00 : 00 / 05: 00 (Preview)

This discussion has ended. Watch the recording on Medflix app,

Choose variables for your study | Befriend the pillars of data

Mar 04 | 12:30 PM

In the 4th module of the Presidents Program on Dissertation-Guidance by IRIA, Kerala, we are delighted to invite you to join an academic feast with Dr. Praveen Nirmalan as he walks you through the lanes and bylanes of choosing variables for your research study. The choice of variables is an integral part that drives the output of your research study. Join us Live as we trek through some great adventure trails!

[Music] good evening everyone i am he is the research mentor in the amma healthcare research guru kochi kerala he has his amma center for diagnosis and preventive medicine so on behalf of ira kerala i welcome you sir presenting i would like to invite dr prabhin himalayan for this talk because he has taken lot of pains to teach the residents because it's very difficult to write it with a particular time so he is help he's helping us how to make that this is easy and how can we write it which will be very helpful for the residents as you all know he is an ophthalmologist well-known ophthalmic surgeon and now he has left the job and come to this teaching group which is very helpful for us as a clinician he is doing this job is very helpful and we are very very very happy to have him in our ira family group because he is a very good helping hand for us over to you dr praveen you can continue a session over to the platform you can view the recording of the previous sessions in the kerala iri club on the netflix app as in when they get uploaded which will be very soon the learning objective is for this particular session it's a short sweet session the learning objectives for this particular session is to understand the principles behind the choice of variables for your research study and so let's start by looking at what are variables now when we hear the term variables we often tend to worry what should we include how many should we include why do we include all this but variables are not variables they're not something to worry about what's what are variables the information we collect to answer the question are known as variables they are called variables because they are able to vary between people or between measurements so that's why they are called variables because they are able to vary why are they important these pieces of information fit into the larger puzzle to answer your particular research question the variables make the path towards your answer smooth if they are appropriately chosen if they are not appropriately chosen if you don't choose the variables carefully you will find it difficult to get the answer that you want the answers that you're seeking you'll find it difficult to interpret the results that you're getting or to even get relevant results so that's why they are important you have to choose them carefully now we can choose variables in different ways why how i would look at it usually is to choose variables in terms of the function of the variables and then to choose where and then to look at the structure of the variables so what do we mean when we say choose variables by the function there are four basic types there the first is the identifiers identifier variables are the variables that are necessary to identify the study and the study participants they are an integral part of every research study every dissertation every thesis the identifier variables the identifiers can include the title of the study can include the study center the name of the study center location center id especially if it's a multicentric study the date and time of the assessment of the particular subject image or sample a subject id or an image id or sample id demographics of the subject that can include the name age and gender the name has to be anonymized later as we are sharing the data and analyzing the data but you can collect that information you can add more demographic variables as you would like to clinical demographic variables as you would like to so those are the identifiers the variables that help you to identify which study is this which center or which department which patient or which sample and some details of that to make you identify this particular person then you have a set of variables which are related to outcomes variables that are essential to understand the outcomes now your research question has got outcomes primary outcomes secondary outcomes those are the answers that you are seeking to the answers that you are seeking from your research question the variables that are essential to understand the outcomes have got a prerequisite that you need to be absolutely clear on what your primary and secondary outcomes are so in a sense these variables the variables looking at the outcome so what we call as steady measures in a sense they force you to obtain absolute clarity on your primary and secondary outcomes they include defining characteristics and criteria for diagnosis screening and prognosis how do we get that clarity for each of the study measures or each of these variables that look for the outcome look at the outcome we want to look at what is this particular variable when do we want to collect it where do we collect it from or which setting is it from an op setting an ot setting is it from a primary healthcare setting secondary healthcare setting a community setting who collects it is it a trained clinician is it a paramedical personal is it someone else a trained study staff why are we collecting this variable and how do we collect this video you have another set of variables based on function which is known as the impact variables now these are variables that impact on other variables necessary to understand the outcomes so such variables change how other variables behave with each other or they change the way other variables act on the outcomes and then the fourth category by function is the fishing net what we call as the phishing it variables that are collected just because well we are doing a study we are anyway collecting information from subjects so let's collect a lot of information and why do we do that because we can avoid repeated visits to collect information especially if it's a population based you don't want to keep going back to the village to collect information when we are going once let's collect all the information that's possible and we can use that for any future study well the fishing net is not at all a good approach and that's because we cannot be sure if the information that we collect now will be relevant for a later time period the study subject must give more time because there are lot more variables to answer and study subject must give more time and that may put off and that will put off the subject from participating in the study because they have to give more time and it will lead to more refuses the use of information for other studies almost always never happens because different questions need different levels of measurement and we'll find when we try to use it for a different study that the particular way or method that we use to measure this variable is not sufficient so usually we don't tend to use it often for any other