Resource allocation at hospital level: the groundwork
G-DRG costing excludes the costs of teaching and research, as these have a wide range among hospitals and not all hospitals have teaching and research units. Accruals, most amortizations, private physician liquidation, capital costs, tax, insurance, and interest are still not part of the tariff calculation, although DRG relevant. This facilitates benchmarking, but contrasts a useful full cost approach and managerial relevance. Although the separation of capital costs (historically reimbursed separately
[
3]) increases national comparability, it does not respect different capital cost requirements resulting from the case-mix of a hospital. Thus, an inclusion in G-DRG costing, as currently developed, is desirable. As it does not seem applicable to completely reduce reimbursement to DRGs and keep the number of DRGs manageable, additional fees for highly specialized services in various degrees (e.g., for expensive drugs and treatment or intensive care) are suitable. This allows a fair reimbursement of highly specialized or expensive services. The exclusion of DRG-relevant costs in G-DRG costing limits efficiency, as reimbursement is then partially not based on actual costs. Transparency is limited, as the sources and magnitude of capital costs cannot be compared among hospitals. Systematic, software-based calculation inequalities have to be incorporated when reflecting a hospital’s DRG costs. Currently, unlike in Australia, hospitals are free to decide on accounting and hospital management software. A certification process such as that applied for the DRG grouping software does not exist for cost accounting programs referring to the InEK calculation
[
36]. Thus, transparency is further limited.
Resource allocation at hospital level: patient-level costing
Participation rates in the InEK calculation scheme are around 16%, as hospitals often do not have the cost accounting prerequisites to calculate at the patient-level
[
37]. Feeder systems in cost-centers such as radiology cannot provide the key cost drivers necessary (see Figure
3) to participate in the calculation. A future cost-modeling approach for non-calculating hospitals might support representativeness without downgrading the accounting standards or making participation in the calculation mandatory – which is hardly achievable because of the costly accounting prerequisites that participation in reimbursement calculation requires. Providing public service weights (as in Australia) for hospitals with less advanced accounting systems or accepting less advanced key cost drivers in some cost modules are a practical solutions to this issue
[
36]. Graded calculation methods, representing the technical accounting capabilities of a hospital, as in the English PbR system, are already well-proven
[
38]. Still, a high percentage of cost distribution is based on LOS, and especially for medical DRGs (conservative therapy), the fraction of directly case-related costs is low compared with operative DRGs. Although key cost drivers allocated to patients improved (see Figure
3), a sophisticated solution has not been implemented yet. The use of key cost drivers based on old and imprecise point systems, originally developed for physician reimbursement, results in imprecise capacity and resource planning (e.g., operating room or staff costs for DRGs). Economically sound decisions become limited. The use of the G-DRG system based on InEK calculation as a pricing system and not only as a budgeting instrument is critical against this background
[
8]. To improve costing at the patient-level, Kaplan & Porter (2011) introduced time-driven activity-based costing (TDABC) in the hospital
[
37,
39]. This allocation method defines nearly all costs as variable and includes time as a key cost driver in a more detailed implementation, such as duration of clinician visits as an example of direct costs or the duration of laboratory tests as an example of indirect costs. In some cost modules, such as physicians in the operating room, TDABC is already implemented in the InEK costing standard. The more detailed the calculations are, the more precisely and efficiently management can act; and management decisions become transparent.
Overall, transparency and efficiency in cost accounting have improved since the introduction of the InEK costing scheme. The calculation manual for the calculation of case costs is in its third version now and has increased costing standards greatly since its introduction. Its implementation is enforced by improved and more rigorous plausibility checks. At the hospital level, the learning effect, paired with developing cost accounting software and documentation requirements, has kept up with increased calculation requirements. After an adaptation phase, the coding of diagnoses and operations and procedures has remained relatively stable
[
13]. Hospitals try to adapt treatments to the DRG-system (reimbursement) by implementing clinical pathways, also oriented on the InEK calculation scheme
[
40], promising higher efficiency
[
41] and increased transparency through better documentation of the course of action.
Tariff calculation at national level: plausibility checks
The claimed advanced cost accounting methods at hospital level contain risks for tariff calculation. The control for tight band-width in cost modules and further plausibility checks can lead to the exclusion of cases from the calculation, even if costs are calculated correctly. If the costs of a case in a single cost module are not in a target corridor based on previous years, the case might be excluded
[
6]. As these corridors are not published, the actual influence and control mechanism of band-widths cannot be defined. Future research is needed to analyze bias caused by band-width control. Too closely meshed plausibility checks might lead not only to data quality improvements, but also to an undesirable trimming of the calculation on InEK plausibility checks. To achieve high data quality, plausibility checks are both a blessing and a curse. On the one hand, they enforce improved cost accounting, resulting in improved data quality. On the other hand, they might push hospitals into default band-widths, possibly not representing the hospitals’ actual cost structure. To improve participation rates, fees for every case passing plausibility checks were introduced, promoting the focus on plausibility checks and questioning the “one hospital” approach, as further bias is introduced. The efficiency goal of DRG introduction is therefore slightly transgressed, although plausibility checks contribute to efficiency overall. Transparency is improved, as calculation errors are reported on a patient basis in every round. However, the background on band-width calculation should be reported better and the kind of calculation errors should be made public (e.g., the top 100 calculation errors) to let hospitals benefit from early avoidance of these calculation errors in future calculations. To further reduce the rate of cancelled data, an automated system should be introduced for reporting error explanations on a case basis, enabling the InEK to produce statistics from the hospitals’ point of view on specialized calculation errors.
