Association of poor housing conditions with COVID-19 incidence and mortality across US counties.

OBJECTIVE: Poor housing conditions have been linked with worse health outcomes and infectious disease spread. Since the relationship of poor housing conditions with incidence and mortality of COVID-19 is unknown, we investigated the association between poor housing condition and COVID-19 incidence and mortality in US counties.
METHODS: We conducted cross-sectional analysis of county-level data from the US Centers for Disease Control, US Census Bureau and John Hopkins Coronavirus Resource Center for 3135 US counties. The exposure of interest was percentage of households with poor housing conditions (one or greater of: overcrowding, high housing cost, incomplete kitchen facilities, or incomplete plumbing facilities). Outcomes were incidence rate ratios (IRR) and mortality rate ratios (MRR) of COVID-19 across US counties through 4/21/2020. Multilevel generalized linear modeling (with total population of each county as a denominator) was utilized to estimate relative risk of incidence and mortality related to poor housing conditions with adjustment for population density and county characteristics including demographics, income, education, prevalence of medical comorbidities, access to healthcare insurance and emergency rooms, and state-level COVID-19 test density. We report incidence rate ratios (IRRs) and mortality ratios (MRRs) for a 5% increase in prevalence in households with poor housing conditions.
RESULTS: Across 3135 US counties, the mean percentage of households with poor housing conditions was 14.2% (range 2.7% to 60.2%). On April 21st, the mean (SD) number of cases and deaths of COVID-19 were 255.68 (2877.03) cases and 13.90 (272.22) deaths per county, respectively. In the adjusted models standardized by county population, with each 5% increase in percent households with poor housing conditions, there was a 50% higher risk of COVID-19 incidence (IRR 1.50, 95% CI: 1.38-1.62) and a 42% higher risk of COVID-19 mortality (MRR 1.42, 95% CI: 1.25-1.61). Results remained similar using earlier timepoints (3/31/2020 and 4/10/2020). CONCLUSIONS AND RELEVANCE: Counties with a higher percentage of households with poor housing had higher incidence of, and mortality associated with, COVID-19. These findings suggest targeted health policies to support individuals living in poor housing conditions should be considered in further efforts to mitigate adverse outcomes associated with COVID-19.

Introduction

Coronavirus disease-2019 (COVID-19) is a rapidly evolving pandemic caused by the novel enveloped RNA beta-coronavirus, Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) [1]. As of April 26, 2020, SARS-COV-2 has infected over 2 million people worldwide and led to >200k deaths [2]. The infectivity potential of SARS-CoV-2 is a function of its high basic reproductive number (R0) which has been estimated to be from 2.2 to as high as 5.7, as compared to R0 of 1.3 for influenza virus [3-5]. Given the high infectivity and lack of a vaccine so far, one of the important mitigation strategies employed worldwide in order to decrease the rate of spread and ‘flatten’ the infection curve has been social and physical distancing [3,6]. However, physical distancing can be compromised if the living environment for various reasons, such as poor housing conditions, prevents the individuals from staying isolated. Problems associated with poor housing, reported as severe housing problems by CDC, include overcrowding, high housing cost, lack of kitchen facilities or lack of plumbing facilities [7,8].

Overcrowding has been associated with spread of respiratory illnesses like tuberculosis and influenza which have aerosol and droplet transmissions, both of which are potential modes of transmission for COVID-19 [9-13]. The World Health Organization (WHO) published housing and health guidelines in 2018, identifying poor housing related environmental risk factors including overcrowding, air and water quality and lack of access to adequate plumbing and sanitation, as factors contributing to the burden of infectious diseases including airborne respiratory illnesses [14]. Another feature of poor housing conditions is high cost burden, defined as more than 50% of the household income spent towards housing cost [8]. Households facing high housing cost in proportion to the household income have to compromise in other aspects of living, such as education, health insurance and food, among others [15]. Prior studies have shown an association between high housing costs and delays in seeking healthcare [16,17]. There are several news reports indicating that minimum wage workers have found it difficult to follow the stay-at-home guidelines, due to a lack of economic resources (e.g. public transportation, choice of working environment) and being part of essential work force that is mostly an exception to stay-at-home guidelines [18-20]. Another aspect of poor housing conditions is lack of plumbing or kitchen facilities and the inherent need to utilize communal facilities (bathrooms or kitchens). Lack of appropriate facilities interferes with the ability to practice good hygiene. Based on the above, people with poor housing conditions have a limited ability to practice effective physical distancing and good personal hygiene; hence, potentially at higher risk of worse outcomes related to a respiratory infectious illness such as COVID-19.

Since the relationship between poor housing conditions and COVID-19 outcomes are currently unknown, we undertook a study to test the hypothesis that the percentage of households with poor housing conditions in US counties are associated with higher incidence and mortality of COVID-19 across 3135 counties in the United States (US). Robust literature had linked poor housing conditions to worse health outcomes [16,21-23]. Conversely, availability of appropriate plumbing facilities and in-home clean water has been linked to a decline in infectious diseases including lower respiratory tract infections [14,24] and highlights the importance of this study.

