Kira Ryskina, M.D. – 2019 IHPI Emerging Scholars Exchange Presentation

– So I have the pleasure of
introducing Dr. Kira Ryskina, who’s visiting us from the
University of Pennsylvania. She’s the first IHPI Emerging Scholar who’s visiting us this year. So Kira started with her
undergrad studies at Yale, and then has basically
been working her way down the East Coast since then. So medical school and
residency in New York, and then to Philadelphia and Penn for her General Internal
Medicine Fellowship, where she’s an Assistant
Professor of Medicine, and a Senior Fellow at the
Leonard Davis Institute. So from her personal
statement when she applied for the program, she said she’s
interested in understanding how physician practice patterns
and organization relates to the cost and quality
of care for aging adults with multiple chronic conditions. And to do this work, she links a variety of survey, administrative billing, and other secondary data sets
to answer these questions, which is like sort of a topic that’s like catnip for this audience. So I know we’re all looking
forward to your talk, and we’re really glad the
IHPI could provide the reason for your first field trip to Michigan. So thanks, Kira. (applauding) – Thank you, so much, for having me, I had an amazing day meeting many of you, and I really appreciate all
the hospitality and advice. So I will talk about
the role of physicians and advance practitioners
in post acute care outcomes of patients discharged to
skilled nursing facilities. These are my disclosures. Most of the work here was supported by the National Institute on Aging. Here are my objectives that I proposed, and as I was putting together the slides, I realized these are really ambitious. I will try to touch on most of them, and I would love to have
a discussion at the end, to be able to talk about implications, your thoughts, and potential next steps. So I’d like to start with this slide. A lot of people are gonna,
“Oh, I’ve seen this before.” For some reason, this really stuck to me, and just basically a
US Census data showing that in 15 years, we’re
gonna have as many adults over the age of 65, as
children under the age of 18. So a lot of people I
talk to about what I do, I have a hard time explaining it. It has to do with skilled
nursing facilities, but they’re really nursing homes. And then, what are nursing
homes, and why do we care? I see this glazing over sometimes, when I talk about nursing facilities. But, there was a recent,
I mean, I think this sort of drives home the point, but also, there’s a recent NPR
study that estimated that of the adults aged 51 to 65
today, over half are going to spend some time in a nursing home. So basically, all of us
are gonna spend some time in a nursing home, if things
go the way they’re going now. So I’m gonna go over some of
the language, because again, it’s very wordy and confusing. And I apologize, because I
keep going back and forth between some of the terms. So what are skilled nursing facilities, versus nursing homes? I’m currently a hospitalist, or somebody who works in the hospital. And I discharge a lot of my patients to skilled nursing facilities. Skilled nursing facilities,
typically are places that are co-located with
the, in a nursing home. There are patients often in the same room, or on the same floor,
sometimes they’re separated. Skilled nursing facilities
have to be certified by Medicare and Medicaid, so they have to meet certain requirements. But there’s really a lot of overlap, in terms of their physical
place, the staffing, as well as regulations. One thing that’s different, is that skilled nursing facilities are typically reimbursed by Medicare. And Medicare pays for 21
days of short-term rehab, or skilled nursing needs. And then, they subsidize up to 100 days per episode. And after that, Medicare, some
of the patients transition into a long-term care,
or require more care. But Medicare doesn’t
pay for long-term care. So those patients are
then either on Medicaid, or are self-pay, or you know, private-pay. And they end up being
nursing home patients, if they require custodial care. For example, there are often
patients with dementia, or some sort of long-term disability that requires long-term
facility custodial care. Why else do we care, other than the fact that we’re all going to be
in a nursing home some day? So a lot of the public
awareness of post acute care has been around the fact
that it’s very costly. Just to kind of drive home the point, these are some estimates from 2016. The payers are different,
and so, the big house is acute hospital care,
public payer spending. So Medicare spent about
140 billion dollars. Medicaid about 50 billion. For SNFs, the skilled nursing
post acute care piece, Medicare paid about 30 billion. And for nursing home, Medicaid
paid about 50 billion. I should also mention that evidence shows that about 75%
of the variation in spending for Medicare has been
attributed to variation and use of skilled nursing facilities. So there was a lot of variation
in the use of this benefit, and it drives a lot of the
