I have some basic questions related to the data, if they were discussed somewhere else already, my apologies and please let me know.
\n1) How much is the standard measuring error on FVC value? (random error/statistical uncertainty)
\n2) How much variations can there be between two measurements (taken a few weeks apart)? E.g. If I were measured 2000 the previous one, could it be possible I get 3000 or 1000 on my next measurement? If so what are the normal causes? I read in a post mentioning that the patients in our data set were not on a anti-fibrosis therapy, but could they engage into some beneficial activities themselves, e.g. start jogging etc? (but I guess we have no way to know it here), or for a rapid drop, it might be due to a cold or something (again we have no clue from the data)
\n3) The time gap (Weeks) between each measurements were smaller/closer for weeks before the last three weeks, they were normally in 2~5 weeks range, but for the last three weeks, they were measured mostly 12~15 weeks gap, I would imagine that as the time gets closer to the end, there would be more frequent measurements, why the opposite?
\n@ahmedhshahin thanks in advance!
", "messages": [ { "id": "1010450", "postDate": "09/14/2020 18:50:05", "content": "I have some basic questions related to the data, if they were discussed somewhere else already, my apologies and please let me know.
\n1) How much is the standard measuring error on FVC value? (random error/statistical uncertainty)
\n2) How much variations can there be between two measurements (taken a few weeks apart)? E.g. If I were measured 2000 the previous one, could it be possible I get 3000 or 1000 on my next measurement? If so what are the normal causes? I read in a post mentioning that the patients in our data set were not on a anti-fibrosis therapy, but could they engage into some beneficial activities themselves, e.g. start jogging etc? (but I guess we have no way to know it here), or for a rapid drop, it might be due to a cold or something (again we have no clue from the data)
\n3) The time gap (Weeks) between each measurements were smaller/closer for weeks before the last three weeks, they were normally in 2~5 weeks range, but for the last three weeks, they were measured mostly 12~15 weeks gap, I would imagine that as the time gets closer to the end, there would be more frequent measurements, why the opposite?
\n@ahmedhshahin thanks in advance!
", "rawMarkdown": "I have some basic questions related to the data, if they were discussed somewhere else already, my apologies and please let me know.\n\n1) How much is the standard measuring error on FVC value? (random error/statistical uncertainty)\n\n2) How much variations can there be between two measurements (taken a few weeks apart)? E.g. If I were measured 2000 the previous one, could it be possible I get 3000 or 1000 on my next measurement? If so what are the normal causes? I read in a post mentioning that the patients in our data set were not on a anti-fibrosis therapy, but could they engage into some beneficial activities themselves, e.g. start jogging etc? (but I guess we have no way to know it here), or for a rapid drop, it might be due to a cold or something (again we have no clue from the data)\n\n3) The time gap (Weeks) between each measurements were smaller/closer for weeks before the last three weeks, they were normally in 2~5 weeks range, but for the last three weeks, they were measured mostly 12~15 weeks gap, I would imagine that as the time gets closer to the end, there would be more frequent measurements, why the opposite?\n\n\n@ahmedhshahin thanks in advance!", "votes": null }, { "id": "1010579", "postDate": "09/14/2020 22:19:24", "content": "The organiser suggest that the pure measurement error is 70 mL, but there is definitely more natural variation that that (even in clinical trials with very strict quality control on spirometry).
\nWho knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal). Additionally, I have the impression that some FVC results are the result of something going wrong with the spirometry (I'd guess it was just not done properly, but given that we do not know what quality control was done on the spirometry, it is hard to know exactly what is going on with some of the weird results).
\nOn an individual level, of course, a lot of fluctuations happen naturally. I would not be shocked, if you told me that results varied with standard deviation of 300 or more mL (even after conditoning on the first FVC value) - which suggests that you should get some results that are 2 times that far from the average trend of a patient.
\nRegarding the spacing of measurements, I have no idea what the underlying process is. However, note that the organizers mentioned that they did not give us the timing of the test set measurements, because that apparently conveyed predictive information. So, presumably, there is something interesting going on. I have no idea what though.
", "rawMarkdown": "The organiser suggest that the pure measurement error is 70 mL, but there is definitely more natural variation that that (even in clinical trials with very strict quality control on spirometry).\n\nWho knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal). Additionally, I have the impression that some FVC results are the result of something going wrong with the spirometry (I'd guess it was just not done properly, but given that we do not know what quality control was done on the spirometry, it is hard to know exactly what is going on with some of the weird results). \n\nOn an individual level, of course, a lot of fluctuations happen naturally. I would not be shocked, if you told me that results varied with standard deviation of 300 or more mL (even after conditoning on the first FVC value) - which suggests that you should get some results that are 2 times that far from the average trend of a patient.\n\n* I see that you saw a post about patients not being on medications, if so I missed that. Do you remember where that was? That would make a big difference and makes me wonder from what time in which country the data is (I think I saw some speculation that this is data from Europe - but especially in Western Europe, I'd have expected various medications to be used unless this data is somewhat older).\n\nRegarding the spacing of measurements, I have no idea what the underlying process is. However, note that the organizers mentioned that they did not give us the timing of the test set measurements, because that apparently conveyed predictive information. So, presumably, there is something interesting going on. I have no idea what though.", "votes": null }, { "id": "1011908", "postDate": "09/15/2020 18:44:32", "content": "\n\nWho knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal).
