Week 10: Not-so-final Week

With my presentation less than two weeks away, my project is coming to a close. Sort of. Technically, this should be the last week of my internship, but I'm planning on coming to the clinic for the rest of April and May because we still need to submit our research paper. Dr. Zieman suggested that we try to contact the patients in our study to ask about their employment, housing, and marriage status. Considering the fact that we are trying to measure the impact of speech therapy on the lives of patients, I desperately want to include this question-and-answer section in our paper. However, I don't know how feasible this is because most of the patients have not had contact with the hospital in years (DV patients have been known to flee the state too). I won't have time to add this data to my school presentation either. It would be a nice addition, anyway. 

This week, I had some technical difficulties with my most recent data, and an entire day's worth of work was lost. Just swallowed up by the void known as OneDrive. I also observed a man who is currently under investigation for the disappearance of his girlfriend. You know, just another day at the office. 

It's kind of funny that all of these things seem so normal to me now after only a few months at the clinic. I've learned so much during my time here. First of all, I learned all about the complexities of traumatic brain injury and its symptoms (and I may or may not have begun diagnosing my family members with concussions whenever they complained of a headache). I learned just how pervasive domestic violence has become in our society, and how much of it goes unreported. I learned why patients might be considered "non-compliant" and how to interact with patients in general. I've gained some serious life skills here. Are there things I could work on? Definitely. But I have plenty of time to sort out my social ineptitude. I'm just glad that I have this experience under my belt now. I look forward to pursuing this research until the very end. Or at least until the paper is finally published.

Week 9

I'm observing patients again, which means I'm not stuck behind my computer all day! 

I mean, it's not like remaking spreadsheets is the absolute worst thing ever. I just don't want grid lines permanently branded into my eyeballs. 

Coincidentally, my very first patients were also here on the day I decided to stop being antisocial.  It was really gratifying to see how they've improved over the course of a few months. I also met some new patients and had a few... new experiences. For the first time, I was forced to leave the room while observing a patient. I didn't have a problem with it because I ask patients as soon as I meet them if they're comfortable with me in the room. Most of them agree, and this one did too. Unfortunately, she became very anxious in the middle of her evaluation and ran out to get some fresh air. It was at this point that I left because I didn't want to risk upsetting her further. I think this made me realize just how long-lasting the emotional trauma of domestic abuse can be. It was a bit unusual that I hadn't seen someone so sensitive yet, considering the subject at hand here. Most of my patients completely brush over the topic of domestic violence when they tell their stories, so seeing this woman panic was kind of a wake-up call. So many of these people are still suffering in silence...

As I'm writing both of my papers, I'm beginning to understand that the personal aspects of working with human subjects are lost in a research report. It's difficult to explain the downcast eyes, the fidgeting, the shame in a scientific journal. The social stigma and the sheer prevalence of domestic violence in the nation are the reasons why I decided to help this population in the first place. Meeting and connecting with these people only encouraged me to continue. I can't translate that into numbers. 

Week 8

I spent most of this week gathering background information for my paper(s) as Kristina reviewed my work. There isn't any prior research on this population, so my scope was a bit broader. Basically anything related to domestic violence, traumatic brain injury, or the relationship between mood and cognitive symptoms is fair game. Dr. Zieman and I agreed on our assigned portions of the paper, and, of course, the order of the authors' names. The latter of the two wasn't really important, but I thought it was interesting that this was even negotiable. 

