Immediate computed tomography or admission for observation after mild head injury: cost comparison in randomised controlled trial

Study design

The study had two parts, one on medical outcome and one on costs. Power calculations were performed for the first part. Though a much smaller study would be needed to detect the expected difference of around 30% of cost between the methods, we extended the cost analysis to the whole study.

In total, 2602 patients (aged ≥ 6) were randomly assigned to computed tomography (n = 1316) or observation in hospital (n = 1286). Emergency departments at 39 hospitals throughout Sweden recruited consenting patients from May 2001 through January 2004. Inclusion criteria for the study were mild head injury defined as loss of consciousness or amnesia, or both; normal score on the Glasgow coma scale (15 points); and normal neurological findings on arrival after the head trauma. The two groups of patients were similar at baseline as regards age, sex distribution, and time from trauma to randomisation. The figure shows the flow of participants through the study. The analysis of outcome data was by intention to treat. Of the 1316 patients allocated to computed tomography, 1292 received it and 117 were admitted for observation; and of the 1286 patients allocated to observation in hospital, 1220 were admitted and 111 also had computed tomography. We assessed survival, functioning, and wellbeing after three months using the extended Glasgow outcome scale (GOS-E).3 The number of patients who had not fully recovered (GOS-E 1-7) was similar for the two strategies of care, even when only the worst outcomes were considered.

Flow of participants through study and the medical results

• Download figure
• Open in new tab
• Download powerpoint

Data collection

Researchers prospectively registered the time of trauma, arrival, computed tomography, admission, discharge, and use of resources in the case report form for each patient. They also prospectively registered clinical course—that is, complications, change in care, and treatment. At follow-up after three months, we posted questionnaires to explore outcome (assessed by the extended Glasgow outcome scale) and included questions on outpatient care, inpatient care, and sick leave from work because of the injury. After reminders and telephone calls, the response rate was 97%. In the remainder we checked survival and hospital care through official registers.

Costing

The Swedish healthcare system is based on administratively and financially independent county councils and municipalities. Every hospital in Sweden, regardless of its category or size, has responsibility for basic medical and surgical care within a defined primary catchment area. The hospitals are university/regional, county, or local. We compared costs of two alternative management strategies from a societal perspective. Use of resources was translated into costs by use of average costs based on accounting data in official statistics from the county councils.

All patients had an outpatient visit to the emergency department—that is, the same cost for both groups. Differences of costs appeared after the allocation to computed tomography or observation in hospital. The cost of one scan was based on a survey of 57 Swedish hospitals.4 The arithmetic mean cost for each hospital category was calculated for office hours (Monday-Friday, 7 am-5 pm) and out of office hours (all other hours and weekends). The lower rate of use during nights and weekends increases the cost per examination. We individually assigned costs of computed tomography depending on the time of examination and the category of hospital (table 1).

The cost of observation in hospital was also individually assigned and was defined as the cost for room and board5 best corresponding to resources used by patients with mild head injury. It includes staff salaries, costs of housing, and drugs but excludes costs for surgery, intensive care, radiology, or laboratory analyses. We could not get relevant accounting data from every included hospital so we used mean costs from several hospitals at each level. Financial reports from 28 Swedish hospitals in 1978, 1988, and 2002 showed that the structure of costs was steady over time. Room and board constituted about 60% of the average cost per day (61% in 2002).6 We therefore multiplied the mean cost per day on surgical wards at regional/university, county, and local hospitals by 0.61 to achieve the cost for room and board (table 2).

For the follow-up period we collected itemised costs for health care from county council estimates and statistics (table 3). Indirect costs were defined as the estimated loss in production (excluding students) caused by days of absence from work, including that of parents as a result of injury in a child, and were calculated individually. Cost for absence was calculated as the average daily labour cost, payroll taxes included, and stratified for sex and age, for patients aged < 65.

