Examining case mix is not new. Other models for case-mix classification have been in use for some time. One of the most commonly used is Diagnosis Related Groups (DRGs) developed at Yale University. In Canada, Case Mix Groups (CMGs) was developed in 1983 by the Canadian Institute for Health Information (CIHI) and later refined. Both systems use weighting, a measure that compensates for the different levels of treatment required for different conditions. In other words, a condition with a case-weight of 2.0 should be twice as expensive to treat as an average condition (1.0 case-weight). So if a hospital treated five patients with case-weighted conditions of 2.0, and their total cost was twice that of another hospital's which treated five average patients, both hospitals would rank evenly in cost efficiency, all other things being equal.
But that's just the problem; things are often not equal. Both systems have been criticized for failing to distinguish grouped cases according to severity. That is, two patients in different hospitals may be in the same case-mix classification-such as coronary bypass-but one patient may have complications, or have other illnesses at the same time (co-morbidities), which would require alternate treatment and consequently higher cost.
The response to this by researchers at Yale was the Refined Diagnostic Related Groups (RDRGs). This system took the original DRG classifications and subdivided them, based on co-morbidities or complications, into different levels of severity. Instead of 500 categories of patients, RDRGs now have nearly 1200.
The Manitoba Method
Hospital Case Mix Costing incorporates RDRG's useful and relevant differences in severity of illnesses. However, no weights like those for CMGs or DRGs were available for RDRGs. We wanted to develop weights based on up-to-date costs.
This was not easy. There is a scarcity of detailed cost information in Manitoba, or in the rest of Canada for that matter. That's because Canadian hospitals operate on a "global budget"-a sum of money based primarily on a hospital's size and the particular services (such as open heart surgery) each provide. This is quite unlike the United States where hospitals bill for the care of each patient. This also means they keep track of every bandage, every suture, every little expense (known as micro cost data) to justify what they charge.
So out of necessity, we went south of the border for this micro-cost data. What we were interested in was not the actual costs so much as the differences in costs between cases. Put another way, the expense for a particular surgery may not have been the same in Canada as it was in the U.S., but if a procedure cost 25% above average in the States, it likely cost 25% above average here. We chose the Maryland database for several reasons; for one, it is very large, but most notably, it included data on all hospitalizations, not just on the elderly and poor as is the case in many states.
Essentially, then, we combined our estimated hospitalization costs with information on the mix of cases at each of Manitoba's acute care hospitals. We used 1991/92 hospital data because that period in Manitoba was relatively free from the major health care changes of subsequent years. Adjustments were also made for other factors-very long lengths of stay, non-acute (not typical) care cases, deaths and transfers-each of which were expected to affect costs.
Our intent was to develop a sound, up-to-date method for assessing the relative efficiency of Manitoba hospitals that could be applied to subsequent years of data.