Association between median household income, state Medicaid expansion status, and COVID-19 outcomes across US counties.

OBJECTIVE: To study the relationship between county-level COVID-19 outcomes (incidence and mortality) and county-level median household income and status of Medicaid expansion of US counties.
METHODS: Retrospective analysis of 3142 US counties was conducted to study the relationship between County-level median-household-income and COVID-19 incidence and mortality per 100,000 people in US counties, January-20th-2021 through December-6th-2021. County median-household-income was log-transformed and stratified by quartiles. Multilevel-mixed-effects-generalized-linear-modeling adjusted for county socio-demographic and comorbidities and tested for Medicaid-expansion-times-income-quartile interaction on COVID-19 outcomes.
RESULTS: There was no significant difference in COVID-19 incidence-rate across counties by income quartiles or by Medicaid expansion status. Conversely, for non-Medicaid-expansion states, counties in the lowest income quartile had a 41% increase in COVID-19 mortality-rate compared to counties in the highest income quartile. Mortality-rate was not related to income in counties from Medicaid-expansion states.
CONCLUSIONS: Median-household-income was not related to COVID-19 incidence-rate but negatively related to COVID-19 mortality-rate in US counties of states without Medicaid-expansion.

Background

Instituted in the 1965, Medicaid has become the largest provider of health insurance in the United States (US) by providing medical access to people with low income and limited resources [1]. The 2010 Affordable Care Act (ACA) sought to decrease the number of uninsured individuals by expanding Medicaid coverage and modifying individual insurance markets [2, 3, 5], but a 2012 Supreme Court decision overturned the requirement that states adopt Medicaid expansion [4, 5]. By January 2020, 14 states had yet to adopt Medicaid expansion [6]. Studies have found that among states who implemented the ACA, the increased access to care has led to early diagnosis of cancers, diabetes, and depression among other health outcomes [7-10], but the relationship between Medicaid expansion and an infectious disease outbreak is unknown.

In 2019, a novel coronavirus (COVID-19), originating from Wuhan City, Hubei Province, China began spreading at an alarming rate [11]. As COVID-19 progressed in the United States (US), the health toll disproportionally impacted African Americans and communities with high prevalence of poor housing conditions [12-14]. In addition, COVID-19 has already been shown to impact individuals with certain pre-existing health conditions at a greater rate [15]. Both federal and state policymakers looked to Medicaid as a central tool in their response to the national emergency [16]. However, whether differences exist in COVID-19 outcomes between communities of Medicaid and non-Medicaid expansion states remains unknown. Moreover, it would be important to quantify differences in outcomes, if any, on the strata that appeared most impacted by COVID-19, the low-income communities, to help the states balance cost versus benefits. This is important because individuals without health insurance coverage are likely to be more vulnerable to the adverse health outcomes related to COVID-19. In the states that did not implement Medicaid expansion, 30 percent of low-income workers were uninsured before COVID-19 [17]. This number was less than half in Medicaid expansion states [17].

We thereby sought to investigate the impact of COVID-19 in the counties nationwide according to their socio-economic status and investigate whether the impact varies by counties of states with Medicaid expansion versus those without Medicaid expansion. We stratified the US counties by its median household income and compared on their COVID-19 incident and mortality rates. We hypothesized that county-level median household income would be inversely related to COVID-19 incidence and mortality rates. Additionally, we hypothesize that a state’s Medicaid expansion status will alter the association between county-level median household income and COVID-19 outcomes.

Methods

Providence Veterans Affairs Medical Center Institutional Review Board, formal waiver of approval due to non-Human subject research, project 1660196.

We conducted a longitudinal, retrospective analysis of data of the US counties and District of Columbia (n = 3142) using 2010–2019 baseline data from the Centers for Disease Control and Prevention and the US Census Bureau and related them to the COVID-19 outcome data from the John Hopkins Coronavirus Resource Center, 2020 [18-21]. Counties from US territories (American Samoa, Guam, Northern Mariana Island, Puerto Rico and US Virgin Islands, n = 105) were not part of the analysis [18]. All data used in this study were publicly available; therefore, the study met the criteria for exemption by the Providence Veterans Affairs Medical Center Institutional Review Board.

Main exposure variables

County-level median household income per annum for each county was collected from the 2018 US Census Bureau’s Small Area Income and Poverty Estimates (SAIPE) [19], and log-transformed (ln X +1) to approximate normality prior to stratification by quartiles. We defined Medicaid expansion states as those that had adopted expansion efforts as of the first case of COVID-19 in the United States on January 20, 2020 (Listing in S1 Table) [6]. Counties in 36 states plus Washington, DC, were included in the Medicaid expansion group, while counties in 14 states were in the non-Medicaid expansion group (S1 Table).

Outcome

The main outcomes of our study were the cumulative COVID-19 incidence rate and mortality rate per 100,000 of the population from January 20th to December 6th, 2020 [21]. The cumulative COVID-19 incidence and mortality rates of the respective US counties were obtained from the John Hopkins Coronavirus Resource Center, divided by the county population and reported as incidence and mortality rates per 100,000, respectively.

