Decision-making in patient flow management: Cognitive strategies and the role of forecasting

Abstract

Hospitals in the UK are struggling to achieve timely admissions of emergency patients to beds. Currently, over a quarter of patients wait longer than four hours between a decision to admit and admission. Given the financial and practical limits on a sustained increase in capacity, it has been proposed that insights about patient flow may yield efficiencies. Patient flow coordinators and managers play a critical role in ensuring patients receive timely treatment in an appropriate type of bed and ward, but we know little about the cognitive aspects of their work. Forecasts of patient admissions are produced by many NHS trusts and have the potential to inform operational flow decisions, but it is not clear how they are used in practice. Thus, this thesis studies the cognitive decision-making strategies of patient flow coordinators and managers and examines their use of bed demand forecasts.

A mixed methods design was used to study individuals’ cognition and contextual influences on their work at acute medical units in two large hospitals in the UK. Data collection used participant observations, semi-structured interviews, chart drawing, and routinely collected admissions data. Data analysis applied the skills-rules-knowledge model of cognition, the decision ladder template, and time series forecasting methods.

It was found that patient flow management has characteristics that differentiate it from many problems previously studied in naturalistic settings. It required continuous control of a dynamic system with fluctuating demand, a changing configuration of patients to beds, and delayed information feedback loops. Consequently, the effects of a decision were tightly interrelated with subsequent decisions but were often not fulfilled for several hours.

Decision-makers used several cognitive strategies to make allocation, prioritisation and capacity decisions: sensemaking, sequential rule-based matching, option evaluation, time projection and feedforward control. The level of cognitive processing, as defined in the skills-rules-knowledge model, varied and this was influenced by the structure of the information environment. For example, the task required a deliberate search for clinical and operational data that were distributed across sources, limiting intuitive decision-making. Time projection involved extrapolating from the current bed state to a situation several hours in the future and was informed by individuals’ tacit knowledge of bed demand patterns. Coordinators used a feedforward control strategy to make proactive decisions to compensate for delays in the patient transfer and capacity adjustment processes. Decision-makers had limited engagement with the daily forecast of admissions, and this is attributed to several factors. The forecast had poor accuracy, was reported without prediction intervals, and did not align with the planning horizons of flow coordinators and managers. The time series analysis demonstrated that accuracy improvements of 37.3% (Trust A) and 23.1% (Trust B) over incumbent methods can be achieved.

This thesis contributes by providing insight into the cognitive work of professionals who are central to the delivery of timely and appropriate hospital treatment. It extends existing studies on patient flow roles by revealing that they alternated between different levels of cognition that included intuitive, conscious, and rule-based strategies. This work contributes to knowledge about the temporal aspects of decision-making by providing evidence that individuals pictured the future bed state over a time horizon of several hours. This is important because previous naturalistic studies have reported anticipatory processes for only seconds or minutes. The analysis posits that coordinators achieved an extended horizon of anticipation by developing a mental model of the patient flow system. There are implications for the development of decision aids to support and share the coordinator’s mental picture. Furthermore, there is scope to integrate managers’ knowledge of influences on admissions with statistical methods to produce more useful forecasts of bed demand.

Date of Award	25 Jun 2025
Original language	English
Awarding Institution	University of Bath
Supervisor	Christos Vasilakis (Supervisor), Julie Gore (Supervisor) & Fotios Petropoulos (Supervisor)