Dissertation committee:
Παυλοπούλου Ιωάννα, Καθηγήτρια, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Αποστολοπούλου Ελένη, Ομότιμη Καθηγήτρια, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Μπενέτου Βασιλική, Καθηγήτρια, Ιατρική Σχολή, ΕΚΠΑ
Τσουμάκας Κωνσταντίνος, Ομότιμος Καθηγητής, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Κατσούλας Θεόδωρος, Αναπληρωτής Καθηγητής, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Γαλάνης Πέτρος, Επίκουρος Καθηγητής, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Τσερώνη Μαριγώ, Επίκουρη Καθηγήτρια, Τμήμα Νοσηλευτικής, ΕΚΠΑ
Summary:
Ventilator-associated pneumonia (VAP) has been recognized as a patient safety threat. The Centers for Disease Control and Prevention (CDC) began conducting VAP surveillance in the 1970s. However, the limitations of VAP diagnostic criteria led to the development of a more objective and potentially automatable approach to surveillance for ventilator-associated conditions and complications. This approach, termed «ventilator-associated events» surveillance, was designed to capture an array of infection- and noninfection-related events in patients receiving mechanical ventilation and was implemented in National Healthcare Safety Network from January 2013.
There are three levels of definition in the VAE algorithm: Ventilator-associated conditions (VAC), Infection-related ventilator-associated complications (IVAC) and Possible ventilator-associated pneumonia (PVAP). This algorithm is not a clinical definition and was not designed for use in patient management, but for hospital surveillance.
Although the economic impact of VAP is considerable, the additional cost of VAE remains unclear. Cost analysis is complicated by the fact that patients acquire infections during their hospital stay, having already spent time at risk without having an infection. This time at risk must be taken into account by treating infections as time-dependent exposures from ICU admission. Multistate models that describe the occurrence of events over time as transitions between multiple states are an attractive tool that accounts for time-dependent bias and competing risks bias. This innovation avoids the serious problems of bias, which can occur in all areas of epidemiology and clinical medicine and results in unbiased estimates, which offer a complete view of the situation prevailing in ICU.
The main purpose of the present study was to estimate the direct additional cost-fixed and variable- of VAE in adult ICU patients.
The individual objectives of the study were:
1. The calculation of VAE incidence density as a time-dependent exposure from admission to the ICU.
2. The cumulative risk assessment of VAE.
3. The estimation of the cumulative risk of the competing events of death and discharge without VAE.
4. The calculation of total mortality.
5. The assessment of patients with VAE discharge and death risk ratio.
6. The calculation of VAE attributable mortality.
7. The calculation of population attributable fraction of mortality.
8. The calculation of VAE additional length of stay.
9. The calculation of VAE additional length of stay for patients who died and for patients who discharged alive.
10. The estimation of the direct cost (fixed and variable) per day of hospitalization in ICU.
11. The estimation of direct additional cost per VAE episode.
12. The estimation of direct additional cost per VAE episode for patients who died and for patients who discharged alive.
13. The estimation of direct additional cost for all VAE.
A prospective study was carried out for the period January 2018-December 2019 in four ICUs of Attica. The sample consisted of all adult patients who received conventional mechanical ventilation in ICU for ≥ 4 days during surveillance and were followed until discharge from the ICU or until death. CDC standard definitions and standardized protocols for the year 2018 were used to diagnose VAE and monitor patients. VAE were defined using a combination of objective criteria including worsening respiratory status after a period of stabilization or improvement in the ventilator, infection or inflammation documentation and laboratory documentation of respiratory infection. Patients should be mechanically ventilated for at least 4 calendar days to meet VAE criteria. The day of start of mechanical ventilation is the first day, while the shortest date of event for VAE (the day of onset of worsening oxygenation) is the third day of mechanical ventilation. The reference period of stabilization or improvement on the ventilator refers to the first two calendar days preceding the first day of increased daily minimum of PEEP (positive end-expiratory pressure) or FiO2 (fraction of inspired oxygen) and must be characterized by ≥ 2 calendar days of stabilization or reduction of daily minimum PEEP or FiO2 values. The minimum daily value of PEEP or FiO2 used to monitor VAE is the lowest value during a calendar day if maintained for at least 1 hour.
