Administrative Memos

200114

FROM: Theodore O. Will, Chief Executive Officer
DATE: Jul 12, 2001
SUBJECT: Hospital-Specific DRG Length of Stay Report for Fiscal Years (FYS) 1999 and 2000
IPRO CONTACTS:

Kathy Terry, Ph.D., Senior Director of Data Analysis & Evaluation, Extension 261

Coding errors resulting in incorrect DRG assignment are a major focus of the Payment Error Prevention Program (PEPP). These errors can occur due to confusion over coding rules, inconsistencies in the medical record, or incomplete or missing documentation. PEPP is designed to reduce Medicare inpatient Prospective Payment System (PPS) payment errors, including errors that lead to over-payment as well as under-payment to hospitals. One way of identifying potential coding payment errors is by examining the length of stay (LOS) associated with each DRG.

In a series of conference calls with New York State (NYS) hospitals, it was requested that IPRO produce a hospital report detailing LOS data for high volume DRGs. To assist hospitals in their efforts to identify and reduce Medicare inpatient payment errors due to coding issues, IPRO has produced the enclosed reports which identify the national, NYS, peer group, and hospital-specific arithmetic mean length of stay (AMLOS) for your hospital's top 20 DRGs by volume for FY 1999 and FY 2000. The reports detail the shortest LOS, longest LOS and percentile points1 associated with those DRGs. Averages are calculated as the AMLOS for all cases in which the patient was discharged to home. All other discharges (transferred, expired, leaving against medical advice (AMA), or non-PPS claims) are excluded. Preliminary analysis identified several cases with atypically high LOS. Since these aberrant cases would artificially inflate the AMLOS, the top 0.1% (LOS>65) of the cases were excluded, leaving 99.9% of the cases for the analyses.

For FY 2000, case listings of low and high outliers are also provided. Outliers were defined as cases falling in the 10% extremities (i.e., equal to or less than the 10th percentile point or, equal to or greater than the 90th percentile point). These case listings will enable you to select and review sample cases in order to identify hospital-specific issues that might have led to these outlier distributions. Hospital review of these cases may identify instances of either over-coded or under-coded DRGs.

Peer, NYS, and National AMLOS data is also provided. The differences between a hospital's AMLOS and those of its peers may be the result of many factors including differences in patient population and services provided. Therefore, IPRO does not suggest that simply differing from your peer hospitals or the statewide average should be cause for concern. However, when there is a large variation from peer hospitals, NYS, or the national average for certain DRGs, a hospital may want to take a closer look to determine if the DRG was coded appropriately.

IPRO is planning a follow-up conference call to provide hospitals with an opportunity to discuss these reports. The call will last one (1) hour. Please select one of the three available dates/times that follow, then call 1-877-806-9727 to register. At registration, you will receive additional call-in information. (NOTE: the same agenda will be followed for all three calls, therefore it is only necessary for you to participate in one session). The calls have been scheduled as follows:

Tuesday, August 14, 2001; 11:00am - 12:00 pm
Wednesday, August 15, 2001; 2:00 - 3:00 pm
Thursday, August 16, 2001; 10:00 - 11:00 am

Please note, the registration period for a call closes the day before the call.

To facilitate discussion, the agenda we will follow has been provided at the end of this document.

RECOMMENDATIONS:

We recommend performing auditing and monitoring of cases that fall in the 10th and 90th percentiles in each of the top 20 DRGs. Medical record review will help identify patterns in regard to payment error issues such as medical necessity, utilization management, coding and/or documentation. The following are some suggested approaches that you may consider in routinely monitoring and analyzing DRG LOS:

Review cases with the shortest and longest LOS for medical necessity and utilization management concerns, e.g., for long LOS cases, is it more appropriate to have transferred the patient to an alternative setting or level of care? For short LOS cases, did the patient require admission/treatment in an acute care setting?
Identify your top 20 DRGs that demonstrate similar diagnoses/procedures but differ in revenue consumption and/or LOS patterns such as DRGs representing conditions with or without complications and/or comorbidities (e.g., DRG 89 vs. 79, DRG 110 vs.111, DRG 138 vs 139, etc.). Does the documentation support the complication and/or comorbidity?
Analyze outlier LOS data by physician, by diagnosis or procedure, or by DRG to help identify possible root cause(s) of the high or low LOS. Cost and resource consumption variations for the same DRG among physician or among coders may help focus training and educational opportunities.
Identify DRGs that are significantly different from your peer, NYS, and/or national AMLOS rates. Reviewing some of these cases may reveal areas requiring further analysis.
Continue to evaluate your patient discharge process. Are there discharge delays that may be avoided by improving the timeliness of transmission of test results or by notifying the family members of discharge plans sooner?

The above recommendations represent a few suggested methods for hospitals to utilize in assessing their data. IPRO is most interested in obtaining hospital feedback in regard to the use of these data reports and case listings, as well as other methods employed to analyze your data. We look forward to your conference call participation.

Should you have any questions in regard to this memorandum or the enclosed reports, please feel free to contact Dr. Kathy Terry.

¹A percentile is a value on a scale of one hundred that marks the percent of the distribution that is equal to or below it. Therefore, percentiles provide information about data distributions. For example, the 10th percentile identifies that 10% of your cases fall below the identified point.