On the statistical validity of standards used in profile monitoring of health care.

In current methods of profile monitoring, standards of acceptability (cut-offs) are set either by consulting panels of experts, or by selecting an arbitrary point (e.g., the 75th percentile) on the profile (statistical distribution). However, experts have only vague ideas of what outcome rates ought to be, while profile statistics stem from samples for which unknown percentages of cases have received acceptable care. Poorly chosen standards could cause profile monitoring to be ineffective, inefficient, or unnecessarily disruptive. A new method proposes to set standards by using statistics for which the percentage of adequate care has been predetermined by examining the process of care. Plans to circumvent the pitfalls involved are described, as are two approaches to estimating the degree of process adequacy from routinely produced outcome rates.