Background
Intensive care consumes a large proportion of healthcare resources. Days in intensive care are substantially more costly than general ward days in hospitals due to increased resource utilization and labour intensity. One major driver of these costs is mechanical ventilation.
Adverse events occurring during hospitalization add a substantial further burden to the healthcare system. The most prevalent hospital-onset conditions are hospital-acquired infections, which absorb substantial resources in hospitals and often require costly treatment [Reference Maragakis, Perencevich and Cosgrove1]. For patients under mechanical ventilation, one of the infections commonly acquired in the hospital is ventilator-associated pneumonia (VAP).
The major aim of this review is to determine the relative impact of mechanical ventilation on the daily costs of intensive care unit (ICU) care. We believe that this information is crucial in a number of contexts but has been neglected given the unavailability of reliable estimates.
We therefore conducted a systematic review regarding the costs of an ICU day and the proportion of these costs attributable to mechanical ventilation and quantified the relative increase in the daily costs using meta regression. In order to illustrate the practical relevance of these findings, we then determine the economic burden of VAP using recent cost information from France and Germany.
Methods
Literature search
On 24 March 2017 we searched MEDLINE via Wolters Kluwer's search interface Ovid (indexed and non-indexed databases), Web of Science via Thomson Reuters, (now Clarivate Analytics), CINAHL via EBSCOhost and the NHS Economic Evaluation Database (NHS EED) via Centre for Reviews and Dissemination, University of York. We performed the searches restricted to the English language and publication year 2000 to the update status at date of search (2017).
The search strategies varied by the database used. For details of the search strategies in all databases, see Supplementary material (Appendix). The searches returned 2839 studies of potential interest. After elimination of duplicates, 2072 studies remained.
Eligibility criteria
Studies were eligible if they gave data for the daily costs of ICU stays for general populations of adults (age ⩾ 18 years) and the added costs of mechanical ventilation. Screening of titles and abstracts found 257 potentially relevant papers. We were able to access full-texts of all articles of interest. Of these, 14 articles provided data on the daily costs of ICU stays, but only five articles [Reference Dasta2–Reference Moran6] fulfilled our inclusion criteria of also providing information regarding the added costs of mechanical ventilation (Fig. 1: PRISMA-chart). No additional relevant articles were found in the reference lists of the 14 articles providing data on the daily costs of ICU stay.
PRISMA flow diagram.
Data extraction
Two authors (KK and PH) checked the titles and abstracts of all potentially relevant English articles, and the full text of these was obtained. The same authors independently checked all the full-text articles for eligibility, and author KK extracted clinical data from the full texts. Details on the number of patients, ICU days, daily ICU costs and corresponding measures of uncertainty (standard deviations or 95% confidence intervals) were derived from the studies for ICU patients currently under mechanical ventilation or not. If necessary, patient days were derived from the proportion of patients ventilated and mean/median length of ICU stay (LoS). In the case of no reported variances, a standard deviation of 20% was imputed. In two cases (Fig. 1) it was assumed that ventilation takes place during 50% of total ICU stay.
Data analysis
Daily ICU costs were log transformed along with their standard errors and analysed using random effects meta regression (command metareg [Reference Harbord and Higgins7–Reference Stanley and Doucouliagos9] in Stata). As the studies' base years and currencies vary, an additional fixed effect on the study level was added to the meta regression. For the individual studies, t-tests based on summary data (command ttesti in Stata) were applied on log transformed daily costs. As the latter costs were used in all analyses, resulting coefficients were exponentiated in order to show the relative effect of ventilation on the daily ICU costs. Please note that the relative increase is always taking place within a country and year, therefore making adjustment for inflation and/or currency exchange rates unnecessary.
Application
The results of the different analyses were applied to recent results by Bluhmki et al. who calculated the excess LoS associated with ICU-acquired pneumonia [Reference Bluhmki10]. Due to methodological issues and/or lack of cost data, many studies analyse the burden of ICU-acquired pneumonia regarding the endpoint LoS but aim to express the results on a pecuniary scale as well. In comparison with the endpoint costs, analysing LoS is advantageous as it enables accounting for the time-dynamic pattern of the exposure. In line with Bluhmki et al. [Reference Bluhmki10] excess LoS was derived using a multistate methodology, which accounted for the time-dynamic pattern of both VAP as well as ventilation status. Their precise formulation further allows decomposing the excess LoS into extra days spent under, and not under ventilation. We emphasise that ignoring the time-dependency of VAP would lead to a substantial overestimation of extra days spent in the ICU [Reference Schumacher11Reference Heister, Kaier and Wolkewitz12]. On the one hand, this decomposition expresses the disease burden and a patient's quality status in more detail but it is also highly relevant from an economic point of view, because ventilation is known to be a major cost driver in the ICU. Thus, we combine our results with summary measures provided by Bluhmki et al. [Reference Bluhmki10] in order to obtain a more precise estimate of the additional costs associated with VAP.
