Postoperative Critical Care: Resource Availability, Patient Risk and Other Factors Influencing Referral and Admission

Although intended for benefit, surgery exposes patients to potential complications. Critical care is thought to protect against the development of these complications, and is recommended for patients at higher risk. However, previous literature suggests that high-risk patients do not consistently receive postoperative critical care. In this PhD thesis, I investigate the supposed misallocation of critical care resources, and seek to answer the following research questions: 1. What is the availability of postoperative critical care? 2. How do clinicians estimate perioperative risk? 3. How accurate are current available risk prediction tools? 4. How do clinicians decide which patients to admit for postoperative critical care? 5. What factors influence their admission? A survey of postoperative critical care availability was conducted in 309 hospitals across the United Kingdom, Australia and New Zealand (NZ). Then, in a subset of 274 of these hospitals, a cohort study enrolling 26,502 patients undergoing inpatient surgery was undertaken. Postoperative critical care availability was found to differ between countries. UK hospitals reported fewer critical care beds per 100 hospital beds (median = 2.7) compared with Australia (median = 3.7) and NZ (median = 3.5). Enhanced care/high-acuity beds used to manage some high-risk patients were identified in around 31% of hospitals. The estimated numbers of critical care beds per 100,000 population were 9.3, 14.1, and 9.1 in the UK, Australia, and NZ, respectively. The estimated per capita high-acuity bed capacities per 100,000 population were 1.2, 3.8, and 6.4 in the UK, Australia, and NZ, respectively. The risk profile of inpatients undergoing inpatient surgery and the incidence of short-term mortality and morbidity outcomes were described. Less than 40% of predicted high-risk patients (defined as having a 5% or higher predicted 30-day mortality) in the cohort were admitted to critical care directly after surgery, regardless of risk model used. Compared with objective risk tools, subjective clinical assessment performed similarly in terms of discrimination, but consistently overpredicted risk. The Area Under the Receiver Operating Characteristic curve (AUROC) for subjective clinical assessment was 0.89, compared to 0.91 for the Surgical Outcome Risk Tool (SORT), the best-performing objective risk tool. However, a model combining information from both objective tools and subjective assessment improved the accuracy and clinical applicability of risk predictions (combined model AUROC = 0.93; continuous Net Reclassification Index [NRI] = 0.347, p