- EpiData-list - University of Manitoba Mailing Lists

Re: Outbreak analysis diary - attack rate table
by epidata-list＠lists.umanitoba.ca 08 Jun '08

08 Jun '08

thanks very much Jens, Hans and Jamie for the suggestions and Alistair for the encouragement. Hans thanks I tried your suggestion for the browse and it worked well. The other suggestions will take a bit longer to work through. I will report back when I do but this post I want to report on the analysis and also the success of creating and combining two data entry files. Attack rate table of exposures As has been said several times on this list but is worth saying again, the 'outbreak analysis' function in Epidata that creates a single attack rate table showing attack rates, relative risks, CIS and pvalues of all exposures of interest against the outcome of 'case' is a wonderful feature. The command is TAB CASE .....all exposures of interest/OA The steps I took were: 1. Click on Browse Data from the menu bar 2. Tick my outcome of interest (Case) and all my exposures of interest (foods etc) 3. Click on Paste (this copies all the selected variables into the section at the bottom of the output screen) 4. I then copied this line into the programme editor, took out the browse at the start and added/OA at the end as below: TAB CASE ANYFOOD COURTL1 COURTL2 JUNGLEPLAY JTOILET BTOILET KIOSK M1NUGGETS M2CHEERIOS M3CHIPS M4SAUCE M5PIZZA M6JUICE M7JELLYICE M8YOGURT M9CHOCCRAC M10CAKE M11LOLLIES M12TEA M13COFFEE M14SAV M15SAND M16HOTDOG M18FISHBIT M19FISHFIL M20FRUITSA M21DANISH M22BURGER/OA Note: Adding EX after OA gives the pvalue I then selected run programme - this creates the attack rate table in the output screen, and also in the default output file. I really like the outbreak analysis feature but on my wish list would be 1. for 'outbreak analysis' to be an option in one of the menus somewhere. I only knew about this feature through someone mentioning it on the list and I had to search through the docuentation to find it. Maybe this was what the 'attack rate plus RR' option was but I could not seem to select this - it stayed greyed out. 2. If a select' has been used for this to show above the attack rate table. I ran the outbreak analysis above again but this time selecting only people who had attended the centre on a certain date. The command I used was Select Day25 = TRUE I know the select has worked because N has gone down from 251 to 66, but there is no indicaiton on the output file or the information under the attack rate table that a 'select' is in place. Thank you all again for your feedback on these postings. I will do one maybe two more - covering the process of sending the data output to a report and a final summary of the main steps and final programme I used. Christine Having more than one data entry file We ended up with 250 questionnaires so decided we would try using several data entry operators to enter the data Christine Roseveare Analyst Regional Public Health 570 9194 027 495 9671 >>> <epidata-list(a)lists.umanitoba.ca> 6/06/2008 8:56 a.m. >>> Dear Christine: Just a quick comment on BROWSE. Did you use the command: set display databrowser=on before issuing the BROWSE field1 field2 command? The default is that the browser is off. Note that this works fine with the public release version, but if you experiment withe (marvelous) test version build 134, there seems to be some hickup that occasionally blocks the program, but I believe to have noted that it works if the command is in the beginning of the program before reading the file (not yet certain). In any case, I appreciate your telling us your experiences, they are useful for many. Apologies for not taking the time to answer to more of your queries, but I am currently a bit taken with traveling between continents. Regards, Hans epidata-list(a)lists.umanitoba.ca wrote: > Today we started doing the analysis of exposures for our outbreak. First we need to create a new variable, as not everyone who is sick meets the case definition for the outbreak. > > Case definition > Our case definition is someone who has diarrhoea and/or vomiting 24-72 hours after the function. > I used the programme editor to define a new variable called Case > I made case a Boolean field so the relative risks would calculate correctly (with a Boolean field the Y appears first in a two by two table and then the N and this is what you want to calculate relative risks correctly). > > Diarrhoea and Vomit - the fields I needed to create the Case field were also Booklean fields. For a long time I had a problem as I was trying many combinations of If Diarrhoea =es or Y or "Y" in my code and it was not working. Then my workmate suggested I try TRUE. I think this is confusing as when you print a Frequency of a boolean Field it prints as Y. Anyway here was the code that worked. > > Define case <Y> > If Diarrhoea =RUE or Vomit = TRUE then case = TRUE > > This deals with the first part of the case definition - that they must have diarrhoea or vomiting. The next part is that they are NOT a case if the symptoms started on the same day. > > The code I ended up using for this was IF onsetdate <=ate("24/05/2008") then case = FALSE. I had to put the date function as well as the date itself or the code would not run. I don't know why that is but any other combination I tried like "24/05/2008" or 24/05/2008 or ("24/05/2008") did not work. > > Onsetdate > In the last post I asked if anyone had some code to create a single onset date from two pieces of information - the vomitdate and the diarrhoea date. My workmate helped me figure out something that worked and this is it. > > If diadate < vomdate then onsetdate =iadate > If vomdate < diadate then onsetdate =omdate > If vomdate =adate then onsetdate = diadate > If vomdate =) then onsetdate = diadate > If diadate =) then onsetdate = vomdate > > Now we have a variables of case and onset date and I created an Epicurv graph using them. It was great. > One more frustration. While we were working on the code we often used browse to check our results. We wrote the browse command into the programme editor but the results of the Browse do not appear in the output so I had to do it over again each time in the output screen. I tried list instead of Browse but this didn't work either. Any suggestions for how you can build the output from a browse or line list command into a programme will be welcome. > > Next posting we will get on to analysing the exposures > > thanks > > Christine > > > -- Hans L Rieder, MD, MPH Jetzikofenstr. 12 3038 Kirchlindach Switzerland Tel: +41 31 829 4577 Mob: +41 79 321 9122 Web: http://www.tbrieder.org -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 1

