The Lancet | The best science for better lives



Web extra materialeDocument 1: Mattress Specification GuidelineAll centres are requested to provide list of mattresses in use within each Hospital and used within the trial.1.All participants will have an electric profiling bed frame as an adjunct to the trial mattress, this was to standardise self-care and assisted reposition functionality across both arms.2.All participants will be randomised to either a High Specification Foam Mattress (HSF) or Alternating Pressure Mattress (APM). Overlay or replacement mattresses may be used. 3.After randomisation an eligible mattress will be sourced and allocated by the clinical research nurse. 4.All mattresses and their use, will comply with the Medical Devices Regulations SI2002/618Mattress specifications: Pressure relieving mattresses can be divided into low tech devices and high tech devices:Low tech devices provide a comfortable surface that redistributes body weight over a large surface area. Examples are: standard foam mattresses, high specification foam mattresses, visco-elastic mattresses, cubed foam, convoluted foam (these aim to redistribute body weight over a larger contact area).High tech devices are dynamic systems which include alternating pressure mattresses where the patient lies on air filled sacs, which sequentially inflate and deflate and relieve pressure at different anatomical sites for short periods. Mattresses included in the trial - Alternating Pressure Mattress:?All mattresses included in the Trial should be currently in use with in the recruiting centre and fully comply with local medical devices and infection control standards.?All mattresses included should be fully automatic, some may have dual therapy e.g. the mattress comprises of a combination of alternating pressure or low air loss. The trial will only include participants nursed on the alternating pressure mode of action.Alternating pressure mattresses should have the following minimum function.Box 1: Alternating mattress minimum function specificationAPM replacementsAPM overlaysCell height19.6 – 29.4 cm8.5 – 12.5 cmCycle time 7.5 – 30 mins7.5 – 30 minsCycle frequency1 in 2, 3 or 41 in 2,3 or 4Mattresses included in the trial - High Specification Foam Mattresses?All mattresses included in the Trial should be currently in use with in the recruiting centre and fully comply with local medical devices and infection control standards.?Mattresses can be high density foam, visco-elastic (memory) foam, combination of both, can be castellated (for ventilation and profiling)?All mattresses will have a cover with the following characteristics: removable, minimum two way stretch, vapour permeable, covered zips as defined in British Standard 3379 [28]. ?All mattresses should be replacement mattresses with a minimum depth of: 150-200 mmMattresses Excluded:Hybrid mattresses with combination therapy, e.g. mattresses comprising of a static layer with alternating cells/low air loss. E.g. Atmos air.Alternating mattresses which have collapsible cells. E.g. Huntleigh Nimbus professional.Low air loss mattresses. E.g. Huntleigh Breeze.Static foam and gel mattressesContinuous static low pressure mattresses including fibregel. Fluid, clay. E.g. Rik.Air filled mattresses. E.g. Repose.Beds and mattresses which have a motorised repositioning option. E.g. Huntleigh Acer.Air fluidised bead beds. E.g. ClinitroneDocument 2: Eligible Care SettingsEligible care settings were any in-patient healthcare facility that admitted acutely ill patients under the care of a medical practitioner e.g. Consultant or General Practitioner. Evidence of acute illness through:a.acute admission to secondary care hospital, community hospital or NHS funded intermediate care/rehabilitation facility.b.secondary care, community hospital or NHS funded intermediate care/rehabilitation facility in-patient with onset of acute illness secondary to elective admission.c.Recent secondary care hospital discharge to community hospital or NHS funded intermediate care/ rehabilitation facility.eDocument 3: Assessment of capacity to consent In relation to the practicalities of recruitment we adopted the following process:The assessment of capacity will relate specifically to decisions pertaining to the research project. Each patient will be assumed to have capacity unless it is established that they lack capacity. Ward nurses identifying patients for study participation, will be asked to consider aspects of capacity before any approach to patients is made and during the information giving stage prior to consent. The Clinical Research Nurse/Practitioner (CRN/P) will assess the patient’s ability to understand what decisions they need to make and why? the consequences of the decision to participate? their ability to understand, use and retain the information related to the decision to participate and be able to communicate their decisions effectively (as specified in the Mental Capacity Act 2005). If there is any concern about capacity the CRN/P will consult further with other members of the attending clinical team and/or relative/carer/friend (as appropriate) and a decision will be made with the relative/carer/friend as to whether the patient is able to provide written consent. Where the patient is thought not to have capacity to consent, a relative, carer or friend who is interested in the patient’s welfare will act as a personal consultee.The relative/carer/friend will be involved in the information and decision making process with the patient and will advise the CRN/P on their presumed wishes and feelings and Consultee Assent will be obtained on behalf of the patient. The relative, carer or friend will be advised to set aside their own views and provide advice on the participation of the patient in the research, taking into consideration the patient’s wishes and interests. Research participants will not be required to do anything which is contrary to any advance decisions or statements that have been made by them in relation to their treatment or any other matter. Advance decisions made by the patient about their preferences and wishes will always take precedence.If, despite taking all reasonable steps, a personal consultee cannot be identified and contacted then a nominated consultee would be approached. This person would have no connection with the research project. They would be nominated by the local Principal Investigator and most likely will be the participant’s lead clinician, or ward manager (each Trust will establish a system for identifying a personal consultee). The consultee would be provided with the information leaflet describing the research study and the role of the consultee and it would be emphasised that they are being asked to act on behalf of the participant, rather than any personal views or feelings.All CRN/P were trained in Good Clinical Practice (GCP). No additional training on assessment of capacity was mandatory and no specific tools to aid assessment were recommended. However, this topic was included at site initiation visits and subsequent investigator meetings. Where there was any doubt regarding a patient’s decisional capacity, this was referred to the local Principal Investigator.eDocument 4:Randomisation processScreened patients who were both eligible for study participation and provided written informed consent/witnessed verbal consent/consultee or nearest relative/Guardian/welfare attorney (Scotland) agreement were randomised. Following confirmation of informed consent, eligibility and completion of baseline assessments (including questionnaires), patients were randomised by an authorised member of the research team at the site using the CTRU automated secure 24-hour telephone randomisation service. Authorisation codes and PINs, provided by the CTRU, were required to access the randomisation system. The following information was provided at randomisation:?Participant details, including initials, gender and date of birth?Site code for research site?Name of person making the randomisation?Confirmation of eligibility ?Confirmation of informed consent.?NHS number?Confirmation of completion of baseline skin assessments.?Details relating to the patient characteristics (see below).Patients were randomised in a 1:1 allocation ratio, to receive either HSF or APM and were allocated a trial number. A computer-generated minimisation programme that incorporates a random element was used to ensure intervention groups were well balanced for the following participant characteristics, details of which will be required for randomisation. ?Centre?PU status (no pressure ulcer, Category 1, Category 2)?Secondary care hospital, community hospital / intermediate care or rehabilitation facility.?Consent (written, witnessed verbal, consultee or nearest relative/Guardian/welfare attorney (Scotland) agreement). The randomisation system included an automated internal check using NHS number to confirm that the patient has not been recruited to the trial previously. This was a risk in this study since the recruitment period was for 3 years and involved adjacent acute and community NHS Trusts.Details of the patients participation in the trial and the randomised intervention were documented in medical and nursing records and the ward staff informed and if necessary asked to change the mattress as soon as possible.eDocument 5: Revised sample size and expected accrualThe trial recruited participants at a much slower rate than anticipated and an unplanned interim analysis and VOI Analysis using confidential data from the trial was requested by the funder and conducted in November/December 2015 on 909 participants. Unblind analyses were reviewed by the DMC and remained confidential, but they informed the Independent TSC that the event rate was much lower (9.9%) than originally estimated. The DMC and TSC asked the Trial Management Group (TMG) who remained blind to the event rate, to consider the minimum clinically relevant differences on varying centred event rates. The TMG noted that a relative differences of 25%, 33.3% and 50% centered on event rates of 15%, 10% and 5% corresponding to absolute differences of 3.75%, 3.3% and 2.5% respectively were considered to be the minimum clinically important differences. The TMG also noted that a relative difference of 33.3%, 40% and 60% centered on event rates of 15%, 10% and 5% corresponding to absolute differences of 5%, 4% and 3% respectively were considered of clinical relevance. In addition, having sufficient precision in the estimate of the treatment effect to conclude futility was also considered important by the TMG.The DMC and TSC supported two requests to the funder: a)a 6 month recruitment extension (requiring no additional funding) to achieve minimum 1996 participants, allowing a clinically relevant difference of 4% to be detected with at least 80% power, assuming an overall event rate of 10%. b)a 14 months recruitment extension (requiring substantial additional funding) to achieve a sample size of 2728 to detect the minimum clinically important difference between mattresses of 3.3% with 80% power, together with a clinically meaningful improvement in precision of a further 3% under the assumption. The 6-month recruitment extension was approved by the funder with no further interim analyses to be conducted. The funder eventually agreed to a final recruitment target of 1996 patients, under a revised assumption of an overall event rate of 10% and difference of 4% between mattress groups (corresponding to a HR of 0.652) with 80% power, requiring 172 events to be observed. This resulted in the trial design being modified to have one final analysis. eDocument 6: Statistical Analysis PlanIntroductionDesignPRESSURE 2 is a randomised controlled trial involving up to 2954 patients comparing High Specification Foam (HSF) and Alternating Pressure Mattresses (APMs). The trial is a multicentre, open, randomised, double triangular group sequential, parallel group trial, with two planned interim analyses. A maximum of 2954 consenting high risk patients from secondary and community in-patient facilities with evidence of acute illness will be randomised via minimisation on a 1:1 basis to receive either HSF or APM in conjunction with an electric profiling bed. Since this is a group sequential trial with two interim analyses the patient numbers are a maximum and may be lower depending on the results of the interim analysis. The group sequential design provides an efficient design through the possibility of early stopping for demonstrating either futility of the trial or inferiority of either mattress. Treatment phase follow-up assessments will be undertaken until the end of the treatment phase (up to a