variation in costs for Medicare. And then finally, nursing facility stays are very common. This is, again, for an elderly population that’s Medicare insured, except
for the nursing home piece. So there were about 10
million hospital stays in 2016 that Medicare paid for. About 25, 2.5 million of
them ended up in a SNF. And then, for the nursing home,
these are long-term stays, so this is 1.3 million
residents in a nursing home. The trends in sending people
to a skilled nursing facility after a hospitalization,
are not really changing. There was a little bit of a
uptrend in the early 2000s. This is a recent paper by
my mentor, Rachel Werner, that looked at institutional
post acute care and included both in-patient rehab and skilled nursing facility care. I will say, that in-patient
rehab has very strict criteria that has to be met, to
get into in-patient rehab. Whereas, skilled nursing
facilities are much looser. So there’s a lot more variation. I don’t have a pretty graph just looking at skilled nursing facility,
but it looks about the same. Most of this facility-based
institutional care is actually in SNFs. And lastly, what do we know? So let’s say we’re paying a lot of money, a lot of people are using this service. That’s great, people are getting rehab, they’re getting stronger,
they need the service. So unfortunately, as
you’ve probably heard, and many of you have published on this, quality of care outcomes at
skilled nursing facility care for short stays, sub acute
rehab, are not great. This slide actually shows a public reporting program
that the government has, Medicare has, called Nursing Home Compare. They have this five star
rating, that looks at staffing, health inspections, and some quality, clinical quality indicators. This program was instituted in response to concerns about poor quality in the skilled nursing
facilities, and poor outcomes. What is shows, is that
between 2009 and 2013, the proportion of facilities
that had the five, or four star ratings, increased a lot. Unfortunately, what we
found, and others before, other researchers also substantiated this, is that the quality that is measured by the five star rating, which shows improvement,
doesn’t really correspond to the things that we care about, like readmissions, mortality. So this slide shows on the Y axis potentially preventable hospitalizations. It’s a little bit of a strange measure, or unusual measure, where we looked at the number of potential
preventable hospitalizations per 100 patients within
each 30-day interval the patients spent in a nursing home. And what you can see here, is that before a five star release, there was really a very
minimal correlation between the five star rating and the number of potential
preventable hospitalization. But, actually, if
anything, it was going in the right direction. Whereas, after five star release, the line is flat, so you’re
at risk of being readmitted from a one star nursing home, is similar to a five star facility. So I’m just gonna leave that,
and like I mentioned before, there’s a lot of data
talking about outcomes of these facilities. And you’re just gonna have to
trust me for a little while, that outcomes of skilled
nursing facility care are generally poor and variable. And I’ll kind of circle
back toward the end, and show you some more of the numbers. I should also acknowledge other efforts that are out
there, that are being tested and implemented, to improve
nursing home care quality, or skilled nursing facility
post acute care quality. There are, obviously,
regulatory and legal efforts, certifications, I mentioned a couple of the public reporting initiatives. And then there are sort of specific multi-factorial interventions
that have been implemented, like INTERACT, telemedicine-type
connected care models. And, of course, like grassroot
quality improvement efforts. A lot of healthcare
organizations have also started looking at post
acute care as part of like ACO, value-based payment
models, trying to improve efficiency of that component
of the hospital stay. So I’m not gonna talk about any of those interventions, and frankly, nothing seems to be the magic pill, just like is common in healthcare. What I was interested,
once I started looking at this space is, what actually
happens when physician, when patients are leaving the hospital and going to a skilled nursing facility, from the perspective of a physician? Working in the hospital and in my training and also in my work in
the clinic previously, I very rarely interacted with a physician who was taking care of may patients in a skilled nursing facility. Once I started studying this
field, I started to work with medical directors and other clinicians who worked in the field. And the things they were
telling me kind of did not correlate very well with
what I was telling patients to expect when I was discharging them, to get a little rehab. So what we did, this paper actually just came out in Health
Affairs, but it’s based on a whole herd of Medicare
beneficiaries discharged between 2012 and 2014 from
an acute hospital stay, who went to a skilled nursing
