Introduction
A CEA was embedded in the POPPI cluster RCT, aiming to evaluate the relative cost-effectiveness of the POPPI intervention versus usual care. The specific objectives were to:
- compare incremental cost-effectiveness at 6 months of the POPPI intervention versus usual care (primary objective)
- estimate the lifetime incremental cost-effectiveness between the treatment groups (secondary objective).
Methods
Overview
A full CEA was undertaken to assess the relative cost-effectiveness of the POPPI intervention versus usual care. The CEA was undertaken in two phases; first, resource use and outcome data that were collected as part of the cluster RCT database linked to the CMP and NHS Digital databases were used to report cost-effectiveness at 6 months. Second, the CEA used the 6-month data to project the lifetime cost-effectiveness of the POPPI intervention versus usual care. The cost analysis takes an NHS and Personal Social Services perspective.93 The economic evaluation is reported in line with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement.94
The CEA used information on HRQoL at 6 months, combined with information on survival status, to report QALYs. Each QALY was valued using the NICE-recommended threshold of willingness to pay for a QALY gain (£20,000), in conjunction with the costs of each strategy, to report the incremental net monetary benefits (INMBs) of the POPPI intervention versus usual care, overall and for the same prespecified subgroups, as for the evaluation of clinical effectiveness. The main assumptions of the CEA were subjected to extensive sensitivity analyses.
Resource use, unit costs, outcomes and cost-effectiveness up to 6 months
Resource use measurement
The resource use categories considered were chosen a priori; according to those for which the differences between the treatment groups were judged as being possible and likely to drive incremental costs, these were resource use associated with the POPPI intervention, hospital admissions (index admission and re-admissions), and visits to outpatients and community health-care services.
Resource use associated with the POPPI intervention
The costs of the POPPI intervention were estimated based on experience of delivering the intervention in a typical critical care unit and considering the rolling out of the POPPI intervention into routine NHS practice. Costing was based on the following guiding principle: key elements of the intervention that were deemed important to outcomes and would be provided in routine practice were considered, and elements that would incur costs in routine practice but were provided free of charge in the cluster RCT were costed and included. The cost items associated with the POPPI intervention were grouped into three elements (Table 28). The base-case intervention costs were estimated using resource use data recorded on the cluster RCT electronic case report form, and informed by expert clinical opinion and the process evaluation, reflecting the most plausible assumption for routine practice in the majority of critical care units (see Table 28), with alternative levels considered in the sensitivity analyses (see Table 30).
| Elements of the intervention | Content/delivery of element | Action | Level of resource use |
|---|---|---|---|
| Element one: creating a therapeutic environment in critical care | Delivery of online training | Take course (including test), co-ordinate delivery in unit | Absorbed in the NHS mandatory training cost |
| Creating a therapeutic environment | Seminars/teaching, core groups meetings/activities, individual activities/actions | ||
| Element two: three stress support sessions for patients identified as acutely stressed | IPAT assessment | Screening patient with questionnaire | 10 minutes of bedside nurse’s (band 5) time |
| Delivery of stress support sessions | Prepare for stress support sessions Carry out stress support sessions Write up stress support sessions | 1.5 hours of the POPPI nurse’s (band 7) time | |
| Three-day training course | Attending the training course Delivering the training course Other costs | Per-patient costs calculated from the actual training costs incurred | |
| Debriefing and support | Trainee/clinical supervisor’s meeting time | 0.5 hours per month of three POPPI nurses’ and a trainer’s (band 8) time, per site | |
| Element three: relaxation and recovery programme for patients identified as acutely stressed | Relaxation and recovery programme | Delivering the programme via tablet computer app, patient booklet and DVD | Absorbed in the NHS routine cost |
Element one (creating a therapeutic environment in critical care) required all clinical critical care unit staff to complete an online training course (the POPPI online training). The online training would be expected to be included in the NHS mandatory training, and completed by staff members on an annual basis, if the POPPI intervention were to be rolled out into routine NHS practice. Creating a therapeutic environment in critical care also required staff to deliver and attend seminars and other educational activities, and to discuss creation and promotion of the therapeutic environment at team meetings. It was considered that this element of the intervention would be built in to the regular training and educational activities of critical care units if the POPPI intervention were to be rolled out into routine NHS practice. Therefore, no additional staff time and costs were considered for this element.
Element two (three stress support sessions for patients identified as acutely stressed) included a staff member assessing a patient for acute stress (using the IPAT), delivery of stress support sessions for patients identified as acutely stressed (i.e. those with an IPAT score of ≥ 7 points), costs of the 3-day training course for the POPPI nurses, and continued debriefing and support for the POPPI nurses following the 3-day course.
