Analysis

Analysis is the process of manipulating raw data in order to put it in a form that is useful for making managerial decisions. Analysis does not create information, it only clarifies information already present in the raw data. This means that the person performing the analysis (or setting up the computer to perform the analysis) must be careful not to use techniques that are inappropriate and thus give spurious results.

While the selection of analysis techniques is beyond the scope of this discussion, some general caveats are possible. The most common error in interpreting data is to analyze too small a sample. There is a fundamental relationship between the size fo the sample that is analyzed and the reliability of the results of the analysis. The more detailed the information that is needed, the more data that must be analyzed. The minimum amount of data necessary for a given amount of detail in the results of the analysis is best determined by consultation with a statistician familiar with the mathematical techniques for determining sample size.

A second problem arises when the data are not matched with proper control data. For example, in the use of incident reports to determine the frequency of medication errors, an analysis would not count the errors that did not result in an incident report being filed. A better technique would be to check the number of medication errors obtained from reported against the treatment of patients on whom incident reports were not filed. As the complexity of the desired information increases, the problem of obtaining a proper control populations becomes more difficult. The process of developing control population is related to the problem of sample size and requires the use of mathematical analysis techniques.

A third analysis problem occurs if the raw data have been biased by the collection techniques. This is a severe problem in the medical environment because the persons who are charged with keeping the medical records are also involved in caring for the patient. This makes the medical record a suspect source of quality control information. For example, it would be unreasonable to expect a health care provider to attribute a patient's demise to the provider's own mistake. More subtle biases, stemming from the emotional involvement of the provider in the patient's fate, can also affect the data. In many situations, the medical records are overly optimistic because the providers will not admit to themselves that the patient is doing poorly.

Overcoming the problem of biases is a twofold endeavor. The most important step is to recognize the existence of biases in the data. Once these have been identified, it should be possible to design collection techniques to countered the biases. In some circumstances, this can be done by careful design of the data collection forms; in other cases, it may required bringing in third parties to collect the data.