Hypotheses Testing (OLS Regressions)

[ahmedelfatmaoui]

1. Within the NA racial group, compared to adult mothers, teenage mothers have lower vaccination rates (including both Tdap, and flu), through which their infants have worse health outcomes (still need to test the second part).

• Dependent variables: Binary for whether a pregnant women within each race is of age lower than 20 years old. Note that each race dummy excludes other races (e.g., Indian American Teenage Dummy is defined within Indian American only and all other races were excluded)

• Independent variable: Flu (columns 1 to 3) is a dummy for whether flu shot is taken; similarly, TDAP is binary for receiving TDAP vaccine shot

• Data period: 2011 to 2019 (TDAP and Flu exist only during this period)

[ahmedelfatmaoui]

2. Between the NA and Native Hawaiian groups, compared to Native Hawaiian teenage mothers, NA teenage mothers have equal vaccination rates (including both Tdap, and flu), through which their infants might have equal health outcomes.

This one cannot be tested due to small sample size for Hawaiian group

[ahmedelfatmaoui]

3. Across all racial groups, NA teenage mothers have the lowest vaccination rates (including both Tdap, and flu), through which their infants have the worst health outcomes.

• Effects on infant outcomes

Note: The outcome variable in columns 1 and 2 is the dummy for baby having low birth weight (weight <2500grams)

[ahmedelfatmaoui]

@paghosh @jzhao2jzhao2jzhao2 @RJaggad

Please see the above tested hypotheses. We can also add medicaid expansion later if needed.

[jzhao2jzhao2jzhao2]

Thank you Ahmed @ahmedelfatmaoui for such prompt results -- You are appreciated by all of us @paghosh @RJaggad !

Rashmi @RJaggad and I have interpreted and discussed your results - each table and each main row, we wanted to check consistency with your @ahmedelfatmaoui and Dr. Ghosh's @paghosh interpretation, before typing our interpretation so that your interpretation will be independent and not influenced by ours.

Might you @ahmedelfatmaoui @paghosh please kindly:

1. Interpret your results - if you prefer a call or Zoom to more efficiently discuss/check consistency between our interpretations, please let me know!
2. Paste your general model specification, and control variables.
3. For hypothesis 1, add ethnicity (Hispanic)?
4. For hypothesis 3, break down results by each race and ethnicity (add Hispanic)?
5. Dr. Ghosh @paghosh , you mentioned over the phone that ACA effective in 2014 might affect results - totally agree! Though we might not want to get complicated by either of the two mechanisms within ACA (1.require most insurance impose no cost-sharing of CDC recommended vaccines, 2.expand medicaid enrollment income eligibility) that might affect results. Because whichever mechanism within ACA, it would be the timing of 2014 that might matter as just a control variable for our current research question focusing on teen moms' vax behavior and babe outcome. What do you think? Should we add ACA as a control variable, like a 2014 dummy variable?
   Or other advice you might have?

After the above is clear, we would have another paper ready to write! Thank you all! June

[paghosh]

@jzhao2jzhao2jzhao2, @RJaggad and @ahmedelfatmaoui: Today, I had a meeting with Ahmed to discuss your fifth comment. Since tomorrow is dedicated to teaching, I am providing a response here instead of scheduling a Zoom call. Below are the details regarding your comments:

1. We analyzed five tables—four focused on different racial groups and one combining all groups. Our findings consistently indicate that the probability of teenage mothers receiving a flu vaccination is 0.46 lower for combined races, although the specific estimates vary by race.

While we also observed a lower likelihood of Tdap vaccination among teenage mothers, this result wasn't statistically significant. Interestingly, we noted that reduced Tdap vaccination rates among teenage mothers might lead to lower birth weights, though not to a higher infant mortality rate. The connection between Tdap vaccination during pregnancy and reduced birth weight requires further exploration, and look forward to learn the biological mechanisms from you.

2. Ahmed will share the complete results in a detailed table.

3 & 4. Regarding Hypotheses 1 and 3, we can include results for each racial group except for Hispanic individuals, as our dataset lacks information on this group. However, we can incorporate data from all other racial categories if that meets your needs.

5. In addressing how the Affordable Care Act (ACA) might have varying impacts across states, we have decided not to use a time dummy. Instead, we will analyze how the ACA influences state-specific time trends differently. Specifically, our approach will involve estimating these state-specific trends and then applying the 2014 ACA policy dummy to estimate the differential effects.

Let me know if there's anything I missed out or if there are any additional things you'd like us to discuss.

[jzhao2jzhao2jzhao2]

@paghosh @ahmedelfatmaoui Thank you millions for your precise interpretations and helpful responses!

1. Your interpretations are what @RJaggad and I discussed and had in our minds Consistent! We're good. We have discussed potential biomechanims to explain why flu and Tdap are different, this is because the two diseases differ in their susceptibility and severity, and flu vaccine has more adverse event concerns -- we will cite health/bio literature to support this. No worries.

2, 3, 4, 5. All sound good! Look forward to 2. and 5. results, and I'll send you Medicaid paper outline today!

[ahmedelfatmaoui]

Hi @paghosh

For the state specific time trend, I added as another fixed effect as shown bellow; I also added three individual control variables (the last 3 variables); Please let me know if you think this is the correct way so that I will update all the tables accordingly.

• Here is how I computed the trend fixed effect in stata: gen post_policy = year>2013 gen time = post_policy*year gen st_yr_trend = time*state_fips

eststo flu4: quietly reg flu na_teen $controls i.year i.state_fips i.st_yr_trend, vce(cluster state_fips)

-

[paghosh]

Hi@ahmedelfatmaoui: The model specification looks good

I have a question regarding the outcome variable. Is it binary? If that's the case, it reflects changes in probability. Could you please provide the average value of the Teen Dummy variable? This would assist in calculating the percentage change. Also, could you please use the logit/probit model to see whether the results still hold?

[ahmedelfatmaoui]

Hi @paghosh Yes, it's binary. I will do the logit for robustness after we're done with all the tables--just for the sake to time. About 6.3% of pregnant women are teen:

I see, it should be interpreted as a probability change since both the outcome and the dependent variables are binary. based on column 4, can we say that being a teen mother is associated with a 1.4% lower probability of receiving a flu shot compared to those who are not teen mothers?

[ahmedelfatmaoui]

Hi @paghosh @jzhao2jzhao2jzhao2 @RJaggad, Here are the updated tables:

1. Within race results

reg_vax_within_indian.csv reg_vax_within_wt.csv reg_vax_within_haw.csv reg_vax_within_black.csv

1. All teen results with heterogeneous effects

reg_vax_teen_hetro.csv reg_vax_teen.csv

3. Infant effects infant_effects_hetro_baby_low_weight.csv infant_effects_tdap.csv infant_effects_flu.csv

[ahmedelfatmaoui]

check if medicaid has any significance