The Utility of the Shock Index for Predicting Survival, Function and Health Status Outcomes in Major Trauma Patients: A Registry-Based Cohort Study

Lena Wikström, Thomas Kander, Belinda J. GabbeTrauma 2022 · DOI

TL;DR

This study investigates whether the Shock Index (SI), a simple ratio of heart rate to systolic blood pressure, can improve existing models for predicting survival and longer-term outcomes in major trauma patients. The findings suggest that while the SI offers a small but statistically significant improvement in predicting survival and functional recovery, its impact on health status prediction is minimal.

What Problem Does This Paper Address?

Traumatic injuries are a leading cause of death and disability globally. Effective trauma care relies on accurate prediction models to guide triage, resource allocation, and patient management. Existing models, like the Trauma and Injury Severity Score (TRISS) and Revised Trauma Score (RTS), often rely on variables like systolic blood pressure (SBP) but don't explicitly account for the physiological state of shock, which is a major contributor to mortality. Hemorrhage leading to hypovolemic shock is a critical concern, and a simple, readily available metric that captures this state could enhance predictive accuracy. The Shock Index (SI), defined as heart rate divided by SBP, has shown promise in predicting mortality and the need for blood transfusions, but its utility in predicting longer-term functional and health status outcomes in major trauma patients remains largely unexplored. This paper aims to fill that gap by evaluating the SI's contribution to existing prediction models for survival, function, and health status.

Methods

The researchers conducted a registry-based cohort study using data from the Victorian State Trauma Registry (VSTR). They developed and evaluated several prediction models. The core approach involved comparing the performance of existing VSTR models (incorporating variables like age, sex, Injury Severity Score (ISS), Head Injury Severity Scale (HISS), and mechanism of injury) with and without the inclusion of the Shock Index (SI).

Key statistical methods included:

  • Discrimination Analysis: Measured using the Area Under the Receiver Operating Characteristic Curve (AUC), which quantifies a model's ability to distinguish between different outcomes (e.g., survival vs. death).
  • Calibration Analysis: Assessed using the Hosmer-Lemeshow (H-L) statistic and calibration curves to evaluate how well the predicted probabilities matched observed outcomes.
  • Model Comparison: Likelihood ratio (LR) tests were used to compare the fit of nested models (models where one is a subset of the other).
  • Regression Models: Binary logistic regression was used for predicting survival and independent living (GOS-E), while linear regression was used for predicting health status (EQ-5D-3L).

The dataset was split into training and testing sets to ensure internal validation. The SI was dichotomized into "low" (SI < 1) and "high" (SI ≥ 1) groups for analysis, based on prior research.

Datasets

The study utilized data from the Victorian State Trauma Registry (VSTR). This registry contains data on all major trauma patients managed within the Victorian trauma system.

  • Data Source: Victorian State Trauma Registry (VSTR).
  • Patient Population: 29,574 adult major trauma patients (aged >15 years) with a date of injury between July 2009 and June 2019.
  • Data Collected: Patient demographics, injury type, injury severity, physiological observations on arrival, hospital length of stay, mortality, and 6-month follow-up data on function and health status.
  • Follow-up: Standardized telephone interviews were conducted for survivors at 6 months post-injury. Post-discharge deaths were identified through linkage with the Victorian Registry of Births, Deaths and Marriages.

No new datasets were introduced; the study relied solely on existing registry data.

Key Insights and Ideas

  • SI as a Modest Predictor: The Shock Index, a simple physiological measure, adds a statistically significant but small improvement to existing trauma prediction models for survival and functional outcomes.
  • Survival vs. Long-Term Outcomes: The SI appears to be a more impactful predictor for short-term outcomes like survival compared to longer-term functional and health status. This makes intuitive sense, as the SI reflects an acute physiological state.
  • Age-Specific Differences: The SI's impact on survival prediction was more pronounced in older adults, suggesting it might be particularly useful in identifying high-risk elderly trauma patients. However, the overall discriminative ability of models was higher in older adults compared to younger adults.
  • Limitations of Existing Models: The study highlights that even established models like TRISS and RTS performed less effectively in this cohort, potentially due to a higher median Injury Severity Score (ISS) compared to previous studies. This underscores the need for continuous model refinement.
  • Health Status Prediction Remains Challenging: The SI had a negligible impact on predicting health status (EQ-5D-3L scores), indicating that this outcome is influenced by a complex interplay of factors not captured by the SI alone.