maximum of 60 days). These will be undertaken by a trained registered healthcare professional/clinical research nurse twice weekly from randomisation up to 30 days, then once weekly up to a maximum of 60 days. The treatment phase is defined as the period from randomisation to discharge or transfer from an eligible in-patient facility or 60 days, or when the patient is considered no longer at high risk, whichever is soonest. A final skin assessment will be undertaken 30 days (-2/+7 days) from the end of the treatment phase.The main trial will be supplemented with a Quality of Life (QoL) sub-study for responsiveness validation of the PU-QoL-P instrument.The trained registered healthcare professional/clinical research nurse will conduct the assessment of skin sites using International guidelines 1. As it is not possible to blind participants, or the Tissue Viability Team (TVT), a validation sub study, using photography with blinded central review and expert clinical assessment of the skin sites, will be carried out to assess any bias in the reporting (over or under -reporting) of category 2 or above pressure ulcers. Aim and objectivesThe aim of this study is to determine the clinical and cost effectiveness of HSF and APM when both are used in conjunction with an electric profiling bed frame in secondary and community in-patient facilities with evidence of acute illness, for the prevention of Category 2 (and above) pressure ulcers. Primary ObjectiveThe primary objective is to compare the time to developing a new category 2 or above pressure ulcer, in patients using HSF to those using APM by 30 days post end of treatment phase. Secondary objectivesTo compare the time to developing a new category 3 or above pressure ulcer, between patients using HSF and those using APMTo compare the time to developing a new category 1 or above pressure ulcer, between patients using HSF and those using APMTo compare the time to healing of pre-existing Category 2 pressure ulcers between patients using HSF and those using APM To determine the impact of HSF and APM on health related quality of life To determine the incremental cost-effectiveness of APM compared to HSF from the perspective of the health and social care sectorsTo compare incidence of mattress change between patients using HSF and those using APMTo compare safety between patients using HSF and those using APM. Secondary validation objectivesTo assess the responsiveness of the PU-QOL-Prevention (PU-QOL-P) instrumentTo determine the extent of under and over-reporting of category 2 and above PUsTo assess the feasibility of using photographs for pressure ulcer assessmentSample size and expected accrualOriginal sample size and expected accrualA maximum of 588 events (patients developing a new Category 2 or above PU), corresponding to 2954 patients, are required for the study to have 90% power for detecting a difference of 5% in the incidence of Category 2 and above pressure ulcers between APM and HSF, assuming an incidence rate of 18% on APM and 23% on HSF, (corresponding to a hazard ratio of 0.759), 2-sided significance level of 5%, and accounting for 6% loss to follow-up. The category 2 or above PU incidence rate for APM of 18% was estimated on the ITT population for PRESSURE 1 and hence the sample size estimate incorporates the effect of non-compliance. The sample size accounts for multiplicity in the interim analyses using Lan-DeMets α and β spending functions.Pressure ulcer incidence rates cannot be estimated accurately for the HSF and the maximum sample size estimate is based on the detection of the smallest relevant difference of 5% (clinical opinion). If the difference is greater than 5% then the trial will have sufficient power to stop early having demonstrated superiority (or inferiority) of the APM; if the difference is lower than 5% then the trial is likely to stop early for futility. The first planned interim analysis to be conducted after 300 patients have developed a new category 2 or above pressure ulcer (~1508 patients overall) corresponds to the earliest time point at which the trial can be stopped for demonstrating overwhelming evidence of efficacy or futility, and also corresponds to the minimum number of events required for conducting the economic evaluation.The second planned interim analysis, conducted after 445 patients have developed a new category 2 or above pressure ulcer (~2236 patients overall) corresponds to the number of expected events required for trial termination under futility (with 434 corresponding to the number of events required for demonstrating superiority or inferiority of APM to HSF)Revised sample size and expected accrualThe PRESSURE2 trial recruited patients at a much slower rate than originally anticipated. An unplanned interim analysis was conducted in November/December 2015 at the request of the funder and the results of this analysis were reviewed by the DMEC. The DMEC and TSC supported a no-cost extension request which was submitted in June 2016 to continue recruitment until the end of November 2016 by which time approximately 1996 patients were expected to be recruited. A costed extension request to continue recruitment until the end of May 2017 was also supported by the TSC and DMEC. The funder approved the no-cost extension request in July 2016 and therefore the final sample size is 2030 (achieved November 2016). Expected sample sizes for secondary validation objectivesTo assess the responsiveness of the PU-QOL-Prevention (PU-QOL-P) instrumentThe approximate sample size for the PUQoL-P sub-study will be 500 patients.To determine the extent of under and over-reporting of category 2 and above PUsA maximum of ~1653 photographs are expected for the central blinded review, which will enable Kappa to be estimated to within a precision of at least ±0.044 (corresponding to the half width of the 95% CI), assuming 65% of photographs are of Category 2 or above PUs and Kappa ≥0.5.Planned analysesThe Data Monitoring and Ethics Committee (DMEC) will review the safety and ethics of the PRESSURE 2 trial at agreed intervals. The DMEC will be presented with detailed open and closed reports (containing aggregated overall data and unblinded data respectively) containing a summary of recruitment, data collection, mattress compliance, a review of adverse events (AEs), related & unexpected SAEs (RUSAEs) and other safety issues. In addition to the above reports, interim analyses have been presented to the DMEC in strict confidence. This committee, in light of the interim data, and of any advice or evidence they wished to request, had the opportunity to advise the Trial Steering Committee (TSC) if there was proof beyond reasonable doubt that the trial should have been stopped in accordance with the planned stopping rules ( REF _Ref466878365 \r \h \* MERGEFORMAT Appendix F).The first planned interim analysis was planned to be conducted after 300 events (~1508 patients) which corresponded to the earliest time point at which the trial could be stopped for demonstrating overwhelming evidence of efficacy or futility, and also corresponded to the minimum number of events required for conducting the economic evaluation. The futility boundaries were constructed as non-binding in order for the DMEC to overrule a decision of stopping early for futility in the event that a futility boundarywas crossed. In the event of the DMEC recommending that the trial was stopped for futility using the pre-defined stopping criteria, an Expected Value of Sample Information Analysis would have been undertaken to assess the value of additional sample information on the effectiveness parameter, to establish whether continuing the trial would have been valuable from the NHS decision makers’ perspectiveThe second planned interim analysis, to be conducted after 445 events (~2236 patients) corresponded to the number of expected events required for trial termination under futility (with 434 corresponding to the number of events required for demonstrating superiority or inferiority of APM to HSF)The final analysis was scheduled to take place after 588 events (~2954 patients) had occurred. A graphical representation of the double triangular group sequential design is provided in REF _Ref466878365 \r \h \* MERGEFORMAT Appendix F.Note that if a stopping boundary was crossed but the DMEC recommend continuing with the trial the stopping boundaries would be recalculated in EAST software assuming the first analysis had not been conducted 2. The overall power and type-I error rate of the trial will not change. This advice was provided by James Wason in response to a request for methodology assistance at the MRC Methodology adaptive designs hub. This is the statistical analysis plan for the final analysis of the PRESSURE2 trial. Unplanned analysesAs explained in section REF _Ref467664378 \r \h \* MERGEFORMAT 1.3.2, recruitment has been slower than anticipated for a number of reasons. The trial team submitted a no cost extension request to recruitment in June 2015 to enable 1996 patients to be recruited by the end of November 2016. The HTA (funder) reviewed this request in August 2015 and asked the DMEC to request the CTRU to conduct an unplanned interim analysis as soon as possible; this was conducted in November and December 2015 and the HTA have since approved the no cost extension request to continue recruitment until the end of November 2016 and a final sample size of 2030 patients was achieved. No further formal interim analyses are planned before the final analysis. Endpoints to be analysedPrimary EndpointThe primary endpoint is time to developing a new Category 2 or above PU from randomisation to 30 days from the end of the treatment phase.Secondary EndpointsTime to developing a PU of Category 3 or above from randomisation to trial completion. Time to developing a PU of Category 1 or above from randomisation to trial completion. Time to healing of pre-existing Category 2 pressure ulcers from randomisation to trial completion Health-related quality of life using SF-12 instrument. Incremental cost effectiveness of APM compared to HSF Mattress change during the treatment phase.Adverse eventsDerivation of primary endpointIn order to derive the time to developing a new category 2 or above pressure ulcer the derivation of whether a participant has developed a new category 2 or above pressure ulcer at each skin site and follow-up assessment needs to be defined. Each participant will have a minimum of 14 pre-specified skin sites (Spine/back, sacrum, left and right buttocks, ischial tuberosities, trochanters (hips), heels, ankles, and elbows) assessed at baseline and every follow up assessment thereafter. There will be the option at each assessment to add other additional skin sites that have not been pre-specified on the CRF. At each assessment:If the skin site is assessed as healthy skin then a 0’ is recorded. If the skin site has skin alterations then an ‘A’ is entered with the relevant sub category or description also entered ( REF _Ref464470433 \r \h \* MERGEFORMAT Appendix B). If there is pressure damage then the international pressure ulcer classification system is used to record the level of pressure ulceration ( REF _Ref464470433 \r \h \* MERGEFORMAT Appendix B). If the skin cannot be assessed, or there is another wound present, then the research nurse will enter ‘N/A’ for the skin assessment with the relevant sub category or description ( REF _Ref464470433 \r \h \* MERGEFORMAT Appendix B). Note that the data collection process above leads to repeated measures for each skin site. Therefore, before deriving whether a patient has developed a new Category 2 or above pressure ulcer (and the time to development), the derivation of whether a new Category 2 or above pressure ulcer has developed since baseline needs to be defined on a skin site basis ( REF _Ref464489367 \h \* MERGEFORMAT Figure 1). This will lead to a dataset that has one record per skin site. This data will then be used to derive the primary endpoint dataset which will have one record per patient that includes a variable denoting whether the patient developed a category 2 or above pressure ulcer at any skin site and the time to development of the first new pressure ulcer or censoring time. A summary of the derivation of the primary endpoint (time to first new Category 2 or above pressure ulcer on a patient level) is provided in REF _Ref467665471 \h \* MERGEFORMAT Figure 2with justification for the derivations in REF _Ref464489392 \n \h \* MERGEFORMAT Appendix C and REF _Ref464489397 \n \h \* MERGEFORMAT Appendix D.. Figure SEQ Figure \* ARABIC 1 Derivation of PU development on a skin site level (one record per skin site per patient)*Patients with a Category 