Prior diagnoses and medications as risk factors for COVID-19 in a Los Angeles Health System

Introduction

Global cases of novel coronavirus disease 2019 (COVID-19) reached 9.8 million in June 2020, with the United States accounting for 2.4 million cases^1,2. While still in the midst of the first acute phase of the pandemic^2–4, knowledge of risk factors associated with COVID-19 susceptibility and severity can shape government policies, identify at-risk populations, guide clinical decision making, and prioritize future COVID-19 research. Existing studies have identified risk factors in hospitalized COVID-19 patients that are associated with death, intensive care unit admission or ventilation. Hypertension, cardiovascular disease, renal disease, and diabetes have consistently been identified as risk factors for a severe outcome in hospitalized patients in China, Italy and the United States^5–14. On the other hand, there have been limited analyses on risk factors that confer an increased susceptibility for being diagnosed with COVID-19 (severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) positivity), and inpatient admission, both of which are important stages of the COVID-19 disease spectrum. In addition, data collection in previous studies were predominantly performed via standardized forms from electronic health records (EHR)^8,14,15 or registry data reported to centralized coordinating centers^7,12. These data are standardized, less detailed, and have limited access to records from previous encounters compared to direct EHR extraction^6,16.

In addition, certain COVID-19 risk factors may be emphasized based on study design and vary due to the geographical and population context¹⁷. Within California, Los Angeles County has seven times more COVID-19 cases (93,000) compared to the second ranked county¹⁸. Los Angeles comprises a racially and ethnic diverse population with 48% self-identified Hispanic or Latinos, 15% Asians, and 9% Black or African Americans spread over 4,000 square miles¹⁹. With hospitals and clinics predominantly located in Los Angeles County, the University of California, Los Angeles (UCLA) Health System cares for individuals from this community. In this retrospective study, our objective was to identify pre-existing risk factors for COVID-19 diagnosis susceptibility, inpatient admission, and severe outcome through the automated EHR extraction of a large metropolitan medical system²⁰. We identified several less established risk factors and also confirmed many known demographic and clinical COVID-19 risk factors.

Results

Study Subjects

The UCLA Health System includes two hospitals (520 and 281 inpatient beds) and 210 primary and specialty outpatient locations predominantly located in Los Angeles County. We leveraged an extract of the de-identified electronic health records from the UCLA Health System known as the Data Discovery Repository (DDR), which contains longitudinal records for more than 1.5 million patients since 2013.

This retrospective analysis included individuals who were tested for SARS-CoV-2 via RT-PCR within the UCLA Health System between March 9, 2020 and June 14, 2020, where the first positive SARS-CoV-2 PCR test was on March 12, 2020. Of the 26,602 individuals PCR tested for SARS-CoV-2 (Tested) (55% female, 44% male; mean (standard deviation, SD) 49 (20) years old), 992 were COVID-19 positive (3.7% of Tested; 51% female, 48% male; mean (SD) 49 (20) years old), 220 were admitted in the hospital (Inpatient) (22% of COVID-19 positive; 43% female, 56% male; mean (SD) 62 (21) years old) and 77 were treated in the intensive care unit or required intubation (Severe) (7.7% of COVID-19 positive, 35% of Inpatient; 38% female, 61% male; mean (SD) 56 (20) years old) (Figure 1).

• Download figure
• Open in new tab

Figure 1.

Flow diagram of subjects included in COVID-19 susceptibility, inpatient and severe analyses for pre-existing phecodes and medications

Prior Diagnoses as COVID-19 Risk Factors

To identify diagnoses that were risk factors for COVID-19 severity groups, we analyzed diagnoses entered in the EHR prior to an individual’s SARS-CoV-2 testing. ICD codes were mapped to ∼1,800 phecodes, which have been shown to represent meaningful and interpretable phenotypes²⁰. We sought COVID-19 susceptibility risk factors by comparing COVID-19 positive individuals to a random sampling of 10,000 individuals who were age- and sex-matched and drawn from the DDR cohort. In addition, we analyzed the presence of prior diagnoses between the COVID-19 positive Inpatient group compared to the COVID-19 positive Outpatient group and between the COVID-19 positive inpatient Severe group compared to COVID-19 positive inpatient not Severe group (Figure 1).