Covariates

Data on age and gender were collected using 2010 US Census Bureau data as the elderly and men have been reported as possessing a higher risk of COVID-19 mortality [19]. The COVID-19 pandemic has been shown to afflict minority races in the US to a greater degree; therefore, we included data for racial composition of counties: percentage of White, Black, and Hispanic residents using US Census Bureau data from 2014–2018. Population density (population per square feet of land area) was calculated from the county population from 2010 US census divided by the square foot area of the county to account for overcrowding in a community. In addition to median-household income, we also abstracted data that were confirmatory of the socioeconomic status of the communities such as unemployment rate (2019), percentage of population age >25 years without high school diploma (2014–2018), and percentage of population age <65 years without health insurance (2018) [22]. Access to care was assessed by number of hospitals per county (2017).

Since diabetes mellitus, obesity, and smoking are known risk factors for worse outcomes in COVID-19 [23], we obtained the percentage of the population aged >20 years diagnosed with diabetes mellitus, with obesity, and percentage of adults who are current smokers from the Centers for Disease Control and Prevention from 2016–2018 [20, 23].

Statistical analysis

Baseline characteristics for the counties were described by mean ± standard deviation (SD) and range for continuous variables and percentage for categorical variables. Counties were stratified by quartiles of log-transformed county-level median household income per annum, which comprised of the following median household income ranges: Q1 ($25,385 - $43,681); Q2 ($43,688 - $50,565); Q3 ($50,568 - $58,838); Q4 ($58,848 - $140,382). Linear regression was used to test for trend of baseline characteristics across the income quartiles.

We used a multilevel mixed-effects generalized linear model with a negative binomial distribution and log link function to study the relationship between quartiles of log-median household income and COVID-19 outcomes across US counties: incidence and mortality, in a separate fashion, using Q4 as the referent. We applied a random intercept for states to account for clustering effect due to similarities in health policy for counties within the same state and specifying an unstructured covariance matrix. Using county population as the offset in the model, the outcomes reported were incidence rate ratios (IRR) and mortality rate ratios (MRR) of COVID-19 across income quartiles of the counties, respectively. In a stepwise fashion, we first adjusted for demographics age over 65 years old, gender, and race (Model 1); followed by population density, diabetes, obesity, current smoking status, state Medicaid expansion status and number of hospitals (Model 2). The percentage of population without high school diploma under 25 years old and population without health insurance were not included in the model given their significant correlation with the median household income per county (r = -0.36, P<0.001) and the Medicaid expansion status (r = -0.63, P<0.001) variables, respectively. We tested for interaction between quartiles of log-median household ‘income quartile-times-Medicaid expansion status’ on COVID-19 outcomes in Model 2. If the interaction was significant, the above analyses were repeated, stratified by counties of states with Medicaid Expansion (n = 1,814) and counties of states without Medicaid Expansion (n = 1,328).

Sensitivity analyses were performed to replace median age in lieu of % over 65 years old. All analyses were performed using STATA/SE version 11.2 software (StataCorp LP, College Station, TX). A 2-sided p-value of < 0.05 was considered significant. Replication data stored in Harvard Dataverse© for public access and replication.

Results

As of December 6, 2020, there were a total of 14,528,356 COVID-19 cases, and 279,115 COVID-19 deaths, across the 3,142 US counties. The mean (SD) for COVID-19 incidence and mortality were 5,155.01 (4308.24) cases and 88.38 (96.46) deaths per 100,000 population per county, respectively.

The characteristics of the 3,142 counties, overall and stratified by four quartiles of log-transformed median household income were presented in Table 1. Overall, 57.7% of counties were located in Medicaid expansion states. Higher median household income quartiles were associated with higher mean county population and population density, number of hospitals and percentage of white residents. Conversely, lower median household income quartiles were associated with higher percentage of elderly residents (65 years or older), of black or Hispanic population, of unemployed, of population without a high school diploma and of people without health insurance. Counties of lower income quartiles were also associated with a higher prevalence of diabetes, obesity and smoking and had a lower likelihood of belonging to a state that adopted Medicaid-expansion.

Table 1

Counties baseline characteristics by log transformed county-level median household income.

