%)

General ICU

86 (65.6%)

4410 (71.4%)

793 (61.6%)

7248 (68.3%)

Coronary Care Unit

20 (15.3%)

738 (12.0%)

172 (13.4%)

1179 (11.1%)

Cardiac ICU

4 (3.1%)

395 (6.4%)

111 (8.6%)

892 (8.4%)

Medical ICU

16 (12.2%)

342 (5.5%)

152 (11.8%)

748 (7.1%)

Surgical ICU

5 (3.8%)

288 (4.7%)

59 (4.6%)

539 (5.1%)

Diureticsc

49 (37.4%)

Ace inhibitors/ ARBsc

of

ro

re

-p

0.006

No N= 10,606

P Value

3350 (31.6%)

<0.001

<0.001

439 (34.1%)

1926 (18.2%)

<0.001

36 (27.5%)

929 (15.0%)

<0.001

202 (15.7%)

1523 (14.4%)

0.20

50 (38.2%)

1264 (20.5%)

<0.001

382 (29.7%)

2411 (22.7%)

<0.001

28 (21.4%)

835 (13.5%)

0.01

246 (19.1%)

1495 (14.1%)

<0.001

77 (58.8%)

2568 (41.6%)

<0.001

836 (65.0%)

5913 (55.8%)

<0.001

0 (0.0%) 15 (11.5%)

34 (0.6%) 312 (5.1%)

169 (1.6%)

0.04

2042 (19.3%)

<0.001

Any vasopressorse

2 (1.5%)

46 (0.7%)

160 (1.5%)

0.23

Norepinepherined

18 (13.7%) 6 (4.6%) 2 (1.5%) 1 (0.8%) 9 (6.9%)

102 (1.7%) 284 (4.6%) 31 (0.5%) 25 (0.4%) 90 (1.5%)

1127 (10.6%)

<0.001

1324 (12.5%)

<0.001

Calcium channel blockersc Antibioticsd Starchesd Albumind

Phenylephrined Vasopressind Any inotropese Dopamined

ur

Beta blockersc

lP

860 (13.9%)

Jo

Drugs received

P Value

Yes N= 1,287 474 (36.8%)

na

Northeast

ICU Typeb

Mortality n (%)

Yes N= 131 56 (42.7%)

26

0.39 0.001 0.31 <0.001 0.99 0.11 0.53 <0.001

11 (0.9%) 429 (33.3%) 25 (1.9%) 412 (32.0%) 258 (20.0%) 175 (13.6%) 11 (0.9%) 138 (10.7%)

299 (2.8%) 75 (0.7%) 403 (3.8%)

<0.001 0.56 <0.001

Journal Pre-proof Variable name

SI ≥0.9

SI <0.9

Variable type

Mortality n (%)

APACHE III scored,*

Mean ± SD

Sepsisf Serum Lactateg

No reading available

45.9 ± 17.6

44 (33.6%)

388 (6.3%)

83 (63.4%) 30 (22.9%)

5378 (87.1) 582 (9.4%)

Moderate 5 to <8 mmol/L

0 (0.0%)

Severe ≥8 mmol/L

0.13 0.63 <0.001 <0.001

70.4 ± 20.1

53.2 ± 17.8

<0.001

663 (51.5%)

2238 (21.1%)

<0.001

661 (51.4%) 240 (18.6%)

P Value <0.001 <0.001 0.13

7641 (72.0%) 1754 (16.5%)

266 (20.7%)

1005 (9.5%)

8 (0.1%)

54 (4.2%)

157 (1.5%)

0 (0.0%)

3 (0.0%)

66 (5.1%)

49 (0.5%)

5 (3.8%)

31 (0.5%)

<0.001

67 (5.2%)

180 (1.7%)

<0.001

49 (37.4%)

640 (10.4%)

<0.001

582 (45.2)

2113 (19.9%)

<0.001

87 (66.4%) 10 (7.6%) 18 (13.7%) 4 (3.1%) 12 (9.2%) 108 (82.4%) 92 (70.2%)

3157 (51.1%) 748 (12.1%) 1392 (22.5%) 368 (6.0%) 508 (8.2%) 4824 (78.1%) 4962 (80.4%)