Epidata analysis on a network with restricted access to c drive
by epidata-list＠lists.umanitoba.ca 08 Jun '08

08 Jun '08

Just a quick question re an issue I have with Epidata analysis and the C drive default directories it uses for some functions. This is the message I get when I load Epidata Analyis * EpiData Analysis default settings file * Edit the next lines to change size or font . set echo=off Access is denied Operation aborted Although I can still use Epidata Analysis I do not get the option to change the settings, I don't see the results of the read command displayed or information about how many records are in the file. I also have to change the default I send the output to as I can not access the default directory. I asked our IT help desk and they say the reason this happens is that I am now allowed access to one of the directories Epidata looks for - an administrator folder in the C drive. I wonder if this might be a problem for others who are in organisations/working on networks where users are not allowed access to administrator folders. Do you have a suggested workaround? thanks Christine Christine Roseveare Analyst Regional Public Health 570 9194 027 495 9671 >>> <epidata-list-request(a)lists.umanitoba.ca> 7/06/2008 5:04 a.m. >>> Send EpiData-list mailing list submissions to epidata-list(a)lists.umanitoba.ca To subscribe or unsubscribe via the World Wide Web, visit http://lists.umanitoba.ca/mailman/listinfo/epidata-list or, via email, send a message with subject or body 'help' to epidata-list-request(a)lists.umanitoba.ca You can reach the person managing the list at epidata-list-owner(a)lists.umanitoba.ca When replying, please edit your Subject line so it is more specific than "Re: Contents of EpiData-list digest..." -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 0

Outbreak analysis diary - Defining cases as a new variable
by epidata-list＠lists.umanitoba.ca 06 Jun '08

06 Jun '08

Today we started doing the analysis of exposures for our outbreak. First we need to create a new variable, as not everyone who is sick meets the case definition for the outbreak. Case definition Our case definition is someone who has diarrhoea and/or vomiting 24-72 hours after the function. I used the programme editor to define a new variable called Case I made case a Boolean field so the relative risks would calculate correctly (with a Boolean field the Y appears first in a two by two table and then the N and this is what you want to calculate relative risks correctly). Diarrhoea and Vomit - the fields I needed to create the Case field were also Booklean fields. For a long time I had a problem as I was trying many combinations of If Diarrhoea = Yes or Y or "Y" in my code and it was not working. Then my workmate suggested I try TRUE. I think this is confusing as when you print a Frequency of a boolean Field it prints as Y. Anyway here was the code that worked. Define case <Y> If Diarrhoea = TRUE or Vomit = TRUE then case = TRUE This deals with the first part of the case definition - that they must have diarrhoea or vomiting. The next part is that they are NOT a case if the symptoms started on the same day. The code I ended up using for this was IF onsetdate <= date("24/05/2008") then case = FALSE. I had to put the date function as well as the date itself or the code would not run. I don't know why that is but any other combination I tried like "24/05/2008" or 24/05/2008 or ("24/05/2008") did not work. Onsetdate In the last post I asked if anyone had some code to create a single onset date from two pieces of information - the vomitdate and the diarrhoea date. My workmate helped me figure out something that worked and this is it. If diadate < vomdate then onsetdate = diadate If vomdate < diadate then onsetdate = vomdate If vomdate =diadate then onsetdate = diadate If vomdate =(.) then onsetdate = diadate If diadate =(.) then onsetdate = vomdate Now we have a variables of case and onset date and I created an Epicurv graph using them. It was great. One more frustration. While we were working on the code we often used browse to check our results. We wrote the browse command into the programme editor but the results of the Browse do not appear in the output so I had to do it over again each time in the output screen. I tried list instead of Browse but this didn't work either. Any suggestions for how you can build the output from a browse or line list command into a programme will be welcome. Next posting we will get on to analysing the exposures thanks Christine -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 2

