Frequently Asked Questions

About the Website

Using the Website

Data Analysis and Reports on MyMigraineJournal.com

About the Website


Who should use this site?

This site should be used by patients who have been diagnosed with migraine headaches, and who, with the advice of their doctor, are interested in creating a migraine journal that can be used to help pinpoint possible headache triggers.. It should not be used to diagnose headaches or any other disorder, and it does not prescribe treatments for headaches or provide medical care. Anyone whose pattern of headaches has changed or worsened should contact their personal physician immediately (and see here or here for headache warning signs).

Return to top

What are the advantages of creating a migraine journal with this online system, as against using the traditional pen-and-paper migraine journal?

We believes this approach has many advantages: (1) The system will automatically remind you every day (through email) to record the information that goes into the journal. (2) It will take less time to enter the information here than it would on paper. (3) You will have a neat and organized symptom calendar to review and share with your doctor. (4) The system will periodically compute statistics to help determine what triggers are associated with headaches. (5) As you add further data to the system, the system will regularly re-analyze the complete dataset so your results should become more accurate over time. (6) This website utilizes a ("hierarchical Bayesian") data-analysis approach which also allows the analysis of each user's data to benefit from the analysis of the total dataset of all users.

Return to top

How did this website come into being?

The website was designed by an experimental psychology professor-- Hal Pashler--who at one time suffered from a migraine condition and experienced first-hand the difficulty of trying to track down migraine triggers. Pashler uses PHP/SQL web technology in his research on human learning and memory, and has an interest in statistical issues. Over the past several years, he became intrigued by the potential of web technology to help people disentangle confusing cause-and-effect questions of daily life, including migraine triggers. The website benefited greatly from statistical insights and suggestions provided by Ed Vul and Mike Mozer. The website was developed by Mountaintop Data Mining, LLC, of Cheyenne, Wyoming.

Return to top

Are my data kept private?

Please see here for details of our privacy policy. We believe the policy is stronger and clearer than most health-related websites or other types of websites (e.g., you can have the system delete your account and your email from the server at any time). However, as noted in our Terms of Service, as a website operating with a modest budget, MyMigraineJournal.com cannot provide the sort of data security that one would expect at a bank or financial institution. Therefore, although it seems unlikely that any hacker would be interested in stealing files from MyMigraineJournal.com, it cannot be absolutely ruled out; thus, if you would view unauthorized access to records containing your migraine headache history and your email as a catastrophic event, please do not use this site.

Return to top

Is this an academic research project?

No, MyMigraineJournal.com is an online service, not an academic research project. It has no affiliation with any university, and does not collect the sort of detailed information on users that would be customary in an academic research study. However, as this website acquires more users, the MyMigraineJournal.com database may potentially have value for academic investigators, and to promote this, we will share the fully anonymized data (i.e., datasets without email, i.p. address, or other potentially identifying information) with any interested and qualified investigators from fields such as statistics, machine learning, and medicine. If this happens and any interesting findings emerge, we will inform our users.

Return to top

Does the website offer medical diagnosis or treatment services?

No, it does not. If you are using this site, you should already have been diagnosed by a physician as a migraine sufferer, and we recommend using the site with the advice of your physician. If you have any questions about the nature of your headaches, or if your headaches have changed or worsened, consult your physician (and see here for a discussion of potentially dangerous headache warning signs.) Naturally, if you use this site, you should continue to consult with your physician regularly for advice on management of your headaches.

Return to top

What pays for this website? Will the site always be free?

The site will be supported by ads and sponsored links (advertising things likely to interest people who suffer from migraine headaches). If this does not provide adequate funding to keep the website in operation, MyMigraineJournal.com may eventually charge a very modest user fee to new users. Even if this should happen, users who register now (in the beta-testing phase) are guaranteed free access to the site as long as it remains in operation.

Return to top

What if I have comments or experience problems with the website?

We are grateful to learn of your comments, suggestions, or any problems that you may run into. Please email suggestions to suggestions@mymigrainejournal.com and report problems to problems@mymigrainejournal.com.

Return to top

 

Using the Website


Do I really need to provide information on days when I don't have a headache?

Yes you do! It will not be possible to learn anything about what triggers are causing your headaches if you only fill out the questionnaire on days when you have a headache. To see why, suppose you ate aged cheese on 40% of days before you got a headache. If you ate aged cheese on about the same fraction of days not followed by a headache, this would obviously not imply that aged cheese is a trigger. On the other hand, if you ate aged cheese on just 10% of the days preceding a headache-free day, then this would strongly suggest that aged cheese is a trigger. For this reason, it is every bit as important to fill out the questionnaire on days when you have no headache, as it is on days when you do have a headache!

