Ventilator-Associated Events Cost in ICU Patients Receiving Mechanical Ventilation: A Multi-State Model

A prospective observational study was carried out in four medical-surgical ICUs in Athens, Greece for the period January 2018 to December 2019. The sample consisted of all adult patients who received conventional mechanical ventilation for ≥4 days during surveillance. Follow-up of patients lasted from admission until discharge from the ICU or until death. In patients with more than one episode of VAE, only the first episode was considered. Exclusion criteria were age below 18 years, duration of mechanical ventilation less than 4 days, and confirmed infection on the day of ICU admission.

The diagnosis of VAEs was based on the criteria of the CDC standard definitions [12]. There are three levels of definition in the VAE algorithm: ventilator-associated conditions (VACs), infection-related ventilator associated complications (IVACs), and possible ventilator-associated pneumonia (PVAP). VACs refer to respiratory status deterioration defined as an increase in daily minimum PEEP ≥3 cm H₂O or FiO₂ ≥0.20 maintained for at least 2 days after a basic period (2 days) of stability or improvement. IVACs, along with the above criteria, include evidence of infection or inflammation defined as a white blood cell count ≥12,000 cells/mm³ or ≤4,000 cells/mm³, temperature >38 °C or <36 °C and a new antimicrobial prescription has been started and continued ≥4 days by the attending physician. PVAPs, in addition to the above criteria, include evidence of lower respiratory tract infection, defined as purulent respiratory secretions or positive culture or more strict criteria, including positive lung histopathological examination, positive pleural fluid culture and other tests, such as Legionella spp [12].

Data was collected daily from the patients’ files, from the forms of the nursing staff as well as from the micro-biological laboratory. For each mechanically ventilated patient, the daily minimum value of PEEP and FiO₂ maintained for at least 1 hour during each calendar day was recorded. An interruption of mechanical ventilation for at least one day, followed by re-intubation or resumption of mechanical ventilation during the same hospitalization was recorded as a new episode of mechanical ventilation. For patients with VAEs, information was collected regarding mechanical ventilation conditions, amount and color of respiratory secretions, bronchial secretion cultures, antibiotic treatment, and chest radiographs.

To estimate the additional LOS associated with VAEs, we used a MSM regarding VAE as a possible intermediate state between ICU admission and discharge or death. In the 4-state model used, admission (patients without VAEs) is state 0, VAE is state 1, and discharge (state 2) or death (state 3) indicates the end of ICU LOS. Patients in state 2 cannot enter state 3 (Figure 1).

Fig. 1.

We assumed that the distribution of time to event followed a recommended exponential distribution. For cases in a given state up to a specified time, the hazard (λ) approximates the probability of moving from one state to another in a short interval, divided by the length of the interval. Based on the 5 hazards from the MSM, it is possible to quantify risk and clinical outcomes of VAE in terms of mortality and LOS (Table 1) [11].

Since the change in LOS due to VAEs is caused by the days after VAEs are acquired, it is necessary to consider the timing of VAEs. Based on the patients’ pathway across available model-states transitions, the MSM provides a weighted average of the LOS. The mean difference in LOS was calculated for each day as the difference between the estimated LOS given whether the intermediate state (VAEs) had been achieved or not by that day. The overall change in LOS was calculated as a weighted average of these quantities, with the weighting determined by the observed distribution of time to onset of VAEs [13].

All patients were followed up until discharge or death in ICU; there was no administrative censoring. VAEs could be one of the following events: VACs, IVACs or PVAPs. Each of the patients was infected from only one of these. The additional LOS is presented overall (VAEs) and stratified by VAEs event (VACs, IVACs or PVAPs). Also, the additional LOS was calculated separately for patients with VAEs who were discharged alive, as well as for patients with VAEs who died in the ICU.

The direct cost of ICU hospitalization was calculated using the bottom-up method, which constitutes a microeconomic approach of analytical distribution of costs for each hospitalized patient according to the use of resources [14]. Direct hospital costs, consisting of fixed and variable cost, were calculated for each patient.

The daily fixed cost was derived from the payroll of the staff and the ICUs operating and maintenance costs. These costs were obtained from data provided by the Greek Ministry of Health based on the annual cost reports of the hospitals participating in the study [15]. ICU LOS was recorded prospectively for each patient and the number of days in the ICU was used to calculate the fixed cost of hospitalization in the ICU. Fixed cost was calculated by multiplying ICU LOS by the corresponding daily ICU costs.

