|
Cost and outcome of mechanical ventilation for life-threatening stroke |
Mayer S A, Copeland D, Bernardini G L, Boden-Albala B, Lennihan L, Kossoff S, Sacco R L |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology The health technology assessed in the study was the use of mechanical ventilation (MV) for life-threatening stroke.
Type of intervention The object of the study was to establish whether the use of MV prolonged meaningful survival in stroke patients. This was addressed by looking at three main questions:
Study population The study population consisted of 'incidence' stroke patients who were treated with MV and met with predetermined entry criteria. The entry criteria was that all were aged over 39 years at the onset of stroke, were hospitalised at the Columbia-Presbyterian Medical Centre for their first stroke, and resided in Manhattan in a household with a telephone.
Setting The study setting was hospital. The economic study was conducted in Manhattan, USA.
Dates to which data relate The effectiveness evidence and hospitalisation resource data were collected between 1 July 1993 and 30 June 1996. Post-hospital resource data were estimated from literature published in 1993 and authors assumptions. The price year was stated to be 1996.
Source of effectiveness data Effectiveness data were derived from a single study and authors assumptions.
Link between effectiveness and cost data The effectiveness evidence and hospitalisation cost data were ascertained from the same group of patients. Both hospitalisation cost data and effectiveness evidence were collected prospectively throughout the study. Post-hospitalisation cost data were estimated using published literature and authors assumptions.
Study sample Power calculations were not performed to determine the study sample size. From the population of 510 patients enrolled into the Northern Manhattan Stroke Study (NOMASS) trial 52 (10%) were treated with MV. Indications for MV were categorised as follows: elective, emergent, pneumonia, seizure, pulmonary oedema and pulmonary embolism. The mean age of the 52 study patients was 65 years, 54% female and 46% male. The ethnic diversity was reported as 60% Hispanic, 27% Black and 13% White. Of the 52 study patients, 20 with cerebral infarction (CI), 24 with intracerebral haemorrhage (ICH) and 8 with aneurismal subarachnoid haemorrhage (SAH), all met the entry criteria and were treated with MV. The study sample was representative of the individuals who lived within the geographical area covered by the hospital. The mean admission National Institutes of Health Stroke Scale score was 17.3 (range: 0 - 30), and the mean Glasgow Coma Scale (GCS) was 9.6 (range: 3 - 15).
Study design This was a single centre population-based cohort study to establish the outcome of MV for life threatening stroke and 30-day survival predictors. Follow-up took place for all patients discharged alive, interviews were conducted at 6-month, 12 month and 24 month periods after the initial stroke. Final follow-up of all survivors took place in July 1999 (six years after initial enrolment). Duration of survival for those lost to follow up and those who survived longer than the six year follow-up period was calculated using the 1996 age, sex and race-specific estimates provided by the National Centre for Health Statistics. The number lost to follow-up was not reported.
Analysis of effectiveness The health outcomes used in the analysis were post-stroke 30-day, 6-month, 1-year and 2-year mortality and survival, and functional status, as measured by the Barthel Index (BI). Survival was estimated by the Kaplan-Meier method. It was not stated whether survival was calculated as the median or mean.
Effectiveness results The effectiveness results were as follows:
30-day mortality was 65% (34/52);
30-day mortality for CI patients = 50% (10/20);
30-day mortality for ICH patients = 71% (17/24); and
30-day mortality for SAH patients = 88% (7/8).
The mean (SD) duration of MV was 7.7 days (10.5);
mean intensive care unit (ICU) stay was 5.9 days (4.4); and
mean hospital stay 18.4 days (17.4).
Of those who died in hospital the mean (SD) interval between stroke and death was 8.9 days (9.9);
with immediate cause of death being neurological in 77% (26 cases); and
due to a medical complication in 23% (9 cases).
25 patients had a do not resuscitate order written during hospitalisation, this was associated with a higher 30-day mortality (84% versus 48%; p=0.007).
Survival was also shown as a Kaplan-Meier survival curve, but only for the first 30-days.
BI scores were:
30-day scores (SD): 95 - 100 = 1%(2); 70 - 90 = 0(0); 40 - 65 = 1%(2); 0 - 35 = 13%(25); unknown = 2(4).
6-month scores (SD): 95 -100 = 15%(2); 70 - 90 = 1%(2); 40 - 65 = 4%(8); 0 - 35 = 7%(13); unknown = 2%(4).
1-year scores (SD) were: 95 - 100 = 1%(2); 70 - 90 = 1%(2); 40 - 65 = 3%(6); 0 - 35 = 7%(13); unknown = 2%(4).