Return to top

How bad is it if I miss a day or a week?

Missing a few days here and there, or even a few weeks here and there, is not a very big deal. Of course, it means that it will take you longer to build up a statistically meaningful dataset, but it does not undermine the value of the data you already entered, nor of the data you will enter in the future. Thus, if you forget to use the system, become too busy to use it for a while, go on vacation, etc., don't worry about it--but do resume inputting data as soon as you can. To get the most value from the site, try to make it a daily habit to input the data--once you get a bit of practice, iit will probably take you only about 30 seconds per day.

Return to top

When I miss a few days, I end up with a bunch of emails from you in my inbox. If I want to fill in some of the missing data (because I remember what I ate, etc. on those days), where do I start? Do I have to work through these old emails one by one?

Actually, it doesn't matter which email you start with. Whether you click on an old email or on the newest email, the system will always start by asking you about the most recent day for which it lacks data. Then (without your needing to click on any more emails) the system will march back in time through the missing days, one day at a time, asking you if you want to provide data on each day in turn. If you ever say "no", that you don't want to provide data on, say June 15, then it will never again ask you about June 15 or any dates prior to that. The system assumes that if you can't remember what happened on June 15, you won't want to be bothered about June 14, June 13, etc. Of course, this may not always be the case. If for some reason, you ever do want to go back and add in (or, for that matter, change) information relating to earlier dates which the system has stopped asking you about, simply go to MySettings/Edit Daily Information and you can change or add anything you want.

Return to top

What if I go on vacation and won't have internet access? Can I suspend the daily emails?

Yes, log into the site, click on MySettings, and uncheck the button labeled "Send daily mail to me". If we get many requests for it, we may add a "Vacation Hold" feature with scheduled resumption (let us know if you would want this or other special features, by emailing us at suggestions@mymigrainejournal.com).

Return to top

What if I realize I made a mistake in some information I entered?

This is no big deal. Just go to MySettings/Edit Daily Information and you can change any of the information you entered.

Return to top

Where does the default list of trigger questions that you provide come from?

When you sign up, you are offered a suggested list of potential triggers. The list is a fairly comprehensive one, drawn from the entire migraine research literature. A subset of the items are pre-checked. The pre-checked ones are items which seem to us to be (a) most commonly believed by patients and scientists to trigger migraines, and (b) action-relevant (e.g., while there is good evidence that weather can trigger migraines, you probably cannot do very much to reduce your exposure to weather, so weather items are not pre-checked). We recommend including all of the pre-checked triggers in your daily questionnaire unless you have some reason not to, but of course this is completely up to you (you can check or uncheck whatever you want). During the sign-up process, you can also add up to 15 other potential triggers of your own devising, and these will be included in your daily questionnaire (note that the responses you will be offered are always from the same set: none, a little, a moderate amount, a lot, so you will probably want to word your questions accordingly).

Return to top

Suppose I am already actively avoiding something, e.g., aged cheese, because I strongly suspect that it triggers my headaches. Should I still include it among my daily Trigger Questions?

Our recommendation is: Yes, include it anyway. Of course, if you will never be exposed to aged cheese in the future, then there is no point having the system ask you about it. However, if you merely try to avoid it, but find that you eat it anyway from time to time (a common scenario), then you should include it in your list of Trigger Questions. If you do that, with continued usage of the system you should obtain a better and better basis for deciding whether to continue to avoid it.

Return to top

You say I can add new trigger questions or new symptom questions at any time. But suppose I add a new question after, say, 3 months of using the site--won't I lose the value of all the data I previously input before the change?

No, you won't lose any data when you add (or delete) a trigger question or symptom question. The system is designed to retain all data and analyze all relevant data.. Thus, if you add a new trigger question or delete an old trigger question, when your data is periodically analyzed, the analysis will proceed one trigger at a time, looking at all the data in the system that is relevant to that trigger. Depending on how many times you have answered a particular question, there may or may not be enough data to provide meaningful results about any given trigger, but adding or subtracting one item has no effect on how the rest of the dataset is analyzed.

Return to top

If I add new trigger questions or new symptom questions, do these get included in the statistical analysis that the website performs to look for triggering? And do they get included in the calendar?

If you add new trigger questions, these will be analyzed to see if they are meaningfully related to headaches. On the other hand, symptom questions play no part in any statistical analysis (regardless of whether they are items you checked off on our suggested question list, or questions made up by you). The statistical analysis attempts to predict the occurrence or non-occurrence of a headache, not which particular symptoms occur or do not occur. However, all symptom question responses are saved in the raw data record that you can download at any time (MyReports/Download Raw Data).