0.006

755 (58.7%) 124 (9.6%) 233 (18.1%) 72 (5.6%) 103 (8.0%)

Operation Prior to ICU Admissionj

15 (11.5%)

ED admission prior to ICU Admissionj

3 (2.3%)

Payera

Medicare Medicaid Commercial Other Unknown

Teaching status (Hospital) a Urban/Rural status (Hospital)a

Yes Urban

ur

Early ventilationi

Jo

Blood transfusionh

202 (3.3%)

No N= 10,606 396 (3.7%) 272 (2.6%) 263 (2.5%)

re

18 (13.7%)

lP

Mild 2 to <5 mmol/L

0.003

Yes N= 1,287 94 (7.3%) 68 (5.3%) 41 (3.2%)

na

Normal <2 mmol/L

64.5 ± 18.5

P Value

of

Milrinoned

Mortality n (%)

ro

Dobutamined

No N= 6,173 66 (1.1%) 33 (0.5%) 11 (0.2%)

-p

Epinepherined

Yes N= 131 5 (3.8%) 2 (1.5%) 0 (0.0%)

<0.001

<0.001

5052 (47.6%) 1454 (13.7%) 2410 (22.7%)

<0.001

724 (6.8%) 966 (9.1%)

0.24

1023 (79.5%)

8267 (77.9%)

0.21

0.004

982 (76.3%)

8464 (79.8%)

0.003

1495 (24.2%)

<0.001

142 (11.0%)

2432 (22.9)

<0.001

214 (3.5%)

0.46

27 (2.1%)

296 (2.8%)

0.15

27

Journal Pre-proof Variable name

SI ≥0.9

SI <0.9

Variable type

Mortality n (%) No N= 6,173

Septic shock

11 (8.4%)

Cardiogenic shock

Mortality n (%) P Value

Yes N= 1,287

No N= 10,606

P Value

83 (1.3%)

<0.001

403 (31.3%)

799 (7.5%)

<0.001

4 (3.1%)

19 (0.3%)

<0.001

73 (5.7%)

200 (1.9%)

<0.001

Hypovolemic shock

0 (0.0%)

40 (0.6%)

0.36

63 (4.9%)

213 (2.0%)

<0.001

Other shock

3 (2.3%)

26 (0.4%)

0.002

63 (4.9%)

200 (1.9%)

<0.001

Blood Loss Anemia

2 (1.5%)

92 (1.5%)

21 (1.6%)

176 (1.7%)

0.94

Fluid and Electrolyte Disorders

76 (58.0%)

1921 (31.1%)

<0.001

869 (67.5%)

4904 (46.2%)

<0.001

Weight Loss

23 (17.6%)

294 (4.8%)

<0.001

338 (26.3%)

1278 (12.0%)

<0.001

Myocardial Infarction

31 (23.7%)

1032 (16.7%)

0.04

214 (16.6%)

1593 (15.0%)

0.13

Congestive Heart Failure

59 (45.0%)

1502 (24.3%)

<0.001

503 (39.1%)

2968 (28.0%)

<0.001

20 (15.3%)

1228 (19.9%)

0.19

292 (22.7%)

2000 (18.9%)

0.001

36 (27.5%)

1235 (20.0%)

0.03

207 (16.1%)

1599 (15.1%)

0.34

12 (9.2%)

283 (4.6%)

0.01

96 (7.5%)

474 (4.5%)

<0.001

COPD

71 (54.2%)

2274 (36.8%)

<0.001

651 (50.6%)

4533 (42.7%)

<0.001

Rheumatoid Arthritis

3 (2.3%)

317 (5.1%)

0.14

81 (6.3%)

598 (5.6%)

0.34

Peptic Ulcer

8 (6.1%)

280 (4.5%)

0.39

78 (6.1%)

562 (5.3%)

0.25

Mild Liver Disease

16 (12.2%)

660 (10.7%)

0.58

213 (16.6%)

1281 (12.1%)

<0.001

Diabetes wo Chronic Comp

44 (33.6%)

1718 (27.8%)

