Targeting a Modifiable Risk Factor: A Causal Framework Informs Strategies to Reduce Pressure Injuries

Speaker(s)

Alderden J1, Krikov S2, Li H3, Johnny J4, Vanderpuye-Orgle J5, Gregg M6, Wilson A7
1Boise State University, Boise, ID, USA, 2Parexel International, Lexington, MA, USA, 3University of California, Berkeley, Oakland, CA, USA, 4Huntsman Cancer Hospital, Salt Lake City, UT, USA, 5Parexel, Inc., LA VERNE, CA, USA, 6Parexel, Austin, TX, USA, 7Parexel International, Waltham, MA, USA

Presentation Documents

OBJECTIVES: Pressure injuries (PrI) are a serious concern in critical care patients, and low serum albumin is frequently identified as a PrI risk factor. However, despite abundant existing literature highlighting the association between hypoalbuminemia and PrI formation, a critical gap exists in understanding a potential causal relationship; this gap is particularly relevant because serum albumin is a modifiable factor. This study employed a formal causal framework to rigorously examine the potential causal effect of albumin on PrI, providing insights for improving PrI prevention strategies.

METHODS: We applied the steps of the roadmap for causal inference to estimate the protective effect of increased Albumin against developing pressure injuries. The data comes from the MIMIC IV dataset, capturing hospital admissions from 2008 to 2019 for almost 300,000 patients. Collaborative targeted maximum likelihood estimation (C-TMLE) was used to estimate the Average Treatment Effect (ATE) of Albumin on the development of pressure injuries.

RESULTS: Of 17,504 eligible cases, 1,566 developed a pressure injury (8.9%). The ATE of Albumin showed a significant protective effect (𝛹 = 0.04021, 95% CI: 0.03082 to 0.0496, p <0.0001). Additionally, when considering Albumin exposure quintiles, the effect estimate is monotonic, increasing the risk with decreasing Albumin quintiles. (presented as Figure 1)

CONCLUSIONS: The causal roadmap helps identify causal effects from data. We must consider the context in which data was generated, including intervening events. Real-world data can contain limitations, so we need robust collection methods and advanced analytics. It's important to interpret the results carefully.

These preliminary findings, while compelling, warrant further investigation into the dynamic aspects of care delivery, such as dynamic patient conditions and intercurrent interventions. Ultimately, this research aims to guide potential intervention experiments to verify if exogenous albumin administration can prevent pressure injuries in clinical settings.

Code

RWD94

Topic

Methodological & Statistical Research, Study Approaches

Topic Subcategory

Artificial Intelligence, Machine Learning, Predictive Analytics, Confounding, Selection Bias Correction, Causal Inference, Electronic Medical & Health Records

Disease

No Additional Disease & Conditions/Specialized Treatment Areas