Variable costs were derived from daily costs of antibiotics and other medications, enteral and parenteral nutrition, blood products, medical supplies, and laboratory/imaging tests. The cost of medication was calculated by multiplying the daily dose of each drug by the days of medication for each patient and then multiplying by the price of the drug per recorded unit (gr, ml, etc.). Hospital costs of drugs, enteral and parenteral nutrition, as well as transfusion costs were derived from relevant data from the Ministry of Health [16]. In order to record the exact dosage of antibiotics and other drugs, a daily review of patient files was performed. The cost of the medical supplies was obtained from the official “Observe Net” website of the Health Supplies Committee [17], which is the medical supplies price observatory for Greek public hospitals. Medical supplies included were endotracheal tubes, tracheostomy tubes, central venous catheters, bladder catheters, pressure ulcer pads, and enteral and parenteral feeding devices. The costs of diagnostic tests were obtained from the official website of the National Organization for Healthcare Services Provision [18]. These costs were then multiplied by the number of diagnostic tests performed on each patient. Diagnostic tests included are electrocardiogram, electroencephalogram, radiographs, ultrasound, CT and MRI scans, as well as microbiological tests (cultures).

The direct additional cost per VAEs episode was calculated by multiplying VAEs additional LOS by the cost per day of ICU hospitalization. The cost per ICU hospitalization day was calculated by dividing the sum of the direct ICU hospitalization costs for all patients by the sum of the ICU days for all these patients. The total direct additional cost of VAEs over the 2-year period was calculated by multiplying the number of VAE cases by the derived additional LOS and the estimated costs per ICU bed day.

Categorical variables were expressed as absolute (N) and relative frequencies (%) and differences between the two groups were compared using the χ² test or the Fisher exact test, where appropriate. Continuous variables were expressed as median and interquartile range, and differences between groups were compared using the nonparametric Mann-Whitney U-test. The Kolmogorov-Smirnov test was used to test for normality. Descriptive analysis of the data was performed using statistical packages for social science version 22 (SPSS Inc., Chicago, IL).

The mean additional LOS of VAEs was calculated using MSM, taking into account the time dependence of VAEs and thus avoiding time-dependent bias. Transitions between states were determined by time-varying hazards, which were estimated using the Aalen-Johansen estimator. The expected LOS associated with VAEs was computed by a function of these transition probabilities. We calculated the standard error for extra LOS by bootstrap sampling using 1,000 replicates. The bias-correlated and accelerated confidence intervals (BCa CIs) were estimated with the method of DiCiccio and Efron [19]. All analyses were performed with R (version 4.2.0) via RStudio (2022.02.2 – 485) using the etm and mvna packages.

The present study responded to the fundamental principles of ethics. Complete confidentiality was maintained regarding the data collected, anonymity secured, and the research protocol was approved by the Hospitals’ Scientific Committees, which has waived the need for informed consent.

During the 2-year study period, 500 patients were hospitalized for ≥4 days under mechanical ventilation in 4 medical-surgical ICUs; 122 were excluded because of confirmed infection on the day of admission. In the final analysis, 378 patients with 9,369 days of ICU hospitalization were included. The characteristics of patients with and without VAEs are shown in Table 2.

In total, 143 of 378 patients (37.8%) developed 143 VAEs (83 VACs, 44 IVACs and 16 PVAPs). Overall, 72 VAEs (50.3%) had an onset date on or after day 5 of mechanical ventilation. The rate of VAEs was 32.36 episodes per 1,000 ventilator-days. The highest incidence rate (14.30 episodes per 1,000 ventilator-days) was observed in VAC, followed by IVAC (6.85 episodes per 1,000 ventilator-days) and PVAP (2.30 episodes per 1,000 ventilator-days).

The risk of VAE (λ₀₁) was 2.6 times greater than the competing risk of death (λ₀₃) and 0.8 times less of competing risk of discharge without VAE (λ₀₂). For all types of VAE it appears that λ₀₂ was greater than λ₁₂. This means that patients without VAE had a higher daily discharge probability, compared to patients with VAE. Also, λ₁₃ was greater than λ₀₃. This means that patients with VAE had a higher daily risk of death in the ICU, compared to patients without VAE (Table 3).

