Case 9: Coding Accuracy BACKGROUND In our recent report, Using Software to Detect Upcoding of Hospital Bills (OEI-01-97-00010), we examined the ability of commercially available software to identify DRG upcoding through analysis of electronic claims data. We used two software products to identify 299 hospitals with a high predicted rate of upcoding. We then had accredited medical records professionals perform a blinded DRG validation on a sample of over 2,600 claims from 50 of these hospitals and a control group of 20 hospitals. In the course of conducting this study, we developed serious concerns about the potential for abuse of the DRG system through upcoding and about the Healthcare Financing Administration’s (HCFA) oversight of the accuracy of DRG coding. Specifically, we found that, although the hospital payment system is functioning well as a whole, the system has significant vulnerabilities to upcoding that can easily be avoided. We also found that, despite these vulnerabilities, HCFA is not performing routine, ongoing monitoring and analysis of DRG coding to detect problematic DRGs, hospitals, and coding situations that require administrative, educational, or law enforcement intervention. Healthcare Coding Accuracy Case Study

 FINDINGS The DRG system is vulnerable to abuse by providers who wish to increase reimbursement inappropriately through upcoding, particularly so within certain DRGs. Our analysis found noticeable, detectable, and curable upcoding abuses among providers and within specific DRGs. © 2018 Laureate Education, Inc. Page 1 of 6 In a focused sample from a group of 299 hospitals that computer software identified as high upcoders, we found that an average of 11% of DRG bills submitted during 1996 were upcoded, versus 5% of bills among a control sample of hospitals. Identifying Hospitals That Upcode OIG experimental sample: hospitals with a high predicted rate of upcoding Average Average Upcoding Downcoding Rate Rate 11.4% 5.1% 5.2% 3.9% OIG control sample: hospitals without (n=50) a high predicted rate of upcoding (n=20) Source: U.S. Department of Health and Human Services, Office of Inspector General. (1998). Using software to detect upcoding of hospital bills (Report no. OEI-01-97-00010). Retrieved from https://oig.hhs.gov/oei/reports/oei-01-97-00010.pdf The average rate of upcoding in the control sample of hospitals (those without a high predicted rate of upcoding) was not statistically different from the average downcoding rate. However, among hospitals that the software predicted would have a high rate of upcoding, the average upcoding rate was more than twice that of downcoding. The difference between upcoding and downcoding in these hospitals suggests intentional abuse of the DRG system by some providers. © 2018 Laureate Education, Inc. Page 2 of 6 Using data from both our focused review and the more broadly representative 1996 DRG validation performed by HCFA’s clinical data abstraction centers (CDAC), we found that certain DRGs are particularly susceptible to upcoding. Three Highly Vulnerable DRGs OIG Experimental OIG Control CDAC Up- Sample Down- Up- Sample Down- Up- Sample Down- coded coded coded coded coded coded DRG 79: Respiratory 37.7% 0.0% 18.5% 0.0% 11.0% 0.7% Infections (n=60) (n=0) (n=5) (n=0) (n=48) (n=3) DRG 416: 21.2% 0.0% 16.7% 0.0% 13.3% 1.1% Septicemia (n=14) (n=0) (n=3) (n=0) (n=49) (n=4) DRG 14: Specific 10.1% 0.0% 6.7% 0.0% 3.5% 0.4% Cerebrovascular (n=10) (n=0) (n=2) (n=0) (n=24) (n=3) Disorders Claims billed for these three DRGs show a clear pattern that exemplifies the upcoding seen in a group of over half a dozen DRGs we examined. These DRGs were © 2018 Laureate Education, Inc. Page 3 of 6 upcoded disproportionately, especially by our experimentally identified upcoding hospitals and also among hospitals from the general population represented by the CDAC review and our control sample. The HCFA does not routinely analyze readily available billing and clinical data that could be used to proactively identify problems in DRG coding. Additionally, the HCFA does not routinely analyze data from the annual validation of DRG coding performed by its clinical data abstraction centers. Since 1995, HCFA has used two specialized contractors called clinical data abstraction centers to validate the DRGs on an annual national sample of over 20,000 claims billed to Medicare. On a monthly basis, the CDACs report detailed data on each claim reviewed to HCFA’s Office of Clinical Standards and Quality. These data include original and validated diagnostic coding, original and validated DRGs, and reasons for any variance between the DRGs. The purpose of this validation effort is to provide HCFA with insight as to the accuracy of DRGs billed to Medicare. Healthcare Coding Accuracy Case Study

However, we found that HCFA performs no routine, ongoing analysis of CDAC data. In our interviews with staff at the two HCFA components that have responsibility for DRGs—the Office of Clinical Standards and Quality and the Center for Health Plans and Providers—staff were unable to identify any routine monitoring and analysis of CDAC data. In our review of HCFA’s instructions to the peer review organizations (PROs), contractors who have statutory responsibility for DRG oversight, we found no instructions advising them to perform regular analysis of CDAC data. We believe that analysis of CDAC data can be of great value to HCFA in overseeing the accuracy of DRG coding. For example, in HCFA’s 1996 DRG validation, © 2018 Laureate Education, Inc. Page 4 of 6 the CDACs found a 4% upcoding rate with estimated net overpayment of $183 million. Some may suggest that overpayments of $183 million in an $80 billion program (less than one-quarter percent) indicate that the DRG payment system does not have major problems with upcoding and warrants no further analysis. However, our analysis presented above shows that upon digging below the immediate surface, upcoding problems are readily apparent. The HCFA does not routinely analyze data from hospitals, despite the fact that these data are ideally suited for monitoring and analysis of DRGs. The HCFA maintains valuable clinical, demographic, and administrative data that form the underlying basis of each of the over 10 million DRG-based claims billed to Medicare each year. Data for each hospitalization include diagnosis codes, procedure codes, beneficiary demographics, admission and discharge detail, cost reporting data, and hospital identifiers for linkage with provider demographics. Whether used on their own to monitor billing patterns and trends or used to further explore potential problem areas identified within CDAC data, data from hospital claims can provide valuable information to assist in HCFA’s oversight of DRG coding. However, we found that HCFA does not make routine use of data from hospital claims for monitoring and analysis of DRG coding. In our interviews with staff at both HCFA’s Office of Clinical Standards and Quality and its Center for Health Plans and Providers, staff were unable to identify any routine monitoring and analysis of DRG billing data. Interviews at HCFA’s Program Integrity unit, within the Office of Financial Management, revealed that HCFA conducts some limited analysis of billing data. Healthcare Coding Accuracy Case Study

© 2018 Laureate Education, Inc. Page 5 of 6 However, this analysis is done on a very broad level, primarily to identify coverage issues. We also reviewed HCFA’s current instructions to the Medicare PROs. We found no instructions to the PROs advising them to perform any routine monitoring and analysis of DRG coding, despite the fact that PROs already have a complete set of inpatient billing data provided to them by HCFA. In fact, HCFA staff told us that the PROs were instructed not to do “coding projects” within their current contract. We did find that PROs are involved in sporadic activity around DRG oversight; however, this activity often is in support of an OIG investigation. © 2018 Laureate Education, Inc. Page 6 of 6 … Healthcare Coding Accuracy Case Study