Economic model of albumin infusion in septic shock: The EMAISS study

The cost‐effectiveness of albumin‐based fluid support in patients with septic shock is currently unknown.

patients with sepsis and septic shock. When these patients require substantial amounts of crystalloids, the use of albumin in addition to crystalloids is suggested for initial resuscitation and subsequent intravascular volume replacement. 7 Albumin is widely used in clinical practice and it is known for some physiological properties other than the maintenance of oncotic pressure, such as antioxidant effects and positive effects on vessel wall integrity. 8 Albumin interacts with the glycocalyx and reduces vascular permeability, 9 thus enhancing microcirculatory function in critically ill patients. 10 Albumin can restore endothelial response to acetylcholine. 11 In the SAFE (Saline versus Albumin Fluid Evaluation) trial, albumin was shown to be as safe and effective as 0.9% saline in a pool of Intensive Care Unit (ICU) patients requiring fluid administration, and a trend toward a mortality reduction was specifically observed in patients with severe sepsis. 12 On the contrary, the use of artificial colloids such as hydroxyethyl starches (HES) has been banned in patients with sepsis, due to safety concerns, including renal function deterioration and mortality. 13,14 Another large trial, the ALBIOS (ALBumin Italian Outcome Sepsis) one, including patients with septic shock as well as patients with severe sepsis, addressed the efficacy of albumin added to crystalloids in improving survival compared to crystalloids alone; no overall mortality benefit was shown in patients with sepsis, but lower 90-day mortality was observed with albumin in the subgroup of patients with septic shock (Relative risk [RR] 0.87; 95% CI 0.77-0.99). 15 However, another French trial did not document any mortality reduction of Human Albumin in patients with septic shock. 16 Some meta-analyses have also stated the benefit of albumin in terms of survival in patients with sepsis 17,18 and specifically with septic shock, 19 while others have concluded to a lack of effect of albumin on mortality. 20,21 An obstacle for the use of albumin in patients with septic shock might be its price, as it can be perceived as an expensive option if only the cost of the product is considered. 22 However, albumin has been described as more cost-effective than crystalloids and HES when total medical costs and complications are considered. 23 In 2007, the COASST study, comprising a population of patients with severe sepsis or septic shock included in the Cub-Réa database, and considering the 4.6% reduction in mortality observed in the SAFE trial, also concluded that albumin is cost-effective in those patients. 24 Considering the results of the ALBIOS trial, 15 the aim of the EMAISS (Economic Model of Albumin Infusion in Septic Shock) study was to assess the cost-effectiveness of albumin in patients with septic shock, under standard medical practice in French ICUs.

| Patients and treatment
The data used in this cost-effectiveness study were collected from the Cub-Réa database. 25 All data are recorded in each ICU with standardized database software. Quality-assurance controls are applied to ensure consistency between the diagnoses and procedures. The data are transmitted anonymously to the coordinating center in a once a year basis.
Patients admitted with septic shock (ICD10 code "R572") to one of the 28 participating ICUs from 1 January 2014 to 31 December 2016 were included. All patients received either epinephrine or norepinephrine. Patients with ascites and likely to receive albumin were excluded. Only the first stay was included in the analysis for those patients admitted two or more times due to septic shock.
Long-term survival was determined from a group of 184 patients admitted to the ICU for septic shock between 1998 and 2000, who fulfilled the Cub-Réa inclusion criteria, 27 and whose 10-year vital status was obtained from the French Epidemiological Centre on causes of deaths (CepiDc). 28 Their characteristics were compared to those included in our study using either Chi-square test or Student t test, as appropriate.

| Model
A decision tree model was developed to estimate ICU costs and health outcomes when using either albumin and crystalloids or crystalloids alone in patients with septic shock, from ICU admission and over a lifetime horizon. Different analyses were made for those patients being discharged alive from the hospital or dying within the hospital ( Figure S1). Albumin administration was assumed to last for a maximum duration of 7 days.
The data obtained from Cub-Réa database for this analysis were age, sex, ICU length of stay (ICU-LOS) for patients who died in hospital, ICU-LOS for patients discharged alive, hospital LOS for patient who died at hospital, hospital LOS for patient discharged alive, mortality rate with crystalloids, and cost of hospital stay per day. The

Editorial Comment
In many countries, there is a demand to demonstrate the economic effectiveness of treatment protocols. Models to calculate the health economic values have emerged. For these, it is the internal and external validity of the patient cohort investigated that determines the effectiveness, medical, or economic. reduction of mortality with albumin compared to crystalloids alone in patients with septic shock was taken from the ALBIOS trial. 15

| Estimation of costs and life expectancy
For this analysis, the perspective of the French National Health System (Caisse Nationale d'Assurance Maladie) was adopted. Medical direct costs incurred during the ICU stay and related to the treatment strategies were included in the analysis, considering both tariffs 29 and prices as costs. The total ICU cost was calculated by

| Base-case analysis
The Incremental Cost-Effectiveness Ratio (ICER), ie, the difference in cost per life-year gained for albumin vs crystalloids alone, was estimated using the mean values of the parameters presented in Table 1 as inputs of our model.

| Sensitivity analysis
In order to appraise the uncertainty of our estimations, several sensitivity analyses were performed including a one-way Deterministic Sensitivity Analysis (DSA) and a Probabilistic multivariate Sensitivity Analysis (PSA). In DSA, the inputs of the model were changed one by one using extreme values for each parameter and the ICER was recalculated. In PSA analysis, except for the price of albumin and the additional daily price for ICU stay, which are fixed, the value of the input variables was simulated from their respective distribution.

