The data used in this study come from the OSCAR study, a multi-centre, randomised, controlled trial of HFOV as compared to conventional MV in patients with moderate-to-severe ARDS. Patients (>16 years) who were undergoing MV were recruited in different-sized ICUs from 29 hospitals across England, Wales and Scotland and were randomised to either HFOV or conventional MV. Patients were eligible if they were expected to require at least two more days of MV and met the definition of moderate or severe ARDS. Patients were excluded if they had been on ventilation for 7 days or more. Inclusion criteria for OSCAR and patient recruitment have been described in detail in Young et al. . The study was approved by national ethics review committees and research governance departments at each centre. Patients or their representatives provided written informed consent.
Characteristics such as age, gender, ventilation prior to enrolment, physiology and other information required to assess illness severity were collected at the time of randomisation. The Acute Physiology, Age, and Chronic Health Evaluation (APACHE II) score was used to measure the severity of illness within the first 24 h after the patient was admitted to an ICU. The APACHE II score is based on several physiological measurements and pre-admission health status and ranges from 0 to 71, with higher scores corresponding to more severe illness. The baseline ratio of the partial pressure of arterial oxygen (PaO2) to the fraction of inspired oxygen (FiO2) was also recorded as a measure of the severity of ARDS. Case report forms (CRFs) were completed by the medical and nursing staff for each day a patient was in ICU. The CRFs recorded the use of antibiotics, sedatives and muscle relaxant drugs as well as information on support for organ systems. Serious adverse events and whether the patient required a chest drain or presented radiological evidence of barotrauma were also recorded. ICU discharge date and location, ICU readmissions and hospital discharge date were available.
Following hospital discharge, questionnaires were sent to surviving patients and their carers at 6 and 12 months. Patients were asked about their use of medical services during the previous 6 months, including primary- and community-based health and social services and residential inpatient stays. The questionnaires also contained questions related to the use of aids and equipment, gross loss of earnings and other major expenses. Patients’ questionnaires also recorded quality of life using the EuroQol-5D (EQ-5D) . Carers’ questionnaires recorded the cost of travel to and from medical services, major expenses and loss of earnings.
Quality of life
Patients’ HRQL was measured at 6 and 12 months using EQ-5D. Health utility weights for each patient were derived from EQ-5D responses using the standard UK specific tariffs . For each time period, mean utility scores were compared to those of an age- and sex-matched reference population .
To obtain mean QALYs at 1 year, quality of life and survival data were combined into a single metric. Patients’ histories were partitioned into several periods from randomisation to 1 year, and a utility value was assigned to each of these periods . For the ventilation period, we multiplied the average number of days ventilated by the utility weight of an unconscious patient reported in the EQ-5D scoring manual (−0.4) . We then calculated the mean utility scores of survivors at 6 and 12 months using all available questionnaires and, assuming a linear change in HRQL between assessments, we calculated mean utility scores at two additional mid-points (3 and 9 months). The mean time in each period was obtained by calculating the area under the relevant Kaplan-Meier survival curve. The sum of the mean survival time in each state multiplied by the utility weights provided an estimate of the quality-adjusted survival at 1 year, expressed in QALYs.
We used a bottom-up micro-costing approach where we assigned unit costs to volumes of resource use for each patient . We analysed costs from the perspective of the health and social care sector (NHS) and from a broader societal perspective. Costing was undertaken for the various pathways of care following ICU admission: initial ICU stay, hospital stay following first ICU discharge, post-hospital NHS resource use and societal costs. ICU and hospital costs were assessed using the bottom-up approach [36, 42, 43].
