Alternative Hospital-Wide Cost Metrics

How Do We Measure Hospital Costliness?

Because the majority of a hospital’s revenue is fixed by fee schedule or related to billed charges almost every hospital in the US is vitally concerned with their relative cost position. Quite naturally, hospital executives are interested in a facility wide metric of cost that can be benchmarked with their prior cost position and more importantly against other peer hospitals. For comparison with other hospitals, the use of publicly available data is required unless there is a shared data arrangement in existence. Hospitals located within a defined market area are unlikely to share data with competitor hospitals so public data becomes a necessity.

In general there are three hospital wide cost metrics that have been cited in the literature and used by hospitals:

  • Cost per adjusted discharge – adjusted for case-mix and wage index (CAD)
  • Cost per adjusted patient day – adjusted for wage index (CPD)
  • Hospital Cost IndexTM (HCI)

Cost per adjusted discharge and cost per adjusted patient day have been used for many years in the hospital industry. They were originally developed in an era when 80% to 90% of a hospital’s total revenue came from acute inpatient services. To account for outpatient activity, total discharges or patient days were restated to reflect outpatient volume. For example, a hospital with $100 million of inpatient revenue and 10,000 discharges that also had $20 million of outpatient revenue would restate its total adjusted discharges to 12,000. This adjusted discharge figure of 12,000 would then be divided into total facility cost to create a cost per adjusted discharge. The same methodology would be applied to define the number of adjusted patient days.

In the days when outpatient business was 10% of total revenue, few problems resulted from using an adjusted discharge or adjusted patient day method. However, the average US hospital now has close to 50% of its total business in outpatient areas, and the percentage is growing. While the growing percentage of outpatient revenue is clearly the primary factor that compromises the utility of a cost per adjusted discharge or cost per adjusted patient day methodology, it is not the only problem. Most of these adjusted measures also try to adjust for case-mix complexity through utilization of a case-mix factor. In the absence of an outpatient case-mix measure, most case-mix adjusted measured use an inpatient DRG case-mix index. Because the relative case-mix complexity of inpatient and outpatient services is not uniform, another potential bias is introduced. For example, a hospital with a major cardiac program may have a relatively high inpatient case-mix score, but its outpatient services may be similar to other hospitals. Adjusting its cost per adjusted discharge measure with its high inpatient case mix would make the hospital appear more cost efficient than it really was.

The HCI measure is a cost metric that is derived from three public-use file datasets: Medicare Cost Reports, Medicare Inpatient Claims -Medpar, and Medicare Outpatient Claims-Hospital Outpatient Prospective Payment System (HOPPS). Two basic measures of cost are used to construct the HCI:

1. Medicare Cost per Discharge – Case Mix and Wage Index Adjusted (MCPD)

2. Medicare Cost per Outpatient Visit – Relative Value Unit and Wage Index Adjusted (MCPV)

The HCI is constructed as follows:



MCPD is a good reflection of inpatient cost. Data for computing this measure can be derived from the public-use files: Medpar and Medicare Cost Reports. This process results in a unique publicly available number for most hospitals in the US. The MCPV is used to assess costliness on the outpatient side of hospital operations. We can construct this measure from public-use files – Medicare Outpatient Claims and Medicare Cost Reports – which makes its availability for most US hospitals a reality. To derive the measure, we divide the cost per visit defined through the Departmental Ratio of Cost to Charges (DRCC) extensions by the relative value units (RVU) of the claim. The RVUs for each claim are based upon Medicare Ambulatory Patient Classification (APC) weights or Medicare defined fee payments divided by the hospital’s APC conversion factor.

The real question that still needs to be answered is which measure is a better predictor of actual hospital costliness. In an earlier study of over 2,000 non-teaching hospitals, we examined the correlations between these three cost measures and the hospital’s Medicare margin. A hospital’s margin on Medicare inpatient and outpatient business should provide a surrogate measure of hospital relative cost. Medicare’s payment to acute-care hospitals is prospectively based for both inpatient and outpatient services for most acute-care hospitals so the primary variable impacting profit on Medicare patient s would be cost. We have excluded from our sample of acute-care hospitals all teaching hospitals and all other hospitals not paid on a prospective basis, e.g., critical-access hospitals. Because Medicare payments are fixed, hospitals with higher margins (defined as Medicare payments less Medicare assigned costs divided by Medicare payments) should have more efficient cost structures. . We also excluded disproportional share (DSH) payments. DSH payments are provided because of payer mix and should not be related to actual costs of hospital operations on a unit basis.

To evaluate the relative accuracy of the three cost measures, we performed a number of statistical tests. The HCI had a correlation with Medicare margin that was 67% better than CPD and 48% better than CAD. While all three measures were significantly correlated with Medicare margins, the HCI exhibited a more positive association. The HCI should produce fewer false signals of relative cost position and would be a better predictor of a hospital’s relative cost position.

If you have any questions, or would like to know more, let us know!

Comments (0)

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>