Variables (2010–2019, Census, CDC)	Overall Mean ± SD (Range) n = 3,142	Log Transformed County-Level Median Household Income Quartiles
		Quartile 1 Mean ± SD n = 786	Quartile 2 Mean ± SD n = 784	Quartile 3 Mean ± SD n = 785	Quartile 4 Mean ± SD n = 787	Linear trend P-value
County-Level Median Household Income, $ (2018)	52,794.41 ± 13,880.12 (25,385–140,382)	38,514.62 ± 3,857.38	47,217.37 ± 1,953.39	54,293.69 ± 2,368.64	71,139.70 ± 13,110.30	<0.0001
Population (2010)	98,174.98 ± 312,433.81	28,645.91 ± 66,314.95	54,073.68 ± 127,899.69	83,421.66 ± 196,745.95	226,261.71 ± 554,409.79	<0.0001
	(82–9,818,605)
Population density per square foot (2010)	216.10 ± 1,231.37	107.48 ± 909.25	120.44 ± 518.46	125.91 ± 305.19	509.83 ± 2181.17	<0.0001
	(0.04–47,505.94)
Median Age (2010), years	40.34 ± 5.06	40.41 ± 5.12	41.17 ± 5.28	40.65 ± 5.17	39.11 ± 4.38	<0.0001
	(21.90–62.70)
Population >65 years (2010), %	15.88 ± 4.19	16.51 ± 3.99	17.07 ± 4.10	16.42 ± 3.98	13.54 ± 3.77	<0.0001
	(3.47–43.38)
Male (2010), %	49.98 ± 2.22	49.98 ± 2.95	50.11 ± 2.31	49.90 ± 1.58	49.91 ± 1.79	0.2134
	(43.20–72.10)
White (2014–2018), %	76.45 ± 20.18	67.03 ± 24.87	79.91 ± 17.72	81.26 ± 16.29	77.60 ± 17.55	<0.0001
	(0.7–100)
Black (2014–2018), %	8.87 ± 14.46	17.86 ± 21.53	7.04 ± 11.23	4.99 ± 8.37	5.60 ± 8.23	<0.0001
	(0–87.4)
Hispanic Latino (2014–2018), %	9.21 ± 13.79	9.54 ± 18.09	8.29 ± 12.81	9.06 ± 12.14	9.95 ± 11.01	0.0979
	(0–99)
Unemployment rate (2019), %	4.00 ± 1.48	4.93 ± 1.71	4.10 ± 1.40	3.69 ± 1.16	3.28 ± 1.05	<0.0001
	(0.7–19.3)
Age >25 years without high school diploma, (2014–2018), %	13.41 ± 6.34	19.22 ± 6.03	13.60 ± 5.08	11.50 ± 4.95	9.31 ± 4.45	<0.0001
	(1.2–66.3)
Age <65 years without insurance (2018), %	11.50 ± 5.04	14.00 ± 4.89	12.37 ± 4.86	10.53 ± 4.67	9.11 ± 4.31	<0.0001
	(2.4–32.2)
Number of Hospitals per county (2017)	1.46 ± 2.56	0.86 ± 0.79	1.23 ± 1.38	1.48 ± 1.83	2.26 ± 4.38	<0.0001
	(0–79)
Age‐adjusted population with diabetes mellitus, age >20 years (2016), %	10.38 ± 3.80	12.71 ± 4.34	10.84 ± 3.64	9.50 ± 2.91	8.45 ± 2.67	<0.0001
	(1.5–33)
Age‐adjusted population with obesity, age >20 years (2016), %	32.76 ± 5.70	35.15 ± 5.79	33.77 ± 5.20	32.38 ± 4.78	29.74 ± 5.55	<0.0001
	(12.3–57.9)
Population with reported smoking (2017), %	17.47 ± 3.63	20.58 ± 3.82	17.93 ± 2.79	16.46 ± 2.51	14.89 ± 2.53	<0.0001
	(6–41)
Number of Counties in states with Medicaid expansion	1,814	321	417	514	562	N/A

The mean number of COVID-19 cases and deaths per 100,000 population across counties from different income quartiles were described in Table 2. Specifically, the cases per 100,000 population attributed to COVID-19 were 5,121.08 ± 2,471.59 for counties in the lowest income quartile as compared to 5,033.77 ± 5,705.18 for counties in the highest income quartile. There was no significant association between COVID-19 incidence and quartiles of household income in unadjusted and adjusted analyses. The only exception was in the fully adjusted model, where counties from income quartile 2 (5,299.32 ± 5338.71 COVID-19 cases per 100,000) had a 10% increase in the risk of COVID-19 incidence compared to counties in the income quartile 4 (IRR 1.10, 95% CI: 1.04–1.17). The interaction between income quartile and Medicaid expansion status’ was not significant for COVID-19 incidence (P values 0.07 to 0.20 Q1-3).

Table 2

Association of SARS-COV-2 outcomes as of December 6, 2020 with county-level median household income quartiles.

	County-Level Median Household Income Quartiles
SARS-CoV2 Outcomes	Quartile 1	Quartile 2	Quartile 3	Quartile 4
(As of December 6, 2020)	N = 786	N = 784	N = 785	N = 787
	IRR / MRR	IRR / MRR	IRR / MRR	IRR / MRR
	(95% CI)	(95% CI)	(95% CI)	(95% CI)
Cases per 100,000 population (mean ± SD)	5,121.08 ± 2471.59	5,299.32 ± 5338.71	5,166.40 ± 2666.65	5,033.77 ± 5705.18
Model 1	0.96	1.05	0.96	REFERENT
	[0.90–1.02]	[0.99–1.10]	[0.91–1.01]
Model 2*	1.04	1.10	1.00	REFERENT
	[0.97–1.12]	[1.04–1.17]	[0.95–1.05]
Deaths per 100,000 population# (mean ± SD)	113.32 ± 87.43	92.21 ± 109.58	75.69 ± 62.84	72.32 ± 112.19
Model 1	1.16	1.15	0.99	REFERENT
	[1.06–1.26]	[1.07–1.24]	[0.93–1.06]
Model 2*	1.22	1.18	1.02	REFERENT
	[1.09–1.35]	[1.08–1.28]	[0.95–1.10]

IRR = Incident Rate Ratio; MRR = Mortality Rate Ratio; 95% CI = 95% Confidence Interval

