Importance of Definitions

INTRODUCTION

On my recent trip to London I visited the House of Lords. Imagine how surprised I was to hear Lords discussing a data-modeling question. A government minister was defending a definition from his
data dictionary. Definitions are usually at the bottom of any project’s priority list. Here I was, seeing the issue elevated for debate in the highest legislative body of the United Kingdom!
It was very educational. I sincerely hope that all of our definitions could withstand such scrutiny.

The discussion that took place in the House of Lords on January 21, 2003 is presented verbatim in Appendix A. Let’s go over the definition and debate and draw recommendations for day-to-day
Information Technology practice.

A DEFINITION UNDER FIRE

The discussion in the House of Lords was around the definition of a hospital bed.

“So what seems to be the problem, a bed is a bed, right?” The following is the official National Health Services definition [1].

“A bed includes any device that may be used to permit a PATIENT to lie down when the need to do so is as a consequence of the PATIENT’s condition rather than the need for active intervention such
as examination, diagnostic investigation, manipulation/treatment, or transport. Cots should be included in statistics about beds where appropriate. It should be noted that:

HOW DOES IT STACK AGAINST THE RULES

Let’s verify the definition against the standard rules [2]

• It communicates what the entity represents and the role it plays in the enterprise. It avoids describing how, where, or when the entity is used, or who uses it. An example violating this rule
  by specifying how the entity is used might be: A bed is a unit of government statistics on hospital usage.
• It states the general meaning followed by the details and exceptions. As the last part it includes a couple of representative examples.
• This definition stands alone; it is not dependent on other definitions to convey its meaning. Circular definitions would violate this rule: BED: It is something used by a PATIENT. PATIENT:
  Someone who uses a BED.
• It is logically organized and articulated in full sentences.

The above is a solid definition that follows the formal rules. This ensures that the intended meaning is delivered to the reader. The next question is what that meaning is and whether it reflects
reality. The text shows the complexity of the problem. In order to classify a bed correctly, it is not enough to simply consider the physical device, but it is also necessary to know how the device
is used. During the debate this stemmed the argument that the same trolley can be moved to a different place in the hospital and be counted differently. This would make statistics ambiguous. But
according to the minister the definition reflects the underlying realities. His statement that “The whole point of having a bed definition is to achieve clarity and consistency of
reporting” emphasizes the strength of the definition and his confidence in it.

IMPORTANCE OF DEFINITIONS

Documenting requires making an extra effort, actually doing a lot of extra work for an extended period of time. This is why an internal conviction of the importance of the work is crucial.

The ministry as a government organization must provide information to the general public. With all the external scrutiny the requirements on the quality of definitions are high. One would guess
that if the health services minister had lost the argument on the weakness of the data dictionary definition he would have made sure that it did not happen again. So just learning of this story
might be enough for a government agency to take a hard look at the state of its data dictionary. Non-profits and inter-organizational groups are in a similar situation with the need to share
information with outside groups.

As for the rest of us, data models are created for internal consumption. There is no incentive to document at the company level, but there is a group of people who depend on quality meta-data for
their survival. Many companies have a team dedicated to Data Administration. Their work is based on well-tested relational theory and expertise gained in multiple projects. They provide value to
the company by designing databases that not only solve today’s problems but can also withstand the test of time. Nowadays major development initiatives are scarce. Under these circumstances
support of existing databases becomes the most important part of Data Administration. This work involves small incremental changes and success depends on a repeat customer. A good definition is a
powerful tool in the fight for this customer.

The very first consumer for the definitions is the Data Administrator (DA) himself. In order to process an incoming change a responsible DA has to understand the overall design and the surrounding
constructs. This is impossible to do with a bare-bones no-definitions model. It means constantly requesting additional information from the business users and developers. This causes developers
perceive the DA as a hurdle to the completion of the project. On the other hand, when armed with good quality definitions, the DA can make incisive decisions, maintain database quality and enforce
proper data perspective. It changes the direction of information flow. Now the DA provides the information and by doing that adds value to the project.

ACQUIRING DEFINITIONS

There are a number of ways to acquire definitions. The ideal option is to create definitions while in the initial database design stage. The recent wave of technological advances required a lot of
projects to be done in a hurry. This led to a significant gap in documentation. The current slow down can be effectively used to fill in the gap and to prepare for the next technological boom.