Materials and methods

We conducted a cross sectional ecological analysis of the 3141 US counties using publicly available data relating poor housing conditions with COVID-19 outcomes. Counties with missing data for poor housing condition (n = 6) were excluded yielding a sample size of 3135 US counties. These counties were Prince of Wales-Outer Ketchikan, Skagway-Hoonah-Angoon, Wrangell-Petersburg and Kusilvak from Alaska, Bedford City from Virginia and Oglala Lakota from South Dakota. Counties in US territories (American Samoa, Guam, Northern Mariana Island, Puerto Rico and US Virgin Islands) were not included in our analysis. Since there is no individual identifying information, the data were in aggregate by county, and publicly available from the Centers for Disease Control (CDC), US census Bureau and John Hopkins Coronavirus Resource Center, the protocol received exemption from the Providence Veterans Affairs Medical Center Institutional Review Board [2,25,26].

Exposure

The exposure of interest was the percentage of households in a county with poor housing conditions which has been published as percentage of households with severe housing problems (2010–2014) by the CDC [27]. These households were identified as having any of the four problems: 1) overcrowding, 2) high housing cost burden, 3) incomplete kitchen facilities, and 4) incomplete plumbing facilities. In further breakdown, overcrowding is defined as more than one person per room, high housing cost as more than 50% of the monthly household income allocated towards housing cost (including utilities), incomplete kitchen facilities as lacking a sink with running water, stove or range, or a refrigerator, and incomplete plumbing facilities as lacking of hot and cold piped water, a flush toilet, or a bathtub/shower [8]. To facilitate interpretation, we studied the poor housing conditions as a continuous variable by units of 5% increase in households with poor housing conditions, which is roughly equivalent to the standard deviation (SD) of the variable. Additionally, we also categorized the counties according to approximate quartiles based on percentage of households with poor housing conditions (rounded values were used as cut off).

Outcome

The outcomes of interest were incidence relative risk and mortality relative risk of COVID-19 related to poor housing condition. For this purpose, we obtained the COVID-19 cases and death data from the John Hopkins Coronavirus Resource Center, where it has been published for “public health, educational, and academic research purposes.” [2] We obtained the cumulative data for three dates of 10-day intervals: March 31st, April 10th and April 21st, which allowed us to test the robustness of our findings across three temporal cross sections of the reported data. April 21st data were utilized in the main analysis. These numbers of cases and deaths per county were subsequently incorporated into an equation with the total population of each county as denominator for standardization to incidence (cases/100,000) and mortality (deaths/100,000) in the regression modelling [26], to calculate the incident rate ratio and mortality rate ratio related to poor housing conditions.

Covariates

We obtained data on county level variables related to COVID-19 spread and outcomes based on literature. The data regarding total population (2010) and population density (population per square miles of land area of a county) were obtained to account for the exposure pool [27]. In order to account for socioeconomic disparity, data on median household income of a county (2016) and percentage of residents without a high school diploma (2013–2017) were included [27].

County demographic data was collected because male sex, older age and percent of racial minority have been linked to a higher risk of, and mortality in, COVID-19 [28-31]. This included percentage of male residents (2010), median age (2010), and percentage of white, black, Hispanic or Latino, Asian, Native Hawaiian or Pacific Islander; and American Indian or Alaska Native residents (2013–2017) [27].

From the very initial reports, COVID-19 was seen to drastically effect individuals with a heavier burden of comorbidities [28,29]. We therefore collected data for percentage of residents diagnosed with diabetes and those with obesity (2015) [27]. Furthermore, Medicare hospitalization data for: hypertension, ischemic stroke, myocardial ischemia, heart failure and dysrhythmia (2014–2016) were obtained as a surrogate for cardiovascular disease burden in the county [27].

The morbidity and mortality associated with COVID-19 has been linked to respiratory failure [32]. We therefore collected variables that affect respiratory health: annual concentration of particulate matter 2.5 μ (pm2.5) (2014) and percentage of residents who are current smokers (2017) [27].

We accounted for the percentage of adults without health insurance under 65 years (2016) and number of hospitals with emergency rooms (ER) (2016) in each county, as surrogates of access to care. The number of hospitals may also influence the number of cases detected in the county. Furthermore, we obtained the total number of tests conducted per state (cumulative up to April 21st) and calculated the ratio of tests to the total population of the state (test density) [33].

Statistical analysis

County covariates were described as mean ± SD and range for continuous variables and as number (%) for categorical variables. Linear regression was used to assess linear trend of the county covariates across the four quartiles.

We utilized multilevel generalized linear models with a negative binomial distribution family and a log link function (with county population as a denominator) to determine the association between county-level prevalence of poor housing conditions and the county-level incidence and mortality of COVID-19 standardized to the population of each county. To account for clustering effect due to policy, social and behavioral similarities across counties within the same state, we applied a random intercept for the state. The covariance matrix was specified as unstructured. We adjusted for different categories of variables in the regression model in a stepwise fashion: 1) population density and test density, 2) demographics (% male, median age, % white), 3) socioeconomic status (median household income, % residents with lack of high school education), 4) respiratory exposure (annual ambient PM2.5, % current smokers), 5) prevalence of comorbidities (% diagnosed with diabetes, % diagnosed with obesity), 6) Medicare hospitalization rates (hypertension, ischemic stroke, myocardial ischemia, heart failure, dysrhythmia), and 7) Access to healthcare (% adults without health insurance under 65 years and number of hospitals with ER). The fully adjusted model included all the aforementioned variables. We reported incidence rate ratios (IRR) and mortality rate ratios (MRR) and 95% confidence intervals (CIs), respectively, which are interpretable as the relative increase in incidence and mortality rates for COVID19, for each 5% increase in households with poor housing conditions.

