Outcomes are only useful in quality assessment when the specific processes of care that relate to them are known. Then, if the outcomes are not as good as they should be, it is clear what aspects of care have to be addressed to try to improve them. In other words, if you do not know how an outcome relates to processes of care, you will not know what to do to improve the outcome when you find that it is poor at a particular hospital.
It also helps to measure outcomes from different perspectives. For example, palliative chemotherapy for metastatic cancer may decrease a patient's tumor burden and potentially prolong life, but it might also cause severe fatigue and weight loss, so the patient's clinical status might improve while functional status declines.
It is also important to use outcomes that can be reasonably related to the health care sys-tem-and the particular part of the system—that one is assessing. It is not reasonable to hold a provider or plan accountable for an outcome, unless the outcome is a direct result of the way care is provided. Sometimes, however, a single outcome may be influenced by many factors over many years, of which health care is only a part. Outcomes for lung cancer, for example, may reflect the quality of care provided over many years, including the quality of smoking prevention and cessation counseling for adolescents and adults. Outcomes for breast cancer may in part depend on the quality of screening and early detection. Given the frequency with which most patients change clinicians or health plans, it could be difficult to relate the quality of any one clinician or plan to some outcomes. Similarly, if one is trying to use outcomes to assess the quality of surgeons treating a sarcoma at various hospitals, it is important to distinguish whether the outcomes are related to the skill of the surgeon, competence of the surgical team, or organizational characteristics of the hospital. One might also want to consider the skill of the medical oncologist prescribing neoadjuvant chemotherapy. For breast cancer, treatment may depend upon an oncologist, a surgeon, and a radiation oncologist. It can be difficult to distribute responsibility among them.
In addition, outcomes should be measured on samples that are large enough to detect differences in quality. Adverse outcomes are often uncommon events, so large samples are needed to detect clinically meaningful differences between hospitals. To detect a difference of 2 percentage points in the rate of catheter infections between two hospitals (e.g., 5 percent for one and 7 percent for the other), each hospital would have to have at least 1,900 catheterized patients.
In summary, many challenges are inherent in using outcomes to measure quality of care. If these are not addressed, it is difficult to determine whether different outcomes observed among the patients of three physicians are attributable to the physicians themselves. Process measures have their own challenges (e.g., one must make sure that there is a proven link between the process and a desired outcome), but such measures can be quite effective in showing whether providers are doing what they should so that their patients have the best chance of achieving good outcomes. There has been more experience using process measures than outcomes measures to assess quality, and many quality assessment systems depend primarily or exclusively on process measures. However, interest in improving outcomes measurement is increasing, so that outcomes might be used along with process measures to provide more useful assessments of health care quality.
In conclusion, to assess quality of care, measures of structure, process, or outcome can be used. If outcomes measures are used, care must be taken to account for differences that might simply reflect differences in other factors, such as patient selection or case mix. If structure or process measures are used, they should be associated with the desired outcomes. In addition, to make inferences about quality, a measure must be compared to a standard.