For all statistical analyses Stata Version 14.0 (Stata Corp, College Station, Texas, USA) was used.
Results
The selection of studies is summarised in Figure 1. Ultimately, five studies were included in the analysis representing a total of 54 766 patients and ~238 037 patient days (Table 1). Four studies included detailed cost figures calculated from the hospital perspective [Reference Moerer3–Reference Moran6], while the other derived daily charges and multiplied these by hospital specific cost-to-charge ratios in order to give the hospital perspective [Reference Dasta2]. As shown in Figure 2, there was strong variability in the relative effect of mechanical ventilation across studies but overall, mechanical ventilation was associated with a 25.8% (95% CI 4.7%–51.2%) increase in the daily costs of ICU care.
Relative increase in daily ICU costs due to ventilation. For the individual studies, t-tests are applied on log transformed daily costs while the overall effect is calculated using random effects meta regression [Reference Sharp8]. All resulting coefficients are exponentiated in order to show the relative effect of ventilation on the daily ICU costs.
Overview of studies
PD = Patient days.
a Patient days derived from the proportion of patients ventilated and mean/median LoS. In two studies (Kahn and Jacobs) it was also assumed that ventilation takes place during 50% of total ICU stay.
b Daily costs were estimated by multiplying daily hospital charges by hospital specific cost-to-charge ratios.
Application
Bluhmki et al. [Reference Bluhmki10] reported a total extra LoS due to VAP of 3.52 days (95% CI 2.46%–4.59%) and found that the ‘excess LoS associated VAP is mainly triggered by excess LoS under ventilation, whereas the excess LoS under non-ventilation is almost negligible’. This group also conducted a more complex analysis allowing for a decomposition of excess LoS under ventilation and non-ventilation. For illustrative purposes and simplicity, we have used the summary measure based on the multistate model not distinguishing between ventilation and non-ventilation.
The study of Lefrant et al. [Reference Lefrant13] provided standardised unit costs (€1425 per ICU day, in 2008 Euros) for patients hospitalised in French ICUs and 65% of their patients received mechanical ventilation. If we further assume that, among these patients, ventilation took place during 50% of ICU stay, 32.5% of all patients days were ventilated ICU days. Combining these results (€1425) with the ventilation-related cost increase of 25.8%, results in daily costs of a single ventilated ICU day of €1654. Unfortunately, the duration of ventilation is not reported in the study making an assumption regarding the mean duration of ventilation necessary. For simplicity, we have assumed that ventilation takes place during 50% of ICU stay. The resulting 32.5% (0.5 × 0.65) share of ventilated ICU days may then be used to combine overall daily ICU costs (€ 1425, including both ventilated and non-ventilated ICU days) with our estimate regarding the ventilation-related cost increase (we have used the formulas €1425 = 0.325 × Cost_ventilated + (1−0.325) × Cost_(non-ventilated) and Cost_ventilated = 1.258 × Cost_(non-ventilated)). However, if we instead assume that ventilation takes place during 75% of ICU stay, the cost of a single ventilated ICU day changes slightly to €1592. Further combination of this cost figure with Bluhmki's data [Reference Bluhmki10] (3.52 days of excess LoS) then results in the financial burden of a single VAP of €5822 (95% CI €4012–€ 7632).
Bock et al. provide standardised unit costs (€1338 per ICU day, in 2011 Euros) for patients hospitalised in German ICUs [Reference Bock14]. Assuming that 50% of patients are ventilated and ventilation takes place during 50% of their respective ICU stay, we may combine these estimates (€1338) with the ventilation-related cost increase (25.8%) to obtain daily costs of a single ventilated ICU day of €1581. Further combination of this cost figure with Bluhmki's results [Reference Harbord and Higgins7] (3.52 days of excess LoS) then results in the financial burden of a single VAP of €5565 (95% CI €3881–€7249).