Model 1: % Population > 65 years, % Male, and % White

Model 2: % Population > 65 years, % Male, % White, Population Density, % Obesity, % Smoking, % Diabetes, Number of Hospitals, Medicaid expansion status according to state policy

*Interaction between income quartiles and Medicaid status was significant (p-value ≤ 0.005) for SARS-COV-2 mortality but not for SARS-CoV2 Cases (p-value ≥ 0.073)

#Differences between means were statistically significant (p-value < 0.0000) for SARS-COV-2 Deaths per 100,000 population but not for SARS-COV-2 Cases per 100,000 population (p-value < 0.6693)

Conversely, there was a significant association between COVID-19 mortality and quartiles of household income. Specifically, the deaths per 100,000 population attributed to COVID-19 were 113.32 ± 87.43, for counties in the lowest income quartile as compared to 72.32 ± 112.19, for counties in the highest income quartile. In the fully adjusted model, counties from income quartile 1 had a 22% increase in the risk of COVID-19 mortality compared to quartile 4 (MRR 1.22, 95% CI 1.09–1.35). Furthermore, the interaction ‘income quartile*Medicaid expansion status’ was significant (P values <0.01, Q1-3), for which subgroup analyses by Medicaid expansion status were conducted. The sensitivity analyses replacing % population over 65 years old with median age of the county population did not significantly change the results.

The comparison of baseline characteristics between counties in Medicaid and non-Medicaid expansion states were described in Table 3. Counties from states with Medicaid expansion had a higher population density, percentage of white residents, median household income, unemployment rate, number of hospitals; and a lower percentage of population who were Black, Hispanic, without high school diploma, without health insurance, with diabetes, with obesity or reported being a current smoker.

Table 3

Counties baseline characteristics by Medicaid expansion status of the state.

Variables (2010–2020, Census, CDC, Johns Hopkins Coronavirus Resource Center)	Counties within States with Medicaid Expansion n = 1,814 Mean ± SD	Counties within States without Medicaid Expansion n = 1,328 Mean ± SD	P-value
Population (2010)	114,149.71 ± 367932.35	76,354.07 ± 212778.78	0.0008
Population density per square foot (2010)	289.89 ± 1599.52	115.30 ± 275.52	0.0001
Median Age (2010), years	40.75 ± 5.10	39.77 ± 4.94	<0.0001
Population > 65 years (2010), %	15.94 ± 4.14	15.79 ± 4.25	0.3138
Male (2010), %	50.03 ± 2.07	49.90 ± 2.41	0.1078
White (2014–2018), %	80.92 ± 18.13	70.33 ± 21.23	<0.0001
Black (2014–2018), %	5.65 ± 10.59	13.28 ± 17.56	<0.0001
Hispanic Latino (2014–2018), %	7.62 ± 11.17	11.38 ± 16.48	<0.0001
Overall, County-level Median Household Income (2018), $	55,512.16 ± 14866.59	49,084.12 ± 11411.09	<0.0001
Quartile 1, County-level Median Household Income (2018), $	38,462.38 ± 3943.10	38,550.68 ± 3800.93	0.7526
Quartile 2, County-level Median Household Income (2018), $	47,417.04 ± 1941.98	46,990.50 ± 1944.14	0.0022
Quartile 3, County-level Median Household Income (2018), $	54,335.76 ± 2317.18	54,213.89 ± 2465.68	0.4935
Quartile 4, County-level Median Household Income (2018), $	72,362.99 ± 14049.30	68,089.64 ± 9787.66	<0.0001
Unemployment rate (2019), %	4.13 ± 1.63	3.81 ± 1.24	<0.0001
Population without high school diploma, aged >25 years (2014–2018), %	11.80 ± 5.54	15.61 ± 6.69	<0.0001
Population without health insurance, aged <65 years (2018), %	8.77 ± 3.26	15.23 ± 4.64	<0.0001
Number of Hospitals per county (2017)	1.58 ± 2.96	1.28 ± 1.86	0.0010
Age‐adjusted population with diabetes mellitus, aged >20 years (2016), %	9.70 ± 3.38	11.30 ± 4.14	<0.0001
Age‐adjusted population with obesity, aged >20 years (2016), %	32.12 ± 5.70	33.63 ± 5.63	<0.0001
Population with reported smoking (2017), %	17.09 ± 3.63	17.98 ± 3.39	<0.0001

The association between household income quartiles and COVID-19 mortality by state Medicaid expansion status was depicted in Table 4. In Medicaid-expansion states, the deaths per 100,000 population attributed to COVID-19 were 92.31 ± 128.60, for counties in the lowest income quartile as compared to 70.20 ± 138.43, for counties in the highest income quartile. On the other hand, for non-Medicaid-expansion states, the COVID-19 deaths per 100,000 population were 138.78 ± 89.11, for counties in the lowest income quartile as compared to 73.36 ± 53.55, for counties in the highest income quartile. In fully adjusted analyses, median household income quartiles were associated with COVID-19 mortality only in counties within non-Medicaid-expansion states, such that counties in the lowest income quartile had a 41% increase in COVID-19 mortality compared to counties in the highest income quartile (MRR 1.41, 95% CI: 1.25–1.59). Contrarily, there were no significant differences in COVID-19 mortality risk by income quartiles in counties within Medicaid expansion states (Fig 1).

