# Benchmark Data Details

## Overview

Thomson Health Care’s methodologies based on the All Patient Refined Diagnosis Related Groups (APR-DRG)- allow us to produce severity-adjusted performance comparisons on length of stay (LOS), hospital charges, and mortality between or across virtually any arbitrary subgroups of inpatients. These patient groupings can be based on DRGs, hospitals, product lines, geographic regions, physicians, and so forth. In order to adjust for differences in diagnosis type and illness severity across the different reporting groups, one can calculate a severity-adjusted value based on the assigned APR-DRG. The severity-adjusted values based on the APR-DRG methodology measure the clinical need for acute hospital resources with respect to LOS and charges, and an “expected” mortality rate relating to the group’s aggregate severity level.

### Normalized LOS and Charge Weights, and Mortality Rates

Thomson Health Care’s Projected Inpatient Database (PIDB) was used to develop the all-payer pediatric LOS and Charge weights, and Mortality rates based on the APR-DRG. **The PIDB is a nationally representative database** containing more than 20 million inpatient discharges annually which are statistically weighted to represent the universe of all short-term, general, non-Federal hospitals in the U.S. Each record is assigned a weight or “projection factor” which indicates the number of discharges it represents in the universe. In order to account for geographic cost-of-living differences, charges are adjusted for each hospital using the CMS wage index. APR-DRG weights are calculated by dividing the average charge (wage-adjusted) or LOS for each APR-DRG by the overall average charge (wage-adjusted) or LOS for patients in the universe of short-term, general, non-Federal hospitals in the U.S. The mortality rates are calculated within each APR-DRG and its mortality indicator using the same normative data. For the development of the pediatric weights and rates for the APR-DRG, data were excluded with ages greater than or equal to 18 years.

## Calculating Severity-Adjusted LOS, Charges, and Mortality

Using the normalized weights and rates, one can calculate severity-adjusted values that can be used in “fair” comparisons. Each variable (LOS, Charges, and Mortality) has its own calculation for creating a severity-adjusted value. One can calculate the severity-adjusted values on the discharge level for all records and then choose different sets of patients to generate the severity-adjusted values on a group level. In addition, group level severity-adjusted values can be generated from group level data if it is more convenient to use than the discharge level severity-adjusted values.

### For Severity-Adjusted LOS:

#### Discharge level

- Assign the APR-DRG and APR-DRG LOS Weight for each discharge
- Severity-Adjusted LOS = (Actual LOS) / (APR-DRG LOS Weight)
- Using the SA-LOS, one can calculate the Average SA-LOS on a group level

#### Group level (slightly different)

- Average APR-DRG LOS Weight for the group of patients is required
- Average Actual LOS for the group of patients is required
- Severity-Adjusted Average LOS = (Avg. LOS) / (Avg. LOS Weight)

### For Severity-Adjusted Charge:

#### Discharge level

- Assign the APR-DRG and APR-DRG Charge Weight for each discharge
- Severity-Adjusted Charge = (Actual Charge) / (APR-DRG Charge Weight)
- Using the SA-Charge, one can calculate the Average SA-Charge on a group level

#### Group level (slightly different)

- Average APR-DRG Charge Weight for the group of patients is required
- Average Actual Charge for the group of patients is required
- Severity-Adjusted Average Charge = (Avg. Charge) / (Avg. Charge Weight)

### For Severity-Adjusted Mortality:

#### Discharge level

- Assign the APR-DRG and the Mortality Rate for the specific APR-DRG and mortality indicator for each discharge
- Average Severity-Adjusted Mortality Rate is simply the average of the rates on the discharge records

## Using Severity-Adjusted LOS, Charges, and Mortality

Using severity-adjusted values for comparisons is straightforward. There are some basic ways to use these values. For the LOS and Charge values, if the severity-adjusted value increases for a group of patients with respect to its actual value this indicates the facility is treating more severe patients within that group. The same can be determined with the charges. This can also be determined by simply looking at the average LOS (or Charge) weight for the group of patients. If the average weight is greater than 1.0 this indicates that the group of patients is more severe with respect to resource utilization. For mortality, one can simply compare the actual mortality rate with severity-adjusted mortality rate and make similar inferences.

