By Richard C. Johnston, MD
Outcomes data are nothing more than the relevant facts regarding our care of a group of patients. Exactly what facts are relevant? That depends on what we will do with the facts.
We have said that we need outcomes data to improve either our selection or implementation of treatments (management strategies, interventions). Therefore, we need to organize the collection and management of the facts in such a manner that we can distinguish reality from noise (bias and confounding). Measurement must be sensitive enough to detect differences that are greater than can be attributed to chance.
We must consider routine clinical care a giant natural experiment and we need some sort of experimental design. We must describe the population before the intervention, describe the intervention and then describe the population after the intervention. A randomized controlled trial is the accepted gold standard for accomplishing this, with inclusion and exclusion criteria that attempt to make the population homogenous, randomization that attempts to make two populations the same to eliminate bias and confounding and rigid control to keep the intervention constant.
All of these aspects are simply not compatible with routine clinical care and are extremely expensive. However, a quasi-experimental design is quite compatible with routine clinical care. The precision with which we describe the populations (groups of patients with the same diagnoses or treatments) pre- and postintervention (treatment) and describe the interventions (treatments) will determine the accuracy with which our selection and implementation of treatments can be judged.
The more facts we know about the patientís pre- and post-treatment, and the more facts we know about the treatment, the greater the precision of the descriptions. Therefore, the more relevant facts that are available, the more accurately our selection and implementation of treatments (our work) can be judged. If our work must be judged, and it is being judged, I think we would want that to be done as accurately as possible.
Therefore, we need to collect data to describe the patient, (demographics, co-morbidities, and expectations) and the patientís state of health (clinical and functional health status and well-being) pre- and post-treatment. The Academyís MODEMS forms and some standard physical exam parameters do this quite well. We also need to describe the treatment with at least the CPT code. More detail needs to be standardized for the future. This turns out to be a very, very large amount of data.
In the practice setting, time constraints on physicians and staff make it difficult to collect data expressly for research purposes. Any research tool employed over and above standard office processes is viewed as an imposition and an unnecessary cost. In most cases collection of research (outcomes) data has represented double work and involved changes in routine. Many of us have experienced the difficulty of trying to compensate for this phenomenon by hiring a special research assistant. He or she is also viewed as an imposition and an unnecessary cost.
Consequently, it is extremely rare that data expressly used for research purposes, no matter how minimal, is consistently collected over a significant time period. Therefore, it is mandatory that all necessary data be collected as an integral part of standard office processes in the routine care of patients in a manner that does not increase the burden of care.
Since outside changes are putting pressure on our time and our income, the standard office processes need to be redesigned in such a way that the burden of care for the physician and/or staff is not increased. This can be accomplished by using the answers to the questions on the patient administered forms and standard physical exam parameters in the MODEMS package, entered into a relational database, and merged with standard or individual physician designed templates to produce clinical notes, insurance claims and bills in an automated fashion. The data in the relational database can then be analyzed to accurately judge our work.