Risk Factors for testing positive for SARS-CoV-2 in a national US healthcare system

Abstract Background Identifying risk factors for SARS-CoV-2 infection could help health systems improve testing and screening strategies. Objectives Identify demographic factors, comorbid conditions, and symptoms independently associated with testing positive for SARS-CoV-2. Design Observational cross-sectional study. Setting Veterans Health Administration. Patients Persons tested for SARS-CoV-2 nucleic acid by polymerase chain reaction (PCR) between March 1 and May 14, 2020. Measurements Associations between demographic characteristics, diagnosed comorbid conditions, and documented symptoms with testing positive for SARS-CoV-2. Results Of 88,747 persons tested, 10,131 (11.4%) were SARS-CoV-2 PCR positive. Positivity was associated with older age (≥80 vs. <50 years: aOR 2.16, 95% CI 1.97-2.37), male sex (aOR 1.45, 95% CI 1.34-1.57), regional SARS-CoV-2 burden (≥2,000 vs. <400 cases/million: aOR 5.43, 95% CI 4.97-5.93), urban residence (aOR 1.78, 95% CI 1.70-1.87), Black (aOR 2.15, 95% CI 2.05-2.26) or American Indian/Alaska Native/Pacific Islander (aOR 1.26, 95% CI 1.05-1.52) vs. White race, and Hispanic ethnicity (aOR 1.52, 95% CI 1.40-1.65). Obesity and diabetes were the only two medical conditions associated with testing positive. Documented fevers, chills, cough, and diarrhea were also associated with testing positive. The population attributable fraction of positive tests was highest for regional SARS-CoV-2 burden (35.3%), followed by demographic variables (27.2%), symptoms (12.0%), obesity (10.5%), and diabetes (0.4%). Limitations Lack of information on SARS-CoV-2 exposures or the indications for testing which may affect the likelihood of testing positive. Conclusion The majority of positive SARS-CoV-2 tests were attributed to regional SARS-CoV-2 burden, demographic characteristics and obesity with a minor contribution of chronic comorbid conditions.


Introduction
Understanding the risk factors for testing positive for SARS-CoV-2 infection could help public health and health system initiatives to target testing, education, and preventive efforts toward those most likely to benefit. Sociodemographic risk factors for a positive SARS-CoV-2 test include advanced age, male sex, Black race, low socioeconomic status, and residence in a high-incidence area.
(1-4) Medical conditions such as chronic kidney disease and obesity may be risk factors for infection with SARS-CoV-2, although results have been inconsistent. (2,3,5) Most published studies examining the correlates of testing positive for SARS-CoV-2 have been conducted within individual health care facilities or regional health care systems rather than in national health systems.(1, 3, 5,

6)
Symptoms of SARS-CoV-2 including fever, shortness of breath, cough, loss of taste, and loss of smell are frequently included in outpatient SARS-CoV-2 screening questionnaires where laboratory or radiology testing is not routinely available. However, symptoms such as dyspnea or cough are common in the general population, lack specificity for SARS-CoV-2, and overlap with symptoms reported by persons with cardiopulmonary diseases. Whether SARS-CoV-2 symptoms might differ for those with underlying cardiopulmonary comorbidity is not known.
The dual aims of this study were to identify the baseline demographic factors and comorbidities associated with testing positive for SARS-CoV-2 infection in the Veterans Health Administration nationally, and to examine the symptoms associated with a positive SARS-CoV-2 polymerase chain reaction (PCR) test. Resource"(7) which includes analytic variables extracted from the CDW for all persons in VHA tested for SARS-CoV-2. We identified all VA patients tested for SARS-CoV-2 nucleic acid by PCR in the inpatient or outpatient setting between 02/28/2020 and 5/14/2020, excluding VA employees. The reasons for SARS-CoV-2 testing were not available in the study data but it is likely that some cohort members were tested as part of routine screening prior to hospital admission (8) or elective procedures. (9) This study was approved by the Institutional Review Board of the Veterans Affairs Puget Sound Healthcare System.

