Interpretability of Temporal Fusion Transformer Vitals Forecasters Under Nursing Shift Handover Documentation Gaps: A Computer Reliability Evaluation

Temporal Fusion Transformer (TFT) models are increasingly proposed for forecasting vital signs in hospital wards, yet their interpretability outputs are typically evaluated on complete, regularly sampled data. In nursing practice, vital-sign series are punctuated by shift handover documentation gaps, during which charting pauses while care continues. Because no observational corpus of handover-gapped vitals is available, the study is framed as a design-characterisation analysis: a before-after-control-impact (BACI) estimator is evaluated under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64) across 2000 replications per design cell, with cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The analysis characterises the operating characteristics of the design rather than reporting empirical TFT performance. Power curves, minimum detectable effects and sampling distributions are derived for the BACI interaction, and the implications for interpretability auditing are stated as propositions. The central argument is that handover gaps function as a structured missingness mechanism whose interaction with attention-based interpretability has not been characterised, and that reliability evaluation of TFT explanations must be designed around that mechanism rather than around complete-case benchmarks.