study in the future fishing net is also not good because it gets more workload you must train data collection personally even if it is you as the investigator you must train to collect more data points more which means more training points you need a larger data and reform which means you need more time for data entry which means you need a larger database which increases the chance for errors so the phishing net approach is not at all a good approach and the most important point is if we will not use all that information after all that effort why collect it you're already overworked as it is why do we just collect it enter it and then not clean it and then not use it in any way so let's not go to a fishing network so basically we'll be looking at three types the identifiers which are used to identify the outcomes or the study measures and the impact variables or variables that impact or act upon other variables in a way that changes their behavior in terms of structure what do we look at we have what we call as quantity variables that involve involves numbers or amounts they can be classified as discrete or continuous a discrete is any numerical variable that you can realistically count such as the coins in your ballot or the money in your savings account hopefully you can count it maybe it's limitless i don't know but hopefully you can count it continuous variables numerical variables that you could never finish counting such as time so you have discrete variables you have continuous variables just keep these terms in mind when i do the analysis i don't really look at if it's a discrete variable or a continuous variable as long as i've just got that concept in the mind it helps me but when i'm creating the data form i want to look at okay how am i collecting this particular variable i collecting it as a discrete variable or as a continuous variable you have the qualitative variables or categorical variables which are non-numerical values or groupings they can be binary that is they have only two categories for example male or female they can be nominal in terms that you have more than two categories but there is no specific order for these categories single family home a condo a tiny home a big home medium home so there's no particular order and you can have those are known as normal variables you can have ordinal variables which are variables in more than two categories that follow a particular order especially quality of life or satisfaction unsatisfied neutral satisfied very poor poor neutral good very good so that's an ordinal variable the other type of variable that we usually come across and that we usually have and collect is an independent variable and a dependent variable so we'll just look at that briefly an independent variable is a singular characteristic that cannot be changed by other variables in your experiment so it's something singular characteristic it cannot be changed by any other variable in your experiment however independent variables can change other variables a dependent variable relays on and can be changed by other components independent variables can influence dependent variables but a dependent variable cannot influence independent variables and we look at an example of this soon so your age is an independent variable it can affect hypertension which is a dependent variable or the outcome the dependent variable your age can affect hypertension but your hypertension cannot change your age so you can see that the independent variable can impact on the dependent variable but the dependent variable cannot change the independent variable then we have what is known as confounders a confounder is a variable that can decide disguise effects of another variable on the outcome they can invalidate your experiment results through bias or suggesting a relationship between variables when there is no relationship [Music] for example you are studying the relationship between exercise which is an independent variable and body mass index as the dependent variable but you do not consider the effect of age on these factors so age can affect the level of exercise or the intensity of exercise it can also affect bmi age thus becomes a confounding variable that changes your results so when we are looking at variables we also have to consider are there confounders that need to be included are there confounders that must be measured now one point we have to keep in mind is that we cannot we cannot control or we cannot assess or measure all confounders we can only look at non-confounders there might be many indirect and many unknown confounders as well which might become obvious when we look at the data but as far as possible as much as possible we need to look at what are the confounders and where do we find that from we find that from a review of the literature and what is previously known from published evidence then we have what is known as a composite variable that's two or more variables combined to make a more complex variable and you can see this often coming in health outcomes like the neonatal mobility discourse perinatal morbidity scores where you're looking at adverse events single adverse events and then you are creating a composite index of morbid mobility so you could look at something like a neurodevelopmental disorder a congenital malformation there is neonatal jaundice you can look at things like that and then create a composite variable that combines all of this and creates a score for mobility that's often used in neonatal and perinatal mobility now the thing with variables is that you have to plan ahead and you have to plan well plan in the sense that identify the function and structure of each variable that you're including they are integral to your plan of analysis what you must think is what is the best way to communicate yourselves what will let you find optimal results and whatever is necessary to communicate optimal results must be there in the list of variables that you are including and if you look at variables in terms of their function more often than not you will find that you are able to include all the necessary variables now suppose when i say plan suppose i want to present results as a comparison of means can i do that if i collect information as categories or groups i cannot derive an average if i collect information as a category or a group so that makes it a bad choice or a bad list of variables so i need to think okay and now i'm going to compare age in people who are obese and people who are normal have a normal bmi if i want to compare age i can look at it in terms of proportions where categories work i can if i want to present the average or i want the mean or if i want to present the median that's the 50th percentile then i need to have the age in a numerical form unless i have it in a numerical form i can't do that i can't create a mean or a median so i need to think how do i want to present this result in a more optimal manner for a better communication and then plan the structure of the variables as well now one thing to remember is that i can create create groups out of continuous or discrete debt numerical data but i cannot convert categorical data into numerical values so if i have a choice i prefer collecting