Missed Nursing Care Among Emergency and Critical Care Nurses: An Item-Level Network Analysis With Computer-Simulated Interventions.

AIM: The primary aim was to estimate the item-level conditional association network of missed nursing care among emergency and critical care nurses. Secondary aims were to identify central, bridge and perturbation-responsive activities and assess sensitivity to dose, parameterisation and dichotomisation. BACKGROUND: Missed nursing care, defined as required care that is omitted, delayed or incomplete, is associated with poorer care quality and adverse patient outcomes. Studies in emergency and critical care have mainly examined total scores, item occurrence or associated factors. How individual omissions are conditionally related and whether frequently missed activities are also central, bridge care categories or respond most strongly to simulated perturbation remain unclear. METHODS: This study was a secondary analysis based on a previously established dataset of 3351 emergency and critical care nurses recruited through stratified cluster sampling from 22 tertiary Grade A hospitals in Hubei Province, China, between January and August 2025. Participants completed Part A of the MISSCARE Survey, comprising 24 missed care activities. Responses were dichotomised as no occurrence or any occurrence. An Ising network estimated conditional associations between binary activities. Expected influence (EI) measured signed overall connectivity, whereas bridge EI (BEI) measured signed connectivity with activities outside each care community. Bootstrap procedures assessed accuracy and stability. Using the NodeIdentifyR algorithm (NIRA), each calibrated node threshold was altered, in turn, in aggravating and alleviating directions, and exact changes in model-implied mean activation were calculated across all 224 possible states. Dose and alternative parameterisation analyses assessed robustness. A Gaussian graphical model (GGM) estimated sparse partial correlations among the original 0-4 scores to assess sensitivity to dichotomisation. RESULTS: The Ising network retained 109 of 276 possible edges, of which 108 were positive. Vital sign monitoring had the highest raw EI (8.97), followed by rehabilitation guidance and focused reassessment/care-plan updating. Pressure ulcer/wound/ostomy care had the highest raw BEI (6.94), followed by rehabilitation guidance and discharge planning/teaching. From a calibrated model-implied baseline of 4.93 active nodes, a 2-SD threshold increase for intravenous/drainage tube care yielded the largest aggravating increase in the mean number of active nodes (+5.67), whereas a 2-SD threshold decrease for emotional support yielded the largest alleviating decrease (-2.70). Emotional support and bathing/skin care were the only activities that remained among the five highest-ranked alleviating nodes at every examined dose in both parameterisations. BEI rankings were highly concordant between the Ising network and GGM (Spearman's ρ = 0.961), with four of the five highest-ranked activities shared. CONCLUSIONS: Missed nursing care formed a predominantly positive network. Frequently missed, central, bridge and perturbation-responsive activities were related but distinct. The NIRA findings are hypotheses derived from a cross-sectional model and do not represent observed intervention effects or causal estimates. IMPLICATIONS FOR NURSING MANAGEMENT: Occurrence frequency alone is insufficient for prioritisation. Centrality, bridge connectivity and simulated responsiveness provide complementary criteria for nursing audits and prospective evaluation. Emotional support and bathing/skin care warrant evaluation in longitudinal studies and intervention trials.