To address knowledge gaps by (i) developing a theoretical understanding of escalation and (ii) identifying escalation success factors. Non-participant observations were used to examine deteriorating patient escalation events. Escalation event data were collected by a researcher who shadowed clinical staff, between February 16th 2021 and March 17th 2022 from two National Health Service Trusts. Events were analysed using Framework Analysis. Escalation tasks were mapped using a Hierarchical Task Analysis diagram and data presented as percentages, frequency and 95% CI. A total of 38 observation sessions were conducted, totaling 105 h, during which 151 escalation events were captured. Half of these were not early warning score-initiated and resulted from bleeding, infection, or chest pain. Four communication phenotypes were observed in the escalation events. The most common was Outcome Focused Escalation, where the referrer expected specific outcomes like blood cultures or antibiotic prescriptions. Informative Escalations were often used when a triggering patient's condition was of low clinical concern and ranked as the second most frequent escalation communication type. General Concern Escalations occurred when the referrer did not have predetermined expectations. Spontaneous Interaction Escalations were the least frequently observed, occurring opportunistically in communal workspaces. Half of the events were non-triggering escalations and understanding these can inform the design of systems to support staff better to undertake them. Escalation is not homogenous and differing escalation communication phenotypes exist. Informative Escalations represent an organizational requirement to report triggering warning scores and a targeted reduction of these may be organizationally advantageous. Increasing the frequency of Spontaneous Escalations, through hospital designs, may also be beneficial. Our work highlights that a significant proportion of escalation workload occurs without a triggering early warning score and there is scope to better support these with designed systems. Further examination of reducing Informative and increasing Spontaneous Escalations is also warranted. Extensive PPIE was completed throughout the lifecycle of this study. PPIE members validated the research questions and overarching aims of the overall study. PPIE members contributed to the design of the study reviewed documents and the final data generated.