Table 4

Subgroup analysis of Medicaid and non-Medicaid SARS-COV-2 mortality rate as of December 6, 2020 with county-level median household income quartiles.

	Log Transformed County-Level Median Household Income Quartiles
	Quartile 1 Medicaid N = 454 MRR [95% CI]	Quartile 2 Medicaid N = 453 MRR [95% CI]	Quartile 3 Medicaid N = 452 MRR [95% CI]	Quartile 4 Medicaid N = 455 MRR [95% CI]	Quartile 1 Non-Medicaid N = 332 MRR [95% CI]	Quartile 2 Non-Medicaid N = 332 MRR [95% CI]	Quartile 3 Non-Medicaid N = 332 MRR [95% CI]	Quartile 4 Non-Medicaid N = 332 MRR [95% CI]
SARS-COV-2 mortality rate
(As of December 6, 2020)
Deaths per 100,000 (mean ± SD)	92.31 ± 128.60	71.83 ± 73.10	77.18 ± 67.47	70.20 ± 138.43	138.78 ± 89.11	104.13 ± 68.60	94.69 ± 79.22	73.36 ± 53.55
Model 1	1.01	1.09	0.94	REFERENT	1.43	1.31	1.15	REFERENT
	[0.88–1.15]	[0.97–1.21]	[0.85–1.03]		[1.30–1.57]	[1.19–1.43]	[1.05–1.26]
Model 2	1.06	1.12	0.97	REFERENT	1.41	1.28	1.14	REFERENT
	[0.90–1.26]	[0.98–1.27]	[0.87–1.07]		[1.25–1.59]	[1.16–1.42]	[1.03–1.25]

MRR (95% CI) = Mortality Rate Ratio (95% Confidence Interval)

Model 1: % Population > 65 years, % Male, and % White

Model 2: % Population > 65 years, % Male, % White, Population Density, % Obesity, % Smoking, % Diabetes, Number of Hospitals, Medicaid expansion status according to state policy

*Interaction between income quartiles and Medicaid status was significant (p-value ≤ 0.005) for SARS-COV-2 mortality but not for SARS-CoV2 Cases

Fig 1

Association of SARS-COV-2 mortality rate ratios (95% confidence intervals) by median household income quartiles, for counties in Medicaid (gray diamond) and non-Medicaid expansion states (orange diamond), referent = Quartile 4.

In fully adjusted analyses, median household income quartiles were associated with COVID-19 mortality only in counties within non-Medicaid-expansion states (orange diamond), but there were no significant differences in COVID-19 mortality risk by income quartiles in counties within Medicaid expansion states (gray diamond).

Discussion

To our knowledge, this is one of the first investigations of the association between median household income with COVID-19 outcomes at the county level, in Medicaid expansion and non-expansion states. We found no significant difference in COVID-19 incidence across counties by income quartiles and when sub-stratified by Medicaid-expansion status. However, we found a significant difference in COVID-19 mortality by county median household income, such that COVID-19 mortality was significant higher in counties from the lower compared to the highest income quartiles, but only in states that did not adopt Medicaid-expansion, and not significantly different in counties from Medicaid-expansion states.

There is ample evidence to support that socioeconomic status is related to health outcomes. Our group has shown that the percentage of population living in poverty in communities was associated with a higher cardiovascular and heart failure mortality [24]. We also showed that counties with higher percentage of households living in poor housing conditions had significantly higher risk of COVID-19 incidence and mortality [14]. In this study, we showed that COVID-19 infection affected communities of distinct income strata in a similar fashion, but with a higher mortality risk in communities of lower household income. Multiple mechanisms have been posited to explain poor health outcomes in low-income population. It is possible that people in lower-income communities have worse health at baseline, receive care at lower quality hospitals, receive differential care within a hospital due to lack of health insurance or poor health literacy, and/or there is a lack of access to care outside of the hospital due to lack of health insurance [25, 26]. In this study, the mechanisms for a higher COVID-19 mortality associated with lower-income quartiles compared to the highest are likely multi-factorial. At the county level, we found a higher prevalence of obesity, diabetes and smoking in lower income communities to support a lower baseline health status of the lower-income communities. We also found a higher prevalence of non-graduation from high school as well as a lack of health insurance in counties within the lowest income-quartiles, which can potentially lead to a lower health literacy and health care access, respectively. All of the above could affect the population behavior including the timeliness towards seeking healthcare when they become ill with COVID-19 as well as post-hospitalization care after discharge.

At the state level, various mechanisms can explain the findings of disparate COVID-19 mortality risk in lower income compared to high income communities, a finding that is significant only in states that did not adopt Medicaid expansion but not significant in Medicaid-expansion states. It is possible that Medicaid expansion is only a marker of the state-level policy towards COVID-19, in terms of mask and social distancing mandates as well as health education and promotion practices for the population, all of which could influence population behavior stated above. In addition, the observed mortality outcome differences across income quartiles in states without Medicaid expansion can also be related to a lower health care access due to lack of insurance after contracting COVID-19, since there was no significant difference in COVID-19 incidence across income quartiles. This is supported by the much higher prevalence of population without health insurance at 14% in the lowest income quartile, compared to 9% in counties from the highest income-quartile (Table 1), a percentage that is twice larger in non-Medicaid expansion states (Table 3).