## Calculating Severity-Adjusted Expected LOS, Charges, and Mortality

Using the normalized weights and rates, one can calculate severity-adjusted expected values that can be used in “fair” comparisons within a facility or across facilities. Each variable (LOS, Charges, and Mortality) has its own calculation for creating a severity-adjusted expected value. One can calculate the severity-adjusted expected values on the discharge level for all records and then choose different sets of patients to generate the severity-adjusted expected values on a group level. In addition, group level severity-adjusted expected values can be generated from group level data if it is more convenient to use than the discharge level severity-adjusted expected values.

### For Severity-Adjusted Expected LOS:

#### Discharge level

- Assign the APR-DRG and APR-DRG LOS Weight for each discharge
- Severity-Adjusted Expected LOS = (National Average LOS) x (APR-DRG LOS Weight)
- Using the SA-ELOS, one can calculate the Average SA-ELOS on a group level

#### Group level (slightly different)

- Average APR-DRG LOS Weight for the group of patients is required
- Severity-Adjusted Expected Average LOS = (National Average LOS) x (Avg. LOS Weight)

### For Severity-Adjusted Expected Charge:

#### Discharge level

- Assign the APR-DRG and APR-DRG Charge Weight for each discharge
- Severity-Adjusted Charge = (National Average Charge) x (APR-DRG Charge Weight)
- Using the SA-ECharge, one can calculate the Average SA-ECharge on a group level

#### Group level (slightly different)

- Average APR-DRG Charge Weight for the group of patients is required
- Severity-Adjusted Expected Average Charge = (National Avg. Charge) x (Avg. Charge Weight)

### For Severity-Adjusted Expected Mortality:

#### Discharge level

- Assign the APR-DRG and Mortality Rate for the specific APR-DRG and mortality indicator for each discharge
- Average Severity-Adjusted Expected Mortality Rate is simply the average of the rates on the discharge records

## Using Severity-Adjusted Expected LOS, Charges, and Mortality

A Standardized LOS Index is produced when the expected LOS divides the actual LOS. This Standardized LOS Index indicates whether a given subgroup has an average LOS that is lower or higher than expected for the selected group. When calculating the Standardized Charge Index we need to consider wage-adjustment when comparing to the actual Charge. So the Standardized Charge Index can either be the (wage-adjusted actual charges) divided by (expected charges), or the (actual charges) divided by (reverse wage-adjusted expected charges). In this document we have not shown the calculation for the reverse wage-adjusted expected charge so the first option is the one that is recommended. For a Standardized Mortality Index, one calculates the actual Mortality Rate divided by the expected Mortality Rate. To simply know the expected number of deaths in the selected group, sum the expected Mortality Rates at the discharge level. These Standardized Indexes are very useful in making fair comparisons, determining if a group of patients are high, low, or moderate risk, and be able to trend over time.

A Standardized Index of 1.0 would indicate no difference between the actual (or observed) and expected average value. If the Standardized Index is less than 1.0, it indicates that the average observed value is lower than expected, while a Standardized Index greater than 1.0 means that the average observed value is higher than expected. A confidence interval can be calculated to assess whether the Standardized Index for a given subgroup is statistically different from 1.0. Given a specified probability (by convention, 95%), the confidence interval shows a range of possible values of the Standardized Index, which are consistent with the data. If the 95% confidence interval includes the value of 1.0 (indicative of no difference), then the Index is not statistically significant at the 95% level. If the confidence interval excludes the value of 1.0, then the Index is statistically significant at the 95% level. Note that any given Standardized Index may be either lower or higher than 1.0 and not be “different enough” to be statistically significant. An Index may or may not reach statistical significance depending on the size of the sample, the variances of the observed and expected values, and the covariance of the observed and expected values.

Copyright Notes:

- 3M owns all copyrights in and to APR-DRG. All rights reserved.
- Thomson Health Care owns the methodology for weights calculations as described in this document.