We conducted several sensitivity analyses to assess the robustness of our findings. 1) In lieu of percent white population in a county, we used the percent breakdown of the minority population: black, Hispanic or Latino, Asian or Native Hawaiian or Pacific Islander and American Indian or Alaska Native residents per county in the fully adjusted model. 2) Quartile Analysis—we also studied quartiles of percent households with poor housing conditions. We tested the quartiles in the fully adjusted model as an ordinal variable (i.e. linear effect across quartiles) and as categorical (dummy) variables (using 1st quartile as referent). 3) Temporal Analysis—Since the main analysis is using the most recent data (April 21st, 2020), we repeated our analyses in two earlier time points for COVID-19 incidence and mortality as outcomes, on March 31st and April 10th, to account for and understand temporal changes, if any, in this association.

A two-sided p value of <0.05 was considered statistically significant. All analyses were conducted in Stata SE statistical software (Stata Corp, Texas, v. 15.0).

Results

Across 3135 US counties, the mean (range) percentage of households with poor housing conditions was 14.2% (2.7% to 60.2%). The modified quartiles of percentage of households with poor housing conditions were 2.7% to 11% (Quartile1), 11.1% to 14.0% (Quartile2), 14.1% to 17.0% (Quartile3) and 17.1% to 60.2% (Quartile4). Till April 21st, there were a total of 144190 confirmed COVID-19 cases assigned to these 3135 US counties and 14887 COVID-19 deaths. The mean (SD) number of cases and deaths of COVID-19 were 255.68 (2877.03) cases and 13.90 (272.22) deaths per county, respectively. Table 1 describes characteristics of counties across the US, overall and stratified by quartiles of percentage of households with poor housing conditions.

