Presentation feedback from group #7

[Yukodeng]

The proposal is well-composed. Good job! Given Parkinson's disease ranked as one of the most common disorders in the US, the significance of the study is self-evident. And here are a few suggestions and comments on the specific aims.

For specific aim 1:

1. It’s great to see sensitivity analyses included in the study plan. It would be wonderful if any simulation method could be elaborated so the whole proposal would be in a more complete status.

Relating to aim 2:

1. I am wondering what is your outcome for this research question and how are you operationally defining it? Is it binary, i,e., disease present vs. no disease? I would suggest stating it more clearly in your revised proposal :)
2. This is a more high-level question. I see that you are assessing a combination of demographic, dietary, and lifestyle-related factors for risk prediction in patients with early-onset PD. However, I wonder to what degree these are actually predictive/ prognostic of the early onset of PD? Because from my (limited) knowledge family history or genetic mutations are major determinants of the disease onset. Perhaps you want to address this in your proposal rationale?

For the experimental approach:

1. I have a quick question on the evaluation metrics. Apart from the standard metrics you’ve mentioned, do you have any other tailored metrics in your mind as well?

[JingnanBai]

Thanks a lot for the suggestions and detailed feedback!

here are answers to aim2 part that:

Relating to aim 2:

1. I am wondering what is your outcome for this research question and how are you operationally defining it? Is it binary, i,e., disease present vs. no disease? I would suggest stating it more clearly in your revised proposal :)
2. This is a more high-level question. I see that you are assessing a combination of demographic, dietary, and lifestyle-related factors for risk prediction in patients with early-onset PD. However, I wonder to what degree these are actually predictive/ prognostic of the early onset of PD? Because from my (limited) knowledge family history or genetic mutations are major determinants of the disease onset. Perhaps you want to address this in your proposal rationale?

1. about outcome

Our outcome is binary (0-1) and it is relied on the professional diagnosis for each patient (which is reliable). We will clearify it in the introduction of the dataset.

2. predictive or prognostic features

Although the exactly cause of Parkinson's disease is not defined yet but it is well-known that it is related to a combination of various factors, including lifestyle, environment and genetics factors. Genetics factors may contribute a lot while other factors may also give some insights about the risk of Parkinson's disease. Good question and we will address more about these preliminary knowledge and information from literature review in rationale!

Thanks again for the suggestions and we would improve our proposal based on that :D

[JingnanBai]

• [x] update proposal to clarify outcome (in introduction part)
• [x] update aim 2 rationale part @JingnanBai

[QYHuang00]

Thanks a lot for the feedback!

here are answers to aim3 part that:

For the experimental approach:

I have a quick question on the evaluation metrics. Apart from the standard metrics you’ve mentioned, do you have any other tailored metrics in your mind as well?

We will be using standard metrics such as:

mean, median, standard deviation, frequencies

Accuracy, Precision, Recall, F1 Score, AUC-ROC

We will also use tailored metrics such as:

Time-to-Detection, Risk Stratification Performance, Patient Impact Score, Model Interpretability and Explainability, Cost-Effectiveness etc.

[kexinwang3]

Thank you for your suggestions!

In terms of sensitivity analysis for specific aim 1, the simplest approach we consider is to change one factor at a time while keeping others at their baseline to observe the effect on the output. This method, however, may not capture the interactions between variables. We may consider approaches such as changing the model structure and other avanced analysis to detect global sensitivity.