Introduction

This concept describes the development and use of a costing method using the Johns Hopkins Adjusted Clinical Groups® (ACG®) system and Relative Cost Weights. 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.

When there is a measure of resource consumption per person (e.g.: MD fees claimed), you can associate a value with each Adjusted Clinical Group® (ACG®) which represents its resource consumption in comparison to other ACGs. This value is the ACG relative weight (also called ACG 'resource intensity weight') and is typically calculated by dividing the mean expenditures for each ACG by the grand mean for all ACGs. MCHP used three measures of resource consumption:


1. physician,
   
2. hospital and
   
3. pharmaceutical expenditures

and we computed relative weights for various combinations of these costs.

Because of limitations in the datasets, outpatient pharmaceutical and emergency department expenditures were excluded. Cancer treatment expenditures incurred at the provincial cancer treatment centre were also excluded.

Physician Costing

Physician expenditures were based on fees submitted (billings) to the medical services plan. These billings represented approximately 90-98% of physician services. (Tataryn, Roos, Black (1994)).

Physicians in Manitoba who practice outside of the city of Winnipeg receive a premium above the standard (i.e. Winnipeg) fee for all services. At the time, the fee premium structure was:


• physicians in the city of Brandon receive an additional 2%
  
• physicians in the rural south of Manitoba receive a premium of +5%
  
• physicians in the rural north of Manitoba get 10% more than the standard fee

These premiums should be accounted for when computing physician costs in Manitoba. The MCHP physician claims datasets contain a FEECODE variable (with values 'B', 'R' or 'N') to be used to flag claims with fee premiums. Please refer to the SAS code (internal access only) for an example of how the physician costs can be adjusted for these premiums.

Hospital Costing

Hospital costs were estimated for ambulatory surgical procedures and inpatient hospital stays. Hospitals in Manitoba are funded by a global budget rather than for individual services. Consequently, we employed a 'top-down' case-mix costing methodology to estimate patient-specific hospital costs. This method starts with total hospital expenditures and then divides this value by a measure of total output. This methodology has been used in various research projects at MCHP (Shanahan et al., 1994, 1996 and 1997).

A case-mix costing approach goes further than per diem costing methods by dividing patients into groups that are clinically meaningful and homogeneous with respect to expected hospital expenditures. Patients are assigned higher weights if they are expected to consume more resources. Weights can be adjusted depending on whether the case is typical or atypical. This method of costing uses total hospital costs as the numerator (as is the case if one were to calculate unadjusted per diem rates), but rather than dividing by the number of patient days to find an "average" cost per day, it is divided by the sum of the case-mix weights to estimate the cost per weighted case (CWC).

It is important to note that:


1. the cost assigned to a case is for a complete course of treatment, and dividing the cost by the length of stay will not accurately reflect the costs of any particular day.
   
2. the cost for a case is an estimate of the average cost for that particular type of case, and may not accurately reflect the actual cost of a specific case.
   
3. the cost for a particular type of case is calculated as a value relative to all other types of cases.
   
4. weights were not developed from Manitoba cost data. Maryland Health Services Cost Review Commission (HSCRC) 1991 and 1992 data were used to calibrate Manitoba weights and costs.

We used the Refined Diagnosis Related Group (RDRG®) classification system to group inpatient cases into clinically meaningful resource use. (Canadian Institute for Health Information 1995) The RDRG® system allows for differing levels of severity based on complications and co-morbidities within similar diagnostic groupings. Relative case weights (RCWs) were developed based on charge data from Maryland and Manitoba lengths of stay (LOS) in hospital. Average LOS (ALOS) and trim point (the point after which a length of stay is determined to be abnormally long) for typical patients were also developed for Manitoba patients for each RDRG®.

For outpatient care, Day Procedure Groups (DPGs) and weights were available from the Canadian Institute for Health Information. (Canadian Institute for Health Information 1994)

Hospital Costing Details

Using 1995/96 patient hospital data, all inpatient hospital days were classified into RDRGs (version 9); each was weighted using Manitoba RCWs. Adjustments were made to the weights for atypical cases (cases involving non-acute days, outliers (LOS > Trim), transfers or deaths). For example, for cases with LOS > Trim, a marginal case weight was added to the [term=103489] Relative Case Weight (RCW) for every day that the LOS was past the ALOS (that is, case weight = RCW + Marginal case weight * (LOS - ALOS)). Marginal case weights were developed for each RDRG®.

For each hospital the case weights were summed. Hospital-specific average case weights were calculated by summing all the case weights in each hospital and dividing by the total number of hospital cases. Hospital average costs per weighted case (CWC) were calculated by dividing the total inpatient dollars by the total hospital case weights. The CWC for a hospital represents an average cost per case adjusted for the types of patients treated in that hospital.

CWC (per hospital) = Total dollars per hospital / Sum of all RCW (per hospital)

To find the cost of a particular case:

Cost of a case = CWC * RCW for that case

Day care surgery costs were estimated using the DPG to classify cases and apply appropriate weights. The DPG weight was then multiplied by the CWC for the hospital providing the care to obtain an estimated cost per case.

