Use the LIS to monitor TAT.
The Cerner Command Language (CCL) ad-hoc report writer lets us monitor our performance as well as analyze and respond to customer complaints quickly and effectively. The LIS's computerized date and time stamping of key events in the laboratory testing life cycle makes our data unquestionably objective.
QC applied to TAT analysis
Quality control principles serve as models for developing new service level measurement techniques. We adapted the QC concept of normal ranges to test TATs to benchmark them in much the way test results are benchmarked. Establishing "normal range" TATs make TAT goals - and performance measurements against them - a reality.
After we were able to extract, analyze, and plot the distributions of our TAT outcomes, we found no one outside the laboratory cares very much about the good outcomes. Our customers instead base their perceptions of our service to a large extent on the poor service outcomes. In typical "management by exception" style, they are concerned mostly with our TAT outliers, how frequently they occur, and how long they are overdue. We began searching for different statistical measures to use in creating our TAT "normal ranges." We always have recognized the laboratory management value of goal-setting and performance-monitoring against goal and so always have built these features into our programs. We established the 95th percentile TAT (which approximates the two-standard deviation mark in a normally distributed population) as a key benchmark for measuring our performance. The 95th percentile TAT has proven to be a good indicator of what our customers will perceive as the "standard" TAT for lab service. We began developing real-time computerized tools to identify turnaround time "outliers" as they occur and alert laboratory management to their existence.
Developing a simple, intuitive statistic that correlates well with our customers' perceptions is very important. Since our customers "manage by exception," quantifying outliers is key. We use a percentage goal to measure how many outliers occurred, and look for a completely satisfactory measure of performance on outlier TATs.
The statistic I currently use is called average lateness. It is calculated by summing the minutes between the goal time and the actual completion time for each test over goal. The sum is divided by the number of late tests to determine an average lateness. Rather than a ratio, users mentally evaluate the significance of the average lateness versus the average TAT. If this statistic is low, it implies the distribution of the late test population probably is clustered close to the goal. Unfortunately, the statistic still doesn't help clarify distribution pattern when average lateness is large.
Work remains to be done evaluating the distribution patterns that give the best and worst perceptions of our service. Naturally the best perception would come from a pattern with minimum lateness and with the late tests clustered close in time. On the negative side, would worse perceptions come from a pattern with wide variations in lateness or from a pattern of high clustering? Using my current statistic, both patterns could yield the same value for average lateness, but the customers' perception of our service might be different.
Data selection programs
After controlling our outliers, we sought better methods for pursuing our TAT analysis and reporting goals. We created three quick, flexible, and powerful on-line tools to use on an as-needed basis for assessing our TAT service. Each serves a specific purpose, but they all use a common structure. They:
* Stay out of the way until needed
* Allow flexibility in creating multi-variant data selection criteria based on frequently requested elements
* Where appropriate, display graphical and numerical representations of data, letting users specify goals against which to evaluate data
* Query against a special TAT database populated with data extracted from our LIS's files.
Operations trend viewer
This program [ILLUSTRATION FOR FIGURE 1 OMITTED] uses the specified selection criteria to extract corresponding data from up to a 22-day period. This capability lets us compare a particular day's performance against a prior day's, including comparisons against up to three previous occurrences of the same day of the week. This program is my most recent and most comprehensive. Display of the selected data can be switched among total, collection, and in-lab turnaround times. Five views of the data are possible: the range of TATs (from the minimum to the 95th percentile time), average times, mode TATs, percentages of goal, and average lateness.
Depending on the view selected, the data is displayed either in a column or an adapted high-low-open-closed graph format. Management goals can be expressed and measured either as TATs (e.g., two hours after an activity begins) or clock times (for example, complete an activity by 10:30). Goals can be established for the total, collection, or in-lab turnaround time. Goal lines are drawn on the display when the display mode (total, collection, or in-lab) matches the mode for which the goal was established. Graph markers are used to locate averages (A), medians (M), average lateness (L), and 95th percentiles (9) visually. The initial display shows the mode, which is the value experienced by the largest number of customers.
These programs are restricted to selecting data from a specific day. Data can be selected for collection or in-lab times, and are grouped by the hour collected or received in-lab. Graphs are created for each hour and the total. The graphs portray the distribution of tests that started their activity during the hour, spanning the range of TATs from the minimum to the 95th percentile.
Operations detail viewer
This program [ILLUSTRATION FOR FIGURE 2 OMITTED] is restricted to selecting data from a specific day. Line-item data can be sorted in the following sequences: accession number, drawn time, in-lab time, completion time, employee ID, and ordering location. Visual cues are provided to help quickly locate outliers. Line-item data can be printed and/or saved to a PC file.
Processing and data extraction
CCL was used to develop all the programs, which were written to run on the Digital Equipment (DEC) VT320 terminals that came with our LIS. The data is extracted from the LIS twice each day - at 3:30 a.m., when the prior day's tests are extracted, and at 1:30 p.m., when the current day's morning work is extracted.
The query programs were designed to take advantage of the quick, efficient data retrieval that is possible using key indexed files. The keys established for the TAT database were selected to support quick extraction based on our primary areas of concern as a hospital-based lab. Existing keys include test mnemonic, work center/testing site, ordering location, and employee ID. Others can be added.
Each program directs the user through the selection criterion fields at the top of the screen. The answers are used to create the multi-variant selection criteria. Based on the user's responses to requests for test mnemonic, testing site, work center, ordering location, ordering priority, hours of the day, day(s) of the week, specimen collection employee ID, and test verification employee ID, the program uses the speediest retrieval index key. Records are selected that qualify based on all the responses provided by the user.
We are pleased with the progress we have made in our quest to develop effective tools to use in monitoring and confirming we are meeting the turnaround time needs of our customers.
William W. Childress Jr. is financial manager in pathology and LIS specialist at the Baylor University Medical Center in Dallas.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||laboratory information systems; thematic apperception test|
|Author:||Childress, William W., Jr.|
|Publication:||Medical Laboratory Observer|
|Date:||Dec 1, 1995|
|Previous Article:||CLIA, stark rollbacks part of Medicare budget proposals.|
|Next Article:||Disenchanted bench tech shepherds 'professional' crusade for MTs nationwide.|