data as a numerical data and then create my categories later now that allows me flexibility as i do the analysis i find that well i categorized ages 11 to 20 21 to 30 extra extra text up to 70. i checked with the results and then i realized that okay i can present this as less than 50 or less than 45 more than now if i have predefined categories i cannot do that if i have age as a numerical value then i can create those categories so i can look at what is more relevant more pertinent and then create the categories that are necessary so it gives me greater flexibility during my analysis the wrong variables the choice of wrong variables either in function or in structure can introduce and will introduce a lot of noise in the analysis and results noise in the data results in more time for data entry cleaning and adds confusion to the entire analysis wrongly character characterized or measured variables lead to a loss of information as if just explained in that example if i categorize or characterize age wrongly then i can't present the mean or median so i lose information so we need to be looking at the results we need to also be considering the workload that is being put into collecting this information as variables and then that workload that you are putting your effort must actually bear some optimal result otherwise it becomes an exercise in futility inappropriately measured variables can and will derail your study completely and that when i say inappropriate measures it includes inappropriate timing and methods of measurement measuring a variable at the wrong time taking a blood pressure measurement at the wrong time taking a blood sugar measurement at the wrong time inappropriate methods of measurement inappropriate personal doing the measurements or untrained people doing the measurements inappropriate units and storage of measurement that all of this can affect and derail your study results what we must realize is that the statistical analysis plan that you are building builds on the structure and function of the variable so they are very very very important so here is something for the road i am on the road as it is so here is something for the road for you what you must check as you look at your list of variables is what value does and you must do this for each variable what value does this variable add to my study is it necessary do i have proof of its utility now the only way for you to do this is if you do a review of literature if you look at the evidence that is out there and look to see if it's been proven um ten times that this variable has no particular relation with this particular question then don't collect it if it has some relation or if it's in a gray zone you definitely want to collect it so what value does this variable add to my study is this necessary do i have proof of its utility very important is it feasible to collect this information and then you must think who will collect it how will it be measured where will it be collected how about training to collect that particular variable if it's something that needs a lot of training and a lot of standardization then you if it's a dissertation you definitely want to think do i really want to go down this road do i have the time to go down this road at the end will i be sitting and sweating so that's something that you have to think very important write down what will i lose if i don't study this variable if i do not include this variable what am i going to lose and that in a way reaffirms the point about why is this particular variable important so you are looking at it from two angles why is it important what will i lose if i don't study this variable then you check is this variable linked to other variables industry is there a relationship that's there do they interact with other variables in any way so in a sense you are looking for confounded step how does if it does this variable influence or impact other variables in the study most important how will i use this information does the method of ascertainment allow me to use this information appropriately now if i will not use this information from this variable or the method of ascertainment does not allow me to use this info i have not measured it appropriately then don't collect it at all because it's an added work and anyone who is doing data entry on their own knows the difficulty of having to enter data in a spreadsheet especially when you won't use it in any way so you have to make sure that it's something that you will use before you collect it if it's a small study with a small sample size it won't be too much of a pain but if the sample size is large let's say 500 600 subjects a thousand subjects and then you're just collecting variables that you won't use at all get so much of entry so much of cleaning so much more chances for errors and so much opportunity to get completely demotivated well as it is you are not completely motivated regarding your thesis in the first instance so don't choose variables if you will not be using that information most important thing is that your list of variables walk you through the process of measurement to answer your question and give a preview of results before you start the study so it actually gives you a preview of results it walks you through the process of measurement it tells you how you can answer your question it tells you everything about your study before you actually start collecting data before you start this so that's the importance of the list of variables now oftentimes we look at a question and then we just write down a list of variables and then we go ahead and collect that and then more often than not during the analysis we get stuck because we're not getting the things that we wanted that we expected that we want to present and then you go back and forth back and forth back and forth and becomes a complete painter that can be taken care of if you plan your list of variables appropriately and so you must spend time thinking over this part this is something that no one writes or talks about in a in any course on research any research methods course this is a practical pragmatic based on painful experience of having inappropriately collected variables having inappropriately created or created inappropriate data collection forms and then sitting and sat and scratched my own head so many times wondering why did i do this now what do i do how do i correct it i'm losing so many patients data now what so there that's that's the way you then start thinking okay i need to be careful at the very beginning and if i am careful at the very beginning i can avoid a lot of heartache later so thank you so much and uh this is a short speed session short in the sense that i just want to introduce thank you dr gravy i think you are finished in half an hour i was just thinking how dedicated you are because you are sitting in the car and you are giving the speech i am really surprised to see your dedication i i think the resident should utilize this because you are giving the talk with such some inspiration and dedication the resident should utilize it i hope everybody would have listened to it i am very happy that uh to see your sincerity thank you so much dr praveen for giving such a wonderful talk