All patients recruited from intervention group sites during the intervention period of the POPPI cluster RCT were assessed for acute stress using the IPAT. Informed by expert opinion and data from the process evaluation, base-case costing for this activity considered that an IPAT assessment session required, on average, 10 minutes of a bedside nurse’s (band 5) time and was costed accordingly. Delivery of stress support sessions to acutely stressed patients involved staff time for preparation (reviewing patient records/medical history and IPAT responses), carrying out of the three stress support sessions and the writing up of key actions/outcomes of the session(s). Expert opinion and data from the process evaluation suggested that the delivery of one stress support session required 1.5 hours of a POPPI nurse’s (band 7) time. Per-patient costs were calculated as per the data recorded on the POPPI cluster RCT electronic case report form on the frequency of stress support sessions (0–3 sessions) received by acutely stressed patients (i.e. those with an IPAT score of ≥ 7 points) in the cluster RCT. At the beginning of the transition period, all POPPI nurses at the intervention group sites attended the 3-day training course, organised centrally. From the actual costs incurred in the cluster RCT, costs per site were calculated, from which per-patient costs were calculated considering the mean annual number of critical care admissions who would probably meet the POPPI cluster RCT eligibility criteria (based on data from the CMP and the cluster RCT screening and enrolment logs). All of the POPPI nurses were allocated a clinical trainer from the POPPI training team to provide ongoing debriefing and support following the 3-day training course. Debriefing and support were focused on enhancing nurses’ skills and discussing individual patient cases. The level of debriefing and support considered per site was that all POPPI nurses (band 7) meet a trainer (health psychologist band 8) for 0.5 hours per month. The opportunity costs of debriefing and support time were calculated. Per-patient costs were calculated considering the mean annual number of critical care admissions who would probably meet the POPPI cluster RCT eligibility criteria (as above).
Element three (relaxation and recovery programme for patients identified as acutely stressed) required delivery of the programme via already-developed materials: a tablet computer app, a patient booklet and a DVD. These costs are negligible and are considered to be absorbed in routine NHS costs.
Hospital stay: index admission
The duration and location of the index hospital admission, that is, the hospital stay following recruitment, was recorded for each patient for up to 6 months post recruitment on the electronic case report form. The total duration of the index admission included the time spent in critical care and on general medical wards. The length of stay in critical care was calculated as the total duration in days (including fractions of days), from the date and time of admission for the stay in critical care during which the patient was recruited until the time of discharge from critical care or death. Within the index admission, the total duration of critical care stay included all of the time spent in critical care between admission to the critical care unit and discharge from acute hospital and included any transfers to critical care units in other hospitals, as well as those in the hospital where the patient was recruited. For each day in critical care, data on the number of organs supported was recorded in the CMP database. Each critical care episode was then assigned a Healthcare Resource Group (HRG) code, applying standard HRG grouper algorithm.95
For the index admission, the total length of stay was calculated as the total duration in days from the date of recruitment to the date of ultimate hospital discharge, or death.
Re-admissions
A hospital re-admission was defined as a further hospital admission following ultimate hospital discharge from the index admission. The information on re-admissions was collected from two sources. First, data on re-admissions to critical care were accessed from the CMP database.63 From the CMP database, information was accessed on the duration of stay in critical care and the total hospital stay, including subsequent transfer to other care areas (e.g. general medical wards) in the same hospital and to other hospitals. Second, information on re-admissions that did not include a further stay in critical care was collated from responses to the health services questionnaire administered to patients surviving to 6 months.
Hospital outpatient visits and community service use
The resource use items considered included the total number of hospital outpatient visits and community service use following discharge from the index admission, but before 6 months. The resource use items considered resources used both for reasons related to the admission to the critical care unit in which the patient was recruited and reasons unrelated to the initial critical care admission (i.e. other health reasons). The items of community service use included visits to the GP, nurses (i.e. from the GP clinic, hospital or a psychiatric nurse), health visitors, occupational therapists, speech and language therapists, counsellors, physiotherapists, psychiatrists, psychologists and critical care follow-up clinics. The levels of resource use were taken from responses to the health services questionnaire administered to patients surviving to 6 months.