3/4/U pressure ulcer should have been excluded from the trial according to the eligibility criteria, however they have been included in this derivation to account for patients that may have been incorrectly recruited to the trial. Figure SEQ Figure \* ARABIC 2 Derivation of time to PU development (one record per patient including whether they developed a PU and time to event/censor)Exceptions in the above derivation are where main skin sites are assessed as N/A - device ulcer. These will be reviewed centrally by the Chief Investigator by reviewing the skin assessment data entered on the database (the paper CRFs may also be referred to), without reference to the allocated mattress and may be assumed to be at least a category 2 pressure ulcer and if confirmed as a Category 2 or above PU will contribute to the primary endpoint in the primary analysis. The impact of this assumption will be assessed through a sensitivity analysis where these device ulcers are not assumed to be a Category 2 pressure ulcer in the derivation of the endpoint. Further to this, the further descriptions for skin sites assessed as ‘N/A’ will be reviewed by the Chief Investigator and Clinical Co-ordinator(s) (without reference to mattress allocation) and the assessment ‘N/A’ may be changed to ‘A’, ‘Category 1’, or ‘Category 2’ for the purposes of a sensitivity analysis. This is to acknowledge the subjectivity that may arise in the assessment of pressure damage and that some skin sites assessed as N/A could have been misclassified. Other skin sitesOther additional skin sites assessed at either the baseline and/or follow up visits will be reviewed without reference to the allocated mattress by the Chief Investigator (CI) and Clinical Co-ordinator/s as to their inclusion in the analysis (e.g. whether damage to the skin site could be pressure related). The skin site level endpoint (i.e. whether a skin site develops a category 2 or above pressure ulcer) for these other additional skin sites will be derived again (assuming a healthy skin assessment at baseline, if the skin site was not recorded at baseline) for the purposes of a sensitivity analysis.Time to development and censor variablesThe time to development of a new category 2 or above pressure ulcer and the value of the censor variable (a variable denoting whether the participant was observed to develop a category 2 or above pressure ulcer or not) will be derived as detailed in table 1. Table SEQ Table \* ARABIC 1 Derivation of the event/censor variable and the number of days to developing a new category 2 or above pressure ulcer (on a patient basis)ScenarioNumber of daysEvent/censor variablePatient is observed to develop a new category 2 or above pressure ulcer(s) before end of follow up Number of days between date of randomisation and date first new category 2 or above pressure ulcer was observed (i.e. date first recorded on CRF)EventPatient dies before end of follow-up not having developed a category 2 or above pressure ulcerNumber of days between date of randomisation and date of last evaluable skin assessment Censoredα at date of last evaluable skin assessment Patient withdraws from the trial before end of follow-up not having developed a category 2 or above pressure ulcerNumber of days between date of randomisation and date of last follow up assessment where an assessment on an evaluable skin site was madePatient is censored α at date of last evaluable skin assessment*Patient is not observed to develop a new category 2 or above pressure ulcer before the end of follow up (including patients who were lost to follow-up)Number of days between date of randomisation and date of last follow up assessment where an assessment on an evaluable skin site was made Patient is censored at date of last evaluable skin site assessment α Death and some withdrawals may be considered as competing risks as they prevent a pressure ulcer occurring, or being observed. The decision on whether specific reasons for withdrawal will be reviewed from a competing risks perspective by the CI blind to treatment allocation. See section REF _Ref466875807 \r \h \* MERGEFORMAT 5.2.3.2 for more details. * where patients were withdrawn due to clinical condition but a skin assessment was made on the same day the data used to derive the patients endpoint will be up until the assessment prior to the patient being withdrawn (i.e. the previous visit). This is due to ethical considerations around the patient’s data being used if they had lost capacity or, for example, they were receiving palliative care. The decision over whether to censor these patients earlier is made in conjunction with the CI (or delegate) and is based solely on the reason for withdrawal, blind to mattress provision, skin status and time since entering the study. Derivation of healing A Category 2 PU will be classified as healed if the same skin site is later recorded as healthy or altered skin. The time to healing will be calculated as the number of days between date of randomisation and date first Category 2 PU is observed to heal, or patients will be censored at date of last evaluable assessment in line with REF _Ref473729001 \h \* MERGEFORMAT Table 1.Missing dataAttempts will be made to retrieve missing data via a thorough data cleaning process. Every effort will be made to obtain complete dates for all key data and missing dates will be monitored.The levels of missing data and reasons for missing will be investigated, in addition to exploring the pattern of missing data over time by treatment group. Exploratory work will be done comparing patients with complete data to those with missing data for various variables including baseline characteristics. This is in order to assess the level of imbalance and help assess the reason for missing. The quantity of missing data will be monitored by treatment group and a summary of the number of patients with missing primary endpoint data and the quantity of missing data by treatment group and centre will be reported. Should key primary endpoint data (i.e. dates) be unavailable at the time of the final analysis, after all possible attempts to retrieve this missing data have been carried out, this data will be will be assessed (in the ways described above) and if assumed to be missing at random (MAR) will be imputed based on the date of the baseline visit and the visit number of the last evaluable skin assessment using the protocol visit schedule such as to comply with the intention-to-treat analysis principle 3. The baseline covariates that will be included in the primary analysis model consists the stratification factors used in the randomisation algorithm (Section REF _Ref467670395 \r \h \* MERGEFORMAT 5.2.3) and as such there will be no missing data for these covariates. However, there will be some baseline covariates that will be included in the primary analysis and secondary analyses such as presence of pressure-related pain for which there may be missing data. If missing data exist for baseline data we will assess whether the missingness is correlated with the outcome and singly impute the baseline data according to White and Thompson 4. Sensitivity analysisA sensitivity analysis will be conducted to assess the robustness of the conclusions to the missing data assumptions. This will consist of a complete case analysis whereby patients with incomplete data for the corresponding analysis will be excluded and the conclusions from the analyses will be compared.PopulationsEligibilityEligibility criteria are detailed in the PRESSURE2 protocol (P:\CTRU\Projects\CHRD\Skin\PRESSURE II\TC\03_Protocol\3c_Current Protocol). Note that eligibility waivers are not permitted. Intention to treat populationThe intention to treat (ITT) population will include all patients randomised to treatment for whom consent to the main trial has been obtained. Allocation to treatment is by minimisation with respect to the following factors: centre, healthcare setting (Secondary care hospital, community hospital / intermediate care or rehabilitation facility), pressure ulcer status (no pressure ulcer/Category 1/Category 2), and consent (written/witnessed verbal/consultee agreement). Patients will be included in all statistical analyses using the correct data (i.e. using the data recorded on the randomisation CRF (F03) rather than the 24 hour system). The ITT population will be analysed by allocated treatment. Per protocol population The per-protocol population will exclude all major protocol violators, this includes:Patients who do not fulfil all major eligibility criteria. Major eligibility criteria includes:Patients must have evidence of acute illnessPatients must be at high risk of PU development Patients must be on an electric profiling bed framePatients must not have previously participated in the trialPatients must not have a current or previous category 3 PU at baselinePatients must be within the acceptable weight limitsPatients for whom consent to the main trial has not been obtainedPatients whose date of consent to the main trial is after the date of randomisationPatients who do not receive their randomised mattress within 2 days of randomisation Patients who spend less than a pre-specified proportion (60%) of their follow up time on their allocated mattress An analysis will be conducted on the per-protocol population by allocated mattress. Per-protocol violators will be monitored by centres who will complete the protocol deviation form (F42) and return it to the CTRU. The Trial Statistician will then present overall summaries of other protocol deviations at Internal Project Team meetings and Trial Management Group (TMG) meetings and a decision will be made over whether other deviations noted warrant exclusion from the per protocol population without reference to the endpoint data or treatment group allocation. Note that a compliance adjusted analysis has been considered with the TSC and DMEC but is not considered achievable with currently available methods ( REF _Ref466878462 \r \h \* MERGEFORMAT Appendix E).Safety populationThe safety population will be defined as all patients who were recruited into the PRESSURE 2 trial and will be used to summarise patient disposition during the trial. This population will be used to summarise any adverse events (AEs) and Related Unexpected Serious Adverse Events (RUSAEs) that occur and will be summarised by allocated mattress and actual mattress at the time of the AE. Time to healing of category 2 PU populationThe analysis of time to healing of pre-existing category 2 PUs will be conducted for the ITT population excluding all patients who did not have a category 2 PU at baseline. Quality of life populationThis will not be evaluated by the Trial Statistician and a separate analysis plan will be written for this population.Blinded validation sub study populationsTo assess over-reportingTo assess over-reporting a sensitivity analysis population will be defined. This will be the ITT population but skin assessments will be replaced with the outcome of the blinded central review where appropriate. That is, for skin sites assessed as category 2 or above by the research nurse, a photo should have been taken of the category 2 or above PU (providing consent has been provided by the participant) and sent for blinded review. The outcome of the central review will replace the skin assessment that took place closest to the date of the photo. If a photo was not taken (either because consent was not provided or because of a logistical reason such as the camera being unavailable) then the research nurse assessment of the category 2 or above PU will remain in the dataset and used in the sensitivity analysis. To assess under-reportingIn addition to photos being taken of category 2 PUs, the CTRU will randomly identify 10% of patients from each centre for whom an independent TVTM at that centre (i.e. they are blind to the research nurse’s skin assessments) will carry out a skin assessment of all 14 skin sites. In addition to these skin assessments the independent TVTM will also take photographs of 2 pressure areas (1 torso and 1limb/other) regardless of the PU status (note that if a PU is present then 1 photo will be of the PU and the second photo will be of a PU-free skin site), and providing the patient has given consent to have the photos taken. These 10% of participants will form the analysis population to assess under-reporting.Data HandlingData monitoringDay to day monitoring for completeness and quality of trial data will be conducted centrally by the Trial Co-ordinator/Data Manager or their delegate. Every effort will be made to ensure that as much data as possible is available and that reasons for unobtainable data are obtained. Missing