Six design and implementation lessons.

One of the great paradoxes of medicine has been the cost-increasing effect of medical technology. Instead of increasing productivity, new medical equipment often adds great diagnostic insight and highly specialized staff. Instead of replacing outmoded tests and procedures, the latest advance in medical technology often tends to be an additional procedure. One area of technology, information technology, is countering this trend in medicine.

There are numerous examples of how computers and databases have been applied to medical care, including cancer and poison control databases, telemetry, and hospital management information systems. Utilization review (UR), conducted on behalf of employers as a part of managed health care programs, is also information technology applied to medical care. Viewed in this manner, UR may be defined as the collection of data about proposed, ongoing, or historical tests, treatments, or procedures; comparison of the data to criteria, or standards of medical practice; and feedback of comparative results to a practitioner.

This definition includes certain key information management components: data collection, reorganization of the data to make them meaningful, analysis (comparison to standards), and reporting (or feedback). Inpatient hospital UR, which has been shown to be cost-effective,(1) provides a familiar context: nurses collect data about a patient over the phone or on site; compare that information to admission criteria and length of stay norms; and then report the results of that comparison, often with consultation from a review physician, to the attending physician in the form of a review decision.

Lessons on How to Design Ambulatory UR

As in the case of inpatient UR, ambulatory UR (AUR) entails data collection, comparison of data to standards, and feedback of results to practitioners. There the similarities virtually end. Each of these three information management steps is complicated by the nature of ambulatory care. Because of this, significant lessons were learned at each step of the AUR program design process employed by Metropolitan Life Insurance Company.

AUR Lesson #1: Claims Data Are Needed to Augment Review Data. Hospital UR usually can be accomplished with just one phone call or a visit to one place. In ambulatory care, the patient may walk out of the doctor's office and visit another doctor, a clinical laboratory, or a radiology department at the local hospital. Even though one primary care physician may be responsible for the patient's travels, a phone call (or a visit, were it economically feasible) to the doctor's office is unlikely to yield all of the data necessary for review. In fact, one might not even know when to make the call. In contrast to the concentrated episode of care in a hospital, ambulatory care episodes of care tend to stretch out over weeks or months.

One place where most of the data describing ambulatory utilization come together is in the patient's claim history. Given the need for linking data over weeks or months of service delivery, we decided that an AUR program needed to give review nurses and physicians access to relevant claims data, a database comprising tens of millions of records. Considering how to use this claims database yielded the next lesson.

AUR Lesson #2: Episodes of Care Should Be Reviewed. The use of data making comparisons to standards in ambulatory care is complicated by two issues: comparison of what and to what? In the hospital, an episode of care is conveniently defined as the care that takes place between admission and discharge. Of course, patients entering hospitals or doctors' offices are not well when they arrive, and when they leave, they are typically not fully recovered. In hospital review, "admission" and "discharge" dates are conveniently available and used to define episodes of care. In ambulatory care, there are no clear indicators of the beginning or the end of an episode--no "length of stay" norms, for example, to use as benchmarks. What is it that needs to be compared in AUR? What are the standards of comparison? What is there to feed back to practitioners?

To resolve these analytic problems, we decided to approximate an episode of care analytic methodology by using intervals of care defined as all services provided by one physician to one patient for one diagnostic category during fixed intervals of time.

We adopted "Patterns of Treatment[R]," (2) a set of standards covering virtually all services and all diagnoses that is available in this format and that is criteria for the maximum monthly, quarterly, and annual frequencies of ambulatory service utilization. "Patterns of Treatment" provides professionally sound standards against which the data collected during review may be compared, but the necessarily simple definition of an episode and the nature of data collected caused yet another journey up the "learning curve."

AUR Lesson #3: System Design Should Be High-Tech and High-Touch. Although an episode of care was defined to encompass only one diagnostic category for both the data collection and the comparative norms, individuals often present more than one diagnosis at a time. As in the case of inpatient review, various nuances of ambulatory care, the severity of illness, and the complexity of each patient's condition need to be considered. "Patterns of Treatment" accomplishes this to some extent; for example, separate standards are provided for routine and complicated pregnancies.

Even though the diagnostic coding of claims data had improved by 1986, and even though it has improved since then, claims data alone cannot fully uncover these textures of patients' conditions. AUR needed a new variety of high-touch professional review to coexist with its high-tech information management design features. The AUR system needed to be designed to give nurse and physician reviewers easy, on-line access to all claims information that could be relevant to a review.

All information that could be relevant turned out to be more than just the data constituting an episode of care under review. We decided to arm the nurses with additional data from the patient's total claim history. The information gleaned from this database would prove to enhance and speed review by helping uncover the "textures" of patients' conditions not evident in the episode data. For example, review nurses could interpret ambulatory services in light of associated hospitalizations.