Unit costs
The unit costs required for valuing the resource use data were taken from national unit cost databases,96,97 and are listed in Table 29. The unit costs associated with the additional staff time required to deliver the POPPI intervention were taken from national sources.96 The costs per critical care bed-day by HRG and general medical bed-day were taken from the ‘Payment by Results’ database.97 Unit costs for hospital outpatient visits and community service use were obtained from a recommended published source for Health and Social Care costs.96 All unit costs were reported in 2015–16 GBP prices.
| Item | Unit cost (£) | Source |
|---|---|---|
| Staff time for delivering the POPPI intervention | ||
| Hospital nurse, band 5 (per hour) | 35 | Unit Costs of Health and Social Care 2016 96 |
| Hospital nurse (POPPI nurse), band 7 (per hour) | 53 | Unit Costs of Health and Social Care 2016 96 |
| Health psychologist, band 8 (per hour) | 60 | Unit Costs of Health and Social Care 2016 96 |
| Hospital costs (bed-day) | ||
| Critical care bed-day (number of organs supported) | ||
| 0 | 759 | NHS Reference Costs 2015–2016 97 |
| 1 | 1031 | NHS Reference Costs 2015–2016 97 |
| 2 | 1399 | NHS Reference Costs 2015–2016 97 |
| 3 | 1619 | NHS Reference Costs 2015–2016 97 |
| 4 | 1794 | NHS Reference Costs 2015–2016 97 |
| 5 | 1977 | NHS Reference Costs 2015–2016 97 |
| ≥ 6 | 2274 | NHS Reference Costs 2015–2016 97 |
| General Medical bed-day | 298 | NHS Reference Costs 2015–2016 97 |
| Outpatient and community health services | ||
| Hospital outpatient | 135 | Unit Costs of Health and Social Care 2016 96 |
| GP practice visit (per visit) | 36 | Unit Costs of Health and Social Care 2016 96 |
| GP home visit (per visit) | 118 | Unit Costs of Health and Social Care 2016 96 |
| GP nurse visita | 11 | Unit Costs of Health and Social Care 2016 96 |
| GP nurse home visita | 18 | Unit Costs of Health and Social Care 2016 96 |
| Hospital nursea | 9 | Unit Costs of Health and Social Care 2016 96 |
| Health visitora | 8 | Unit Costs of Health and Social Care 2016 96 |
| Health visitor home visita | 18 | Unit Costs of Health and Social Care 2016 96 |
| Occupational therapista | 8 | Unit Costs of Health and Social Care 2016 96 |
| Physiotherapista | 8 | Unit Costs of Health and Social Care 2016 96 |
| Psychiatrista | 16 | Unit Costs of Health and Social Care 2016 96 |
| Psychologista | 13 | Unit Costs of Health and Social Care 2016 96 |
| Counsellora | 8 | Unit Costs of Health and Social Care 2016 96 |
| Speech and language therapista | 8 | Unit Costs of Health and Social Care 2016 96 |
Mortality and health-related quality of life
Information on the date and time of deaths was used to calculate the survival time up to 6 months for each recruited patient. HRQoL at baseline (i.e. at the time of consent) was self-completed by patients using a visual analogue scale (a health thermometer with scores ranging from 0 to 100). HRQoL at 6 months was measured using the EuroQol-5 Dimensions (EQ-5D) (a generic measure), which requires patients to describe their health on five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The EQ-5D with five levels (EQ-5D-5L)56 was chosen, in which patients are required to state whether they have ‘no problems’, ‘slight problems’, ‘moderate problems’, ‘severe problems’ or ‘extreme problems’ for each dimension. Permission for reuse of the questionnaire was obtained from EuroQol [https://euroqol.org (accessed 1 July 2019)] and the EQ-5D-5L questionnaire was administered to survivors as part of the 6-month follow-up questionnaire (see Chapter 5). The responses to the EQ-5D-5L questionnaire were used to report each patient’s described health, which was then valued according to health state preferences from the general English population to calculate EQ-5D utility scores,98 anchored on a scale from 0 (death) to 1 (perfect health).99 QALYs at 6 months were calculated by valuing each patient’s survival time by their HRQoL at baseline and at 6 months according to the ‘area under the curve’ approach.100 For 6-month survivors, QALYs were calculated using the HRQoL at baseline and at 6 months, applying linear interpolation. For decedents between recruitment and 6 months, a linear interpolation was applied between the baseline HRQoL and the date of death, when a zero HRQoL was applied.
Cost-effectiveness
The CEA followed the ITT principle, and reported incremental costs, QALYs and cost-effectiveness up to 6 months, according to treatment group. Missing data for resource use and HRQoL were handled with multivariate imputation by chained equation,101 assuming that the data were MAR, conditional on baseline covariates, resource use and observed end points.70 Patients who did not return or fully complete the EQ-5D-5L questionnaire administered at 6 months had their EQ-5D-5L scores imputed from those survivors who did fully complete the questionnaire. Similarly, for those eligible patients who did not return the health services questionnaire, information on the use of outpatient services up to 6 months was imputed from those patients who completed the health services questionnaire. The resultant estimates were combined with Rubin’s rules, which recognise uncertainty both within and between imputations.102
Total costs up to 6 months were calculated by combining the resource use with unit costs. The CEA used a GLMM at the individual patient level (patients nested within sites and within treatment group/time period), with random effects at site level. The analysis adjusted for prespecified baseline covariates at both patient and site level (see Chapter 4 for further details), with the primary effect estimate being the interaction (difference in difference) between treatment groups and time period.70 We reported adjusted mean differences between the treatment groups in 6-month HRQoL, QALYs and costs, together with 95% CIs. The differences in average costs and QALYs between the treatment groups were used to calculate the INMB of the POPPI intervention versus usual care. We valued the incremental QALY according to the NICE-recommended willingness-to-pay threshold for a QALY gain (£20,000) and subtracted from this the incremental cost. INMBs were reported overall, and for the same prespecified subgroups as the clinical evaluation (see subgroup analyses of the primary outcome for the clinical evaluation in Chapter 7).
