Edit Seer Signature Service with airSlate SignNow

Fax signed seer

since then there are some events and expansions presently about just over a quarter of the US population is covered by one of the seer registries there's a new expansion that is not available to see your Medicare yet because just because of the timing that will add three more states as well as greater representation of American Indians so the seer programs as you can see are progressively working to make sure that population representation and the diversity of the u.s. population are represented in the registries but the registries are selected part in part largely on their ability to conduct high-quality cancer surveillance as well as to contribute to the population representativeness so that combination is how the seer has systematically expanded this shows the present seer registry so if you were to order seer Medicare data these are the registers you would get you will see there are several city registries such as Seattle and Detroit as well as Atlanta there was an expansion in 1980 in 1988 that added Los Angeles San Jose and Georgia the rest of Georgia and then in 2000 a further expansion that included the rest of California Louisiana Kentucky and New Jersey so we have very solid representation nationally geographically and demographically as well there will be the new expansion has added some states we've added New York Wisconsin the Seattle registries larger Idaho and so on as well as Mormons Cherokee Nation for example so you can see this progressive change I point this out to you for two reasons one to help you understand the strengths and the excitement that we have over working with fear but also so that when we come to questions about design that you're aware that that the registries have changed and your study design may need to take registry changes into account depending on what you choose to do and over what period of time when we look at this year registry so this is comparing cases diagnosed in 2010 with the 2010 census what we'll see is that the seer areas represent a diverse population but at the blue bar is the seer areas the gray bar is the census but we do see that there's still an over-representation of white people over others particularly on black American India and well it may be representative is very small and often too small to be able to do in-depth studies so we're very excited about the 2018 expansion the seer registries are more urban concentrated than rural and so we need to take that into account when we're doing inference and talking about representativeness and again the addition of some rural states will be helpful with it with a new expansion we do have about what we'd expect form representativeness of Jett by gender so we don't see any differences there and so then understanding how the cases are identified by the registry the registries use two methods these active and passive case finding so the active case finding is notification from facilities who are required to the review of sorts documents such as pathology reports passive would include other sources of information that might be that might be received it is important to notice that the seer program relies heavily on facilities such as hospitals and so one place where the seer registry is worried and we agree is when treatment and diagnosis is provided only in physician offices and that's because that reporting might be less complete so if you're thinking of a situation such as simple skin cancer you know where it's often treated largely in physicians offices the registries may be less complete than you would like and we would definitely want to consider whether we're seeing consider and information between seer and Medicare and how to how to understand that the cases are extracted by trained professionals working for either a hospital or registry and this is important they use a consistent process involving agreed-upon established rules as a user I can tell you very clearly that understanding the rules and occasionally taking time to look them up can be very helpful so we will see a code and often say hmm I wonder what that means but because these are consistent established rules it means they're established codebook and we have sources so we can look them up and say now I understand how this is used and how it compares to something else that sounds really similar and I can understand the patterns that I'm seeing so I encourage you all to make sure that you take the time to find your codebook and to know where they are and at least look them over so you understand how the abstractors view the medical records because of course medical records are rarely consistent even within a same facility but certainly nationally so the abstractors have the challenge of taking very inconsistent information and making it consistent so the rule books the abstraction manuals are an incredible resource the data collected by the seer registries fall into four categories demographic information so about the people who are diagnosed with cancer information about the cancer itself how it was identified and details about the tumor we will go into those more later there's information about the treatment and information about survival and those are the very large categories of information that the seer registries collect your registries also demographic residents at diagnosis these are population-based registries so one of the first steps that the registries do is confirm that this person actually is part of the registry catchment area they collect gender sex age at diagnosis year and place of birth race or Hispanic origin and marital status a couple of these variables are not ones we routinely are able to get marital status for example is not a variable we are able to routinely obtained from other places so many people find it quite interesting and an opportunity to be able to use in the place of birth information is is available I think there are always our questions about it because of course the the abstractors don't go out and seek information that isn't there so if the information isn't in a chart they won't have it and so we one of the things that needs that you need to think about when considering on formations