facility for post-acute care, and all of their physician
billing data to Medicare. So what you can see here, is that on the X axis, it’s the number
of days since SNF admission to the first visit by a physician that we saw a claim for. The Y axis is the proportion of stays. So to summarize, about 10% of discharges, we actually didn’t see
any physician visits. They had other Part B billing, but not for a physician
visit during their SNF stay. And while most, about half of the visits occurred in the first 48 hours
after admission to a SNF, there were other visits that occurred on day three or four, et cetera. So we were on one hand,
surprised to see that that was a little longer, we knew that a lot of these
patients weren’t gonna be seen the next day, or even in two days. This sort of potentially made sense to us, if sicker patients are being
triaged and being seen first. You know, I’m a physician, so I understand that I can’t see every single patient at the same time, fairly, you know. So we wanted to estimate the risk, or the, if we modeled the
probability of being seen during their SNF stay,
and also, model the time to their first visit. Based on all the patient characteristics that we could assess, things like how many times
they went to the hospital in the prior year, how
long their hospital length of stay was, Alex Houser (mumbles) Index. So I kind of tried to
capture if physicians are triaging sicker patients, to be seen sooner, can we see
that in the data, clinically? And then at the same time,
we included variables that were nursing home characteristics. So if physicians were seeing
sicker patients sooner, we expected to see those differences in the facts from our models. And I unfortunately don’t
have the space to show you the patient characteristics. But the reason I’m not
showing those to you, is because they’re in the paper, but there’s not a lot of variation across clinical characteristics. It’s not like people with
higher Alex Houser Index we’re being seen sooner. There were no demographic
differences that we could see. So there was nothing
really striking there. What was striking to
me, were the differences based on location of the nursing home. So rural, oh! (laughing and clapping) Rural–
(people mumbling) Did I do something? – [Woman] No. – So rural nursing homes were, you were less likely to be seen period. And your time to first
visit, the number of days, you had a higher IRR for that. There was also geographic differences and differences based on
the size of the facility. So patients who were
in the smaller facility were less likely to be seen and also had more days to the their first visit, kind of adjusting for all the other things that we put in the model. And then, when we tried to publish this, a really valid question
that we were interested in, we looked at before, but we saw just we weren’t looking at it in the most rigorous way,
was so what are the outcomes? This is all great, but
we don’t really care about physician visits, we care about re-hospitalizations and mortality. And the reason we were
hesitant to submit this, was because, it’s obviously
there’s indigent need between being seen by a physician, and like going back to the hospital. So clearly, some patients
might rightfully be sent to the hospital before being seen, because they have a new
issue, where they need to be tested for something like that. So we didn’t attempt to tease
that out in this analysis. I think the findings are
still a little striking, and I wanted to show these to you. So we looked at re-hospitalizations within 30 and seven days, and deaths within 30 and seven days. And this also speaks to the
high risk of poor outcomes of going to SNF overall. And then this measure
of successful discharge from SNF to communities, is
really the ideal outcome. The reason we tell
people they should go to, so they can rehab, is so that they can get stronger and go home. The orange bar is for patients
who didn’t have any visits and the blue bars are for patients who had at least one physician visit. As you can see, there is
a dramatic difference. We, you know, to kind of
try to address this issue of patients being, like reverse causality, we looked at the length of stay for patients who didn’t have any visits and we found that most of them
stayed for at least a week. So this wasn’t like a
early discharge to the SNF, that got sent back the next day. They seemed to have enough
time to see a physician. But again, I should, this is
completely unadjusted outcomes. So that brought up the next question of who are these physicians and are there maybe physicians who are more focused on nursing home care, or skilled nursing facility care? So we were lucky to have access to a public use file that’s downloadable from the internet called,
actually, public use file. But it was provider utilization data, at the line item level. It’s aggregated by the NPI, the Hippocratic Code and
the place of service. It’s now available
through 2016, I think when we looked at this, it
was only through 2015. It’s, again, a publicly available file, so any hick pick level observations with 10 or fewer