Resource use associated with initial stay in ICU was assessed daily based on the number of organs supported, respiratory support, whether the patient was on renal replacement therapy, whether he/she required X-rays (to check for barotrauma) or a chest drain (pneumothorax) and use of medicines. The quantities of resource use were multiplied by their corresponding unit costs, and the sum per patient for the entire stay was calculated. The unit costs of ICU resources including cost per organ supported, radiology and the cost of pneumothorax were taken from the National Schedule of Reference Costs . The costs of medicines used were taken from the British National Formulary (BNF) . The daily cost of ventilation was based on a fixed and per patient cost. We assumed that ventilation machines would be used for 5 years; annual maintenance and the costs of single use circuits were also included.
Once patients were discharged from ICU, the cost of step-down and the cost of ICU readmissions were calculated using the number of days until death or discharge multiplied by the cost of the level of care required, based on various nationally available references [44, 46, 47]. If a patient was discharged to another hospital following ICU, the cost of transport was taken as that of an emergency transfer. Serious adverse events and corresponding unit costs were assessed on an individual basis. Patient-reported data on resource use were collected at 6 and 12 months. NHS resource use included further inpatient care, outpatient care, primary- and community-based care and aids and equipment provided by the NHS. The costs of attendance at medical services were calculated using national reference costs multiplied by the number of times a patient attended [44, 47]. Inpatient stays were based on the number of days admitted multiplied by the corresponding unit cost . The cost of aids and equipment was taken from the NHS supply chain cost for each individual item. Costs incurred by patients and their carers included cost of travel, loss of earnings and patient out-of-pocket expenses for aids, equipment and extra expenses (home adaptations). The cost of travel to and from appointments for both carers and patients was based on the distance in miles provided in the questionnaires multiplied by the cost per petrol mile as provided by HM Revenue & Customs (HMRC) . Patients and carers were asked to give the gross amount lost in earnings in the 6 months covered by each questionnaire. Cost of aids and equipment purchased by patients was directly reported in the patient questionnaire. All costs were adjusted to 2012 prices using the Hospital and Community Healthcare Services (HCHS) index published by the Personal Social Services Research Unit (PSSRU) . Unit costs used in the analysis and more details on the costing strategy are provided in the supplemental appendix.
The primary outcomes were 1-year survival, quality of life and resource use. Resource use was measured in terms of duration of treatment in both the ICU and the hospital and in terms of health care and societal costs at 1 year, divided into the various care pathways, i.e. ICU stay, hospital stay and post-discharge period. These outcomes were presented for various subgroups of patients based on baseline demographic, physiological, and clinical characteristic, including ventilation strategy (HFOV or conventional ventilation). Total cost per 1-year survivor was obtained by dividing the sum of ICU, hospital, and post-hospital costs for all patients by the number of patients remaining alive 1 year post-randomisation. The 1-year incremental cost effectiveness ratio (ICER) of HFOV as compared to conventional MV was also calculated. The ICER was obtained by dividing the difference between the average 1-year costs in the HFOV group, and the average 1-year cost in the conventional ventilation group by the difference between average health outcomes (QALYs) gained in the HFOV group and those gained in the conventional ventilation group .
Patient characteristics are presented using means and standard deviations (SD) for continuous variables and proportions and percentages for categorical variables. We compared baseline characteristics of 1-year survivors and non-survivors using Student’s t tests for continuous variables and chi-squared tests for categorical variables. No corrections were made for multiple testing. Cost data are presented using means, 95 % confidence intervals, median and interquartile range. HRQL at 6 and 12 months and quality-adjusted survival were summarised using means and 95 % confidence intervals. HRQL of survivors was compared to population reference values using Student’s t tests. We first estimated mean post-hospital costs based on complete cases. To obtain total costs at 1-year for each patient, missing data on post-hospital costs among 1-year survivors was addressed by multiple imputation using chained equations. We used a truncated model in which costs were constrained to be equal or above zero and generated ten datasets. Estimates from each imputed dataset were combined following Rubin’s rule . All p values were two-sided, and a significance level of less than 0.05 was used. R 3.0.2 and Stata 12 (Statacorp, College Station, TX, USA) were used for the analyses.