BEING ATTENDED BY

Dr. PREETI PRAKASH & 182 others

SPEAKERS

dr. Ramesh Shenoy

Dr. Ramesh Shenoy

Consultant Radiologist | Kochi

+ Details
dr. Praveen Nirmalan

Dr. Praveen Nirmalan

AMMA Center for Diagnosis and Preventive Medicine | Research Mentor, AMMA Healthcare Research Gurukul, Kochi, Kerala

+ Details
dr. Venugopal .

Dr. Venugopal .

Consultant Radiologist

+ Details
dr. Rijo Mathew

Dr. Rijo Mathew

Consultant Radiologist | Kochi

+ Details
dr. Gomathy Subramaniam

Dr. Gomathy Subramaniam

Professor and HOD, Radiology, Malabar Medical College, Kozhikode

+ Details
dr. Ramesh Shenoy

Dr. Ramesh Shenoy

Consultant Radiologist | Kochi

+ Details
dr. Praveen Nirmalan

Dr. Praveen Nirmalan

AMMA Center for Diagnosis and Preventive Medi...

+ Details
dr. Venugopal .

Dr. Venugopal .

Consultant Radiologist

+ Details
dr. Rijo Mathew

Dr. Rijo Mathew

Consultant Radiologist | Kochi

+ Details
dr. Gomathy Subramaniam

Dr. Gomathy Subramaniam

Professor and HOD, Radiology, Malabar Medical...

+ Details

About Medflix

Medflix is a new platform by PlexusMD, India's most active and trusted doctor community. On Medflix, you can discover live surgeries, discussions, conferences and courses from some of the top doctors and institutions across the world. Join clubs in your areas of interest and access hundreds of amazing live discussions everyday.