2-year scores (SD) were: 95 - 100 = 0(0); 70 - 90 = 1%(2); 40 - 65 = 1%(2); 0 - 35 = 6%(12); unknown - 1%(2).
Univariate analysis showed 30-day mortality to be significantly influenced by three variables: day of intubation GCS score; neurological deterioration after intubation; and neurological improvement after intubation. Multiple logistic regression showed both neurological deterioration after intubation and day of intubation GCS score to be independent predictors of 30-day mortality. Full results are available in table format within the paper.
Clinical conclusions Two thirds of those patient mechanically ventilated died in hospital within 30-days, 17% survived to 2-years, 1 with a BI of 70-90, 1 with a BI of 40-65 and 6 with BI's of 0-35.
Methods used to derive estimates of effectiveness The authors' assumptions were used to derive some estimates of effectiveness.
Estimates of effectiveness and key assumptions The assumption was made that all patients would have died on the day of their stroke if they had not been mechanically ventilated (zero-cost, zero-life assumption). The assumption was also made that survival of less than 30-days was equal to zero. Patients who were followed for at least 6-months were assumed to have a stable future level of disability (BI) if they were subsequently lost to follow-up or survived beyond 2-years. If a patient had no BI measurement after 14 days they were assigned a default value of modified Rankin Scale (mRS) = 4 (BI 40 - 65).
Measure of benefits used in the economic analysis The measures of benefit used in the study were both life-years saved and quality adjusted life years (QALYs). To obtain QALYs, utility values from a published cost-effectiveness analysis, which used the mRS were adapted. The authors converted there BI scores to mRS according to published criteria as follows: BI 95 - 100 = mRS 0 to 2; BI 70 to 90 = mRS 3; BI 40 to 65 = mRS 4; BI 0 to 35 = mRS 5.
Direct costs Discounting was carried out at a rate of 3% annually for the duration of the study (6 years). The price year was given as 1996, and both costs and cost estimates were adjusted accordingly. All costs were rounded up to the nearest $100. Resources were measured from hospital records during the study period and hospitalisation costs were obtained from the hospital accounting system during the same period. The hospitalisation costs did not include physician fees, since, as physicians were billed separately, these costs were too difficult to estimate. Post-hospital costs, (i.e., inpatient and outpatient rehabilitation costs) were estimated from both expert opinion and published literature. The costs of additional hospitalisation for recurrent stoke or other medical problems were not included. Unit cost and resource use were not reported separately. It was assumed that no costs would be incurred under the assumption that, had patients not been mechanically ventilated, they would have died.
Statistical analysis of costs Costs were not treated in a stochastic way.
Indirect Costs Indirect costs were not included; due to the perspective of the study they were not required.
Sensitivity analysis Initially a one-way sensitivity analysis was carried out in order to establish how sensitive both cost per life year and cost per QALY were to variations in different variables including: cost, survival, discounting, and utility variables. To test the plausible extremes of the variables within the model all such variables were adjusted to test the best and worse case scenarios. For the purpose of sensitivity analysis the discount rate was varied within the range 0% - 5%; justification for the range chosen was not stated. The default mRS of 4 which was assigned to patients who had no measure of BI after 14 days was tested with a range of mRS 0 to 2 and mRS 5.
Estimated benefits used in the economic analysis Estimated benefits were discounted at an annual rate of 3%. In the baseline analysis 1.72 life-years and 0.37 QALYs were saved per patient mechanically ventilated.
Cost results The mean cost of hospitalisation after MV was $27,300 (+/-21,500). The cost of hospitalisation per patient discharged alive was $83,400.
Synthesis of costs and benefits The incremental cost-effectiveness of MV was $37,600 per life year saved and $174,300 per QALY saved, under the base case scenario. A detailed sensitivity analysis was conducted and results were well documented within the paper. Both benefit measures were found to be mildly sensitive to survival and discounting and moderately sensitive to economic assumptions. Cost per QALY was found to be extremely sensitive to utility assumptions.
Authors' conclusions The authors concluded that MV is a cost-effective treatment for prolonging life but not as cost-effective in preserving quality of life when compared to other critical care and cerebrovascular interventions.
CRD COMMENTARY - Selection of comparators The comparator of no treatment was justified as the normal comparator when evaluating ICU care for acute, life-threatening diseases. The choice of comparator may bias the result against MV in that, as the authors acknowledge, without MV the patient would not receive absolutely no care. It also raises ethical issues, which might make this comparator inapplicable.
