How difficult is TISSNET

Comparison of the cost relevance in the DRG system of the Apache II, SAPS II, TISS-28 scores and the IMKB to develop an optimized score for predicting the costs of cardiac surgery cases

The cost pressure in hospitals is steadily increasing. This leads to the desire of the hospitals to know the costs of the treatment cases at an early stage in order to better distribute the available financial resources economically and medically if necessary. It is difficult to determine the treatment costs of patients in a timely manner. The gold standard in Germany is the InEK cost calculation, according to which a hospital determines the costs for treatment cases according to a Germany-wide procedure. However, this is a retrospective method that is used to calculate the previous year's costs. This method is therefore not suitable for recognizing short-term cost developments and being able to react to them. Thus, several questions arose for this work: Can costs be predicted using medical scores? Which medical scores should be included in the study? Can the selected scores be calculated retrospectively with high data quality? Which of the scores is best suited for cost forecasting? Is it possible to make a high-quality forecast of the costs? Is it possible to reduce the score to a few items in order to minimize the work involved in the calculation? 12,733 treatment cases were selected, each of which had a CABG, a heart valve operation or a combination of both procedures. For the Business Intelligence (BI) software Qlikview, programs were developed that retrospectively from the medical data for the years 2008-2013 and 2014 for all treatment cases using four medical scoring systems (SAPS II, APACHE II, TISS-28 and the IMKB) and in 2015 automatically calculated the points. The calculation of the scores from the medical data was possible without any problems. The TISS-28 was selected through correlation analyzes because of the four selected scores, its results showed the highest correlation to costs. Two models were tested to predict costs. It turned out that a model that uses the entire TISS-28 to predict total costs was the most suitable. With the data from the years 2008 to 2013, the years 2014 and 2015 were forecast. Since two different medical directors were responsible for the treatment process in these two years, it could also be proven at the same time that the TISS-28 was robust enough in its cost forecast to reliably forecast the costs even with different treatment strategies. The number of items in the TISS-28 was not reduced in order to simplify the calculation of the score, as all items showed a significant correlation to the costs and the reduction might create a source of error that could not be foreseen.


The immense pressure of hospital costs continues to increase. This leads hospitals to want to be able to calculate the case treatment costs early on, so as to distribute the available finances in an economic and medically sound fashion. It is difficult to estimate the treatment costs of patients in advance. The gold standard in Germany is the InEK cost calculation, which enables hospitals to estimate the costs per case in accordance with a nationally defined procedure. However, this procedure is based retrospectively on the costs from the previous year. Using it, it is therefore not possible to recognize and react to short-term cost developments. For this reason this work looks at several related questions: Are medical scores able to predict costs? Which medical scores should be investigated? Is it possible to calculate the selected scores retrospectively with high data quality? Which of the scores is most suited to cost prognosis? Is high-quality projection of the costs possible? Can the score be reduced to a small number of items to simplify the calculation? A total of 12,733 treatment cases were selected in which coronary artery bypass grafting (CABG) or a heart valve operation or a combination of the two was performed. Programs were developed for the BI software Qlikview that were used to automatically calculate the points for four medical scoring systems (SAPS II, APACHE II, TISS-28 and IMKB) retrospectively from the medical data for the years 2008, 2013, 2014 and 2015. Calculation of the scores from the medical data was possible without problems. Using correlation analyzes the TISS-28 was selected since, of the four scores, it showed the highest correlation with the costs. Two models were tested to predict the costs; a model using the whole TISS-28 to predict the total costs was found to be the more suitable. Therefore the data from the years 2008 2013 were used for predictions for the years 2014 and 2015. In these two years the medical directorship of the institution changed, so that at the same time it was established that the TISS-28 was robust enough in its cost projection to correctly predict the costs even under different treatment strategies. The number of items of the TISS-28 was not, after all, reduced, since all items showed a significant correlation to the costs and reducing them might have introduced an unforeseeable source of error.