TL;DR
This paper outlines a protocol for a systematic review aiming to identify patient-specific variables that predict unplanned ICU admissions. By synthesizing existing research, the authors hope to inform the development of more accurate predictive models, ultimately improving patient outcomes by enabling earlier intervention.
What Problem Does This Paper Address?
Hospitals frequently see unplanned admissions to Intensive Care Units (ICUs). These admissions are often preceded by a period of clinical deterioration, and patients admitted this way have a higher mortality rate than those admitted directly. While Early Warning Scores (EWS) exist to flag deteriorating patients, they often lack sufficient positive predictive value, leading to too many false alarms and potentially missed critical events. The authors hypothesize that incorporating additional patient data beyond basic vital signs could improve the predictive power of these systems, but it's unclear which specific patient variables are most effective. This review aims to systematically identify these predictive variables.
Methods
The paper details a protocol for a systematic review, adhering to PRISMA-P guidelines. The methodology involves:
- Database Searching: Comprehensive searches of MEDLINE, EMBASE, CINAHL, and the Cochrane Library will be conducted.
- Study Selection: Two independent researchers will screen titles and abstracts, followed by full-text review for eligible articles. Disagreements will be resolved through discussion or a third party.
- Inclusion Criteria: Quantitative studies published in peer-reviewed journals, focusing on adult patients admitted to general ICUs, with data published from January 2000 onwards. Studies must describe a statistical relationship between a patient-derived variable and unplanned ICU admission.
- Exclusion Criteria: Qualitative studies, case studies, grey literature, editorials, letters, practice guidelines, and abstract-only reports will be excluded. Studies focusing on narrow patient cohorts (e.g., trauma, sepsis) or ICU readmissions will also be excluded.
- Data Extraction: Two authors will independently extract data, with a tool called DistillerSR used for management.
- Quality Assessment: A scoring system adapted from the Newcastle-Ottawa Scale (NOS) will be used to assess the risk of bias in included studies.
- Data Synthesis: A qualitative synthesis of results will be presented descriptively, with summary comparison data (odds ratios) extracted where possible.
Datasets
Not discussed in the paper. The paper describes a protocol for a systematic review of existing literature, not the creation or use of new datasets for training or evaluation.
Key Insights and Ideas
- Beyond Vital Signs: The core idea is that current EWS, primarily based on vital signs, are insufficient. The review seeks to identify other patient-specific data points that could significantly improve predictive accuracy.
- The Need for Specificity: The goal is to move beyond general "deterioration" and pinpoint variables that specifically predict the need for ICU admission, which is a critical step for timely intervention.
- Foundation for Future Models: This systematic review is intended to be a foundational step for developing improved predictive models, potentially integrating these identified variables into future clinical decision support systems.
Key Results
As this is a protocol for a systematic review, no results from the review itself are presented. The paper outlines the planned methodology and the expected contribution to the field.
Limitations and Future Work
- Limitations of Included Studies: The quality assessment section acknowledges that the review will assess the risk of bias in the included studies, implying that limitations within the existing literature (e.g., study design, comparability of groups) will be a factor.
- Future Work: The primary future work envisioned is the development of improved models for predicting clinical deterioration and unplanned ICU admission, directly informed by the findings of this systematic review. The authors suggest that the results may also be useful for researchers looking to enhance existing predictive tools.
Summary updated