The uncertainty around the differences in average costs and QALYs between the treatment groups was illustrated on the cost-effectiveness plane.103 We estimated the incremental costs and QALYs with a multilevel regression model (described above). To express the uncertainty in the estimation of the incremental costs and QALYs, we used the estimates of the means, and variances from the regression model, to generate 500 estimates of incremental costs and QALYs from the joint distribution of these end points, assuming asymptotic normality. We then plotted these incremental costs and QALYs on the cost-effectiveness plane. We also reported cost-effectiveness acceptability curves, by calculating the probability that, compared with usual care, the POPPI intervention is cost-effective given the data, at alternative levels of willingness to pay for a QALY gain.
Base-case assumptions and subsequent sensitivity analyses
The main assumptions made in the base-case scenario and how each was relaxed in sensitivity analyses are detailed below and summarised in Table 30.
| Assumption | Base case | Sensitivity analysis |
|---|---|---|
| Nurse’s time for IPAT assessmenta | 10 minutes per patient | 5 minutes per patient |
| Nurse’s time for IPAT assessment | 10 minutes per patient | 20 minutes per patient |
| Nurse’s time for delivering stress support sessionb | Each stress support session requires 1.5 hours | Each stress support session requires 1 hour |
| Nurse’s time for delivering stress support session | Each stress support session requires 1.5 hours | Each stress support session requires 2 hours |
| Re-admissions from health services questionnairesc | Included in the analysis | Excluded from the analysis |
| HRQoL at time of consentd | HRQoL measured at time of consent was applied | Zero HRQoL at time of consent was applied |
| Distributional assumptionse | Costs and QALYs normally distributed | Costs and QALYs gamma distributed |
| Unit-level standardised mortality ratiosf | Not included in the analysis | Adjusted for in the analysis |
The results of the sensitivity analysis are reported as mean INMBs with corresponding 95% CIs.
Subgroup analyses
We undertook prespecified subgroup analyses (as per analysis of clinical effectiveness) to report INMBs according to age (quartiles), sex (males vs. females), socioeconomic status (quintile of the IMD76), duration of delirium (patients with no CAM-ICU defined delirium versus patients with CAM-ICU defined delirium less than the overall median duration of delirium versus patients with CAM-ICU defined delirium greater than or equal to the overall median delirium duration), STAI-6 score (quartiles), surgical status (emergency/urgent surgery vs. elective/scheduled surgery vs. non-surgical), overall intervention implementation score for intervention group sites (tertiles) and predicted PSS-SR score (quintiles). The base-case analysis was repeated, adjusting for adherence using a structural mean model105 with an instrumental variable of randomised allocated treatment to estimate the efficacy (adherence-adjusted causal effect, as per the analysis of the primary clinical outcome, see Chapter 4) of the stress support sessions among those patients consenting to the IPAT assessment and stress support sessions, assessed as being at a high risk of psychological morbidity (i.e. an IPAT score of ≥ 7) and receiving at least two stress support sessions.
Lifetime cost-effectiveness
Lifetime cost-effectiveness was projected by summarising the relative effects of alternative strategies on long-term survival and HRQoL as compared with that of an age- and sex-matched general population.106,107 The survival of the patients who survived up to 6 months was extrapolated over the lifetime. This extrapolation compared the survival and HRQoL of POPPI cluster RCT patients with those of the age- and sex-matched general population. There is evidence to suggest that critical care survivors face a higher probability of death108 and lower quality of life109,110 after the critical care episode than the (age- and sex-matched) general population. However, there is no clear evidence as to the magnitude and duration of excess mortality in general critical care patients. In the POPPI cluster RCT, the survival probability of patients at 1 year was similar to those of the age- and sex-matched general population. There is evidence of a decrement in quality of life for up to 5 years following discharge from critical care.109 In the POPPI cluster RCT, the HRQoL of POPPI survivors at 6 months was ≈87% of that of the age- and sex-matched general population.111 We therefore assumed a HRQoL decrement in the first year, but with improvement over 5 years to match the quality of life of the age- and sex-matched general population. After 5 years, we applied HRQoL values for the age- and sex-matched general population. Lifetime QALYs were reported by combining life-years and HRQoL.