Table 1

Variable	Overall Mean ± SD (Range) n = 3135	Quartiles of Percentage Households with Poor Housing Conditions
		Quartile 1(2.7%— 11%) Mean ± SD(Range)n = 679	Quartile 2(11.1%—14%) Mean ± SD(Range)n = 912	Quartile 3(14.1%—17%) Mean ± SD(Range)n = 830	Quartile 4(17.1—60.2%) Mean ± SD(Range)n = 714
Total Population (2010)	98418.71 ± 313273.6 (82–9818605)	20417.61 ±35194.54 (478–365169)	51102.02 ± 82563.37 (286–1223348)	93130.64 ± 158212.9 (662–1503085)	239180.8 ±603202.7 (82–9818605)
Population Density, n/100,000	345.18 ± 3792.37 (0.04–168127.7)	43.16 ± 110.13 (0.16–2014.21)	183.96 ± 1372.22 (0.17–37333.46)	581.41 ± 6459.65 (0.09 ± 168127.7)	560.31 ± 3459.51 (0.04–79500.17)
Number of Tests (State level)	91539.76 ± 101898.8 (7618–633861)	NA	NA	NA	NA
Test Density (number of tests/ total population) (State Level)	± 0.01 (0.01–0.04)	NA	NA	NA	NA
Male (2010), %	49.97 ± 2.22(43.2–72.1)	50.37 ± 2.30(44.8–68.3)	49.89 ± 1.90(44.5–64.7)	49.90 ± 2.27(43.2–72.1)	49.78 ± 2.41(45.3–66.9)
Median Age (2010), years	40.34 ± 5.04(22.4–62.7)	42.53 ± 4.27(23.5–55.8)	41.13 ± 4.26(26.3–62.7)	39.78 ± 4.53(24–54)	37.91 ± 5.97(22.4–55.9)
White (2013–2017), %	76.78 ± 20.04(0.6–100)	87.3 ± 13.27(5.8–100)	83.01 ±14.07(11.3–99.8)	75.73 ±17.95(12.8–99.1)	60.03 ±23.10 (0.6–99.3)
Black (2013–2017), %	8.87 ± 14.45(0–86.9)	2.72 ± 6.60(0–60.2)	6.26 ±9.94(0–71.9)	10.52 ± 14.34 (0–86.3)	16.12 ± 20.27(0–86.9)
Hispanic/Latino (2013–2017), %	9.08 ± 13.72(0–99.1)	6.60 ± 11.22(0–93.8)	6.57 ± 10.15(0–88.4)	8.78 ± 12.75(0–86.5)	14.97 ± 18.39(0–99.1)
Asian Population or Native Hawaiian or Other Pacific Islander Population (2013–2017), %	1.36 ± 3.10(0–62.7)	0.56 ± 0.83(0–6)	0.85 ± 1.39(0–17.3)	1.28 ± 2.08(0–27.4)	2.85 ± 5.55(0–62.7)
American Indian/Alaska Native (2013–2017), %	1.67 ± 7.11(0–82.2)	± 3.07 (0–54.6)	1.12 ± 3.53 (0–39.6)	1.32 ± 4.64 (0–55.7)	3.56 ± 13.36 (0–90.3)
Diagnosed Diabetes, Age-Adjusted Percentage, 20+, (2015), %	9.83 ± 2.24(3.6 ± 18.3)	9.11 ± 1.96 (4.6–16.6)	10.05 ± 2.01 (5.1–17.4)	10.19 ±2.16 (4.1–16.6)	9.84 ± 2.68 (3.6–18.3)
Obesity, Age-Adjusted Percentage, 20+. (2015), %	31.96 ± 4.72(13.5 ± 49.7)	32.40 ± 3.77 (18.3–46.1)	32.72 ± 3.91 (16.7–43.8)	32.24 ±4.55 (14.7–49.7)	30.25 ± 6.07 (13.5–46.9)
Acute Myocardial Infarction Hospitalization (2014–2016) Rate per 1,000 Medicare Beneficiaries	11.82 ± 3.81(3.6–36.1)	11.05 ± 3.58 (3.9–31.5)	12.33 ± 3.92 (3.6–36.1)	12.29 ±3.85 (4.8–29.2)	11.35 ± 3.64 (5–26.9)
Dysrhythmia Hospitalization (2014–2016) Rate per 1,000 Medicare Beneficiaries	58.52 ± 16.91 (16.6 ± 136.9)	51.59 ± 17.13 (16.6–114.6)	60.46 ± 16.72 (16.9–119.5)	62.64 ±16.28 (17.3–136.9)	57.87 ±15.55 (17.3–116.1)
Heart Failure Hospitalization (2014–2016) Rate per 1,000 Medicare Beneficiaries	50.89 ± 19.60 (9.3–148.3)	43.30 ± 18.82 (9.3–131.2)	52.70 ± 18.91 (9.9–138.2)	55.34 ± 19.47 (11.1–138.3)	50.65 ± 18.94 (10.2–139.8)
Hypertension Hospitalization (2014–2016) Rate per 1,000 Medicare Beneficiaries	132.10 ± 40.01 (32.1–283.2)	114.20 ± 38.47 (42.4–254.1)	135.23 ± 37.93 (38.6–276.2)	141.29 ± 39.33 (38.5–283.2)	134.47 ± 39.64 (32.1–249.5)
Ischemic Stroke Hospitalization (2014–2016) Rate per 1,000 Medicare Beneficiaries	12.68 ± 3.39 (3.6–35.7)	11.61 ± 3.51 (4.4–35.7)	13.13 ± 3.32 (4–33.8)	13.33 ± 3.30 (3.6–27.8)	12.36 ± 3.17 (3.7–22.4)
Median Household Income (2016), In thousands of $	49.52 ± 12.89 (22–134.6)	51.56 ± 11.61 (24.5–114.7)	49.39 ± 12.23 (24.9–134.6)	48.67 ± 12.81 (23.1–115.5)	48.75 ± 14.65 (22–110.8)
Residents without High School Diploma, Ages 25+, (2013–2017), %	13.81 ± 6.48 (1.1–58.7)	11.65 ± 5.89 (1.1–47.5)	13.64 ± 5.95 (1.8–58.7)	14.32 ± 6.19 (1.9–41.8)	15.50 ± 7.36 (1.3–51.2)
Adults who are current smokers (2017), %	17.45 ± 3.58 (6–41)	16.49 ± 3.11 (7–31)	17.69 ± 3.13 (8–27)	17.93 ± 3.42 (6–30)	17.51 ± 4.46 (8–41)
Particulate Matter <2.5 μm (2014) Annual Average Ambient Concentrations, ug/m3	9.02 ± 1.97 (3–19.7)	8.39–2.12 (3.4–12.6)	9.32 ± 1.84 (3–13.3)	9.21 ± 1.77 (3.4–14.3)	9.03 ± 2.05 (3.7–19.7)
Adults without Health Insurance, Under Age 65, (2016), %	11.12 ± 4.92 (2.1–33.5)	9.94 ± 4.83 (2.9–31.1)	10.68 ± 4.75 (3.0–30.9)	11.37 ± 4.65 (2.1–33.5)	12.51 ± 5.18 (2.6–31.0)
Hospitals with Emergency Department (2016)	1.15 ± 1.89 (0–46)	0.68 ± 0.70 (0–4)	0.92 ± 0.99 (0–15)	1.13 ± 1.27 (0–10)	1.91–3.36 (0–46)
COVID 19 Cases per County (Up to 3/31/2020)	58.04 ± 836.02 (0–43119)	3.10 ± 11.48 (0–141)	8.82 ± 23.09 (0–325)	27.17 ± 98.98 (0–1591)	207.31 ± 1733.53 (0–43119)
COVID-19 Cases per County (Up to 4/10/2020)	156.35 ± 1851.53 (0–92384)	8.95 ± 31.12 (0–431)	26.58 ± 64.54 (0–825)	85.00 ± 313.08 (0–4511)	540.69 ± 3,824.68 (0–92384)
COVID-19 Cases per County (Up to 4/21/2020)	255.68 ± 2877.03 (0–144190)	15.58 ± 51.43 (0–615)	46.82 ± 117.92 (0–1635)	155.48 ± 560.40 (0–9621)	860.10 ± 5,933.84 (0–144190)
COVID-19 Deaths per County (Up to 3/31/2020)	0.97 ± 17.36 (0–932)	0.06 ± 0.33 (0–5)	0.20 ± 0.70 (0–7)	0.59 ± 3.12 (0–70)	3.26 ± 35.98 (0–932)
COVID-19 Deaths per County (Up to 4/10/2020)	5.76 ± 107.42 (0–5820)	0.27 ± 1.25 (0–19)	0.87 ± 2.63 (0–27)	2.73 ± 13.53 (0–282)	20.58 ± 223.13 (0–5820)
COVID-19 Deaths per County (Up to 4/21/2020)	13.90 ± 272.22 (0–14887)	0.62 ± 2.78 (0–39)	1.93 ± 5.68 (0–74)	7.44 ± 33.88 (0–506)	48.91 ± 565.70 (0–14887)