data, with the exception of participant reported data will be chased until it is received, confirmed as not available, or the study is at final analysis. In addition, the CTRU also monitor consent forms centrally. Any issues with data collection will be discussed at Internal Project Team meetings and, if appropriate, at Trial Management Group meetings.The CTRU reserves the right to intermittently conduct source data verification exercises on a sample of participants. Source data verification will involve direct access to participant healthcare records and other relevant investigation reports at participating centres. Data validationThe database validates most dates and specific data items in line with the pre-programmed validation rules in real time, as data are entered. The key data items relate to the data required for the primary endpoint. The key data items list (saved here: P:\CTRU\Projects\CHRD\Skin\PRESSURE II\TC\12_DataManagement\Key Data Item List\Key Data Item List.xlsx) for PRESSURE2 will be 100% checked for validation by the Trial Co-ordinator / Data Manager or their delegate.SAS will be used to validate the data and check for any missing or inconsistent data. The data will be downloaded and read into permanent SAS data sets. The names and contents of the variables can be found on the annotated final database specifications in the Statistician’s Trial File. Other checks to be performed include:Checking eligibility of all randomised participantsChecking for consistency between stratification factors provided to the 24hr randomisation system and the randomisation CRF (F03)Checking that participants allocated to the validation sub study have had the skin verification sub study CRF (F33) completed Checking that participants who have had a category 2 or above pressure ulcer reported have had the photography CRF (F32) completed for each category 2 or above pressure ulcer reportedChecking that participants who have had a category 2 or above pressure ulcer reported at baseline have had the duration and area CRF (F34) sufficiently completed for each category 2 or above pressure ulcer reported at baselineSequential datesChecks for unusual and outlying dataChecks for inconsistent dataChecks for missing data Other checks as deemed appropriateAny suspicious or inconsistent data identified via these checks will be noted and an e-mail sent to the Trial Co-ordinator / Data Manager responsible for the study. The Clinical Co-Ordinator/Trial Co-Ordinator / Data Manager or their delegate will check such inconsistencies against the participant forms. If there has been an error in data input causing such inconsistencies this will be corrected on the database. If there has been no error in data input, a query requesting clarification will be sent to site. Details of corresponding changes will be documented. Data AnalysisGeneral calculationsUnless otherwise stated, all percentages will be calculated using the total number of patients or forms expected as the denominator (i.e. including all patients with missing data for that variable). All percentages, means, medians, interquartile ranges (where appropriate) and also ranges will be rounded to 1 decimal place (or 1 significant figure for numbers less than 1), whilst standard deviations will be rounded to 2 decimal places (or 2 significant figures for numbers less than 1). P-values will be rounded to 4 decimal places (those less than 0.0001 will be displayed as <0.0001), whilst parameter estimates, standard errors, hazard ratios and corresponding confidence intervals will be reported to 2 decimal places (or 2 significant figures for numbers less than 1). All analyses will be carried out using SAS unless otherwise stated.AnalysisScreening summariesScreening data, including age, gender, ethnicity and current mattress type will be summarised using frequencies and summary statistics. The screening process will be summarised using a flow diagram. If one of the screening questions, for example ‘Eligibility assessed?’, is missing and no further questions have been completed then it will be assumed that the answer to the corresponding screening question is ‘No’. Else, if any subsequent screening questions have been completed then it will be assumed that the answer was ‘Yes’. The screening database and main database will be linked, where possible, using screening numbers and trial numbers. The screening logs will be monitored on an on-going basis to ensure that they are as accurate as possible, but it is anticipated that there will be some differences. Since the screening log is anonymised it would be very difficult to work out how to correct the differences, and so ball park figures will be given, i.e. if the screening log has identified 550 as being randomised, but there are actually 560 randomised participants then it will be assumed that these extra 10 participants are not on the screening log and will therefore be added to each stage of the screening process. Similarly if the screening log has identified 560 patients as being randomised, but there are actually 550 randomised participants then it will be assumed that these extra 10 patients on the screening log had consented but were not actually randomised.Baseline characteristicsBaseline patient and clinical data as recorded on the randomisation (F03) and baseline assessment form (F02) will be tabulated using frequencies and summary statistics for each treatment group and overall. Missing or unobtainable data will be included as missing unless data are available from the 24-hour randomisation form. Summaries of the number of incorrect data on the 24-hour randomisation form compared to F03 will be produced. Primary endpoint analysisThe Primary endpoint analysis will be conducted on both the ITT population and on the per-protocol population (Appendix 4).The incidence rate of patients developing a new category 2 or above pressure ulcer will be summarised by treatment group together with the number of new category 2 or above pressure ulcers per patient and the location of new category 2 or above pressure ulcers. In addition, the incidence rate of new category 2 or above pressure ulcers will be summarised for each covariate planned to be included in the primary analysis using frequencies for each treatment group and overall. After feedback from research nurses that Incontinence Associated Dermatitis (IAD) could be considered as skin alterations rather than ‘N/A’, summaries of whether the incidence rate of patients developing a new category 2 or above PU would change assuming IAD to be equivalent to skin alterations will be produced. It is not appropriate to conduct a sensitivity analysis on this data as IAD is often categorised in line with pressure ulcer reporting and may in fact be a category 2, and we do not collect this additional category data for IAD. In addition, it is anticipated that there will be very few ‘other’ skin sites (see section 2.2.1.1) that are considered non-evaluable. Summaries on how the primary endpoint might change if these ‘other’ skin sites were included in the derivation of the endpoint. Proportional hazards and no competing risksIf proportional hazards can be assumed and there are considered to be no (or minimal) competing risks then aCox Proportional Hazards (Cox PH) model will be fitted to the primary endpoint, with the following minimisation factors fitted as fixed effects: healthcare setting, pressure ulcer status and consent, in addition, the following covariates will also be fitted as fixed effects: presence of pain on a healthy, altered or category 1 skin site, conditions affecting peripheral circulation. The effect of adding treatment group to this model will be assessed using a likelihood ratio test. Centre will be fitted as a random effect. The hazard ratio for the treatment effect and adjusted confidence interval corresponding to the nominal p-value will be presented. The proportional hazards assumption and adequacy of the Cox PH model will be explored through examination of:Plot of Schoenfeld residuals versus timePlot of ln(cumulative hazard) versus timeFormal test of whether treatment effect varies over time by fitting an interaction of treatment and ln(time)“Assess” statement in the SAS PHREG procedureThe probability of patients (using cumulative incidence functions) developing a Category 2 or above pressure ulcer over the length of trial participation in each group and adjusted confidence intervals corresponding to the nominal p-value will be presented.Monitoring will also be conducted in East software and the analysis output will also be presented in a Double Triangular Group Sequential Design diagram to aid interpretation of the results.Proportional hazards and competing risksIt is recognized that there may be competing risks, for example death or withdrawal from follow up which prevent the event of interest from occurring or being observed, and the reason for this may be such that the patients risk of developing a pressure ulcer has changed. These will be considered upon inspection of the data, by the CI who will be blind to treatment group, and if considered to be substantial, and the hazards of the cumulative incidence function are proportional then a Fine and Gray model 5 will be fitted to the primary endpoint, with the following minimisation factors fitted as fixed effects: healthcare setting, pressure ulcer status and consent, in addition, the following covariates will also be fitted as fixed effects: presence of pain on a healthy, altered or category 1 skin site, conditions affecting peripheral circulation. The effect of adding treatment group to this model will be assessed using the likelihood ratio test. Centre will be fitted as a random effect. The model assumptions will be checked according to 5.In addition, for patients who have died or withdrawn, the number of days between the last evaluable assessment and the date of death/withdrawal will be summarised, by treatment arm. Non proportional hazards and no competing risksIt is recognised that the proportional hazards assumption may not hold, therefore in this scenario if there are considered to be no (or minimal) competing risks then the Restricted Mean Survival Method (RMST) 6 will be used to compare the mean survival times between treatment groups. SAS software does not have an in built function to conduct this analysis therefore STATA will be used as described in 6,7 and the relevant model assumptions will be checked. Non proportional hazards and competing risks If competing risks are considered to be substantial and the proportional hazards assumption does not adequately hold for the Fine and Gray model then the methods for dealing with this data are limited. In this scenario we will conduct Gray’s test which is an analogue of the log rank test to compare two or more groups in the presence of competing risks 8. The disadvantage to this method is that it will not provide an estimate of the treatment effect, but we will explore whether there are confounding variables that may affect the proportional hazards assumption in order to conduct an analysis that provides an estimate of the treatment effect. We will also conduct the analyses described in sections REF _Ref466875807 \r \h \* MERGEFORMAT 5.2.3.2 and REF _Ref466875811 \r \h \* MERGEFORMAT 5.2.3.3 as sensitivity analyses. Sensitivity analysis based on treatment phaseThe time to pressure ulcer development within the treatment phase only will be derived (i.e. ignoring the 30 day post treatment phase visit) and a sensitivity analysis will be conducted on this endpoint to evaluate whether there is evidence of a treatment effect within the treatment phase. The analysis will be conducted as in line with sections REF _Ref466876074 \r \h \* MERGEFORMAT 5.2.3.1 to REF _Ref466876076 \r \h \* MERGEFORMAT 5.2.3.4.Sensitivity analysis for non-evaluable participantsThe analysis conducted on the primary endpoint will be repeated to assess the robustness of the missing data assumptions for participants who did not have an evaluable endpoint and were included in the primary analysis through imputation of dates; A complete case analysis will be conducted where patients will be excluded if the date of their last evaluable skin assessment is missing. Moderator & mediator analyses: Moderator analysesModerator analyses will explore whether the treatment effect depends on baseline characteristics of the patient. These will be conducted by building separate models for each of the following potential predictors of response (time to developing Category 2 or above pressure ulcer):pre-existing pressure ulcersCategory A skin statusdiabetesagemobilitysensory perceptionmacro and micro circulatory functionnutritional statusskin