Table of Contents
Administrative time is very important in medical records. It differs from measurement time by its lack of a close relationship with physical measurements. Beats per minute and isotope half life are typical of measurement time. The short story:
Patient was admitted at 6AM. Patient went into surgery at 7AM and out at 9AM.
has examples of administrative time. There are three point times identified (6AM, 7AM, and 9AM). There are also two time intervals identified: admission service interval (started 6AM, stopped unknown) and surgery service interval (started 7AM, stopped 9AM).
The proper recording and use of administrative time is different in important ways from the treatment of measurement time. IHE document sharing metadata makes extensive use of administrative time. The discussions around IHE ITI change proposal show that we lack a common understanding of administrative time.
Administrative time takes two forms:
- Points in time, e.g., report creation time.
- Intervals, e.g., the time spent in an operating room, expressed as a start and stop time.
These are recorded and compared when using medical records.
When a radiology report is given a creation time of 150303, that does not mean that it was created at that instant. It probably took a minute or two to create the report. That creation interval is not of administrative interest and is not preserved. The time assignment is somewhat arbitrary, but it’s usually within the interval of creation. (Note: all example times will use the ISO shorthand of "hh", "hhmm" and "hhmmss" notation for administrative time.)
Precision is important in administrative time and a fundamental part of administrative time notation. Administrative times come in a range of precisions: seconds, minutes, hours, days, months, or years. Lower precision points in time are defined in terms of sets. The point time 1503 represents a member of the set of all times in the range 150300 to 150359.999… .
Accuracy is rarely important for administrative time, and is not captured in document metadata. IHE ATNA specifies that all IHE systems shall be able to provide time that is accurate to 1 second. This does not apply to data provided by patients or other systems. ATNA does not require that precision be to the level of seconds. So it remains permissible to report something like admission time to just minute or hour precision. But NTP easily available and provides a typical accuracy of a few milliseconds.
The combination of accuracy and precision rules can have some unexpected results. An event that is reported as taking place at 153003 could actually have taken place any time in the interval 153002 to 153004.999… . Precision allows the time interval to be 153003.000… to 153003.999… and accuracy allows another second in both directions. The resulting range is 153002.000… to 153004.999… .
Since millisecond accuracy is easily available, the accuracy issues are generally ignored. The only issue that matters in practice for administrative time is precision.
Intervals are used when there is an administrative interest in the time interval during which something happened. Intervals come in three kinds:
- A fully specified interval, which has both a start time and a stop time
- A partially specified interval with only a start time. For example, when a patient has been admitted to the hospital and is still in the hospital, the hospital service interval has only a start time. There will eventually be a stop time, but it is not yet known. For various reasons this will be treated as having a stop time of "the end of time".
- A partially specified interval with only a stop time. This will similarly be treated as having a start time of "the beginning of time"
Precision applies to the start and stop time of intervals much like it does for points in time.
For the start and stop time, the actual interval started or stopped at one of the points in time that is found within the precision range of the start time. So that when an interval is documented as starting at 1500, it might actually have started at any time between 150000 and 150059. This also leads to the odd result that an interval that started and stopped at 1500 is meaningful. This interval took place entirely in the time interval between 150000 and 150059.
Accuracy is generally ignored for administrative intervals because it is an insignificant factor.
When comparing point time A with point time B there are three possible results:
- A is before B
- A is at the same time as B
- A is after B
Precision considerations make the rules somewhat complex. It is not a simple arithmetic comparison. It is more like a set operation.
Each point time can be considered to be the set of all times that are within the precision range. Then the rules become:
- A is before B if the intersection of sets A and B is null, and the individual times in A are before B.
- A is the same time as B if the intersection of sets A and B is not null
- A is after B if the intersection of sets A and B is null, and the individual times in A are after B
Using simple arithmetic can give the wrong answer because it ignores precision. You can use ordinary arithmetic comparisons, but you need to compare multiple times. The earliest possible member of A and the earliest possible member of B can be determined. Then the latest possible member of A and the latest possible member of B can be determined. Then these can be compared to determine the set operation results.
It is easy to make the mistake of using simple arithmetic because it is common to use phrases like "A is less than B" interchangeably with "A is before B". The strictly accurate terminology uses "is before" when dealing with administrative time. For the math geeks, administrative time is not even a strictly ordered set. This makes mapping it onto ordinary arithmetic treacherous. Precision variations result in non-ordered relationships like 15 == 1500 < 1504 < 1550 == 15.
When comparing point time B with interval A, there are five possible results:
- B is before A
- B is within A
- B is after A
- A is within B
- A overlaps B
Both precision and partially specified intervals affect the comparison. This is why it is best to use the "beginning of time" and "end of time" as substitutes for missing endpoints of intervals. If you do this, then the rules correspond to these set operations:
- B is before A if the intersection of sets A and B is null, and the individual times in B are before A.
- B is within A if the intersection of sets A and B is B.
- B is after A if the intersection of sets A and B is null, and the individual times in B are after A.
- A is within B if the intersection of sets A and B is A.
- A overlaps B if the intersection of sets A and B is not null, and neither A nor B.
One of the issues for the CP is deciding what to do if A is within or overlaps B. The technical framework does not discuss the situation where the time interval is shorter than the point time precision.
When comparing interval time with interval time there are five possible results. These results are not mutually exclusive:
- A is within B
- A overlaps B
- A is before B
- A is after B
- A is equal to B
Again, precision and partial specification make these set operations. The rules are
- A is within B is the intersection of A and B is A. If A is within B, then A also overlaps B.
- A is equal to B if the two sets are the same.
- A overlaps B if the intersection of A and B is not null.
- A is before B if the intersection is null, and the points in A are before the points in B.
- A is after B if the intersection is null, and the points in A are after the points in B.
IHE has a variety of queries. These may compare a point time with an interval, or they might compare an interval with an interval. The notation used for this has led to some confusion and is the source for the CP. The meaning for most of the comparisons has been:
- Match when the point A in the metadata is within an interval B in the query. Interval B may be partially specified.
- Match when the interval A in the metadata is within/overlaps an interval B in the query. Either interval may be partially specified. One clarification needed is whether the queries meant "is within" or "overlaps".
The discussion above shows that there are usage issues that must be clarified in addition to the notational considerations.
- When a point is compared with an interval in a query, what should be done for the situation where the interval is within the point or overlaps the point? What kind of problem will lead to comparing a more precise time interval with a lower precision point? The first example that I could think of is the query for patients born in January 1920. What should be done with a patient whose birth date is given as 1920? This patient could have been born at any time during 1920. The records are not more precise. Should this be considered a match? What are other cases?
- When intervals are compared with intervals, should the match be for "is within" or "overlaps". One example is the query for "documents about patients in surgery between 0800 and 0900". Should that include a patient who went into surgery at 0730 and left at 0830? What are all the other cases?
- Explain how to phrase the usage queries using the XDS (and other) query notations.
- If XDS (and other) notations cannot be used for all typical queries, consider the alternatives and decide whether to revise XDS requirements or add queries. One alternative is to explain how a different queries can be used together with client processing to accomplish the desired result. For example, the interval comparison result "is within" can always be split into a query for "overlaps" followed by client selection of the "is within" results from that larger set.