Comparing Health Outcomes of Privileged US Citizens With Those of Average Residents of Other Developed Countries

Key Points Question Are the health outcomes of White US citizens living in the 1% and 5% richest counties better than the health outcomes of average residents in other developed countries? Findings In this comparative effectiveness study of 6 health outcomes, White US citizens in the 1% and 5% highest-income counties obtained better health outcomes than average US citizens but had worse outcomes for infant and maternal mortality, colon cancer, childhood acute lymphocytic leukemia, and acute myocardial infarction compared with average citizens of other developed countries. Meaning For 6 health outcomes, the health outcomes of White US citizens living in the 1% and 5% richest counties are better than those of average US citizens but are not consistently better than those of average residents in many other developed countries, suggesting that in the US, even if everyone achieved the health outcomes of White US citizens living in the 1% and 5% richest counties, health indicators would still lag behind those in many other countries.

This appendix focuses on the data analysis for acute myocardial infarction (AMI).We first consider the micro-level analysis for the 65+ population in three countries with high-quality data from the U.S., Denmark, and Norway, and then the OECD data for ages 45+.

Micro-level claims data for the U.S., Denmark, and Norway
For the U.S. data, AMI was both the primary admitting diagnosis and the patient's first AMI hospitalization using ICD-9 codes 410.xx [except 410.x2].The Master Beneficiary Summary File is used to identify dates of death for patients in this sample.We calculate the shares of patients by age group (65-69, 70-74, 75-79, 80-84, 85+) and sex who died within 30days of their first AMI hospitalization observed during 2013-14.
For the US data, the 30-day case fatality rate is 13.5% (14.6% for women, 12.4% for men; mortality for women is higher because the age distribution for AMI is more heavily weighted towards older ages for women).For whites in the top 5% of counties, the case fatality rate is 12.4%; thus the ratio (or adjustment factor for whites in the highest-income US counties to the US average is the ratio, or 0.92. We collect similar data from Denmark (2012-14) and Norway (2013-15) for the entire populations age 65+.In both countries, the sample was created using ICD10 codes I21.1-I21.4,and I21.9.In Norway, the case fatality rate (weighted by US population weights for age and sex) is 10.2% (11.7% for women, 8.8% for men), N = 20,496.The corresponding measures for Denmark is 10.7% (12.8% for women, 8.8% for men), N = 15,567.

OECD estimates and potential biases
We rely primarily on the publicly available OECD data age 45+ for the U.S. and 10 comparison high-income countries with available data for 2013-14.We are limited to 2013-14 in the OECD data because there is no more recent data available for U.S. 30-day mortality (case fatality) following the AMI.To calculate rates for the 45+ population in high-income counties, we apply the ratio 0.92 (derived above) to the overall OECD data for the U.S.That is, our estimate for the highest-income 5% of U.S. counties, 8.1, is equal to.92*8.8 per 100 AMI admissions; for the highest-income 1%, it is 8.4 (.95*8.8).
For OECD data, we combined two years of data and implemented confidence intervals reflecting both the reported annual confidence intervals and the two-year variability in rates across years.(We assumed a binary distribution for mortality at the individual level, which allowed us to recover sample sizes from each year's sample; this was then used to create confidence intervals for the combined two-year samples under the assumption that the samples in each year were independently drawn and equal in number).
As noted above, we used the linked hospitalization data to better capture true 30-day mortality, but there are more recent "unlinked" OECD data from 2016 that suggest a somewhat better ranking for the U.S. compared to other countries reporting data in that year.However, as has been pointed out by Drye et al. (2012), the link between in-hospital mortality and the corresponding 30-day mortality rate can be tenuous, most notably because differences in average length of stay differs so much across hospitals (as in Drye et al, 2012), or across countries.For example, even if 30-day mortality rates were identical in the U.S. and Great Britain, it is more likely that an AMI patient will die in hospital in Great Britain because average length of stay for AMI patients was 9.8 days compared to just 4.8 days in the U.S in 2010, the most recent data available (https://stats.oecd.org/Index.aspx?QueryId=51881).
Because there are no linked national databases for the general U.S. population between ages 45-64, the OECD used data from the Healthcare Cost and Project (HCUP) a select group of states that are able to track readmissions from one hospital to another; one may then estimate case-fatality from either mortality occurring during the initial hospital admission, or during a readmission, but not if the death occurs outside of the hospital.
For this reason, U.S. case-fatality estimates are likely to be biased downward compared to other countries because they miss out-of-hospital deaths.For example, when we use the simple adjustment described above for the top 5% of counties to the U.S. average of 8.8%, the implied 30-day mortality rate is 8.1 percent, identical to the corresponding rate in Denmark.Yet as we have seen from the micro-level Danish (and Norwegian) universal data, mortality rates in the age 65+ population are consistently lower than U.S. mortality rates for higher-income counties and zip codes.If instead we calibrate not from the potentially biased U.S. data, we instead calibrate from the (e.g.) Danish data, we would predict an age 45+ mortality rate equal to 9.4% for the top 5% income counties, 9.6% for the top 1% of counties, and 9.0% for the top 5% of zip codes; all of these substantially larger than the median of comparison countries (8.2%).
And while these estimates may themselves be biased upward because we are using fee-forservice Medicare claims data (rather than all enrollees including managed care patients), our alternative approach reinforces the earlier finding that even for privileged Americans, outcomes following AMI are no better than the median of the comparison countries.

eTable 2 .
Comparison of Infant Mortality Rates Comparison of 5-Year Survival Rates for Breast and Colon Cancer and Childhood ALL

100 AMI hospitalizations) Overall 65-69 70-74 75-79 80-84 85+
*The adjustments to US data by income group are made using the Medicare claims data; see text and eAppendix 1 for details.