- Email: [email protected]
The cost-effectiveness of diagnostic management strategies for adults with minor head injury
Injury, Int. J. Care Injured 43 (2012) 1423–1431
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
The cost-effectiveness of diagnostic management strategies for adults with minor head injury M.W. Holmes *, S. Goodacre, M.D. Stevenson, A. Pandor, A. Pickering School of Health and Related Research, The University of Sheffield, United Kingdom
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 18 July 2011
Study objective: To estimate the cost-effectiveness of diagnostic management strategies for adults with minor head injury. Methods: A mathematical model was constructed to evaluate the incremental costs and effectiveness (Quality Adjusted Life years Gained, QALYs) of ten diagnostic management strategies for adults with minor head injuries. Secondary analyses were undertaken to determine the cost-effectiveness of hospital admission compared to discharge home and to explore the cost-effectiveness of strategies when no responsible adult was available to observe the patient after discharge. Results: The apparent optimal strategy was based on the high and medium risk Canadian CT Head Rule (CCHRhm), although the costs and outcomes associated with each strategy were broadly similar. Hospital admission for patients with non-neurosurgical injury on CT dominated discharge home, whilst hospital admission for clinically normal patients with a normal CT was not cost-effective compared to discharge home with or without a responsible adult at £39 and £2.5 million per QALY, respectively. A selective CT strategy with discharge home if the CT scan was normal remained optimal compared to not investigating or CT scanning all patients when there was no responsible adult available to observe them after discharge. Conclusion: Our economic analysis confirms that the recent extension of access to CT scanning for minor head injury is appropriate. Liberal use of CT scanning based on a high sensitivity decision rule is not only effective but also cost-saving. The cost of CT scanning is very small compared to the estimated cost of caring for patients with brain injury worsened by delayed treatment. It is recommended therefore that all hospitals receiving patients with minor head injury should have unrestricted access to CT scanning for use in conjunction with evidence based guidelines. Provisionally the CCHRhm decision rule appears to be the best strategy although there is considerable uncertainty around the optimal decision rule. However, the CCHRhm rule appears to be the most widely validated and it therefore seems appropriate to conclude that the CCHRhm rule has the best evidence to support its use. ß 2011 Elsevier Ltd. All rights reserved.
Keywords: Minor head injuries Cost-effectiveness Diagnostic management strategies Quality adjusted life-years Economic modelling
Introduction Background Head injury accounts for around 700,000 emergency department (ED) attendances each year in England and Wales,1 90% of which may be categorised as apparently minor on the basis of having a Glasgow Coma Score (GCS) of 13–15 at presentation to hospital.2 The costs of liberal CT scanning and hospital admission are therefore substantial. However, if restriction of investigation
* Corresponding author at: School of Health and Related Research, Regents Court, 30 Regent Street, Sheffield S1 4DA, United Kingdom. Tel.: +44 0114 2220745; fax: +44 0114 2724095. E-mail address: [email protected] (M.W. Holmes). 0020–1383/$ – see front matter ß 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2011.07.017
and observation leads to delayed treatment of neurosurgical injury then the consequences may be devastating and include death or severe long-term disability. The main challenge in the management of minor head injury is identification of the minority of patients with significant intracranial injury, especially those who require urgent neurosurgery. Increased access to CT scanning in recent years has reduced the risk of missed pathology but has raised concerns about increased radiation exposure and inappropriate use of health care resources. Clinical decision rules have been developed in an attempt to limit the use of CT scanning whilst limiting the risk of missed pathology.1,3–7 It is not clear, however, which existing clinical decision rule achieves the best balance of increasing the benefits of scanning whilst reducing costs and harms, and how these rules compare with CT scanning all, or no patients.
1424
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Two previous studies have compared the cost-effectiveness of decision rules but both have limitations.8,9 One did not include the risk of cancer from CT scans and none were from the perspective of the United Kingdom (UK) National Health Service (NHS). Furthermore, these analyses did not explore decisions beyond CT scanning, such as whether or not to admit a patient with a normal CT scan. Data from the UK have suggested that increased use of CT scanning for minor head injury was associated with increased hospital admission.10 This may be due to admission of patients with non-neurosurgical abnormalities on CT scan or for patients with normal CT scans, but without a responsible adult to observe the patient after discharge. Aims of this investigation The goals of this investigation were to: (1) determine the optimal existing CT scanning diagnostic strategy in terms of costeffectiveness; (2) determine the cost-effectiveness of admission strategies consequent upon CT scanning. Materials and methods Diagnostic performance of clinical decision rules A systematic review of the literature was undertaken to evaluate the diagnostic performance of clinical decision rules.11 The literature review identified that most clinical decision rules for adults had estimates of diagnostic parameters from validation cohorts, although often in different settings from the derivation cohort. Decision rules would be expected to perform better in the derivation cohort and additionally in a validation cohort within the same setting as the derivation cohort. Therefore a validation study undertaken in a different setting may provide a more appropriate estimate of diagnostic performance. However, using different cohorts to estimate parameters for different decision rules could introduce selection bias. For these reasons we decided to use sensitivity and specificity data from a validation study by Stein et al., as this was a large, unselected cohort in which all of the main clinical decision rules were validated.12 Table 1 shows the sensitivity and specificity of each strategy used as probabilities in the model and to which decision nodes these apply.
The mathematical model For the model building process, we first identified the pathways for the model using the clinical opinion and modelling expertise within our team and with consideration of the data that was identified in a systematic review.11 Once the pathways were agreed we decided that a decision tree type model would be most appropriate. The model was constructed using Simul8 software (Simul8 Corporation). The model estimates the costs and qualityadjusted life-years gained (QALYs) for the following diagnostic strategies: discharge all patients without testing (theoretical), CT all patients (theoretical), CT only if GCS less than 15 on arrival (theoretical), Canadian CT Head Rule high risk3 (CCHRh), Canadian CT Head Rule high or medium risk3 (CCHRhm), Neurotraumatology Committee of the World Federation of Neurosurgical Societies4 (NCWFNS), New Orleans Criteria5 (NOC), National Emergency XRadiography Utilization Study II6 (NEXUS II), National Institute for Health and Clinical Excellence1 (NICE) and the Scandinavian rule.7 The analysis was conducted for patients aged 40 and 75 years presenting to the ED. A lifetime horizon was used with mean life expectancy taken from UK interim life tables,13 averaging the value for males and females. The economic perspective of the model is the UK National Health Service (NHS) and Personal social services (PSS) in England and Wales. This perspective was chosen as it conforms to the National Institute for Health and Clinical Excellence (NICE) guidelines for the methods of Technology Appraisals.14 A simplified schematic of the decision-analytic model is presented in Fig. 1. Each strategy was applied to a hypothetical cohort of 10,000 patients attending the ED with isolated, closed minor head injury (i.e. with no other serious injuries). A proportion of the cohort was assumed to have an intracranial lesion requiring immediate neurosurgery (an intracranial haematoma requiring evacuation); a further proportion was assumed to have an intracranial lesion that did not initially require immediate neurosurgery. For convenience, we refer to the former patients as having a neurosurgical lesion and the latter as having a non-neurosurgical lesion, although it should be recognised that the latter may ultimately receive neurosurgical intervention. The remainder would have no intracranial haemorrhage. These proportions were estimated from the study of patients with minor head injury by Smits et al. (lesion requiring
Table 1 Sensitivity and specificity of CT decision rules in correctly determining that a CT scan was required predicated on the assumption that CT scanning has 100% sensitivity and 100% specificity in identifying significant intracranial lesions.* Decision node Sensitivity for NS injurya
1-Sensitivity for NS injuryb
Sensitivity for NNS injuryc
Sensitivity or specificity of each strategy to detect intracranial injury (95% CI) Discharge all 0 1 0 CT all 1 0 1 Abnormal arrival GCS 0.91 (0.84–0.95) 0.09 0.72 (0.68–0.76) CCHRh3 0.99 (0.94–1.0) 0.01 0.97 (0.94–0.98) 3 CCHRhm 0.99 (0.94–1.0) 0.01 0.99 (0.97–1.0) 4 NCWFNS 0.99 (0.94–1.0) 0.01 0.95 (0.93–0.97) 5 NOC 0.99 (0.94–1.0) 0.01 0.99 (0.97–1.0) NEXUS II6 1.0 (0.97–1.0) 0 0.97 (0.94–0.98) NICE1 0.98 (0.93–1.0) 0.02 1.0 (0.99–1.0) Scandinavian7 0.99 (0.94–1.0) 0.01 0.95 (0.92–0.97)
1-Sensitivity for NNS injuryd
Specificitye
1-Specificityf
1 0 0.28 0.03 0.01 0.05 0.01 0.03 0 0.05
1 0 0.97 0.51 0.47 0.47 0.33 0.47 0.31 0.53
0 1 0.03 0.49 0.53 0.53 0.67 0.53 0.69 0.47
(0.96–0.98) (0.49–0.52) (0.46–0.48) (0.46–0.48) (0.32–0.34) (0.46–0.48) (0.30–0.32) (0.52–0.54)
NS, neurosurgical; NNS, non-neurosurgical. Beta distributions were used in the probabilistic sensitivity analysis for all parameters. All letters refer to Fig. 1. a The probability that a neurosurgical lesion is correctly identified, these patients receive prompt treatment. b The probability that a neurosurgical lesion is not correctly identified, these patients receive delayed treatment. c The probability that a non-neurosurgical lesion is correctly identified, these patients are admitted and those that deteriorate are given prompt treatment. d The probability that a non-neurosurgical lesion is not correctly identified. These patients are discharged and those that deteriorate are given delayed treatment. e The probability that patients with no intracranial lesion are correctly identified and discharged. f The probability that patients with no intracranial lesion are incorrectly identified and are given a CT scan. * All estimates have been extracted from Stein et al.12
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1425
Fig. 1. Model pathways.