This high-touch-inspired design feature, however, raised the ante on high-tech system design requirements. As the complexity of system design increased, so did the challenge of integrating AUR with claims payment. One particular challenge, our last design requirement, arose because of the volumes of data and transactions involved.

AUR Lesson #4: Ambulatory Review Entails High-Volume Review and High-Volume Claim Payment. Inpatient review has caused significant adjustments among insurers' claims operations because of the necessity to add input into medical necessity determinations and to keep track of benefit plan authorization requirements. To this set of administrative requirements, ambulatory care and AUR add the need to simultaneously process vast volumes of claims and claims data.

While the average person covered under an employment-based benefit plan has about a 1 in 15 chance of being hospitalized during a year, each person covered under such a plan might have about four ambulatory physician visits during a year--not to mention associated visits for laboratory tests, x-rays, and the like. For Metropolitan, about 11 percent of claims have to do with inpatient services, while 89 percent are associated with ambulatory care.

The challenge here was to review a large volume of relatively low dollar value claims in a cost-effective manner. For the AUR program, "cost-effective" meant meeting two objectives:

* Saving more than the cost of review.

* Paying the claim correctly and as promptly as the adjudication process will allow.

While these design objectives were established at the outset, their achievement was only possible in the continuously evolving operational stage of AUR.

Lessons on How to Implement AUR The high-tech/high-touch design requirements of AUR result in the process portrayed in the figure to the right. Steps l and 2, claims submission and data entry, and the final step, processing the claim, are familiar to most employers as a standard claim payment sequence. AUR adds three additional steps, two of which are familiar components of inpatient UR.

Step 3, mainframe processing, is unique to AUR and entails three components. First, the computer creates 1-, 3-, and 12-month(3) intervals of care by combining all of the patient's claim records over the past year for the same doctor and the same diagnosis. This episode of care is then compared to the "Patterns of Treatment." Finally, exceptions, claims that "fail" the computer screens, are referred to registered nurses for review.

Like standard inpatient UR processes, AUR entails RN review (Step 4) and, if the review nurse and practitioner cannot agree on a resolution, MD review (Step 5). As noted above, access to complete patient claim history during these steps permits increased sensitivity to the "textures" of an episode of care--both clinical and with respect to benefit plan determinations. Often, the additional data so provided permit release for payment of a claim with no further action. If the available data do not clear up potential problems with a claim, additional information is requested from the responsible provider in writing or via telephone.

Consideration of the latter phases of the review process, nurse and physician involvement, serve well to introduce the implementation issues learned. These lessons have to do with some of the complexities of the program design and with the novelty of AUR itself.

AUR Lesson #5: AUR Requires Close Cooperation of Claims and Review Personnel. AUR presents a new set of objectives for any insurer. To the historical goals of prompt, accurate claim payment are added the imperatives of managed care: enhanced determination of medical necessity and appropriateness, and increased profiling of individual provider performance. Both claims and medical management personnel need to contribute to meeting these objectives.

Even though the electronic AUR system permits physical separation of review nurses from claims payment staff, we located ambulatory review nurses in each of our 19 claims offices nationwide. Not only did this proximity enable nurses to transfer some of their expertise to claims approvers, as will be discussed in Part II, but it also engendered a mutual understanding of each others' potential contribution to success in managed care.

Perhaps the greatest key to success in operationalizing AUR was an early recognition by all parties of the need to focus AUR on only the most aberrant cases. At the beginning, there arose a potentially serious problem with the computer excessively "kicking out" claims that failed the review criteria but that turned out to be payable. One might apply the statistical term "false positive" to such a situation. Considering the turnaround time objectives that claims offices face, there could have been no worse result for AUR than to slow down the claim process--only to have the claim ultimately paid as originally submitted. The cooperation of claims approvers and review staff was tested by "false positives," but several years of improvements to the system have overcome the problem. These improvements included adjusting claim selection and review criteria and "flagging" particularly aberrant providers for 100 percent review.

Continuous review and fine-tuning of review criteria have focused review on claims and physicians who most warrant review. Today, only 6.4 percent of claims eligible for AUR are suspended and, of those, about a third are false positives. Perhaps more important, under the current AUR process, 0.9 percent of physicians account for 81 percent of all AUR-declined charges, and 91.6 percent of all doctors who submit claims have no charges declined. This phenomenon of a relatively small number of physicians accounting for most AUR activity introduces a final implementation lesson: the need for communication.

