HOSPITAL
COSTS
PREDICTION

General information

- overview
- list of affections
- list of operations

Products

- hospital costs
- day surgery
- readmissions
- reoperations

Contact us

- contact us
- order form

Aim	To predict the costs of inpatients
Definition of the indicator	Cost = total amount of resources consumed divided by the number of inpatient discharges.
SQLape categories	Inpatients are allocated to one or several affections or operations. Each category is defined by inclusion and exclusion criteria, detailed in the manual “SQLape® - hospital costs prediction”. Inclusion criteria are based on the Swiss surgical intervention nomenclature (CHOP-XI, a Swiss adaptation of the ICD-9-CM) for operations and on the International classification of diseases (ICD-10) for affections, including optional additional codes (German adaptation of ICD-10). Four types of exclusion are applied, in order to reduce the redundancy of information : 1. if many interventions have been performed on the same organ, only the most invasive is taken into account; 2. a health problem (diagnosis) related to an operated organ is not taken into account; 3. some surgical or medical categories are excluded if they are associated with other specified categories 4. some surgical or medical categories are excluded if they are associated with specified combination of categories. Examples: A predominant operation (organ transplant or caesarean) removes all other operations or affections. In the absence of such predominant operations, predominant affections (leukaemia, premature newborns, obstetrical conditions remove also all other affections and operations. In other cases, all major affections – i.e. justifying hospitalisations, for instance: acute myocardial infarct or encephalitis – and all other operations are all retained. Non major affections are considered only in the absence of an other category. More examples to illustrate these hierarchical rules can be found in the manual “SQLape® - hospital costs prediction”. The SQLape® grouper allocates to each stay a principal category (indicated by the letter p in the grouper output), using the following order: predominant operations, predominant affections, major operations, major affections, minor affections (ou sont les minor operartions?). In addition, the most vital systems takes precedence (prematurity, respiratory, nervous, cardiac, hepatic, digestive, urinary, musculoskeletal systems, etc.). The attribution of a primary category in thus independent of the order in which health problems are coded in medical records.
Expected costs	The expected cost of an inpatient discharge is equal to the sum of the cost weights of SQLape® categories allocated to the stay. A correction is brought when a patient is transferred from or to an other hospital, provided that the two following conditions are met : 1) the length of stay is shorter than expected according to the case mix; 2) the patient is transferred in or from another acute care setting (transfer to a nursing home is not taken into account for instance). The expected cost weights of each inpatient stay are automatically computed by the grouper.
Option	Hospitals or financial allocation authorities may be interested to compute cost weights by themselves. Several methods can be applied, especially to identify the best practice hospitals and to exclude outliers. The optional “cost management tool” provides a simple approach to split the costs of hospital stays into several category affections or operations (cost control); this tool also provides specific cost weights, computed by the observed mean the data set after excluding outliers (greater or less than six time the observed mean, determined with four iterations). Technical support is also given to develop other cost weights computation rules.
Scientific validation	Cost weights have been computed with a random sample of Swiss hospital data (one University hospital and many public hospitals of various size). A cross validation has then be applied, to compare expected and observed values on another random sample (coefficient of determination R^2): SQLape® - without length of stay adjustment : R^2 = 0,55 - with length of stay adjustment : R^2 = 0,64 DRGs groupers - without length of stay adjustment : R^2 = 0,38 to 0,44 - with length of stay adjustment : R^2 = 0,52 to 0,56 For more details : 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.
Implementation	For more details about required data and material, format, software package implementation, click on the left button : Manuel technique (French) Technisches Handbuch (German)

Last update: July 9, 2009