moistureEach of these models will take the form of the primary analysis model with the addition of the potential moderator and the interaction of the treatment variable with the moderator. If the interaction is observed to be statistically significant in the model then the variable will be said to moderate the treatment effect, although the exploratory nature of these analyses will be recognised in the interpretation of the results. Mediator analysesMediator analyses will explore the relationship between treatment group and the following potential mediator variables:length of staytime on allocated mattresspatient turning*use of specialist cushions*heel protectors*protective dressings*The data for the above variables indicated by * have been collected in a simple way and therefore there is a query over the validity and reliability of these data. Further, the length of stay and time on allocated mattress may be dependent on outcome (e.g. stay and time on allocated mattress may be longer/shorter if the patient develops a PU). Therefore the data will first be summarised by source (e.g. patient reported/records), by centre and by PU outcome to explore whether there are any systematic differences. If there are considered to be systematic differences then further analysis of these data will not be conducted because of the risk of bias. If there do not appear to be systematic differences that the Baron and Kenny method 9 will be used to explore whether any mediators exist. Secondary endpoint analysisTime to developing a PU of at least Category 3 The time to development of a category 3 or above PU will be derived in line with the derivation of the primary endpoint but adjusted for the development of a more severe pressure ulcer (i.e. Category 3, see REF _Ref473728748 \r \h \* MERGEFORMAT Appendix G) rather than a Category 2. The analysis on this endpoint will be conducted in line with the analysis outlined in section REF _Ref466877177 \r \h \* MERGEFORMAT 5.2.3Time to developing a PU of at least Category 1 The time to development of a category 1 or above PU will be derived in line with the derivation of the primary endpoint but adjusted for the development of a less severe pressure ulcer (i.e. Category 1, see REF _Ref473728755 \r \h \* MERGEFORMAT Appendix H) rather than a Category 2. The analysis on this endpoint will be conducted in line with the analysis outlined in section REF _Ref466877177 \r \h \* MERGEFORMAT 5.2.3. 5.2.2.3Time to healing of pre-existing Category 2 pressure ulcersThe time to healing of pre-existing Category 2 pressure ulcers will be analysed in line with the analysis methods described in section REF _Ref466877177 \r \h \* MERGEFORMAT 5.2.3. Note that for this endpoint patients who do not have their Category 2 PU heal during the intervention period and by 30 days post trial completion will be censored at 30 days post trial completion, or else at the date of withdrawal, death or loss to follow-up, accounting for competing risks in the same way as described in section REF _Ref466877972 \r \h \* MERGEFORMAT 2.3.2.The incidence rate of healing of pre-existing category 2 PUs will be summarised by treatment group together with the number of healed pre-existing category 2 PUs per patient and the location of healed pre-existing category 2 PUs. In addition the incidence rate of healing of pre-existing category 2 PUs will be tabulated for each covariate to be included in the analysis using frequencies and summary statistics for each treatment group and overall. SafetyAEs and SAEs classified as related to the mattress or resulting from administration of any research procedures will be listed and summarised by allocated mattress and mattress at the time of the AE. Validation sub study analysisComplianceThe consent rates for having photographs taken of skin sites will be summarised, at baseline and during follow up, by treatment arm, by centre and overall. Compliance with sending photographs will be summarised by treatment arm, by centre and overall and will include how frequently photographs were: unable to be assessed by the central review team (for example because they were blurry or the grey scale card was not present) not taken by the research team for logistical or research team reasonsnot sent to the CTRU for logistical or research team reasons. Assessing risk of over-reporting (Photographs of all Category 2 PUs)A sensitivity analysis on the primary endpoint will be conducted such that the model used in the primary analysis will be applied to the population defined in section REF _Ref473790608 \r \h \* MERGEFORMAT 3.7.1.Assessing risk of under-reporting (10% sample)For each skin site assessed by the independent TVTM, 2 by 2 tables will be produced by arm and overall to summarise whether the skin assessments made by the PI and the skin assessments made by the research nurse/registered health care professional (RHCP), agree on the category 2 or above PU status. In addition, a 2 by 2 table summarising the agreement overall skin sites (i.e. on a patient basis) will be produced. The following statistics will be reported for each of these tables: Kappa and Prevalence and Bias Adjusted Kappa.Reliability of photographsSummaries, by arm and overall, of the outcome of the blinded central review of photographs compared to the clinical skin assessment made by either the research nurse or independent TVTM will be produced, and corresponding agreement statistics (Kappa and Prevalence and Bias Adjusted Kappa) will be reported for each of the following cases:Assessments drawn from the photographs (via blinded central review) and the skin assessments made by the independent TVTM, for each skin site for which a photograph was taken and overall area (i.e. torso/limb)Assessments drawn from the photographs taken by the research nurse/RHCP because a Category 2 or above PU was observed and the skin assessments made by the research nurse/RHCP for each skin site for which a photograph was taken. Where poor agreement is observed the data will be explored, for example the number of days between the photograph taken by the independent TVTM and the clinical assessment made by the research nurse/RHCP will be summarised because these may not have been conducted on the same day and skin appearance can change quickly. Any additional skin descriptors (for N/A or alterations) will also be examined. Patient DispositionThe number of patients who completed the study, died, withdrawn (with reasons where available), and who were lost to follow-up will be presented by centre, treatment arm and overall. The length of treatment phase and length of overall time in the study will also be summarised by treatment arm and overall. Protocol violators/deviationsProtocol violations and protocol deviations will be described by treatment arm and overall. This includes those found not to satisfy the eligibility criteria after randomisation and those who do not receive their randomised mattress throughout the treatment phase. Reasons for those violating the eligibility criteria will be summarised by treatment arm and overall. Mattress complianceA summary of the time to receiving allocated mattress, reasons for not receiving mattress on the day of randomisation (day 0) and the number of mattress changes together with reasons will be summarised by treatment arm and overall. The number of days and the proportion of time spent on the randomised mattress during the treatment phase will be summarised using descriptive statistics and frequency distributions by treatment arm and overall. In addition, for those allocated to APM, a summary of the number of times the mattress is observed to not be working correctly together with reasons will be presented. CONSORT diagramThe CONSORT diagram 10 will be produced to summarise the course of patients throughout the study and the analysis populations included reasons for exclusion from analyses (e.g. derivation of the per-protocol population). Reporting and dissemination of the resultsThe success of the trial depends upon the collaboration of all participants. For this reason, credit for the main results will be given to all those who have collaborated in the trial, through authorship and contributorship. Uniform requirements for authorship for manuscripts submitted to medical journals will guide authorship decisions. These state that authorship credit should be based only on substantial contribution to: conception and design, or acquisition of data, or analysis and interpretation of datadrafting the article or revising it critically for important intellectual contentand final approval of the version to be publishedand that all these conditions must be met ().In light of this, the Chief Investigator, co-applicants and relevant senior CTRU staff will be named as authors in any publication. In addition, all collaborators will be listed as contributors for the main trial publication, giving details of roles in planning, conducting and reporting the trial.To maintain the scientific integrity of the trial, data will not be released prior to the first publication of the analysis of the primary endpoint, either for trial publication or oral presentation purposes, without the permission of the Trial Steering Committee. In addition, individual collaborators must not publish data concerning their participants which is directly relevant to the questions posed in the trial until the first publication of the analysis of the primary endpoint.Approval of Analysis PlanClinical Trials Research Unit (CTRU)The following final analysis plan, February 2017, for the PRESSURE 2 study has been approved by the following personnel. Any signed amendments to the plan will be filed with this document.Trial Statistician: ________________________________________________________________________Date: _________________________________________________________________________________Supervising Statistician: __________________________________________________________________Date: _________________________________________________________________________________Data Manager: _________________________________________________________________Date: _________________________________________________________________________________Trial Co-ordinator: _________________________________________________________________Date: _________________________________________________________________________________Clinical Co-ordinator: _________________________________________________________________Date: _________________________________________________________________________________Clinical Co-ordinator: _________________________________________________________________Date: _________________________________________________________________________________Clinical Co-ordinator: _________________________________________________________________Date: _________________________________________________________________________________CTRU Project Delivery Lead: _______________________________________________________________Date: _________________________________________________________________________________Chief Investigator: _______________________________________________________________________Date: _________________________________________________________________________________Additional information:Graphical representation of trial designSkin classification incorporating the NPUAP/EPUAP PU Classification SystemCategoryDescriptionCategory 0Healthy intact skinNo skin changesCategory AAlterations to intact skinAlterations to intact skin. Please specify with sub-category code:001 = Blanching redness which persists011 = Papery thin002 = Bruising – red hue012 = Cracks/calloused003 = Bruising – purple hue013 = Spongy004 = Scar014 = Macerated005 = Oedema015 = Scratches006 = Cellulitis016 = Rash007 = Lymphodema017 = Scab008 = Discoloration – ischaemia018 = Induration009 = Discoloration – cyanosis019 = Heat010 = Dry/flaky999 = None of the above, please describeCategory 1Non-blanchable erythema of intact skinIntact skin with non-blanchable erythema of a localised area usually over a bony prominence. Discoloration of the skin, warmth, oedema, hardness or pain may also be present. Darkly pigmented skin may not have visible blanching. Category 2Partial thickness skin loss or blisterPartial thickness loss of dermis presenting as a shallow open ulcer with a red pink wound bed, without slough. May also present as an intact or open/ruptured serum or sero-sanginous-filled blister.Category 3 Full thickness skin lossFull thickness tissue loss. Subcutaneous fat may be visible but bone, tendon or muscle are not exposed. Some slough may be present. May include undermining and tunnelling.Category 4Full thickness tissue lossFull thickness