individuals, are excluded. So there’s a limitation with that. So small volume services, are
missed for these physicians. However, these are all visits. So they’re specific visit
codes for nursing home care and you can actually
separate skilled nursing wing and a hospital, I mean nursing home wing. Based on the place of service. So we felt that physicians who are seeing, making fewer than 10 visits,
could be ignored in this study. We didn’t know where to start defining a
nursing home specialist. There’s no specialty. There is a (mumbles) or a long term, post acute and long term care society that has about 5,000 members, but they provide some training, but they clearly don’t cover everyone who takes care of patients
in the nursing home. So we wanted to use this billing approach to define a nursing home specialist. And we adopted what hospitals did, which is setting the threshold of at least 90% of line
items for that year, for nursing home services. This is something that
Medicare is now using to define a hospitalist and their claims. So we just went with that and I’ll tell you a little bit more later about whether that made sense or not. We limited this to generalist physicians and general medicine,
family medicine, geriatrics and general practice and
advanced practice providers, physician’s assistants
and nurse practitioners. We’ll also use some data
from the long term care focused database to get the
number of occupied beds, in each nursing home. So to normalize this across
different facilities, because facilities were
opening and closing and the number of
occupied beds also varied, over this time period. We created this measure of the number of nursing home physician specialists and advanced practice providers per 1,000 occupied
Medicare certified beds. And this is what we saw,
which was surprising to me, but also intriguing. So between 2012 and 2015, the number of prescribing clinicians who specialize in nursing home
practice, increased by 34%. There’s a total of about 30,000 clinicians who were taking any, who
submitted any billing for a nursing home care, during that time. And about 7,000 of them were exclusively, basically more than 90% of the billing was from the nursing home. The blue bars show overall, rates of specialists per occupied beds. The orange bars are the physicians. Then the green bars are
the advanced practitioners. So you can see most of the increases among advanced practitioners, most of those are nurse practitioners. We also saw, we tried to see whether this is kind of different areas, catching up to respond
to the increased use of post acute care. And the changes are definitely
somewhat corresponding to like rise in post acute care use in the North-west, I think, but it’s kind of all over the place and one thing that you can see, is that all the red spots
are an increase in SNFists, so there is definitely more red than blue. It didn’t result in a
homogeneous level of SNFisism. (laughing) I had to say that. And then we also looked specifically at post acute care specialists. So these are people who, plurality of their claims are for the post acute care services provided in nursing home,
versus long term care. So the implications of this were, I mean we were kind of excited that there’s clearly a trend here, as somebody who works in the hospital and interacts with hospitalists a lot, the hospitalist trajectory came to mind. This is obviously, if anything, it’s at the very beginning of that, but it’s certainly interesting to study. So prescribing clinicians who
treat nursing home patients, increasingly specialize in
this practice, it seems, at least relative to other
services, provided to Medicare fee-for-service beneficiaries. But whether this is a response
to increased pressures to improve quality, whether it’s driven by
physicians, or the nursing homes, was not something we knew about and also. So basically, our next step was to ask, is there any correlation with outcomes, or any kind of quality measures? So we did this ecologic analysis, also using publicly available data, to measure the association between regional nursing home performance on clinical measures and the prevalence of nursing home specialists. We used nursing home compare data base and selected six quality measures that we deemed to be relevant to, or be under control of
a prescribing clinician, like a physician, or a nurse practitioner. Everything was aggregated at the HRR, the Hospital Referral Region level. And again, weighted by nursing home beds. And then we used the same
provider public use file, to measure, to identify
nursing home specialists. The models were adjusted
for a commonly used nursing home characteristics
and we had some data about the level of acuity, or case mix for each nursing home. Also using LTC focus data. So this was again, very
much just a correlation. But it was interesting that, it was going in the right
direction and it was significant. We looked at anti-psychotic medications for short and long state, those
are two separate measures. Restraints, catheter use and depression and