neurosurgery 0.53%, 95% CI 0.33–0.85%; lesion initially not requiring neurosurgery 7.1%, 95% CI 6.26–8.05%).15 The number of patients undergoing CT scanning was dependent on the diagnostic strategy and ranged from 0% (discharge all patients without testing) to 100% (CT scan all patients). The probability of detecting a neurosurgical lesion was determined by the sensitivity of the diagnostic strategy. We assumed that patients with a neurosurgical lesion which was detected on CT would receive prompt evacuation of their intracranial haematoma (before any deterioration occurred); where a patient with a neurosurgical lesion did not have a CT scan, delayed evacuation (after deterioration had occurred) was assumed. We assumed that a proportion of patients with a nonneurosurgical lesion would deteriorate over the following 48 h and require intervention (critical care support and/or neurosurgery), whilst the remainder would remain well. These proportions were taken from a study by Fabbri et al.16 We assumed that if the
strategy led to CT being performed and the lesion detected then the patient would be admitted to hospital and would receive prompt appropriate treatment. If the strategy did not lead to CT being performed we assumed that the patient would be discharged home and would receive delayed treatment. We assumed that patients without an intracranial lesion remained well. The proportion of these patients receiving unnecessary CT was determined by the specificity of each strategy. Table 1 shows the sensitivity and specificity parameters of each strategy in correctly determining that a CT scan was required predicated on the assumption that CT scanning has 100% sensitivity and 100% specificity in identifying significant intracranial lesions. Only relevant lesions were those related to the head injury (i.e. we did not consider incidental findings unrelated to the injury). The CT all strategy therefore had 100% sensitivity and 100% specificity for both neurosurgical and non-neurosurgical lesions. The discharge all strategy had zero sensitivity and 100% specificity.
1426
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Although patients with an initially normal CT can subsequently develop significant injury and deteriorate, the rate of this occurrence is very low and unlikely to influence the costeffectiveness of strategies for selecting patients for CT.17
from the UK Personal Social Services Research Unit (PSSRU).23 The cost of glioma including the cost of palliative care for terminally ill patients was taken from the only relevant and reliable data source identified.24 Costs used in the model are shown in Table 2.
Patient outcome
Quality of life utility values
Patient outcome is measured by the Glasgow Outcome Score (GOS) with the following categories: GOS 1, Dead; GOS 2, Vegetative state; GOS 3, Severe disability; GOS 4, Moderate disability; GOS 5, Good recovery. The model allocated each patient to a GOS category based on whether they had an intracranial lesion and how quickly it was treated. This involved estimating the probabilities that patients with neurosurgical and non-neurosurgical lesions would end up in each GOS category depending upon the extent of treatment delay. For prompt treatment of patients with a neurosurgical injury we meta-analysed the data from 5 studies identified from a systematic review of the literature.11 For delayed treatment of patients with a neurosurgical injury we estimated GOS outcomes using a study by Haselberger et al., which showed the association between outcome and time delay from loss of consciousness to operation.18 For patients with a non-neurosurgical lesion where the lesion is detected on CT and the patient admitted and treated appropriately, we based estimates of GOS outcomes on a study by Fabbri et al.16 For patients with a non-neurosurgical lesion where a CT is not performed and the patient is discharged home without appropriate treatment, we did not identify any studies that reported the effect of treatment delay upon outcome. We therefore assumed that the delay had a similar effect as in the treatment of neurosurgical lesions and adjusted the GOS outcomes from Fabbri et al., accordingly.16 The distribution amongst GOS outcomes is provided in Table 2.
A literature review was conducted to identify studies that estimated health related utility values for GOS states. Two studies were found. Smits et al. obtained long-term GOS outcomes and quality of life (QoL) scores using the EQ-5D questionnaire from a subset of patients (n = 87) from the CT in head injury patients study.9 Aoki et al. used standard gamble methods to elicit quality of life utilities for GOS states 2–5, from 140 members of staff and students at a hospital in Japan.25 These studies were assessed for methodological compliance with the NICE reference case which stipulates that utilities should be measured in patients using a generic and validated classification system for which reliable UK population reference values, elicited using a choice based method such as the time trade-off or standard gamble are available.14 The Smits et al. study was considered to comply most closely with the NICE reference case and therefore the utility values from this study were used.9,14 The Smits et al. study did not report the age distribution of those patients used to estimate QoL utilities.9 We assumed that QoL values within GOS outcomes 3–5 were not age-related. The QoL utility values for death and persistent vegetative state were assumed to be 0. The utilities used in the model are shown in Table 2. Costs and utilities were discounted at an annual rate of 3.5%, as recommended by NICE.14
Costs
CT scans expose the patient to radiation which causes cancer in a proportion of patients. The estimated risk of a tumour from a single CT scan is taken from a study by Stein et al., which also provides an estimated QoL decrement risk per CT Scan.26 This study was limited to patients aged 20 years and younger; data from ages 5–20 years reported in Stein et al., were used to predict values for tumour risk and QALY loss for older patients. The statistical relationship resulted in tumour risk being equal to 0.0077 multiplied by age (in years) 1.076 (R2 = 0.92). The predicted tumour risk from a single CT scan for ages 35 and above was 0.0001 which corresponds with the best available evidence for the lifetime risk of cancer in adult patients.27 The predicted QoL decrement estimate from a single CT scan equalled 0.0401 multiplied by age (in years) 0.675 (R2 = 0.99), giving disutilities of 0.003 and 0.002 for ages 40 and 75 years respectively. The standard deviation associated with these values was predicted and were 0.0008 and 0.0007, at 40 and 75 years respectively. The costs of cancer associated with a CT scan were assumed to equal the risk of a tumour multiplied by the mean cost of cancer (£26,738).24 The Stein et al. study reports that the latency between radiation exposure and tumour diagnosis is over five years in the majority of cases.26 We have assumed a mean latency period of ten years with the cost of cancer subject to discounting.
Costs included in the model are the direct costs of diagnostic management including the costs of investigation, CT scanning, hospital admission and the subsequent costs of providing neurosurgical treatment, the cost of intensive care, nursing home care, rehabilitation for the severely disabled and the cost of glioma. For each hospital care cost, the most appropriate Health Resource Group (HRG) version 4 code was identified and cross referenced with the UK Department of Health National Schedule of Reference Costs (NSRC).19,20 The NSRC provides summary statistics (mean, range, inter-quartile range) from all Trusts for each HRG. The NSRC captures the overall cost of a treatment episode including length of stay. Patients with outcome GOS 5 were assumed well and did not incur any further costs. A literature review was conducted to find long-term costs of care and rehabilitation for patients whose outcomes are represented by GOS states 2–4. Only one study was identified which aimed to identify the health and social care services used by young adults aged 18–25 years with acquired brain injury.21 The study estimated average costs per person in four groups of patients, two of which correspond closely to the descriptions of GOS states 3 and 4. We were unable to find any cost data for older people and have therefore assumed the costs are the same as for age 18–25 years. The study did not provide cost data for patients in a vegetative state. We have therefore based our estimates on expert opinion assuming two weeks in intensive care followed by four months of rehabilitation before transferring to a nursing home for the rest of the patient’s life. (Sophie Duport, Royal Hospital for Neuro Disability, Personal communication, 2010). The Multi-Society Task Force on Persistent Vegetative State reported the mean length of survival for adults in a vegetative state (GOS 2) as 3.6 years which was used in the model.22 Nursing home costs were taken from the NSRC20 whilst rehabilitation costs were taken
Cancer risk from CT scans
Additional clinical questions In addition to an evaluation of the published strategies the following four clinical questions were assessed. 1. Is it cost-effective to admit patients with a normal CT scan compared to discharge home with a responsible adult? These patients are estimated to have a very low (0.006%) risk of deterioration.17
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1427
Table 2 Costs, parameters and distributions used in the PSA. Parameter Clinical outcomes Neurosurgical injury Non-neurosurgical injury Neurosurgical lesion: GOS outcomes after immediate intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Neurosurgical lesion: GOS outcomes after late intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Non-neurosurgical lesion: GOS outcomes after immediate intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Non-neurosurgical lesion: GOS outcomes after late intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 QoL utilities GOS 3 GOS 4 GOS 5 Age related effect of a single head CT scan on tumour occurrence Age 40 and 75 Age related effect of a single head CT scan on QoL decrement Age 40 Age 75 Cancer latency (years) PVS mean survival Age 40 and 75 Costs ED visit CT scan Admission with no deterioration or neurosurgery. Neurosurgical intervention after deterioration. Neurosurgical intervention before deterioration. Long-term costs GOS 4 Long-term costs GOS 3 GOS 2 intensive care GOS 2 rehabilitation GOS 2 nursing home Cost of cancer
Mean probability or percent (%)
95% CI probability or percent (%)
Distribution
Source
0.0053 0.0710
0.0033–0.0085 0.0626–0.0805
Beta Beta
15 15 11
81.0% 9.3% 3.2% 2.7% 3.8%
74.7–86.1% 5.6–13.9% 1.2–6.3% 0.9–5.5% 1.6–7.9%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
57.0% 6.8% 12.0% 9.9% 14.3%
7.3–87.5% 0.8–12.4% 0.9–38.2% 0.7–33.2% 1.1–43.1%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
81.2% 11.1% 6.8% 0% 0.9%
73.2–87.2% 6.6–18.1% 3.5–12.9% 0–3.2% 0.2–4.7%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
55.9% 8.2% 27.6% 2.6% 5.4%
7.2–85.8% 1.0–15.7% 2.1–74.1% 0–15.8% 0.2–24.7%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
0.15 0.51 0.88
0.06, 0.28 0.39, 0.63 0.74, 0.97
Beta Beta Beta
0.0001
0.00009–0.00011
Normal
0.003 0.002 10
0.009–0.005 0.001–0.004 5–10
Beta Beta Normal
3.59
3.23–3.95 95% CI £67–170 £80–117 £490–997 £3,605–6616 £3758–6374 £15,444–20,592 £30,510–40,680 £12,781–17,561 £25,164–33,552 £804–1072 £24,064–29,412
Normal
18
16
16
9
26 26
£126 £100 £847 £5805 £5273 £17,160 £33,900 £15,469 £27,960 £893/week £26,738
Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma
26 22
19,20 19,20 19,20 19,20 19,20 21 21 19,20 23 20 24
PSA, probabilistic sensitivity analysis; GOS, Glasgow Outcome Score; CI, confidence interval; QoL, quality of life; PVS, persistent vegetative state. For probabilities of each strategies sensitivity and specificity see Table 1.