that obtained by series is is something that we would expect to be reliably in medical records and to this extent the quality of the medical record affects the quality of the data the information about the cancer includes when and how it is diagnosed so date of diagnosis only month and year are released day of diagnosis is not released we will talk about that more later on this morning but also there they also include information on method of diagnostic confirmation was it a clinical diagnosis the doulas affirmed the more that we would hear it I think somebody needs to muse we also have information about who reported the case to seer which category not a person so was it reported by a hospital was reported by a nursing home did we get it from a death certificate or autopsy and this information again we often use to determine to help us provide get some insight and to determine how likely we are that this is at information the seer registry provides us with this information that we can make the decision about who we want to include in our study and then we also understand the content of sequence of cancers is this their first cancer in which case of sequence temp was actually zero zero so the first and only was it their first of multiple cancers or was it somewhere else was it their second or third cancer this is important as we look at treatment because we made clinicians will make different treatment decisions depending on somebody's history I'm going to take this opportunity to point out that there are some cancers where people can have multiple primaries breast is one colon is another and there are others where people can have multiple independent tumors that would be distinguished from being a recurrence so as you think about that and make a decision about sequence number take that into account likewise if somebody is first diagnosed with in-situ disease for example DCIS or to the city disease of the breast and then later is diagnosed with invasive they should have two records one further in city diagnosis and one for their invasive diagnosis in that case we might not interpret that sequence number in exactly the same way as we would if someone were for example first diagnosed with breast cancer and then diagnosed with um with lung cancer or brain cancer or something else the tumor information is very very important and very useful so we know the cancer type we know the stage at diagnosis we know the histology increasingly we know tumor markers so for example for breast cancer ers estrogen receptor progesterone receptor and her two status are now available we know lymph node testing and positivity so we know so we can distinguish between people who have no positive nodes but in fact none were tested from people who have no positive nodes but had 20 tested that may change how we view the quality of their work up but it also may change our understanding of the treatment patterns we know tumor grade we know size and we know extent of disease or spread the exact details of all of these will have changed over time with a definite trend toward increasing information and also they will vary from tumor to tumor so just because you look up kidney cancer and you can't find full detail across all of these things the detail you want does not mean that if you look up gastric cancer you won't find more detail so my big lesson here and again from experience is that whenever I look at a new tumor I take the time to actually dig into these variables and I've often been surprised at the level of detail and the number of variables that are available for me to work with also have basic information about the first course of treatment so what treats first course of treatment and this is an important definition and I really want to highlight that that it's important that you remember this tape it to your wall memorize it the first course is defined as all methods of treatment accorded in the treatment planned and administered to the patient before disease progression or recurrence so as we will talk about in some cases we end up with with mismatches between seer and Medicare but we if we understand what is included and we can think clinically how long would this take it sometimes makes things easier seer collects more information and we I do on emphasize system is released as part of the seer Medicare linkage and this has to do with confidentiality with quality with consistency they will quote lead they release though information categories about surgery and information about radiation therapy and this is covered in quite a bit of detail in segment four later on this morning here it doesn't routinely collect or doesn't routinely release detailed information about all treatments information about chemotherapy information about treatment beyond the first course so if the first chemotherapy didn't work and made a second chemotherapy seer does not routinely release that fear does not routinely release cancer recurrence information or the exact timing how many days from diagnosis to receipt of chemotherapy is not something that you that you will routinely find from Stier I talked earlier about the importance of the manual here are some websites the seer website itself seer.cancer.gov is important there are documentation for the seer data there's training information there coding manuals and there's a competitive staging manual the other thing that can work is using your favorite web browser and typing in Cerre or sometimes I'll use seer cancer because se er sometimes comes up with Sears or people who are willing to do willing to dump predict your future but again these are widely available and from the registries themselves as well as from the NCI the Medicare program now is an equally complex and interesting and valuable program for researchers Medicare as you all know is a federal health insurance plan available to qualifying elders aged 65 and older and select disabled adults so this is the population that is covered by the Medicare program the Medicare program covers for most people as an entitlement hospital care limited skilled nursing facility care and hospice and this is sometimes called Part A about 98% of part of elders have Part A and this is an entitlement which means if