You may have read recently about a novel measurement for healthcare value called the Healthcare Value Process Index (HVPI). 

The HVPI captures value according to Michael Porter’s definition in the New England Journal of Medicine for Healthcare Value. When we first put forward the new index, the team at The Surgical Lab acknowledged that some challenges exist regarding using the HVPI for continuous data. These included the situation where the data you have for system capability are not normally distributed. For more on that, take a look at the excerpt at the bottom of this entry.

Importantly, we suggested several ways to use the index with non-normal data including data transformation with a Box-Cox or Johnson transformation. 

We also described other alternatives to capturing outcomes with continuous data variables. One of the ones we considered was using the DPMO or Defects Per Million Opportunities. This would not be contingent on data normality and would represent those sub-optimal outcomes. For several reasons, including the fact that the DPMO would place focus on bad outcomes instead of positive performance, we did not put the DPMO forward.

In a recent discussion with other experts at the Lean Six Sigma World Conference, conversation returned to the DPMO, and inspired us to return to this metric so as to develop a more robust measure that would capture both non-normal and normally distributed data. 

In an upcoming entry, we revise the numerator of the HVPI to incorporate DPMO so that it can represent outcomes in non-normal and normally distributed data sets. Here we recommend changing the numerator to 1 minus the percent DPMO for your system.

This new index would read as (100)(1-%DPMO) / (COPQ in thousands), and would be workable for continuous data whether normally distributed or no.

It could even be made to work for discrete data...