Summary Week 11 — Transcript
Full transcript
- 0:02this week we learned about
- 0:05standardization as a technique to adjust
- 0:10for
- 0:11confounding when we're comparing two
- 0:14different
- 0:15populations that have
- 0:18different uh composition differences so
- 0:21populations that have differences in age
- 0:24distribution socioeconomic distribution
- 0:27different demographic differences across
- 0:29po
- 0:30calulations when we look at results from
- 0:33public health
- 0:35agencies or descriptive studies we are
- 0:39often looking at rates that have been
- 0:42standardized in order to be able to
- 0:45compare and draw conclusions about any
- 0:49excess categories or any excess
- 0:53demographic groups that may have higher
- 0:57rates than
- 0:58another standardization helps to remove
- 1:02differences in the composition between
- 1:05po populations that are being compared
- 1:09by mathematical procedures so it's a
- 1:13technique that allows us to anal
- 1:16analytically adjust for a biasing factor
- 1:21on the basis of differences in
- 1:24population
- 1:27composition we uh learned about
- 1:30standardization as involving data from
- 1:33two
- 1:34populations are study populations or
- 1:37populations of Interest which are the
- 1:40ones we are going to
- 1:41standardize and then a standard or
- 1:45reference population that we will
- 1:47standardize
- 1:49against so I taught you about two forms
- 1:52of standardization direct
- 1:54standardization and indirect
- 1:58standardization Direct standardization
- 2:00is always what we would prefer to do
- 2:03because it allows us to then compare
- 2:07rates between study populations that
- 2:09were adjusted to the same standard
- 2:12population so for example if I wanted to
- 2:16compare rates in covid infections
- 2:20between different states in the United
- 2:23States I could standardize them all to
- 2:25the US population as my standard
- 2:28population and then I could compare
- 2:32those rates between states even though
- 2:34they all might have differences and
- 2:36they're underlying population
- 2:39distributions or if I wanted to look at
- 2:43rates
- 2:45of suicide
- 2:47mortality or mortality from different
- 2:52cancers I could do that by adjusting
- 2:55them to the same standard
- 2:57population the choice of a standard
- 2:59population is somewhat arbitrary there
- 3:02are different situations that inform our
- 3:05choice of a standard population but most
- 3:09commonly we're going to choose it as a
- 3:11larger population so for example if I'm
- 3:14standardizing rates across Colorado
- 3:17counties I would choose Colorado as a
- 3:20whole as the standard population or if I
- 3:24wanted to compare rates across States I
- 3:26would choose the us as a whole as the
- 3:29state standard
- 3:31population but in order to do direct
- 3:34standardization we have to have Stratus
- 3:37specific rates from our population of
- 3:40Interest the one that we are
- 3:43standardizing so if I wanted to look at
- 3:46rates in uh covid infection and I'm
- 3:50concerned about age different ages of
- 3:53the population as being a potential
- 3:56confounding Factor then I would need
- 4:00Stratus specific rates of covid
- 4:02infection for the different age
- 4:06categories and for a variety of reasons
- 4:09sometimes we don't have these so that's
- 4:12why we have to turn to indirect
- 4:15standardization but direct
- 4:17standardization is always preferable
- 4:20because it allows us to answer the
- 4:22question how many events would occur
- 4:25among the standard population if they
- 4:27had the same rates as the study
- 4:30population and once we calculate our
- 4:33adjusted rates we can compare them to
- 4:35each other if they were adjusted to the
- 4:38same standard
- 4:40population the situation in which we
- 4:43cannot do direct standardization is if
- 4:46we don't have the stratus specific rates
- 4:49so maybe I don't know what the age
- 4:52specific rates uh covid infection are
- 4:56but I'm still worried about differences
- 4:58in population
- 5:01composition and um then I would need to
- 5:04turn to indirect standardization so we
- 5:07use this when we know the total number
- 5:09of cases in our study population but we
- 5:12don't have Stratus specific rates some
- 5:16of the reasons that we may not have
- 5:18these rates is because we just don't
- 5:20have that data available to us we're
- 5:22turning to a surveillance source that
- 5:25doesn't provide Stratus specific rates
- 5:28but most commonly the reason we don't
- 5:31have these strus specific rates is
- 5:33because they're based on small numbers
- 5:35and when we calculate rates based on
- 5:37small numbers they're very unstable they
- 5:40lack
- 5:41precision and the difference between
- 5:44adding one or two more cases can result
- 5:47in a huge jump in that rate which we
- 5:50don't think is really reflective of
- 5:53changes in the population but it's a
- 5:56problem with Precision because we're
- 5:58dealing with really small number
- 6:00numbers so we turn to indirect
- 6:03standardization which uses strata