tissue loss with exposed bone, tendon or muscle. Slough or eschar may be present. Often includes undermining or tunnelling. Category UUnstageableFull thickness skin loss in which actual depth of the ulcer is completely obscured by slough (yellow, tan, grey, green or brown) and/or eschar (tan, brown, or black) in the wound bed. Category N/A Not applicableSpecify with sub-category code: 001 = Amputation007 = Device-related ulcer002 = Bandage in situ008 = Surgical wound/bruising003 = Cast in situ009 = Traumatic wound/bruising004 = Dressing in situ010 = Dermatological skin condition e.g. eczema005 = Incontinence associated dermatitis011 = Unable to assess006 = Other chronic wound999= None of the above, please describe Derivation of skin site level outcome Table SEQ Table \* ARABIC 2 Derivation with justification of whether a new category 2 or above pressure ulcer develops on a skin site basis (i.e. for each separate skin site)ScenarioBaseline skin assessmentFollow up skin assessmentsNew category 2 or above pressure ulcer?Reasoning1Category 0, 1 or ANo follow upUnknownNo follow up assessments to identify whether the skin site does or doesn’t develop a new category 2 or above pressure ulcer4Category 0,1, or AAt least one follow up assessment is category 2 or aboveYesThe skin site was assessed as healthy, altered or a category 1 pressure ulcer at baseline. A category 2 or above pressure ulcer is reported during follow up and therefore we conclude that a new category 2 or above pressure ulcer has developed.3Category 0, 1 or AAll “actual” follow up assessments are category 0,1, or A (although some could be ‘N/A’ or missing)NoThere has been at least one follow up assessment from which we could say no new category 2 or above pressure ulcer developed5Category 0,1, AAll follow up assessments are ‘N/A’ or missingUnknownThere are no follow up assessments to determine whether or not this skin site develops a new category 2 or above pressure ulcer2Category 2 or aboveAny number of follow up skin assessments have taken placeN/AThe baseline skin assessment shows the skin site to have an existing category 2 or above pressure ulcer therefore a new category 2 or above pressure ulcer cannot develop at this skin site6N/AAny number of follow up skin assessments have taken placeUnknownThe baseline assessment cannot inform whether this skin site develops a new category 2 or above pressure ulcerExceptions in the above derivation are where main skin sites are assessed as having a device ulcer. These will be reviewed centrally without reference to the allocated mattress and may be assumed to be at least a category 2 ulcer and could contribute to the primary endpoint in the primary analysis. Derivation of patient-level outcomeTable SEQ Table \* ARABIC 3 Derivation with justification of whether a participant develops a new category 2 or above pressure ulcerNew category 2 or above pressure ulcer on a skin site level basis?New category 2 or above pressure ulcer on a participant level basis?ReasoningUnknown or N/A (for all skin sites)UnknownParticipant has no evaluable skin sites (e.g. there were either no follow up assessments or the baseline skin assessments were insufficient)At least one yesYesParticipant develops a new category 2 or above pressure ulcer on at least one skin siteNo (for all skin sites)NoParticipant does not develop new category 2 or above pressure ulcer during follow up at any skin siteNo’s and unknown’sNoBased on the evaluable skin sites the participant did not develop a new category 2 or above pressure ulcer Compliance adjusted analysesThe TSC and DMEC have previously recommended conducting analyses in addition to the per protocol analysis whereby the causal effect of the two mattresses are assessed taking into account compliance with allocated mattress and respecting the randomisation process. However, because PRESSURE2 is a pragmatic trial where the trial interventions are used in practice there are multiple considerations in terms of ‘treatment switching’. These include situations where the patient could receive their allocated mattress; switch to the alternative trial mattress or a non-trial mattress and, switch back again to their allocated mattress multiple times. This can be due to ward transfers or the patients’ level of risk changing for example. This complex level of mattress compliance has been discussed with methodologists including Ian White who was previously contacted regarding how to deal with mattress compliance in the original PRESSURE1 trial 11 and they were unable to find a solution. Ian White is keen to explore this again with the PRESSURE2 data but it is deemed to be infeasible to do for this unplanned interim analysis. A pragmatic solution at this stage is to include patients who have spent a minimum proportion of their treatment phase spent on their allocated mattress in the per protocol population. Scenarios to consider at interim analysisDerivation of Category 3 PU development on a skin site level *Patients with a Category 3/4/U pressure ulcer should have been excluded from the trial according to the eligibility criteria, however they have been included in this derivation to account for patients that may have been incorrectly recruited to the trial. Derivation of Category 1 PU development on a skin site level*Patients with a Category 3/4/U pressure ulcer should have been excluded from the trial according to the eligibility criteria, however they have been included in this derivation to account for patients that may have been incorrectly recruited to the trial. eDocument 7: Skin sites of Pressure Ulcers7a: Baseline skin sites and classificationsAPMHSFOverallSkin classification all skin areas*Normal9443(66.4%)9438(66.5%)18881(66.4%)Category A3764(26.5%)3805(26.8%)7569(26.6%)Category 1212(1.5%)197(1.4%)409(1.4%)Category 291(0.6%)86(0.6%)177(0.6%)Suspected Deep Tissue Injury0(0.0%)1(0.0%)1(0.0%)Not applicable 596(4.2%)520(3.7%)1116(3.9%)Missing124(0.9%)144(1.0%)268(0.9%)Totals14230(100%)14191(100%)28421(100%)Skin sites with pre-existing Category 2 PUSpine/back3(3.3%)1(1.2%)4(2.3%)Sacrum23(25.3%)23(26.7%)46(26.0%)L Buttock17(18.7%)18(20.9%)35(19.8%)R Buttock16(17.6%)17(19.8%)33(18.6%)L Ischial3(3.3%)0(0.0%)3(1.7%)R Ischial4(4.4%)0(0.0%)4(2.2%)L Trochanter1(1.1%)0(0.0%)1(0.6%)R Trochanter0(0.0%)0(0.0%)0(0.0%)L Heel6(6.6%)7(8.1%)13(7.3%)R Heel10(11.0%)6(7.0%)16(9.0%)L ankle1(1.1%)0(0.0%)1(0.6%)R ankle1(1.1%)2(2.3%)3(1.7%)L Elbow2(2.2%)5(5.8%)7(4.0%)R Elbow2(2.2%)3(3.5%)5(2.8%)Other2(2.2%)4(4.7%)6(3.4%)Total91(100.0%)86(100.0%)177(100.0%)Duration of most severe pre-existing PU (days)Mean (S.D.)29.6(42.44)67.6(179.73)48.3(129.61)Median (range)14(1.0, 182.6)7.5(1.0, 730.5)10(1, 730.5)< 7 days10(14.3%)13(17.3%)23(15.9%)8 to 28 days7(10.0%)7(9.3%)14(9.7%)More than 28 days10(14.3%)6(8.0%)16(11.0%)Missing43(61.4%)49(65.3%)92(63.4%)Total number of patients with a PU70(100.0%)75(00.0%)145(100.0%)Area of most severe pre-existing PU (mm2)Mean (S.D.)462.3(1357.4)245.4(392.4)352.0(989.2)Median (range)68.6(3.7, 5369.8)123.7(3.1, 1652.5)76.3(3.1, 5369.8)Number of patients with non-missing data293059Number of patients with missing data414586Left (L) Right (R) Standard Deviation (SD)7b: Location of all new Pressure UlcersLocationAPMHSFOverallSpine/back1(1.1%)6(4.8%)7(3.3%)Sacrum18(20.2%)26(21.0%)44(20.7%)Buttock L18(20.2%)24(19.4%)42(19.7%)Buttock R15(16.9%)23(18.5%)38(17.8%)Ischial Tuberosity L1(1.1%)3(2.4%)4(1.9%)Ischial Tuberosity R2(2.2%)3(2.4%)5(2.3%)Trochanter (Hip) L0(0.0%)2(1.6%)2(0.9%)Trochanter (Hip) R1(1.1%)0(0.0%)1(0.5%)Heel L12(13.5%)15(12.1%)27(12.7%)Heel R9(10.1%)13(10.5%)22(10.3%)Ankle L3(3.4%)2(1.6%)5(2.3%)Ankle R3(3.4%)2(1.6%)5(2.3%)Elbow L2(2.2%)0(0.0%)2(0.9%)Elbow R4(4.5%)5(4.0%)9(4.2%)Total89(100%)124(100%)213(100%)eDocument 8: Additional Results Tables8a: Time to development of new PU Category ≥2 during treatment phaseCovariateLevel of covariateIncidence rateReference levelHR point EstimateHR 95% Wald Confidence limitsWald P-valueTreatmentHSF79/1016 (7.8%)----0.0190*APM53/1013 (5.2%)vs HSF0.660.46 to0.93Skin statusNo PU95/1648 (5.8%)----0.0089PU Category 122/236 (9.3%)vs No PU1.831.11 to3.02PU Category 215/145 (10.3%) vs No PU1.981.12 to3.50Consent typeWritten84/1404 (6.0%)----0.7400Witnessed verbal25/303 (8.3%)vs Written1.140.73 to1.77Consultee agreement23/322 (7.1%) vs Written1.180.72 to1.92SettingSecondary care hospital85/1416 (6.0%)----0.8670Community hospital29/379 (7.7%) vs Secondary care hospital1.120.73 to1.73NHS intermediate care/ rehabilitation facility18/234 (7.7%) vs Secondary care hospital1.010.60 to1.71Pain on a healthy, altered or PU Category 1 skin siteNo51/890 (5.7%)----0.2081Yes78/1084 (7.2%)vs No 1.370.94 to1.99Unable to assess1/30 (3.3%)vs No0.440.06 to3.48Missing2/25 (8.0%)vs No2.570.56 to11.73Presence of condition affecting peripheral circulationNo97/1567 (6.2%)----0.5671Yes34/455 (7.5%)vs No1.130.76 to1.67Missing1/7 (14.3%)vs No2.650.32 to21.80*P-values obtained from corresponding likelihood ratio tests for the effect of treatment is 0.0176Table 8b: Time to development of new PU Category ≥1 by 30 day final follow-upCovariateLevel of covariateIncidence rateReference levelHR point EstimateHR 95% Wald Confidence limitsWald P-valueTreatmentHSF190/1013 (18.8%)----0.0741*APM160/1016 (15.7%)vs HSF0.830.67 to1.02Skin statusNo PU272/1648 (16.5%)----0.0301PU Category 150/236 (21.2%)vs No PU1.521.11 to2.09PU Category 228/145 (19.3%) vs No PU1.180.79 to 1.75Consent typeWritten222/1404 (15.8%)----0.0140Witnessed verbal59/303 (19.5%)vs Written1.150.86 to1.53Consultee agreement69/322 (21.4%) vs Written1.521.15 to2.01SettingSecondary care hospital226/1416 (16.0%)----0.0970Community hospital67/379 (17.7%) vs Secondary care hospital0.950.72 to 1.26NHS intermediate care/ rehabilitation facility57/234 (24.4%) vs Secondary care hospital1.351.01 to1.82Pain on a healthy or altered skin siteNo147/943 (15.6%)----0.0063Yes198/1029 (19.2%)vs No 1.381.11 to1.71Unable to assess2/30 (6.7%)vs No0.280.07 to1.15Missing3/27 (11.1%)vs No1.310.40 to4.36Presence of condition affecting peripheral circulationNo259/1567 (16.5%)----0.3258Yes90/455 (19.8%)vs No1.190.93 to1.51Missing1/7 (14.3%)vs No1.850.26 to13.12*P-values obtained from corresponding likelihood ratio tests for the effect of treatment is 0.0733 Table 8c: Time to development of new PU Category ≥3 by 30 day final follow-upCovariateLevel of covariateIncidence rateReference levelHR point EstimateHR 95% Wald Confidence limitsWald P-valueTreatmentHSF18/1013 (1.8%)----0.5498*APM14/1016 (1.4%)vs HSF0.810.40 to1.62Skin statusNo PU22/1648 (1.3%)----0.0288PU Category 13/236 (1.3%)vs No PU0.850.24 to2.98PU Category 27/145 (4.8%) vs No PU3.201.33 to7.71Consent typeWritten16/1404 (1.1%)----0.0335Witnessed verbal6/303 (2.0%)vs Written1.680.66 to4.28Consultee agreement10/322 (3.1%) vs Written2.971.31 to6.74SettingSecondary care hospital26/1416 (1.8%)----0.3045Community hospital3/379 (0.8%) vs Secondary care hospital0.430.13 to1.41NHS intermediate care/ rehabilitation facility3/234 (1.3%) vs Secondary care hospital0.610.18 to2.10Pain on a healthy, altered or PU Category 1 skin siteNo11/890 (1.2%)----<0.0001Yes19/1084 (1.8%)vs No 2.000.93 to4.32Unable to assess0/30 (0.0%)vs No0.000.00 to 0.00Missing2/25 (8.0%)vs No5.901.19 to 29.32Presence of condition affecting peripheral circulationNo22/1567 (1.4%)----<0.0001Yes10/455 (2.2%)vs No1.490.70 to3.15Missing0/7 (0.0%)vs No0.000.00 to 0.00*P-values obtained from corresponding likelihood ratio tests for the effect of treatment 0.5530 Table 8d: Analysis model for time to healing of pre-existing PUsCovariateLevel of covariateHealing rateReference levelHR point EstimateHR 95% Wald Confidence limitsWald P-valueTreatmentHSF45/75 (60.0%)----0.5990*APM44/70 (62.9%)vs HSF1.120.74 to 1.68Consent typeWritten63/102 (61.8%)----0.9193Witnessed verbal14/23 (60.9%)vs Written1.080.65 to1.81Consultee agreement12/20 (60.0%)vs Written1.120.57 to2.19SettingSecondary care hospital71/111 (64.0%)----0.3093Community hospital8/20 (40.0%)vs Secondary care hospital0.550.26 to1.18NHS intermediate care/ rehabilitation facility10/14 (71.4%)vs Secondary care hospital0.910.44 to1.86Presence of condition affecting peripheral circulationNo20/38 (52.6%)----0.0469Yes68/106 (64.2%)vs No0.590.36 to0.97Missing1/1 (100.0%)vs No0.560.31 to1.04*The p-value from the corresponding likelihood ratio test was equal to 0.6122eDocument 9 Mattress allocation, compliance and changesAPMHSFOverallAllocated mattress received on day 0Yes491(48.3%)660(65.2%)1151(56.7%)No523(51.5%)349(34.5%)872(43.0%)Mattress log not returned2(0.2%)4(0.4%)6(0.3%)Total1016(100%)1013(100%)2029(100%)If no, reasons why notLogistical reasons e.g. mattress unavailable or awaiting delivery499(95.1%)301(86.2%)800(91.7%)Clinical