2. Is it cost-effective to admit patients with a normal CT scan compared to discharge home without a responsible adult? 3. Is it cost-effective to admit patients with a non-neurosurgical lesion on CT scan? These patients have a significant (13.5%) risk of deterioration.17 4. Does the optimal strategy change when a patient is discharged with no responsible adult? It is assumed that patients with a missed neurosurgical or non-neurosurgical lesion who deteriorate will not receive treatment and die. This contrasts with the base case where patients who deteriorate are admitted and receive delayed treatment. Analysis undertaken Deterministic results are presented first using incremental cost-effectiveness ratios (ICER).
Univariate sensitivity analysis was undertaken to explore the impact of changing each parameter in the model to its lower then higher confidence interval and in altering the discount rates for both costs and benefits to 0% and 6% per annum. Probabilistic sensitivity analysis (PSA) was undertaken to explore the impact of the joint uncertainty surrounding all model parameters.28 Definition of cost-effectiveness terms The ICER measures the relative value of two strategies and is calculated as the difference in mean costs divided by the difference in mean benefits. Where a strategy is less effective and more expensive than its comparator it is dominated. Extended dominance occurs when a combination of two alternative strategies can produce the same QALYs as a chosen strategy but at a lower cost.29
1428
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Strategies that are neither dominated nor extendedly dominated constitute the cost-effectiveness frontier, and the ICER is reported for these strategies compared with the next least effective strategy. In a PSA each parameter in the model is assigned a distribution which encapsulates the uncertainty within the parameter. Beta distributions were used for probabilities as this distribution is bounded between 0 and 1, Gamma distributions were used for costs as the Gamma can describe the potential skewness of costs, Dirichlet distributions were used for GOS states as the states are correlated and must sum to 1, normal distributions were used for all other parameters. For each model iteration (1000 simulations each with 10,000 patients) each model parameter is randomly sampled from the relevant distribution. Results are presented as mean and incremental costs and QALYs, and ICERs. A costeffectiveness acceptability curve (CEAC) is calculated using the probability that a strategy has the highest net benefit (NB) for each iteration of the model for a range of willingness to pay (WTP) values.30 Net benefit is defined as: NB = WTP QALY-cost. Table 2 shows the parameters and distributions used in the PSA.
Table 3 Probabilistic analysis for age 40.
Discharge all Abnormal arrival GCS CT all NCWFNS NICE Scandinavian NOC NEXUS II CCHRh CCHRhm
Mean costs
Mean QALYs
ICERa
£3305 £2992 £2963 £2928 £2932 £2923 £2929 £2922 £2919 £2916
18.6669 18.6859 18.6897 18.6907 18.6907 18.6908 18.6908 18.6909 18.6910 18.6913
Dom Dom Dom Dom Dom Dom Dom Dom Dom Dominant strategy
QALY, quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, costeffectiveness; Dom, dominated. a Compared with next last effective treatment on the CE frontier.
Univariate sensitivity analysis
Results
For all ages, no parameter change altered the optimal strategy decision.
Deterministic analysis
Probabilistic sensitivity analysis
For adults aged 40, the CCHRhm strategy had an ICER of £3879 when compared to the Scandinavian strategy (the only other strategy on the cost-effectiveness frontier). All other strategies were either dominated or extendedly dominated. For adults aged 75, the CCHRhm strategy had an ICER of £10,397 when compared to the Scandinavian strategy (the only other strategy on the costeffectiveness frontier). All other strategies were dominated. Although the CCHRhm strategy was estimated to be optimal the absolute values for costs and QALYs were similar for all selective CT strategies.
Table 3 shows, for adults aged 40 years at presentation, the mean costs and QALYs per patient and which decision rules lie on the cost-effectiveness frontier. The strategies are ordered by ascending effectiveness (QALYs gained). Incremental costs and QALYs are not shown as the CCHRhm strategy dominates all other strategies, although the selective CT strategies had broadly similar costs and QALYs. Fig. 2 shows, for age 40, the CEAC which plots the probability of each strategy being the most cost-effective against values of WTP for a QALY ranging from zero to £50,000. Typical thresholds for
Fig. 2. Cost-effectiveness acceptability curve for adult aged 40.
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1429
Table 4 Probabilistic analysis for age 75.
Discharge all Abnormal arrival GCS CT all NICE New Orleans NCWFNS Scandinavian NEXUS CCHRh CCHRhm
Mean costs
Mean QALYs
Incremental cost
Incremental QALYs
ICERa
£1718 £1546 £1574 £1541 £1539 £1532 £1527 £1529 £1525 £1526
7.8295 7.8374 7.8383 7.8390 7.8390 7.8391 7.8392 7.8392 7.8393 7.8393
n/a n/a n/a n/a n/a n/a n/a n/a n/a £0.17
n/a n/a n/a n/a n/a n/a n/a n/a n/a 0.1704
Dom Dom Dom Dom Dom Dom Dom Dom – £2666
QALY, quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, cost-effectiveness; Dom, dominated. a Compared with next last effective treatment on the CE frontier.
Table 5 Clinical question 4, probabilistic analysis for age 40 discharged without a responsible adult. Strategy
Mean cost
Mean QALY
Incremental cost
Incremental QALY
ICERa
Discharge all Abnormal arrival GCS NCWFNS Scandinavian NEXUS II CCHRh CT all NICE NOC CCHRhm
£2053 £2838 £2917 £2911 £2914 £2909 £2964 £2922 £2923 £2910
18.5517 18.6624 18.6878 18.6879 18.6895 18.6895 18.6897 18.6901 18.6903 18.6907
n/a n/a n/a n/a n/a n/a n/a n/a n/a £857
n/a n/a n/a n/a n/a n/a n/a n/a n/a 0.1390
– ExtDomb Dom Dom Dom ExtDomb Dom Dom Dom £6167
QALY, Quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, cost-effectiveness; Dom, dominated; ExtDom, extendedly dominated. a Compared with next least effective treatment on the CE frontier. b Extendedly dominated by CCHRhm and discharge all.
decision-making in the UK are £20,000–30,000 per QALY14. Strategies that never reach a probability of 10% have been removed for visual clarity. At all WTP values, the CCHRhm strategy had a higher probability of being cost-effective than any other strategy. However, a comparison of the best two strategies at age 40 indicated that there was considerable uncertainty in the results: in 63% of simulations CCHRh dominated CCHRhm, in 9% CCHRhm dominated CCHRh, in 16% CCHRhm was more expensive but provided more health benefits and in12% CCHRh was more expensive but provided more health benefits. When the CCHRhm was compared to the NEXUS II and NOC diagnostic rules, similar levels of uncertainty were seen. Table 4 reports the analysis for adults aged 75 years. Mean QALYs and mean costs are both lower than in the analysis for 40 year old adults, reflecting reduced life expectancy. All strategies except for CCHRh and CCHRhm are dominated. The ICER for CCHRhm versus CCHRh is £2666. Again, the results for selective CT strategies are broadly similar. The CEAC for age 75 is not shown as it is similar to the CEAC for age 40. At all WTP values, the CCHRhm strategy had a higher probability of being cost-effective than any other strategy. However, there was considerable uncertainty regarding the optimal strategy. Assessment of additional clinical questions The following results were found for the additional clinical questions: Question 1. A deterministic analysis indicates that hospital admission for patients with a normal CT scan compared to discharge home with a responsible adult gains on average only 0.00001 QALYs and costs an additional £441 per patient. The
ICER for admitting all patients compared to discharge home is approximately £40 million. Question 2. A deterministic analysis indicates that hospital admission for patients with a normal CT scan compared to discharge home without a responsible adult gains on average only 0.0002 QALYs and costs an additional £468 per patient. The ICER for admitting all patients compared to discharge home is approximately £2.5 million. Question 3. A deterministic analysis indicates that the admission strategy for those with a non-neurosurgical lesion costs approximately £340 less and gains 0.004 QALYs compared to discharge home and therefore dominates discharge home. Question 4. Table 5 shows the main probabilistic analysis for a 40 year old adult repeated with the assumption that patients who are discharged home have no responsible adult and are not brought to medical attention when they deteriorate. The discharge all strategy is cheaper and less effective than in the main model and is no longer dominated by other strategies. The CCHRhm strategy is most cost-effective with an ICER of £6167, compared with the discharge all strategy. Discussion Decision-analysis modelling is limited by both the need to make assumptions and the available data. Our estimates of the effect of delayed treatment upon intracranial pathology in particular are based on very limited observational data. In the UK, hospitals do not collect and record data on the use of resources by individual patients. The process of allocating costs starts with the annual financial returns of a hospital and this expenditure is then reallocated to patient treatment services in a ‘top down’ manner. This process varies across NHS Acute trusts and consequently some resources have
1430
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