urinary tract infection. Depression, obviously, actually, is going in not a good direction, but we hypothesize that maybe, or we explain this as perhaps, diagnosing more depression, or identifying and treating depression, which is what would flag you for that measure is a good thing. And then the journal asked us to split it up into physicians
and nurse practitioners. And the anti-psychotic
medication use was less and the urinary tract infection, I’m sorry, the urinary
catheter use was also less, whereas it looks like
the depression measure association was stronger
for the nurse practitioners, but not the other measures. So this was sort of encouraging. Obviously, it was a very
simplistic correlation in trends, geographically. We again, tried to think about depression, what it meant and we’re not too bothered by the fact that there was more depression correlated with more
SNFists, because again, it seems like the measure
itself is a little bit flawed. But ultimately, we wanted
to really get to the bottom of the question if your family member, or if I were going to a nursing home for sub acute rehab and I had the option to choose between a SNFist and non-SNFist, who would provide better care? Or who would result in better outcomes? So that’s the main focus of my talk. I wanna measure the effect, we tried to measure the effect of physician and advanced
practitioner specialization on post acute care outcomes and we wanted to also measure the cost associated with
being treated by a SNFist, versus a non-SNFist. Based on this prior work that we’ve done, we hypothesized that SNFists, actually will result in
better care at lower costs, or better outcomes at lower costs. Very briefly, about how we did this. We analyzed discharges
from acute care hospitals to skilled nursing facilities
for, this is short stay, for Medicare fee-for-service beneficiaries over 66 years of age. Physicians and advanced practitioners with at least 90% of claims
from nursing home care, were considered
specialists and we did some (mumbles) analyses I’m
gonna show you in a minute, that looked at different
thresholds of specialization. Patients were assigned to
the attending clinician, based on during their SNF stay, the plurality of claims
that were submitted and then that physician
was identified as a SNFists or a non-SNFist, based on their entirety (mumbles) of claims. And those were the three outcomes I’m gonna talk about in
a second, a little more. So to do this, we used MedPAR, which is part E claims for
SNFs and hospital stays. Minimum Data Set is a
resident assessment data set. It has a lot of rich,
clinical information, treatment information, as well. As well as some social factors, that is commonly used to study
the nursing home population. These assessments have to be done like five, 14 and 30 days after admission for a short stay and then quarterly, or I think 60 days and then quarterly? So there is a lot of data there and we use that admission
assessment usually. And then MD-PPAS is a data set that is, I’m not sure how commonly it’s used. It’s physician level data set that provides the aggregate
of the number of beneficiaries they saw in a year and also the
number of services provided. So it’s sort of the
denominator that we used for each physician to
determine that proportion of claims that they submitted
from the nursing home. We tried to get part B
data for all the physicians who saw these patients
and it was something like, because Medicare goes by beneficiaries, so it was 15 million
beneficiaries were touched by the physicians we’re talking about. I guess now I’m not surprised about that, but that was an interesting one, I heard that the first time. So the study went from
January 2012 to December 2014. And we included two
million hospital discharges from 14,000 nursing homes. They’re about 16 to 17
thousand nursing homes, depending on how you count
certain Medicare certified, versus Medicaid certified. And we had about 50,000
NPIs in the sample. So let me tell you a little
bit more about the outcomes. We wanted to focus on like a
global measure of good care and we kind of just went
with well accepted things that Medicare and other
health systems use. So we used the Medicare 30 day re-hospitalizations specification, which is an unplanned
readmission to a hospital within 30 days of SNF entry. Successful discharge to community, which is defined as
discharged to community within 100 days of admission to SNF and the patient remained alive and was not readmitted
to a hospital or SNF, or another facility for at least 30 days. And we also estimated
payments for 60 days. We actually started from
the date of admission to the hospital because, we felt that that’s how payers really
look at the episode of care between the hospital and
followed by a SNF discharge. We also wanted to look
at hospital payments as a way to compare the populations. And so the 60 day payments,
they were also adjusted so if somebody spent more time in a SNF than 60 days, we adjusted the total payment,
for the number of days that was still in that episode. So the payments included
facility and professional claims, from Part A and B,
