Analytical approach:
The two cost-effectiveness analyses were based on two randomised controlled trials where Step 2 was nested in Step 1. The time horizon was the duration of the trials (one year for both). The authors stated that the perspective was that of the UK National Health Service.
Effectiveness data:
The primary measure of effectiveness for the trials was the difference in Neck Disability Index (NDI) score, which measures pain-related activity restrictions in 10 areas. For cost-effectiveness analyses the primary measure of benefit was based on the EQ-5D utility scores (a secondary effectiveness measure of the trials). Other secondary measures included the Short Form 12 (SF12) version (a measure of health-related quality of life), patient self-rated benefit (higher scores being better) and serious adverse events.
Baseline values were obtained via a registration document completed by staff at the emergency department. Postal questionnaires were sent to patients at four, eight and 12 months after the initial attendance to the emergency department. For Neck Disability Index scores patients were asked to recall events over the past month. Adjustments were made for ethnic origin, grade of whiplash associated disorder at initial attendance and clustering within emergency departments for Step 1 and for Step 2 use of Neck Disability Index, grade of whiplash associated disorder at initial attendance and clustering within emergency department and therapists.
Monetary benefit and utility valuations:
Utilities were measured using EQ-5D responses from the patient postal questionnaires. Quality-adjusted life-years (QALYs) were calculated as the area under the baseline-adjusted utility curve, assuming linear interpolation between utility measurements. Baseline EQ-5D scores were derived from published general health population estimates, matched for age.
Measure of benefit:
The measure of benefit was the quality-adjusted life-year (QALY). There was no discounting of benefits as the time horizon was only one year.
Cost data:
Resource use was measured for all hospital in-patient and outpatient service use, diagnostic tests, prescribed medications and community health resource use. Items were measured using the questionnaires mailed to participants at four, eight and 12 months. Costs were attached to each resource use item. Costs were obtained from standard national sources including the Personal Social Service Research Unit (PSSRU), NHS Reference Costs 2008 and NHS Information Centre Prescribing Support Unit. Work days lost were calculated by asking patients in the questionnaire about the number of work days lost recalled in the past four months. Work days lost were analysed using baseline covariate adjustment for employment status and multiple linear regression with bootstrapping because of highly skewed data. Costs were reported in 2009 GBP (£) prices. There was no discounting as the time period was only one year.
Analysis of uncertainty:
The non-parametric bootstrap method was used to construct cost-effectiveness acceptability curves to show the probability that each alternative was cost-effective across a range of willingness to pay thresholds for an additional QALY. Sensitivity analysis was conducted on the Step 2 analysis to investigate the mediating role of medication of the effect estimates for physiotherapy.