you qualify for Medicare you get this coverage they also have two levels of optional coverage the first level is called Part B and that covers outpatient care meaning care that is provided in a facility setting for people who don't decide their so for example emergency rooms day surgery in a hospital where you go in in the morning leave at the end of the day it also covers physician and provider care and home health care and this again is called Part B in about 94 percent of elders enroll in Part B and then there's another optional coverage which is Part D which provides prescription medication coverage Part D as some of you may know began in 2006 as a benefit seer Medicare releases Part D data starting in 2007 and the reason that they made that decision is because the 2006 Part D data had some problems they were not as comprehensive and there were concerns about completeness that were resolved by the 2007 release and this decision of not releasing 2006 is quite common and most people who use Part D data opted not to use the 2006 and as we said most people have Part A and Part B and fewer people have Part C so if you notice in this graph part a is the blue column 98.6% habit Part B is the green column and Part D is the gold column you'll also notice there are two rows one is called fee for service and one is called managed care Medicare also has two ways of processing and coordinating claims the fee for service programs sometimes called Original Medicare involved and I still think it's backwards it should really say service for fees in other words providers provide a service a hospital provides a hospitalization and then once they provide it they are paid for that to give a service they get fees back the managed care plans Medicare contracts with plans and the plans are responsible for providing all care and coordinating those benefits this is about a third of all Part A enrollees or in managed care and about two thirds are in fee for service the challenge and we will talk about this repeatedly is managed care plans do not release their claims so with managed care we know there and managed care we can see the Sears side but we cannot see the Medicare details so we will be talking about this repeatedly throughout the course about how to Matt how to handle that in your data and the implications of different decisions I point out also that Part D enrollment is much more common in managed care than it is in the fee-for-service program typically the Part D is bundled into a managed care program and with Part D it's a different decision so keep this in mind when you're deciding when you're deciding what to do so broadly the Medicare data includes two types of information they include enrollment data so who is in the program because if you're not enrolled in Medicare then we have no expectations of seeing the claims and then we have information from bills provide submitted by providers and processed by Medicare so we know what Medicare paid for or what people asked to be paid for for those Medicare beneficiaries so the way it works in the fee-for-service is that whenever a Medicare enrollee Steve's health care a bill is sent to the Medicare program to be paid the bill will contain information that will help the bill be paid but it's honestly almost exactly the same information we want who provided the care when was the care provided so we can exact dates why the care was provided so diagnosis what care was provided so was there a procedure what was done was it an office visit was it a surgery was it a biopsy and it also contains information about how much was paid for the care and so those are the big categories of elements that we look for we'll be covering costs in much more detail in segment 5 they're an opportunity but are also very can be very tricky so if you're interested in cost of care or payments please make sure that you attend segment 5 the data are organized based on bill type so they're not necessarily arranged in a way that's intuitive for most researchers so there's there are seven claims based service files there's a file called the medpar which includes the hospital and skilled nursing facilities debt claims there's the outpatient file and that includes that's non inpatient facility claims so it's largely hospital outpatient departments plus a few more there's the MCH file nasir medicare calls if you have used medicare data is often called the carrier file there is the home health agency file there's a hospice file there's a DME or durable medical equipment file and so that will contain durable medical equipment that wasn't provided by a physician and there's a pharmacy file so you can see that you cannot see already and we'll talk about it more that you will often have to look across several files to find a category of care or a type of service because these are organized not by not by what happened but by where it happened and who is paid for the money in Medicare and as I said we will cover this more later but at a very high level the fields and the Medicare claims files are divided into I would say two groups the charges which are what the provider the hospital the clinic the physician asked to be paid sometimes they ask exec for what they know they will get and sometimes they ask for more the charges are there and we also know the payments we know what they were paid or were approved to be paid so we know for example that the hospital might be approved to be paid up to ten thousand dollars but we will also see that there was a co-payment that the patient was responsible for so those are the two big categories of money in the Medicare program there are some details of the Medicare program that I think make them particularly useful for studies particularly longitudinal studies of Cancer Care so first of all they contain all claims regardless of where somebody is so well the seer registries are population-based based on where they live at time of diagnosis the Medicare data will continue to follow them so if somebody who lived in Iowa was diagnosed with a lung cancer in Iowa decided I hope there's nobody from Iowa on here that they were tired of the winters and they were to move to Arizona we would see the cancer diagnosed in Iowa but we would