Outcome definition: positive or negative SARS-CoV-2 status
Patients were considered positive for SARS-CoV-2 if they had at least one positive PCR test during the study period and negative if all of their SARS-CoV-2 PCR tests were negative. Final adjudication of SARS-CoV-2 status was performed by the VA National Surveillance Tool, which is intended to be the single, authoritative data source for determination of positive and negative cases within VHA. The index date was defined as the date of the earliest positive (for those with at least one positive test) or negative (for those with only negative test results) test, unless the patient had been admitted to a VA hospital during the preceding 15 days, in which case the date of admission served as the index date. SARS-CoV-2 testing was considered to have been conducted in the inpatient setting if the test was obtained within a day before hospital admission.
A c c e p t e d M a n u s c r i p t Predictors: Demographic, pre-existing illnesses, and symptoms Sociodemographic characteristics included age, sex, race, and ethnicity, urban versus rural residence and regional SARS-CoV-2 burden and time period of testing. Urban versus rural residence was based on zip codes using the Rural-Urban Commuting Areas system, with urban areas defined as census tracts with at least 30 percent of the population residing in an urbanized area. Regional SARS-CoV-2 burden was operationalized as the number of cases per million in each participant's state of residence as of 4/10/2020,(10) categorized as <400, 400-999, 1000-1999, ≥2000. The time period of testing was based on quartiles of the cumulative number of tests performed during the study period.
Comorbid conditions were defined using ICD-10 codes recorded in VA administrative data during the 2-year period on or before the index date and body mass index (BMI). 25 The Charlson Comorbidity Index (CCI) provides an estimate of overall burden of comorbidity ( Table 2). Patients were categorized as being underweight (BMI <18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25-29.9) or having stage I (BMI 30-34.9) or stages II-III (≥35 BMI) obesity.
Documented symptoms thought to be related to SARS-CoV-2 were identified by VINCI analysts based on a combination of natural language processing of text notes in patients' electronic medical records (i.e., outpatient visit notes, admission notes, progress note, discharge summaries) stored as "text integration utilities" (TIU) notes in CDW, and searching for relevant ICD-10 codes in VA administrative data,(7) occurring on or within 30 days prior to the index date ( Table 3). We did not include information on the frequency of loss of smell or taste as these were not widely recognized as being symptoms of SARS-CoV-2 during the study period and were rarely reported.
A c c e p t e d M a n u s c r i p t

Statistical Analysis
We used multivariable logistic regression models to identify independent risk factors for testing positive for SARS-CoV-2 infection, with a significance threshold set at a p < 0.05. Analyses were adjusted for the aforementioned measured sociodemographic characteristics, comorbid conditions, test setting (inpatient or outpatient), and all documented symptoms (Tables 1-3). We also examined the sum of symptoms individually associated with a positive test in unadjusted analyses.
Analyses examining symptoms were stratified by presence or absence of heart failure, chronic obstructive pulmonary disease and asthma, comorbid conditions with prominent respiratory symptoms, with potential interactions examined a likelihood-ratio test. Additional supplementary analyses were conducted after stratifying by test setting (inpatient vs. outpatient).
Multivariate population attributable fractions (PAF) for each major risk factor were estimated by averaging over randomly selected permutations of the PAF when other risk factors were sequentially removed from the model. The number of permutations was sufficient to approximate the true average to 0.1%. Confidence intervals were calculated using Monte Carlo simulation (500 iterations over 5000 samples). The PAF corresponds to the estimated adjusted fraction of the positive SARS-CoV-2 tests that would not have occurred if the risk factor was not present (e.g. BMI ≥35).(11)

RESULTS
Of 88,747 persons tested, 10,131 (11.4%) had at least one positive SARS-CoV-2 PCR test during the study period, (Figure 1) among whom 27,062 (30.5%) were tested within a day before   Indian, Alaska Native, Native Hawaiian or Pacific Islander, 7.0% were of unknown race, and 9.3% were of Hispanic ethnicity.

