Quantifying the healthcare costs of treating severely bleeding major trauma patients: a national study for England

Introduction Severely bleeding trauma patients are a small proportion of the major trauma population but account for 40 % of all trauma deaths. Healthcare resource use and costs are likely to be substantial but have not been fully quantified. Knowledge of costs is essential for developing targeted cost reduction strategies, informing health policy, and ensuring the cost-effectiveness of interventions. Methods In collaboration with the Trauma Audit Research Network (TARN) detailed patient-level data on in-hospital resource use, extended care at hospital discharge, and readmissions up to 12 months post-injury were collected on 441 consecutive adult major trauma patients with severe bleeding presenting at 22 hospitals (21 in England and one in Wales). Resource use data were costed using national unit costs and mean costs estimated for the cohort and for clinically relevant subgroups. Using nationally available data on trauma presentations in England, patient-level cost estimates were up-scaled to a national level. Results The mean (95 % confidence interval) total cost of initial hospital inpatient care was £19,770 (£18,177 to £21,364) per patient, of which 62 % was attributable to ventilation, intensive care, and ward stays, 16 % to surgery, and 12 % to blood component transfusion. Nursing home and rehabilitation unit care and re-admissions to hospital increased the cost to £20,591 (£18,924 to £22,257). Costs were significantly higher for more severely injured trauma patients (Injury Severity Score ≥15) and those with blunt injuries. Cost estimates for England were £148,300,000, with over a third of this cost attributable to patients aged 65 years and over. Conclusions Severely bleeding major trauma patients are a high cost subgroup of all major trauma patients, and the cost burden is projected to rise further as a consequence of an aging population and as evidence continues to emerge on the benefits of early and simultaneous administration of blood products in pre-specified ratios. The findings from this study provide a previously unreported baseline from which the potential impact of changes to service provision and/or treatment practice can begin to be evaluated. Further studies are still required to determine the full costs of post-discharge care requirements, which are also likely to be substantial. Electronic supplementary material The online version of this article (doi:10.1186/s13054-015-0987-5) contains supplementary material, which is available to authorized users.


Introduction
Trauma is a leading cause of death and disability around the world. An estimated 15 % of major trauma patients present to hospital with severe bleeding, yet 40 % of all major trauma deaths can be attributed to haemorrhage [1]. Bleeding patients are therefore a clinically significant subgroup of the trauma population and have been the focus of recent research on trauma physiology, haemostatic agents, near-patient coagulation testing, damage control resuscitation, and consolidating best-practice management [1][2][3][4][5][6][7][8]. The costs to the healthcare provider of treating severely bleeding trauma patients are likely to be sizeable but have received less attention. Despite calls for better information on these costs, and whilst estimates for blunt and penetrating major trauma in England and Wales have been published, the cost burden to the National Health Service (NHS) of treating severe bleeding in trauma has yet to be fully quantified [9][10][11].
Detailed knowledge of the costs of severe bleeding in trauma is essential for a range of stakeholders. For health service managers, understanding the key cost drivers can help with the development of targeted cost reduction strategies. For health economists, such data can be used as inputs into studies evaluating the costeffectiveness of new bleeding cessation interventions [12]. For policy planners, the nation-wide cost burden of severe bleeding in trauma can be compared with that of other conditions, and potential future costs, such as those resulting from an aging population, can be modelled.
As part of a broader prospective cohort study investigating the incidence, treatment, and outcomes of severe traumatic bleeding, a detailed patient-level assessment of the resource use and costs required to treat severely bleeding major trauma patients arriving at NHS hospitals was performed. Specific objectives of the costing component of the study that is presented in this paper were to: 1) generate estimates of patient-level resource use and healthcare costs, 2) to explore whether costs vary across particular clinical subgroups, and 3) to scale up patient-level figures to a national level for England using Hospital Episode Statistics (HES) data.