The mean number of COVID-19 cases and deaths per county for all time points, March 31st, April 10th and April 21st increased across increasing quartiles of percentage of households with poor housing conditions (all p’s <0.001). Similarly, increasing quartiles of percent households with poor housing conditions were associated with higher county population, population density, percentage of minority residents, percentage of residents without high school diploma, prevalence of diabetes, Medicare hospitalization rates for hypertension, ischemic stroke, heart failure and dysrhythmia, percentage of current smokers, annual PM2.5 levels, percentage of adults <65 years without health insurance and number of hospitals with ER facilities (all p’s <0.001). Conversely, a decrease in the median household income and prevalence of obesity was observed across increasing quartiles of percent households with poor housing conditions (both p’s <0.001).

Using the COVID-19 data from April 21st, we found that for each 5% increase in poor housing condition per county, there is a 59% increase in the relative risk of COVID-19 incidence (IRR 1.59, 95% confidence interval [CI]: 1.49–1.70) (Table 2). The association was only slightly attenuated after adjustment for an extensive list of county covariates (Table 2). The IRR for the fully adjusted model (Model VIII) was 1.50 (95% CI: 1.38–1.62) (Table 2). Secondary analyses categorizing the exposure into quartiles (Fig 1a and 1b) or using data from March 31st and April 10th yielded comparable results (Table 2).

Table 2

Relative risk of county Covid-19 incidence as of April 21, 2020 related to percent households with poor housing conditions.

Model	Incidence Rate Ratio, IRR (95% Confidence Interval)
I- Percentage Households Living with Poor housing conditions	1.59 (1.49–1.70)
II- Model I + Population Density and Test Density	1.58 (1.48–1.68)
III- Model II + Demographics (Male, Age, % White)	1.31 (1.21–1.42)
IV- Model III + Socioeconomic Factors (Median Household Income, % Residents with Lack of High School Education)	1.48 (1.38–1.60)
V- Model IV + Respiratory Exposure Variables (Annual PM25 and Percentage of Residents who reported Smoking)	1.55 (1.44–1.67)
VI- Model V + County Prevalence of Comorbidities (Diabetes and Obesity)	1.52 (1.41–1.65)
VII- Model VI + Hospitalization Rate per 1,000 Medicare Beneficiaries (HF, Dysrhythmia, HTN, Ischemic Stroke, MI)	1.51 (1.40–1.63)
VIII- Model VII + % Adults without insurance + Number of Hospitals with ER/county (Fully Adjusted Model)	1.50 (1.38–1.62)
Sensitivity Analyses
IX- Model VIII with Minority Demographics (Male, Age, Black%, Hispanic%, Asian%. Native Americans% in lieu of White%)	1.43 (1.32–1.54)
X- Model VIII with poor housing conditions Quartiles as an Ordinal Variable	1.25 (1.18–1.32)
XI- Model VIII for March 31st, 2020	1.65 (1.51–1.81)
XII- Model VIII for April 10th, 2020	1.24 (1.15–1.34)

Note: Incidence Rate Ratios are from: Multilevel Generalized Linear Negative Binomial Models allowing random intercepts for States. A 5% increase in households with poor housing conditions represents approximately 1-SD change.

Fig 1

Relative risk increase in incidence and mortality of Covid-19 with each quartile of percentage of households with poor housing conditions.

Similarly, using the COVID-19 data from April 21st, we found a 63% increase in the relative risk of COVID-19 mortality for each 5% increase in poor housing condition per county, (MRR 1.63, 95% CI: 1.48–1.79) (Table 3). The association was only mildly attenuated after adjustment (Table 3), and remained highly significant in the fully adjusted model (Model VIII), with an MRR of 1.42 (95% CI 1.25–1.61) (Table 3). Secondary analyses categorizing the exposure into quartiles (Fig 1a and 1b) or using data from March 31st and April 10th yielded comparable results (Table 3).

Table 3

Relative risk of county Covid-19 mortality as of April 21, 2020 related to percent households with poor housing conditions.

Model	Mortality Rate Ratio, MRR (95% Confidence Interval)
I- Percentage Households Living with Poor housing conditions	1.63 (1.48–1.79)
II- Model I + Population Density and Test Density	1.62 (1.47–1.78)
III- Model II + Demographics (Male, Age, White%)	1.25 (1.12–1.40)
IV- Model III + Socioeconomic Factors (Median Household Income, % Residents with Lack of High School Education)	1.41 (1.26–1.58)
V- Model IV + Respiratory Exposure Variables (Annual PM25 and Percentage of Residents who reported Smoking)	1.49 (1.32–1.67)
VI- Model V + County Prevalence of Comorbidities (Diabetes and Obesity)	1.44 (1.27–1.63)
VII- Model VI + Hospitalization Rate per 1,000 Medicare Beneficiaries (HF, Dysrhythmia, HTN, Ischemic Stroke, MI)	1.44 (1.27–1.64)
VIII- Model VII + %Adults without insurance + Number of Hospitals with ER/county (Fully Adjusted Model)	1.42 (1.25–1.61)
Sensitivity Analyses
IX- Model VIII with Minority Demographics (Male, Age, Black%, Hispanic%, Asian%. Native Americans% in lieu of White%)	1.37 (1.21–1.56)
X- Model VIII with poor housing conditions Quartiles as an Ordinal Variable	1.22 (1.11–1.34)
XI- Model VIII for March 31st	1.36 (1.10–1.67)
XII- Model VIII for April 10th	1.31 (1.14–1.51)

Note: Mortality Rate Ratios are from: Multilevel Generalized Linear Negative Binomial Models allowing random intercepts for States. A 5% increase in percent households with poor housing conditions represents approximately 1-SD change.

