I am always keen to learn about new ways to enable me to examine how effective the processes I am part of are at achieving end results. When a process changes, does it change the outcome for the better, or for the worse?
Today saw me attend a basic 1 day course on “Statistical Process Control” (SPC) and I confess that I was very nearly put off by the title. I’m glad I went on to do the course as I think I learnt something new and useful.
What is it about?
SPC (at this level) described a simple method to examine common types of information available in a healthcare setting. By applying SPC to the information, it should be possible to understand:
- Why data varies – is it due to a special case, or are we describing “normal” variability?
- What does the information say about our performance against a given target?
- If we make a change to our processes, how does this influence our outcomes?
I used it to look at the last 25 people who attended for a “tier 2” assessment and were allocated a clinic appointment with a clinician to access opiate substitute treatment. The control chart (see below) shows the number of days between tier 2 assessment and clinic appointment on the “y” axis, with patient id number for consecutive patients on the “x” axis. The green arrow is a target of 21 days. The mean wait (green line) falls below this at 13 days and the 2 dotted red lines show the upper and lower confidence intervals.
Confidence intervals are calculated at 3 standard deviations from the mean. I understand that the consensus in SPC circles is that 3 standard deviations should be used so that it is very unlikely that special cases, those outside the red lines, are part of the normal population. In this example the confidence intervals extend below zero – clearly it’s impossible to see someone in the clinic before they have presented requesting treatment, so we ignore this.
Special Cases
Special cases are examples of variation that could require special explanation. The rules are as follows:
- Any point which lies outside one of the confidence intervals.
- A run of 7 points all above or all below the centre line, or all increasing/all decreasing.
- Any unusual patterns or trends within the control limits – this one is a bit hard, but essentially if there is change from a little variability around the mean to a lot of variability around the mean, or vice versa there is something funny going on!
- The proportion of points within the middle third of the region between the control limits differs excessively from the other two thirds.
In my data I did not have any special cases. If I had a point for patient 6 that was at 40 days, this would be a special case and would require explanation – for example the patient might have failed to attend the first clinic appointment they had been offered, perhaps they had been in hospital when they had had a triage assessment and their hospital stay had been for some considerable time.
How do we calculate the confidence intervals?
Find the mean for the values that you have collected.
Calculate the moving range values. This is the difference between a value and its preceding value.
Calculate the mean of the moving range values.
1 standard deviation = the mean of the moving range/d2 (where d2 = 1.128)
The confidence intervals are +/- 3 standard deviations from the mean.
Capability Index
This tells you how capable your current process is in relation to a target.
Capability index (Cpk) = (Target – Mean Value)/3 x standard deviation.
- A Cpk >/=1 indicates that there will be at least 99.865% inside the specified target.
- For a Cpk <1 you need to read off a % figure from a table.
- For a negative Cpk, less than 50% will be inside the target.
Some quick applications for this technique
- Examine waiting times for treatment using this approach and when nurse prescribers come into the system, see what difference it makes.
- Do the same and see what difference it makes when test on arrest starts up.
- Use it to look at numbers of referrals – does test on arrest make a difference?
- Look at hepatitis immunisations. What percentage of prescribed clients were immunised at least once in May, June, July, August etc.? Keep going, what happens when we have 4 trained nurses complementing the doctors in immunising people?
