Using Data To Define Quality Performance Based Networks
People in Workers’ Compensation are beginning to power up their data to gain insight and objective decision support to structure their provider networks. To do that, physician and other provider performance is evaluated based on actual performance evidenced in the data. That seems simple enough on the surface, but it is fraught with challenges. A few are described here, along with a case description of fraud by data proliferation.
Evaluating the data to determine provider performance quality is tricky. For instance, who among those treating a claimant should be held most responsible for claim outcome? Which provider is the so-called primary provider? Is it the first provider to see the claimant, the provider who has charged the most money, or the one who saw the claimant most frequently? There is no specific indicator in the data denoting primary provider, nor do providers generally self-identify in that way unless they are involved in a formal gatekeeper arrangement. Consequently, for analytic purposes a decision must be made regarding provider influence in the claim, aka, primary provider.
Distinguishing Individual Providers
Another common problem is that individual providers are often not differentiated in the data. Many payers accept bills “as is,” meaning they do not require the billing entities to specify individuals. Typically, individual physicians and other providers are camouflaged under the organization’s Tax ID. In the past, that was adequate because the purpose of the bill was to pay and record the transaction. But that is no longer good enough because of the demand for analytics.
Bills are now a significant piece of the data required for provider performance analytics. Therefore, for individual treating providers, the National Provider Identification number (NPI) or state license number is needed to recognize single medical doctors or other professionals treating claimants. Unfortunately these identifiers are usually not included in the data. Withholding payment is the most powerful method of generating compliance and payers have that power.
Moreover, among data issues, deliberate identity proliferation is even more damaging to accurate provider performance analytics.
Medical fraud surfaces in many forms. Duplicate billing, up-charging, and optimizing charge codes and diagnostic codes (up-coding) are among the most common, but now newly creative methods are being employed by a few. Perpetrators are obfuscating the data to conceal their poor performance by proliferating their identities in the data.
By altering names or addresses slightly, thereby adding to their number, providers are able to cause the system to recognize each variation as a separate entity. That way, multiple provider records are created in the data, even though they are really all the same individual. Proliferating provider records in a data set effectively skews the results of performance analytics.
A Case Of Data Proliferation
Provider identify proliferation was discovered recently when a monthly billing report for an organization was analyzed. Fifty (50!) different name and address iterations for the same medical provider Tax ID were discovered. This had been attempted previously, but this time, the effort was extreme.
Is this provider representing themselves carelessly? Probably not. The provider knows computer systems consider data literally, so each submission would generate a new record, the hoped-for result. Without investigation, the provider’s billing will not be questioned, yet when the provider’s performance is analyzed, the results will be distorted and inaccurate.
The provider vendor will be paid because all 50 iterations have an acceptable Tax ID. However, the problem surfaces when executing provider performance analytics. Different claims are attached to the 50 different records for the provider rather than consolidated in one record for the provider. Performance indicators are distributed across the faux entities rather than consolidated for the single provider, thereby distorting performance results, a new-age form of medical fraud.
As with many forms of fraud, the solution is to discover and subvert the effort early. Evidence-based quality networks composed of quality individual providers cannot be created using such distorted data. Payers should monitor their data to discover and expose such behavior as it occurs.
Payer systems are culpable, as well. Systems should be designed or updated so that multiple record entry is thwarted, either through administrative procedures for data entry or simple technical methods. Including individual identifiers such as National Provider Identification and state license numbers will add to the solution, forcing accuracy in provider records.
For the case described here, an additional solution was implemented. The multiple provider identities were merged electronically by the analytics company, thereby integrating the occurrences for this perpetrating provider. As a result, the provider’s performance can be analyzed as a whole rather than in fragments.
Because claims actually associated with this provider are distributed across the multiple artificial provider records in the data, analysis of performance is inaccurate. Not surprisingly, when this provider’s data was merged and re-analyzed, the provider ranked in the lowest performance quartile. Gotcha.