Discussion

To our knowledge, this is the first nationwide study to investigate county level association of COVID-19 incidence and mortality with percentage of households facing poor housing conditions in the US. Our study showed that with each 5% increase in percent households with poor housing conditions, there was a 50% higher risk of COVID-19 incidence and a 42% higher risk of COVID-19 mortality across US counties. Findings remained similar in three different time points and after accounting for county-level population density and state test density, demographics, socioeconomic status, prevalence of comorbidities, respiratory exposure, lack of health insurance and number of ER facilities.

Of the four factors categorized under poor housing conditions, overcrowding and a lack of access to adequate plumbing and sanitation offered the most direct explanation for the higher incidence and mortality of COVID-19. The 2003 spread of severe acute respiratory syndrome (SARS) epidemic in Hong Kong was worst in the Amoy Garden estate which was overcrowded and had significant plumbing and sanitation problems [34]. An initial study from China investigating the initial COVID-19 outbreaks also showed that 79.9% of outbreaks occurred indoors, almost all in apartment settings [35]. Evidence from the Influenza epidemic of 1918 showed not only increased spread but also increased severity of the disease as a result of overcrowding [36]. Overcrowding and inadequate plumbing may lead to repeated exposure and potentially a higher viral inoculum, which had been linked to worse COVID-19 clinical outcomes [37-39]. A more severe COVID-19 disease process may offer a potential explanation of the higher mortality. Health education via print or social media targeted at the population at risk to improve awareness of preventative measures and hygiene should be employed to counteract the potential risks in overcrowded facilities [40]. Moreover, to mitigate indoor airborne transmission of COVID-19, investment in engineering controls to improve ventilation, prevent recirculation and using air cleaning filters and disinfecting mechanisms have been proposed to be potential solutions to mitigate indoor transmission, applicable to communal living (apartments), transportation (e.g. bus, train or stations) and work spaces [41]. Moreover, a lack of appropriate plumbing and kitchen facilities in their residence would require the residents to use communal facilities, thereby increasing social contact. It is important to note that SARS-CoV-2 is detectable for up to 72 hours on plastic and stainless steel materials, whereas influenza virus is detectable for only 24–48 hours after [42,43]. This poses another public health risk for transmission to other individuals sharing a crowded space, in the absence of meticulous hygiene. Establishment of hygiene protocols and increased availability of mobile bathrooms and cleaning supplies at communal facilities should also be considered to help mitigate the COVID-19 spread [44].

Another factor of poor housing is high housing cost, which translates into a lack of resources when it comes to seeking healthcare as well as being able to stay at home [15,17,19,45]. Our study showed that counties with the highest percentage of residents with poor housing conditions also had the lowest median household income and highest percentage of residents without a high school diploma. However, an important feature of our study is that although linked to economic status, the associations between poor housing conditions and COVID-19 incidence, and mortality, were independent of the racial composition, median household income, lack of education and lack of health insurance, as the relative risk was only mildly changed and remained highly significant after adjusting for these factors.

Our findings have health policy implications as they identify a particularly vulnerable population to be at heightened risk and also the potential pathways for public health interventions during the current COVID-19 pandemic. In addition, our study adds to a robust body of evidence for other disease processes, which has shown that inadequate housing is a public health hazard especially in relation to infectious diseases and highlights the importance of finding short (e.g. better access to clean water and bathrooms) and long-term (e.g. overcrowding, cost) solutions to problems surrounding poor housing to help contain or mitigate the spread of COVID-19.

The strength of this study is that this is a nationwide report of 3135 counties across US, which allows for a large sample size and generalizability of our findings. Furthermore, to our knowledge, this is the first study to establish an association between the incidence and mortality of COVID-19 and poor housing conditions. The limitations of the study also merit consideration. First, the county-level covariate data utilized were from earlier time period, and hence may have weakened the strength of the associations. However, we utilized the most updated results publicly available. Furthermore, the assumption that county age structure and ethnic composition does not quickly change over the span of several years is the current approach shared by the US Census methodology (every 10 years). However, the consistent results after accounting for extensive list of covariates and various sensitivity analyses, support the robustness of the findings. Due to limitations of the data, we could not separate the distinct elements (e.g. overcrowding, cost, plumbing, kitchen) that comprised poor housing for better understanding of the problem and targeting of policies. This is also a cross-sectional ecological analysis and does not lend itself to causal inference. Finally, despite careful adjustments and inclusion of covariates, residual confounding cannot be excluded.

Conclusion

In a nationwide analysis of US county-level data, counties with a higher percentage of households with poor housing had higher incidence of, and mortality associated with, COVID-19. These findings suggest targeted health policies to support individuals living in poor housing conditions should be considered in further efforts to mitigate adverse outcomes associated with COVID-19.

(CSV)

Click here for additional data file.