Three hundred eighty-six phecodes present prior to SARS-CoV-2 testing were associated with COVID-19 diagnosis susceptibility and included known risk factors such as cardiovascular disease, type 2 diabetes, hypercholesterolemia, hypertension, and chronic renal disease (Table S3) when adjusting for age and sex (Methods). To identify phenotypic risk factors not correlated with and not already explained by major known COVID-19 risk factors, we controlled for these major known COVID-19 risk factors (coronary heart disease, congestive heart failure, chronic obstructive pulmonary disease, type 2 diabetes, hyperlipidemia, hypertension, obesity, and chronic renal disease^{7,8,10,13–15,27}). We identified 80 phecodes observed prior to SARS-CoV-2 testing that were associated with COVID-19 susceptibility (Figure 2, Table S4). Less well-established significant COVID-19 diagnosis susceptibility risk factors included dementias (dementia (OR 5.2 [3.2-8.3], p=2.6 × 10⁻¹⁰), Alzheimer’s disease (OR 7.7 [3.9-15], p=1.2 × 10⁻⁸)), mental health conditions (major depressive disorder (OR 2.1 [1.6-2.8], p=1.1 × 10⁻⁶), anxiety disorder (OR 1.7 [1.4, 2.0], p=5.3 × 10⁻⁷)), and vitamin D deficiency (OR 1.8 [1.4-2.2], p=5.7 × 10⁻⁶) (Figure 2, Table S4).

• Download figure
• Open in new tab

Figure 2.

Significant prior phecode risk factors for COVID-19 susceptibility and inpatient correcting for age, sex and known risk factors (Model 2) grouped by phecode category. For phecode categories, dermatologic, congenital anomalies, and neoplasms were combined to form dermatologic. Neurological and sense organs were combined to form neurologic. heme = hematopoietic, psych = mental disorders.

Individuals with dementia and Alzheimer’s disease may be at greater risk of contracting COVID-19 as they are older and have more comorbidities²⁸. In the DDR sample, patients over 65 years old with dementia have on average 2.3 known risk factors (standard deviation (SD) = 1.8) versus patients over 65 years old without dementia who have on average 0.98 known risk factors (SD = 1.4) (p<0.001). However, these comorbidities are less likely to confound these analyses, as we controlled for age and known comorbid risk factors. Individuals with dementia may also have limited access to accurate information and live in facilities that have high infection rates^28–32.

Mental health conditions, including depression and anxiety, have not been described as prior risk factors for COVID-19 susceptibility. This may be due to the known association of mental health conditions and health outcomes including pneumonia and barriers to accessing health care^33,34. We evaluated if these mental health associations were present in the UCLA Health system. Using the DDR Sample controlling for age and sex, we found that both depression and anxiety were positively associated with risk of influenza (depression OR 4.5 [2.6, 7.1], p<0.001; anxiety OR 2.2 [1.4, 3.2], p<0.001). Access to health care was likely not a contributing factor in our health system as patients with depression or anxiety compared to those without depression or anxiety had a greater number of health system encounters, after controlling for age and sex (20 encounters per year depression or anxiety vs 10 encounters per year without depression or anxiety, OR per 10 encounters per year 1.5 [1.4, 1.6], p<0.001).

