Week 4

Dealing with Data Issues

If you guys remember, last week I pretty much sat in front of a computer for 35 hours and wrote down the heights and weights of around 630 patients.  So basically, this week, along with some other exciting experiences, was mainly focused on "cleaning up the data".

So in the beginning of the week, I thought it would be a good idea to organize the data that had issues. And that means more Excel sheets! 

If you look back at the week 2 post, we had 4 main issues: missing data, data outside our time parameter, patients who died within 1 year of transplant, and patients who received multiple transplants. I made a spreadsheet for patients with each of the following issues. Each one of these spreadsheets has an "Excel Number" column, which corresponds to the row the patient info was in on the original data collection sheet. I also had to black out patient names for obvious reasons.

So this was the spreadsheet for patients who have received multiple transplants, including the transplant type and when this secondary transplant was received in relation to the one we care about. There are two main types of transplants: autologous and allogeneic. Autologous stem cell transplant involve giving a patient his/her own stem cells (which are first extracted then frozen) and allogeneic stem cell transplants are when you receive healthy stem cells from an outside donor. 

 It is not uncommon for a patient to receive an autologous stem cell transplant followed by an allogeneic stem cell transplant. Although patients who receive an auto transplant are still at a high risk for infection post-transplant, there are far less negative side effects when compared to an allo transplant (no graft versus host disease for auto transplants!). Because patients who receive multiple allo transplants are usually in a much worse state going into their second or third transplant, this might screw up our data. Therefore, if a patient received an auto transplant prior to their allo transplant, we kept the data. But if a patient had multiple allo transplants, we only used the data from their first allo transplant. This lowered the number of patients we had from around 633 to 533.

Next, we have the patients who were missing data. Sadly, there is no way to work around this so we were pretty much forced to remove them from the study. Since the data I was collecting was from 2008 to present, there were some patients who were missing 100 day of 1 year weights simply because it hasn't been 100 days or 1 year past there transplant. After taking these patients out, we went from 533 patients to 475.

 This was the chart for patients who had data outside out time parameter. I decided to include how far off the data was that way, we would know how far we had to expand out time parameter if we needed more data. Turns out, we have even after taking out these patients so we will most likely just not include them. This left us with 409 patients.

Finally we have patients who died early. I sorted out the patients who have passed into 4 categories as you can see in the spreadsheet. We decided not to take these patients out of the study. We can at least calculate pre-transplant BMI with all of these patients and if we see a trend among these patients, perhaps that would tell us we need to intervene if patient weights are too low pre-transplant.

Sorting the Data 

So the last thing I had to do this week was determine the different BMI classes and percent change in body weight classes. First, let's talk about BMI. I decided to use a pre-existing BMI classification that the World Health Organization uses. If you wanna read more about exactly how BMI is calculated click here.

 I broke down BMI using the 4 main classes (underweight, normal, overweight, and obese), but I didn't break down the data into those subclasses, simply because we don't have enough data to break it down that far.
Next, I had to break down the percent BMI change (this is actually the exact same as percent body weight change) and there is no "standard" for this. For now, I broke it down into X>=0%, 0%<X<5%, 5%<X<15%, X>=15%. I might have to adjust these depending on how much data falls into each of these categories.
I hope this longer post made up for the short post last week. Although I might do some work over spring break, I won't be posting again to week 5. Until then, bye!