decision e.g. participants clinical condition11(2.1%)32(9.2%)43(4.9%)Patient request8(1.5%)13(3.7%)21(2.4%)Other reason/reason unknown/missing5(1.0%)3(0.9%)8(0.9%)Total523(100%)349(100%)872(100%)If no, mattress the patient onAPM or other ‘high tech’ mattress39(7.5%)336(96.3%)375(43.0%)HSF or other ‘low tech’ mattress481(92.0%)10(2.9%)491(56.3%)Other1(0.2%)2(0.6%)3(0.3%)Missing2(0.4%)1(0.3%)3(0.3%)Total523(100%)349(100%)872(100%)Allocated mattress received within two days of randomisationYes828(81.5%)826(81.5%)1654(81.5%)No186(18.3%)183(18.1%)369(18.2%)Missing2(0.2%)4(0.4%)6(0.3%)Total1016(100%)1013(100%)2029(100%)Mattress compliance (%) during treatment phaseMean (S.D.)72.8(35.81)72.8(37.81)72.8(36.8)Median (range)92(0, 100)100( 0, 100)95(0, 100)IQR(50.0, 100)(47.1, 100)(50.0, 100)Missing246Frequency distribution0.0%94(9.3%)110(10.9%)204(10.1%)0.0% to <20.0%74(7.3%)78(7.7%)152(7.5%)20.0% to <40.0%51(5.0%)50(4.9%)101(5.0%)40.0% to <60.0%59(5.8%)51(5.0%)110(5.4%)60.0% to <80.0%80(7.9%)57(5.6%)137(6.8%)80.0% to 100.0%656(64.6%)663(65.4%)1319(65.0%)Missing2(0.2%)4(0.4%)6(0.3%)Total1016(100%)1013(100%)2029(100%)Changed from randomised mattress at least onceYes222(24.1%)220(24.4%)442(24.2%)No698(75.7%)679(75.2%)1377(75.5%)Mattress log not returned2(0.2%)4(0.4%)6(0.3%)Total922(100%)903(100%)1825(100%) Reason for first change from randomised mattress Participant requested mattress change - to aid movement20(9.0%)0(0.0%)20(4.5%)Participant requested mattress change - mattress not comfortable90(40.5%)28(12.7%)118(26.7%)Participant requested mattress change - participant no longer at risk1(0.5%)0(0.0%)1(0.2%)Ward led mattress change - participant no longer at risk4(1.8%)1(0.5%)5(1.1%)Ward led mattress change - to aid rehabilitation29(13.1%)5(2.3%)34(7.7%)Ward led mattress change - participant comfort5(2.3%)17(7.7%)22(5.0%)Ward led mattress change - participant clinical condition3(1.4%)130(59.1%)133(30.1%)Ward led mattress change - participant safety/health4(1.8%)2(0.9%)6(1.4%)Ward led mattress change - reason unknown0(0.0%)2(0.9%)2(0.5%)Ward led mattress change - in error1(0.5%)0(0.0%)1(0.2%)Ward Transfer40(18.0%)20(9.1%)60(13.6%)Technical fault11(5.0%)0(0.0%)11(2.5%)Mattress is required by another patient3(1.4%)0(0.0%)3(0.7%)Home leave2(0.9%)2(0.9%)4(0.9%)Slept in chair1(0.5%)1(0.5%)2(0.5%)Hospital Transfer0(0.0%)2(0.9%)2(0.5%)Reason unknown8(3.6%)10(4.6%)18(4.0%)Total222(100.0%)220(100.0%)442(100.0%)eDocument 10: Mediator Analysis - Repositioning during treatment phase by mattress allocationeDocument 11: Health economics analysisAimThe aim of the within-trial economic evaluation was to assess the incremental cost-effectiveness of APM compared to HSF in the prevention of pressure ulcers in high-risk patients at 30 days post-treatment phase (maximum 90 days post randomisation), from the perspective of the National Health Service (NHS) and Personal Social Services (PSS). MethodsIn line with the main trial methods, the within-trial economic analysis adopted an intention-to-treat (ITT) perspective. The analysis used quality-adjusted life years (QALYs) as the main outcome and adopted the perspective of the UK NHS and PSS. Utility values were derived from the EQ-5D-5L.12 The EQ-5D-5L responses were converted to health-state utility values using the UK tariff.13,14 In addition to the EQ-5D-5L sensitivity analyses were undertaken using QALYs estimated from the PUQoL-UI15,16 (a PU condition specific utility measure) using the appropriate algorithm 15151515151515151413.Quality of lifeQuality of Life questionnaires were administered by a trained clinical research practitioner at baseline and at weeks 1, 3 and at 30 days after the end of the treatment phase. The values obtained from each of two instruments (EQ-5D-5L and PU-QoL-UI) were multiplied by duration (t) in each health state to generate quality-adjusted life years (QALYs). An area under the curve (AUC) approach was adopted for estimating QALYS. Where an individual died during the trial we have assumed that their utility value is 0 from the date of death to trial end. We assume a linear transition to this value from their last completed outcome measure questionnaire.Resources useResource use was obtained using a Health Care Resource Utilisation questionnaire comprising abstracted data from healthcare records and researcher administered questionnaire. These include the length of stay in-hospital, use of hospital outpatient facilities, contact with community-based health care services and utilization of supported living such as care and nursing homes. Utilisation of health and social care was combined with appropriate unit cost information obtained from national sources such as the PSS Research Unit Costs of Health and Social Care.14 All costs were adjusted by the health care price index to February 2017. Unit costs and their sources are shown in tables 1-3. Neither costs nor QALYs were discounted given the time period was 90 days post randomisation.Missing data Multiple imputation was used to provide data for all 2029 patients as only 267 patients had fully completed the EQ-5D-5L at all four-time points. The multiple imputation assumed missing at random and include pressure ulcer status, setting, peripheral circulation and presence of pain, BMI and gender as control variables. Cost and QALYs were also adjusted for baseline imbalances. A multiple regression analysis was used for this propose. Utility at baseline (for outcomes only), pressure ulcer status, setting, peripheral circulation and presence of pain were used as control variables. Sensitivity analysisAlternate scenarios were explored, to test the robustness of the main trial analysis results. The effect of not imputing missing data was considered with an analysis that includes only complete cases. Additionally, in the event of an imbalance at baseline, the effect on cost-effectiveness was evaluated.The level of sampling uncertainty around the ICER was determined via a probabilistic sensitivity analysis (PSA), using a non-parametric bootstrap. The bootstrapped estimates were plotted on the cost-effectiveness plane to illustrate the uncertainty surrounding the cost-effectiveness estimates and the Cost-Effectiveness Acceptability Curve (CEAC) to show the probability the probability of APM or HSF being cost-effective as a function of the willingness to pay threshold (λ). Mean net benefit is reported. Table 1: Inpatient care unit costs of the resource usedResourceUnit of measureUnit costComment/SourceInpatient careMedicineBed day?271Weighted average per ward of elective inpatients excess bed day NHS reference cost 2015-2016: Organisation level source data.(DoH, 2016)Surgery?296Orthopaedics?300Oncology?397Rehabilitation unit?241Cardiology?318Neurology?347High dependency unit?1,026Assumed critical care requiring support for 0 or 1 organNHS reference cost 2015-2016.(DoH, 2016)Intensive care unit?1,595Assumed critical care requiring support for 2 or more organsNHS reference cost 2015-2016. (DoH, 2016)Medical teamConsultant: MedicinePer consultation?97Time per consultation assumed 32.9 minutes based on the duration of a surgery consultation reported in Curtis (2013) Consultant: Psychiatrist?99Consultant: Surgeon?100Registrar?51Physiotherapist?21Annual salary is the same for hospital or community based medical professionalsTime per consultation assumed 32.9 minutes based on the duration of a surgery consultation reported in Curtis (2013) Dietitian?20Occupational therapy?21Physiologist (counsellor)?27Speech therapist?20Social worker?79Assume one hour of client-related work (including qualification) reported in Curtis & Burns (2016) TestsCT ScanPer test?108NHS reference cost 2015-2016 (DoH, 2016)MRI?148NHS reference cost 2015-2016 (DoH, 2016)X-rays?252016/17 National prices and national tariff workbook ( Monitor, 2016)Echocardiogram?67NHS reference cost 2015-2016 (DoH, 2016)Ultrasound?53NHS reference cost 2015-2016 (DoH 2016)Bloods?3NHS reference cost 2015-2016 (DoH 2016)Mattresses typeHSFPer day?0.08Cost per-mattresses ?169 with an assumed durability of six years.Source: NICE 2014 APM?2.05Cost per-mattresses ?3,742 with an assumed durability of five yearsSource: NICE 2014Table 2: Care homes and rehabilitation unit costsResourceUnit of measureUnit costComment/SourceCare homesResidential care homePer week?638.05Not-for-profit residential care homes for 65 and older age group. (Curtis & Burns, 2016)Nursing home?888.22Not for- profit nursing home for 65 and older age group. (Curtis & Burns, 2016)Rehabilitation unitHospital assessmentPer stay?331Cost per patient stay was estimated by adding the per-stay costs plus the per day costs times the number of days the patient stays in the rehabilitation unit (Curtis & Burns 2016)Discharge?536Health services provided by CART coordinator?428Cost of typical episode?2,957Wages; overheads; capital costsPer day?101Table 3: Community care unit costsResourceUnit of measureUnit costComment/SourceOutpatient careGPSurgeryPer consultation?369.22 minutes per consultation (Curtis & Burns 2016)?3.9 per minute (Curtis & Burns 2016)Clinic?6717.2 minutes per consultation (Curtis & Burns 2015)?3.9 per minute (Curtis & Burns 2016)Telephone?287.1 minutes per consultation (Curtis & Burns 2015)?3.9 per minute (Curtis & Burns 2016)Home visit?9223.4 minutes per consultation including 12 minutes travel time (Curtis & Burns 2015)?3.9 per minute (Curtis & Burns 2016)NurseSurgeryPer consultation?14?56 per hr assuming a face to face contact ratio 1:0.30 (Curtis & Burns 2015)15 minutes per consultation (Curtis 2014)Clinic?14?56 per hr assuming a face to face contact ratio 1:0.30 (from Curtis & Burns 2015)15 minutes per consultation (Curtis 2014)Telephone?6?56 per hr assuming a face to face contact ratio 1:0.30 (from Curtis & Burns 2015)6 minutes per consultationHome visit?23?56 per hr assuming a face to face contact ratio 1:0.30 (Curtis & Burns 2015)25 minutes per consultation including travel time (Curtis 2014)Physiotherapist*SurgeryPer consultation?2132.9 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Clinic?1523.3 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Telephone?813.1 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Home visit?28Assumed the same duration as surgery consultation plus 12 minutes of travel time (as outpatient GP consultation)?38 per hour with qualifications (Curtis & Burns 2016)Counsellor*SurgeryPer consultation?2732.9 minutes per consultation (Curtis 2013)?49 per hour with qualifications (Curtis & Burns 2016)Clinic?1923.3 minutes per consultation (Curtis 2013)?49 per hour with qualifications (Curtis & Burns 2016)Telephone?1113.1 minutes per consultation (Curtis 2013)?49 per hour with qualifications (Curtis & Burns 2016)Home visit?37Assumed the same duration as surgery consultation plus 12 minutes of travel time (as outpatient GP consultation)?49 per hour with qualifications (Curtis & Burns 2016)Occupational therapy*SurgeryPer consultation?2132.9 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Clinic?1523.3 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Telephone?813.1 minutes per consultation (Curtis 2013)?38 per hour with qualifications (Curtis & Burns 2016)Home visit?28Assumed the same duration as surgery consultation plus 12 minutes of travel time (as outpatient GP consultation)?38 per hour with qualifications (Curtis & Burns 2016)Outpatient consultationsSurgeryPer consultation?131NHS reference cost 2015-2016 (DoH 2016)Orthopaedics?118NHS reference cost 2015-2016 (DoH 2016)Oncology?152NHS reference cost 2015-2016 (DoH 2016)Cardiology?128NHS reference cost 2015-2016 (DoH 2016)Respiratory?156NHS reference cost 2015-2016 (DoH 2016)Urology?106NHS reference cost 2015-2016 (DoH 2016) Neurology?177NHS reference cost 2015-2016 (DoH 2016)Ophthalmology?91NHS reference cost 2015-2016 (DoH 2016)Rehabilitation?126NHS reference cost 2015-2016 (DoH 2016)Medicine ?168NHS reference cost 2015-2016 (DoH 2016)A&E?138NHS reference cost 2015-2016 (DoH 2016)Radiology ?85NHS reference cost 2015-2016 (DoH 2016)Rheumatology?144NHS reference cost 2015-2016 (DoH 2016)Social CareMeal on wheelsHotPer meal?8.6Community meals LeedsFrozen?4.0Community meals LeedsCare workerPer patient visit?13.5Assumes a 30-minute visit (Sixty-three per cent of local authority commissioned home care visits lasted 16-30 minutes (PSSRU, 2016) – we take the upper value) at cost of ?27 per face to face visit at weekend (Curtis & Burns, 2016). We assume that this is paid by the social care sectorSocial WorkerPer hr?79Assume one hour of client-related work (including qualification) reported in Curtis & Burns (2016)* Annual salary is the same for hospital or community based medical professionalsResultsCosts and Utility values estimates for both interventions