Best method to calculate right skewed healthcare cost?

[Sujithpeta]

Hello,

I've patients healthcare annual cost and I would like to calculate the mean, but, the data is highly skewed to the right making the mean ~22 times the median. Below is the summary statistic for the data.

Mean =  $10,312, Median = $1497, SD = 23,970, Max = $242,502, Min = $0

I would really appreciate if someone can guide me to resource, preferably SAS code example. 

Thanks in advance!

[PGStats]

What properties do you want your measure of cost to have?

• meanCost has the property that meanCost * TotalNumberPatients = TotalCost
• medianCost has the property that Cost < medianCost for TotalNumberPatients/2 patients and Cost > medianCost for TotalNumberPatients/2 patients.

Other measures have other properties. It all depends on what you need.

[Reeza]

---

@Sujithpeta wrote:

Hello,

I've patients healthcare annual cost and I would like to calculate the mean, but, the data is highly skewed to the right making the mean ~22 times the median. Below is the summary statistic for the data.

Mean =  $10,312, Median = $1497, SD = 23,970, Max = $242,502, Min = $0

I would really appreciate if someone can guide me to resource, preferably SAS code example. 

Thanks in advance!

---

Code for what? 

In my experience when the data is that variable, we usually report three statistics, the Median, 50% above/below which is easy to explain and then the 90% and 99th percentiles. These tell you how much 90% of people are paying and 99% to understand the distribution some more. If you're in Canada check CIHI cost reports, not sure what US has. You can probably look at the CIHI reports to know what to measure. 

[Sujithpeta]

I’ve come across sources that point out using different methods like transforming the cost to log or use two part models. I’m looking for SAS code example for any of those methods.