\n
Agree, although if that's the case then the effort of making predictions seems like a bit futile to me, we know in general the lung FVC decreases in the last three measures, but if there are other factors influencing it (e.g. maybe exercise, diet, rest, work condition etc) then we don't have ways knowing it, even if we make use of the noisy images.
\nSee the discussion here about these patients were not treated
", "rawMarkdown": "> Who knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal).\n\nAgree, although if that's the case then the effort of making predictions seems like a bit futile to me, we know in general the lung FVC decreases in the last three measures, but if there are other factors influencing it (e.g. maybe exercise, diet, rest, work condition etc) then we don't have ways knowing it, even if we make use of the noisy images.\n\nSee the discussion [here](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/178237#993891) about these patients were not treated", "votes": null }, { "id": "1012975", "postDate": "09/16/2020 12:36:35", "content": "Hi!
\nThank you for your questions!
I hope this helps. Again, I will seek a more accurate answer from our clinical leads.
", "rawMarkdown": "Hi!\nThank you for your questions!\n1. It is reported to be around 70 mL, can't get the reference now though.\n2. I am not a clinician so it is quite difficult to say. But FVC is known to have some noise and this is one of the drawbacks of the FVC tests. However, even with this noise and outliers that can happen due to several reasons (some of them are related to the patient as you mentioned), FVC remains widely considered for IPF diagnosis and is routinely used as a primary endpoint from a clinical perspective. Actually, this is one of the reasons that make us wonder if we can get a better estimate from the CT scans. I will try to get a more accurate answer from our radiology lead though.\n3. I think it can vary widely according to the patient's general state. This is one of the reasons that we didn't make the time of the last three visits available to your models. It could lead to information leakage (if this patient had several visits in a short time, this might mean that he is deteriorating. Models could use this information while it shouldn't, so we preferred to keep it hidden)\n\nI hope this helps. Again, I will seek a more accurate answer from our clinical leads.", "votes": null }, { "id": "1013089", "postDate": "09/16/2020 13:50:26", "content": "Hi Ahmed, thanks for your reply!
\nThe organiser suggest that the pure measurement error is 70 mL, but there is definitely more natural variation that that (even in clinical trials with very strict quality control on spirometry).
\nWho knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal). Additionally, I have the impression that some FVC results are the result of something going wrong with the spirometry (I'd guess it was just not done properly, but given that we do not know what quality control was done on the spirometry, it is hard to know exactly what is going on with some of the weird results).
\nOn an individual level, of course, a lot of fluctuations happen naturally. I would not be shocked, if you told me that results varied with standard deviation of 300 or more mL (even after conditoning on the first FVC value) - which suggests that you should get some results that are 2 times that far from the average trend of a patient.
\nRegarding the spacing of measurements, I have no idea what the underlying process is. However, note that the organizers mentioned that they did not give us the timing of the test set measurements, because that apparently conveyed predictive information. So, presumably, there is something interesting going on. I have no idea what though.
", "votes": null, "replies": [ { "id": 1011908, "author_name": "samshipengs", "author_url": "", "post_date": "09/15/2020 18:44:32", "content": "\n\nWho knows whether all of that is natural variation in the disease course from day-to-day, disease progression, the influence of medications*, other factors like other diseases a patient may have varying over time (various respiratory conditions are very seasonal).
\n
Agree, although if that's the case then the effort of making predictions seems like a bit futile to me, we know in general the lung FVC decreases in the last three measures, but if there are other factors influencing it (e.g. maybe exercise, diet, rest, work condition etc) then we don't have ways knowing it, even if we make use of the noisy images.
\nSee the discussion here about these patients were not treated
", "votes": null, "replies": [] } ] }, { "id": 1012975, "author_name": "ahmedhshahin", "author_url": "", "post_date": "09/16/2020 12:36:35", "content": "Hi!
\nThank you for your questions!
I hope this helps. Again, I will seek a more accurate answer from our clinical leads.
", "votes": null, "replies": [ { "id": 1013089, "author_name": "samshipengs", "author_url": "", "post_date": "09/16/2020 13:50:26", "content": "Hi Ahmed, thanks for your reply!
\n