Policies are, of course, carried out by individuals. The personnel who provide clinical care, maintain engineering systems, and otherwise do the day-to-day work of the organization often have considerable discretion in following policies and recommended practices. In recent years, many health care and other organizations have attempted to design systems to minimize opportunities for unwanted variations in work practices. For example, some facilities have installed electronic monitoring systems that check whether doors to tuberculosis isolation rooms are closed, consistent with policy. Nonetheless, for many activities, universal standardization or monitoring of work practices would be viewed as impossible, offensive, counterproductive, or excessively expensive. Thus, individual adherence to organizational policies continues to be a concern.
Mailed surveys are a relatively inexpensive way of collecting information about a large number of geographically dispersed institutions. When the surveyed institutions are familiar with both the surveying organization and the kinds of questions asked and when the topic is viewed as important, voluntary questionnaires can generate respectable response rates of 70 percent or more.
The potential limitations of survey data are, however, familiar. If all members of a population are not surveyed and the sample of the population is not properly selected, the subset chosen may be unrepresentative of the population. This limits generalizations from the surveyed population to the larger population. Whether surveys are directed to a universe or a representative sample, the lower the response rate, the greater the concern that responses will be unrepresentative. In addition, survey questions may be deliberately or unintentionally biased or otherwise formulated in ways likely to produce inaccurate and unrepresentative responses. Even if the questions are sound, those who respond may intentionally or unintentionally provide inaccurate or insufficient information. Bias is a particular concern if those surveyed know that important policy decisions may hinge on the survey results. Organizational surveys may also be misdirected to and returned by individuals who lack the knowledge to respond accurately.
In 1992, while increases in tuberculosis cases and case rates were still being recorded, CDC surveyed hospitals about their tuberculosis control practices. Questionnaires went to all 632 federal, state, and local public hospitals in the United States and to a 20 percent random sample