large variability around their mean cost. However, these estimates are the best available and this variability in included in the PSA. We did not use age related QoL utilities. However, it is likely that the QoL lost through the ageing process will be proportionately comparable across all management strategies and the conclusions will be unaltered. Another potential limitation is using data from ages 5 to 20 years to predict QALY loss in adults. It is possible that our predictions are not sufficiently taking into account the effects of discounting which could mean our predictions are overestimating QALY loss in adults, especially the 75 year old patient. This limitation, however, is expected to have little effect as any inaccuracies around the QALY loss are likely to be small but would favour those policies that perform fewer CT scans. Productivity losses were not included as they are difficult to estimate and subject to substantial uncertainty. It is likely that including productivity losses would only strengthen our overall conclusions because they would simply add to the high costs of missing serious injury that drives the cost-effectiveness of liberal CT use. Finally, we assumed that the various strategies would be implemented in a consistent way. This may not be the case in practice and studies of the implementation of decision rules in practice are required. These limitations noted, our economic analysis showed that use of CT scanning as determined by a clinical decision rule is not only more effective than not investigating but is also cost saving. Selective CT use according to a clinical decision rule was also costeffective compared to CT for all patients. The most cost-effective rule was the CCHRhm criterion,3 however there was considerable uncertainty in the results when compared to the other rules. The costs and outcomes associated with each rule were broadly similar, so the superiority of the CCHRhm may simply reflect a small difference in the estimate of diagnostic accuracy which was not statistically significant in the primary data.12 However, the Canadian rules appear to be the most widely validated and have estimates of diagnostic accuracy that are reasonably consistent across a number of cohorts.3 Two USA studies assessed the costeffectiveness of various decision rules.8,9 Although the comparator strategies were different from our study, both USA studies agreed that the CCHR decision rule was cost-effective compared to other rules. Our analysis showed that admission of patients with a normal CT scan would not be cost-effective regardless of whether there was a responsible adult available to observe the patient. These results are based upon data suggesting a very low (0.006%) risk of deterioration and it is assumed that patients are clinically well and would not benefit from general hospital care.17 The conclusion that patients with a normal CT scan should not be admitted to hospital does not apply to those with, for example, repeated vomiting or significant amnesia who might benefit from symptomatic treatment, nursing care or a safe environment. However, if the patient is orientated, comfortable and able to selfcare then our analysis suggests that hospital admission for observation is not a cost-effective use of health care resources. Hospital admission for those with a non-neurosurgical lesion on CT was cost saving as the costs of long-term care for those who deteriorated and received late treatment outweighed the costs of hospital admission. This analysis was limited by the lack of a standard definition as to what constitutes a significant nonneurosurgical lesion on CT and the limited data relating to outcomes from non-neurosurgical lesions. The prognosis of different non-neurosurgical lesions varies markedly, so costeffectiveness could potentially be improved by selecting those at highest risk of deterioration for admission whilst discharging
those at lower risk. Currently, however, we do not have sufficient data to evaluate this approach. Conclusions Our economic analysis confirms that the recent extension of access to CT scanning for minor head injury is appropriate. Liberal use of CT scanning based on a high sensitivity decision rule is not only effective but also cost-saving. It is recommended therefore that all hospitals receiving patients with minor head injury should have unrestricted access to CT scanning for use in conjunction with evidence-based guidelines. Provisionally the CCHR decision rule appears to be the best strategy although there is considerable uncertainty around the results. Incremental changes in both costs and QALYs are very small compared to the other selective CT strategies. However, the CCHR appears to be the most widely validated rule and therefore it does not seem inappropriate to conclude that, currently, the CCHR rule has the best evidence to support its use. Hospital admission appears to be cost-effective for patients with an intracranial lesion on CT scanning but not for those with a normal CT. Conflict of interest The authors have no commercial associations or sources of support that might pose a conflict of interest. Acknowledgements This project was undertaken for the UK National Institute for Health Research Health Technology Assessment Program. The findings of this project, including the findings presented in this article, will be published as a report in the Health Technology Assessment monograph series: http://www.hta.ac.uk/project/ 1765.asp. The study sponsors had no involvement in the study design; collection, analysis and interpretation of data; the writing of the manuscript or the decision to submit the manuscript for publication. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR HTA. References 1. National Institute for Health and Clinical Excellence (NICE). Head injury – triage, assessment, investigation and early management of head injury in infants, children and adults (NICE Clinical Guideline No. 56). London: NICE; 2007. 2. Swann IJ, Teasdale GM. Current concepts in the management of patients with so-called ‘minor’ or ‘mild’ head injury. Trauma 1999;1:143–55. 3. Stiell IG, Wells GA, Vandemheen K, Clement C, Lesiuk H, Laupacis A, et al. The Canadian CT Head Rule for patients with minor head injury. Lancet 2001;357:1319–96. 4. Servadei F, Teasdale G, Merry G, Neurotraumatology Committee of the World Federation of Neurosurgical Societies. Defining acute mild head injury in adults: a proposal based on prognostic factors, diagnosis, and management. J Neurotrauma 2001;18:657–64. 5. Haydel MJ, Preston CA, Mills TJ, Luber S, Blaudeau E, DeBlieux MC. Indications for computed tomography in patients with minor head injury. N Eng J Med 2000;343:100–5. 6. Mower WRH, Hoffman HJ, Herbert M, Wolfson AB, Pollack C, Zucker M. Developing a decision instrument to guide computed tomographic imaging of blunt head injury patients. J Trauma Injury Infect Crit Care 2005;59:954–9. 7. Ingebrigtsen T, Romner B, Kock-Jensen C, Ingebrigtsen T, Romner B, Kock-Jensen C. Scandinavian guidelines for initial management of minimal, mild, and moderate head injuries. The Scandinavian Neurotrauma Committee. J Trauma Injury Infect Crit Care 2000;48:760–6. 8. Stein SC, Burnett MG, Glick HA. Indications for CT scanning in mild traumatic brain injury: a cost-effectiveness study. J Trauma Injury Infect Crit Care 2006;61:558–66. 9. Smits M, Dippel DW, Nederkoorn PJ, Dekker HM, Vos PE, Kool DR, et al. Minor head injury: CT-based strategies for management – a cost-effectiveness analysis. Radiology 2010;254:532–40.
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431 10. Goodacre S. Hospital admissions with head injury following publication of NICE guidance. Emerg Med J 2008;25:556–7. 11. Harnan SE, Pickering A, Pandor A, Goodacre SW. Clinical decision rules for adults with minor head injury: a systematic review. J Trauma 2011;71(1):245–51. 12. Stein SC, Fabbri A, Servadei F, Glick HA. A critical comparison of clinical decision instruments for computed tomographic scanning in mild closed traumatic brain injury in adolescents and adults. Ann Emerg Med 2009;53:180–8. 13. Office for National Statistics (ONS). United Kingdom, interim life tables 1980–82 to 2006–2008. London: ONS; 2008. 14. National Institute for Health and Clinical Excellence (NICE). Guide to the methods of technology appraisal. London; NICE: 2008. Available at: http:// www.nice.org.uk/media/B52/A7/TAMethodsGuideUpdatedJune2008.pdf [accessed 12.05.11]. 15. Smits M, Dippel DW, Nederkoorn PJ, de Haan GG, Dekker HM, Vos P, et al. Predicting intracranial traumatic findings on computed tomography in patients with minor head injury: the CHIP prediction rule. Ann Intern Med 2007;146:397–405. 16. Fabbri A, Servadei F, Marchesini G, Stein SC, Vandelli A. Observational approach to subjects with mild-to-moderate head injury and initial non-neurosurgical lesions. J Neurol Neurosurg Psychiatry 2008;79:1180–5. 17. Af Geijerstam JL, Britton M. Mild head injury: reliability of early computed tomographic findings in triage for admission. Emerg Med J 2005;22:103–7. 18. Haselsberger K, Pucher R, Auer LM. Prognosis after acute subdural or epidural haemorrhage. Acta Neurochir (Wien) 1988;90:111–6. 19. National Health Service, Health Resources Grouper version 4. Available at: http://www.ic.nhs.uk/services/the-casemix-service/using-this-service/ reference/archived-past-groupers-and-documentation/reference-cost/ hrg4-2008-09-reference-costs-grouper [accessed 12.05.11].
1431
20. Department of Health. National schedule of reference costs 2007–08 for NHS trusts. London: Department of Health; 2010. 21. Beecham J, Perkins M, Snell T, Knapp M. Treatment paths and costs for young adults with acquired brain injury in the United Kingdom. Brain Inj 2009;23: 30–8. 22. The Multi-Society Task Force on PVS. Medical aspects of the persistent vegetative state – second of two parts. N Engl J Med 1995;333(2):130. 23. Curtis L. Unit costs of health and social care 2009. University of Kent: Personal Social Services Research Unit (PSSRU); 2009. 24. Garside R, Pitt M, Anderson R, Rogers G, Dyer M, Mealing S, et al. The effectiveness and cost-effectiveness of carmustine implants and temozolomide for the treatment of newly diagnosed high-grade glioma: a systematic review and economic evaluation. Health Technol Assess 2007;11:1–221. 25. Aoki N, Kitahara T. Management of unruptured intracranial aneuysm in Japan: a Markovian decision analysis with utility measurements based on the Glasgow Outcome Scale. Med Decis Making 1998;18:357–64. 26. Stein SC, Hurst RW, Sonnad SS. Meta-analysis of cranial CT scans in children. A mathematical model to predict radiation-induced tumors. Pediatr Neurosurg 2008;44:448–57. 27. National Radiological Protection Board (NRPB). X-rays, how safe are they?Oxon: NRPB; 2001. 28. Claxton K, Sculpher M, McCabe C, Briggs A, Akehurst R, Buxton M, et al. Probabilistic sensitivity analysis for NICE technology assessment: not an optional extra. Health Econ 2005;14:339–47. 29. Cantor SB. Cost-effectiveness analysis, extended dominance and ethics: a quantative assessment. Med Decis Making 1994;14:259–65. 30. Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of costeffectiveness acceptability curves. Health Econ 2001;10:779–87.