We can force people to do definitions for us, entice them to help us, do it ourselves, or combine the above approaches.

• Find an influential sponsor to create a special program/project/committee. One possible way is establishing a data stewardship program. The data stewardship creates ‘proxy’ data
  modelers out of business users, and makes them responsible for handling the enterprise-wide glossary of terms. For example in the Rational Unified Process the role of Business System Analyst could
  effectively take over as a data steward in order to maintain the data dictionary.
• Publish existing definitions and establish a feedback procedure. For many people it is easier to fix existing text than to create a brand new piece of information. It also appeals to people to
  be able to correct someone. Make the procedures easy to use and remember to recognize individuals who contribute. Specialized tools can facilitate the process of publishing definitions and provide
  an easier interface for editorial work. One such tool is Wiki [3]. The Wiki concept revolves around “open editing”. It keeps track of who makes changes and when those changes are made.
• Do-it-yourself. There is a lot that you can accomplish even without organizational and management support. Start with original design documents if any. On each occasion and under the smallest
  pretense ask questions. During each modification effort ask questions about all of the data elements involved. Ask for screenshots and request process walkthroughs. Draw information from all
  available sources including books, articles, vendors’ and competitors’ sites, previous experience, and plain common sense.

REVIEWS

When you feel that you have grasped enough to understand the overall picture you can do an overview of the data model for the development team. This is usually when people start telling you a lot
of things that they seem to have forgotten.

You can also include a peer review in your process. It is a great idea to use the expertise of other DA’s on your project. However, be aware that the reviews can get very personal and strain
relationships within the team. Make them formal with all the rules published ahead of time. Define specific roles, including a reviewer, facilitator, etc. Our reviews might benefit from the Peer
behavior in the House of Lords. Just a cursory observation shows that each new statement is started with a formal greeting. This constant acknowledgment of an opponent’s value makes
conversation more formal and seems to minimize unnecessary personal undertones.

Our work undergoes reviews by application teams, business users, upper management, etc. In the above example a completely independent body did the review. The government position was attacked from
several angles. Opponents used examples and scenarios to make their points. The stakes were heightened by the fact that the whole debate was recorded and transcribed verbatim. It is important to be
able to articulate and defend your position in front of a group of people. The required skill set should be developed and promoted. For example, dedicate a weekly meeting where team members take
turns to present and defend various topics. Topics outside of the current project work would remove personal concerns and make discussions easier.

CHANGE TRACKING

The debate in the House of Lords concentrated on the impact of the definition on government statistics. The main question was whether an improvement in statistical results could be caused by
changes in the definitions. When the question of changing the bed definition was parried, it was expanded to a more general line of questioning on possible renaming.

Model versioning and keeping a maintenance log helps trace by whom, when, and on which request a change was made. Also, if you have definitions in different stages of readiness it is helpful to
document the current state of those definitions.

CONCLUSION

The debate in the House of Lords is a case study of the importance of data administration. One the one hand it is a good example in itself. It shows how data dictionary definitions were used to
establish and defend government policies. On the other hand, it provides another chance to reflect on the state of our day-to-day practice. It is not enough to just design databases. Good
documentation is crucial for data administration especially during database maintenance. The ability to defend one’s position is very important as well. When all these components are present,
Data Administrators become indispensable. This article has presented some techniques and tools that should help you maintain accurate documentation as well as enhance your value as Data
Administrators to the organization. So let’s do our part in improving the quality of definitions and be ready to defend them at any level.

REFERENCES

1. Data Standards Definition Request HD1211 on 14 Jan 03 www.nhsia.nhs.uk
2. The data modeling handbook: a best-practice approach to building quality data models Michael C. Reingruber, William W. Gregory. ISBN 0-471-05290-6, 1994.
3. What is Wiki www.wiki.org
4. Lord’s Hansard from January 21st of 2003

APPENDIX A

Baroness Noakes asked Her Majesty’s Government:

Lord Hunt of Kings Heath: My Lords, the definition of a “bed” is published in the NHS Information Authority data
dictionary. The definition is used to aid statistical returns. No recent change has been made in the definition.

Baroness Noakes: My Lords, I thank the Minister for that reply. Will he confirm that, in NHS-speak, a “bed” can be a
trolley or even a chair? On 20th November 2002, he said that,

Was he referring to extra beds or to trolleys and chairs?