Sociodemographic factors
Characteristics independently associated with a positive SARS-CoV-2 test included male sex (adjusted odds ratio (aOR) 1.45, 95% confidence interval (CI) 1.34-1.57), older age (age ≥ 80 vs. 18 Persons living in geographic regions with a higher burden of SARS-CoV-2 disease at the time of our study had a significantly higher risk of testing positive (living in a state with >2,000 versus <400 cases/million: aOR of 5.34, 95% CI 4.97-5.93). Living in an urban versus rural area was also associated with testing positive for SARS-CoV-2 (aOR 1.78, 1.70-1.87).

Comorbid Conditions
Most (68.1%) of those tested had a significant burden of comorbidity with a Charlson comorbidity score ≥ 1. (Table 2) Among the individual comorbid conditions examined, only diabetes was associated with a higher odds of testing positive (aOR 1.10, 95% 1.04-1. 16). Overall burden of comorbidity measured with the Charlson score, and many individual chronic conditions were associated with a lower risk of testing positive including a cancer, coronary artery disease, congestive heart failure, dialysis, cirrhosis, COPD, obstructive sleep apnea, alcohol dependence, and drug dependence.
Results were similar after adjustment for all documented symptoms (Tables 1-2). In analyses stratified by test setting (inpatient versus outpatient), the association of diabetes with a positive A c c e p t e d M a n u s c r i p t SARS-CoV-2 test was of greatest magnitude in the outpatient setting, while this association was attenuated and not statistically significant among those tested in the inpatient setting. Table 3). There were some differences in the chronic conditions associated with decreased odds of a positive SARS-CoV-2 test between inpatients and outpatients, although those with cancer, COPD, alcohol dependence and drug dependence had a lower risk of a positive test in both settings.
In analyses stratified by whether patients had a diagnosis of COPD, asthma or heart failure vs. none of these conditions, dyspnea was independently associated with testing positive (aOR 1.09, 95% CI 1.13-1.30) among those without any of these conditions but there was a negative association between dyspnea and testing positive among those with one or more of these conditions (aOR 0.77, 95% CI 0.69-0.84, p-value for interaction <0.001). (Table 4) Associations of fever and cough with testing positive were also attenuated in people with COPD, asthma or heart failure (p-value for interaction <0.01).
A c c e p t e d M a n u s c r i p t DISCUSSION Among persons tested for SARS-CoV-2 in a large national US healthcare system representing all 50 states, the largest number of positive tests were attributed to regional SARS-CoV-2 burden, followed by racial-ethnic group, overweight or obesity, male sex, and age ≥ 50 years, with less than 1% attributable to diabetes. Fever, chills, cough and diarrhea were associated with increased risk of testing positive for SARS-CoV-2 whereas dyspnea, sore throat, and abdominal pain were associated with a lower risk.
The high proportion of positive tests attributed to living in states with a greater SARS-CoV-2 burden after adjusting for individual pre-existing factors underscores the importance of public health measures to mitigate spread of the disease. (12) In our cohort, race and ethnicity were significantly associated with testing positive for SARS-CoV-2. Black cohort members had more than twice the odds of testing positive than white persons and, consistent with prior studies, we also found a higher risk of testing positive for SARS-CoV-2 among Black, American Indian and Hispanic cohort members. (2,(13)(14)(15) Hypothesized reasons for the racial and ethnic disparities include differences in underlying chronic medical conditions and socioeconomic factors that increase exposure to the virus due to being less able to social distance at home. (13,16) Our results likely support the importance of social determinants of health as our analyses adjusted for underlying health conditions and there was likely adequate access to testing since the proportion of Black persons in our cohort (26.5%) is actually higher than Black Veterans getting care in the Veterans Health Administration (15.5%). (17) Consistent with prior studies, older age, male sex, and obesity were associated with a positive SARS-CoV-2 test, (2,3,5) and may be related to differences in the immune response to SARS-CoV-2 (18)(19)(20). The relationship between obesity and testing positive