Return to top

How long will it take before I can learn something about what is triggering my headaches?

The system will send you by email an update on your data after several months of usage, and at regular intervals thereafter. However, we recommend that you plan to use the system for 3-6 months before expecting definitive results. How long it takes to get any reliable information about triggering depends upon many things, including the strength of any triggering effects that may exist, your rate of headaches, and the proportion of days on which you are exposed to any given candidate trigger. If the triggering is very strong and if you have frequent headaches as well as regular exposure to a given trigger, evidence for triggering could emerge pretty quickly. However, this will probably occur only rarely. Naturally, the need to collect data over a considerable time derives from mathematics (statistics), not from any limitation in the website: the same statistical issue would also prevent anyone from drawing reliable conclusions based on a paper-and-pencil journal used for the same period of time.

Return to top

Can the site help me figure out whether a potential trigger is not causing my headaches, and perhaps help me to stop avoiding things I need not avoid?

Yes. MyReports will report when it has become probable that a potential trigger is not causing any meaningful increase in your headache risk. If you decide to stop avoiding a potential trigger based on an analysis in MyReports, we strongly urge that you do not stop recording daily exposure to it. As you get more exposure to it because you aren't avoiding it anymore, your certainty about whether it really is a trigger or not will then continue to grow. But keep in mind that MyMigraineJournal only provides statistics, not medical advice; you will need to decide what to avoid and what not to avoid in consultation with your doctor and based on all available information.

Return to top

Can't you prompt me to fill in the daily questionnaire by some means other than email?

We may consider adding such features. If you would like to receive a prompt by some means other than email, please give us your specific suggestions at suggestions@mymigrainejournal.com .

Return to top


Data Analysis and Reports on MyMigraineJournal.com


What kinds of statistical analyses will the website do to try to figure out what triggers may be causing my headaches?

First, a nontechnical answer: Whenever the data are analyzed, the system examines each individual trigger factor, and basically asks the question: Are higher values of this trigger (e.g., eating more aged cheese) associated with a higher probability of getting a headache on the same or the following day? By assessing the degree to which such a pattern is found, together with the variability of the data, the system determines which of four situations holds, and reports back to you accordingly: (1) the data indicate that this trigger factor probably is raising your headache risk to a meaningful degree; (2) this trigger factor probably is not affecting your headache risk to a meaningful degree; (3) this trigger factor is probably lowering your headache risk to a meaningful degree (although this sounds implausible, protective effects should not be ruled out in advance); and (4) there are insufficient data to conclude that any of the first three situations holds, in which case you need to use the system more before a meaningful judgment can be provided about this potential trigger.

Now a more technical answer for those interested: separately for each candidate trigger, a logistic regression analysis is carried out on all data pairs of the form <trigger_i, headache_i> (where trigger_i means "trigger exposure on a weighted average of day i and day i-1" and headache_i means "headache on day i". This analysis yields a beta coefficient and a standard error of the beta coefficient for each trigger. Using these two quantities, we carry out parametric bootstrapping to estimate the probability distribution of the difference between p(headache_i | maximum exposure to trigger on day i and day i-1) and p(headache_i | minimum exposure to the trigger on day i and day i-1). We set a minimum threshold for a practically meaningful difference, and report that a triggering effect is probably present if, given the data, there is a 75% or greater chance of an effect equal to or exceeding this minimum threshold. The system reports that a candidate trigger is probably not affecting your headache risk if there is a 75% chance that the difference does not exceed the threshold. When neither an effect of a trigger (positive or protective) nor the probable lack of such an effect can be concluded, the software tells you you need to use the system longer before any judgment can be made about the trigger factor.

Note that the bootstrapped distribution of the probability difference described above, which serves as the basis for the summary conclusions provided to users as described above, is a Bayesian posterior distribution conditioned on the data assuming a uniform prior probability distribution of beta. However, once there is an adequate quantity of data from multiple users who have provided data on the same trigger, the code computes the prior distribution for beta from the posterior distributions derived from the aggregated population of previous users (assessed in periodic batch analyses), rather than assuming uniform priors. This improves performance, by allowing the analysis at the individual level to be informed by the regularities contained in the entire user database, without assuming that triggering operates in a homogeneous fashion across people (see Gelman & Little, Survey Methodology, 1997, 23(2), for a hierarchical logistic regression model applied to polling data-- this paper partly inspired the design of the current system).

Return to top

Do you look for complex effects of triggers ("interactions") or long-range triggering?

