I could not find information about whether the smoking status in the CSV is meant to represent patients' status as of the date when they had the baseline CV or when they were last checked by a doctor, ie their final FVC read - which can take up to a year or more from the baseline CT.
\nThis definition is relevant because a patient can quit smoking between baseline CT and the subsequent FVC readings (and actually they should! But that's another story….) and that could be correlated with the outcome.
\nSo I would be grateful to know from the Kaggle team the answer to the question above. Many thanks in advance!
\nPS.: fellow Kagglers, please feel free to point out to me the answer in case this was indeed provided in the descriptions and I missed it. Good luck to all!
I could not find information about whether the smoking status in the CSV is meant to represent patients' status as of the date when they had the baseline CV or when they were last checked by a doctor, ie their final FVC read - which can take up to a year or more from the baseline CT.
\nThis definition is relevant because a patient can quit smoking between baseline CT and the subsequent FVC readings (and actually they should! But that's another story….) and that could be correlated with the outcome.
\nSo I would be grateful to know from the Kaggle team the answer to the question above. Many thanks in advance!
\nPS.: fellow Kagglers, please feel free to point out to me the answer in case this was indeed provided in the descriptions and I missed it. Good luck to all!
Hi Douglas,
\nThanks for your interest in the competition. Smoking status is based on the patients baseline smoking status - which is the date of the baseline CT.
\nYour point is correct, but in practice, we generally don't control for changes in smoking status over the study period. In fact, unless smoking the focus of study, patients tend to be grouped as \"ever\" and \"never\" smoker.
\nI hope this helps.
\nBest,
\nSimon
@simonlfwalsh Thank you for the reply - it does clarify. I was afraid smoking, if measured later in the process (ie, after the baseline CT), could represent data leakage but I'm happy that it does not.
", "rawMarkdown": "simonlfwalsh Thank you for the reply - it does clarify. I was afraid smoking, if measured later in the process (ie, after the baseline CT), could represent data leakage but I'm happy that it does not.", "votes": null } ], "comments": [ { "id": 989052, "author_name": "simonlfwalsh", "author_url": "", "post_date": "08/28/2020 14:12:09", "content": "Hi Douglas,
\nThanks for your interest in the competition. Smoking status is based on the patients baseline smoking status - which is the date of the baseline CT.
\nYour point is correct, but in practice, we generally don't control for changes in smoking status over the study period. In fact, unless smoking the focus of study, patients tend to be grouped as \"ever\" and \"never\" smoker.
\nI hope this helps.
\nBest,
\nSimon
@simonlfwalsh Thank you for the reply - it does clarify. I was afraid smoking, if measured later in the process (ie, after the baseline CT), could represent data leakage but I'm happy that it does not.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "987016": "I could not find information about whether the smoking status in the CSV is meant to represent patients' status as of the date when they had the baseline CV or when they were last checked by a doctor, ie their final FVC read - which can take up to a year or more from the baseline CT.\n\nThis definition is relevant because a patient can quit smoking between baseline CT and the subsequent FVC readings (and actually they should! But that's another story....) and that could be correlated with the outcome. \n\nSo I would be grateful to know from the Kaggle team the answer to the question above. Many thanks in advance!\nPS.: fellow Kagglers, please feel free to point out to me the answer in case this was indeed provided in the descriptions and I missed it. Good luck to all!", "989052": "Hi Douglas, \n\nThanks for your interest in the competition. Smoking status is based on the patients baseline smoking status - which is the date of the baseline CT. \n\nYour point is correct, but in practice, we generally don't control for changes in smoking status over the study period. In fact, unless smoking the focus of study, patients tend to be grouped as \"ever\" and \"never\" smoker. \n\nI hope this helps. \n\nBest, \nSimon", "989140": "simonlfwalsh Thank you for the reply - it does clarify. I was afraid smoking, if measured later in the process (ie, after the baseline CT), could represent data leakage but I'm happy that it does not." }, "source": "meta" }