re: ongoing outbreak
by epidata-list＠lists.umanitoba.ca 05 Jun '08

05 Jun '08

I think this opportunity for open sharing of ideas and experiences is very good and unique In general I discourage the use of boolean variables. In particular since these in many softwares can have only the values true (Y) or false (N), and not the value "no value". With the introduction of value labels it is far more efficient to use integer variables, e.g. for the current outbreak: qes file food items intaken : v1 food item 1 (e.g. soup) # v2 food item 2 (e.g. vanilla icecream) # etc Preparation of chk file is then easy. 1. add checks (#3 in entry) 2. define a value label (+) for v1: 0 no 1 yes 9 missing 3. further define "9" as missing for v1: F9 (edit) and add missingvalue all 9 3. after defining that go to field 1 and press ctrl+c 4. go to v2 (and all other food items) and press ctrl+v now all fields have 0 and 1 defined and 9 as missing value. 5. enter your data 6 read data into analysis define cases: strategy 1: define case # case = 0 if (vomiting = 1) then case = 1 * remove those vomiting before 24 hours or after 72 hours if (vomiting=1) and (hourvomit < 24) then case = 0 if (vomiting=1) and (hourvomit > 72) then case = 0 * alternatively if we have a fixed date of exposure on 21st of may) if (vomiting=1) and (datevomit < dmy(22,5,2008) then case = 0 if (vomiting=1) and (datevomit > dmy(24,5,2008) then case = 0 * all with diarrhoea are cases: if (diarrhoea = 1) then case = 1 * date of symptoms: define dateonset <dd/mm/yyyy> If diadate < vomdate then onsetdate = diadate If vomdate < diadate then onsetdate = vomdate If vomdate =diadate then onsetdate = diadate If vomdate =(.) then onsetdate = diadate If diadate =(.) then onsetdate = vomdate * now we are ready for the first simple analysis: tab case v1 v2 v3 .... etc - all exposure variables..... /oa /t epicurve onsetdate case epicurve onsetdate case /by=sex The problems reported the latest days, such as: using boolean variables changing from string to numerical etc. a Can be handled by combining entry and analysis - as was shown on the list by others: a. rewrite the qes file and some conversions are done by EpiData Entry when possible b. take the file into analysis - generate the new numerical variables, e.g. gen i v1 = (upper(v1string) = "Y") c. save the data (savedata newfile) d. with entry create a qes file from the rec file (in Entry - menu tools) e. change the format and labels in qes file and save the qes file f. enter more data - entry will see that the qes file is new and adapt format of rec file. But the best strategy is to have a standard qes file and rec file from the beginning and then adapt to the current outbreak. Two examples of this is shown in the "field guide under way" in http://www.epidata.org/wiki/index.php/Field_Guide Exciting to get more news on the ongoing analysis "down under" (seen from my perspective). regards Jens Lauritsen EpiData Association

1 1

Outbreak analysis diary - Symptoms and Epicurves
by epidata-list＠lists.umanitoba.ca 04 Jun '08

