Data Quality

Quality data are at the heart of quality healthcare. It is well known that poor data can lead to incorrect diagnoses, prescription errors, or surgical errors with tragic consequences. Similarly, the day-in, day-out consequences of poor data are enormous as well, leading to added time and expense throughout the system. In short, improving data quality is essential.

There are many approaches to defining data. The one that is often used for data quality recognizes that data consist of two interrelated components: data models and data values. Data models define entities, which are real-world objects or concepts, attributes, which are characteristics associated with entities, and relationships among them. As an example, each reader is an entity, and his or her employer is interested in attributes such as name, date of birth, and specialty. Relationships may include report manager and subordinates. A data value is the specific realization of an attribute/relationship for a specified entity. For example, a member of a medical research team may be assigned the specialty “statistician.” Clearly, data, per se, are abstract. Data records are the physical manifestations of data in paper files, forms, spreadsheets, databases, and so forth.

Physicians are uniquely positioned to initiate data quality efforts and have much to gain by doing so. However, most are unfamiliar with the thinking that underlies data quality management: As in healthcare, the steps one takes to improve data quality are rooted in the scientific method. Thus, this entry focuses on physicians. The first part summarizes three key principles of data quality management and the second part offers eight simple prescriptions that physicians can follow to make immediate improvements. These will not, of course, address all the data quality issues that currently afflict healthcare. But they form a solid beginning: the data quality equivalents of the age-old dictum, “First, do no harm.”