Methods
The work was conducted in collaboration with the Trauma Audit Research Network (TARN), the independent monitor of trauma care in England and Wales [13]. TARN collects information from hospitals in England and Wales on all presenting major trauma patients (those with a hospital length of stay of 72 hours or more, and/or requiring intensive or high dependency care, and/or where a death occurred in hospital). Data coordinators at hospitals record detailed patient-level information on treatments and care received, and mortality at hospital discharge and at 30 days. All data are anonymised and entered into TARN's Electronic Data Collection and Reporting (EDCR) system.
Twenty two hospitals (21 in England and one in Wales) submitting data to TARN were recruited to the prospective study. The general methodology and main study outcomes have been submitted for publication, and are briefly described as follows [14]. Study hospitals were selected to ensure adequate geographical dispersion and a balanced representation of large multi-specialty trauma centres (12 centres) and medium-sized hospitals with trauma units (10 centres). Following Research and Development approvals, the study enrolled consecutive patients with severe traumatic bleeding at the different hospitals, where severe traumatic bleeding was defined by a transfusion need for four units or more of red cells and activation of a massive haemorrhage protocol (MHP). A MHP supports the immediate and simultaneous administration of different blood products in pre-specified ratios, and its activation serves as an indicator that the degree of bleeding is considered significant enough to warrant multi-component transfusion therapy for its attempted control. Only index hospital admissions were sought, and so patients transferred into a study hospital after receiving treatment elsewhere were excluded. Blood component transfusion data were not routinely collected by TARN, and so for the purposes of the study, additional fields for units of blood components issued, transfused, and wasted, and any transfusion-related complications reported were added to the EDCR system. The follow-up period for patient mortality was extended to 12 months postinjury. Study data collection took place between 1 April 2009 and 31 March 2011. The 12-month follow-up period for the last patient entering the study ended on 31 March 2012.
The National Research Ethics Committee South Central Oxford B gave full ethical approval (approval number: 09/ H0605/58) for the study. TARN already had Patient Information Advisory Group approval for data collection to include individual patient's NHS number, age, gender, ethnicity, incident postcode, and sector postcode for home, and so no further permissions were required. A separate application to the National Information Governance Board Ethics and Confidentiality Committee and NHS Medical Register Information Centre allowed the processing of patient identifiable information without consent and the collection of one-year survival and mortality data for participants.
Overall aims of the study as a whole were to: describe the characteristics of patients with major traumatic injuries and bleeding admitted to trauma-receiving emergency departments (ED); develop an estimate of the national incidence of patients requiring four or more units of packed red blood cells (PRBCs); report on the use of PRBCs and fresh frozen plasma (FFP); estimate NHS treatment costs; and describe mortality during 24 hours, at 30 days, and one year.
Resource use data Figure 1 illustrates the care pathways taken by patients in the study cohort. For each patient, data were available on mode of emergency transport to hospital, duration spent in the ED, and Accident and Emergency (A&E) consultant and/or trauma team attendances.
From ED presentation until hospital discharge, transfer, or death, data on haematology tests (full blood counts and clotting investigations) and imaging investigations (X-ray, computed tomography, magnetic resonance imaging, and ultrasound) were collected, along with neurologist, general surgeon, and cardiologist consultations, and the use of pro-coagulant medications (recombinant Factor VIIa and Vitamin K).
Information was collected on the number of units of blood components (PRBCs, FFP, platelets, and cryoprecipitate) and pro-haemostatic concentrates (such as prothrombin complex concentrate (Beriplex)) administered up to 24 hours following presentation, and from hour 25 to 30 days (or discharge, transfer, or death if earlier). Blood banks provided data on units of each blood component issued to a patient that were then wasted, and on any transfusion-related complications. Volumes of intravenous (IV) fluids given up to 24 hours following presentation were also recorded.
For patients admitted to the hospital from the ED, time (hours) spent in theatre, intensive care units (ICU), and on general wards was documented. Inpatient days spent receiving ventilation were also recorded. A small number of patients were transferred directly from the ED to another acute hospital, however no inpatient resource use data were collected from receiving hospitals. Similarly for patients surviving the index hospital admission who were then discharged to another acute hospital or institution, acute care was considered to be ongoing but further information was not collected. Patients discharged to nursing homes and rehabilitation units entered a post-acute phase of care, but the duration of time spent in these facilities was unknown.
The occurrence and duration of re-admissions to initial treating hospitals were recorded. It was assumed that re-admission checks were conducted frequently for each study patient, and as the final study dataset was received from TARN after the time since trauma injury exceeded 12 months for all patients, all re-admissions to initial treating hospitals over 12 months were recorded.