which includes hospital, SNF, physician and
other provider payments. Unfortunately, we didn’t
have outpatient claims, those are ED claims. We didn’t have hospice claims, but hospice patients
were excluded from this. And we didn’t have a couple, Part D medication costs were not included, although most of them are billed under Part
A for the facilities. All right, key and
variable interest again, is whether the nursing
home attending clinician, who was taking care of that patient, specialized in nursing home practice. And we modeled the specification included a lot of patient characteristics,
demographics, race. We used Alex Houser (mumbles) variables, because we were looking at readmissions. Medicare actually uses Charleston. We used Charleston just to check and it didn’t make a difference. And then there’s a long list of variables that I can provide later, is in the paper. I think it’s in the appendix
of clinical variables from MDS and a few variables from claims that we used to try to risk adjust. We used nursing home fixed effects. So in effect, we were comparing patients in the same nursing homes
that were being cared for by a SNFist to patients
in the same nursing home that were under the care of a non-SNFist. We also tested random effects models and tried to estimate like a hierarchical random effect model. But we had 50,000
physicians, so that sort of, that didn’t work, but they’re a simple random
effect models also tested. So this is our patient characteristics. Pretty much what you see
for a patient population that is in sub acute care, or in a SNF. I don’t have much to mention here. I wanted to mention a little
bit about the clinicians that we saw, because I
think it’s really important and it’s not I think there’s a lot more to be done in this area. So there are major
differences between SNFists and non-SNFists. This is at the state level, because all of our analyses
were at the state level. But we are working on a paper to quantify this at the physician level, which obviously makes more sense. But not surprisingly, a
large proportion of SNFists were nurse practitioners, that’s something that
we saw in prior slides. A lot of them were physiatrists, who are PMNR physicians, that I actually don’t
interact very closely with. So I had to look up whether
this is even makes any sense and apparently a lot of PMNR physicians, actually are primary care physicians in skilled nursing facilities. Obviously, they are also
for sure taking care of acute rehab patients,
but I didn’t know that they worked in SNFs. And then the distribution of
specialties is very different, as you can see for non-SNFists. And then we wanted to quantify
the number of clinicians that were seeing these patients. So for SNFists stays,
there were more clinicians actually seeing the patient
and they were providing more, the number of visits was higher, which kind of maybe is
what we would expect. They have more time to
spend with the patients, or what we would want if we had a SNFist. The number of SNFs with any
of that physician’s patients basically how many SNFs does a SNFist, versus a non-SNFist go to? And this also somewhat makes sense. So people who don’t specialize
in nursing home care, actually go to slightly
more nursing homes, but have fewer patients in each. And the last measure the
proportion involves SNF, is it’s made to that patients
SNF, is higher for a SNFist, which I think also kind of
makes sense, on a face value. So this is the big finding. So these are adjusted 30
day re-hospitalizations for non-SNFists, they
were about 1.5% higher. So this is like an absolute difference in the rate of hospitalization. Successful discharge to
community was not as impressive. Still, SNFists were doing
slightly better job, but only .8% absolute difference and the base line rate
was obviously much higher. And then the interesting finding that we had some trouble explaining, or wrapping our heads around, was the payments. So you would expect that
a lower re-hospitalization rate would produce some sort
of savings for the SNFists. And I apologize, I couldn’t
put all the numbers in there, but again, SNFists are yellow bars and non-SNFists are blue bars. And the number indicates a difference SNFists minus non-SNFists. So SNFists actually had higher costs. They had higher costs for overall. Slightly higher costs
for the hospital stay, much higher costs for the SNF stay and then going along with the findings of better discharge to community and lower re-hospitalization
they had lower, their cost savings associated with those two re-hospitalizations
and readmissions to SNF. Part B costs which included the professional claims and
visits were slightly higher. As we would expect. We were trying to understand why were the costs so much higher for SNFists for that SNF stay. So what we did was we adjusted, added to the adjusted model rugs and it looked like the differences were
attenuated a little bit and then we looked closely at the rugs that were used for stays under SNFist care and those were systematically higher rugs. So they were, these patients were getting higher intensity rehab care. Or, they were being charged