Over the past decade, studies have shown that in the states that expanded Medicaid coverage, there were improvements in diagnosis, management and mortality of chronic conditions [7–10, 27–29]. Further studies have also investigated the impact on disease mortality rates in Medicaid expansion states on a nationwide scale [30, 31]. In end-stage renal disease, patients had improved 1-year survival rates in Medicaid expansion states [31]. Similarly, a decrease in cardiovascular mortality was observed in states after Medicaid expansion. This was considered to be a benefit of improved access to healthcare for low income individuals by raising the Medicaid eligibility threshold to 138% of the federal poverty level [30, 32]. We believe similar mechanisms may in part explain the differences in COVID-19 mortality at communities of different income strata, especially in non-Medicaid-expansion states. A review of literature shows that individuals without health insurance are less likely to seek health care even when in need [33]. In contrast, it has been shown that when they could afford care, individuals were more likely to utilize healthcare resources [23, 34–36]. Therefore, while a proportion of population in high income communities are able to afford insurance regardless of state Medicaid expansion status, exemplified by similar mortality rates between counties in the highest-income quartiles between Medicaid vs. non-Medicaid expansion states, the highest mortality rate gap is observed in the lowest-income quartiles. Thus, the lack of access to health care is another potential mechanism for COVID-19 mortality disparity in low-income communities from non-Medicaid expansion states.

Limitations and strengths

The strength of this study is that it is a nationwide study, that utilized cumulative and representative data of US communities in 2020 suitable to assess outcomes as it relates to socio-economic status. Study limitations include its observational design, inability to conclude causality and the potential for residual confounding despite our careful control of known confounders. For example, the use of crude mortality and COVID-19 incidence rates instead of age-adjusted rates, to account for diversity in age distribution in a county, is a limitation and may introduce confounding. As such, we adjusted for age, gender, race and comorbidities of the county population in the final model to minimize the residual confounding. We are aware that policies regarding social distancing and mask mandate may influence the outcomes, it is difficult to incorporate these into the analyses given the ever-changing nature of these policies through-out the year and the disparate execution of these mandates at the regional level. Instead, we used the cumulative outcome approach to study the Medicaid-expansion policy that was unaltered during 2020. Although some states adopted Medicaid expansion into their state constitution during 2020 (Missouri and Oklahoma), none of them achieved implementation stage during 2020.

Conclusions and implications

Median-household-income was not related to COVID-19 incidence but negatively related to COVID-19 mortality in US counties of states without Medicaid-expansion. It was unrelated to mortality in counties of states that adopted Medicaid-expansion. Future studies are needed to untangle which state policies have the most impact in the attenuation of the excessive COVID-19 mortality risk associated with socioeconomically disadvantaged communities.

List of Medicaid expansion and non-Medicaid expansion states as of January 1, 2020.

(DOCX)

Click here for additional data file.

21 Dec 2021

Submitted filename: plos one review comments-plos one.docx

Click here for additional data file.

11 Jan 2022

PONE-D-21-39999

Association between Median Household Income, State Medicaid Expansion Status, and COVID-19 Outcomes Across US Counties

PLOS ONE

Dear Dr. Wu,

Thank you for submitting your manuscript for review by PLOS ONE. After careful consideration, we have decided that your manuscript does not meet our publication criteria and must therefore be rejected.

As noted, this manuscript is a resubmission of PLOS ONE submission PONE-D-21-03802, which was previously rejected due to concern about the contents of the manuscript. Unfortunately, we do not feel that these concerns have been sufficiently addressed.

In light of the remaining concerns, the manuscript does not meet our publication criteria requiring that conclusions are presented in an appropriate fashion and are supported by the data.

I am sorry that we cannot be more positive on this occasion, but I do hope that you will find these comments useful when deciding how to proceed with your manuscript.

With best wishes,

James Mockridge, PhD

Division Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

- - - - -

For journal use only: PONEDEC3

24 Feb 2022

The response to the reviewers was included as a word file at the end of the uploaded documents

Submitted filename: plos one review comments-plos one.docx

Click here for additional data file.

15 Jun 2022

29 Jun 2022

Author’s Reply to Reviewer’s Comments

Journal Requirements:

1/ Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

R/ We apologize for the not adhering by PLOS ONE's style requirements. We have altered the format of the manuscript to reflect the updated PLOS ONE's style requirements, including those for file naming.

2/ Thank you for stating the following financial disclosure: Yes, this project is supported by the Department of Veterans Affairs, Health Services Research Award IRP-20-003 (WW, project PI). Please state what role the funders took in the study. If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript." If this statement is not correct you must amend it as needed. Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

R/ We apologize for the omission regarding the role of the funders in the study. We have adjusted the financial disclosure statement to include a statement clarifying the role of the funders in the study.