also see all of their claims including those from when they lived in Iowa when when they move to Arizona and even if they went back to Iowa for summers we would still see those claims so the claims follow the person and aren't going to toes that a population-based combined with a person based analysis ability to analyze the Medicare program also will have fee-for-service care provided from enrollment until death so that means we can look both at what conditions or how what care somebody received prior to diagnosis as well as care they received after diagnosis or after first course of treatment we can look we can we have a better sense and we'll be talking about this about comorbidity about how patients manage multiple chronic conditions simultaneously so this again makes sure that we have a much more comprehensive view of things that we might otherwise and what that and that allows us by using days to really understand what is going on and what what what conditions a person is juggling there are some things that Medicare data do not contain and these are important to remember first of all the reason a test was done is not included and the results of the test are not included so we cannot know was this test done in response to some clinical signers symptom or was it a routine follow-up we don't we do not have test results we do not have for example PSA test for prostate cancer and sometimes very importantly we don't have pathology so we don't so Medicare won't provide histology but they also don't provide surgical margin so we cannot tell if a tumor was completely receptive or incompletely respected we only know there was a resection in some cases that information would be very important to try to categorize and distant distinguish cases there are behavioral risk factors that are that could be really really useful that we rarely get and I want to note here that there are codes for every single one of these things I'm going to mention but it's really important to remember that just because it can be coded doesn't mean it will be and so it's really important not to over interpret these items and to do some checks to say does this pattern make sense but some that are not available or are we want to be very careful with one is obesity so many of the surgeons that I work with one ask questions like is our do surgeons do different things for obese patients and for non obese patients we cannot know that smoking is important for those of us who study smoking-related cancers this can be frustrated or when they wanted to button distinguish between somebody for examples lung cancer among smokers and lung cancer among non-smokers smoking will be coded sometimes and I personally believe that if the smoking is coded you can assume the person smoked but if the coding if the smoking is not coded I am not completely sure that I believe that person was an smoker so that's the challenge family history isn't coded and alcohol use among other things so if you want to use a behavioral risk factor as a stratification variable or an explanatory variable you need to think very carefully about whether it's going to give you the information you need others saying that routinely that isn't routinely found in the data is that is related to under diagnosed conditions so if something isn't diagnosed in a chart it almost never shows up in a bill which means we don't have it to work with and the two biggest examples of that for me are dementia and osteoporosis the absence the presence of a dementia claim we can be pretty confident that somebody thinks person has dementia but the absence of a dementia claim in no way should be interpreted as a person is cognitively intact likewise osteoporosis depression there are a number of conditions that are under diagnosed that well may be present so if you're looking for those either as risk factors or as outcomes take the time to consider how likely it is that somebody with that condition would actually have their conditions showing up in a medical record so then but what can you find and and what we can find is really exciting and provides a lot of opportunity you can provide you can you can get pre diagnosis medical conditions that might influence treatment so comorbidities you can see diagnostic services such as ultrasounds mammography CTS we can find surgical we can find detailed data about the surgery what type of surgery it was done we find other treatment we will see the chemotherapy claim so as I said earlier fear doesn't routinely release information about chemotherapy but Medicare pays for it which means using the Medicare data you can understand chemotherapy they provide information about the providers whether it was a teaching hospital bed size in from some information about volume in ambulance services we know consult not just oncology we can find rehospitalization rates the use of rehabilitation services other care that comes at the same time the one thing we will often see is during the sort of what we call the peri diagnostic period right around the time that somebody is getting worked up for their cancer and getting prepped for surgery often other conditions will be found as well and so we will often see cardiac testing diabetes monitoring and so on we will also see treatment received after the completion of cancer treatment and payments for all of these so you can see that although there are things that we cannot find there an awful lot of really important and really interesting things that we can and so one so the dance that we need to use and that we will be talking about is how you find the opportunity and how you make sure that this opportunity matches your research question I want to point out now that the Medicare data has got a tricky structure and that's important because it means that you need to spend a little bit of time understanding your data this is not a typical rectangular file so for each person we will have a ped staff which and don't ask me with the acronym is patient something file and it will have the summary data will have their demographics it will have for each cancer the cancer type the stage the months and year of diagnosis and so on each person will have multiple medpar records that's multiple hospitalizations