for high intensity rehab care. And then the other interesting
thing that we tested, we used different thresholds
and well I’ll just tell you. This is how we did it, there
are other ways of doing it, but it’s kind of interesting
that, so the darker bars are 90% of your claims
are from the nursing home and the very light bars are only 30% of your claims
are from a nursing home. We kind of tried to
address the question of, is there too much specialization? Is that potentially bad,
what’s the sweet spot? And you see for readmissions, it sort of starts dropping off at the 50%. But for successful discharge
to community interestingly, the highest difference, most benefit that we
were getting for SNFists was at the 80% threshold. So there’s more work
to be done in this area and I think it might speak
also to the mechanism of how these physicians
or advanced practitioners well physicians, are needy,
getting these outcomes. And then one last thing
I wanted to mention, that also requires further digging into, was that we stratified
our findings into stays where the primary
clinician was a physician, versus stays where a primary clinician was an advanced practitioner and actually saw that most of our findings were concentrated among the
stays where the advanced, the physician was, the clinician was a physician and not
an advanced practitioner. So there are numerous
limitations to this paper. We tried to account for
differences across SNFs by including SNFist facts, but we couldn’t really
adjust for patient selection. Conceptually, we think that
if there is a SNFist in a SNF, perhaps it would be, perhaps unusual, or we wouldn’t expect them to select out healthier
patients for themselves, because that’s just not
why they would be there, but that’s something else that we need to work on in future studies. There might be other unobserved clinical, social, or time varying hospital factors that we couldn’t control for. Again, this is generalizable
only to nursing homes that have both SNFists and non- SNFists and this SNFist movement,
if it is a movement, is so, so early in the process, that a very small
proportion of nursing homes have both types of clinicians. It’s unclear how, most
importantly I guess, it’s unclear how the specialists achieve, what is the mechanism for this association that we observed and
is it really related to the specialists, their
knowledge, the time they spent. Is it somehow related to other staff that is working with
them, versus other people? And lastly, there were some data sets that we didn’t have access to, like emergency room visits,
observations, stays. We also didn’t include any clinic visits, outside of the nursing home. So in summary, I think I’m
gonna leave you with this, that it looks like there’s
a change in the makeup of prescribing clinicians
who treat this population. One in five now are specialists, but there is a lot of variability. Specialization, it looks like, might be associated with better outcomes in mixed staff facilities, but at higher cost. And in part, those costs are due to more intensive rehab services,
which may be a good thing, but also probably requires more studying. And then lastly, based on
these kind of dose effect sub analyses, we would in the future, probably use a lower
threshold for specialization of about 70% is probably where the most, the widest difference between
specializing clinicians and non-specializing clinicians was found, at least in our data. So I would say that based on our data, which is really early, health
systems aiming to improve post acute care outcomes,
might consider strategies that shift their patients toward
a specialist, a SNFist. Especially, given all
the efforts in reducing 30 day readmissions. I was really impressed by the
effect size in that measure. It also, one of the concerns
of having a specialist, is that they might wanna keep
that patient for a long time, leading to an institutionalization
of these patients and that’s not what we saw, based on the successful
discharge to community measure. But there is a lot of heterogeneity across clinicians and
specialization thresholds and all that would be included
in the next steps for us. These are the next steps
that I’m thinking of. I’m sure there are many and I know we’re kind
of running out of time. I wanted to get your thoughts on this. And lastly, I just wanted to provide you with some updated slides with 2016 data. 2017 data, I think is
coming up in a month. So since 2012, there was like a 54% change in the number of specialists
per occupied beds. And in case, just as a
sidebar, we did check. There wasn’t a marked change
in the number of occupied beds, so it’s really moving
on the numerator side. I wanted to thank my mentors
and other, the data sources and the Leonard Davis
Institute of Health Economics for supporting this work and for having this exchange program and obviously, I wanna thank
IHPI for having me here. (applauding) – [Woman] Thank you, (mumbles) thank you for this presentation. This is near and dear to my heart, as you know, as we talked