Contents lists available at ScienceDirect
Injury journal homepage: www.elsevier.com/locate/injury
The cost-effectiveness of diagnostic management strategies for adults with minor head injury M.W. Holmes *, S. Goodacre, M.D. Stevenson, A. Pandor, A. Pickering School of Health and Related Research, The University of Sheffield, United Kingdom
A R T I C L E I N F O
A B S T R A C T
Article history: Accepted 18 July 2011
Study objective: To estimate the cost-effectiveness of diagnostic management strategies for adults with minor head injury. Methods: A mathematical model was constructed to evaluate the incremental costs and effectiveness (Quality Adjusted Life years Gained, QALYs) of ten diagnostic management strategies for adults with minor head injuries. Secondary analyses were undertaken to determine the cost-effectiveness of hospital admission compared to discharge home and to explore the cost-effectiveness of strategies when no responsible adult was available to observe the patient after discharge. Results: The apparent optimal strategy was based on the high and medium risk Canadian CT Head Rule (CCHRhm), although the costs and outcomes associated with each strategy were broadly similar. Hospital admission for patients with non-neurosurgical injury on CT dominated discharge home, whilst hospital admission for clinically normal patients with a normal CT was not cost-effective compared to discharge home with or without a responsible adult at £39 and £2.5 million per QALY, respectively. A selective CT strategy with discharge home if the CT scan was normal remained optimal compared to not investigating or CT scanning all patients when there was no responsible adult available to observe them after discharge. Conclusion: Our economic analysis confirms that the recent extension of access to CT scanning for minor head injury is appropriate. Liberal use of CT scanning based on a high sensitivity decision rule is not only effective but also cost-saving. The cost of CT scanning is very small compared to the estimated cost of caring for patients with brain injury worsened by delayed treatment. It is recommended therefore that all hospitals receiving patients with minor head injury should have unrestricted access to CT scanning for use in conjunction with evidence based guidelines. Provisionally the CCHRhm decision rule appears to be the best strategy although there is considerable uncertainty around the optimal decision rule. However, the CCHRhm rule appears to be the most widely validated and it therefore seems appropriate to conclude that the CCHRhm rule has the best evidence to support its use. ß 2011 Elsevier Ltd. All rights reserved.
Keywords: Minor head injuries Cost-effectiveness Diagnostic management strategies Quality adjusted life-years Economic modelling
Introduction Background Head injury accounts for around 700,000 emergency department (ED) attendances each year in England and Wales,1 90% of which may be categorised as apparently minor on the basis of having a Glasgow Coma Score (GCS) of 13–15 at presentation to hospital.2 The costs of liberal CT scanning and hospital admission are therefore substantial. However, if restriction of investigation
* Corresponding author at: School of Health and Related Research, Regents Court, 30 Regent Street, Sheffield S1 4DA, United Kingdom. Tel.: +44 0114 2220745; fax: +44 0114 2724095. E-mail address: [email protected] (M.W. Holmes). 0020–1383/$ – see front matter ß 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.injury.2011.07.017
and observation leads to delayed treatment of neurosurgical injury then the consequences may be devastating and include death or severe long-term disability. The main challenge in the management of minor head injury is identification of the minority of patients with significant intracranial injury, especially those who require urgent neurosurgery. Increased access to CT scanning in recent years has reduced the risk of missed pathology but has raised concerns about increased radiation exposure and inappropriate use of health care resources. Clinical decision rules have been developed in an attempt to limit the use of CT scanning whilst limiting the risk of missed pathology.1,3–7 It is not clear, however, which existing clinical decision rule achieves the best balance of increasing the benefits of scanning whilst reducing costs and harms, and how these rules compare with CT scanning all, or no patients.
1424
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Two previous studies have compared the cost-effectiveness of decision rules but both have limitations.8,9 One did not include the risk of cancer from CT scans and none were from the perspective of the United Kingdom (UK) National Health Service (NHS). Furthermore, these analyses did not explore decisions beyond CT scanning, such as whether or not to admit a patient with a normal CT scan. Data from the UK have suggested that increased use of CT scanning for minor head injury was associated with increased hospital admission.10 This may be due to admission of patients with non-neurosurgical abnormalities on CT scan or for patients with normal CT scans, but without a responsible adult to observe the patient after discharge. Aims of this investigation The goals of this investigation were to: (1) determine the optimal existing CT scanning diagnostic strategy in terms of costeffectiveness; (2) determine the cost-effectiveness of admission strategies consequent upon CT scanning. Materials and methods Diagnostic performance of clinical decision rules A systematic review of the literature was undertaken to evaluate the diagnostic performance of clinical decision rules.11 The literature review identified that most clinical decision rules for adults had estimates of diagnostic parameters from validation cohorts, although often in different settings from the derivation cohort. Decision rules would be expected to perform better in the derivation cohort and additionally in a validation cohort within the same setting as the derivation cohort. Therefore a validation study undertaken in a different setting may provide a more appropriate estimate of diagnostic performance. However, using different cohorts to estimate parameters for different decision rules could introduce selection bias. For these reasons we decided to use sensitivity and specificity data from a validation study by Stein et al., as this was a large, unselected cohort in which all of the main clinical decision rules were validated.12 Table 1 shows the sensitivity and specificity of each strategy used as probabilities in the model and to which decision nodes these apply.
The mathematical model For the model building process, we first identified the pathways for the model using the clinical opinion and modelling expertise within our team and with consideration of the data that was identified in a systematic review.11 Once the pathways were agreed we decided that a decision tree type model would be most appropriate. The model was constructed using Simul8 software (Simul8 Corporation). The model estimates the costs and qualityadjusted life-years gained (QALYs) for the following diagnostic strategies: discharge all patients without testing (theoretical), CT all patients (theoretical), CT only if GCS less than 15 on arrival (theoretical), Canadian CT Head Rule high risk3 (CCHRh), Canadian CT Head Rule high or medium risk3 (CCHRhm), Neurotraumatology Committee of the World Federation of Neurosurgical Societies4 (NCWFNS), New Orleans Criteria5 (NOC), National Emergency XRadiography Utilization Study II6 (NEXUS II), National Institute for Health and Clinical Excellence1 (NICE) and the Scandinavian rule.7 The analysis was conducted for patients aged 40 and 75 years presenting to the ED. A lifetime horizon was used with mean life expectancy taken from UK interim life tables,13 averaging the value for males and females. The economic perspective of the model is the UK National Health Service (NHS) and Personal social services (PSS) in England and Wales. This perspective was chosen as it conforms to the National Institute for Health and Clinical Excellence (NICE) guidelines for the methods of Technology Appraisals.14 A simplified schematic of the decision-analytic model is presented in Fig. 1. Each strategy was applied to a hypothetical cohort of 10,000 patients attending the ED with isolated, closed minor head injury (i.e. with no other serious injuries). A proportion of the cohort was assumed to have an intracranial lesion requiring immediate neurosurgery (an intracranial haematoma requiring evacuation); a further proportion was assumed to have an intracranial lesion that did not initially require immediate neurosurgery. For convenience, we refer to the former patients as having a neurosurgical lesion and the latter as having a non-neurosurgical lesion, although it should be recognised that the latter may ultimately receive neurosurgical intervention. The remainder would have no intracranial haemorrhage. These proportions were estimated from the study of patients with minor head injury by Smits et al. (lesion requiring
Table 1 Sensitivity and specificity of CT decision rules in correctly determining that a CT scan was required predicated on the assumption that CT scanning has 100% sensitivity and 100% specificity in identifying significant intracranial lesions.* Decision node Sensitivity for NS injurya
1-Sensitivity for NS injuryb
Sensitivity for NNS injuryc
Sensitivity or specificity of each strategy to detect intracranial injury (95% CI) Discharge all 0 1 0 CT all 1 0 1 Abnormal arrival GCS 0.91 (0.84–0.95) 0.09 0.72 (0.68–0.76) CCHRh3 0.99 (0.94–1.0) 0.01 0.97 (0.94–0.98) 3 CCHRhm 0.99 (0.94–1.0) 0.01 0.99 (0.97–1.0) 4 NCWFNS 0.99 (0.94–1.0) 0.01 0.95 (0.93–0.97) 5 NOC 0.99 (0.94–1.0) 0.01 0.99 (0.97–1.0) NEXUS II6 1.0 (0.97–1.0) 0 0.97 (0.94–0.98) NICE1 0.98 (0.93–1.0) 0.02 1.0 (0.99–1.0) Scandinavian7 0.99 (0.94–1.0) 0.01 0.95 (0.92–0.97)
1-Sensitivity for NNS injuryd
Specificitye
1-Specificityf
1 0 0.28 0.03 0.01 0.05 0.01 0.03 0 0.05
1 0 0.97 0.51 0.47 0.47 0.33 0.47 0.31 0.53
0 1 0.03 0.49 0.53 0.53 0.67 0.53 0.69 0.47
(0.96–0.98) (0.49–0.52) (0.46–0.48) (0.46–0.48) (0.32–0.34) (0.46–0.48) (0.30–0.32) (0.52–0.54)
NS, neurosurgical; NNS, non-neurosurgical. Beta distributions were used in the probabilistic sensitivity analysis for all parameters. All letters refer to Fig. 1. a The probability that a neurosurgical lesion is correctly identified, these patients receive prompt treatment. b The probability that a neurosurgical lesion is not correctly identified, these patients receive delayed treatment. c The probability that a non-neurosurgical lesion is correctly identified, these patients are admitted and those that deteriorate are given prompt treatment. d The probability that a non-neurosurgical lesion is not correctly identified. These patients are discharged and those that deteriorate are given delayed treatment. e The probability that patients with no intracranial lesion are correctly identified and discharged. f The probability that patients with no intracranial lesion are incorrectly identified and are given a CT scan. * All estimates have been extracted from Stein et al.12
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1425
Fig. 1. Model pathways.