each hospitalization is a single record so in this patient a might have two hospitalizations or might have more each hospitalization is a single record that's in a rectangular structure they will also have multiple records for examples in the outpatient file in the outpatient the carrier the hospice the DME are all divided up into multiple segments so each patient will have a claim for each claim a patient has the claim will be divided up by services so an outpatient record for patient a might be divided up into two or three records one for each of their services that they received so you can see we have an enrollment record one per person you have hospitalizations one record per hospitalization and we have other claims records which are going to be divided up into one per service so what we often have to do is roll our outpatient records up to a claim level and then roll them up again to a person level as we are defining our variables so the bottom line about the Medicare data is and the Medicare data are rich source of information about cancer care and outcomes in the elderly they are complex and it's really important to take your time to think about it because you can spend an awful lot of time misunderstanding codes and and being confused by things if you remember though that the Medicare Medicare is an insurance company and that the organization of what they pay and how they pay is actually defined by the insurance function and that that is the perspective from which the data are collected I personally have found that by keeping that in mind I have been able to make sense and to understand what I'm seeing and why I'm seeing it there are resources and ads with the seer data that's really important with the Medicare data that you take the time when you see things that don't make sense excuse me to look into it Medicare you is it is actually a great information about the program it's written for enrollees so it's in plain English but it's very clear and very helpful the medicare.gov website has a tremendous almost overwhelming amount of information I also encourage you to go to resdac org resdac is a research data assistance center is funded by CMS to provide technical assistance for users of Medicare and Medicaid data and there are a lot of documents and links and pointers to information about the Medicare program on that website so then putting it all together what this year Medicare does and the vision of why it was created is combining information about incident cancers and their to detail tumor information with detailed information about treatments before and after cancer diagnosis allows people to measure treatments and outcomes for clinically defined groups of cancer patients so linking the seer Medicare data together the way it is done these personal identifiers for cancer patients are matched against CMS's Center for Medicare and Medicaid Services master enrollment file from that their unique Medicare numbers found that numbers sometimes called the hip and that hick is used to extract claims for the people in seer who are found to be Medicare beneficiaries CMS sends these file to IMS which is messages nci's programming contractor and IMS creates the analytic file they remove the unique Medicare number and overwrite it with their seer case number so from this again we start with this year registries we match them to CMS we use the unique Medicare ID to collect the claims that the unique Medicare ID is never released as part of the seer Medicare data the Registry's sent all mci all of their cases for the linkage the match rates are exceptional I think the most recent linkage had a match rate of ninety five point seven percent of the people aged 65 and older in the Medicare data so if you remember our estimate is is that about ninety eight percent of people who are in this age range are part of Medicare so this is an exceptional match rate with complex data and and opportunities for mismatches all around so again we are I am constantly impressed when I see these numbers and I'm not just saying that because Lindsay is on the call there is variation across registries between about ninety one percent and ninety nine point two percent that that involves a number of things including variation across populations in how either Medicare enrollment is and perhaps information about quality of data that is that is found in the medical records when we look at the cancers that are in there between the period 2004 and 2013 there were 1.1 million people were found in unique people who are found in the data with a cancer the breast cancers is about twelve point seven percent of the data 5.2 percent had bladder cancer ten and a half percent colon cancer 15 percent lung cancer 16 percent prostate cancer and so on so you can see these end up being incredibly large numbers and this is if used appropriately it can be an incredible advantage for asking questions and asking detailed questions and allowing us to focus on subgroups were normally they would be hard to understand so thinking about what is included in this year Medicare data your Medicare data includes 100 percent of the people in the seer data who are found to be in Medicare they're also they have the ability if you want to cut the population the other way they also include a non cancer control sample and what that is is a 5% sample of Medicare eligible people and they take the cancer cases and flag the ones who are part of the 5% population sample so if you want to focus on cancer cases you can do that but if you say no I want to cut the other way and I want to understand risk factors for cancer you can actually use the 5% samples that would combine non-cancer your cancer and other cancer cases and work it so this again is additional flexibility that provides opportunities that aren't typically available and this we will talk about this more in the next segment sometimes people will call me up and they'll still express concern because they found a cancer diagnosis in the controlled population and they're worried that maybe they don't understand the data right and so it's important to recognize that there will always be people in the community in the non cancer population that have cancer diagnoses and there are