are shown in Table 4, 5, 6 and 7. Only a few costs category estimates were statistically significant between the two treatment arms: intensive care unit treatment; social worker, overall tests costs, GP telephone consultations, overall community nurse consultations, clinic and overall community physiotherapist consultation; orthopaedic, cardiology and total outpatient specialist consultation; X-rays. The APM mattress was more expensive than the HSF (?28.80 vs ?1.05) and the difference was statistically significant (p<0.001). However, the main driver was the total in-patient care costs in which the mean costs for the HSF was higher (?2,888 vs ?2,810). In terms of Utility, both the EQ-5D-5L and PUQoL-UI show no statistically significant differences between the arms of the trial across each of the four-time periods (baseline, week 1, week 3 and 30 days post-treatment phase). Table 4. Average per patient hospital and residential care costs Cost category APMHSFP value (Wilcoxon)MeanMedianSDMeanMedianSDIn-patient careMedicine834.690.001806.96814.640.001535.370.51Surgery94.40.00598.16117.450.00803.500.91Orthopaedics211.550.00903.10226.590.00855.320.50Oncology31.780.00367.0111.990.00217.260.16Rehabilitation unit1,352.480.003421.901,294.560.003222.630.89Cardiology93.110.001111.8991.270.001067.900.84Neurology118.830.001462.4755.210.00939.430.30High Dependency Unit42.620.00743.1558.890.001633.240.51Intensive care unit113.120.001460.78276.270.003280.450.07*Total in-patient care costs2,810.081897.004513.132,888.681897.005372.530.54Medical teamConsultant44.060.0075.0848.060.0073.690.22Registrar56.690.00190.2771.800.00390.670.69Physiotherapist104.6962.64123.90108.3183.52131.260.31Dietitian9.000.0025.529.4240.0030.800.81Occupational therapist40.050.0080.2939.37520.8779.680.60Physiologist (counsellor)5.470.0040.515.0490.0030.800.89Speech therapist11.530.0049.9913.610.0046.670.12Social worker17.020.0058.1611.330.0040.450.038**Total medical team costs288.55176.00346.28306.97180.52548.590.71TestsCT Scan11.8850.0039.2612.880.0041.900.68MRI6.1910.0036.796.7020.0031.840.37X-rays7.7130.0023.0119.8560.0026.270.10Echocardiogram0.100.002.663.102190.002.620.98Ultra sound0.740.006.22.7200.006.1430.95Bloods5.9003.139.5567.143.1315.8290.19Total tests costs32.530.0070.5937.410.0079.320.07*Mattress Mattress type28.814.3426.31.050.540.930.00Residential careCare Home177.070.00936.60140.20.00603.080.19Nursing Home285.020.00719.53140.250.00967.1510.63Rehabilitation Unit 324.620.001503.61432.660.001722.340.13Total residential care cost786.720.001820.84837.100.001959.860.97 *Significant at 10%, **significant at 5%Table 5 Average per patient community cost Cost category AMPHSFP value (Wilcoxon)MeanMedianSDMeanMedianSDGPSurgery6.5940.0020.745.2980.0018.6780.42Clinic2.0970.0021.000.4170.005.3050.38Telephone8.7130.0028.1286.4940.0021.020.018**Home visit65.620.00109.1564.780.0093.780.44Total GP costs4.7790.0030.372.5750.0015.3020.665NurseSurgery.75850.004.0931.4520.0011.870.87Clinic.02910.00.6410.0870.001.9210.99Telephone.0700.00.72370.03480.00.4420.47Home visit57.190.00155.4783.260.00209.50.089Total nurse costs58.050.00155.3284.830.00209.310.065*PhysiotherapistSurgery2.0390.0028.122.8950.0030.680.40Clinic1.790.0013.970.6170.006.910.01**Telephone.34710.003.290.3780.003.910.81Home visit39.910.00107.2958.300.00146.670.03Total physiotherapist costs44.100.00110.7262.190.00148.620.060*CounsellorSurgery0.000.000.000.000.000.00-Clinic0.000.000.000.000.000.00-Telephone0.000.000.00.0900.001.990.31Home visit1.1460.0018.93.2280.003.7490.41Total counsellor costs1.14660.0018.93.3180.004.9710.411Occupational therapistSurgery0.21690.003.1611.9440.0026.500.3Clinic0.65080.009.360.0300.00.6800.10Telephone0.1650.001.775.2630.003.390.60Home visit14.6940.0046.7120.680.0087.560.38Total occupational therapy cost15.720.0047.5422.90.0091.090.306Home careFrozen meals2.110.0017.033.280.0018.690.23Hot meals4.990.0043.533.340.0028.570.64Laundry0.610.005.580.6980.005.510.64Care worker190.220.00640.9147.210.00367.640.38Social worker7.5080.0031.2613.490.00106.460.53Total Home Care205.450.00645.92168.030.00168.030.56*Significant at 10%, **significant at 5%Table 6 Outpatient servicesCost category AMPHSFP value (Wilcoxon)MeanMedianSDMeanMedianSDOutpatient specialists Surgery6.500.0055.626.210.0045.330.84Orthopaedic11.210.0042.1621.370.0071.920.003**Oncology3.460.0038.464.700.0045.140.41Cardiology3.450.0023.917.420.0038.960.07*Respiratory2.250.0021.191.600.0015.770.75Urology4.820.0050.865.0210.0060.160.67Medicine 11.490.0057.7615.250.00115.540.76Neurology1.830.0017.920.360.008.0350.10Rehabilitation1.300.0012.781.300.0015.10.73Ophthalmology2.070.0015.972.250.0015.370.66Rheumatology.290.006.540.590.009.220.56A&E1.720.0017.792.000.0020.820.99Radiology0.880.0010.220.350.007.730.17Total Outpatient specialists51.330.00123.4168.450.00167.570.05**TestsCT Scan 0.000.000.000.440.006.920.15MRI0.610.009.500.000.000.000.15X-rays0.100.001.600.410.003.560.09*Ultra-sound0.550.006.350.330.005.370.41Bloods0.000.000.000.0130.000.200.15Total outpatient tests1.260.0011.501.190.009.400.28*Significant at 10%, **significant at 5%Table 7. Utility values (EQ-5D, PUQOL-UI)Time pointMean (SD, median)APMHSFP value (Base case; Wilcoxon)EQ-5D-5LBaseline0.34(0.23, 0.33)0.34(0.22, 0.32)0.34(0.23,0.33)0.87Week 10.41(0.22, 0.40)0.41(0.21,0.40)0.41(0.22,0.40)0.50Week 30.47(0.19, 0.47)0.47(0.18,0.47)0.46(0.20,0.47)0.6530 days after end of treatment phase0.52(0.22, 0.53)0.52(0.21,0.54)0.52(0.22,0.53)0.49PUQoL-UIBaseline0.60(0.16, 0.60)0.60(0.16,0.60)0.60(0.16,0.60)0.96Week 10.65(0.14, 0.66)0.65(0.14,0.66)0.65(0.14,0.65)0.52Week 30.69(0.12, 0.71)0.69(0.11,0.71)0.69(0.12,0.70)0.6530 days after end of treatment phase0.68(0.13, 0.69)0.69(0.13,0.70)0.69(0.13,0.69)0.28Once baseline adjustments have been made the deterministic analysis shows the mean total costs of APM and HSF are ?4533 and ?4646 respectively with mean QALYs of 0.128 and 0.127 respectively. Thus, APMs dominate HSF as APM has lower costs and higher QALY values. Similar results are seen in the probabilistic analysis where APMs dominate HSF (mean total costs of APM and HSF are ?4533 and ?4646 and mean QALYs of 0.128 and 0.127 respectively) (Table 8). Within this analysis, APM has a 99% probability of being cost-effective at a threshold of ?20,000. Figures 1 and 2 show the cost-effectiveness plane and CEAC. Table 8 Results of within-trial analysisStrategyTotal CostTotal QALYIncremental CostIncremental QALYICERNB Pr. of cost-effective(?20,000 threshold)ResultDeterministicWithin-trial (adjusted for baseline costs and QALYs)APM?4,4820.128-?1,918-Cost-effectiveHSF?4,6210.127?138-0.0010-?136,171-?2,077-DominatedProbabilistic within trial (adjusted for baseline costs and QALYs)APM?4,5330.128-?1,9790.99Cost-effectiveHSF?4,6460.127?113-0.0011-?101,699--?2,1140.01DominatedFigure1: Cost effectiveness planeFigure 2: CEACThe sensitivity analyses show mixed results. The complete case analysis showed, unlike the previous analyses, that HSF was cost-effective. An additional analysis was undertaken using only complete cases for the baseline and 30 days after discharge. This analysis increased the number of complete cases to 934 (APM: 460 and HSF: 474 vs 118 for the APM and 149 for the HSF mattress when for complete cases using the four-time point measurements). The results here, when baseline adjustment was made, showed HSF was the most cost-effective; conversely when no baseline adjustment was made APM was the cost-effective strategy. ITT analysis without baseline adjustment found APM was the cost-effective alternative. The results of the sensitivity analyses using QALYs derived from the PUQoL-UI similarly found differing results.ConclusionOverall the results of the within-trial analyses were mixed. The ITT analyses found the APM to be the most cost-effective strategy. However, the results were not robust to the sensitivity analyses. In particular, the complete cases found HSF to be cost-effective. In addition, the difference between the QALY gains of the two arms was very small (~0.001) equating to around half a quality-adjusted life day in both the within trial and lifetime model analyses. Similarly, the difference between the costs accrued in each arm was relatively modest with few cost categories showing a statistically significant difference between the arms. Interestingly, the difference in cost between the two mattresses was statistically significant, with the APM attracting a higher mean cost (?28.80) compared to the HSF (?1.05) (p<0.001); but these costs represented only a small proportion of the total mean costs. In fact, the cost driver for both arms was, as anticipated, in-patient stays and no statistical difference was observed between these costs between arms (mean ?2810.08 for APM vs ?2888.68 for HSF p= 0.54 (Table 4). Small differences in utility values derived from the EQ-5D-3L have often been attributed to a lack of sensitivity in the instrument.17 In an attempt to address this, we used the EQ-5D-5L. The EQ-5D-5L was developed to improve sensitivity.18 However, recent preliminary evidence suggests that use of the EQ-5D-5L rather than the EQ-5D-3L ‘causes a decrease in the incremental QALY gain from effective health technologies and therefore technologies appear less cost-effective’17 Despite the current uncertainty between the choice of EQ-5D measure, a strength of this analyses is the use of data collected using the PUQoL-UI; a preference based measure developed to assess the impact of pressure ulcers on health-related quality of life.15 The sensitivity analyses using this measure produced similar results to the primary analyses giving us confidence in our results and the conclusions drawn.References for all eDocuments1.NPUAP/EPUAP/PPPIA. Prevention and Treatment of Pressure Ulcers: Clinical Practice Guideline. Osborne Park, Western Australia: Cambridge Media; 2014.2.Lan KK, Lachin JM, Bautista O. Over-ruling a group sequential boundary--a stopping rule versus a guideline. Stat Med 2003; 22(21): 3347-55.3.Agency EM. Guideline on missing data in confirmatory clinical trials. 2010.4.White IR, Thompson SG. Adjusting for partially missing baseline measurements in randomized trials. Stat Med 2005; 24(7): 993-1007.5.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94(446): 496-509.6.Royston P, Parmar MK. The use of restricted mean survival time to estimate the treatment effect in randomized clinical trials when the proportional hazards assumption is in doubt. Stat Med 2011; 30(19): 2409-21.7.Royston P, Parmar MK. Restricted mean survival time: an alternative to the hazard ratio for the design and analysis of randomized trials with a time-to-event outcome. BMC Med Res Methodol 2013; 13: 152.8.Gray RJ. A Class of K-Sample Tests for Comparing the Cumulative Incidence of a Competing Risk. Ann Stat 1988; 16(3): 1141-54.9.Baron RM, Kenny DA. The Moderator Mediator Variable Distinction in Social Psychological-Research - Conceptual, Strategic, and Statistical Considerations. J Pers Soc Psychol 1986; 51(6): 1173-82.10.. (accessed 20/12/2016.11.Nixon J, Cranny G, Iglesias C, et al. Randomised, controlled trial of alternating pressure mattresses compared with alternating pressure overlays for the prevention of pressure ulcers: PRESSURE (pressure relieving support surfaces) trial. BMJ (Clinical research ed) 2006; 332(7555): 1413.12.EuroQol. EQ-5D-5L. 2017. (accessed 13/10/17.13.Devlin NJ, Shah KK, Feng Y, Mulhern B, van Hout B. Valuing health-related quality of life: An EQ-5D-5L value set for England. Health economics 2017.14.Curtis L, Burns A. Unit Costs of Health and Social Care 2016. 2016. (accessed 10/10/17.15.Nixon J, Nelson E A, Rutherford C, et al. Pressure UlceR Programme Of reSEarch (PURPOSE): using mixed methods (systematic reviews, prospective cohort, case study, consensus and psychometrics) to identify patient and organisational risk, develop a risk assessment tool and patient-reported outcome Quality of Life and Health Utility measures. Programme Grants for Applied Research 2015; 3.16.Czoski MC, Meads DM, McCabe C, et al. A Utility Algorithm forthe Pressure Ulcer Quality of Life - Utility Instrument (Puqol-Ui). Value in health : the journal of the International Society for Pharmacoeconomics and Outcomes Research 2014; 17(7): A513.17.Wailoo A, Hernandez Alava M, Grimm S, et al. Comparing the eq-5d-3l and 5l versions. What are the implications for cost effectiveness estimates? . 2017. (accessed 03/11/17.18.Herdman M, Gudex C, Lloyd A, et al. Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L). Quality of life research : an international journal of quality of life aspects of treatment, care and rehabilitation 2011; 20(10): 1727-36. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download