- 6:06specific rates from a standard
- 6:08population a larger population in which
- 6:12the stratus specific rates are stable
- 6:14because they're based on larger numbers
- 6:17and we use the total number of observed
- 6:21cases from our study
- 6:23population this is very common in a lot
- 6:27of uh looking at hand
- 6:30rates where we might not we might have a
- 6:33small number of total cases thank
- 6:35goodness but um we don't have large
- 6:39enough numbers to calculate Stratus
- 6:42specific rates or those rates are just
- 6:44unstable because they're based on very
- 6:46few cases so when we use indirect
- 6:49standardization what we calculate is
- 6:52called a standardized morbility sorry
- 6:56mortality or incidence ratio so an SMR
- 7:00or an
- 7:02S and this is calculated as the ratio of
- 7:06observed versus expected number of cases
- 7:10The observed number of cases come from
- 7:13our study population and the expected
- 7:16number of cases come from the rates that
- 7:19we're able to get from our standard
- 7:22population applied to the stratus
- 7:25specific numbers for uh demographics in
- 7:29our study
- 7:30population it answers a slightly
- 7:33different question than we can answer
- 7:35with direct
- 7:36standardization and it answers a more
- 7:38limited question so indirect
- 7:42standardization where we calculate smrs
- 7:45or S IRS answers the question what is
- 7:49the excess mortality in the study
- 7:51population compared to the reference
- 7:55population so for example if I know the
- 7:58total number number of cases of
- 8:01pancreatic cancer in Colorado but I
- 8:04don't know what those rates are
- 8:07according to age group and I think that
- 8:09age could be an important confounding
- 8:12factor I could use the US population as
- 8:16the standard population in which there
- 8:18are enough cases to generate age
- 8:22specific rates and I apply that to the
- 8:26demographic composition of my study
- 8:28population
- 8:30Colorado according to the age groups and
- 8:33I calculate the expected number of cases
- 8:36for each age group that would occur if
- 8:40Colorado had the same age specific rates
- 8:43as my standard population of the
- 8:46US and then I calculate my standardized
- 8:51incidence
- 8:52ratio and I can interpret it as the
- 8:57amount of mortality in my study
- 9:00population relative to my standard
- 9:05population one important limitation of
- 9:08indirect standardization where we
- 9:10calculate smrs and SS is that we cannot
- 9:14compare them to each other even if
- 9:17they're adjusted to the same standard
- 9:20population the reason is because we're
- 9:24using the rates from the standard
- 9:27population in indirect Ste
- 9:29ization so they are not coming from our
- 9:34study population so if I calculate an
- 9:38smmr for
- 9:40Colorado using the us as the standard
- 9:43population and I calculate another one
- 9:46for New Mexico again using the us as the
- 9:49study population with indirect
- 9:52standardization I could not compare that
- 9:55SMR from Colorado to that of New Mexico
- 9:59C because indirect standardization has a
- 10:03more limited interpretation the only
- 10:06interpretation is the excess mortality
- 10:10excess or deficit mortality in the study
- 10:13population compared to the standard
- 10:17population so this week I walk through
- 10:21these calculations in the lecture make
- 10:23sure that you look at the flow of the
- 10:26calculation it can be a little confusing
- 10:28to think about these concepts of a
- 10:30standard population and a study
- 10:34population and then also thinking about
- 10:36our Stratus specific rates from these
- 10:40populations and our Stratus specific
- 10:43demographic composition so the challenge
- 10:47in mastering these
- 10:48calculations is understanding the inputs
- 10:52necessary for direct and indirect
- 10:55standardization and the appropriate
- 10:58interpretations to Aid in helping you
- 11:01master the flow of these calculations we
- 11:04have an applied EP exercise this week
- 11:07please take time to work through that
- 11:09exercise you need practice applying
- 11:11these
- 11:12calculations and then I will host a a
- 11:17virtual live recitation section next
- 11:21week to walk through these calculations
- 11:24as a group remember your applied EP
- 11:26exercises are graded on compl completion
- 11:30not correctness so don't worry if you're
- 11:34not able to work through all these
- 11:36calculations I want you to attempt them
- 11:39and turn in your answers and then we'll
- 11:41review them together as a group to make
- 11:44sure that you understand the flow of
- 11:46these
- 11:47calculations as always I'm available in
- 11:50office hours on Thursdays at noon and
- 11:53I'm happy to work through
- 11:55additional um examples at that time
About this transcript
This page contains the full transcript of Summary Week 11 by Tessa Crume, generated from the public captions YouTube serves with the video. The transcript has 1,409 words across 245 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.