To project lifetime costs attributable to the initial episode of critical illness, we considered the re-admission (to critical care and general wards) costs recorded up to 1 year post recruitment in the POPPI cluster RCT. Mean annual re-admission costs in critical care and general wards were calculated for patients who survived for at least 6 months and were not censored between 6 and 12 months. For the outpatient and community service costs, we applied the 6-month costs to the subsequent 6 months for patients who survived for at least 6 months and were not censored between 6 and 12 months, and calculated mean annual outpatient and community costs. These mean costs were applied annually for up to 5 years, after which time the HRQoL per patients in one POPPI trial was assumed to be at the levels of the age- and sex-matched population. All future costs and life-years were discounted at the recommended rate of 3.5%.93
Results
Cost-effectiveness up to 6 months
Resource use
For the index hospital episode, the mean length of stay in critical care was longer for patients recruited at intervention group sites than for those recruited at control group sites in the baseline period. In the intervention period, the mean length of stay in critical care from the index admission was lower for patients recruited at the intervention group sites than for those in the usual care group (Table 31). Length of stay in general medical wards from the index admission was higher for patients recruited from intervention group sites than for patients in the usual care group, in both baseline and intervention periods. The proportions of patients re-admitted were similar between treatment groups. The average length of stay in critical care from re-admission was lower in the intervention group than in the usual care group in both baseline and intervention periods. The mean length of stay in general medical wards from re-admissions was higher in the intervention group than in the usual care group in both time periods. The average total length of stay prior to 6 months post recruitment was 31.27 days in the intervention group and 27.52 days in the usual care group in the baseline period. In the intervention period, the average total length of stay prior to 6 months was 33.95 days in the POPPI intervention group and 31.16 days in the usual care group. Overall, the total length of stay was higher in the POPPI intervention group sites than in the usual care group sites, but the intervention reduces the differences in length of stay by ≈1 day in the intervention period compared with the baseline period.
| Admission | Time period | |||
|---|---|---|---|---|
| Baseline | Intervention | |||
| Intervention (n = 283) | Usual care (n = 284) | Intervention (n = 340) | Usual care (n = 446) | |
| Index admission, mean (SD) | ||||
| Days in critical care | 12.55 (12.13) | 11.46 (9.63) | 12.74 (10.86) | 13.05 (12.98) |
| General medical bed-days | 17.10 (24.95) | 13.92 (18.82) | 19.51 (28.42) | 16.96 (25.56) |
| Re-admissions | ||||
| Re-admissions, n (%) | 32 (11.31) | 36 (12.68) | 26 (7.65) | 32 (7.17) |
| Days in critical care, mean (SD) | 0.67 (2.75) | 1.21 (5.28) | 0.58 (3.80) | 0.67 (2.75) |
| General medical bed-days, mean (SD) | 0.94 (6.70) | 0.93 (4.70) | 1.13 (6.64) | 0.94 (6.70) |
| Total length of stay up to 6 months (days), mean (SD) | 31.27 (31.04) | 27.52 (25.06) | 33.95 (33.66) | 31.16 (31.96) |
Table 32 summarises the resource use reported from responses to the health services questionnaire for all patients recruited during the baseline and intervention periods. The results are presented for the samples with complete information. The number of complete responses/eligible patients at 6 months are as follows:
- baseline period –
- intervention group, 160 out of 245 (65%)
- usual care group, 160 out of 259 (62%)
- intervention period
- intervention group, 187 out of 314 (60%)
- usual care group, 257 out of 415 (62%).
| Resource use | Time period, mean (SD) | |||
|---|---|---|---|---|
| Baseline | Intervention | |||
| Intervention (n = 160) | Usual care (n = 160) | Intervention (n = 187) | Usual care (n = 257) | |
| Inpatient days (general medical) | 5.38 (14.35) | 4.60 (11.83) | 9.53 (28.39) | 6.82 (19.39) |
| Outpatient visits | 4.33 (5.40) | 5.49 (6.70) | 4.69 (6.18) | 5.09 (6.13) |
| GP contacts | 4.69 (5.52) | 4.08 (5.36) | 4.05 (4.77) | 3.59 (5.06) |
| Nurse contacts | 3.97 (5.85) | 3.75 (6.51) | 4.39 (7.83) | 3.31 (5.36) |
| Health visitor contacts | 1.41 (4.76) | 0.59 (1.76) | 1.49 (6.57) | 2.5 (16.64) |
| Occupational therapist contacts | 0.70 (1.54) | 1.37 (6.28) | 1.35 (6.08) | 1.49 (7.71) |
| Speech therapist contacts | 0.27 (1.13) | 0.08 (0.48) | 0.06 (0.31) | 0.19 (0.92) |
| Physiotherapist contacts | 1.30 (2.25) | 0.75 (1.89) | 1.22 (2.49) | 1.87 (6.63) |
| Psychiatrist contacts | 0.11 (0.43) | 0.07 (0.36) | 0.15 (0.72) | 0.22 (1.04) |
| Psychiatric nurse contacts | 0.19 (1.45) | 0.05 (0.31) | 0.17 (1.00) | 0.02 (0.19) |
| Psychologist contacts | 0.06 (0.34) | 0.10 (0.75) | 0.04 (0.28) | 0.21 (1.04) |
| Counsellor contacts | 0.36 (1.59) | 0.05 (0.31) | 0.22 (0.92) | 0.41 (1.51) |
The average number of inpatient days reported from admissions other than those to critical care were higher in the intervention group than in the usual care group in both the baseline and the intervention periods. Patients recruited at intervention group sites had higher average numbers of contact with GPs and nurses and lower outpatient visits than those in the usual care group. All other community care contacts up to 6 months were higher or similar in the intervention group than in the usual care group in the baseline period. In the intervention period, all other community care contacts were lower in the intervention group than in the usual care group.
