In short, Numerical Text Analysis, Topic Analysis, and N-grams and Correlations would help us achieve the study goals. Although the way this data is presented and interpreted may vary greatly as the study progresses, the fundamentals should remain the same. Moreover, as we begin to collect data, what the analyses might look like will become clearer.
Numerical text analysis involves looking at the frequency at which certain words or phrases appear. The focus group script seems structured such that categories will form naturally as the information is collected (namely, questions one, five, seven, and nine seem to form these categories.) We may use a combination of factors to see what distinguishes practices that fall into any of the categories we decide. For example, we may look at the words/phrases that appear more frequently when participants answer question nine broken down by their answer for question one.
flowchart LR A[Question 1] --> B[Established clinic] A --> C[New clinic] A --> D[Up and coming clinic] A --> E[Idea in development] C --> |Question 9| F[outcomes, progression, prognosis, etc.] D --> |Question 9| G[cost, financial, startup, loans, etc.]
From the oversimplified example above, we were able to see that up and coming clinics use financial terms when answering what they would like to change about their practice, where new clinics use terms that pertain to patient outcomes.
Topic analysis is an unsupervised classification algorithm for documents, similar to clustering for numerical data. Latent Dirichlet Allocation (LDA) is one algorithm used for this purpose. LDA assumes that each document is a mixture of topics, and each topic is a mixture of words. Once the topic model has been trained, we can see words associated with each topic, and the topics associated with each document. Although topic analysis can be used to label and organize incoming documents, word-topic probabilities is what will be of interest to us. In particular, we can look at the log ratio of word-probabilities to tell which words drove the clustering process (i.e., the words that almost exclusively belonged to one topic over the others.)
At this time, I don’t think document-topic probabilities will significantly help us. In the context of this research, we could think of documents as answers to individual interview questions with a mixture of topics; however, the output won’t contain helpful information, as knowing the topic makeup of individual answers won’t help answer any of the research questions.
Here is an example from Text Mining with R of what this analysis might look like:
Topic 2 seems to have been formed by words associated with politics, while Topic 1 seems to have been formed by words associated with economics. Because these terms have such high probabilities of belonging to one group over the other, the Gibbs sampler likely established the topics around them. In a sense, this approach will end up giving us similar ideas as what word frequency would.
N-grams allow us to examine the co-occurrence of words. This is analogous to correlation for numerical variables. Calculating a \(\phi\)-coefficient is also possible, which is similar to the Pearson correlation coefficient for binary data. The \(\phi\)-coefficient is interpreted in the same way typical correlations do.
Here is a typical n-gram from Text Mining with R:
And here is what we could expect from the correlation
https://rpubs.com/chelseyhill/672546
https://www.tidytextmining.com/sentiment.html
https://bookdown.org/daniel_dauber_io/r4np_book/mixed-methods-research.html
Study the existing structure of multidisciplinary lupus nephritis clinics at various centers in North America, including barriers and strengths to implementation of such clinics. Data will primarily be qualitative in nature, obtained through interviews and questionnaires,
Lupus combined clinic focus group script:
Do you have a rheumatology-nephrology combined care lupus clinic at your center? Please self-categorize as: established clinic (\(\geq\) 5 years), new clinic (\(<\) 5 years), up and coming clinic (expected to happen within the next 12 months), idea in development (more than 12 months out), used to have one in the past but no longer do, have not been able to get traction for the idea, or have never tried to make it happen.
What healthcare professionals are part of your combined lupus clinic? 2a - Potential follow up question – what HCPs would you want in your lupus clinic?
What barriers, if any, have you faced in starting/running a combined lupus clinic? 3a - Will prompt with examples if needed: institutional buy-in, divisional buy-in, lack of staff, billing issues, EMR, time, financial…)
How have you addressed these barriers? 4a – Are there some barriers that you have been able to overcome? 4b – Are there some barriers that you have not been able to overcome, and why?
What is your clinic’s model? 5a – Does the clinic used “shared care,” or does each patient see the same provider at every visit? 5b – What do you see as the advantages of your model? 5c – What do you see as the disadvantages of your model?
Do you utilize a patient care navigator/dedicated lupus nurse for the combined clinic? If so, please describe this person’s role. 6a – If you don’t have a combined lupus clinic at your institution, do you have a patient care navigator/dedicated lupus nurse?
What patients are seen in your lupus clinic? (For instance – all lupus patients? Only LN patients? Only class 4 LN? Only complex LN?)
When you think about your lupus clinic, what works? What aspects of your clinic would you recommend to other lupus clinics? 8a – For those of you who don’t have a combined lupus clinic, have you found other ways to provide the same ancillary services to your lupus patients as you have heard described by those with a dedicated lupus clinic?
When you think about your lupus clinic, what does not work well? What would you like to change?
Do you have ongoing goals for your clinic? New elements you hope to incorporate over time?
Do you collect any outcomes data to measure the impact of your lupus clinic? 11a – Have you integrated research into your lupus clinic?
What outcomes, if any, do you believe are improved by your lupus clinic?