RDD using all immigrants

[jzhang722]

Hi all,

I'm just copy and pasting what I put in the email a couple weeks ago.

Replicating Mette’s refugee papers: Sample: using a sample of 18-49 year immigrants Years: immigrants arriving in 1997-2001 Lookout period: 3-5 years after arrival Outcomes: same sensitive prescriptions from before and some crude measures of utilization (number of outpatient days, indicator for ED or inpatient visit) Method: RDD plots--plot mean (or boxplot) outcome by year-quarter of arrival (e.g., 1998Q4, 1999Q1, etc.)

An open question is what fraction of all immigrants arriving between 1997-2001 are refugees?

Later: @soniabhalotra also suggested adding an extra layer of differencing by including some control group. This can be matched natives based on region, age, and education, or when we get the refugee indicator, these can be non-refugee immigrants. I don’t know the Mette papers well, but I would be shocked if they did not use a control group at least for a placebo check.

[jzhang722]

Thanks @Lina011 I think for this next iteration let's just do two simple things:

1. Can we double check the numbers, including code and comparison with code @tramir has, and with maybe published reports/papers, check sample restrictions, etc. We're sure you're doing it right, but doesn't hurt to double check. We think ER/Inpatient rates are higher than outpatient?
2. Can we do just the utilization ones for natives? Don't need prescription drugs yet.

If there are other things I've missed please add them @mdaysal @soniabhalotra @tramir.

Refugee register: hoping for early June.

[mdaysal]

Thanks, Jonathan. I don’t have anything else to add.

[soniabhalotra]

Thanks, I agree too

[Lina011]

Hi everyone,

Here are the revised graphs, thank you once again, @tramir for looking through the code! For this version, I made sure that I had one entry per person and I kept people who did not have the hospital visit in the sample.

I tried to find statistics to check the numbers, but it was quite difficult as the papers I found were conditional on hospitalization and focused on different regions and ages or used different surveys. I tried to exploit stats from the Statistics Denmark website (https://www.statbank.dk/statbank5a/SelectTable/omrade0.asp?SubjectCode=1&PLanguage=1&ShowNews=OFF). I could use ages 20-49 and it looks like the share for inpatient and ER is quite similar with the numbers here.

Also, I was thinking about using natives who arrived back in the country in the same period. I checked that there is a fraction of natives who arrived in the country at the same time as immigrants (1997-2001), so I was wondering if it could also be informative to check the visits for those natives too. What do you think about that?

Another thing I am thinking of is to calculate the utilization rates for immigrants and natives with poisson regression based on gender, age, and education. Afterward, I could also match immigrants and natives based on these characteristics as Sonia proposed before.

Lastly, I think I will add the information on the number of days/nights spent in the hospital and top 10 ICD codes to compare the reasons of hospitalization between two groups.

[jzhang722]

Thanks @Lina011 you make a good point that we should have thought through what would "arrival date" for natives mean. I suppose we could plot mean utilization for 3 year intervals without any regard for arrival year.

So the x axis would be 1998-2000, 1999-2001, 1999-2002, etc and the y axis would be average utilization in those years for all natives.

[mdaysal]

I agree with this

[jzhang722]

All of Denmark, ages 35-39 in 2008 (only year I can find both hospital and population easily)

Number of hospital patients: 37,775

Population: 386,883

--> 9.7%

You have ~0.5 for immigrants (is this 50%?) and 10% for natives. Are immigrants really use 5x as likely to use inpatient hospital visits??

[jzhang722]

A few questions I will ask on our call today:

1. How are you defining natives?
2. Why is the ER for 1st gen immigrants so different from what you showed 2 weeks ago? How do the samples differ?
3. Can you please change the labeling in https://github.com/jzhang722/Denmark-Immigrant-Health/issues/4#issuecomment-2124901110 (title and y axis)?
4. I'm still confused why ER mean < hospitalization mean in https://github.com/jzhang722/Denmark-Immigrant-Health/issues/4#issuecomment-2124901110

[Lina011]

All of Denmark, ages 35-39 in 2008 (only year I can find both hospital and population easily)

Number of hospital patients: 37,775

Population: 386,883

--> 9.7%

You have ~0.5 for immigrants (is this 50%?) and 10% for natives. Are immigrants really use 5x as likely to use inpatient hospital visits??

@jzhang722 sorry, I am a bit confused about 0.5 for immigrants as inpatient visits three years after the reform is around 0.15 for them.

[Lina011]

just to address a couple of points before we talk: 1) the definition of natives and immigrants comes from the official definition of Statistics Denmark: Immigrant: An immigrant is defined as a person born abroad whose parents are both (or one of them if there is no available information on the other parent) foreign citizens or were both born abroad. If there is no available information on either of the parents and the person was born abroad, the person is also defined as an immigrant. Descendant: A descendant is defined as a person born in Denmark whose parents (or one of them if there is no available information on the other parent) are either immigrants or descendants with foreign citizenship. If there is no available information on either of the parents and the person in question is a foreign citizen, the person is also defined as a descendant. Person of Danish origin: A person – regardless place of birth – where at least one parent is Danish citizen and also born in Denmark. 2) there was a mistake in my code as I conditioned on people having a hospital visit. We can disregard the previous graphs for hospital visits. 3) sorry, I did not understand how you would prefer to change the labeling. 4) I think it is higher in the case of natives: ER (~0.14) and inpatient (~0.10), but similar for the immigrants.

[Lina011]

updated graphs for natives below

[jzhang722]

Summary of findings in June 2024 before we put the project on hold.

We explored the idea of an RDD using January 1999 language training for refugees.

We do not see obvious discontinuities for all immigrants in inpatient visits, perhaps there is something with ED visits. There seems to be something in 3Y outpatient, but disappears with 5Y outpatient.

Healthcare utilization of natives look relatively stable over the time period.

We next need to zoom in on only refugees (~10% of immigrant sample).

[mdaysal]

Thanks for the summary!

N. Meltem Daysal Associate professor Department of Economics University of Copenhagen Denmark Website: www.meltemdaysal.com

On 6 Jun 2024, at 10.47, Jonathan Zhang @.***> wrote:

Summary of findings in June 2024 before we put the project on hold.

We explored the idea of an RDD using January 1999 language training for refugees.

We do not see obvious discontinuities for all immigrants in inpatient visits, perhaps there is something with ED visits. There seems to be something in 3Y outpatient, but disappears with 5Y outpatient.

Healthcare utilization of natives look relatively stable over the time period.

We next need to zoom in on only refugees (~10% of immigrant sample).

— Reply to this email directly, view it on GitHubhttps://github.com/jzhang722/Denmark-Immigrant-Health/issues/4#issuecomment-2152728872, or unsubscribehttps://github.com/notifications/unsubscribe-auth/APS6N6VVONZVKBWL4TZPFJ3ZGBZA3AVCNFSM6AAAAABHWMHLJSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNJSG4ZDQOBXGI. You are receiving this because you were mentioned.Message ID: @.***>