Tariff calculation at national level: inlier calculation
That the calculation of length of stay (LOS) thresholds is highly important concerning incentives for providers has been shown in detail
[
42], also implying that coding issues can be reactions to inlier calculation
[
43]. Still, normatively derived upper and lower LOS thresholds imply systematic failures and possibly underfinancing
[
44,
45]. The deductions from reimbursement rates due to short stay are not calculated based on cost accounting data
[
44,
45]. They follow a “main effort concept,” including deductions for services not part of the main effort
[
45]. Therefore, other countries such as England do not use lower trim points
[
46]. Upper trim points were introduced to lower the risk of the hospital in complicated cases; however, they also suffer from their normative derivation. A non-normative costing-based calculation as in the U.S. (“cost outliers”
[
47]) or a specific case-mix for each care day (Victoria, Australia
[
48]) might be a solution for upper and lower trim points, also embracing rare DRGs. Otherwise, supererogation can lead to a reduced effective reimbursement rate
[
44]. This partial “system failure” can have dramatic costing impacts in some cases, as outliers are not in the minority. The 2011 G-DRG system had an average of 22.3% outliers (standard deviation 12.7%, ranging from 0% to 83.2% outliers within DRGs)
[
49]. Inadequate trim point calculation opens up the discussion of a greater DRG differentiation, implying unintended single case reimbursement and less practicability in the grade of differentiation
[
13]. That a higher grade of differentiation might not improve welfare has already been shown in econometric models
[
50]. Still, consistent outliers can be a sign of the need for further DRG differentiation. Expanding additive components in the DRG calculation, as currently exercised, or outlier-calculation based on costs might partly resolve this issue and increase economic homogeneity
[
32]. As the generation of outliers is a necessity to reduce the risk for providers and to create medically and economically homogeneous groups (
R
2
is about 0.1 lower for all cases compared with inliers; see Figure
4), a focus should be set on the influence of the accounting system to define outliers. Further, the explained part of the variance is much higher when referring to costs compared with LOS (normative derivation)
[
34]. By using the different modules in the InEK matrix in combination with the date of cost occurrence for outlier calculation, the less accurate normative derivation could be replaced. Thus, the current system has high transparency through normative derivation, but serious flaws concerning the efficiency of the calculation.
Tariff calculation at national level: the "one hospital" approach
The “one hospital” approach causes a bias in costs (regional price index, pay scale, etc.), which can be resolved only partly by the negotiated base rate in every state concerning reimbursement. The InEK only adjusts according to wage index/union rates in the regions of former East Germany
[
6]. There is no adjustment in the reimbursement for geographical variations in case-costs – an obvious disadvantage for high-cost regions. Besides, owing to the different composition of DRG costs (e.g., labor costs, material costs), DRGs are affected by this non-adjustment to different degrees, which might result in a preference for less labor-intensive DRGs in high wage index areas and vice versa. For example average household income in the 16 German states (Länder) varied between € 4,253 and € 2,617 in 2008
[
51]. Although a unique base rate calculation and few regional adaptations support competition, they might contradict the care mandate of every German hospital and undermine the security of full health care supply in every region. The convergence phase of the base rate resulting in a future nationwide base rate shows that the “one hospital” approach, supporting competition, is the chosen route.
Further, case weights are always 2 years old when published. The G-DRG reimbursement of 2012 was calculated with the 2011 G-DRG scheme, based on data collected in 2010. The out-of-date issue affects only relative cost-data (case-mix), as base rates are negotiated for every year in every state. But the quality of tariff calculation suffers from out-of-date relative cost-data and especially from insufficient regional cost adaptation. For management implications, non-adjustment in the “one hospital” model is resolved by putting a higher relevance on the relative cost matrix (the case-mix of a case distributed to the cost matrix), which has to be multiplied by a base rate to get the actual cost. For internal management decisions the base rate can then be adapted to the question that has to be answered. For example, the actual costs published by the InEK are not used as a reference for a hospital’s cases, but the relative costs calculated by dividing the InEK cost matrix for a DRG by the allocation base (the InEK calculation base rate, see “Tariff calculation at national level; the “one hospital” approach”).