The highlight of my week was spending time with the clinic's speech therapist. I was finally able to observe the speech therapy process with a few of her patients. This was actually incredibly helpful since I had no knowledge about the treatment beforehand. The first patient (Patient 1) returned for a follow-up visit, while the second (Patient 2) was evaluated for the first time. Both were tested on word association and attention. I'll start with Patient 1. For her first test, she was given a letter and 1 minute to come up with as many words as possible that started with this letter. The test was repeated two more times with different letters. Overall, there was improvement since her last visit. The next test involved filling in the blanks with letter combinations that complete each word. Each row had the same letter combination somewhere in the word (i.e. skirt, mask, risky). The last word association test was also a worksheet which involved filling in the blanks. Patient 1 would choose several words that related to a given word and correct her work after every row. The alternating/selective attention tasks were a bit more interactive. One was an app which required Patient 1 to choose the correct shape or color based on a command. For example, the app would display "Color: triangle" and Patient 1 would have to choose a green star rather than a triangle. In the final test, Patient 1 was asked to sort through a deck of cards by flipping any cards that were odd and red or even and black face-up. In most of the tests, Patient 1 realized that she improved her scores by talking herself through the task or writing reminders on a Post-it note. This is what is so effective about speech therapy; it trains patients to develop skills that are applicable to their own lives. Patient 2 was tested more holistically since it was his first evaluation. In addition to the other word association tests, he was required to associate words with images. For the first task, he was shown a series of images and was required to name them as fast as possible. In the next test, he was shown a symbol key, which stated the names for each individual symbol. He was then shown a phrase created with these symbols, which he would decipher and read aloud. For example, if ☼= "sun", ↑= "above", and ⚘⚘= "trees", then ☼↑⚘⚘= "sun above the trees". 

Example of Visual-Auditory Learning Test

In the next test, the speech therapist read several short stories to Patient 2, and he was required to recount the stories using as many of the same words as possible. He was also asked to recall a series of numbers backwards. The last test was confusing even for me. The speech therapist would read a phrase with objects and numbers. Patient 2 would have to repeat the objects first and the numbers second, but in the same order as the original phrase. For example, "frog two hat five" would become "frog hat two five". These are just a few of many tests commonly used in speech therapy programs. The point is that consistent practice is required for these cognitive tasks to have any impact on patient recovery. For obvious reasons, patients who never return after their evaluation don't benefit from the skills that are practiced in therapy. Perhaps the most interesting fact that I learned by observing these sessions is that the door is intentionally left open for each patient visit. This may seem counterintuitive (and in direct violation of HIPAA), but the speech therapist claims that "the real world isn't quiet" so her practice sessions shouldn't be either. I think this focus on real-world application is what leaves a lasting impact on patients. 

Week 7

Google is wonderful. 

With absolutely no experience in statistics, I had to start from the basics and research exactly what "generalized linear model" or "binary data" mean and how they're applicable to my data. After attempting a GLM with an Excel add-in, I quickly realized just how out of my depth I was. The add-in-- or the 14-day trial, rather --was actually produced by an independent company, so there were very few instructions available online. It turns out I was missing input values that were required to generate the model. Needless to say, I had trouble interpreting my data in this format. I talked it over with Kristina, and she advised me to scrap the idea because I didn't have all of the necessary information to create an accurate model. Instead, she suggested that I compare the slopes of the study and control groups, which are actually visible in my original graphs. Let's hope the slopes test works out better than this generalized linear mess.

In other news, Dr. Zieman will be presenting my preliminary data at a conference this weekend. She mentioned that she's worried someone might steal our data before the paper is published, which completely blew my mind because I couldn't understand what anyone would want with my measly Excel graphs. I still can't believe that my work is actually making an impact in educating other professionals. Dr. Zieman also proposed that I visit a shelter to meet with domestic violence victims themselves. I think it's important to understand the emotional and personal circumstances of people with violence-related TBI. It's easy to associate the patients with numbers and symptoms, since I don't even have their names. However, seeing what their lives are like day-to-day after recovery reminds me why I wanted to do this study in the first place. I haven't been to a shelter yet, but I was lucky enough to meet one of the patients from my study here at the clinic. She came in with her service dog (who was lovely, by the way) for one of her final follow-up visits. She was one of two patients who completed the speech therapy program by attending all 10 visits. After the usual assessment, Dr. Zieman asked her a few questions about how she thought her life had changed post-therapy. This woman, a former writer, had experienced a 30-year dry spell because of the cognitive deficits resulting from her injuries. Now, she's in the process of writing two books about her experiences, and plans to participate in NaNoWriMo in November. She claimed this was made possible largely by her speech therapy treatment. Sure, this was just one case where everything went according to plan. Not all patients react the same way. Yet, hearing her story made me realize the significance of this study. Quantitatively proving that these therapies are effective would also prove that these patients are worth the time and money necessary for further research. I usually end my posts by expressing my goals and hopes for this project, but this is the most important one by far. 

