Wes Weller recently developed a regression model for predicting health care costs that has as one of its variables the relative cost of each day of the week.
Below is a table showing how the predicted costs compare to actual costs (the R Squared is above .95 in this example). The relative costs by weekday for this data was:
Health Plans are often interested in knowing what the “actual” monthly health care costs are rather than “reported” costs. Removing the effects of the following variables is necessary before you can know what the actual health care costs are:
1. A variable number of work days or weekdays in the month
2. The effects of both major holidays and minor holidays.
3. The underlying seasonality of the month once the effects of items 1 and 2 above are removed.
4. Secular trend (which is much more measurable once items 1, 2 and 3 are accounted for).
In addition, it may be necessary to,
5. Remove prior month misstatements (e.g. errors in incurred but unpaid estimates).
6. “Shock” claims or other nonrecurring events such as epidemics, etc.
It’s often a waste of time to dissect “reported” costs without first removing these misleading influences.
Please give Wes a call if you are interested in talking to me more about this kind of thing.
(916) 505 – 2587