a number of reasons for these and ultimately they provide insight that goes in both directions you may also find cancer cases where you can't find any cancer claim and in some of it is that the person so going with the first an on cancer case may have a cancer diagnosis because the person may have been diagnosed with the cancer prior to moving to the area in particular I sometimes worry about people snowbirds as we call them who move back and forth so if somebody's a resident in Iowa let's say and they are in Florida for the winter and they're diagnosed with a cancer depending on the communication and tear patterns Iowa may not know about it it's not that Iowa's not doing their best is that nobody from Florida told them and the records didn't move it up so those people would show up would look like they were non cancer cases but in fact they do have a cancer it just wasn't diagnosed in the area so the registry didn't pick it out there will be some cases at the registry mrs. it may be somebody who were their hospital didn't report it it may be that there was a mistake and how the record number or name was transcribed and nobody could find it and then we also have to recognize that there can be misdiagnosis and claims although technically claims should not include rule-out diagnoses they sometimes do so sometimes we will see mammograms coded as breast cancer reality is the diagnosis should be rule out breast cancer but often is coded with a cancer diagnosis in ative diagnosis when in fact there's no base for one so in those cases the mistake will be in the claims so when you're looking at your control populations it's important to recognize that you will see cancer diagnoses there and you'll need to spend time trying to decide how to interpret it the converse which is not on this slide is will there ever be cases who have no diagnosis and the answer there is very much yes it could be somebody who in addition to their Medicare also as private insurance and the private insurance paid for it it could be somebody who who goes to the VA and all of their care is coordinated through the VA and it is it doesn't return back and it also could be a mismatch every there will be linkage matches where the wrong records are combined what you will see as we go forward is that these are all incredibly rare in the range of one to two percent so it's really important not to get stuck on the one to two percent problems but really look at the 95 percent and then just do checks and sensitivity analyses to make sure you're interpreting things correctly when you're looking at people so earlier we talked about sort of the complex structure of files and and so on what I want to make sure you understand is that every person has a ten digit ID known as the Reg case or seer ID that will be used across files consistently so every single record for a particular person will have the exact same your ID the first two digits of that ID reflect the registry number and then the rest of it is a unique number controls will have similar structured IDs where the first two digits will be will reflect the population that they come from and the other and the rest of the digits are unique numbers that will not be found on link to another person when we think about the seer Medicare data combined we've got the enrollment files the ped staff which is the enrollment and cancer data that a file called some de nom which is data for non cancer controls we have the utilization files that I talked about earlier the medpar this carrier file the hospital outpatient file hospice home health durable medical equipment and prescription drugs and then there are other useful information other files that you may choose to pull in there's a hospital file there's a file called a CCW flag file and those are those are chronic conditions flags that we will be talking about later on this morning there are also our data from other linkages that can be pulled in there's MDS which stands for minimum data set which is nursing home quality and there's a system health care quality so those can all be pulled in as needed and as they support your study design Cerre data included this year Medicare data are in the what's called the ped sceptile here's our at my acronym patient entitlement and diagnosis summary file I want to point out this is not in the same format as this year public access format so if you're used to that file this file is rearranged and you will need to take time to understand how it's similar and how it's different but this is an individual level file so each person has up to ten incident cancer diagnoses per person headset has client has flags that talk about Medicare enrollment and HMO eligibility those will be important to subset cases they also have they also are linked to social you economic data from the census as a zip code or census tract level does this is what is included in the ped stuff there's sometimes people get confused about the data and what they can do so I want to review that quickly so often we what happens is we want to either look at the effect of a policy or we need to understand the impact of a policy change on what we are doing so for example the language year 2000 2017 so we would say I want the 2017 data was actually released in 2018 the incident cases were from the year 2000 the newest incident cases were from 2014 and 15 had claims through 2016 so what's happened sometimes is somebody will call me up and say I wanted others a new treatment that just came online in 2016 and I want to look at it with the 2017 data because it sounds like that should work but in fact will often say well when would it actually have been covered by Medicare and are you sure that you will be able to see it if the incident cases aren't from that year so just keep this in mind I always have to sort of look at this table myself and try to say you're out when I can do it and I sometimes have had to say I love the idea I need to do it but I'm going to have to wait until the next linkage when everything lines up so we but and the reason for some of these lags is we have to allow this fear registries time to identify cases and abstract the data they need to be submitted to the NCI linked with Medicare and the data