3/ Thank you for stating the following in the Acknowledgments/ Funding Section of your manuscript: "This project is supported by the Department of Veterans Affairs, Health Services Research Award IRP-20-003 (WW, Project PI)". Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form. Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows: "Yes, this project is supported by the Department of Veterans Affairs, Health Services Research Award IRP-20-003 (WW, project PI)" Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

R/ Thank you for bringing this to our attention. We have removed all information relating to the funding source in the Acknowledgments section and other areas of the manuscript. Additionally, we have updated the cover letter to include information regarding funding. We would like the online submission form to state the following: “This project is supported by the Department of Veterans Affairs, Health Services Research Award IRP-20-003 (WW, Project PI). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

4/ We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere. Please clarify whether this publication was peer-reviewed and formally published. If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

R/ While awaiting information a decision from PLOS ONE, the authors of the manuscript added portions of the manuscript to the preprint server ‘medRxiv’, an option endorsed by PLoS ONE. The preprints on medRxiv are initial reports, and as such, it is not peer-reviewed or considered formally published.

5/ Please amend your list of authors on the manuscript to ensure that each author is linked to an affiliation. Authors’ affiliations should reflect the institution where the work was done (if authors moved subsequently, you can also list the new affiliation stating “current affiliation:….” as necessary).

R/ We apologize for the oversight about the affiliations of the authors of the manuscript. We have amended the list of authors to reflect the authors affiliations during the time when the work was done.

6/ Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information

R/ We have amended the in-text citations to match the PLOS ONE's style requirements. Also, the Supporting Information files are at the end of the manuscript.

8/ Thank you for including your ethics statement on the online submission form: "Providence Veterans Affairs Medical Center Institutional Review Board, formal waiver of approval due to non-Human subject research, project 1660196." To help ensure that the wording of your manuscript is suitable for publication, would you please also add this statement at the beginning of the Methods section of your manuscript file.

R/ We have revised the beginning of the Methods section of the manuscript to include the ethics statement from the Providence Veterans Affairs Medical Center Institutional Review Board.

9. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

R/ We have reviewed the reference list to ensure that it is complete and correct. During our review, we did not come across any cited papers that have since been retracted.

Author’s Responses to Reviewer's Questions:

Reviewer #1: I am satisfied by the revisions. In particular, the limitations to causal inference are now clearly acknowledged.

R/ Thank you for taking the time to provide insight into our work.

Reviewer #2: “Dear PONE Journal Team of editorials, I am thankful for the chance given to me to review the manuscript titled “Association between Median Household Income, State Medicaid Expansion Status, and COVID-19 Outcomes Across US Counties”. The article has benefit for measuring the association between income, medical expansion across different countries in US even if the spelling for countries is ‘counties’. Here by the way what does across the US countries mean? Is US one or more country with fifty plus states? Anyways, the following are my comments;”

R/ We appreciate your review of our work. After thorough review of the manuscript, we were unable to identify any occurrences of the use of ‘countries’. The analysis included 3142 United States counties (a political and administrative division within the 50 states) including the District of Columbia.

• “Is incidence and mortality the only outcome? Why didn’t you focus on it mainly? Why not risk of being infected by COVID 19? Then the continuum of illness through the diseases process? Why not recovery? What are the complications? Try to address multiple whys?”

R/ We appreciate your review of our work. In order to avoid over-extending the scope of paper and distract the reader, we have chosen to focus only on two main COVID outcomes, the COVID incidence and COVID mortality. As such, the incident rate ratio reported in our results would be considered an epidemiologically acceptable estimate of the relative risk of being infected with COVID and the mortality risk ratio would be an estimate of the worst complication of COVID, i.e. death. We did not address the recovery portion of COVID since it would be outside the scope of the current paper.

• “Is the median income an established/factful US income before or during the era of the pandemic? Why median? Why not high income? Why not low income?”

R/ Given that the unit of analysis being the county, the median income of the households within the county is commonly used as an estimate of the socioeconomic status for the population of the county. Since median income would vary by the socioeconomic status of the population of the county, it would allow us to compare the COVID incidence and mortality between “high income” counties versus “low income” counties. Together they answered our main study questions of “the impact of COVID-19 in the counties nationwide according to their socio-economic status and investigate whether the impact varies by counties of states with Medicaid expansion versus those without Medicaid expansion.”

• “Did US have different Medicaid expansion for the high, medium and low wealth index?”

R/ Excellent question. Medicaid provides a source of health insurance for low-income households and/or those with disability, among other factors. However, the income and qualification thresholds for enrollment into Medicaid differ by state, which depend on the state government’s policy. If a state has adopted the policy of Medicaid expansion, the requirements for Medicaid eligibility are less and at lower income thresholds, which makes it easier for people to enroll in health insurance versus states who did not adopt Medicaid expansion.

• “How do you see the association between the income, Medicaid and COVID outcomes? If so let us take it as a sort of continuum of care, have you calculated the dropout rate per each level?”

R/ Based on the above, it is possible that COVID incidence or mortality may disproportionately affect counties of lower socioeconomic status from states that did not adopt Medicaid expansion because less people in that state would have access to health insurance. Less access to health insurance can lead to worse COVID complications, such as death, because the access to and the continuum of, care, are impaired. Therefore, we first compared the COVID incidence and mortality amongst counties of different socioeconomic status (based on median household income) to study this association. Mortality rates were calculated for each quartile of the median household income. We then looked at the impact of Medicaid expansion on this association by comparing the COVID mortality amongst counties of different socioeconomic status within states that adopted Medicaid expansion and those that did not.