neurosurgery 0.53%, 95% CI 0.33–0.85%; lesion initially not requiring neurosurgery 7.1%, 95% CI 6.26–8.05%).15 The number of patients undergoing CT scanning was dependent on the diagnostic strategy and ranged from 0% (discharge all patients without testing) to 100% (CT scan all patients). The probability of detecting a neurosurgical lesion was determined by the sensitivity of the diagnostic strategy. We assumed that patients with a neurosurgical lesion which was detected on CT would receive prompt evacuation of their intracranial haematoma (before any deterioration occurred); where a patient with a neurosurgical lesion did not have a CT scan, delayed evacuation (after deterioration had occurred) was assumed. We assumed that a proportion of patients with a nonneurosurgical lesion would deteriorate over the following 48 h and require intervention (critical care support and/or neurosurgery), whilst the remainder would remain well. These proportions were taken from a study by Fabbri et al.16 We assumed that if the
strategy led to CT being performed and the lesion detected then the patient would be admitted to hospital and would receive prompt appropriate treatment. If the strategy did not lead to CT being performed we assumed that the patient would be discharged home and would receive delayed treatment. We assumed that patients without an intracranial lesion remained well. The proportion of these patients receiving unnecessary CT was determined by the specificity of each strategy. Table 1 shows the sensitivity and specificity parameters of each strategy in correctly determining that a CT scan was required predicated on the assumption that CT scanning has 100% sensitivity and 100% specificity in identifying significant intracranial lesions. Only relevant lesions were those related to the head injury (i.e. we did not consider incidental findings unrelated to the injury). The CT all strategy therefore had 100% sensitivity and 100% specificity for both neurosurgical and non-neurosurgical lesions. The discharge all strategy had zero sensitivity and 100% specificity.
1426
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Although patients with an initially normal CT can subsequently develop significant injury and deteriorate, the rate of this occurrence is very low and unlikely to influence the costeffectiveness of strategies for selecting patients for CT.17
from the UK Personal Social Services Research Unit (PSSRU).23 The cost of glioma including the cost of palliative care for terminally ill patients was taken from the only relevant and reliable data source identified.24 Costs used in the model are shown in Table 2.
Patient outcome
Quality of life utility values
Patient outcome is measured by the Glasgow Outcome Score (GOS) with the following categories: GOS 1, Dead; GOS 2, Vegetative state; GOS 3, Severe disability; GOS 4, Moderate disability; GOS 5, Good recovery. The model allocated each patient to a GOS category based on whether they had an intracranial lesion and how quickly it was treated. This involved estimating the probabilities that patients with neurosurgical and non-neurosurgical lesions would end up in each GOS category depending upon the extent of treatment delay. For prompt treatment of patients with a neurosurgical injury we meta-analysed the data from 5 studies identified from a systematic review of the literature.11 For delayed treatment of patients with a neurosurgical injury we estimated GOS outcomes using a study by Haselberger et al., which showed the association between outcome and time delay from loss of consciousness to operation.18 For patients with a non-neurosurgical lesion where the lesion is detected on CT and the patient admitted and treated appropriately, we based estimates of GOS outcomes on a study by Fabbri et al.16 For patients with a non-neurosurgical lesion where a CT is not performed and the patient is discharged home without appropriate treatment, we did not identify any studies that reported the effect of treatment delay upon outcome. We therefore assumed that the delay had a similar effect as in the treatment of neurosurgical lesions and adjusted the GOS outcomes from Fabbri et al., accordingly.16 The distribution amongst GOS outcomes is provided in Table 2.
A literature review was conducted to identify studies that estimated health related utility values for GOS states. Two studies were found. Smits et al. obtained long-term GOS outcomes and quality of life (QoL) scores using the EQ-5D questionnaire from a subset of patients (n = 87) from the CT in head injury patients study.9 Aoki et al. used standard gamble methods to elicit quality of life utilities for GOS states 2–5, from 140 members of staff and students at a hospital in Japan.25 These studies were assessed for methodological compliance with the NICE reference case which stipulates that utilities should be measured in patients using a generic and validated classification system for which reliable UK population reference values, elicited using a choice based method such as the time trade-off or standard gamble are available.14 The Smits et al. study was considered to comply most closely with the NICE reference case and therefore the utility values from this study were used.9,14 The Smits et al. study did not report the age distribution of those patients used to estimate QoL utilities.9 We assumed that QoL values within GOS outcomes 3–5 were not age-related. The QoL utility values for death and persistent vegetative state were assumed to be 0. The utilities used in the model are shown in Table 2. Costs and utilities were discounted at an annual rate of 3.5%, as recommended by NICE.14
Costs
CT scans expose the patient to radiation which causes cancer in a proportion of patients. The estimated risk of a tumour from a single CT scan is taken from a study by Stein et al., which also provides an estimated QoL decrement risk per CT Scan.26 This study was limited to patients aged 20 years and younger; data from ages 5–20 years reported in Stein et al., were used to predict values for tumour risk and QALY loss for older patients. The statistical relationship resulted in tumour risk being equal to 0.0077 multiplied by age (in years) 1.076 (R2 = 0.92). The predicted tumour risk from a single CT scan for ages 35 and above was 0.0001 which corresponds with the best available evidence for the lifetime risk of cancer in adult patients.27 The predicted QoL decrement estimate from a single CT scan equalled 0.0401 multiplied by age (in years) 0.675 (R2 = 0.99), giving disutilities of 0.003 and 0.002 for ages 40 and 75 years respectively. The standard deviation associated with these values was predicted and were 0.0008 and 0.0007, at 40 and 75 years respectively. The costs of cancer associated with a CT scan were assumed to equal the risk of a tumour multiplied by the mean cost of cancer (£26,738).24 The Stein et al. study reports that the latency between radiation exposure and tumour diagnosis is over five years in the majority of cases.26 We have assumed a mean latency period of ten years with the cost of cancer subject to discounting.
Costs included in the model are the direct costs of diagnostic management including the costs of investigation, CT scanning, hospital admission and the subsequent costs of providing neurosurgical treatment, the cost of intensive care, nursing home care, rehabilitation for the severely disabled and the cost of glioma. For each hospital care cost, the most appropriate Health Resource Group (HRG) version 4 code was identified and cross referenced with the UK Department of Health National Schedule of Reference Costs (NSRC).19,20 The NSRC provides summary statistics (mean, range, inter-quartile range) from all Trusts for each HRG. The NSRC captures the overall cost of a treatment episode including length of stay. Patients with outcome GOS 5 were assumed well and did not incur any further costs. A literature review was conducted to find long-term costs of care and rehabilitation for patients whose outcomes are represented by GOS states 2–4. Only one study was identified which aimed to identify the health and social care services used by young adults aged 18–25 years with acquired brain injury.21 The study estimated average costs per person in four groups of patients, two of which correspond closely to the descriptions of GOS states 3 and 4. We were unable to find any cost data for older people and have therefore assumed the costs are the same as for age 18–25 years. The study did not provide cost data for patients in a vegetative state. We have therefore based our estimates on expert opinion assuming two weeks in intensive care followed by four months of rehabilitation before transferring to a nursing home for the rest of the patient’s life. (Sophie Duport, Royal Hospital for Neuro Disability, Personal communication, 2010). The Multi-Society Task Force on Persistent Vegetative State reported the mean length of survival for adults in a vegetative state (GOS 2) as 3.6 years which was used in the model.22 Nursing home costs were taken from the NSRC20 whilst rehabilitation costs were taken
Cancer risk from CT scans
Additional clinical questions In addition to an evaluation of the published strategies the following four clinical questions were assessed. 1. Is it cost-effective to admit patients with a normal CT scan compared to discharge home with a responsible adult? These patients are estimated to have a very low (0.006%) risk of deterioration.17
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1427
Table 2 Costs, parameters and distributions used in the PSA. Parameter Clinical outcomes Neurosurgical injury Non-neurosurgical injury Neurosurgical lesion: GOS outcomes after immediate intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Neurosurgical lesion: GOS outcomes after late intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Non-neurosurgical lesion: GOS outcomes after immediate intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 Non-neurosurgical lesion: GOS outcomes after late intervention GOS 5 GOS 4 GOS 3 GOS 2 GOS 1 QoL utilities GOS 3 GOS 4 GOS 5 Age related effect of a single head CT scan on tumour occurrence Age 40 and 75 Age related effect of a single head CT scan on QoL decrement Age 40 Age 75 Cancer latency (years) PVS mean survival Age 40 and 75 Costs ED visit CT scan Admission with no deterioration or neurosurgery. Neurosurgical intervention after deterioration. Neurosurgical intervention before deterioration. Long-term costs GOS 4 Long-term costs GOS 3 GOS 2 intensive care GOS 2 rehabilitation GOS 2 nursing home Cost of cancer
Mean probability or percent (%)
95% CI probability or percent (%)
Distribution
Source
0.0053 0.0710
0.0033–0.0085 0.0626–0.0805
Beta Beta
15 15 11
81.0% 9.3% 3.2% 2.7% 3.8%
74.7–86.1% 5.6–13.9% 1.2–6.3% 0.9–5.5% 1.6–7.9%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
57.0% 6.8% 12.0% 9.9% 14.3%
7.3–87.5% 0.8–12.4% 0.9–38.2% 0.7–33.2% 1.1–43.1%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
81.2% 11.1% 6.8% 0% 0.9%
73.2–87.2% 6.6–18.1% 3.5–12.9% 0–3.2% 0.2–4.7%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
55.9% 8.2% 27.6% 2.6% 5.4%
7.2–85.8% 1.0–15.7% 2.1–74.1% 0–15.8% 0.2–24.7%
Dirichlet Dirichlet Dirichlet Dirichlet Dirichlet
0.15 0.51 0.88
0.06, 0.28 0.39, 0.63 0.74, 0.97
Beta Beta Beta
0.0001
0.00009–0.00011
Normal
0.003 0.002 10
0.009–0.005 0.001–0.004 5–10
Beta Beta Normal
3.59
3.23–3.95 95% CI £67–170 £80–117 £490–997 £3,605–6616 £3758–6374 £15,444–20,592 £30,510–40,680 £12,781–17,561 £25,164–33,552 £804–1072 £24,064–29,412
Normal
18
16
16
9
26 26
£126 £100 £847 £5805 £5273 £17,160 £33,900 £15,469 £27,960 £893/week £26,738
Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma Gamma
26 22
19,20 19,20 19,20 19,20 19,20 21 21 19,20 23 20 24
PSA, probabilistic sensitivity analysis; GOS, Glasgow Outcome Score; CI, confidence interval; QoL, quality of life; PVS, persistent vegetative state. For probabilities of each strategies sensitivity and specificity see Table 1.
