Case studies

Business impact

This IT-driven solution — based on IBM Text Analysis and Knowledge Mining — helps medical professionals better understand the progress of patient groups on the hospital-care continuum. It also provides insight into potential treatment alternatives. Finally, the system helps control healthcare costs.

Issue

In Japan, citizens’ healthcare is covered by a public health insurance system subsidized by the nation’s municipalities. Patients freely choose which clinics or hospitals to visit. However, faced with rapidly increasing healthcare costs, the nation is examining why some hospitals operate more efficiently than others. In addition, Japan is taking a closer look at why individual departments in certain hospitals operate more efficiently than corresponding departments in other medical facilities. Length of stay for similar treatments, for example, varies widely from one medical facility to the next, as do physician fees for similar treatments.

Because of this disparity, the Japanese government has decided to institute a new reimbursement system. Called diagnosis procedure combination (DPC), its goal is to provide set reimbursement fees for specified services, without reducing the quality of treatment. To accomplish this, DPC will have to improve hospital administration across the country.

Hospital administrators and medical personnel are now charged with obtaining, managing, understanding and efficiently acting upon patient information stored in the nation’s medical facilities. This information can be found in such disparate, disconnected systems as electronic medical records, medical imaging reference systems, PDAs and PCs.

Executive summary

IBM Research, working together with Global Business Services (GBS), teamed with the National Cancer Center in Japan to develop a prototype program for Medical Text Analysis and Knowledge Mining for Clinical Decision Intelligence. (MedTAKMI-CDI). MedTAKMI-CDI is an online analytical processing system that enables the gathering, interpretation and analysis of clinical data stored in multiple sources. MedTAKMI-CDI currently handles information on roughly 7,000 patients receiving treatment at the National Cancer Center. MedTAKMI-CDI employs a multi-level model for information gathering. Different modules within that framework cover everything from simple demographic information to treatment timetables.

Compiling and analyzing warehouses of medical data, MedTAKMI-CDI helps provide hospitals with a plethora of information about patient groups. For example, hospitals can learn which patient groups (those with similar diagnoses, similar laboratory test results or ages) perform best under which therapies — including surgery, radiation, chemotherapy and various combinations thereof. The information gleaned from analysis of these patterns helps generate analytic rules. These rules prove useful in helping medical staff determine how to best and most cost effectively treat given groups of patients.

What IBM did

IBM researchers, working with GBS consultants, developed an IT-based solution that takes into account patients, medical processes and the interaction between them, using metaschema for fast responses to queries.

A typical MedTAKMI-CDI engagement begins with medical professionals deciding on a group of patients to study. These professionals may choose, for example, to examine men over age 55 suffering from stage two colon cancer. Medical personnel put forth hypotheses about existing best practices; the potential significance of deviation from standard treatment regimes; and other factors affecting health outcomes. MedTAKMI-CDI’s suite of analytical functions tests the original hypotheses and helps provide insight into how to improve treatment strategies. In addition, MedTAKMI-CDI helps spot clinical care patterns and brings to light patient characteristics prevalent enough to form patterns. From this information, predictive rules about improved strategies for care are developed. And finally, a chronological view helps physicians monitor treatment and patient-group outcomes on daily, monthly, quarterly and yearly bases.

Capabilities applied

IBM's TAKMI solution has long enjoyed a strong reputation in the retail, customer service and customer-relationship management arenas. While IBM researchers did leverage these capabilities in the creation of a medical TAKMI solution, they faced formidable challenges in expanding them to cover the myriad vagaries of the human organism.

Consider TAKMI at the retail level. There are very few checkpoints in the retail stock-to-sale continuum. A product is needed at a certain retail location. It is ordered. It is stocked, then sold. It is re-ordered, if necessary.
The human body is not a sweater. In developing the MedTAKMI-CDI solution, IBM researchers had to take into account each patient’s life history, medical records, medical conditions, current and prior treatments, laboratory test results, current health status and projected health status. To do so, IBM researchers designed data warehouse systems for clinical information management, then implemented interconnected analytic functions for knowledge discovery and rule generation.

As a result, physicians are now able to analyze a broad array of patient information, from simple demographics to mission-critical clinical path patterns. Medical professionals may also uncover health trends by viewing ongoing events chronologically — choosing from daily, monthly and quarterly distributions. This helps improve the correlations among cancers, tumor markers and the efficacy of clinical treatments.

While MedTAKMI-CDI is still under intense development — evolving to fulfill the many information requirements of physicians and hospital administrators — it may prove an invaluable tool in helping hospitals provide their patients the best care possible, while assisting Japan in controlling its healthcare costs.