about this earlier today. And one thing I really like about it is, I think a lot of people in
the medical community forget that nursing homes and
skilled nursing facilities, physicians and clinical acute people still have a very important role and often it’s a bit out
of sight, out of mind for a lot of us and you’re
bringing attention to it. A couple comments and just
a couple more thoughts about the future. First of all, I wanna applaud you for using the public use
files in a creative fashion. I have a junior faculty member here who were really trying to
maximize the use of free data and there’s some really
good stuff out there, so that’s great. And then, on your last slide,
about the rise of SNFists, the timeline correlates with ACO. ACO expansion, I would love
to see that across areas with high ACO concentration versus lower. (mumbles) We’re really actually showing, when you look at specific
rehab populations, less benefit of the ACO movement for post acute use, although studies that have
taken generalized populations have shown reduced costs. We’re not seeing it in groups who really use a lot of post acute, so there’s something interesting going on in that environment. And then just to go way back,
to one of your early slides, I’m really curious about
the no visits work. There’s a couple reasons. First of all, from a
regulatory standpoint, it’s– – It doesn’t make sense. – [Woman] Well not that. So the question is really, are you able to capture, I’m sorry, this is a really weedy,
technical question, that there are types of providers
completely in rural areas, whose claims don’t show up
in our normal carrier file. So I’m wondering if you’re
able to pick those up and being a little cautious
about making a comment that in rural communities,
people aren’t getting good care, they’re not getting any visits, are we actually missing
day to day examples with people of (mumbles)
in rural health centers and actually critical access hospitals, which could have swing beds and all these other kinds of things. So I wonder if you could just, in addition to taking the compliments and thoughts for the future, speak a little bit to that rural finding. – Yeah, so when we saw those results, the first thing I did
was, I went to Rachel and I said, “This doesn’t make sense. “What’s wrong with my data?” Right? Like and then we dug deeper into this. We talked to ResDAC and
that didn’t go anywhere but, (laughing) so, but, here’s what I know. So we did like a pretty thorough thing, where we looked at whether
they had any Part B claims. And they do. We also looked at, how long they were in the nursing home. We excluded people, or we
did a sensitivity analyses that didn’t have time to get into here, where we excluded anyone
who was in a nursing home in the prior year, to kind
of weed out people who, the physician might know. We also included clinic visits. So we included Part B claims. And kind of to just step back again, in terms of regulations,
federal regulations require a physician visit within 30 days. But they also require that a physician is involved in the care plan and that plan should be in
place in about seven days. So, and then I worked in a nursing home and there were certainly
patients who were admitted and sent back to the hospital before I knew anyone had
got their hands on them. Which might be a perfectly appropriate and then there were patients
who were admitted on a Friday, but really didn’t get, so
they might’ve been seen, like people waved to them from the door, or just looked in and
made sure they were okay, but didn’t actually put
in the full Medicare visit that requires a physical exam and like a full thoughtful assessment. So that’s, I think your
point is very valid. There might be some visits, or some care that physicians are not billing for, because they’re not meeting all the Medicare requirements for billing, or they’re billing it for it later, like they’re providing the
services that are needed, but they’re making the
full assessment later when they have time. – [Woman Those visits I mentioned would be in the outpatient hospital file, they wouldn’t be in the carrier file. – Right, right. The only thing that makes me think that, like it can’t all be, none
of those explanations, like each explanation
can explain some of this, but then there’s like a counter argument. So if these people are
really getting visits, then why are they having
terrible outcomes? If they’re all getting sent
back to the emergency room, well first of all, why didn’t
the physician see them first, but maybe they weren’t
there, but then why are they, like they spent enough time in the nursing home it looks like, to be seen before getting
sent back, on average. So I agree, I guess, but I think it just requires more digging and
hopefully we’ll be able to, what I’m trying to figure out now is, how to tease out and
see if we can actually match the population somehow, or use methods to compare outcomes by teasing out like this indigency issue. (mumbling) (laughing) (mumbling) (applauding)

Leave a Reply

(*) Required, Your email will not be published