Introduction

This concept describes the costing method developed for The Direct Medical Costs of Low Birth Weight Infants project. NOTE: For an overview of general costing methods (sources of data, types of costs, and approaches) and the methods of costing specific health services (hospitals, physicians, home care, personal care homes and pharmaceutical/prescription drugs) that have been used in MCHP research over time, please see the Costing Methods: An Overview of Costing Health Services in Manitoba concept.

The project The Direct Medical Costs of Low Birth Weight Infants used costing methodology to measure and compare the direct costs of pre-term to full-term newborns. The costs of both medical claims data and hospital separations were calculated. As we were comparing two birth cohorts, babies born in 1987 and 1993, over seven years of life each, all costs had to be converted from actual (current) costs into real (constant) dollars.

Medical Claims

The variable FEE in the medical claims file is the amount paid to (or drawn back from) the physician on behalf of the patient for the services rendered by the physician according to prescribed fee schedules. It is the original, unadjusted dollar value, which must be converted into constant dollars for comparisons using a price index.

How we calculated a price index:

We developed a price index for physician services in order to do the conversion to constant 1995 dollars. The year 1995 was used to be consistent with the hospital claims, since the cost list assigns prices at constant 1995 dollars. We calculated the mean price per procedure from MB Health data (using fiscal years), then used the proportion of total costs in 1995 for each procedure to calculate the weighted price for each (fiscal) year. Then, to convert current year billings for physician services to 1995 dollars, we divided the physician costs for each year by the index number for that year. So for each year of medical claims, the fee for a given claim is FEE95 = FEE/index where index is the price index for the year the claim was made.

Hospital Separations

In theory is not difficult to cost hospital separations because the cost list provides the cost associated with each RDRG®. However, the cost list is based on the version 7.0 RDRGs®. In our case the data spanned 1987-2000, so we did not have a consistent RDRG® grouper for all years. A uniform RDRG® grouper has to be applied to all years of data. This is an iterative process involving the addition and subtraction of the 5th digit (or code) to DX01 to make it groupable. Once this is done and the majority of separations are assigned version 7.0 RDRGs, it is possible to merge with the cost list by RDRG® and calculate the costs using the standard cost list methodology.

Notes:
1. We costed inpatient claims only. Outpatient claims were a small proportion of costs relative to the inpatient hospital costs. Because Day Procedure Groups (DPGs) were not available for all years of our study, we excluded outpatient procedures in the costing.

2. ER visits are difficult to cost. As such, we only considered ER nursing costs at $47.62/visit.

3. Some RDRGs® cannot be costed using the algorithm because they are missing average LOS, trim point, and marginal cost per day. In these cases we used the cost per case only, since no other information is available.

4. We did not distinguish between transfer/death separations and other separations. They were all costed using the same algorithm. The effect of transfers/deaths was assumed to be negligible.

5. Our analyses were not done using fiscal years but rather age of the subject. Information was summarized as follows: events occurring during initial hospitalization, subsequent age 0 hospitalization, age 1, age 2, … age 6. There was some concern regarding long stays over the first birthday inflating age 0 costs and deflating age 1 costs. For example, a baby admitted at birth and staying in hospital for 600 days would have the costs for all 600 days assigned to their initial visit, though part of their stay was incurred while they were one year of age. To more fairly distribute costs to the age at which the cost was incurred, long stays (> 30 days) over the subjects' first birthday were divided into age zero days (attributed to initial or subsequent hospitalization as appropriate) and age one days.

Here is some code for the hospital costing:


** keep only inpatient hospital claims **;
data inhospclm;
set hospital ;
if transact = '1';
run ;

** one RDRG not in cost list so convert and treat as the next class that is available (check for missings in data before do merge) **;
data inhospclm ;
set inhospclm ;
if rdrg = '4593' then rdrg = '4591' ;
run ;

proc sort data = inhospclm ;
by rdrg ;
run ;

** get cost list **;
data costlist ;
length rdrg $4. ;
set cost.rdrg ;
run ;

proc sort data = costlist ;
by rdrg ;
run ;

** merge cost list with hospital claims by rdrg **;
data inhospclm ;
merge inhospclm (in = a)
costlist (in = b) ;
by rdrg ;
if a and b ;
run ;


** cost the inpatient hospital stays **;
data inhospclm ;
set inhospclm ;

** method 2 does not use death/transfers in defn **;
** first line prevents missing values **;
if avglos = . then coststay = casecost ;

** continue with all others **;
else if los <= trim then coststay = casecost ;
else if los > trim then coststay = casecost + ((los-avglos)*margcost) ;
run ;