Discussion
Intensive care absorbs a large proportion of health care spending in industrialised countries [Reference Halpern and Pastores15]. Reliable information on the magnitude and variability of these costs is essential in order to guide spending decisions and as a starting point in identifying possible saving potentials. This is especially topical in light of variability between countries in resource use and concerns about demand elasticity [Reference Gooch and Kahn16–Reference Negrini19].
The results of the systematic review and meta regression analysis showed that mechanical ventilation is associated with a substantial increase in the daily costs of ICU care. Moreover, we observed a strong variability in the relative effect of mechanical ventilation across studies. However, the number of appropriate studies was rather limited and we were unable to find relevant more recent studies. As technology, procedures and indications for ventilation have changed in the meantime, our findings may be limited in scope and further studies on this topic are needed.
In the estimation of the additional LoS of adverse events during hospitalization, it is crucial to treat the exposure as time-varying to avoid a time-dependent bias [Reference Schumacher11Reference Heister, Kaier and Wolkewitz12]. This bias occurs when it is implicitly assumed that the conditions are already present on admission and leads to an overestimation of the incremental effect [Reference Heister, Wolkewitz and Kaier20]. Bluhmki et al. were aware of this fact and correctly estimated the excess LoS associated with VAP using multistate models accounting for the time-dynamics of ventilation status and VAP. At the same time, Bluhmki et al. [Reference Bluhmki10] considered that ignoring the time-dependency of VAP would lead to an excess LoS of 15 days, which represents a substantial overestimation in comparison with the more precise effect estimate of 3.52 days. This might also be the case in other studies, when cost figures are available, but ignoring the time-dependency of the exposure might have led to a substantial overestimation of the true effect [Reference Heister, Kaier and Wolkewitz12Reference Heister, Wolkewitz and Kaier20–Reference Kaier, Wolkewitz and Heister22].
Our cost calculations were based on the simplifying assumption that the entire excess LoS can be attributed to ventilation; thus, we may have slightly overestimated the costs. However, we refer to the result that this excess seems to be mainly triggered by ventilation [Reference Bluhmki10]. While LoS may be of intrinsic interest as an indicator for the burden on the patient level, it is a rather incomplete proxy for costs as the care intensity of the respective patients is not taken into account. Nonetheless, many studies multiply the extra LoS derived from multistate models with time fixed average daily costs, or standardised unit costs [Reference Stewardson23]. Not taking care intensity into account, however, leads to an underestimation of the costs, as post-infection daily costs are higher than average daily costs. Only a few studies use more accurate average cost per intensive care day, however only to estimate the financial burden of hospital-acquired infections, so they could not be used in our analysis [Reference Shepard24–Reference Sosa-Hernández26]. For severe infections such as pneumonia, however, ventilation is a major driver of ICU cost and should be taken into account when measuring the financial burden of adverse events in ICU settings.
Conclusions
Our study contributes to a more detailed understanding and measuring of the costs of intensive care and mechanical ventilation by providing first estimates and discussing methodological particularities. Being an area with relatively little empirical evidence to date, more studies on the daily costs of mechanical ventilation and intensive care are duly needed.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0950268819001900
Data
The study is based on publicly available data.
Acknowledgements
None.
Authors' contributions
KK and MW conceived of and designed the study. EM designed and refined the search strategy and conducted the search. KK and PH were the primary abstract reviewers. KK and PH reviewed the full texts and selected the articles for inclusion. KK performed the data extraction. KK and TH designed and implemented the statistical analysis. KK wrote the primary draft, with EM, PH and TH participating. TB participated in the analysis and provided additional data and analysis. All authors critically reviewed the manuscript, provided final approval and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Financial support
This research project receives support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no 115523 | 115620 | 115737 resources of which are composed of financial contribution from the European Union Seventh Framework Programme (FP7/2007-2013) and EFPIA companies in kind contribution. The research leading to these results was conducted as part of the COMBACTE-MAGNET consortium. For further information please refer to www.COMBACTE.com.
TH and MW are recipients of grants from the German Research Foundation (grant no. WO 1746/1-2 to MW and grant no. KA 4199/1-1 to TH). KK has received funds from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115737-2 (Combatting bacterial resistance in Europe – molecules against Gram negative infections (COMBACTE-MAGNET)).
The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
Conflict of interest
The authors declare no conflicts of interest.
Ethical standards
This work did not involve any contact with patients or handling of patient data beyond the summary outcomes reported in the studies evaluated.
Open access
