In general, physicians are very data driven. They are trained to read, interpret, and form opinions about changes in care based on the latest published scientific data. The use of these data to influence and change general clinical medical practice is well-established and remains the primary methodology accepted by most physicians; however, when these scientific data are applied at the practice level, they are often met with intense doubt, scrutiny, and mistrust.
In general, physicians are very data driven. They are trained to read, interpret, and form opinions about changes in care based on the latest published scientific data. The use of these data to influence and change general clinical medical practice is well-established and remains the primary methodology accepted by most physicians; however, when these scientific data are applied at the practice level, they are often met with intense doubt, scrutiny, and mistrust. Physicians will question any new data that they deem not to be sound methodologically. One immediately hears rebuttals such as “My patients are sicker,” “You can’t compare apples to oranges,” and “But, I’m different.” The origin of such statements must be understood so that systems and tools can be developed and used to present practice data in ways that will lead to meaningful change on the practice level, when such change is desired.
Currently, the primary source of clinical practice data come from the claims submitted for reimbursement. Claims-based data contain two primary fields: ICD-9 codes to characterize what was wrong and CPT codes to show what was done. Most of the rest is pure demographic information. While this is a very imperfect representation of the practice of medicine for most practitioners, it usually seriously understates the situation in oncology.
So, let me describe two different situations, which are much simpler than those generally encountered by oncologists. The first is an episode of a fractured leg. All of the following events are rather easily captured by claims data: 1) Patient breaks leg; 2) Patient goes to emergency department; 3) Patient undergoes radiographic evaluation; 4) Patient sees orthopedist, who performs the appropriate surgical procedure; 5) Patient has follow-up physician, radiology, and, possibly, rehab treatments; 6) Episode ends.
In the world of comparative data, this 6- to 12- week episode is rather easily captured within brackets and all relevant complications and charges are easy to compare across physician groups.
The second situation is that of an individual with diabetes, heart disease, and kidney failure. Although this patient has a much longer and protracted course than that of the patient with the fractured leg, the primary diagnoses and associated complications are relatively standardized and easily captured within the current coding and billing systems. Thus, comparability of the practice of two physicians treating such a patient is not terribly controversial. An episode might look something like this: 1) Patient has history of diabetes, as shown by previous billing codes and pharmacy prescriptions; 2) Patient presents to emergency department with acute myocardial infarction and is admitted; 3) Patient has prolonged hospitalization characterized by dialysis and intensive care; 4) Patient recovers and is discharged with no further need for dialysis; 5) Episode ends (obviously the patient still has diabetes and heart and kidney disease).
Now, contrast these scenarios with those of oncology patients. The two most valuable pieces of information that “stratify” the severity of an oncology patient’s illness are not claims-based or captured by the current billing systems. The stage of the disease, via classical T, N, and M, is the primary predictor of most outcomes in solid tumor oncology. While the M stage usually can be inferred by the appropriate combination of “history of ” codes combined with “secondary malignant neoplasm of ....” codes, the earlier stages generally cannot be differentiated in this way. For example, the episode of a larger stage I breast cancer can look deceivingly similar to that for a stage III breast cancer. In both scenarios, the patient is diagnosed, receives chemotherapy, undergoes surgery, receives radiotherapy, and the episode ends. Outcomes- based analysis without this information will consistently lead to incorrect comparative statements. For other cancers, the claims-based system does not adequately capture prognostic tumor information, even within a given stage (eg, HER2/neu status and estrogen receptor status for breast cancer and prostate-specific antigen and Gleason scores for prostate cancer). The lack of this very critical information also leads to incorrect conclusions in any comparative analysis and when generating a “report card.”
Oncologists are well aware of the complexity of applying ICD-9 and CPT codes to cancer. Many situations exist in which the billing and claims system is woefully inadequate (ie, patient presents with peritoneal carcinomatosis and is treated for ovarian cancer with no prior history thereof). Unfortunately, the use of electronic medical records is not yet widespread or standardized enough at all levels to allow for the easy collection, reporting, and analysis of this sort of data. The episode grouper methodology, best practice in the healthcare payer industry, is good, but inadequate for oncology. So what steps can be taken to remedy this?
Implement and use an EHR. Although the perfect solution does not exist, the standardization of documentation and coding associated with the usage of an EHR will reduce variability across groups and allow for more accurate comparative datasets.
Become acquainted with ICD-10 codes. Although ICD-10 codes will not fix all of the problems associated with the accurate representation of the current status of a patient with cancer, they represent an improvement, and again, will lead to increased standardization of reporting.
Participate in voluntary reporting initiatives. The use of the previous voluntary G codes, although intended to accomplish many of the aforementioned objectives, was not widely adopted. Similarly, participation in subsequent voluntary reporting initiatives has been less than spectacular. Ultimately though, the practice gained in the efficiencies of doing so will pay off when voluntary becomes mandatory.
Maintain an active dialogue with those who do your coding. Make sure that the people who code your billings either have the clinical knowledge to reflect the situation you are trying to portray or have a mechanism to have such dialogues with you on a regular basis. Institute the appropriate quality-control mechanisms to ensure that this is occurring.
Take home message
None of the four aforementioned solutions are necessarily inexpensive or easily accomplished, but the payoff will be a more complete and accurate representation of your clinical practice, and, in the comparative world, help decrease the chance that your “apples” will be compared with someone else’s “oranges.”
Ron Walters, MD, MBA, MHA, MS, is Associate Vice President of Medical Operations and Infor- matics at the University of Texas M. D. Anderson Cancer Center, Houston