| RE SULTS
During the 3-year study period, a total of 86 152 ICU stays and 8504 patients with septic shock (11% of the entire cohort) were recorded in the Cub-REA database. Patients with septic shock related to infected ascites and second or later stays after a first ICU stay were excluded leaving 6406 cases that fulfilled the inclusion criteria ( Figure 1) Figure   S2). Differences between the patients admitted during the two periods (2014-2016 and 1998-2000) were found in terms of mean age, SAPS II score, presence of comorbidities and percentage of patients requiring mechanical ventilation, as described in Table S1.
The RR ratio of mortality reported in the ALBIOS trial (0.87) was assumed, leading to a mean increase in 0.49 years in survival with albumin added to crystalloids compared to crystalloids alone.
The use of albumin was calculated to have an incremental cost of €480, compared to crystalloids. Consequently, the cost per life-year gained with albumin was estimated to €974 ( Table 2).
The scatterplot of ICER performed for 2000 simulations is presented in Figure S3. Depending on the willingness-to-pay threshold set at €20 000 or €30 000 per life-year saved, the probability of albumin being cost-effective was 95% or 97%, respectively, as shown in the acceptability curve ( Figure 2).
As cost-effectiveness can be sensitive to different factors, a oneway sensitivity analysis according to several factors, such as the age of patients or length of hospital or ICU stays, is presented in Table 3.
The results of DSA combined with PSA are presented in Figure 3 and Figure S4 for age, S5 for sex and S6, S7, S8, S9 for the impact of LOS and survival status. The probability of albumin being cost-effective seemed to be especially sensitive to the risk ratio of mortality (albumin/crystalloids). The analyses also show that the ICER with albumin plus crystalloids compared to crystalloids was sensitive to the variation of ICU-LOS ratio (albumin/crystalloids) (S10). Other factors have a moderate impact on the probability that albumin is cost-effective, especially when €10 000 is considered a willingness-to-pay threshold, are the annual septic shock-specific mortality (S11), the number of Albumin vials infused (S12), and cost of hospital stay (S13). Parameters of a beta distribution are derived from the n and the (%). Thus, g = n and h = n+(100 − (%))/(%).
Abbreviations: ATIH, agence technique de l'information sur l'hospitalization; ICU, intensive care unit; LOS, length of stay; RR, risk ratio. The minimum and the maximum values are hypotheses.

| D ISCUSS I ON
Cost of healthcare interventions is certainly an issue today 32 and it will need to be even more considered in the future. 33 In the case of patients with sepsis, its incidence and expenses are extremely high, as it is the mortality rate after discharge. 34   In the COASST study, which also analyzed the cost-effectiveness of albumin in patients with sepsis registered in the Cub-Réa database, 24  In the EMAISS study, the computation of mortality reduction We did not include the EARSS trial 16 in the analysis because although some results were made public that did not confirm the results of Albios, the study was never published and could not be subjected to rigorous scientific appraisal. It should be emphasized that, in the ALBIOS trial, albumin administration was not restricted to the first hours but also included albumin supplementation for 28 days after enrollment. These data, together with the lack of effects in patients enrolled with early sepsis, suggest that there are beneficial effects associated with albumin use in relation to its ancillary functions, rather than only to its primary oncotic properties. 36 Another concern is the assumption that Cub-REA patients did not receive Albumin. We don't have the type and volume of fluid received by patients in CUB-REA. However, the French recommendations are not in favor of colloids and albumin either for vascular filling 37 or for prevention of acute renal failure. 38 41 which is close to the incremental cost in our study.
In this last paper, the additional cost per life saved was estimated to be $14 384, whereas our cost of €974 was estimated per life-year gained. Applying the survivors' life expectancy at the Cub-Réa ICU in our model (around 10 years) to their ratio would result in an ICER close to ours (around $14 000 for 10 years). Albumin has also been shown to reduce the incidence of some complications or morbidity compared to crystalloids 9 and these differences could also have an important impact on direct healthcare costs.
A potential adverse effect of albumin on renal function was not considered, 23 as no clear signal of harm has been reported in previous studies, except for the huge amount of albumin with very high resulting oncotic pressure, 9 which is not the case of the patients included in the EMAISS study.

| CON CLUS ION
On the basis of the risk reduction observed in the septic shock subgroup analysis of the ALBIOS dataset, albumin-based fluid support may be cost-effective for patients admitted to the ICU with septic shock.

ACK N OWLED G EM ENTS
Jordi Bozzo, Francisco Mota, and Jemina Moretó (Grifols) are acknowledged for medical writing assistance and preparation of the manuscript.

CO N FLI C T O F I NTE R E S T
B Guidet received honorarium for lecture, by Grifols.

AUTH O R S ' CO NTR I B UTI O N S
BG initiated, conducted, and finalized the Work. IG, JR, and PA conducted the economic analysis. PA is the administrator of CUB-REA database and conducted the long-term follow-up study. BG, IG, JR, and PA contribute to the draft of the manuscript. All authors read and approved the final manuscript.

E TH I C S A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
The Cub-Réa database is compliant with the French law, and has been approved by the French data protection authority (Commission Nationale de l'Informatique et des Libertés, CNIL).

CO N S E NT FO R PU B LI C ATI O N
Not applicable.

AVA I L A B I LIT Y O F DATA A N D M ATE R I A L
The dataset supporting the conclusions of this article is available upon request to the president of CUB-REA network (bertrand.gui-det@aphp.fr).