Interpreting Missing Data Patterns in the ICU
share
PURPOSE: Clinical examinations are performed on the basis of necessity. However, our decisions to investigate and document are influenced by various other factors, such as workload and preconceptions. Data missingness patterns may contain insights into conscious and unconscious norms of clinical practice. METHODS: We examine data from the SPOTLIGHT study, a multi-centre cohort study of the effect of prompt ICU admission on mortality. We identify missing values and generate an auxiliary dataset indicating the missing entries. We deploy sparse Gaussian Graphical modelling techniques to identify conditional dependencies between the observed data and missingness patterns. We quantify these associations with sparse partial correlation, correcting for multiple collinearity. RESULTS: We identify 35 variables which significantly influence data missingness patterns (alpha = 0.01). We identify reduced recording of essential monitoring such as temperature (partial corr. = -0.0542, p = 6.65e-10), respiratory rate (partial corr. = -0.0437, p = 5.15e-07) and urea (partial corr. = -0.0263, p = 0.001611) in patients with reduced consciousness. We demonstrate reduction of temperature (partial corr. = -0.04, p = 8.5e-06), urine output (partial corr. = -0.05, p = 7.5e-09), lactate (partial corr. = -0.03, p = 0.00032) and bilirubin (partial corr. = -0.03, p = 0.00137) monitoring due to winter pressures. We provide statistical evidence of Missing Not at Random patterns in FiO2 and SF ratio recording. CONCLUSIONS: Graphical missingness analysis offers valuable insights into critical care delivery, identifying specific areas for quality improvement.
READ FULL TEXT