**HEALING SCORE – DEFINITION**

Like any index, the healing score should fulfill certain mathematical requirements, especially transitivity and proportionality to the measured outcome. The index should generate a scale that enables a comparison of improving and deteriorating health status. Its interpretation would be simple if a stable evolution corresponded to zero, and if a certain percent of improvement corresponded to the same percent of deterioration. Furthermore, the index should be robust; above all it should be independent of the scaling measure (months or weeks).

To comply with the required mathematical properties, SQLape^{®} computes the healing score H as follows:

H = log(I)/log(F),

where I represents the initial costs and F the final costs. To provide a robust measure of the initial cost, we computed the average cost of the first half of cumulated costs of each episode. Similarly, the final cost was defined by the mean monthly cost of the last fifth of cumulated costs.

The proposed index is independent of the measurement unit of the costs, since total costs disappear from the computation, being both numerator and denominator. It does not depend on the time scale, since I and F measurements are independent of the length of the episodes. In other words, the proposed index measures the evolution of the costs in the same manner regardless of whether the episode is short or long, and whether the illness is costly or not. The logarithm expression of the index warrants its transitivity. Moreover, it is centred on zero when costs remain stable (i.e. I = F). Finally, notice that improvement and worsening of health status is measured on the same scale (i.e. thank to the properties of the logarithm function).

Our working hypothesis is that, in a country with good access to healthcare services, healthy people will consume fewer health resources than people with poor health status. As a result, during the course of an episode of illness, decreasing healthcare costs would suggest that health status improves, while increasing costs would be correlated to a worsening health condition. In other words, costs variation with respect to time is a good tracer of the disease’s course. The main assumption underlying our working hypothesis is that physicians are parsimonious in providing care – proposing costlier services to patients with more serious conditions for a specific disease.

Costs are allocated to episode of diseases according to the following cost accounting method.