files need to be created and all of those pieces take steps you actually think about all of the steps that happen the fact that incident cases from 2015 and their claims are released in 2018 is pretty amazing but it does mean you're going to have to stop and think about it and decide whether the timing works for your particular question so as a bit of a pitch just remember the strengths of using this here Medicare data is first of all the large number of cases currently about 1.8 million people age 65 and older it includes they include longitudinal measures of healthcare from the time of Medicare coverage through death or through censoring they span most clinical areas where healthcare is delivered they represent a diversity of geographic areas across the US so one of the things that that allows us to do is to make sure that this isn't just a local phenomenon their population based so often we worry for single institution studies which can offer off an academic is is this just what academic institutions do or just this what the path the practice pattern is in in Omaha or is this in fact a broader phenomenon we're looking at they include data on co-occurring conditions as well as conditions that were present before or after diagnosis and we have the opportunity for a cancer-free control population so in some cases we were trying to look at the long-term consequences of cancer care we really want to compare them to people who wouldn't have had cancer at all and this allows us to do it there are some limitations of these data and these are important limitations to keep in mind as you're balancing whether or not these are the right data and whether or not the limitations outweigh the strengths first of all the people under age 65 includes the the Social Security Administration categorized disabled people this isn't all people with disabilities this is people who the Social Security Administration has said are permanently disabled and then there always is a data lag that we've got to keep in mind there are non covered services so there are things the Medicare program does not cover and those will not be in the data over time the Medicare program is certainly expanded their preventive services but there are things cosmetic surgery not for reconstruction for example that will not be in the data so if those things matter to you it's important that you keep that in mind they're not covered by the Medicare program for example from the VA or the Indian Health Service will not be in these data if you've got a population that you think is going to be a heavy va pop using population this might not be the right data set for you and then if you remember managed care so people who are in managed care the only data we will find for them is their Part D data this is not a clinical database we don't know we don't know the reasons for tests so people will often say well the only justifiable reason for this test is whatever but in fact we don't know that for sure we don't know that we don't know the exact reason for the test we don't know the results of the tests we cannot tell if treatment was offered but not received because because somebody turned it down or because of a scheduling problem and symptoms and behaviors are inconsistently reported the other piece in this matters in some situations is these are observational data so these treatments are not randomly assigned so if you care about random assignment you're going to have to look into things such as an instrumental variable approach or other matching or other statistical approaches to to approximate randomization and there it may be hard to control for unobserved variation and then which gets me to the final point which is you don't know what you don't know there's out there often things going on that we can't see that you have to infer by using by comparing groups and saying if I'm right would this be the pattern I saw the data are tricky and if you remember from earlier most people have more than one each one of each type of Medicare record in a calendar year many events result in multiple separate bills many of the bills are divided into sub records so that managing the data flows is one too many too many too many structured takes time and it's important particularly if you're counting things if you're trying to figure out how many x-rays did somebody have understanding the structure and making sure you manage it correctly is really important understanding Medicare payment policy may be needed to understand what you're seeing and ultimately you're going to need to create rules to translate bills so there is no variable that says were they diagnosed by ultrasound you're going to have to create that variable but I also want to point out some work that's been done by our group that could have only been done based on sere Medicare data we found higher rates of hip fracture after radiation therapy for women with cervical and rectal cancer and for men with prostate cancer that needs population base needed large numbers we also found Howard higher rates of rectal cancer for men receiving radiation therapy for prostate cancer we found a reduction in androgen deprivation therapy for non indicated use after payment reduction but no change in use for indicated conditions again these are all things that we found with these data that we would not have been able to find without you notice it's a mix of treatment outcomes and studying payment policy of understanding treatment patterns and these are all excuse me amazing opportunities that are possible with these data so for this reason I'm delighted that you are all here to learn about it and hopefully we can make your use of the seer Medicare data a little easier and we now have time for questions and I understand that there have been some popping up I see them on my screen a little bit so what I can do will just take a few minutes here I'll read the question and Beth can kind of provide her insight into each I think one of the first questions were how how are some of the states selected for inclusion in the registries the states applied to be part of the registries and they applied and they made the case that they were capable of meeting the seer standards that they had population-based