0.15

324 (25.2%)

2769 (26.1%)

0.47

Peripheral Vascular Disease Cerebrovascular Disease Dementia

0.97

ro

-p

re

lP

na

Charlson defined chronic conditionsk

ur

Elixhauser defined chronic conditionsf

of

f

Jo

Shock

Yes N= 131

28

Journal Pre-proof SI ≥0.9

SI <0.9

Variable type

Mortality n (%)

Diabetes w Chronic Comp

15 (11.5%)

Paraplegia

Mortality n (%) P Value

Yes N= 1,287

No N= 10,606

P Value

666 (10.8%)

0.81

105 (8.2%)

966 (9.1%)

0.26

14 (10.7%)

273 (4.4%)

<0.001

52 (4.0%)

458 (4.3%)

0.64

Renal Disease

28 (21.4%)

804 (13.0%)

0.005

221 (17.2%)

1178 (11.1%)

<0.001

Cancer

21 (16.0%)

643 (10.4%)

0.04

210 (16.3%)

1179 (11.1%)

<0.001

Severe Liver Disease

11 (8.4%)

184 (3.0%)

<0.001

106 (8.2%)

417 (3.9%)

<0.001

15 (11.5%) 1 (0.8%)

310 (5.0%) 29 (0.5%)

224 (17.4%) 16 (1.2%)

729 (6.9%) 85 (0.8%)

HIV AIDS

-p

Metastatic Cancer

of

No N= 6,173

ro

Yes N= 131

0.001 0.63

ur

na

lP

re

Abbreviations: SI, Shock Index; SD, Standard deviation; ICU, Intensive care unit; APACHE, Acute physiology, age, chronic health evaluation Because of rounding, categories will not always add to 100% *Median [25th, 75th]: Age, 65 [53, 75]; APACHE III, 50 [39, 64] APACHE III score includes physiology, chronic health investigation, and age variables a At index hospital visit; b At ICU admission; c From ICU admission day -1 to -90 d During ICU stay; e From ICU admission day -1 to -7 f During the index hospital visit g During the first 72 hours from ICU admission h During the first 48 hours from the index hospital admission i Hospital admission to ICU admission + 24-hours j From hospital admission to ICU admission k Prior to index hospital visit

Jo

Variable name

29

<0.001 0.10

Journal Pre-proof Table 2. Adjusted marginal effects for Shock Index ≥0.9 for cumulative minutes’ exposure.

Per every 4-h 0-h >0 to 2-h >2 to 8-h >8 to 24-h >24 to 48-h >48 to 72-h >72 to 168-h

OR (95% CI): 1.058 (1.046 - 1.070) 14.4% (13.2% to 15.6%) 17.7% (16.0% to 19.3%) 19.6% (18.0% to 21.2%) 23.8% (22.2% to 25.4%) 26.7% (24.2% to 29.2%) 30.0% (26.9% to 33.0%) 31.8% (29.2% to 34.5%)

lP

na

Jo

ICU Lengthof-Stay (Survivors)

Hospital Length-ofStay (Survivors)

OR (95% CI): 1.043 (1.037 - 1.049) 5.2% (4.3% to 6.0%) 5.9% (4.5% to 7.4%) 6.5% (5.3% to 7.8%) 7.8% (5.9% to 9.6%) 8.1% (6.1% to 10.2%) 9.6% (6.9% to 12.3%) 9.2% (6.8% to 11.6%)

Per every 4-h 0-h >0 to 2-h >2 to 8-h >8 to 24-h >24 to 48-h >48 to 72-h >72 to 168-h

OR (95% CI): 1.021 (1.012 - 1.031) 3.2 days (3.1 to 3.3) 3.7 days (3.6 to 3.9) 3.9 days (3.7 to 4.1) 4.4 days (4.1 to 4.6) 5.0 days (4.6 to 5.3) 6.4 days (6.0 to 6.7) 8.4 days (7.9 to 8.9)

Per every 4-h 0-h >0 to 2-h >2 to 8-h >8 to 24-h >24 to 48-h >48 to 72-h >72 to 168-h Per every 4-h