The overall ICU mortality was 31.2% and the competitive risk of ICU discharge alive was 68.8% (260/378). Death hazard ratio was 1.582 (95% CI: 1.101–2.278), which means that patients with VAE had a 58.2% increased risk of death, compared to patients without VAE. The discharge hazard ratio was 0.614 (95% CI: 0.509 – 0.960), which means that patients with VAE had a 38.6% reduced probability of discharge, compared to patients without VAE (Table 4).

The attributable mortality of VAEs was 21.2%. PVAPs had the highest attributable mortality (26.3%), followed by VACs (15.7%), and IVACs (10.0%). The population attributable fraction was 25.6%. By type of VAE it appears that overall mortality in the study population would have been reduced by 11% in the absence of VAC and by 3.6% and 2.1% in the absence of IVAC and PVAP, respectively.

The crude additional LOS for VAEs was 17 days (27.5 days in patients with VAEs versus 10.5 days in patients without VAEs, p<0.001).

Figure 2 illustrates the estimated LOS of patients with VAE still hospitalized, versus those still hospitalized and without VAE, for each day of the study interval up to 30 days after ICU admission. For each day under study, the expected LOS of patients with VAE appears longer than the expected LOS of patients still without VAE on the same day. These differences resulted in an excess LOS of 6.55 days (95% BCa CI: 2.76–9.73); 3.30 additional ICU days regarding death; and 3.25 additional days regarding discharge (Table 5).

Fig. 2.

The direct cost per day of ICU hospitalization was €492.80. Fixed and variable costs per ICU day were €135.20 and €357.60, respectively. Regarding the variable cost, the cost of antibiotics was higher (€239.84) followed by the cost of other drugs (€102.06), diagnostic tests (€7.80), medical equipment (7.26 €) and transfusion (0.64 €).

The direct additional cost per VAE episode was €3,227.84 (Table 6). Variable cost accounts for 73% (2,342.28 / 3,227.84), while fixed costs account for 27% (885.56 / 3,227.84) of direct costs. With reference to the variable cost subcategories, antibiotics account for the largest percentage (67.1%) followed by other drugs (28.5%). The total direct additional cost for the 2-year period was €461,581.12 for VAEs; €283,453.30 for VACs; €157,636.60 for IVACs; and €87,363.52 for PVAPs.

Our results should be interpreted in the context of some possible limitations. First, this study was performed in 4 medical-surgical ICUs, and our results may not be generalizable to other hospitals, given differences in staffing, patient population, and available resources and structures. Second, the economic consequences of VAE related morbidity and mortality from a societal perspective (e.g. lost productivity) were not considered. Because of this perspective, the time horizon of the analysis is limited to the period of hospitalization in the ICU. However, VAEs impose a significant burden on other environments as well. After leaving the ICU, patients are transferred to hospital clinics. In addition to the costs to the health care sector, there may be costs to the patient and their family due to lost work hours. Further analysis could therefore be considered to extend this perspective beyond the ICU [26].

Despite the limitations, our study has several strengths as it is a prospective study, which presents an accurate mapping of the clinical and economic impact of VAEs, based on CDC standard definitions and protocols for VAEs diagnosis and patient monitoring [12]. Our findings highlight the problem of VAEs and although they reflect the Greek ICU situation they are useful for other countries as well. Some follow up after the results are sent to the managers or authorities will be profitable in order to see what we can do to reduce these costs.

In a time of cost containment, estimating the financial burden of VAEs is a key tool for infection control professionals competing for resources [27]. It has become clear that a more structured approach to costing VAEs in the ICU is needed to enable a better comparison between published data [28]. However, cost estimation is not straightforward. The methodology we used differs from those used in previous studies. First, we used a MSM, which provides a conceptual framework for appropriate statistical analyses, such as time-dependent exposure, competing risks, and other issues, such as estimating attributable mortality and LOS [11]. Second, our method approaches micro-costing techniques, which allow the calculation of the total cost of each species and each patient, knowing the exact resource consumption for each patient [27]. We relied on a methodologically rigorous cost analysis, including not only fixed costs, but also the costs of antibiotics, other drugs, transfusions, diagnostic tests, and medical supplies. Additionally, we used actual costs incurred by the hospital rather than charges, as this is a more accurate measure [29].