To estimate VAE additional length of stay we used a four state model, in which «ICU Admission» was state 0, «VAE» was state 1, «Discharge» was state 2 and «Death» was state 3 and we modeled the risks between them. In our model the transition between states was determined by the hazard function λij, which estimates the daily risk of transition from state i to state j. the risk of transition from state i to state j is defined as: λij = Number of patients moving from state i to j / Summed patient-days in state i. only the following 7 simple numbers are needed to estimate the hazards of the multistate model: (1) the number of patients who acquire VAE, (2) the number of patients who are discharged without VAE, (3) the number of patients who are discharged with VAE, (4) the number of patients who die without VAE, (5) the number of patients who die with VAE, (6) the total patients days without VAE, and (7) the total patient days with VAE. We manipulated VAE as a time dependent variable. That is, at the time (day) the patient acquired VAE, his status changed from not exposed to this infection to exposed. From then on, these patients were considered exposed until their discharge from ICU. The four-state model basic analysis was based on 5 hazards: VAE hazard, discharge hazard without VAE, death hazard without VAE, discharge hazard with VAE and death hazard with VAE.
Descriptive analysis of the data was performed using the IBM SPSS Statistics package, Version 22. All tests were two-sided and statistical significance was defined as p < 0.05, with a power of 95%. The four-state model was used to estimate hazards, VAE occurrence and mortality.
VAE incidence density was derived by dividing the number of patients with VAE by the days at risk without VAE. Days at risk were the sum of length of stay for patients without VAE and length of stay until VAE occurred for patients with VAE.
To calculate the cumulative risk of VAE, the competing risks of discharge and death without VAE were taken into account, while at the same time as the total mortality in the ICU, the competing risk of discharge alive from the ICU was also studied.
The direct cost of hospitalization in the ICU was calculated using the bottom-up method, which is a microeconomic approach of analytically allocating costs to each patient according to the resources use. For each patient, the direct hospital cost were calculated, consisting of the fixed cost, as derived from staff payroll and the operation and maintenance cost of the department and machines, and the variable cost, as derived from medication, medical material and diagnostic tests.
The direct additional cost per VAE episode was calculated by the formula: VAE additional length of stay x Cost per day of ICU hospitalization.
The cost per day of ICU hospitalization was calculated using the formula: Sum of direct cost of ICU hospitalization for all patients/Sum of patient days for all these patients.
The total direct additional cost of VAE over the 2-year period was calculated by the formula: Number of patients with VAE x VAE additional length of stay x Cost per day of ICU hospitalization.
VAE mean additional length of stay was calculated taking into account the time of VAE onset (hospital days before VAE belong to the «uninfected group» rather than the «infected group» to avoid time-dependent bias. Mean additional length of stay standard error was calculated by bootstrap sampling using 1000 replications. Bias-corrected and accelerated confidence intervals (BCa CI) were calculated using DiCiccio and Efron’s method and adjusted for both bias and skewness in the bootstrap method. Additional length of stay was calculated for all VAE, but also by event (VAC, IVAC or PVAP), separately for patients discharged alive and for patients who died in the ICU. The four-state model analyses of VAE additional length of stay were performed with the statistical program R (Version 4.2.0) through RStudio (2022.02.2-485) using the etm package.
During the 2 year study period, 500 patients were hospitalized for ≥ 4 days with mechanical ventilation in 4 multipurpose ICUs of Attica for a total of 12,624 days. The 122 patients with diagnosis «infection at admission» were excluded from the study. In the final analysis 378 patients with 9,369 patient-days were included. The majority of patients were male (58.7%) with a median age of 60 years. The most common diagnosis was trauma (22.5%), followed by neurological (19%) and pulmonary disease (18%).
Of 378 patients 143 (37.8%) developed 143 episodes of VAE. Of all VAE, 58% were VAC, 30.8% IVAC and 11.2% PVAP, respectively. 49.7% of VAE were early onset and 50.3% were late onset.
The mean incidence of VAE (19.40 episodes per 1,000 ventilator days) was 4.3 times higher compared to that of the control population. By the type of VAE, the highest incidence (11.26 episodes per 1,000 ventilator days) was observed in VAC, followed by IVAC and PVAP with an incidence of 5.97 and 2.17 episodes per 1,000 ventilator days, respectively. From this comparison, it is deemed necessary to develop and implement active VAE surveillance and control programs with an emphasis on the set of VAE prevention bundle.
