Treatment of Cancellations

[oscar peterson]

When projecting 'large' group medical costs, how should one handle claims experience for members who cancelled during the experience period?

One could argue that restricting the calculation to active members might predict future costs more effectively by aligning the experience period more closely with future circumstances. On the other hand, if cancelled members have systematically worse experience -- such that excluding them is actually material -- one should wonder whether the 'more morbid cancellation' group could be borne again next year out of the persisting, active, currently-somewhat-healthier members. Does that make sense?

Any thoughts?

[Dr T Non-Fan]

If it's legitimate for cancels having worse than average experience, then it is legitimate for cancels having better than average experience. You should do both or neither.

[oscar peterson]

Absolutely, the treatment should not be one-sided, if done at all. I suppose that one could back out the credibility-adjusted experience of cancelled members, in addition to lowering the credibility on those members who remain?

[Dr T Non-Fan]

The credibility of the remaining members should be automatically adjusted according to the remaining experience. In the extremes, if the departing members were 100% of the group's claims, then the remaining group should have a 0.0 credibility factor and would probably be priced with manual rating. If the departing members were 0% of the group's claims, then the remaining group's credibility would remain unchanged.
This would mean an additional line of explanation as to why the claim amount is different from other published and released results.

Another issue is the turnover rate. We're discussing members' dropping coverage, but what of their replacements, if any? Most companies have rules about not allowing new employees to have coverage, or that the rates will be repriced if there is some percentage change in the rating demographics. I think that a more flexible rating system that accounts for all the demographic changes in real time would be a lot more accurate and automatic. Problem with it is that the employer would never know the insurance costs from month to month, and it would encourage age/sex/area/health-status discrimination in real time, instead of getting shocked at renewal.