Short answer: Not yet, but we hope to do so before long (and there is an important reason to wait a while). More detailed answer: The initial build of the website looks for single-trigger effects on headache risk that occur the same day or the next day after exposure to a given trigger. When we have a sufficiently large total dataset, we intend to explore the possibility of longer-term triggering and interactions among triggers (e.g., the possibility that some users might have headaches when exposed to fatigue and dehydration, but not from either one alone). If we find evidence that such effects are real, we will extend the hierarchical logistic regression approach described above to assess the strength of such effects within individuals. We approach this issue cautiously because (a) we are not aware of any published reports that assert the existence of long-range or interactive triggering (although we are also not aware of any credible efforts to test for this), and (b) while it would not be hard for us to test for such linkages within each user's individual dataset, in our judgment this would run an unacceptably high risk of detecting illusory patterns by chance ("false alarms" or "type 1 errors"). As much as possible, we want our technology to help people avoid falling for spurious correlations--not to promote them.

Return to top

What will the reports I get from MyMigraineJournal look like?

After several months of usage you will receive an email from admin@mymigrainejournal.com (with the subject header "Report from MyMigraineJournal") reporting the results of an analysis of your entire dataset. The report will provide summary statistics, including the overall proportion of days that you have had migraines. For each trigger, the report will indicate the observed difference between headache rate following exposure to the trigger versus overall headache rate, and indicate whether the data are reliable enough to indicate that a real effect is likely to be present (see the question immediately above this one). Here is a sample of what a report looks like.

Return to top

What if I want to do my own data analyses?

The website was designed to empower you to do whatever you want with your own data. You can download the complete daily record of symptoms, triggers, headache narratives, etc at any time (MyReports/Download Raw Data). The data are provided as a Microsoft Excel spreadsheet; from there, you should be able to convert it into any other format you want. If you find anything interesting, please let us know.

Return to top

Whenever I get a headache, the system asks if I want to make any narrative comments about the headache. What do you do with this text information? Do you analyze it statistically?

The system doesn't analyze the text, but merely includes it in your raw data file that you can download at any time (My Reports / Download Raw Data).

Return to top

A technical question: What do you do about possible colinearity of predictor variables?

This is an interesting issue. Let's take a concrete example. Suppose you eat rye crackers every time you drink red wine, and suppose red wine causes you to get migraines. If you were to include rye crackers among your potential triggers, the statistical analysis performed by MyMigraineJournal.com would report back to you that rye crackers (as well as red wine) are likely triggers for your headaches. There are statistical methods, e.g., stepwise logistic regression, that can be helpful in distinguishing a "mediated" link (the crackers) from a direct causal linkage (the wine). Currently, the site does not use stepwise methods because (a) they are likely to require an impractical amount of data from each user, and (b) it seems to us that the cost of potentially detecting mediated as well as direct causes is relatively minor (in the example given, this would likely mean that you would avoid both red wine and rye crackers.) Moreover, we suspect that the only way to tell conclusively which of several correlated variables is causing your headaches would be to run experiments in which you expose yourself to one trigger and then the other; we doubt you will want to run such experiments.

Return to top

Miscellaneous


Do you have any plans to extend this technology beyond migraine headaches?

Yes. We are building another more general-purpose system that allows a user to record the occurrence of any problem they want--whether it be a child's behavioral problem, a gastrointestinal complaint, fights with their mother-in-law, difficulty sleeping, or anything else--along with whatever daily events they think might possibly be triggering the problem. The statistical data-mining technology will be similar to that offered here on MyMigraineJournal.com, but it will allow more flexibility and customization.

Return to top

Can you guarantee that using MyMigraineJournal.com will help me avoid headaches, or at least stop avoiding things that I don't need to avoid?

No. It may be that your headaches are not triggered by any common migraine triggers.  And even if they are, using this site may not succeed in identifying them. This site should allow you conveniently to build an accurate record of headaches and triggers, and we believe that the statistical analysis offered here is probably more accurate and powerful than what you can readily obtain by any other means. However, it is much too early to say how often users will be able to translate this into fewer headaches, or less unnecessary avoidance of things that needn't be avoided.

Return to top

What if the site goes out of existence? Will my data suddenly disappear?

First, we have no present plans to terminate the site. Second, as mentioned above, you can download all your raw data in Excel format (triggers and headaches) at any time you wish (MyReports / Download Raw Data). That way, you can keep everything archived on your own computer. Third, if we ever take down the website, we will try to notify all users in advance and give them a chance to download their data beforehand.

Return to top

Health-related information on MyMigraineJournal.com is not offered as a substitute for medical advice or treatment for any medical conditions. You should promptly seek professional medical care if you have any concern about your health, and you should consult your physician before making health-related decisions based on the information here. Copyright © 2009 MyMigraineJournal.com. All rights reserved.