Introduction

Trial-based economic evaluations rely on a number of methods for estimating resource use and costs. The use of routine data has hitherto been limited, with accuracy of coding, confidentiality, ownership and access having been previously identified as significant barriers to access. The Department of Health in England has recommended the use of Patient Level Information and Costing Systems (PLICS) to understand financial drivers at patient, specialty and hospital levels.
These provide an opportunity for estimating secondary care costs within economic evaluations.

Background
As part of a randomised controlled trial comparing the use of multiple daily injections of insulin with pumped infused insulin in newly diagnosed paediatric Type 1 Diabetes Mellitus patients, we investigated the availability and feasibility of PLICS data for estimating diabetes-related hospital inpatient stays.

Method
We obtained consent to access patients’ electronic records from 15 participating sites. Diabetes-related, patient-level data were recorded, on: HRG codes PA67Z and PA68Z, lengths of hospital stay, total cost and, where available, full disaggregation of PLICS data on items such as critical care and drug costs. Inpatient stays were costed using 3 methods based on: (i) bed days alone; (ii) Payment by results (PBR) National Tariff reimbursement; and (iii) PLICS-reported total cost. Confidence intervals were calculated using non-parametric bootstrap
analysis with 10,000 replications.

Results
Data relating to 82 hospital admissions were obtained for 74 patients at 5/15 sites. The remaining hospitals (10/15) were still in the process of setting up their PLICS systems and could only provide routine patient admissions or legacy finance database outputs in time for the study. The diabetes-related inpatient stays (N = 67/82 episodes) were comprised of the codes PA67Z (admission related to diabetic ketoacidosis) (12/67) and PA68Z (admission related to diabetes mellitus, without ketoacidosis or coma) (55/67). Mean costs (95% confidence intervals) for the diabetes-related codes were: (i) bed days: £662
(£587, £741); (ii) PBR: £1252 (£1230, £1278); and (iii) PLICS: £1839 (£1339, £2425). Disaggregated PLICS costs comprised medical/ specialist nursing staff (47%), wards/overheads (30%), critical care (8%), other clinical supply and services (6%), pharmacy/drug costs (5%), therapies (1%) and pathology (1%) with the remainder comprising blood supplies, imaging, operating theatre and other
diagnostic tests.

Conclusion
There is no agreed gold standard for estimating inpatient costs for economic evaluations. Reliance on bed day costing alone risks underestimating the total cost of an inpatient stay, especially if the daily rate does not account for staff, critical care, wards and overhead costs. PBR, whilst giving a more accurate cost based on hospital reimbursement, lacks granularity and does not include unbundled costs such as critical care, expensive drug costs and overheads which are reimbursed at a local level. PLICS outputs have sufficient detail to account for these shortfalls and could provide a more robust method of inpatient cost estimation in trial-based economic evaluations, especially where the stay involves additional expensive bundles of care such as long operation times and intensive care admission.