Total costs
The cost of the POPPI intervention was £140 per patient, which represents 0.5% of the total cost of treatment for the intervention group in the intervention period. The intervention cost included the cost of the IPAT assessment (£6), delivery of stress support sessions (£109), training (£4) and debriefing (£21). The net result of combining the intervention, critical care, general medical ward, and outpatient and community care costs is that the intervention group had higher mean total costs per patient than the usual care group, in both baseline and intervention periods (Table 33). The mean total cost per patient in the baseline period was higher in the intervention group (£27,824) than in the usual care group (£26,176). In the intervention period, the mean total cost per patient was similar between the treatment groups (£30,100 for the POPPI intervention group and £29,618 for the usual care group).
| Costs | Time period, mean (SD) | |||
|---|---|---|---|---|
| Baseline | Intervention | |||
| Intervention (n = 283) | Usual care (n = 284) | Intervention (n = 340) | Usual care (n = 446) | |
| Intervention costs | 140 (128) | |||
| Hospital costs | ||||
| Index admission | ||||
| Critical care | 19,221 (19,183) | 17,424 (15,195) | 19,573 (18,083) | 20,495 (21,627) |
| General medical ward | 5095 (7435) | 4149 (5608) | 5814 (8471) | 5055 (7618) |
| Re-admissiona | ||||
| Critical care | 910 (3897) | 1679 (7486) | 873 (6877) | 782 (4378) |
| General medical | 279 (1995) | 277 (1401) | 337 (1979) | 170 (1364) |
| Outpatient and community costsa,b | 2319 (4356) | 2646 (4551) | 3363 (7026) | 3118 (5609) |
| Total costs up to 6 monthsa,b,c | 27,824 (23,793) | 26,176 (20,957) | 30,100 (25,403) | 29,618 (27,287) |
Health-related quality of life
The health status profiles reported from responses to the EQ-5D-5L questionnaires administered at 6 months are summarised by treatment group and time period in Table 34. The results are presented for the samples with complete information; the number of complete responses/eligible patients at six months are as follows: baseline period: intervention group 188/245 (77%); usual care group 197/259 (76%); intervention period: intervention group: 248/314 (79%); usual care group: 320/415 (77%). At six months, the proportion of patients who reported ‘no problems’ for each dimension of the EQ-5D in the POPPI intervention group was lower than for the usual care group, in both baseline and intervention periods.