Costs
Patient-level costing was conducted from the perspective of the UK NHS, with costs expressed in 2012 to 2013 UK Pound Sterling. Table 1 shows the unit costs used to value the resource use data collected, and Additional file 1 provides supplementary detailed information on the costing methods used. Unit costs from national sources were used in the first instance, supplemented with local costs when necessary.
Emergency transport to the ED was costed. All patients were assigned the cost of an ED visit, and an A&E consultant contact when recorded. Time spent in the ED was used when costing the intervention of the trauma team, which included general and orthopaedic surgeons (see Additional file 1 for a full list team members). The team was assumed to be present for up to a maximum of one hour into a patient's time in the ED, after which patients were assumed to be stable, receiving intensive care in resuscitation, and awaiting the availability of an inpatient bed. This time was costed using the appropriate proportion of an ICU bed day cost [15].
As IV fluids are supplied in 500 mL and 1 L bags and blood components as unit bags, both were costed to allow for wastage (for example if two and a half units of PRBCs were transfused, a cost of three units was assigned). Added to each unit of blood component transfused and wasted was the estimated cost of hospital transfusion laboratory staff involved in component issuing.
Time in theatre was costed to allow for the running cost of the theatre (including surgeons and nursing staff ). Hours in ICU were costed using an ICU bed day cost, assuming no organ support [15]. The incremental daily cost of ventilation was taken from the published literature (Table 1) and applied to the number of ventilated days recorded for each patient [16]. Non-ICU ward  days were costed using a national excess bed day cost averaged across multiple trauma episodes from the NHS Reference Costs [15]. An ambulance cost was assigned to patients transferred or discharged to another acute hospital or institution. A series of assumptions were made about the duration and intensity of acute care (ICU, ventilation, and ward) provided at receiving hospitals (see Additional file 1 for full details).
It was assumed that patients discharged to a nursing home would have been resident for the remainder of their 12-month study period, and this time was costed [17]. Discharges to rehabilitation units were costed based upon the published literature, which suggests an average length of stay of 33 days at a cost of £2,758 (Table 1) [17]. ICU and ward stay data were documented for re-admissions to the initial treating hospital in the 12 months following injury, and were costed using the appropriate bed day costs.

Data analysis
A total of 12 % of resource use data were missing (a breakdown by variable is shown in columns two and three of Additional file 1: Table A1). Exploration of the data suggested patients with missing data were more likely to have been treated at large major trauma centres than at medium-sized trauma units, and were also less likely to have been discharged to another hospital or institution. Data were assumed to be missing at random, and multiple imputation (MI) with chained regression equations (which included hospital and discharge location as predictors) was used to impute 20 values for each missing data point, essentially creating 20 different datasets (for full details see Additional file 1) [18].
Categorical variables were summarised as frequencies and percentages, and continuous variables using means and standard errors (SE). Parametric 95 % confidence intervals (Cis) are presented for total cost estimates. Rubin's rule was used when summarising data across the five datasets created using MI [19]. When calculating measures of precision, this approach allows the variability both within and between imputed datasets to be accounted for. All analyses were conducted in Stata Version 12.1 (StataCorp, College Station, Texas, USA).

One-way sensitivity analysis
A number of potential sources of uncertainty exist in the analysis. The cost of an unventilated bed day on ICU was increased from £619 to £1,200 (sensitivity analysis (SA)1). It was assumed that half of the 174 inpatients for whom a theatre admission was possible but unrecorded, did undergo surgery, but that these data had been missing (SA2). For these patients, theatre duration was inferred conditional upon age, sex, type of injury, and Injury Severity Score (ISS). It was assumed that all 92 patients for whom no trauma team attendance was recorded, were treated by the team in the ED but that these data had been unrecorded (SA3). Finally, readmissions to other (non-study) hospitals are likely to have occurred but were not captured by the study. The impact of 100 % and 200 % increases in the number of re-admissions observed over the study period was also assessed (SA4 and SA5).

Subgroup analysis
Subgroup analyses were not specified in the study protocol, however costs were summarised for a number of clinically relevant groups, including patients classified as suffering major haemorrhage (receiving four to nine units of PRBCs in 24 hours), and those suffering massive haemorrhage (receiving 10 or more units of PRBCs in 24 hours). Costs were also estimated according to injury severity (an ISS score of 15 or more indicative of more severe injury), and injury type (blunt versus penetrating). Mean cost differences, parametric 95 % CIs, and t-tests were used when comparing costs between subgroups. A P value of less than 0.05 was considered to be statistically significant. For completeness, in-hospital mortality for subgroups was also reported and compared. The Χ 2 test was used when comparing proportions of deaths in each subgroup, and a P value of less than 0.05 was again considered to be statistically significant.