reach and that they had an organizational structure and they also said they were willing to have their data shared as part of this fear consortium question a second question is if someone has multiple cancers but their age of diagnosis for their first cancer may have been prior to Medicare diagnosis would that cancer what that individual included in the data and I apologize for not being clear here for each cancer there will be a series of variables so for cancer one and again for cancer two and again for cancer three we will know the cancer type the year of diagnosis age of diagnosis the stage the histology the tumor size the node so those are repeated up to ten times so for each cancer we will know and that also means we can tell so for example how far how what was the timing between the diagnosis of cancer one in cancer two so when we were looking at the study of second cancers we sort of said well what are the different mechanisms that second cancers can happen and how long would it take for a treatment related second cancer to occur and then we could use the timing between the diagnosis of cancer one in cancer two to set to categorize at whether we thought it was treatment related or not okay another question has to do is just the availability of the data and you would touched on this but maybe to rehire in a seer registry but they travel outside of the area to receive their their treatment would they still appear in the data so somebody if somebody lived in a seer registry and they were diagnosed in the seer registry so their first diagnostic workup happened in a seer registry or often the seer registries have strong relationships with adjacent hospitals so sometimes there can be border crossing so somebody lives on right near the border between a seer state and another state and in fact the closest town closes large town would be across the river or across the state border they will have arrangements with those hospitals to do it so using a Minnesota example if somebody lives in the southwest corner of Minnesota they may go to Sioux Falls South Dakota for their cancer care because that's in fact the largest hospital that's nearby and the Minnesota Cancer Registry has very strong relationships with Sioux Falls to make or that the South Dakota Hospital let's the Minnesota hospital know about these cancer diagnoses and all of the cancer registries have similar arrangements and for this reason there's also a number of questions about prognostic factors such as ER and PR status or other tumor markers are those available in the data and where would we find them so as I said the tumor markers will vary from tumor to tumor for example with estrogen receptor status there are I think now five levels so it's after the receptor positive borderline negative not tested unknown there are series that for prostate cancer the Gleason scores are categorized in ways and you may or may not like how it's categorized or what data are available but the record layouts will have detailed information about what information is there and what values are possible but it's really going to depend tumor by tumor but those would all be in the peds F file and many of them appear in site-specific factors which it may not be an intuitive place to look but that would be one of the places apart from having a separate variable where you might look okay another question that we've had it there seems to be a few questions around the link tier Medicare data versus the unlink your Medicare the unlinked to your public use files and whether or not those are the same data and the answer is that there for the people 65 and older they are large they're the same people there will be different variables that are released in each and for different programmatic and scientific reasons the structure of the files is different so you can't merge the sphere public use data with the seer medicare data those IDs will not line out and so sometimes we've got studies that we will use the public use data because it's got it allows for something we want to do that the seer Medicare wouldn't and often it's the reverse so they're similar they're based on exactly the same cases they have different case IDs you can't combine them and you really are going to have to for your tumor that you're interested in and for your period of time look them over and decide which one is the right source for you if you also have separate request processes that we'll be talking about later today okay you introduce the concept of managed care why it is this important and how do we know when a patient is lost to manage here so managed care as we said is important because people who receive their Medicare benefits through managed care will not have claims in the claim filed other than for medications for Part D and for hospice those are the only two exceptions and so it's important to me because I want to make sure I remove them because I would I wouldn't want to mistakenly classify them as not treated so it's so we can distinguish them between people where we can't see the treatments they got and we can see the treatments they got and they didn't get it and so that's why managed care is important in the PDF file there there hm there they call them HM HM Oh in the files but their HMO enrollment Flags that can be used and we'll be talking about that in the next segment and about how to use them but this is the reason because it is such a important factor we will talk about this repeatedly it's a mistake that sometimes people make because they don't understand the consequences of including people with managed care in their analysis but again there are flags in the pad staff and in the some de nom so that you can make a decision about whether to include them or not thank you alright we have a number of other questions that are very treatment or cohort specific so we'll do a number of things hopefully a number of your questions will be answered throughout the day as we cover a variety of other additional topics however if your your questions are not answered we are working to find a mechanism to answer these questions and get them posted after the webinar so don't worry we will get your questions answered it may just be a question of timing