County-Age High-Risk Ratio Derivation

This page documents how the simulator’s county- and age-specific high-risk ratios are constructed from state-level survey estimates and county-level comorbidity burden. These values are model inputs, not fitted simulator code.

In the simulator, high risk means higher risk of hospitalization or death after infection. It does not mean higher risk of becoming infected.

The generated input files are:

data/<STATE>/state_<STATE>_high-risk-ratios-flu-only.csv
data/<STATE>/county_<STATE>_high-risk-ratios-flu-only.csv

State Age Estimates

The state-level high-risk proportion is estimated separately for each of the five simulator age groups:

Simulator age group	Primary source
0-4	National Survey of Children’s Health
5-17	National Survey of Children’s Health
18-49	Behavioral Risk Factor Surveillance System
50-64	Behavioral Risk Factor Surveillance System
65+	Behavioral Risk Factor Surveillance System

The adult estimates use BRFSS conditions associated with severe influenza outcomes. The pediatric estimates use NSCH survey weights and pediatric conditions associated with severe influenza outcomes. The condition set follows CDC influenza high-risk guidance as closely as the source surveys allow.

Let \(s\) index states and \(a\) index simulator age groups. The state survey stage produces:

\[r_{s,a}^{state},\]

the estimated proportion of people in state \(s\) and age group \(a\) who are in the influenza high-risk hospitalization group.

County Burden Adjustment

State survey data are not detailed enough to directly estimate stable county-age high-risk proportions for every county. The county input therefore starts with the state-age estimate and redistributes relative burden across counties using CDC PLACES county comorbidity measures.

Let \(B_c\) be the county comorbidity burden score after selecting available PLACES measures, standardizing them, and combining them into a county-level relative burden index. The county factor is normalized within each state so that it shifts risk between counties without changing the state-age target:

\[F_c = \frac{B_c}{\sum_{c' \in s} \omega_{c'} B_{c'}},\]

where \(\omega_c\) is the county population share used for the state normalization. A value above 1 means the county has higher comorbidity burden than the state average; a value below 1 means lower burden.

The provisional county-age high-risk proportion is:

\[\tilde r_{c,a} = r_{s,a}^{state} F_c.\]

The final value is bounded to remain a valid proportion:

\[r_{c,a} = \min\left(\max\left(\tilde r_{c,a}, 0\right), 1\right).\]

The resulting county values are hypothesis-driven allocations of state-level risk to counties. They are intended to give the simulator county-specific hospitalization-risk structure when county-age survey estimates are not available.

Interpretation

The county high-risk ratio affects how each county-age population is split between low-risk and high-risk hospitalization groups:

\[N_{c,a}^{high} = N_{c,a} r_{c,a}\]

\[N_{c,a}^{low} = N_{c,a} \left(1-r_{c,a}\right).\]

Those groups then use the simulator’s high-risk and low-risk disease progression probabilities. The high-risk ratio does not modify contact rates, mobility, susceptibility, or the probability of initial exposure unless a separate input derivation explicitly uses it.

Relationship To Initial Exposure Fitting

The county-age initial exposure fitting method uses \(r_{c,a}\) to estimate low-risk county-age population:

\[L_{c,a} = N_{c,a}\left(1-r_{c,a}\right).\]

That makes the high-risk-ratio file an input to the external allocation math used for the Delaware example. The allocation itself is documented in County-Age Initial Exposure Fitting.

Producing The Files

The files are produced by the input-preparation scripts:

Step	Script	Result
4a	`scripts/4a_flu_state_high_risk_by_age.R`	State-and-age influenza high-risk proportions from BRFSS and NSCH.
4b	`scripts/4b_flu_county_high_risk_by_age.R`	County-age high-risk ratios using the state estimates and CDC PLACES burden.

The script provenance and full input pipeline are summarized in Data Preparation And Analysis Scripts.