VAE incidence density (32.36 episodes per 1,000 ventilator days) taking VAE as a time-dependent exposure from admission was 1.7 times higher compared to the mean incidence and by type of VAE the highest incidence density was observed in VAC (14.30 episodes per 1,000 ventilator days). These findings emphasize the importance of distinguishing between the time before and after the VAE onset and the different time at risk in the denominator, that is, the total number of length of stay in patients without VAE plus the number of days until VAE onset in patients with VAE, to calculate VAE incidence density.
The high rates of attributable mortality (21.2%) and population attributable fraction (25.6%) indicate the necessity of implementing active surveillance programs to identify patients at risk and targeted interventions that will contribute to VAE reduction.
Investigating VAE and ICU mortality competing risks we found that the risk of VAE (37.8%) was 2.6 times greater than the competing risk of death (14.5%) and 0.8 times less of competing risk of discharge (47.6%). Patients with VAE had 58.2% increased risk of death and 38.6% reduced risk of discharge, compared with patients without VAE. The risk of ICU death (31.2%) was 0.5 times less than the competing risk of discharge alive (68.8%)
The following conclusions are drawn from the above results: the multi-state model allows the simultaneous analysis of competing risks of discharge and death without VAE with VAE risk and the competing risk of ICU discharge alive with ICU mortality, to avoid competing risk bias, which may lead to misleading estimates.
Investigating VAE impact on additional length of stay revealed that VAE crude additional length of stay was 17 days, while VAE mean additional length of stay after applying the four-state model was 6.55 ± 1.78 days. The mean additional length of stay was similar in VAE patients who died and VAE patients who were discharged alive (3.30 ± 1.17 and 3.25 ± 1.37 days, respectively). The longest additional length of stay among patients discharged alive was in VAC (4.50 ± 1.93 days), while among patients who died in the ICU it was in PVAP (7.09 ± 5.38 days). Our findings are difficult to compare, as the relevant literature is quite limited.
From the above results we conclude that the use of appropriate statistical methodology, which addresses issues as time-dependent exposure bias, leads to the avoidance of VAE additional length of stay overestimation.
From the study of VAE economic burden, it emerged that the direct cost per day of ICU hospitalization was € 492.80, the fixed and variable costs per day were € 135.20 and € 357.60, respectively, while the direct additional cost per VAE episode was € 3,227.84.
Regarding cost categories, variable cost account for 73%, while fixed cost account for the remaining 27% of direct cost. Regarding variable cost subcategories, antibiotics account for 67.1%, followed by other drugs with 28.5%.
PVAP had the highest direct additional cost per episode (€ 5,460.22), followed by IVAC with € 3,582.65 and VAC with € 3,415.10 per episode, respectively. Also, PVAP patients who died (€3,493.95) and VAC patients who discharged alive (€2,217.60) had higher additional cost.
The total direct additional cost for the two-year period was € 461,581.12 for all VAE. The total direct additional cost was higher for VAC (€ 283,453.30), followed by IVAC and PVAP with total direct additional cost of € 157,636.60 and € 87,363.52, respectively.
Given the fact that 30-70% of healthcare-associated infections are preventable, the study hospitals would have an overall benefit of 326 to 651 days of free ICU beds and a financial benefit of € 161,553.39 to € 323,106.78, respectively.
The above results confirm the importance of estimating the real cost of VAE using CDC standard definitions and standardized protocols in continuous active surveillance data collection, application of micro-costing for analytical cost allocation, and appropriate statistical analysis through the multi-state model to avoid additional length of stay and cost overestimation.
As a final conclusion, the present study using appropriate statistical methodology is the first at national and international level to highlight the high real burden of VAE and confirm the importance of simultaneous analysis of VAE and competing events when studying VAE risk, as well as the simultaneous analysis of death and discharge alive risk when ICU mortality is studied.
The robust and valuable findings of our study highlight the commitment of healthcare professionals to the international goal of «zero tolerance for VAE» and hospital administrations to increase resources in infection control to reduce the burden of VAE, increase safety and providing quality care to the ICU patients of our country at the lowest possible cost.