The association of low vitamin D levels with COVID-19 disease has been proposed because rates of infection are higher in countries at higher latitudes^35–37. To further investigate if vitamin D insufficiency was associated with COVID-19 susceptibility, we evaluated 25-hydroxyvitamin D laboratory values. Indeed, we found that the minimum level of 25-hydroxyvitamin D one year prior to SARS-CoV-2 testing was significantly lower in COVID-19 positive cases compared to the DDR sample, when correcting for age and sex (mean (SD) 29.1 ng/mL (11.7) COVID-19 positive vs 30.4 ng/mL (13.0), OR per 5 ng/mL 0.92 [0.84, 0.99], p=0.045), but not significant after correcting for age, sex and known risk factors (OR per 5 ng/mL 0.99 [0.88, 1.0], p=0.28) (Methods). These differences in vitamin D levels may not be clinically significant at an individual level, but support that vitamin D deficiency across the cohort increases the risk of COVID-19 diagnosis.

Additional significant prior risk factors associated with COVID-19 diagnosis susceptibility included phecodes representing common symptoms of COVID-19^8,9,15. These included respiratory symptoms (cough (OR 2.7 [2.3, 3.3], p<2.2 × 10⁻¹⁶)), musculoskeletal symptoms (chronic pain (OR 1.9 [1.5, 2.3], p=7.2 × 10⁻⁸), and malaise and fatigue (OR 1.9 [1.5, 2.4], p=3.3 × 10⁻⁸)) (Figure 2, Table S4).

Phecodes present prior to SARS-CoV-2 testing significantly associated with the COVID-19 positive Inpatient group were renal disease (renal dialysis, acute renal failure), respiratory disease (pneumonia, pleurisy), and sepsis (Table S5). Controlling for known risk factors, 22 significant phecodes were identified as risk factors for the COVID-19 positive Inpatient group. Among these, less well-established risk factors included individuals with a history of acidosis (OR 7.4 [3.1, 18], 6.9×10⁻⁶), sepsis (OR 8.4 [3.9, 19], p=2.5×10⁻⁸), kidney replaced by renal transplant (OR 9.7 [2.8, 39], p=3.2×10⁻⁴), and disorders involving the immune mechanism (OR 6.0 [2.3, 16], p=2.7×10⁻⁴) (Figure 2, Table S6).

As acidosis and sepsis are typically diagnosed during relatively severe inpatient admissions, these factors may represent patients with a severe hospitalization prior to SARS-CoV-2 testing. Previous studies suggested a more rapid COVID-19 clinical progression in renal transplant patients^38–40, but they did not investigate if renal transplant was associated with increased COVID-19 inpatient admission risk. Lastly, the “disorders involving the immune mechanism” phecode maps to ICD codes including miscellaneous immunodeficiency diseases (but not including humoral or T-cell defects). Immunodeficiency has yet to be shown as a COVID-19 risk factor for inpatient admission.

Even after controlling for chronic renal disease and other major known COVID-19 risk factors, multiple phenotypes associated with renal disease remained significantly associated with COVID-19 inpatient admission risk including kidney replaced by renal transplant as described above, acute renal failure (OR 4.4 [2.0, 9.7], p=1.5×10⁻⁴), renal dialysis (OR 11 [3.5, 39], p=2.5×10⁻⁵), and anemia in chronic kidney disease (OR 16 [4.8, 63], p=2.2×10⁻⁶) (Figure 2, Table S6).

A previous history of thrombocytopenia was also significantly associated with COVID-19 inpatient admission risk (OR 4.1 [1.9, 8.8], p=1.3×10⁻⁴). Many prior studies showed that thrombocytopenia during hospitalization is a risk factor for severe COVID-19 outcome^8,9,15,41,42. We confirmed via laboratory test results that the minimum platelet level one year prior to SARS-CoV-2 testing was lower in COVID-19 inpatient cases compared to the COVID-19 outpatient cases, when correcting for sex, age and known risk factors (mean (SD) 173 ×10³/μL (82) COVID-19 Inpatient vs 240 ng/mL (66) COVID-19 Outpatient; OR per 10 ×10³/μL 0.89 [0.86, 0.93], p<0.001) (Methods).