Another negative aspect of the “one hospital” model is the unadjusted bias that can be induced by the voluntary participation of hospitals, as the choice of hospitals to take part in the calculation can have many different incentives. In the most favorable case, their motivation to take part is image, the fee for every calculated case, or the wish to compare themselves with a nationwide benchmark, and thus the use of the calculation for internal management decisions. In the worst case, the hospital already uses the InEK cost accounting scheme or equivalent systems for internal strategic management decisions and decides on whether the participation might affect its own future reimbursement positively or negatively. This incentive is especially strong when a hospital knows that it delivers a high percentage of overall cases for the calculation of a DRG, or for hospital chains, where the calculation of one hospital affects the reimbursement of others. Hospitals that are already very efficient have a low incentive to reduce their future reimbursement by delivering beneficial cases. They have the overall vicious circle nature of the system in mind when deciding about participation.
The fact that structure, ownership, and size of the calculating hospitals in general do not reflect the German hospital market is the final problem of the current “one hospital” approach
[
11]. There is an overrepresentation of medium and large hospitals, as small hospitals are possibly not able to achieve the costly, IT and accounting standards required. Concerning ownership, a biased sample can further affect costs, as the incentive to be efficient depends on ownership structure (private for-profit, private non-profit, public). Most of the literature concludes that private for-profit and private non-profit hospitals are less cost efficient than publicly owned hospitals
[
52‐
54]. However, the fraction of actual calculation hospitals (ca. 50% private non-profit, 10% private for-profit, 40% public) does not correspond with the fraction of potential calculation hospitals (ca. 42% private non-profit, 25% private for-profit, 33% public) concerning ownership in recent years
[
55]. Although the fraction of publicly owned calculation hospitals has stayed the same since the introduction of the G-DRG system, the private non-profit fraction has increased by ca. 6% and the private for-profit fraction has decreased by that amount
[
22,
25‐
30,
55]. Hospitals that have an incentive to improve efficiency also have an incentive to participate in G-DRG calculation, as this is the only generalized system to support management with cost accounting. As a result, the calculated costs tend to be higher than the true German mean.
To compare the profitability of calculating hospitals and non-calculating hospitals in further research might verify the impact of DRG calculation on management quality and system dynamics best, although the mentioned limitations concerning incentives, structure, ownership, and size have to be borne in mind and rethought by policymakers. Limitations in representativeness and non-adjustment affect benchmarking and strategic reactions on reimbursement rates concerning the elective case-mix. They increase the insecurity of hospital management on published DRG costs. Hospitals react to changes in the case fees catalog by changing their elective case-mix
[
48] – a regionally shaped DRG supply situation can develop, and a vicious circle is initiated concerning the motivation for participation. In contrast, the U.S. has a system that adjusts reimbursement rates for urban and rural areas, for a disproportionate share of poor patients, for regional wage levels, and for teaching
[
56,
57]. The English Payment by Results (PbR) -system uses market force factors (MFF) to adjust to the regional cost situation
[
16,
58]. The adoption of adjustment mechanisms as in the PbR system or the U.S. system would be beneficial for the G-DRG scheme concerning incentives to participate; however, it would lead to overall efficiency losses on account of less competition.
The G-DRG cost accounting system helps to organize the elective DRG-portfolio. Only departments/DRGs with a positive perspective are established or developed; the system protects from misdirected investments, but also favors DRGs with a high yield. The comparability and reproducibility resulting from a standardized G-DRG tariff calculation scheme are of great interest. Transparency and efficiency of tariff calculation are seriously transgressed by the non-representative calculation sample and the motivation to participate. One option to adjust for non-representativeness in the long run is to make the participation of hospitals in InEK cost accounting compulsory. Without reducing the quality of cost accounting by forcing hospitals with less advanced costing abilities to participate, the most important step is to introduce a score that is assigned to the cost modules in the matrix, representing the quality of the allocation methodology for that cost module. The English patient-level information and costing system (PLICS) uses such scores to allow for a high participation rate
[
59]. To motivate hospitals to reach a high score, the current case fees for cases calculated correctly could depend on score size. Hospitals with less advanced methods can also use key cost driver statistics from hospitals using advanced calculations (official relative value units) to distribute their costs on cases. For an overview of the impact of cost accounting modalities on transparency and efficiency, see Table
2.
Table 2
Assessing the G-DRG cost accounting scheme
Resource allocation at hospital level |
The groundwork | Medium standard/improvements necessary | High standard/small improvements possible | Inclusion of all DRG-relevant costs |
Cost-center accounting | High standard | High standard | - |
Patient-level costing | Medium standard/improvements necessary | High standard/small improvements possible | Improving key cost drivers and further introduction of TDABC |
Tariff calculation at national level |
Plausibility checks | High standard/small improvements possible | High standard/small improvements possible | Improving transparency on reasons for calculation errors |
Inlier calculation | Low standard/improvements necessary | High standard | Combining normative derivation with the cost outlier concept |
The “one hospital” approach | Low standard/improvements necessary | Low standard/improvements necessary | Increasing participation by a lower costing standard parallel to the currentstandard, to reduce participation bias |