04 Jun '08

Today I am trying to see if I can use Epidata Analysis to do the first three analysis tasks. 1. How many people are sick 2. What is the symptoms profile 3. Epidemic curve Tasks 1 and 2 are nice and easy. a) click Analysis b) Click on Frequency c) Select Sick from the Select Variable Menu d) Tick row percentages I can do the same with the symptoms profile (Diarrhoea, Vomiting etc) What I don't like is the no comes first on the table that is produced, even though it is a Boolean field. What would be on my 'wish list' is if all the symptoms I select could appear in the one table as exposures do when you select outbreak analysis for attack rates Epidemic Curve The epidemic curve starts badly. I start by asking for a histogram of diadate (date of onset of diarrhoea) and even though this is a date field (and shows as dates if I do a frequency) the graph shows 76115, 76116 etc on the X axis instead of 27/05/2008, 28/05/2008. Then I remember there is an 'epicurve' graph and this works well. I would like to be able to show diatime on there as well (time of day people were sick) but remember this is not possible yet. Never mind - dates will do for now. My question for today is - can someone help with some code to create a single onset date? There are two variables - one called diadate and one called vomitdate - the onset date should be whichever one of these is the earliest. We could calculate this manually but it would be great if we could build it in to either epidata (we still have at least 100 records to enter that check code could help with) or epi analysis Thanks Christine Christine Roseveare Analyst Regional Public Health 570 9194 027 495 9671 >>> <epidata-list-request(a)lists.umanitoba.ca> 27/05/2008 5:04 a.m. >>> Send EpiData-list mailing list submissions to epidata-list(a)lists.umanitoba.ca To subscribe or unsubscribe via the World Wide Web, visit http://lists.umanitoba.ca/mailman/listinfo/epidata-list or, via email, send a message with subject or body 'help' to epidata-list-request(a)lists.umanitoba.ca You can reach the person managing the list at epidata-list-owner(a)lists.umanitoba.ca When replying, please edit your Subject line so it is more specific than "Re: Contents of EpiData-list digest..." -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 0

Outbreak analysis with Epidata Analysis diary - 1 Data issues - text to Boolean
by epidata-list＠lists.umanitoba.ca 04 Jun '08

04 Jun '08

We have an outbreak to deal with. Up to several hundred people are possiblly sick after attending a local sports venue that also hosts children's birthday parties. Like many outbreaks it comes at a difficult time (they seem to love happening on Christmas Eve or the day before Easter). This one we learn about the day before a holiday weekend and I have the day off the next day. Someone has designed a questionnaire asking our usual questions (Demographics, symptoms and exposures) and the phone calls have already started. I help someone create an Epidata questionnaire to enter the data and promise to help with the analysis on my return. There are two main things we have to do in our outbreak analysis 1. Describe the pattern of illness by person place and time (how many people are sick with what symptoms, when did they get sick and for how long) 2. Discover if certain 'exposures' - e.g. eating certain foods are associated with illness. Today I come back and find I have a problem I must fix before we start the analysis. My workmate has changed my file so all the Boolean fields I used to record the exposure information are now text fields. (She did this because she wanted to be able to put "U" for unknown. I explain we need the data in Boolean fields for our two by table analysis and if the answer is unknown to leave the field blank. She has entered 66 records already so I have a problem - how do I change the data from text to Boolean as changing a field type in Epidata from text to Boolean would lose all the data? The way I work out is: (using the example outbreak.rec for the file that has data entered as text) 1. Use Save as with outbreak.qes to make a new file called outbreak2.qes. Change this file so the txt fields are now boolean fields 2. Create a new blank rec file from outbreak2.qes called outbreak2.rec 3. Export outbreak.rec to Excel as outbreak2.xls 4. Use search and replace to change the Y, N and U that have been entered to 1,0 and blank 5. Save Outbreak.xls to a text file outbreak2.csv (CSV) 6. Import outbreak2.csv as a text file ticking the option to not use first line. This 'pulls' the data into outbreak2.rec and the boolean fields with the 1 and 0 showing as Yes and No This seems a long way round so maybe someone has a better way to fix this problem? Next posting I will start with the analysis thanks Christine Christine Roseveare Analyst Regional Public Health 570 9194 027 495 9671 >>> <epidata-list-request(a)lists.umanitoba.ca> 26/05/2008 5:01 a.m. >>> Send EpiData-list mailing list submissions to epidata-list(a)lists.umanitoba.ca To subscribe or unsubscribe via the World Wide Web, visit http://lists.umanitoba.ca/mailman/listinfo/epidata-list or, via email, send a message with subject or body 'help' to epidata-list-request(a)lists.umanitoba.ca You can reach the person managing the list at epidata-list-owner(a)lists.umanitoba.ca When replying, please edit your Subject line so it is more specific than "Re: Contents of EpiData-list digest..." -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 0

Outbreak analysis with Epidata Analysis diary
by epidata-list＠lists.umanitoba.ca 04 Jun '08