We have expressed all costs at 2003 prices and inflated cost data from earlier years to 2003 with the Swedish price index for wages and for county councils.¹⁰¹¹ Costs were not discounted as the follow-up time was limited to three months. We converted costs in Swedish kronor (SKr) to euros adjusted for purchasing power parity at the rate of 1 SKr = €0.092—that is, the average for 2003 (the exchange rate of euro v SKr increased 1.75% from 2003 to 2005).12

Statistical analysis

We compared costs between the two treatment strategies for arithmetic mean costs, as suggested for pragmatic randomised trials.13 The t test for independent groups, modified for unequal variances, is often used but requires approximately normal distributions. However, if the data are not normally distributed, the t indicator can be used by means of a bootstrapping technique and thereby reliable P values and confidence intervals can be achieved.¹⁴¹⁵ Hence, we applied naive groupwise bootstrapping (that is, imputing the mean of the observed values for each missing observation) with n = 1000 resampled estimates. For a sensitivity analysis, we use the first and third quartile costs for the two most important cost drivers—that is, cost for computed tomography and cost for observation in hospital at hospitals in the three categories

Acute stage and complications

For patients randomised to the computed tomography strategy, the total cost during the acute stage was €607 704 (table 4), or €461 per patient (table 5). For patients admitted for observation the corresponding figures were €869 743 or €677 per patient. Hence, the average cost per patient was 32% or €216 less in the computed tomography group (95% confidence interval - 272 to - 164; P < 0.001). The sensitivity analysis showed the difference to be at least €189 (first quartile cost) and at most €279 (third quartile cost) per patient.

Costs for complications and readmissions are included in the above figures. Four neurosurgical operations for subdural haematoma (one in the computed tomography group and three in the observation group) were performed among all 2602 patients. A similar number of days of observation on a neurosurgical ward and days of readmittance for sequelae were needed in both groups. The computed tomography strategy generated more neurosurgical consultations, mostly by telephone. Because there were few complications, their contribution (neurosurgical consultation, operation and care, and observation) to the total cost per patient was limited to €26 in the computed tomography group and €36 in observation group.

Follow-up

During the three month follow-up period the cost for the computed tomography group was €337 943 v €305 596 for the observation group (table 4). This corresponds to a cost per patient of €257 v €237 (table 5).

Total cost

The total average cost of the acute stage, including complications, was €718 per patient for the computed tomography group v €914 per patient for the observation group (table 5), a difference of €196 (95% confidence interval −281 to −114; P < 0.001). Sensitivity analysis, based on the first and the third quartiles, respectively, showed a difference from €185 to €214 per patient in favour of the computed tomography strategy. Hence, the lower cost of the computed tomography strategy in the acute stage remained largely unchanged during the follow-up period.

International comparison of cost data

A study in the Netherlands showed that nursing costs account for 32% to 64% of the total cost for a day in hospital.16 In the same study, the joint costs of overheads, equipment, accommodation, and housekeeping were 22% to 30% of the total cost. The ratio of average daily costs of hospitals in different categories also resembled that of Swedish hospitals. Hence, the cost structure used for calculation of observation in hospital in this analysis is also found in other countries.

We found that using computed tomography rather than admission for observation in the acute phases reduced costs by about 32%. This is comparable with estimates from model studies in the United States, Spain, and Norway.^17–19 Our own decision tree analysis based on Swedish hospital data showed similar results, suggesting a cost reduction of 36%.2

About 17 000 people with mild head injury are admitted to hospital each year in Sweden, equivalent to about 175/100 000 inhabitants.20 In about 22% of cases, current clinical management includes computed tomography.21 Out of 17 000 patients, around 15 000 are aged ≥ 6 years—that is, the age groups included in our study. If the computed tomography strategy were to be implemented, there would be a yearly increase in demand of about 130 scans per 100 000 inhabitants (two to three a week at an emergency department serving 100 000 inhabitants). Such an increase seems possible to handle within the present capacity for computed tomography in Sweden. The savings of about 165 days of hospital stay (three to four a week) per 100 000 inhabitants could be used for other patients for whom there is a shortage of bed days. A similar decrease in admissions in hospital by introduction of the computed tomography strategy has been estimated for the United States.22

Observation in hospital will probably always be necessary for some patients, such as elderly people living alone. For others, the computed tomography strategy represents positive value for patients, with the opportunity to receive medical assurance and to be discharged several hours earlier. For most patients with mild head injury, this represents a satisfactory and cost effective treatment option.