Estimated day surgery cost for a case = DPG weight * CWC for that hospital

For each patient, the inpatient and day procedure costs, if any, were combined.

Pharmaceutical Expenditures

Pharmaceutical expenditures were total drug costs (including professional dispensing fees) in the fiscal year.

For more information on pharmaceutical costs/expenditures, please see the Calculating Costs/Expenditures for Pharmaceuticals concept.

Source of Global Budgets

The primary source of financial data was the Statistics Canada HS-1 database, supplemented with data from other sources:


• Hospital Statistics Part 1 (HS-1) : Prior to 1995/96 all hospitals annually filed HS-1 data collection forms with Statistics Canada. The HS-1 consisted of hospital costs and statistics in an aggregate form.
  
• Financial Information Systems (FIS): used to provide audited and inventory-adjusted cost data for drugs and medical and surgical supplies for the rural hospitals
  
• Laboratory and Imaging Services (LIS): provides diagnostic services for many rural hospitals.
  
• Maryland Health Services Cost Review Commission (HSCRC) 1991 and 1992 data.

Comparison of Methods

Table 1 - ACG Relative Expenditure Data for Physician Services, presents ACG relative expenditure data for three different groups of costs. The first group contains the cost of physician services provided to the Manitoba population and data from a parallel study in British Columbia (BC) (Reid et al., 1998). In the BC study, ACG resource intensity weights were developed for a 5% stratified random sample of BC residents (n=171,157) who were continuously enrolled in the province's health plan in 1995-96. In contrast to physician expenditure data in Manitoba, the expenditures in BC excluded payments for laboratory and radiology services.

The second group contains expenditure data for physician and hospital services to Manitobans and data from patients in two different USA settings. The third group contains expenditure data for physicians (MD), hospital and pharmaceutical expenditures in Manitoba.

For these analyses, values for outliers, defined as individuals with expenditures greater than 3 standard deviations above the mean in each ACG, were trimmed (n = 14,542) (1.5%) for physician expenditures; 9,349 (1.0%) for total (physician + hospital expenditures). The validity of these weights was assessed by comparing them with similar weights developed in other Canadian and U.S.A. jurisdictions.

The Manitoba resource intensity weights appeared to have substantial face validity (see Table 1 for more information). Those ACG categories with the greatest morbidity burdens (e.g., ACG 5070 with 10+ Ambulatory Diagnostic Groups (ADG) combinations, age > 16 and 4+ major ADGs) had the highest weights and those with the lowest morbidity burdens (e.g., ACG 1600 - Preventive and Administrative) had the lowest weights. On the whole, ACGs with ten or more ADGs had the highest weights followed by ACGs with 6-9 ADGs, ACGs with 4-5 ADGs, and ACGs with 2-3 ADGs. For the majority of ACGs, there were relatively few differences between the Manitoba and BC weights. Forty-four ACG categories (54%) had absolute differences of less than 0.2 (range 0.1-6.0). The overall variation between ACG physician expenditures in Manitoba and BC were very similar. In Manitoba, there was a 52-fold variation between the least and most costly ACG compared to a 46-fold variation in the BC study, after excluding ACG 5200 (Non-users) and ACG 1600 (Preventive and Administrative).

The table also presents relative expenditure data for physician plus hospital services for Manitoba compared to similar data from two USA sources:


1. non-disabled enrollees in Minnesota's Medicaid program in 1995 (n=290,888) and
   
2. enrollees in a large U.S. staff-model Health Maintenance Organization (HMO) in 1994 (n=71,520).

For these expenditures, there appeared to be a similar high degree of variability across ACGs in the U.S. and Manitoba populations. In the HMO population, there was a more than 500-fold variation from the most to the least resource intensive ACG and in the Minnesota Medicaid population there was a more than 200-fold variation. Among Manitoba residents, the variation was just over 100-fold.

In general, the ranking of the resource intensity weights was similar across populations; those ACGs with the highest weights in Manitoba also had the highest weights in both U.S. enrolled populations. The weights for Manitoba appear to follow the HMO weights more closely than the Medicaid weights with small differences between the weights for most ACGs. Notable exceptions were the psychosocial ACGs (ACGs 1400, 1500, 2600) which appeared relatively less expensive in the HMO compared to Manitoba. This surprising finding was likely related to the fact that mental health services were 'carved-out' of the basic HMO benefit package.

Also, the weights for ACGs for relatively sick infants (i.e., those with major ADGs) were lower in Manitoba than in the HMO. This finding may have been related to the imprecise nature of the Manitoba infant physician and hospital expenditure data and/or less intensive care for very small infants. There appeared to be some large differences in the relative expenditures when Manitoba weights were compared to the Medicaid weights, especially in the most resource intense categories. The origin of these differences was unclear but may have been related to the fact that Medicaid populations face access problems for primary care services and rely more heavily on emergency room services.