INPATIENT COSTS

Aim: Analyze the cost of inpatient stays. Stays are defined by the interval between the admission of the patients and their discharges from acute somatic care hospitals. If different stays are merged for reimbursement purpose (SwissDRGs rules), each stay is taken into consideration in its original form (stay definition).

Numerator: The total amount of resources consumed (without investment costs) is computed from detailed cost data. These costs are then allocated according to the nature of the charges, determined from the accounting nomenclature (see table below), so that the sum of the expected partial costs is equal to the expected overall costs.

Denominator: All discharges in acute somatic care units. Stays assigned to Swiss statistics codes M500, M900, M950, M990 (only if the average length of stay exceed 10 days) are excluded. Stays with a zero length of stay are also excluded (only among stays grouped according to SwissDRGs rules).

Expected values: The expected costs per illness and per procedure are calculated based on costs observed in Switzerland during the years 2021 to 2023, excluding chronic care hospitals (average length of stay > 30 days) and those with a daily cost exceeding CHF 10,000. Expected costs are calculated excluding extreme cases. Since 2023, a new method allowing to distribute the costs among the different diseases (and interventions) of the patients was introduced. This a a big step forward, since it provides analyses much more clinically interpretable by doctors and a precise estimation of additional costs explained by iatrogenic complications. This was made possible thanks to the new Iterative Proportion Repartition (IPR) method. This new approach has beeen scientifically validated, with promising results:
– easy to compute, non-parametric (few assumptions)
– outperforming other current methods (linear regression and general linear models)
– providing a proportional repartition of non-additive costs due to interactions among multimorbid conditions.

We explained it in detail in the following text:

Eggli Y. Invitation à découvrir un nouveau monde (post-DRGs). Chardonne, SQLape s.à.r.l., 2023.

Cost of living varies significantly by canton; for example, it is higher in Geneva than in the Jura. The expected costs have been adjusted according to the average salaries* of the cantons grouped according to the following regions:

• Cantons with high salaries: BS, BL, GE, VD, ZG, ZH: 1.11
• Cantons with mountains: GL, GR, NW, OW, SZ, UR, VS: 0.90
• Other cantons: AG (1.02) ; LU (0.99) ; TG (0.96) ; SO (0.95) ; SH (0.94) ; AR, FR, and SG: 0.92, BE (0.89), TI (0.88), JU and NE (0.85).

To remove this adjustment, simply divide the expected costs by the wage index. The canton is determined based on the canton of residence of the majority of the hospital’s patients.

*This salary index correlated much better with observed/expected costs ratio by hospital (0.59 correlation), than other measures of cantonal wealth, such as gross domestic product, rentals or the amounts of inter-cantonal equalizations for example.

Interpretation: Expected costs are computed based on average costs per diagnosis and procedure category among SwissDRG dataset of the years 2021-2023.

Caption of cost analysis tables.

Output files: Results are given by hospital and site in Cost.xlsx file. The results are detailed by nature of costs (medical, nursing, hostellery, imaging, etc.) in two output files:
– NATURE-GLOBAL.xlsx, globally;
– NATURE-DETAIL.xlsx, for all diseases and interventions encompassing at least 15 cases.

These results are consistent with the overall cost analysis results provided by the standard SQLape tool, after excluding outliers. Slight discrepancies are possible, due to the exclusion of extreme cases per disease and per procedure, which may differ depending on whether costs are allocated globally or partially (by type of expense). It is also important to note that costs are calculated here globally per hospital and not per site as in standard tools.

Costs are detailed by hospital stay in Analysis.txt. General information about SQLape output files can be found here.

Strength of the indicator: In comparisons with DRG groupers, the results obtained with SQLape® are more robust to coding habits. In particular, they do not depend on the choice of the code considered as the main diagnosis. The grouper also better takes into account complex patients with multiple severe co-morbidities or interventions. Several hospitals prefer using SQLape® grouper to provide costs per patients’ categories which are clinically relevant to physicians. Although the patients’ classification is based on more than 900 groups to adjust for case mix accurately, results can be summarized in about 200 medical groups and 180 surgical groups easy to interpret.

Another advantage is that SQLape® does not over-adjust for complications. Indeed these conditions (including diagnoses usually associated to complications or immediate causes of death) are not taken into account in the computation of expected values. As a results, while DRGs would fail to detect over-costs due to complications, SQLape® would screened these costs.

Limitations: Some cantons have many rehabilitation or acute geriatric beds (up to 30% of acute somatic care beds), other very few. This might cause distortions in inter-regional comparisons. A solution might consist in aggregating acute and rehabilitation stays.

Scientific validation: Marazzi A, Gardiol L, Duong HD. New approaches to reimbursement schemes based on DRGs and their comparison. Health Services Management Research 2007; 20(3):203-210.

Rousson V, Rossel JB, Eggli Y. Estimating health cost repartition among diseases in the presence of multimorbidity. Health Serv Res Manag Epidemiol 2019;6;2333392819891005.

Rossel JB, Rousson V, Eggli Y. A comparison of statistical methods for allocationg disease costs in the presence of interactions. Stat Med 2021;40(14):3286-3298.