Key Results

  • Survival to Hospital Discharge: Adding the SI to the VSTR model (Sex + Age + ISS + HISS + Mechanism) improved the AUC from 0.797 to 0.807 (p < 0.001) in the training dataset. This represents a marginal increase in predictive accuracy.
  • Independent Living (GOS-E at 6 Months): The SI also led to a small but significant improvement in AUC for predicting independent living, increasing from 0.795 to 0.799 (p < 0.001) when added to the VSTR model.
  • Health Status (EQ-5D-3L at 6 Months): The inclusion of the SI resulted in only a marginal improvement in R² (from 0.109 to 0.115) and a slight reduction in RMSE, indicating a very small enhancement in explaining the variance in health status.
  • Older Adults: For older adults (≥65 years), adding the SI to the VSTR model significantly improved survival discrimination (AUC increased from 0.681 to 0.695).
  • Younger Adults: For younger adults (<65 years), the SI also improved survival discrimination (AUC increased from 0.820 to 0.824).
  • Calibration: None of the models demonstrated acceptable calibration across the board, with calibration curves showing some under- or over-estimation at different prediction percentiles.

Limitations and Future Work

  • Exclusion of Patients: Patients who were deceased on hospital arrival (with SBP or HR equal to zero) were excluded because the SI could not be calculated. This means the SI's utility in the most critically ill patients at the point of arrival was not assessed.
  • Dichotomization of SI: The SI was dichotomized into "low" and "high" groups for analysis. While this simplified the models, it may have led to a loss of information inherent in the continuous SI values.
  • Inter-Hospital Transfers: Substituting missing vital signs from the primary hospital with data from the definitive hospital for patients who were transferred could introduce bias, as these patients might be more physiologically stable upon arrival at the definitive care center.
  • Responder Bias: As with many follow-up studies, there's a potential for responder bias, where certain groups (e.g., those injured in intentional events, younger individuals, or less seriously injured patients) might be over- or under-represented in the follow-up data.
  • Future Work: The authors recommend external validation of these findings using different datasets and trauma populations. They also suggest exploring the inclusion of other factors in the VSTR predictive model to further maximize discrimination and prediction capacity, potentially looking at models used in other regions like the UK or Canada. Further research into the applicability of models to specific contexts, like the Australian trauma system, is also warranted.

Summary updated

Abstract

The shock index (SI; heart rate/systolic blood pressure) has been found to predict mortality in trauma patients. The aims of this study were to establish whether the SI improved the prediction of an existing model for both mortality and longer-term outcomes in major trauma patients. In total, 29,574 adult (&gt;15 years) major trauma patients were included from the Victorian State Trauma Registry with a date of injury from July 2009 to June 2019. Outcomes of interest were survival to hospital discharge and function and health status at 6 months post-injury. Survival and function were analysed using measures of discrimination and calibration, whereas health status was assessed with R2 and MRSE measures. The area under the receiver operating characteristic curve (AUC) of the VSTR survival model improved when the SI was added (AUC 0.797 (0.787–0.807) versus AUC 0.807 (0.797–0.816), p &lt; 0.001). For the prediction of functional outcome 6 months post-injury, the inclusion of the SI increased the AUC marginally (AUC 0.795 (0.786–0.803) versus AUC 0.799 (0.791–0.807), p &lt; 0.001). When predicting in-hospital mortality and health status 6 months post-injury, including the SI resulted in a slightly better fit to the registry-risk adjustment model. In the future, external validation and the exploration of other models to improve prediction outcomes are warranted.
BibTeX
@inproceedings{wikstrom2022utility,
  author = {Lena Wikström and Thomas Kander and Belinda J. Gabbe},
  title = {{The Utility of the Shock Index for Predicting Survival, Function and Health Status Outcomes in Major Trauma Patients: A Registry-Based Cohort Study}},
  booktitle = {Trauma},
  year = {2022},
  doi = {10.3390/traumacare2020023},
}
Figure