04 Jun '08

I use Epidata to help with the analysis of outbreaks of communicable disease. Up until now I have used Epidata to design the questionnaires and enter the data and Epi info to analyse the data, but for the last outbreak I used Epidata Analysis to do some of the analysis (to produce the attack rate table). This was much better than using epi info. I told Jens I might try using Epidata analysis for all the outbreak analysis in the future and he suggested I write some short notes about what I do, and post them to the list as the basis for a future field guide. We now have a brand new outbreak to analyse so I will do this and hope you will contribute with comments and suggestions if you do similar tasks or are interested. Thanks Christine Christine Roseveare Analyst Regional Public Health 570 9194 027 495 9671 >>> <epidata-list-request(a)lists.umanitoba.ca> 26/05/2008 5:01 a.m. >>> Send EpiData-list mailing list submissions to epidata-list(a)lists.umanitoba.ca To subscribe or unsubscribe via the World Wide Web, visit http://lists.umanitoba.ca/mailman/listinfo/epidata-list or, via email, send a message with subject or body 'help' to epidata-list-request(a)lists.umanitoba.ca You can reach the person managing the list at epidata-list-owner(a)lists.umanitoba.ca When replying, please edit your Subject line so it is more specific than "Re: Contents of EpiData-list digest..." -- The information contained in this email and any attachments is confidential and may be legally privileged. If you have received this message in error, please notify the sender immediately and remove all copies of the message, including any attachments. Any views or opinions expressed in this email (unless otherwise stated) may not represent those of Hutt Valley DHB. Thank you. ** Disclaimer added by HVDHB **

1 0

Re: EpiData-list Digest, Vol 55, Issue 19 Missing values is not NA
by epidata-list＠lists.umanitoba.ca 29 May '08

29 May '08

Dear friend Ziad, NA is common expression to identify missing values. If you are working with R or S-Plus then NA is NA. That means that all functions or commands that you want to specify NA, the expression is also NA. However, in Stata or Epidata entry the NAs are epecified by dots (.) and sometimes by empty quotes (" ") depending if the variable of interest is formated as numeric or text. So, my guess is that your check command will work fine if do like this... AFTER ENTRY IF v64=. THEN LET V65=9 & V66=9 & goto V67 ENDIF END Best regards, Pedro Emmanuel > Today's Topics: > > 1. If..then (epidata-list(a)lists.umanitoba.ca) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Wed, 28 May 2008 08:47:34 -0700 > From: epidata-list(a)lists.umanitoba.ca > Subject: [EpiData-list] If..then > To: epidata-list(a)lists.umanitoba.ca > Message-ID: > <897a5cf60805280847u3b99c66ak6f49e3855b879141(a)mail.gmail.com> > Content-Type: text/plain; charset=ISO-8859-1 > > Hi, > might be easy question, i am trying to write condition under check > file to write value of 9 in 2 places if one cell equal to NA (code > below), it does not work (and no error message), any tips? > > AFTER ENTRY > IF v64=NA THEN > LET V65=9 & V66=9 & goto V67 > ENDIF > > > Thank you in advance > best regards > ziad > > > ------------------------------ > > ________________________________________ > EpiData-list(a)lists.umanitoba.ca > http://lists.umanitoba.ca/mailman/listinfo/epidata-list > > > End of EpiData-list Digest, Vol 55, Issue 19 > ******************************************** >

1 0

re: If..then
by epidata-list＠lists.umanitoba.ca 28 May '08

28 May '08

Just split in three lines: after entry IF v64=NA THEN LET V65=9 V66=9 goto V67 ENDIF regards Jens Lauritsen EpiData Association

1 0

Epidata Analysis and diagnostic test accuracy analysis
by epidata-list＠lists.umanitoba.ca 27 May '08

27 May '08

Hi all, I did look at epidata website but for any unknown reasean I could not find the pdf manual of EpiData analysis. I use a lot the epidata entry but neve tryed before the analysis. I would like to know if any of you could tell me if EpiData analysis has resources to analyse diagnostic test accuracy such as sensitivity, specificity (and theses estimates in different prevalences), estimate the area under the ROC curve, draw ROC curves, compare ROC curves form different tests? I also did try to download/install the software and look for theses abilaties, but here at the institute where I work none of the users are coputers administrator (security reasons) and install "non standart softwares" is just not that simple! Thanks a lot in advance! -- Abraço forte e que a força esteja com você, Pedro Emmanuel

1 1