Lord Hunt of Kings Heath: My Lords, the definition of a bed is the same today as it was when the previous government
published it in 1992, when the NHS data dictionary was developed. As for the broad definition, to be a bed, the main purpose must be to
allow patients to lie down for rest or recovery. Beds, trolleys, couches and chairs used for treatments are not counted as beds. The noble Baroness was also quite wrong on actual bed number. For the first time in
many years, we have seen an increase in the number of general and acute beds. They are beds; they are not trolleys.

Baroness Gardner of Parkes: My Lords, will the Minister explain the statement in the newspaper this week that hospitals will no longer be allowed to rename corridors as wards?
Is there a general renaming process going on or being resisted within the department?

Lord Hunt of Kings Heath: My Lords, not at all. For statistical purposes and for the purposes of accurate returns regarding our accountability to Parliament, it is important
that NHS organisations have a consistent approach in reporting their statistics. We clarified the point that the noble Baroness raises in order to ensure clarity and consistency of reporting. Our
aim was to stop hospitals calling patients to say that they were admitted if they were not admitted to what could be described as a ward area. It was a case of tightening up the reporting
criteria.

Lord Walton of Detchant: My Lords, accepting the definition of a bed which the noble Lord has given, can he explain
why it is that three hospitals recently built under the private finance initiative, Cumberland Infirmary, Carlisle, the University Hospital of North Durham and Bishop Auckland General Hospital,
have many fewer beds than the hospitals they replaced?

Lord Hunt of Kings Heath: My Lords, overall the number of beds in general and acute categories is increasing. The
number of beds in individual hospitals is decided before it is decided under which route they should be financed. Therefore, it would be
wrong for my noble friend to draw a conclusion that PFI hospital schemes are by themselves leading to a reduction in bed numbers. We aim to
increase the number of beds. We are doing that. In the past two years we have seen the first increase in the number of general and acute
beds for many, many years. We are also increasing the number of day care beds. We
are expanding the NHS, not contracting it.

Lord Clement-Jones: My Lords, the issue of beds and their definition arose in the context of whether or not the
Government have achieved their targets on the elimination of mixed sex wards. Have the Government eliminated those wards? Have they achieved the targets that they have set out on many occasions and
which have slipped over the past few years?

Lord Hunt of Kings Heath: My Lords, 95 per cent of NHS trusts meet the additional criteria set for mental health facilities; 98 per cent of NHS trusts provide single-sex
sleeping accommodation for planned admissions; 93 per cent of NHS trusts provide properly segregated bathroom and toilet facilities for men and women; and over 98 per cent of NHS wards meet our
guidelines. The remainder will comply once current building projects are completed.

Lord Roberts of Conwy: My Lords, the noble Lord referred to the increase in the number of beds. Can he say what the
increase has been in the occupancy rate?

Lord Hunt of Kings Heath: My Lords, the latest occupancy rate that I have for general and acute beds is 86 per cent in
the year 2000–01. Some believe that the rate should come down. In guidance on emergency care the figure of 82 per cent is indicated as a reasonable figure. Clearly, we shall keep that under
review.

Baroness Strange: My Lords, can the Minister tell us whether a bed has to have four feet flat on the floor or whether
it can have wheels?

Lord Hunt of Kings Heath: My Lords, beds can have wheels as that makes them easier to move. I am tempted to read to
noble Lords the definition of a bed in the NHS data dictionary. However, I am conscious that some noble Lords wish to speak on other
matters this afternoon. Therefore, I shall desist.

Earl Howe: My Lords, is it not the case that a trolley need not in some circumstances be a trolley in an accident and emergency department and that a trolley can, indeed, be a
bed in a day care unit? Is it not true that a trolley can in fact mean anything that the NHS decides that it means?

Lord Hunt of Kings Heath: My Lords, the noble Earl is being a little unfair to the National Health Service. The whole point of having a bed definition is to achieve clarity and consistency of reporting. Beds in day units are
counted separately from beds in the general and acute categories. Clearly, in a day unit a patient may have pre-assessment, treatment, and
post-operative recovery in the same bed. That is why such beds are categorised
separately. I make it clear that trolleys in A&E departments are not defined as beds and nor are trolleys in treatment rooms and
wards.