• E-mail
• Print
• RSS

Select the right data for your ongoing records review

HIM Connection, May 4, 2003

Want to receive articles like this one in your inbox? Subscribe to HIM Connection!

Dear Colleagues:

When you are planning your ongoing records review program, you have to make decisions about the data you are going to study. You may select data are selected for various purposes, including the following:

• determining compliance with rules, regulations, policies and procedures
• determining whether there is a steady-state trend in the data, or whether there are unexpected changes, which could indicate trouble
• preparing reports on which administrative decisions are based

Data are used to draw conclusions. If the conclusions are to be valid, you should understand some basic statistical principles.

Most data are analyzed using some sort of statistical procedure. The mean (average) might be used to determine patient length of stay. There is interest in the proportion (percentage) of delinquent records. The statistical methods used for these parameters are based on the following two crucial assumptions:

• the data being studied are normally distributed
• the sample represents a simple random sample of the population under consideration

Normal distribution: Population is usually distributed according to the "standard normal distribution." Many people know this distribution as the familiar bell-shaped curve. Health information administrators who work with the data should be cognizant of this when working with statistical methods or procedures based on statistical methods. If you are unsure, seek out a colleague or associate who is knowledgeable in these things to help work on the data with you.

Simple random sample: The simple random sample implies scientific sampling. Scientific sampling involves setting up the group of records from which the sample is to be drawn, and then drawing the records from this population randomly. Performing this correctly requires a random number table or a random number generator program to choose which records will be in the sample. Not doing so admits bias into the sampling process (the choosing process itself may influence the outcome of the statistical procedures).

Bias refers to extraneous, unplanned, or unforeseen factors that in some way affect the results of the data. Here's one example: Suppose your organization includes a hospital, a physician practice, a rehab center, and a long term care facility. A questionnaire is prepared for the patients in the long-term care nursing facility to determine whether their level of satisfaction is equal to the care they receive. When the questionnaire is prepared, it is administered to patients who are elderly, and for the most part, not fluent in English but speak Russian, Spanish and Portuguese. The language of the questionnaire introduces bias. Likewise, if mainly English-speaking patients complete the questionnaire, it might not give a true or representative picture of the total population in the facility. It is comforting to know, however, that many statistical procedures can be used in confidence when working with your data. Here are some suggestions:

• Select your samples randomly.
• Work with proportions (percentages) as much as possible instead of means (averages) because statistical tests dealing with means also require knowledge of the population "variance." In most real-life situations, no one knows the population variance, so the results are guesses. Statistical tests done with proportions do not require knowledge of the population variance. Therefore, they produce reliable results.
• Ensure that your sample size is 30 or higher. Statistical research has shown that a sample size of 30 or higher permits valid results and conclusions even when the population is not normally distributed or the nature of the distribution of the population is unknown.

This week's HIM Connection was excerpted from the book and CD-ROM set, Automating Ongoing Records Review: New Strategies for JCAHO success. Designed to be used with basic software packages like Word and Excel, this resource will help you save time and improve your productivity and consistency through automating your ORR.

Click here for more information or to order your copy.

Sincerely,
Laura Motta
Editorial Assistant

Want to receive articles like this one in your inbox? Subscribe to HIM Connection!

• E-mail to a Colleague
• Print
• RSS
• Archive