persisted even after adjustment for diabetes and hypertension. The finding that current or past smokers were at lower risk for testing positive for SARS-CoV-2 than those who had never smoked was surprising. Our results are consistent with a study by De Lusignan et al. that found a similar 50% reduction in odds of a positive test among outpatients in a large primary care clinic network in the United Kingdom. (2) A c c e p t e d M a n u s c r i p t De Lusignan suggested several potential reasons for this finding including the fact that smokers may be more likely to be tested due to chronic cough, smoking may reduce the nasopharyngeal viral load decreasing PCR test sensitivity, and that nicotine may downregulate angiotensin-converting enzyme 2 receptors used by SARS-CoV-2 for cell entry. Also, the nicotinic acetylcholine receptor may be The Centers for Disease Control lists several symptoms of SARS-CoV-2 infection including fevers, chills, cough, dyspnea, fatigue, muscle aches, headache, new loss of taste or smell, sore throat, congestion or runny nose, nausea or vomiting, or diarrhea. Although commonly used to screen outpatients in health care settings, little is known about whether these symptoms predict a positive SARS-CoV-2 test result. We found that fevers, chills, cough and diarrhea were associated with a positive test, whereas patients with dyspnea, sore throat and abdominal pain were less likely to test positive. Interestingly, our findings suggest that the association between some symptoms and testing positive may vary as a function of patients' underlying comorbid conditions. For example, in our cohort dyspnea was associated with increased risk of a positive test in those without COPD, asthma or heart failure, but negatively associated with testing positive if those conditions were present. Complex interactions between underlying medical conditions and SARS-CoV-2 infection may therefore lead to different manifestations of symptoms, or chronic symptoms may influence the indication for testing.
Strengths of our study include analysis of data from a national integrated healthcare system, the large number of persons tested in both the outpatient and inpatient settings, availability of information on pre-existing chronic medical conditions, the large number of demographic characteristics and comorbidities investigated, and the availability of information on documented symptoms. However, our results in the predominantly male veteran population may not be generalizable to other populations and groups. Other limitations relate to the use of ICD-10 codes to ascertain comorbid conditions. Symptoms were assessed using natural language processing combined with ICD-10 codes, which rely on documentation of symptoms and the performance characteristics of these definitions is not known yet. We focused on predictors that would be A c c e p t e d M a n u s c r i p t available clinically in the outpatient setting, and therefore did not include laboratory data or imaging results in the models. We also did not have access to detailed information on social determinants of health such as household size or infection prevention behaviors (e.g. social distancing, mask use) which may affect the risk of becoming infected with SARS-CoV-2. In addition, our results likely reflect institutional policies and practices related to testing that might not be generalizable to other health systems.
In summary, the majority of positive SARS-CoV-2 tests within a national US healthcare system were attributed to geographic burden of SARS-CoV-2, male sex, advanced age, belonging to a racial/ethnic minority group and obesity, with only a minor contribution from underlying health conditions. The results of this study have implications for public health practice and for developing rational strategies for SARS-CoV-2 testing and screening.   A c c e p t e d M a n u s c r i p t M a n u s c r i p t M a n u s c r i p t * Adjusted for all the sociodemographic characteristics shown in Table 1 and the comorbid conditions shown in Table 2.
** Adjusted for all the sociodemographic characteristics shown in Table 1, the comorbid conditions shown in Table 2 and the symptoms shown in Table 3. M a n u s c r i p t * Adjusted for all the sociodemographic characteristics shown in Table 1 and the comorbid conditions shown in Table 2.
** Adjusted for all the sociodemographic characteristics shown in Table 1, the comorbid conditions shown in Table 2 and the symptoms shown in Table 3.