National cost estimates
The HES database provided information on the number of patients (by gender and for 10-year age bands) who presented to all hospitals in England over the study period, and who would have met the TARN eligibility criteria for major trauma [20]. Using all TARN patients (bleeding and non-bleeding) from the 22 study centres, estimates were then made (again by gender and 10-year age bands) of the proportion of major traumas that are associated with severe bleeding (meeting the eligibility criteria of this study). These proportions were assumed to be nationally representative, and were applied (following adjustment for potential under-reporting to TARN) to the HES data to estimate the total number of major trauma cases with severe haemorrhage in England. This same approach was also used for the subgroup of patients classified as having massive haemorrhage (receiving 10 or more PRBCs units in the first 24 hours following presentation). Mean total healthcare costs per patient estimated by this study for patients with severe haemorrhage treated at English hospitals only (n = 414 out of 447), and for the subgroup with massive haemorrhage, were summarised using the same gender and 10-year age categories and multiplied by the appropriate estimated national case numbers to give an estimate of the costs of severe and massive haemorrhage in major trauma patients in England. Uncertainty around these estimates was explored by increasing and decreasing the estimated national case numbers by 5 %, and also by increasing and decreasing the mean total cost estimates for each age and gender subgroup by 10 % and 20 % respectively.

Results
A total of 441 patients met the study entry criteria. Almost three quarters of patients were male, and the mean age of the cohort was 41.9 years. Results showed that 80 % of patients suffered blunt injuries and the mean ISS for the cohort was 29.5. Mortality results showed that 117 out of 441 (26.5 %) patients died whilst in hospital, 120 out of 436 (27.5 %) died within 30 days, and 145 out of 403 (36.0 %) died within 12 months. Figure 1 shows patient pathways for these patients and Table 2 shows demographics and injury characteristics.

Post-acute care resource use and costs
Out of 316 patients, 40 (13 %) surviving the inpatient admission at the presenting hospital were discharged to a rehabilitation unit, and just two (0.6 %) were discharged to a nursing home (Fig. 1). Table 4 shows the cost of this care averaged across the whole cohort. Mean total costs to 12 months were £20,591 (SE: £848; 95 % CI: £18,924 to £22,257) per patient (Table 4).

Re-admissions to original treating hospital at 12 months
A total of 125 patients died during their acute period of care (117 during admission to the presenting hospital, and eight soon after being transferred or discharged to another acute hospital). Amongst the remaining 316 patients, there were 26 (8 %) re-admissions in the 12 months following trauma. Mean re-admission ICU stay was 0.8   Table 4 shows re-admission costs averaged across the whole cohort.
One-way sensitivity analysis Figure 2 shows total healthcare costs and a breakdown of key costs by category for the various sensitivity analyses. Results were most sensitive to a doubling of the daily cost on the ICU (SA1); the mean overall total cost per patient increased from £20,591 (95 % CI: £18,924 to £22,257) to £25,215 (95 % CI: £23,052 to £27,379). Severely injured patients (ISS ≥15) were over 40 % more costly to treat than patients with less severe injuries (£21,870 versus £15,233, mean total cost difference £6,637 (95 % CI: £2522 to £10,753, P <0.01) and appeared five times more likely to die in hospital (proportion 0.31 (112 out of 356) versus proportion 0.06 (5 out of 85), P <0.01) than less injured patients.

National cost estimates
Of the 49,859 major trauma cases presenting to hospitals each year in England, 7,783 (15.6 %) were estimated to have severe haemorrhage (Table 6). Associated costs were estimated to be £148,293,657. Within this figure, 32 % of patients were estimated to have had massive haemorrhage, for which the estimated cost was £56,406,200.
Assuming a 5 % increase and decrease in national case figures altered the total severe haemorrhage cost to £155,708,340 and £140,878,974, respectively, and the massive haemorrhage costs to £59,226,510 and £53,585,890, respectively. A 10 % increase and decrease in mean total costs for each age and gender patient subgroup in Table 6 altered the total severe haemorrhage cost to £163,123,023 and £133,464,291, respectively, and the massive haemorrhage costs to £62,046,820 and £50,765,580, respectively. With a 20 % increase and decrease, corresponding figures were £177,952,388 and £118,634,926 for severe haemorrhage, and £67,687,440 and £45,124,960 for massive haemorrhage.