2. Is it cost-effective to admit patients with a normal CT scan compared to discharge home without a responsible adult? 3. Is it cost-effective to admit patients with a non-neurosurgical lesion on CT scan? These patients have a significant (13.5%) risk of deterioration.17 4. Does the optimal strategy change when a patient is discharged with no responsible adult? It is assumed that patients with a missed neurosurgical or non-neurosurgical lesion who deteriorate will not receive treatment and die. This contrasts with the base case where patients who deteriorate are admitted and receive delayed treatment. Analysis undertaken Deterministic results are presented first using incremental cost-effectiveness ratios (ICER).
Univariate sensitivity analysis was undertaken to explore the impact of changing each parameter in the model to its lower then higher confidence interval and in altering the discount rates for both costs and benefits to 0% and 6% per annum. Probabilistic sensitivity analysis (PSA) was undertaken to explore the impact of the joint uncertainty surrounding all model parameters.28 Definition of cost-effectiveness terms The ICER measures the relative value of two strategies and is calculated as the difference in mean costs divided by the difference in mean benefits. Where a strategy is less effective and more expensive than its comparator it is dominated. Extended dominance occurs when a combination of two alternative strategies can produce the same QALYs as a chosen strategy but at a lower cost.29
1428
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
Strategies that are neither dominated nor extendedly dominated constitute the cost-effectiveness frontier, and the ICER is reported for these strategies compared with the next least effective strategy. In a PSA each parameter in the model is assigned a distribution which encapsulates the uncertainty within the parameter. Beta distributions were used for probabilities as this distribution is bounded between 0 and 1, Gamma distributions were used for costs as the Gamma can describe the potential skewness of costs, Dirichlet distributions were used for GOS states as the states are correlated and must sum to 1, normal distributions were used for all other parameters. For each model iteration (1000 simulations each with 10,000 patients) each model parameter is randomly sampled from the relevant distribution. Results are presented as mean and incremental costs and QALYs, and ICERs. A costeffectiveness acceptability curve (CEAC) is calculated using the probability that a strategy has the highest net benefit (NB) for each iteration of the model for a range of willingness to pay (WTP) values.30 Net benefit is defined as: NB = WTP QALY-cost. Table 2 shows the parameters and distributions used in the PSA.
Table 3 Probabilistic analysis for age 40.
Discharge all Abnormal arrival GCS CT all NCWFNS NICE Scandinavian NOC NEXUS II CCHRh CCHRhm
Mean costs
Mean QALYs
ICERa
£3305 £2992 £2963 £2928 £2932 £2923 £2929 £2922 £2919 £2916
18.6669 18.6859 18.6897 18.6907 18.6907 18.6908 18.6908 18.6909 18.6910 18.6913
Dom Dom Dom Dom Dom Dom Dom Dom Dom Dominant strategy
QALY, quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, costeffectiveness; Dom, dominated. a Compared with next last effective treatment on the CE frontier.
Univariate sensitivity analysis
Results
For all ages, no parameter change altered the optimal strategy decision.
Deterministic analysis
Probabilistic sensitivity analysis
For adults aged 40, the CCHRhm strategy had an ICER of £3879 when compared to the Scandinavian strategy (the only other strategy on the cost-effectiveness frontier). All other strategies were either dominated or extendedly dominated. For adults aged 75, the CCHRhm strategy had an ICER of £10,397 when compared to the Scandinavian strategy (the only other strategy on the costeffectiveness frontier). All other strategies were dominated. Although the CCHRhm strategy was estimated to be optimal the absolute values for costs and QALYs were similar for all selective CT strategies.
Table 3 shows, for adults aged 40 years at presentation, the mean costs and QALYs per patient and which decision rules lie on the cost-effectiveness frontier. The strategies are ordered by ascending effectiveness (QALYs gained). Incremental costs and QALYs are not shown as the CCHRhm strategy dominates all other strategies, although the selective CT strategies had broadly similar costs and QALYs. Fig. 2 shows, for age 40, the CEAC which plots the probability of each strategy being the most cost-effective against values of WTP for a QALY ranging from zero to £50,000. Typical thresholds for
Fig. 2. Cost-effectiveness acceptability curve for adult aged 40.
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
1429
Table 4 Probabilistic analysis for age 75.
Discharge all Abnormal arrival GCS CT all NICE New Orleans NCWFNS Scandinavian NEXUS CCHRh CCHRhm
Mean costs
Mean QALYs
Incremental cost
Incremental QALYs
ICERa
£1718 £1546 £1574 £1541 £1539 £1532 £1527 £1529 £1525 £1526
7.8295 7.8374 7.8383 7.8390 7.8390 7.8391 7.8392 7.8392 7.8393 7.8393
n/a n/a n/a n/a n/a n/a n/a n/a n/a £0.17
n/a n/a n/a n/a n/a n/a n/a n/a n/a 0.1704
Dom Dom Dom Dom Dom Dom Dom Dom – £2666
QALY, quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, cost-effectiveness; Dom, dominated. a Compared with next last effective treatment on the CE frontier.
Table 5 Clinical question 4, probabilistic analysis for age 40 discharged without a responsible adult. Strategy
Mean cost
Mean QALY
Incremental cost
Incremental QALY
ICERa
Discharge all Abnormal arrival GCS NCWFNS Scandinavian NEXUS II CCHRh CT all NICE NOC CCHRhm
£2053 £2838 £2917 £2911 £2914 £2909 £2964 £2922 £2923 £2910
18.5517 18.6624 18.6878 18.6879 18.6895 18.6895 18.6897 18.6901 18.6903 18.6907
n/a n/a n/a n/a n/a n/a n/a n/a n/a £857
n/a n/a n/a n/a n/a n/a n/a n/a n/a 0.1390
– ExtDomb Dom Dom Dom ExtDomb Dom Dom Dom £6167
QALY, Quality adjusted life year; ICER, incremental cost-effectiveness ratio; CE, cost-effectiveness; Dom, dominated; ExtDom, extendedly dominated. a Compared with next least effective treatment on the CE frontier. b Extendedly dominated by CCHRhm and discharge all.
decision-making in the UK are £20,000–30,000 per QALY14. Strategies that never reach a probability of 10% have been removed for visual clarity. At all WTP values, the CCHRhm strategy had a higher probability of being cost-effective than any other strategy. However, a comparison of the best two strategies at age 40 indicated that there was considerable uncertainty in the results: in 63% of simulations CCHRh dominated CCHRhm, in 9% CCHRhm dominated CCHRh, in 16% CCHRhm was more expensive but provided more health benefits and in12% CCHRh was more expensive but provided more health benefits. When the CCHRhm was compared to the NEXUS II and NOC diagnostic rules, similar levels of uncertainty were seen. Table 4 reports the analysis for adults aged 75 years. Mean QALYs and mean costs are both lower than in the analysis for 40 year old adults, reflecting reduced life expectancy. All strategies except for CCHRh and CCHRhm are dominated. The ICER for CCHRhm versus CCHRh is £2666. Again, the results for selective CT strategies are broadly similar. The CEAC for age 75 is not shown as it is similar to the CEAC for age 40. At all WTP values, the CCHRhm strategy had a higher probability of being cost-effective than any other strategy. However, there was considerable uncertainty regarding the optimal strategy. Assessment of additional clinical questions The following results were found for the additional clinical questions: Question 1. A deterministic analysis indicates that hospital admission for patients with a normal CT scan compared to discharge home with a responsible adult gains on average only 0.00001 QALYs and costs an additional £441 per patient. The
ICER for admitting all patients compared to discharge home is approximately £40 million. Question 2. A deterministic analysis indicates that hospital admission for patients with a normal CT scan compared to discharge home without a responsible adult gains on average only 0.0002 QALYs and costs an additional £468 per patient. The ICER for admitting all patients compared to discharge home is approximately £2.5 million. Question 3. A deterministic analysis indicates that the admission strategy for those with a non-neurosurgical lesion costs approximately £340 less and gains 0.004 QALYs compared to discharge home and therefore dominates discharge home. Question 4. Table 5 shows the main probabilistic analysis for a 40 year old adult repeated with the assumption that patients who are discharged home have no responsible adult and are not brought to medical attention when they deteriorate. The discharge all strategy is cheaper and less effective than in the main model and is no longer dominated by other strategies. The CCHRhm strategy is most cost-effective with an ICER of £6167, compared with the discharge all strategy. Discussion Decision-analysis modelling is limited by both the need to make assumptions and the available data. Our estimates of the effect of delayed treatment upon intracranial pathology in particular are based on very limited observational data. In the UK, hospitals do not collect and record data on the use of resources by individual patients. The process of allocating costs starts with the annual financial returns of a hospital and this expenditure is then reallocated to patient treatment services in a ‘top down’ manner. This process varies across NHS Acute trusts and consequently some resources have
1430
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431