Validity of estimate of measure of effectiveness Study physicians obtained effectiveness data prospectively through review of study sample medical records and neurological examination. The effectiveness measures chosen by authors allowed them to address the three main aims of the study:
to determine how frequently stroke patients are mechanically ventilated;
to identify predictors of 30-day mortality; and
to evaluate the cost-effectiveness of MV for stroke in terms of patients discharged alive.
The study results may however have been weakened by the assumptions that were made in order to conduct the study.
The zero-cost, zero-life assumption, although stated to be standard practice, the assumption that less than 30-days mortality = zero life-years, and the assumptions about treatment of losses to follow-up all have an impact and may bias the results obtained.
The authors suggested that mortality predictors might have been biased due to elective withdrawal of life-support. It is also plausible that other concepts of care, which can vary between institutions, may bias the mortality predictors.
It was not clearly stated within the paper whether mean or median survival rate was used in calculations. The choice (due to the high 30-day mortality rate) would effect results. The authors stated that the regression analysis was of low power. However, the main predictors might be confounded by treatment factors and it is not clear why 30-day mortality was used and not 30-day survival, although the authors do state why they made the choice.
Validity of estimate of measure of benefit Two measures of health benefit were used within the study; life-years and quality adjusted life years (QALY). This allowed the authors not only to look at the question of MV prolonging life but also to address the question of the quality of that prolonged life. The authors converted from BI to mRS according to published criteria, any conversion can lead to deviations in results, and this will depend on the quality of the published criteria. The authors also used utility values from a published cost-effectiveness study; all transitions of data from one form to another can lead to error being introduced. Again the amount of error will be dictated by the quality of the publications from which the information has been taken. The authors also highlight that 'severe disability' has a utility value less than that of death; the negative utility value makes the cost per QALY results very sensitive to any change in value.
Validity of estimate of costs
The quantities and unit costs are not reported separately within the paper. Some of the costs used within the study were obtained prospectively from the hospital accounting system, but the authors highlight the problems incurred when attempting to cost physicians services and for this reason they were not included in the costing. The authors highlighted some of the study weaknesses in the paper, one of which is that only 42% of the total costs were directly measured, the rest being estimated from national databases. This may therefore not be a true reflection of the costs incurred by the study population, thus questioning the external validity of the study. Recurrent hospitalisation costs were also excluded and no records of how often patients are readmitted was included. If patients are frequently readmitted then the costs may be much higher than actually reported within the study. No indirect costs were included but, given the perspective from which the study was undertaken, they were not methodologically necessary. They would however, be significant if the study were conducted from a societal perspective.
Other issues The authors chose to enhance the results of their study by comparing the cost-effectiveness of MV to other published studies for critical care and cerebrovascular interventions. This is a valid comparison that allows the decision maker to evaluate the impact of the choices they make when allocating resources. The issue of generalisability to other settings was not really addressed and the study population may not have been representative of the population at large. The authors' conclusions were within the scope of the analyses and many of the limitations were reported within the paper.
Implications of the study The study highlights the differences between cost-per life years and cost per QALY, and these differences should be noted by the decision maker. The comparisons made by the authors show that MV might be cost-effective in prolonging life when compared to cost per life year saved of other critical care and cerebrovascular interventions. It did not, however, compare well with cost per QALY saved by other critical care and cerebrovascular interventions. The study allows decision makers to evaluate MV as a treatment compared to no treatment, within the caveats mentioned above, and then to address where resources will generate the greatest benefit, taking into account the opportunity cost of any decision taken. The authors suggest that 'large multicentre studies of diverse patient populations are needed to more precisely define the predictors of mortality or good functional outcome.'
Source of funding Supported by grants from the National Institute of Neurological Disorders and Stroke (RO1 NS 27517, RO1 NS 29993, and T32 NS 07153).
Bibliographic details Mayer S A, Copeland D, Bernardini G L, Boden-Albala B, Lennihan L, Kossoff S, Sacco R L. Cost and outcome of mechanical ventilation for life-threatening stroke. Stroke 2000; 31(10): 2346-2353 Indexing Status Subject indexing assigned by NLM MeSH Adult; Aged; Aged, 80 and over; Cohort Studies; Cost-Benefit Analysis; Female; Glasgow Coma Scale; Health Care Costs /statistics & Humans; Life Support Care /economics; Logistic Models; Male; Middle Aged; New York City; Quality of Life; Quality-Adjusted Life Years; Respiration, Artificial /economics /statistics & Stroke /economics /mortality /therapy; Survival Rate; Treatment Outcome; numerical data; numerical data /utilization AccessionNumber 22000001598 Date bibliographic record published 28/02/2003 Date abstract record published 28/02/2003 |
|
|
|