| EQ-5D-5L component | Time period, n (%) | |||
|---|---|---|---|---|
| Baseline | Intervention | |||
| Intervention (n = 188) | Usual care (n = 197) | Intervention (n = 248) | Usual care (n = 320) | |
| Mobility | ||||
| No problems | 55 (29.26) | 66 (33.5) | 90 (36.29) | 123 (38.44) |
| Slight problems | 51 (27.13) | 55 (27.92) | 44 (17.74) | 67 (20.94) |
| Moderate problems | 43 (22.87) | 37 (18.78) | 57 (22.98) | 60 (18.75) |
| Severe problems | 31 (16.49) | 31 (15.74) | 42 (16.94) | 52 (16.25) |
| Extreme problems | 8 (4.26) | 8 (4.06) | 15 (6.05) | 18 (5.63) |
| Self-care | ||||
| No problems | 109 (57.98) | 137 (69.54) | 161 (64.92) | 207 (64.69) |
| Slight problems | 41 (21.81) | 29 (14.72) | 37 (14.92) | 53 (16.56) |
| Moderate problems | 24 (12.77) | 22 (0) | 28 (11.29) | 41 (12.81) |
| Severe problems | 12 (6.38) | 5 (2.54) | 12 (4.84) | 14 (4.38) |
| Extreme problems | 2 (1.06) | 4 (2.03) | 10 (4.03) | 5 (1.56) |
| Usual activities | ||||
| No problems | 47 (25.00) | 68 (34.52) | 68 (27.42) | 104 (32.50) |
| Slight problems | 59 (31.38) | 60 (30.46) | 68 (27.42) | 84 (26.25) |
| Moderate problems | 45 (23.94) | 38 (19.29) | 59 (23.79) | 68 (21.25) |
| Severe problems | 20 (10.64) | 22 (11.17) | 28 (11.29) | 31 (9.69) |
| Extreme problems | 17 (9.04) | 9 (4.57) | 25 (10.08) | 33 (10.31) |
| Pain/discomfort | ||||
| No problems | 45 (23.94) | 50 (25.38) | 73 (29.44) | 86 (26.88) |
| Slight problems | 61 (32.45) | 68 (34.52) | 69 (27.82) | 117 (36.56) |
| Moderate problems | 52 (27.66) | 50 (25.38) | 65 (26.21) | 75 (23.44) |
| Severe problems | 20 (10.64) | 19 (9.64) | 30 (12.10) | 25 (7.81) |
| Extreme problems | 10 (5.32) | 10 (5.08) | 11 (4.44) | 17 (5.31) |
| Anxiety/depression | ||||
| No problems | 79 (42.02) | 89 (45.18) | 112 (45.16) | 144 (45.00) |
| Slight problems | 51 (27.13) | 57 (28.93) | 53 (21.37) | 102 (31.87) |
| Moderate problems | 32 (17.02) | 31 (15.74) | 56 (22.58) | 46 (14.37) |
| Severe problems | 10 (5.32) | 10 (5.08) | 18 (7.26) | 21 (6.56) |
| Extreme problems | 16 (8.51) | 10 (5.08) | 9 (3.63) | 7 (2.19) |
The mean EQ-5D-5L utility scores were similar between the treatment groups. Patients recruited at intervention group sites had higher all-cause mortality in the baseline period than the usual care group, but in the intervention period, the all-cause mortality was similar between both groups. The resultant mean QALYs were also similar across the treatment groups (Table 35). The net effect is that the intervention has a very small effect on QALY in favour of intervention in the intervention period.
| EQ-5D-5L/mortality/QALYs | Time period | |||
|---|---|---|---|---|
| Baseline | Intervention | |||
| Intervention (n = 283) | Usual care (n = 284) | Intervention (n = 340) | Usual care (n = 446) | |
| EQ-5D-5L (survivors),a mean (SD) | 0.661 (0.303) | 0.698 (0.268) | 0.668 (0.302) | 0.690 (0.279) |
| All-cause mortality, n (%) | 38 (13.43) | 25 (8.80) | 26 (7.65) | 31 (6.95) |
| QALY,a mean (SD) | 0.263 (0.132) | 0.285 (0.115) | 0.274 (0.120) | 0.291 (0.112) |
Cost-effectiveness
The intervention reduced the mean differences in costs by ≈£755 in the intervention period compared with the baseline period, but the mean effect was surrounded by statistical uncertainty. The mean incremental QALY gain for the POPPI intervention versus usual care was small (0.004), and with a 95% CI that included zero (Table 36). The INMB for the POPPI intervention versus usual care was positive at £835, but with considerable statistical uncertainty (95% CI –£4322 to £5992).
| Cost effectiveness | Time period, mean (SD) | Incremental effect, mean (95% CI) | |||
|---|---|---|---|---|---|
| Baseline | Intervention | ||||
| Intervention (n = 283) | Usual care (n = 284) | Intervention (n = 340) | Usual care (n = 446) | ||
| Costs (£) | 27,824 (23,793) | 26,176 (20,957) | 30,100 (25,403) | 29,618 (27,287) | –755 (–5883 to 4374) |
| EQ-5D-5L (survivors) | 0.661 (0.303) | 0.698 (0.268) | 0.668 (0.302) | 0.690 (0.279) | 0.009 (–0.063 to 0.081) |
| QALYs | 0.263 (0.132) | 0.285 (0.115) | 0.274 (0.120) | 0.291 (0.112) | 0.004 (–0.023 to 0.031) |
| INMB (£)a | 835 (–4322 to 5992) | ||||
When the uncertainty in the incremental costs and QALYs is represented on the cost-effectiveness plane, the majority of the points are in those quadrants that show that the POPPI intervention, on average, has lower costs and gains QALYs, although the magnitude of these average QALY gains were small (Figure 21). The probability that the POPPI intervention is more cost-effective than usual care, given the data, is ≈60% when willingness to pay for a QALY gain is zero; this probability is never > 65%, irrespective of how much society is willing to pay for a QALY gain (Figure 22).
