On 2014 Apr 17, Farrel Buchinsky commented:

Fascinating work and interesting use of large administrative database with checking against medical record. There was almost double the incidence and prevalence in the Medicaid group. So the obvious association is between lower socioeconomic status and higher RRP burden.

Can the database and analysis confirm or refute (or apply probability to) various hypotheses? For instance:

1) In 2006 many people had no insurance yet many children were eligible for state-provided insurance and could get it. Could dysphonia and dyspnea drive uninsured people into Children's Health Insurance Program (CHIP). The hypothesis may or may not be true, but even if it is true it may be way too small a number to account for the difference. It would be interesting to characterize date of enrollment as a function of date of incidence. Also remember that symptoms often precede diagnosis by a long time (about a year).

2) Does low SES ("Medicaid") cause RRP or does RRP cause low SES ("Medicaid") or does another factor "Y" cause both low SES("Medicaid") and RRP? Does data structure permit one to see if an individual moved from commercial to Medicaid as a function of disease incidence or prevalence.

3) Does the data structure permit one to explore association between frequency of interventions (as merely one metric, albeit flawed, of aggressiveness) and type of insurance?

4) In 2006, what number of the 20 million Americans less than 5 years old had no insurance and what effect would/could that number have on the estimates.

5) In my state at least, a child can qualify for Medicaid based on disease and independent of socioeconomic status. How widespread is it and can that be modeled into the data?

On 2014 Apr 17, Farrel Buchinsky commented:

Fascinating work and interesting use of large administrative database with checking against medical record. There was almost double the incidence and prevalence in the Medicaid group. So the obvious association is between lower socioeconomic status and higher RRP burden.

Can the database and analysis confirm or refute (or apply probability to) various hypotheses? For instance:

1) In 2006 many people had no insurance yet many children were eligible for state-provided insurance and could get it. Could dysphonia and dyspnea drive uninsured people into Children's Health Insurance Program (CHIP). The hypothesis may or may not be true, but even if it is true it may be way too small a number to account for the difference. It would be interesting to characterize date of enrollment as a function of date of incidence. Also remember that symptoms often precede diagnosis by a long time (about a year).

2) Does low SES ("Medicaid") cause RRP or does RRP cause low SES ("Medicaid") or does another factor "Y" cause both low SES("Medicaid") and RRP? Does data structure permit one to see if an individual moved from commercial to Medicaid as a function of disease incidence or prevalence.

3) Does the data structure permit one to explore association between frequency of interventions (as merely one metric, albeit flawed, of aggressiveness) and type of insurance?

4) In 2006, what number of the 20 million Americans less than 5 years old had no insurance and what effect would/could that number have on the estimates.

5) In my state at least, a child can qualify for Medicaid based on disease and independent of socioeconomic status. How widespread is it and can that be modeled into the data?