large variability around their mean cost. However, these estimates are the best available and this variability in included in the PSA. We did not use age related QoL utilities. However, it is likely that the QoL lost through the ageing process will be proportionately comparable across all management strategies and the conclusions will be unaltered. Another potential limitation is using data from ages 5 to 20 years to predict QALY loss in adults. It is possible that our predictions are not sufficiently taking into account the effects of discounting which could mean our predictions are overestimating QALY loss in adults, especially the 75 year old patient. This limitation, however, is expected to have little effect as any inaccuracies around the QALY loss are likely to be small but would favour those policies that perform fewer CT scans. Productivity losses were not included as they are difficult to estimate and subject to substantial uncertainty. It is likely that including productivity losses would only strengthen our overall conclusions because they would simply add to the high costs of missing serious injury that drives the cost-effectiveness of liberal CT use. Finally, we assumed that the various strategies would be implemented in a consistent way. This may not be the case in practice and studies of the implementation of decision rules in practice are required. These limitations noted, our economic analysis showed that use of CT scanning as determined by a clinical decision rule is not only more effective than not investigating but is also cost saving. Selective CT use according to a clinical decision rule was also costeffective compared to CT for all patients. The most cost-effective rule was the CCHRhm criterion,3 however there was considerable uncertainty in the results when compared to the other rules. The costs and outcomes associated with each rule were broadly similar, so the superiority of the CCHRhm may simply reflect a small difference in the estimate of diagnostic accuracy which was not statistically significant in the primary data.12 However, the Canadian rules appear to be the most widely validated and have estimates of diagnostic accuracy that are reasonably consistent across a number of cohorts.3 Two USA studies assessed the costeffectiveness of various decision rules.8,9 Although the comparator strategies were different from our study, both USA studies agreed that the CCHR decision rule was cost-effective compared to other rules. Our analysis showed that admission of patients with a normal CT scan would not be cost-effective regardless of whether there was a responsible adult available to observe the patient. These results are based upon data suggesting a very low (0.006%) risk of deterioration and it is assumed that patients are clinically well and would not benefit from general hospital care.17 The conclusion that patients with a normal CT scan should not be admitted to hospital does not apply to those with, for example, repeated vomiting or significant amnesia who might benefit from symptomatic treatment, nursing care or a safe environment. However, if the patient is orientated, comfortable and able to selfcare then our analysis suggests that hospital admission for observation is not a cost-effective use of health care resources. Hospital admission for those with a non-neurosurgical lesion on CT was cost saving as the costs of long-term care for those who deteriorated and received late treatment outweighed the costs of hospital admission. This analysis was limited by the lack of a standard definition as to what constitutes a significant nonneurosurgical lesion on CT and the limited data relating to outcomes from non-neurosurgical lesions. The prognosis of different non-neurosurgical lesions varies markedly, so costeffectiveness could potentially be improved by selecting those at highest risk of deterioration for admission whilst discharging
those at lower risk. Currently, however, we do not have sufficient data to evaluate this approach. Conclusions Our economic analysis confirms that the recent extension of access to CT scanning for minor head injury is appropriate. Liberal use of CT scanning based on a high sensitivity decision rule is not only effective but also cost-saving. It is recommended therefore that all hospitals receiving patients with minor head injury should have unrestricted access to CT scanning for use in conjunction with evidence-based guidelines. Provisionally the CCHR decision rule appears to be the best strategy although there is considerable uncertainty around the results. Incremental changes in both costs and QALYs are very small compared to the other selective CT strategies. However, the CCHR appears to be the most widely validated rule and therefore it does not seem inappropriate to conclude that, currently, the CCHR rule has the best evidence to support its use. Hospital admission appears to be cost-effective for patients with an intracranial lesion on CT scanning but not for those with a normal CT. Conflict of interest The authors have no commercial associations or sources of support that might pose a conflict of interest. Acknowledgements This project was undertaken for the UK National Institute for Health Research Health Technology Assessment Program. The findings of this project, including the findings presented in this article, will be published as a report in the Health Technology Assessment monograph series: http://www.hta.ac.uk/project/ 1765.asp. The study sponsors had no involvement in the study design; collection, analysis and interpretation of data; the writing of the manuscript or the decision to submit the manuscript for publication. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR HTA. References 1. National Institute for Health and Clinical Excellence (NICE). Head injury – triage, assessment, investigation and early management of head injury in infants, children and adults (NICE Clinical Guideline No. 56). London: NICE; 2007. 2. Swann IJ, Teasdale GM. Current concepts in the management of patients with so-called ‘minor’ or ‘mild’ head injury. Trauma 1999;1:143–55. 3. Stiell IG, Wells GA, Vandemheen K, Clement C, Lesiuk H, Laupacis A, et al. The Canadian CT Head Rule for patients with minor head injury. Lancet 2001;357:1319–96. 4. Servadei F, Teasdale G, Merry G, Neurotraumatology Committee of the World Federation of Neurosurgical Societies. Defining acute mild head injury in adults: a proposal based on prognostic factors, diagnosis, and management. J Neurotrauma 2001;18:657–64. 5. Haydel MJ, Preston CA, Mills TJ, Luber S, Blaudeau E, DeBlieux MC. Indications for computed tomography in patients with minor head injury. N Eng J Med 2000;343:100–5. 6. Mower WRH, Hoffman HJ, Herbert M, Wolfson AB, Pollack C, Zucker M. Developing a decision instrument to guide computed tomographic imaging of blunt head injury patients. J Trauma Injury Infect Crit Care 2005;59:954–9. 7. Ingebrigtsen T, Romner B, Kock-Jensen C, Ingebrigtsen T, Romner B, Kock-Jensen C. Scandinavian guidelines for initial management of minimal, mild, and moderate head injuries. The Scandinavian Neurotrauma Committee. J Trauma Injury Infect Crit Care 2000;48:760–6. 8. Stein SC, Burnett MG, Glick HA. Indications for CT scanning in mild traumatic brain injury: a cost-effectiveness study. J Trauma Injury Infect Crit Care 2006;61:558–66. 9. Smits M, Dippel DW, Nederkoorn PJ, Dekker HM, Vos PE, Kool DR, et al. Minor head injury: CT-based strategies for management – a cost-effectiveness analysis. Radiology 2010;254:532–40.
M.W. Holmes et al. / Injury, Int. J. Care Injured 43 (2012) 1423–1431 10. Goodacre S. Hospital admissions with head injury following publication of NICE guidance. Emerg Med J 2008;25:556–7. 11. Harnan SE, Pickering A, Pandor A, Goodacre SW. Clinical decision rules for adults with minor head injury: a systematic review. J Trauma 2011;71(1):245–51. 12. Stein SC, Fabbri A, Servadei F, Glick HA. A critical comparison of clinical decision instruments for computed tomographic scanning in mild closed traumatic brain injury in adolescents and adults. Ann Emerg Med 2009;53:180–8. 13. Office for National Statistics (ONS). United Kingdom, interim life tables 1980–82 to 2006–2008. London: ONS; 2008. 14. National Institute for Health and Clinical Excellence (NICE). Guide to the methods of technology appraisal. London; NICE: 2008. Available at: http:// www.nice.org.uk/media/B52/A7/TAMethodsGuideUpdatedJune2008.pdf [accessed 12.05.11]. 15. Smits M, Dippel DW, Nederkoorn PJ, de Haan GG, Dekker HM, Vos P, et al. Predicting intracranial traumatic findings on computed tomography in patients with minor head injury: the CHIP prediction rule. Ann Intern Med 2007;146:397–405. 16. Fabbri A, Servadei F, Marchesini G, Stein SC, Vandelli A. Observational approach to subjects with mild-to-moderate head injury and initial non-neurosurgical lesions. J Neurol Neurosurg Psychiatry 2008;79:1180–5. 17. Af Geijerstam JL, Britton M. Mild head injury: reliability of early computed tomographic findings in triage for admission. Emerg Med J 2005;22:103–7. 18. Haselsberger K, Pucher R, Auer LM. Prognosis after acute subdural or epidural haemorrhage. Acta Neurochir (Wien) 1988;90:111–6. 19. National Health Service, Health Resources Grouper version 4. Available at: http://www.ic.nhs.uk/services/the-casemix-service/using-this-service/ reference/archived-past-groupers-and-documentation/reference-cost/ hrg4-2008-09-reference-costs-grouper [accessed 12.05.11].
1431
20. Department of Health. National schedule of reference costs 2007–08 for NHS trusts. London: Department of Health; 2010. 21. Beecham J, Perkins M, Snell T, Knapp M. Treatment paths and costs for young adults with acquired brain injury in the United Kingdom. Brain Inj 2009;23: 30–8. 22. The Multi-Society Task Force on PVS. Medical aspects of the persistent vegetative state – second of two parts. N Engl J Med 1995;333(2):130. 23. Curtis L. Unit costs of health and social care 2009. University of Kent: Personal Social Services Research Unit (PSSRU); 2009. 24. Garside R, Pitt M, Anderson R, Rogers G, Dyer M, Mealing S, et al. The effectiveness and cost-effectiveness of carmustine implants and temozolomide for the treatment of newly diagnosed high-grade glioma: a systematic review and economic evaluation. Health Technol Assess 2007;11:1–221. 25. Aoki N, Kitahara T. Management of unruptured intracranial aneuysm in Japan: a Markovian decision analysis with utility measurements based on the Glasgow Outcome Scale. Med Decis Making 1998;18:357–64. 26. Stein SC, Hurst RW, Sonnad SS. Meta-analysis of cranial CT scans in children. A mathematical model to predict radiation-induced tumors. Pediatr Neurosurg 2008;44:448–57. 27. National Radiological Protection Board (NRPB). X-rays, how safe are they?Oxon: NRPB; 2001. 28. Claxton K, Sculpher M, McCabe C, Briggs A, Akehurst R, Buxton M, et al. Probabilistic sensitivity analysis for NICE technology assessment: not an optional extra. Health Econ 2005;14:339–47. 29. Cantor SB. Cost-effectiveness analysis, extended dominance and ethics: a quantative assessment. Med Decis Making 1994;14:259–65. 30. Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of costeffectiveness acceptability curves. Health Econ 2001;10:779–87.