Week 6

I completely forgot to mention in my last post that our final paper will be submitted to a medical journal for publication. That sounds like something that would be impossible to forget, but I've been a bit scatterbrained (as usual). 

I met with Kristina, the statistician, this week to clear up some of my questions. Her verdict: I didn't do everything wrong. So that's always good news. On the other hand, I still have my work cut out for me with some of these individual symptom graphs. See, I can't just take the difference between the correlation coefficients and call it a day. I also need to decide whether the difference between study and control is statistically significant. Kristina recommended a Generalized Linear Model (GLM), which is used to analyze linear regressions like mine. Apparently, this model is commonly utilized by medical researchers who are studying extremely rare diseases, and thus are restricted to a very small population of subjects. This method will account for error resulting from my ridiculously tiny sample size, as well as the fact that there are multiple variables involved (symptom severity vs. number of visits, study vs. control, etc.). Typically, this model is created using an actual statistics program such as SPSS. I have Excel. I'll figure it out. 

Obviously, I want to analyze my data in the most accurate way possible. Regardless of my attempts to combat sample size error, the fact still remains: this study is underpowered. We knew this ahead of time. However, we're not looking for anything revolutionary. In reality, the only novel aspect of this study is the fact that no other prior research has been done on this specific population (domestic violence victims with TBI who undergo speech therapy). Our goal is to change that. We want to encourage other doctors to conduct further research on this population despite all of the other variables that make this research difficult. Most of all, we want to prove that the treatments provided here at BNI are even working. Hopefully, that's what this GLM will demonstrate. 

Week 5

This week, I've realized why (1) research is all trial and error and (2) small sample sizes are the bane of my existence. 

I finished up the graphs that I mentioned in my last post. I created individual scatter plots for my data, in which I listed number of visits (independent variable) on the x-axis and symptom severity (dependent variable) on the y-axis. Then, I generated a best fit line for each set of data to compare the study and control groups. Since speech therapy only addresses cognitive issues, I just designed graphs for each of the 8 cognitive symptoms, as opposed to all 23 physical, behavioral, and cognitive symptoms. Also, I created a separate graph for average cognitive symptom severity vs. number of visits. This provided a look at the overall improvement in patients for both the study and control groups. The only problem was that the control group (who received no therapy) appeared to improve more than the study group (who did receive some therapy). I took a glance at our raw data to see which study patients could be skewing the graphs, and it turns out one of the study patients only completed their first speech therapy visit. This visit is typically just another evaluation, so it really wouldn't have made much of a difference in symptom severity. Because I have such a small population, even one patient could affect everything. As a result, we decided to omit this patient, cutting down my sample size yet again. I know, I know. Reducing my 7-subject population even more sounds like the worst possible thing I could do... But it worked. After removing one patient from the control group-- just to even things out-- the graphs looked much better. 

Fig.1

 Fig. 2

Fig. 1 is before patient omission. Fig. 2 is after. (Note: The x-axis should actually be labeled "Number of Visits". Whoops.)

Recreating each graph over and over again helped me see how much patience is required in professional research (hint: it's a lot). I'm just glad that we managed to make our ridiculously small sample size work. Apparently this type of analysis is common with medical research because some extremely rare diseases only affect a tiny portion of the global population. It's nice to know I'm not the only one who's faced this challenge.