AUR Lesson #6: Communication to Employers and Providers Is Essential. AUR, like any groundbreaking technology, can cause concern among those whom it affects. Communication about the program in advance of implementation can be a critical success factor.

In the case of physicians, the sudden capability of review professionals to peer into previously "safe" office practices often proved disconcerting. By communicating details about the program to all potentially affected physicians, we gained important allies: the 90+ percent of physicians who are unaffected by the program. When AUR began identifying new types and amounts of aberrant practice patterns, one of the first reactions of the affected physicians was to complain to their colleagues. A few comments along the lines "Oh, I haven't found the program to be a problem" helped quiet the complaints.

Communication to employees and corporate human resources departments was also important, particularly prior to the growth of managed care. If the AUR program declines a claim in a nonmanaged care environment, it is not uncommon for the physician to look to the patient for payment. Careful consideration before program implementation of how such situations are to be handled can smooth the way quite a bit.

AUR can also play a role in educating patients to be better consumers of medical care. In nonmanaged care situations, where AUR can identify aberrant practice patterns only after services are rendered and claims submitted, a common action is to issue a "warning letter" to the physician, with a copy to the patient. Such a letter would advise that future services may be denied for payment and could stimulate some constructive dialogue between the patient and the physician.

Summary

Increasing utilization of ambulatory care services presents a challenge for managed care: how to review the appropriateness of a high volume of low dollar value services. By combining sophisticated claims data analysis with traditional components of utilization review, we were able to design and implement a cost-effective ambulatory utilization review program. After uncovering new, "high-tech" episode of care analytic methodologies and review standards, we learned that high-touch components of review were also needed in the AUR program design. These components necessitated more data than usual to support review.

Internally, the high volume of ambulatory claims necessitated improved working relationships and sensitivity among claims and medical management personnel. This was essential to making review cost-effective. Externally, the program required careful attention to provider and employer communications to ensure smooth implementation.

Today, our AUR program reviews about 600,000 ambulatory claims in an average month, or about 30,000 claims per day, and the program saves almost $5.00 for each dollar of program administration expense. How these savings are achieved, and some unintended program results are covered in Part II of this article.

Footnotes

1. Wickizer, T. "The Effect of Utilization Review on Hospital Use and Expenditures: A Review of the Literature and an Update on Recent Findings." Medical Care Review 47(3):327-63, Fall 1990.

2. "Patterns of Treatment[R]" was licensed for use by Concurrent Review Technology, Inc., Shingle Springs, Calif.

3. While these are the standard intervals, in some instances, "lifetimes" are used. For example, CPT-4 codes 99201 to 99205 comprise office medical services for "new" patients who should be coded as "established" patients (99211 to 99215) for subsequent visits to the same physician.

Further Reading

The following additional sources of information on computerization in clinical practice were obtained through a computerized search of databases. For further information on citations, contact Gwen Zins, Director of Information Services, at College headquarters, 813/287-2000.

Covert, K., and Green, K. "Group Practices Expand Electronic Communications." Medical Group Management Journal 39(6):26-30, Nov.-Dec. 1992.

Pomiecko, E. "CompreLink--A Physician to Physician Communication Network." Journal of Medical Systems 16(2-3):87-90, June 1992.

McGee-Cory, J., and Hantho, L. "Medical Groups and Hospitals. Win/Win Collaboration Using Technology." Medical Group Management Journal 38(4):34-6, July-Aug. 1991.

Williams, S., and Burlington, S. "Integrating Medical Records: A Joint Venture." Topics in Health Record Management 8(3):48-55, March 1988.

Kaiser, L. "The Next Medical Frontier: Computer and Robotic-Enhanced Health Care." Group Practice Journal 35(6):5-6,10-2, Nov.-Dec. 1986.

Fox, R. "Connecting Lab to Client Facilities: Link or Sink." Healthcare Informatics 10(3):28-30,32, March 1993.

Dabney, B. "The Impact of Information Technology on Group Practice." Medical Group Management Journal 39(6):56-9,77, Nov.-Dec. 1992.

Gans, D. "Medical Group Information Systems." Medical Group Management Journal 36(2):11,55-6, March-April 1989.

Whinnery, S. "Electronic Claims Submission Helps Group Practice Cut Costs." Group Practice Journal 37(4):26-7,58, July-Aug. 1988.

Paul Frankel, MD, PhD, was Vice President and National Medical Director; Robert Chernow was Senior Vice President, and William Rosenberg is Assistant vice President, Managed Care Services Group, Metropolitan Life Insurance co., Westport, Conn. Dr. Frankel is now President of Life Extension Institute, new York, N.Y. Mr. Chernow is now President, Value Health Information Group, Avon, Conn.