• “Use scientific methods of writing as per the standard. There are inconsistencies. Language, tense, sentences, ideas, paragraphs and the general manuscript needs to be revised?”

R/ We agree with the reviewer and have reviewed the manuscript to ensure the use of scientific methods of writing.

• “Check the whole statistics again”

R/ We appreciate your review of our work. We performed a thorough review of the statistics, we did not find any items in need of correction at this point.

We appreciate and thank the reviewer for his/her comments, to which we provided a point-to-point answer in the reply to reviewer’s document, all of which, made our paper a much improved one.

Tsikata Apenyo & Wen-Chih Wu

Submitted filename: Response to reviewers.docx

Click here for additional data file.

15 Jul 2022

15 Jul 2022

Author’s Reply to Reviewer’s Comments

We appreciate and thank the reviewers for their comments, to which we provided a point-to-point answer below:

Reviewer #3:

The revised manuscript is generally well-presented and presents an interesting topic. Many of my comments are minor. My biggest concern is that the authors proceeded with the outcome of crude mortality and crude incidence rate. I know that US counties are very diverse in terms of age distribution and as such, the age-adjusted mortality rates and age-adjusted incidence rates would have been the better metrics. Please provide enough justification for not age-adjusting rates. Having said that, I have seen plenty of studies examining the crude rate, primarily because of the lack of data. If that is the case, please provide this information in the limitation section.

R/ We agree with the reviewer that age-adjusted mortality rates and age-adjusted incidence rates would have been the better metrics. However, the data available does not have age-adjusted rates for which we have added this as a limitation in the discussion section (pg. 13, second to last paragraph). We also stated that we adjusted for age, gender, race and comorbidities of the county population to minimize the residual confounding.

Key findings /Questions: household income should be replaced with county-level median HH income.

R/ “household income” replaced with “county-level median household income” in Key findings /Questions

Abstract:

Objective: What you have as the objective in the abstract is not really an objective. That’s background information.

R/ We replace the sentence with “To study the relationship between county-level COVID-19 outcomes (incidence and mortality) and county-level median household income and status of Medicaid expansion of US counties”

Methods: Use the easy-to-understand word format for the date given the global readership of the journal. Avoid star; either use the proper multiplication sign or have it in word.

R/ We changed the date format to “January-20th-2021 through December-6th-2021”. We deleted the star and changed it to “Medicaid-expansion-times-income-quartile interaction”

Results: insert “rate” after COVID-19 incidence and COVID-19 mortality

R/ “rate” was inserted after COVID-19 incidence and COVID-19 mortality through out to be read as “incidence-rate” and “mortality-rate”.

Background: On your hypothesis, household income should be replaced with county-level median HH income.

R/ We replaced “household income” with “county-level median household income” in the hypothesis

Outcome: See my general comment on the top

R/ “rate” was inserted after COVID-19 incidence and COVID-19 mortality through out to be read as “incidence rate” and “mortality rate” or “incidence and mortality rates”.

Statistical Analysis:

For the sake of completeness, also mention that these income ranges are per annum.

“Using county population as the denominator in the model”: well there is no such thing as “denominator” in the model. Either you specify the rate (covid outcomes divided by denominator population) as the outcome, or most likely given the NB model, you specified the population as an offset. Make this clearer.

R/ “Median household income” was changed to “County-level median household income per annum” under Main Exposure Variables. We also made the same change to “county-level median household income per annum” under the statistical analysis. We also changed the word “denominator” to “offset”.

I see the value of having random intercepts for counties but not for the reason the authors suggest (i.e. to account for clustering, especially in conjunction with an unstructured covariance matrix). I would have used the robust standard errors clustered at the state level. But I am not holding this relatively small issue against the authors for their enormous effort.

R/ We thank the reviewer for their understanding.

“If the interaction proved to be significant”: note that you are not proving significance. I’d

rephrase this sentence.

R/ the sentence was rephrased to “If the interaction was significant…”

Table 1 title” Again, avoid giving an impression as if the income is household level. Even if the software gives you P=0, this is theoretically not possible. Write something like P<0.0001.

R/ We have changed to “County-Level Median Household Income” for titles and sub-titles of the tables. We changed the P values=0 to P<0.0001.

Thank you sincerely for the insightful comments which made our paper a much improved one.

Sincerely,

Tsikata Apenyo & Wen-Chih Wu

Submitted filename: Response to reviewers-7-15-22.docx

Click here for additional data file.

21 Jul 2022

Association between Median Household Income, State Medicaid Expansion Status, and COVID-19 Outcomes Across US Counties

PONE-D-21-39999R3

Dear Dr. Wu,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Robert Jeenchen Chen, MD, MPH

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: (No Response)

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: (No Response)

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: (No Response)

Reviewer #3: (No Response)

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: (No Response)

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #3: Authors addressed all concerns I had. Thank you, and good luck with the next step of the publication.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #3: No

**********

2 Aug 2022

PONE-D-21-39999R3

Association between Median Household Income, State Medicaid Expansion Status, and COVID-19 Outcomes Across US Counties

Dear Dr. Wu:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Robert Jeenchen Chen

Academic Editor

PLOS ONE