13 Jul 2020

PONE-D-20-15293

Association of Poor Housing Conditions with COVID-19 Incidence and Mortality Across US Counties

PLOS ONE

Dear Dr. Wu,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

I apologize for the delay; it has been very difficult to obtain reviewers during the pandemic given the abundance of manuscripts and manuscript review requests.  Please respond to the reviewer comments on a point-by-point basis and revise the manuscript accordingly.

Please submit your revised manuscript by Aug 27 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Jeffrey Shaman

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional 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

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. 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).

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

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. 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: Partly

**********

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

Reviewer #1: No

**********

3. 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: Yes

**********

4. 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: Yes

**********

5. 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: Overall review and general recommendation:

Understanding within country disparities of COVID-19 cases and mortality is important for informing tailored approaches to the pandemic. The authors present a cross-sectional, ecological study of the association between poor housing conditions and outbreak intensity (number of cases and number of deaths) in 3135 US counties.

The paper addresses the important topic of structural determinants of COVID-19 and has the potential to be of high impact. However, there are several critical issues with the methods and interpretation. Therefore, I recommend major revisions. Below I detail my concerns.

Major comments:

1. It is unclear why number of cases and deaths are used instead of standardized rates. Presenting and comparing standardized measures would strengthen the paper.

2. Because the outcome does not seem to be standardized (is defined as number of deaths and infections), the interpretation of the outcome (IRR and MRR) is unclear. For example, a 5% increase in the percent of households with poor housing conditions is associated with a 50% higher risk of COVID-19 incidence. And yet, incidence seems to be defined differently in this paper – if I understand correctly, incidence is defined as number of deaths (and not deaths per population). Thus, the result could be interpreted as a 5% increase in the percent of households with poor housing conditions is associated with a 50% increase in the probability of each additional death per county (correct?).

On page 7, line 120, the authors state total population is used as the denominator, but the outcome seems to presented as counts (number of cases and number dead) throughout the paper. I recommend using rates or further expanding on the decision to use count data.

3. In adjusted analyses, the authors control for population density, but this won’t adjust (or standardize for comparison) for the underlying total population per county from which cases arise (due to different areas of the counties).

Minor comments:

1. Introduction: Overall, the Introduction is very brief. The rationale for the measurement of poor housing conditions is described in the Discussion. I think it would be helpful to readers to have this background at the beginning of the manuscript. In particular, the reason high housing costs is included would be informative, and seems distinct from the structural issues, such as shared toilet or kitchen facilities.

2. Page 8, lines 129-132: How were variables chosen for inclusion in the final model?

3. Page 15, lines 184-185: The inclusion of additional dates is a strength of this study. However, I would expect increasingly higher R (and thus exponentially more cases) in the counties with poorer housing conditions, but the finding is that the results are consistent across time. This could be further explored.

4. The Conclusion calls for targeted health policies to support individuals living in poor housing conditions. I recommend restructuring the Discussion a bit – moving the motivation for the exposure variables to the Introduction and/or Methods, and discussing the interpretation and implications of the findings in the Discussion. In particular, what kind of interventions might be successful? Are there interventions with demonstrated success in other countries – for example, meal delivery?

**********

6. 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: Yes: Danielle N. Poole

[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.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

21 Sep 2020

Dear Editor-in-Chief, PLoS One:

We thank the reviewer and the editors for their insightful comments and suggestions. We have tried to thoroughly address each of the questions/comments by the reviewer and we believe the manuscript has been made stronger by incorporating these changes. Please find enclosed a point-by-point response to the reviewer’s comments.

Reviewer #1: Overall review and general recommendation:

Understanding within country disparities of COVID-19 cases and mortality is important for informing tailored approaches to the pandemic. The authors present a cross-sectional, ecological study of the association between poor housing conditions and outbreak intensity (number of cases and number of deaths) in 3135 US counties.

The paper addresses the important topic of structural determinants of COVID-19 and has the potential to be of high impact. However, there are several critical issues with the methods and interpretation. Therefore, I recommend major revisions. Below I detail my concerns.

Major comments:

1. It is unclear why number of cases and deaths are used instead of standardized rates. Presenting and comparing standardized measures would strengthen the paper.

Response - We apologize for our lack of clarity. Our analysis does represent standardized rates, standardized to the denominator population in each county. The outcome was a relative risk increase in incidence and mortality attributable to poor housing conditions.

We utilized STATA® commands to get a standardized measure, by using the total population of the county as a denominator (page 7). Therefore, while the input variable is indeed the total number of cases or deaths in the county, the output is standardized by incorporating the total population in the equation. The “exposure(n)” in the command used (below) allows for calculation of rates within each county in the regression model.

We utilized the following STATA command

Equation 1- “meglm death poorhousing, exposure (n) || state:, covariance (unstructured) family (nbinomial) link (log) irr”

Where “death” is the number of deaths on the specified date in the county, “poorhousing” is the percentage of population living in poor housing conditions, “n” is the total population of the county. This allows to model the incidence of Covid-19 cases or deaths per given number of county population (e.g., per 100,000). The model output is a relative risk - either incidence rate ratio or mortality rate ratio.

2. Because the outcome does not seem to be standardized (is defined as number of deaths and infections), the interpretation of the outcome (IRR and MRR) is unclear. For example, a 5% increase in the percent of households with poor housing conditions is associated with a 50% higher risk of COVID-19 incidence. And yet, incidence seems to be defined differently in this paper – if I understand correctly, incidence is defined as number of deaths (and not deaths per population). Thus, the result could be interpreted as a 5% increase in the percent of households with poor housing conditions is associated with a 50% increase in the probability of each additional death per county (correct?).

