This is the fourth article in a series about Clinical Service Quality Management in Health Call Centers.
Most organizations that are concerned with quality will conduct periodic reviews of their own work. Quality experts pan the concept of inspectors, for good reason: reliability should be built into processes. However, to build reliability into processes, techniques are needed to measure process characteristics.
Why 'Standard' Sampling Methods Miss the Target
In the clinical contact center world, a sample point is the patient interaction. The interaction consists of transaction data, a call recording and the patient’s experience. Often the sampling starts with the transaction data. This can be segmented and profiled in many different ways – by guideline, disposition, or call times to name a few.
Typically, the sampling methodology ranges from “let’s look at a bunch of stuff from some time period, group of people, type of work, or some combination thereof, to taking a random sample using a 95% confidence interval.
Intuitively, we lean toward getting a “representative” sample to audit an operation. A random sampling technique may be employed, pulling enough interactions to achieve a 95% confidence interval based on an estimated error rate. The sample audit “set” gets assigned to someone to pull every month, review the data, listen to the recording, and maybe call the patient. Problems are then reported back and corrective actions are taken.
This strategy may be more work and not as effective as we would like. Let’s state our intent and test it against our methods. For most of us, the intent is to find the problems and correct them before patient safety is impacted.
A confidence interval sampling strategy will produce a representative sample, given a normal distribution in the base data. The chance of getting samples that are “okay” is very high and the chance of finding “errors” is very low. In this instance, we really don’t want a representative sample, because the sample produces a low probability of answering the question: “Where are the problems that impact patient safety?”
Utilizing Control Charts to Identify Samples that Need Attention
One of the best tools to answer our question is the control chart. There are over a dozen types of control charts. Your measurement goals and data drive your chart selection. Control charts can be used on the full set of data for any period and category. For example, “disposition frequency” by nurse, time of day, or customer can be analyzed.
Generally, control charts segment data into four zones: one, two, and three standard deviations (sigma) from the mean and then everything else (the outliers). The chart is informing us about the variability of the process. The more variability (the larger the three zone) means less consistency. Low process consistency means a higher error rate. We assume the mean of the process is in compliance. Given this logic, we start analyzing the interactions furthest from the mean, which have the highest probability of being non-conforming. This methodology increases the probability of addressing our original question: “Where are the problems?”
Traditional Call Sampling and Auditing Isn’t Enough
In today’s world of high volume healthcare call centers, quality programs need to be highly focused. The sheer volume can be daunting: techniques are needed to see emerging problems in order take corrective action quickly and effectively. The control chart accomplishes this and is one of the many tools SironaHealth employs to ensure patient safety.
