This model-based analysis is the first study which examined the cost-effectiveness (costs of averted catheter-related complications) of ultrasound guided versus landmark oriented central venous catheterization in German adults. In the base case, the use of UG for a central venous catheterization compared with LM reduces the costs by €179 per procedure and is associated with on average 0.14 less complications. In line with previous studies from other health systems [10, 11], our results confirm, the use of UG for central venous catheterization being more effective in averting complications and less costly than the LM. These results did not change when various sensitivity analyses were applied. Replacing LM by UG could result in at least 8000 avoided complications annually which might correspond to a minimum of cost savings of €10.2 million each year for the German SHI.
Our model has several strengths. First, the analysis considers the correlation between the number of cannulation attempts and the total complication rate. Incorporating the increased likelihood of a complication in relation to the number of cannulation attempts enables a more realistic reflection of the clinical pathway of a central venous catheterization. Unfortunately, the Cochrane review  and the underlying studies do not specify whether the number of arterial puncture was measured on the first or second attempt. However, results of the univariate sensitivity analyses showed that a variation of the incidence of arterial puncture affects the cost-effectiveness ratio by ≤2%, i.e., UG would be more effective and less costly.
Second, the cost estimation of catheter-related complications was based on the recently used claims data from a hospital of maximum care (n = 55,000) which can be assumed to reflect the usual reimbursement practice of German hospitals. Using data from this hospital enabled us to estimate and include the costs of rare events  such as nerve injury or pneumothorax.
Third, the costs of treatments for complications were calculated with respect to regulation principles and specific features of the German DRG-System. More specifically, rules of coding, grouping and accounting for complications were taken into account [18, 23,24,25].
Finally, the target population of the model was in line with patients included in the Cochrane-review  where we took clinical data. In addition, the chosen time frame of the analysis is appropriate to reflect the clinical consequences of UG.
In contrast, modelling studies tend to exhibit weaknesses because of constraints of resources and information availability. In our analysis, to some extent, the use of claims data for estimating several cost parameters might lead erroneous results . For example, the measured catheter-related complications might be subject to up-coding, i.e. the severity or number of complications is increased to improve reimbursement . Therefore, it cannot be excluded that the costs calculated for catheter-related complications were overestimated to some degree. To avoid an overestimation of costs we validated our model with experts. In addition, for the sensitivity analyses a conservative range of ±50% was used for handling uncertainty . Supposed that these costs are still overestimated, the incremental cost-effectiveness ratio would be biased towards a cost saving impact of UG.
With regard to the input data of the model, there are some limitations resulting from the clinical evidence used for the model that may call our results into question. According to the authors of the Cochrane review  the internal validity of the included trials was insufficient due to several methodological deficiencies. In most of the studies, blinding of participants, operators and assessors were judged as unclear. Whilst personnel blinding was not possible due to the nature of intervention, a possible blinding of the outcome assessors and the patients would have reduced the risk of bias. In addition, in 80% of the included studies random sequence generation and allocation concealment was considered to be at moderate or high risk of bias.
Furthermore, some meta-analyses calculated in the Cochrane review were affected by study heterogeneity which may result from different sources. In particular, the meta-analyses for the success rates of the cannulation showed a substantial heterogeneity (> 50%). A qualitative examination conducted by the authors of the review revealed differences in the study populations, technical and methodological differences as potential reasons for this heterogeneity . Specifically, a successful central venous catheterization depends on the operator’s expertise in cannulation. In studies of the Cochrane review,  the ‘learning curve’ of the operators varied from high-expertise practitioners  to operators with limited experience . Because of the lack of information, we could not assess the impact of the practitioners’ expertise on the cost-effectiveness of UG.
The external validity of our study may be called into question because information about the number of excluded patients in clinical studies of UG and the reasons for their exclusion are sparse [7, 26, 32, 33]. Since in the base-case we had to rely on the inclusion criteria of the clinical studies included in the Cochrane review, the patient population in clinical practice might be at higher risk of complications. These complications might be affected by various patient characteristics such as anatomic variations, comorbidities or age. Anatomic variations such as the position or/and diameter (< 7 mm) of the IJV may complicate a central venous catheterization. For example, in 54% of the adults the IJV lies over the carotid artery  or, in 13–18% of individuals the IJV is about 5 mm . If we assume that in 54%  of the adults (> 40 years) the IJV lies over the carotid artery, then our model would only refer to approximately 22% (n = 124,000) of the patients catheterized per year. Additionally, different comorbidities such as coagulative disorders, uremia or obesity may make a cannulation more difficult [3, 7]. To assess the external validity of our study we undertook several sensitivity analyses. The base-case results did not change when a study population with higher incidence for complications was assumed in structural sensitivity analyses (i.e. UG was still more effective in averting complications and less costly). Moreover, the use of UG averted even more complications (i.e., increase of 200% by tripled complication rates) compared to the base-case results.
Similarly, central venous catheterization is a challenging procedure in the pediatric population . Various anatomic anomalies or a large range in the size of vessels may complicate a central cannulation compared to adults . For example, 18% (n = 50) of patients < 6 years had anomalous venous anatomy that may account for some difficulties in the catheterization . However, including both adults and children in the study population did not change the overall conclusions on clinical and economic benefits of UG.
Furthermore, in accordance with clinical studies, in our analysis central venous catheterization was assumed to be performed by the real-time ultrasonography and via the IJV. A scenario of patients catheterized by real-time and indirect technique did not alter the conclusions of the analysis and would even improve the results in favor of the UG.
In addition, the model assumed a central venous catheterization to be performed in intensive care or an operating theatre environment. It goes beyond the scope of this study to judge whether the conclusions derived are valid for emergency departments, where catheterizations might also be performed .
Since judging external validity is often more a clinical than a methodological expertise, future studies should be designed and reported in a way which allows clinicians to judge to whom they can reasonably be applied . Thus, we aimed to report the determinants of our model and of the evidence which it is based on to allow clinicians and other decision makers to reach their conclusions.
Despite some limitations which might affect the cost-effectiveness of UG, our results are in line with analyses [10, 11] from other health systems. In contrast to previous analyses, our analysis considers more catheter-related complications (e.g. pneumothorax, nerve injury) resulting in more averted complications compared to the results of other models. With respect to the expected costs of a central venous catheterization, our analysis confirms the cost saving potential of the UG in other studies. Nevertheless, there are some differences in the included costs, in particular the costs of UG intervention (e.g. UG machines).
Relevant costs to be included depend on the taken perspective , especially on the reimbursement procedures in the setting. Because the costs of UG intervention are not charged to the SHI, we did not include them in our analysis. Supposed that the SHI would reimburse the costs of the intervention, UG would be still a cost saving option up to €179 per catheterization (Additional file 6). This amount is approximately fivefold of the estimated costs which are borne by the hospital considered in this analysis. Since relevant costs and their estimation are subject to country specific variation , the variation of cost savings between different countries is unavoidable.
In addition, recent evidence indicates advantages of UG over chest x-ray because UG can be performed faster and does not subject patients to radiation . Including these costs may improve the cost-effectiveness result in favor of UG.