COEF (95% CI): 1.037 (1.034 - 1.040) 7.5 days (7.1 to 7.8) 8.8 days (8.4 to 9.1) 9.0 days (8.5 to 9.4) 9.7 days;(9.4 to 10.1) 10.4 days (9.8 to 11.0) 12.0 days (11.1 to 12.9) 14.4 days (13.2 to 15.6)

ur

Myocardial Injury

Per every 4-h 0-h >0 to 2-ha >2 to 8-hb >8 to 24-h >24 to 48-h >48 to 72-h >72 to 168-h

ro

Acute Kidney Injury

of

Adjusted Probability or Mean Effect (95% CI) 3.8% (3.1% to 4.4%) 5.1% (4.3% to 6.0%) 6.5% (5.5% to 7.4%) 8.0% (7.1% to 9.0%) 12.3% (11.0% to 13.6%) 11.7% (9.1% to 14.2%) 15.9% (12.5% to 19.2%)

-p

In-Hospital Mortality

Cumulative hours of exposure to SI ≥0.9 0-h >0 to 2-h >2 to 8-h >8 to 24-h >24 to 48-h >48 to 72-h >72 to 168-h

re

Outcome

COEF (95% CI): 1.023 (1.020 - 1.026)

Abbreviations: SI, Shock Index; OR, Odds ratio; COEF, coefficient; CI, confidence interval p Value <0.001 unless noted;

30

Journal Pre-proof

ur

na

lP

re

-p

ro

of

p=0.24; bp=0.005

Jo

a

31

Journal Pre-proof Table 3. Adjusted marginal effects for Time-Weighted Average of Shock Index ≥0.9 exposure

In-Hospital Mortality

0

Adjusted Probability or Mean Effect (95% Confidence Interval) 3.8% (3.2% to 4.5%)

Q1: >0 -0.001a

5.1% (4.1% to 6.2%)

Q2: 0.001-0.006

6.2%; (5.2% to 7.1%)

Q3: 0.006-0.019

7.5%; (6.6% to 8.5%)

Q4: 0.019-0.057

8.3%; (7.5% to 9.1%)

Q5: 0.057-1.138

15.2%; (13.1% to 17.2%)

0

14.8% (13.7% to 16.1%)

Acute Kidney Injury

of

Quintiles

Outcome

19.4% (17.7% to 21.1%)

ro

Q1: >0 -0.001 Q2: 0.001-0.006

21.2% (19.6% to 22.7%)

-p

Q3: 0.006-0.019

25.3% (23.4% to 27.2%)

0

5.3% (4.4% to 6.2%) 6.3% (4.8% to 7.9%)

Q2c: 0.001-0.006

6.9% (5.2% to 8.5%)

na

Q1b: >0 -0.001

Q3: 0.006-0.019

7.8% (6.2% to 9.5%)

Q4: 0.019-0.057

7.9% (6.1% to 9.6%)

ur Jo

Hospital Lengthof-Stay (for Survivors)

24.8% (22.9% to 26.8%)

Q5: 0.057-1.138

lP

Myocardial Injury

re

Q4: 0.019-0.057

23.1% (20.9% to 25.2%)

Q5: 0.057-1.138

7.7% (6.0% to 9.4%)

0

7.7 days (7.3 to 8.1)

Q1: >0 -0.001

9.4 days (9.0 to 9.8)

Q2: 0.001-0.006

10.0 days (9.5 to 10.4)

Q3: 0.006-0.019

9.8 days (9.4 to 10.3)

Q4: 0.019-0.057

9.9 days (9.8 to 10.4)

Q5: 0.057-1.138

10.0 days (9.4 to 10.7)

p Value <0.001 unless noted a p=0.001; bp=0.117; cp=0.006

32

Journal Pre-proof Highlights



A single abnormal Shock Index (SI) value has limited predictive value



Increased exposure to SI ≥0.9 is associated with increased mortality and morbidity

ur

na

lP

re

-p

ro

clinically comparable in predicting mortality

of

Cumulative exposure to SI ≥0.9 and systolic blood pressure ≤100-mmHg were

Jo



33

Figure 1

Figure 2

Figure 3