Discussion
This study has generated estimates of healthcare costs associated with major trauma patients with severe bleeding. Strengths of the study and the cost estimates are data collection across a broad range of hospitals and   Scaling up data to a national level has suggested around 7,800 severely bleeding adult major trauma patients present to hospitals in England each year, at a total cost to treating NHS hospitals of £148,300,000. Sensitivity analyses showed there to be some uncertainty around this estimate, however costs generally remained high. Bleeding trauma patients are therefore a small yet high cost group compared to other types of bleeding patients. By comparison, a recent study of healthcare costs in acute upper gastrointestinal haemorrhage suggested around 57,000 patient presentations per year in the UK, costing a total of £155,500,000 in initial hospital care [21]. Whilst clinicians and policy makers may find the figures in this paper helpful, it should be acknowledged that they are representative of a single country only, and furthermore a country, which in the NHS, has a unique healthcare financing arrangement and system. The detailed breakdown of resource use will allow readers to determine whether the healthcare provided is representative of their own settings, however caution should still be exercised when generalising the findings beyond the NHS. Additionally, and as with any costing study, the nature of the data presented, that is the point estimates isolated from contextual variability that may exist, should also be borne in mind.
Confidence intervals around total costs in this study were estimated using the Central Limit Theorem (CLT) approach, which relies on sample mean costs being normally distributed. The suitability of CLT when data are skewed (which cost data invariably are) has been questioned, and the bootstrap approach has been proposed as an alternative. Recent studies have suggested however, that even with high levels of skewness, provided the study sample is moderate to large (n >50), CLT can still provide accurate estimates of the standard error and accompanying confidence intervals [22]. With a study sample of 441, this approach to confidence interval estimation was considered appropriate.
As few studies have comprehensively assessed the costs associated with bleeding trauma patients, it is difficult to set the estimates presented here within the  [23]. With high estimated average costs per additional case (£16,720 per patient aged 65 years and above), this implies significant additional costs of £11,200,000 for in-hospital treatment alone. Future public health initiatives in trauma prevention programmes might be targeted at the elderly, to offer a potential means of reducing this future cost burden.
The resource use and cost estimates generated by this study should prove useful as a baseline for a variety of stakeholders. Scaling up blood component wastage data from this study (Table 3) to national levels, suggests that approximately 5,759 units of the PRBCs and 6,538 units of the FFP issued to severely bleeding trauma patients annually are wasted, at costs of £710,194 and £179,526, respectively. Although challenging to modify in an emergency environment, information about wastage should be presented in multidisciplinary meetings between trauma and transfusion laboratory staff.
These data also provide a baseline from which to ascertain the cost-effectiveness of new interventions or transfusion protocols where the potential incremental cost implications can be complex and difficult to anticipate. For example, analysis of this study cohort (details available on request) showed almost a quarter of patients did not receive FFP alongside PRBCs which is now recommended in haemostatic resuscitation guidelines [11,24]. Indeed a recently published trial advocated use of FFP, platelets, and PRBCs in a 1:1:1 ratio as a safe means of reducing death from exsanguination at 24 hours [8]. Whilst increasing adherence to such protocols is likely and can improve haemostasis rates, the impact on PRBC transfusion rates and overall patient survival is less clear. The data here for example, show that increasing or reducing PRBC transfusion by just one unit per patient for example could cost or save almost £1,000,000 (7,783 units at £123.31 per unit) at the national level. Furthermore, it is essential that additional costs that may manifest through any improved patient survival are also fully acknowledged. For example, patients in this cohort surviving the first 24 hours, incurred average healthcare costs that were £18,884 (95 % CI: £14,892 to £22,876, P <0.01) greater than patients who died within the first 24 hours (data not shown).
This study has limitations. There were missing data (Additional file 1: Table A1), but largely for low cost items, and established statistical techniques were used for imputation. Also, only care received by patients at study hospitals was considered, and so re-admissions to other hospitals were not captured. Sensitivity analysis was used to explore the potential impact of an underestimating of hospital readmissions, and showed that even a three-fold increase in the rate observed increased total costs only slightly. Finally, the cost estimates presented here can only be considered an underestimate of the true costs to the NHS of treating severely bleeding trauma patients, as they do not include any day-case, outpatient, and primary care contacts during the 12 months following injury, which for this patient group are likely to be numerous. A recent study estimating comprehensive 12-month healthcare costs of a cohort of trauma patients reported that post-discharge costs accounted for 40 % of the total annual costs incurred [25]. If this finding were transferrable to the UK, total 12-month costs to the NHS of treating a major trauma patient with severe bleeding could be as high as £33,000.

Conclusions
Severely bleeding major trauma patients are a small yet high cost subgroup of all major trauma patients, and costs are projected to rise further as a consequence of an aging population, and as evidence continues to emerge on the benefits of early and simultaneous administration of blood products in pre-specified ratios. The findings from this study provide a previously unreported baseline from which the potential impact of changes to service provision and/or treatment practice can begin to be evaluated. Further studies are required to determine the costs of post-discharge out-of-hospital care requirements, which are also likely to be substantial.

Key messages
This is the first study to use detailed resource use data from TARN, the independent monitor of trauma care in England and Wales, to estimate comprehensive treatment costs for severely bleeding trauma patients. The average cost of initial hospital treatment for a severely bleeding trauma patient is estimated to be £19,770 (SE: £810), of which almost 62 % is