In addition to these graphical representations, I need a way to quantify the impact of speech therapy on DV TBI patients. I analyzed the slope and correlation coefficient for each data set to better understand this cognitive impact. The slope is pretty self-explanatory. Rise over run and all that jazz. The correlation coefficient, on the other hand, is a bit more complicated. This measurement, r, indicates the relationship between two values. It has a range between -1 and +1, where -1 shows a perfect negative correlation and +1 shows a perfect positive correlation. 0 signifies no correlation. In layman's terms, a negative correlation indicates that one variable increases as the other decreases (and vice versa). A positive correlation indicates that both variables will either increase or decrease. For our data, we hoped to see a negative correlation, in which symptom severity would decrease over time as patients continued to come in for speech therapy. We also hoped to see a significant difference between the study and control groups, demonstrating that speech therapy is actually worthwhile. The difference between the two is clearly visible in Fig. 2, but I need more concrete evidence than just "eyeballing" it. This is where I've reached the limits of my knowledge of statistics. Comparing the correlation coefficients for study and control involves a more extensive test, which I've never encountered. I contacted a local statistician to go over what I have. Hopefully, she will provide some suggestions for my data analysis methods. 

After finishing up at the clinic this week, I went with my mom to a meeting held by the National Association of Hispanic Nurses. Let me clarify: I don't plan on becoming a nurse. Still, I'm so glad that I went because some of the topics that were presented are absolutely applicable to the population that I'm working with. For example, one of the speakers discussed the term "non-compliant patient". In the eyes of most medical professionals, a non-compliant or non-adherent patient refuses to take their prescribed medication or follow through with their recommended therapy. Even I mentioned the word "refuse" in one of my earlier posts. Now I'm beginning to understand that there are other factors to be considered here. Lack of education, financial status, domestic problems, and other circumstances can prevent a patient from following their treatment regimen. On top of all of the physical and mental symptoms that arise from a TBI, DV patients have the added concern of their physical safety and/or the safety of their children. There's more to it than just a simple refusal. In fact, that's the very reason why doctors, social workers, and therapists are all necessary to treat these patients. Social and medical issues must be addressed together. 

Week 4

It's week 4 and I'm finally starting data collection.

Regardless of what my badge says, I don't have access to the clinic's database like all of the other employees because of my age. So, instead, Dr. Zieman and the clinic's program coordinator have to print off stacks (and I mean stacks) of patient information and white-out their names before giving them to me. I'm extremely grateful that they're willing to take the time to even do this. 

As far as the "screening" process goes, patients were chosen carefully by Dr. Zieman based on several criteria. Age was the biggest factor. We sorted through patient files to age-match each study group patient to another control group patient. This ensures that the mean age for both groups will not affect our results. Other factors such as history of substance abuse or history of mental disorder did not necessarily disqualify patients from the study, but they had to be considered. There are only 300 patients total that are treated here at the clinic, and only about 40 of those are reported domestic violence victims. Only 7 of the 40 DV patients have attended speech therapy more than once. 

This cuts down my target population to below 10, which is a problem.

It's a generally agreed upon fact that statistical validity is dependent on sample size. With such a small group (14 patients with both the study and control), I fear that my results will be inconclusive. Since I don't have the resources to expand my study, there's really no other option but to work with the population that I have. I'm hoping for the best. 

Actually compiling the data isn't too difficult. Essentially, I've been reading through each patient file and noting their age, sex, number of visits, and symptoms in an Excel spreadsheet. Their physical, behavioral, and cognitive symptoms were reported by the patients themselves using a severity scale (from 0 to 6). Each time the patients visited the clinic, they filled out this same symptom questionnaire. My spreadsheet records the symptoms on a 0 to 6 scale for each individual visit (for both the study and control groups). After completing this spreadsheet, I plan on graphing the data to compare the experimental and control groups.

Patient Symptom Severity Sheet

I haven't been living up to my role as clinical observer much this week because I've been busy collecting all of my data. However, I did get to shadow Dr. Zieman as she injected pain medication into her patient's skull. Apparently head injections can drive a grown man to tears. You learn something new every day.