FIGURE 21 Uncertainty in the mean costs (GBP) and QALY differences and their distribution for the POPPI intervention versus usual care. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)
FIGURE 22 Cost-effectiveness acceptability curve, reporting the probability that the POPPI intervention is cost-effective (within 6 months) at alternative willingness-to-pay thresholds for a QALY gain. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)
The estimated INMBs were similar for the scenarios considered in the sensitivity analyses (Figure 23). For example, the INMB is robust and remains ≈£835 regardless of the alternative assumptions considered in the sensitivity analysis, that is whether additional staff time is required to deliver the IPAT assessment or the stress support sessions, an alternative assumption for baseline HRQoL or an alternative distributional assumptions. Similarly, excluding the health services questionnaire costs had only a small impact on the mean INMB compared with base case INMB (£1346 vs. £835).
FIGURE 23 The sensitivity analysis that reports the mean (95% CI) incremental net benefit (at £20,000 per QALY) within 6 months according to alternative assumptions compared with the base case. Vertical line indicates no difference in net monetary benefits between comparator groups. SMR, standardised mortality ratio. HSQ, health service questionnaire. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)
The results of the subgroup analysis are presented in Figure 24, and show that the INMBs were similar across all subgroups. For the ‘duration of delirium’ subgroup, the mean INMB associated with the POPPI intervention, compared with usual care, increased when delirium was of longer duration. For all subgroups, as for the overall results, the CIs around the INMB included zero.
FIGURE 24 The INMB (95% CI) (GBP) at 6 months, by subgroups, at the recommended willingness to pay threshold stipulated by NICE (£20,000 per QALY). Vertical line indicates no difference in net monetary benefits between comparator groups. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)
Lifetime cost-effectiveness
Long-term survival
The Kaplan–Meier survival curves show that when the time horizon was extended beyond 6 months, for those with survival data available, the probability of survival was similar across treatment groups in both the baseline and the intervention periods (Figure 25). In the intervention period, survival was similar between the treatment groups at 6 months and over a longer follow-up period. In the baseline period, survival in the intervention group was lower than the usual care group at 6 months, but the survival difference between the randomised groups was minimised over time, even though a large proportion of cases were censored beyond 1 year.
FIGURE 25 Kaplan–Meier survival curves. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)
Long-term health-related quality of life
The long-term CEA required the HRQoL to be estimated over time. At 6 months, HRQoL was similar across randomised groups and periods. We used HRQoL from the general population at the age of 61 years (which was the median age of POPPI cluster RCT 6-month survivors) to predict the long-term HRQoL of patients recruited to the cluster RCT. HRQoL of POPPI survivors at 6 months was 13% lower than that of the age- and sex-matched general population.112 We used linear interpolation to predict the mean HRQoL for each treatment group between 6 months and 5 years; therefore, after 5 years, the mean for each randomised group was similar to that for the age-matched general population.
Long-term costs
To project lifetime costs attributable to the initial critical care episode, we considered the mean re-admission costs (in critical care, general wards, and outpatient and community services) up to 1 year, estimated from the cluster RCT electronic case report form, the CMP database and the health services questionnaires. The mean costs for each time period were calculated for those patients who survived to at least 1 year. For each intervention period, these mean costs were similar (£7043 for baseline and £7969 for the intervention period). These mean costs were used to impute mean costs between years 1 and 5. After year 5, we assumed no additional morbidity costs.
Cost-effectiveness
Table 37 presents the resultant lifetime QALYs, lifetime costs and INMBs. Overall, at the NICE-stipulated threshold of £20,000 per QALY, the INMB was £4158, but with a wide 95% CI that included zero. The cost-effectiveness acceptability curves show that that the probability of the POPPI intervention being cost-effective is ≈70% at the NICE-stipulated threshold of £20,000 per QALY (Figure 26).
| Lifetime cost-effectiveness | Time period, mean (SD) | Incremental effect, mean (95% CI) | |||
|---|---|---|---|---|---|
| Baseline | Intervention | ||||
| Intervention (n = 283) | Usual care (n = 284) | Intervention (n = 340) | Usual care (n = 446) | ||
| Costs (£) | 56,319 (25,020) | 56,193 (22,115) | 64,406 (26,482) | 64,254 (28,823) | 362 (–5077 to 5801) |
| QALYs | 10.58 (6.01) | 11.52 (5.82) | 10.89 (5.29) | 11.81 (5.51) | 0.226 (–0.447 to 0.899) |
| INB (£)a | 4158 (–10,354 to 18,670) | ||||
FIGURE 26 Cost-effectiveness acceptability curve, reporting the probability that the intervention is cost-effective (at lifetime) at alternative willingness-to-pay thresholds for a QALY gain. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)