On page 7, line 120, the authors state total population is used as the denominator, but the outcome seems to presented as counts (number of cases and number dead) throughout the paper. I recommend using rates or further expanding on the decision to use count data.

Response - We again apologize for the lack of clarity on the issue of standardized rate (standardized to the underlying population). As we indicated in the response to comment #1, the total population was incorporated into an equation as a denominator in modeling the incidence of case and deaths, thus the IRRs and MRRs reflect the relative risk increase in the incidence and mortality rates for 5% increase in the percent of households with poor housing conditions. Please see “Equation-1” in response to comment #1 above. We have now added appropriate units in lines 171, 176, 180 and 186 for further clarification.

We have also added the following description to the methods section for further clarity :

“The outcomes of interest were incidence relative risk and mortality relative risk of COVID-19 attributable to poor housing condition. For this purpose, we obtained the COVID-19 cases and death data from the John Hopkins Coronavirus Resource Center, where it has been published for “public health, educational, and academic research purposes.”(2) We obtained the cumulative data for three dates of 10-day intervals: March 31st , April 10th and April 21st, which allowed us to test the robustness of our findings across three temporal cross sections of the reported data. April 21st data were utilized in the main analysis. These numbers of cases and deaths per county were subsequently incorporated into an equation with total population of a of each county as a denominator for standardization to incidence (cases/100,000) and mortality (deaths/100,000) in the regression modelling(26). Therefore, we were able to calculate the incident rate ratio and mortality rate ratio attributable to poor housing conditions. “

And therefore, the interpretation would be “each 5% increase in the percentage of households with poor housing conditions is associated with a 50% increase in incidence rate per given population (e.g. per 100,000)

3. In adjusted analyses, the authors control for population density, but this won’t adjust (or standardize for comparison) for the underlying total population per county from which cases arise (due to different areas of the counties).

Response – indeed as the reviewer points out rightly, adjusting for population density will not account for the total population per county. In addition to standardizing the regression equation output to the county population, we have also included population density as an additional adjustment in the regression equation to account for “crowding” at county level since our ongoing hypothesis is that transmission could be higher in highly densely populated areas.

Minor comments:

1. Introduction: Overall, the Introduction is very brief. The rationale for the measurement of poor housing conditions is described in the Discussion. I think it would be helpful to readers to have this background at the beginning of the manuscript. In particular, the reason high housing costs is included would be informative, and seems distinct from the structural issues, such as shared toilet or kitchen facilities.

Response- We thank the reviewer for this suggestion and have moved the rationale of individual components of poor housing conditions to the introduction: how high housing cost (more than 50% of the household income going towards housing cost) leads to a trade off from other aspects of living including health insurance and education and has been shown to contribute to delays in seeking healthcare. These four problems have been grouped together and reported as poor housing conditions by CDC. (In introduction: p.4, paragraph 2)

2. Page 8, lines 129-132: How were variables chosen for inclusion in the final model?

Response - All variables utilized in the study were informed by literature review and included in the final model. We have incorporated county level variables related to COVID-19 spread and outcomes based on literature referenced for each variable (Covariates in Materials and Methods).

3. Page 15, lines 184-185: The inclusion of additional dates is a strength of this study. However, I would expect increasingly higher R (and thus exponentially more cases) in the counties with poorer housing conditions, but the finding is that the results are consistent across time. This could be further explored.

Response – We agree that the incidence would be expected to increase with time (until effective measures are instituted and implemented) as is seen with number of cases and deaths in the descriptive statistics (Table 1). However, the relative risk of COVID-19 incidence and mortality related to poor housing conditions should remain fairly stable within the three 10-day intervals used in this study since there have not been significant effective strategies that disproportionately affected one group versus another of the main exposure in the COVID-19 risks during the study period. This is evidenced by the consistent IRR and MRR across different time points.

4. The Conclusion calls for targeted health policies to support individuals living in poor housing conditions. I recommend restructuring the Discussion a bit – moving the motivation for the exposure variables to the Introduction and/or Methods, and discussing the interpretation and implications of the findings in the Discussion. In particular, what kind of interventions might be successful? Are there interventions with demonstrated success in other countries – for example, meal delivery?

Response- We have moved the motivation for the exposure variable (and each component thereof) to introduction as recommended.

- We have added possible interventions to discussion, including increased education, improved ventilation and disinfection, increased availability of supplies and mobile bathrooms (lines 227-231, 247-248)

We hope you will find our revisions satisfactory and our manuscript will be acceptable for publication in your esteemed journal.

Respectfully,

Khansa Ahmad, MD & Wen-Chih Wu, MD

Submitted filename: reviewCOVIDplosone-hw-SEedits-hw_clean_SEedits.docx

Click here for additional data file.

14 Oct 2020

Association of Poor Housing Conditions with COVID-19 Incidence and Mortality Across US Counties

PONE-D-20-15293R1

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,

Jeffrey Shaman

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

19 Oct 2020

PONE-D-20-15293R1

Association of Poor Housing Conditions with COVID-19 Incidence and Mortality 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

Prof. Jeffrey Shaman

Academic Editor

PLOS ONE