Clinical pathways

Healthcare professionals, managers and administrators, always seek to provide timely, high quality health services. However, as stated by John Ovretviet [Guinane97], the many potential benefits often fail to be realized due to poor project planning and management.

“People and perfect processes make a quality health service – a poor quality service results from a badly designed and operated process, not from lazy or incompetent health care workers.”

To meet the need of providing high quality health services, managed care plans are desirable. The concepts of clinical pathways (or integrated care pathways) were thus initiated in the early 1990, and can be defined as below [HAIPAP98, Ireson97].

“Clinical pathways are multidisciplinary care plans, in which diagnosis and

therapeutic intervention are performed by physicians, nurses, and other staffs for a particular diagnosis or procedure.”

For exa mple, Figure 2.1 shows a pathway of cholecystectomy [Guinane97]. The pathway begins with the preadmission process, which mainly involves preadmission testing and anesthesia consult, goes though a number of assessments, surgery, and physicians’ orders, and ends with a follow-up visit in the surgeon’s office.

Preadmission Admission Pre-op Surgery Immediate

Figure 2.1 A pathway of cholecystectomy

Clinical pathways, as the one shown in Figure 2.1, are driven by physician orders, and clinical industry and local standards of care. Once the pathways are created, they are viewed as algorithms in that they offer a flow chart format of the decisions to be made

and the care to be provided for a given patient or patient group. Therefore, clinical pathways are developed for the following purposes [Guinane97]:

- Provide explicit and well-defined standards for care.

- Help reduce variations in patient care (standardize care).

- Help improve clinical outcomes.

- Support training.

- Provide a means of continuous quality improvement in healthcare.

- Support clinical audit.

- Support the use of guidelines in clinical practice.

- Help empower patients.

- Help manage clinical risk.

- Help improve communications between different care sectors.

- Disseminate accepted standards of care.

- Provide a baseline for future initiatives.

- Not prescriptive: don't override clinical judgment.

The application of clinical pathways is an efficient approach to analyzing and controlling clinical care processes. In today’s competitive health care environment, due to the fact that competition advantage of a healthcare institution relies not only on outstanding professional quality but also on the agile clinical care processes, the concept of clinical pathways attracts much attention of managers in large hospitals around the world [Ireson97].

From the discussion above, it can be seen that clinical pathways aim to have medical staffs doing the care services in the right order [NELH]. Take the cholecystectomy

pathway in Figure 2.1 as an example. Care activities are sequenced on a timeline so that physicians can make suitable orders in accordance with the test results in the preadmission step; anesthetic can be executed during the performance of surgery on the basis of anesthesia consult. Best practice, without rework and resource waste, performs if the arrangement is in the right order.

This concept of clinical pathways shows great promise on detecting service providers’

fraud and abuse. A care activity is very likely to be fraudulent if it orders suspiciously.

For example, since physicians perform treatments in terms of test results, treatments following no diagnosis/test activities are doubtful; since physicians prefer performing simple, noninvasive tests before performing more complex and/or invasive tests, there is a high possibility tha t the same set of care activities performed in a different order is fraudulent or abusive.

Extensively, to accurately estimate the likelihood of a care activity performed on a particular patient, we must take into account the other activities performed on the patient. For example, while single ambulant visit is normal, repeating events are problematic, especially in the case that averaging length of pathway instances is small.

On the contrary, a kidney transplant that rarely occurs should not be considered so unlikely if the patient has already undergone a series of diagnostic tests typically used to detect kidney disease.

Such observation, therefore, initiates an interesting idea that the clinical structures, including care activities and their execution orders, has the potential to discriminate

between normal and fraudulent practices. Explorative analysis of our dataset¹ supports this argument. In our data set, for example, about 40% fraudulent instances contain a structure of repeating ambulant visits while only 6% normal instances do so.

In this research, we are thus motivated to exploit the discriminating power of clinical structures.

Confusion often arises over the differences between clinical pathways and packages of care. Clinical pathways are ele ments, or chunks of care service. Each chunk of service is developed into a clinical pathway, setting out detailed processes, i.e., a collection of activities, and done as a whole [NELH]. A Package of Care may contain one or more clinical pathways selected for a particular patient or target patient group.

It describes the whole range of care given to that patient or patient group, usually for one episode of care.

Many factors, such as patients’ conditions, physicians’ preferences, and management cost may influence the selection of clinical pathways in a package of care given to a particular patient. Besides, different medical institutes often enforce different pathways, as there does not yet exist a universally best practice for a disease.

Therefore, each patient may have a different practice (an instance). For our purpose−exploiting the clinical structures to discriminate between normal and fraudulent instances, it is necessary to find structures from practice instances since a complete definition of care package does not exist.

Therefore, we conceive to discover structures from practices−normal and fraudulent instances. To exploit the discriminating power of the discovered structures, we adopt

1 Detailed descriptions of the data set are given in Chapter 4.4.

an induction scheme to construct the detection model. Based on the detection model, new coming instances can be detected automatically and systematically. In this research, we explore the advantages of knowledge discovery, rather than knowledge engineering, to detect service providers’ fraud and abuse.