We did not find any phecodes significantly associated with the COVID-19 Severe group, although our power was low (N=56 COVID-19 positive with Severe outcome and at least one encounter prior to SARS-CoV-2 testing) compared to the COVID-19 Inpatient group (N=177) and COVID-19 positive group (N=843).

Discussion

Identifying COVID-19 risk factors along the entire disease spectrum can inform patient care, public policies, and future research aimed at improving outcomes and reducing health care disparities. We leveraged our ability to query a de-identified electronic health records to determine COVID-19 risk factors in the UCLA Health System⁶¹. We found previously reported health disparities in the Hispanic or Latino, and non-white populations and confirmed known co-morbidities such as cardiovascular disease and chronic renal disease that are risk factors for COVID-19 susceptibility and inpatient admission. We also identified several less well-established risk factors such as pre-existing dementia, depression, vitamin D deficiency, and previous renal transplant. Furthermore, we found that immunosuppressive medication was associated with a decreased COVID-19 positive testing risk, but an increased inpatient admission risk.

This study differs from previous studies in that it was performed with direct EHR extraction as opposed to cases series, manual extraction, standardized form extraction from electronic health records, or registries^{7,8,10,12,14,15,62}, allowing detailed phenotyping from extended historical data. Recent preprints have used direct EHR extraction from collaborations between the United States and South Korea¹⁶ and the United Kingdom⁶. The detailed phenotyping and increased power from these studies identified COVID-19 mortality risk factors in the United Kingdom not reported in manual EHR extraction studies including stroke, dementia, and autoimmune diseases (rheumatoid arthritis, lupus, psoriasis).

Significant risk factors for becoming critically ill (intensive care unit admission or death) among hospitalized COVID-19 patients include older age, cardiac disease (cardiac injury, coronary artery disease), pulmonary disease (chronic obstructive pulmonary disease, acute respiratory distress syndrome), renal disease (chronic renal disease, acute kidney injury), hypertension, diabetes, hyperlipidemia, and obesity^{9–11,14,63,64}. Previously reported comorbid risk factors for severe outcome in COVID-19 hospitalized patients are also prior risk factors for COVID-19 diagnosis susceptibility and inpatient admission in our study. However, we find risk factors that are distinct to COVID-19 susceptibility and inpatient admission. For example, the risk of testing positive for COVID-19 increased with cardiovascular disease (coronary artery disease, congestive heart failure), hypertension, and hypercholesterolemia while risk factors for inpatient admission, were predominantly renal disorders (chronic renal failure, acute renal failure) and thrombocytopenia, the latter of which was confirmed via laboratory test results.

We identified prior risk factors that are less well-established for COVID-19 susceptibility including dementia, major depressive disorder, anxiety disorder and vitamin D deficiency. Individuals with dementia may be at greater risk of contracting COVID-19 as they are older and have more comorbidities, although we corrected for these in our analysis. We suspect that other environmental factors are contributing, including living in facilities which have had high infection rates, difficulty adhering to public health recommendations and limited access to accurate information^28–32.

Studies on the mental health impact of COVID-19 have shown an increased rate of depressive symptoms^65,66. Depression and anxiety as prior risk factors for COVID-19 susceptibility have not been previously described. Previous studies have correlated mental health condition with respiratory disease including pneumonia and pneumococcal disease³³. We confirmed in the UCLA health system that individuals with depression and anxiety have an increased risk of influenza. These associations may be due to diminished effort for personal protection and barriers to accessing health care^33,34, although we did not find that patients with depression or anxiety had less UCLA Health System encounters. Given that dementia and mental health conditions are risk factors for COVID-19 susceptibility, healthcare organizations and communities should consider strategies in addition to social distancing to reduce this risk, as social distancing may contribute to the risk⁶⁷.

There has been speculation using country-wide aggregate data that low vitamin D may be associated with COVID-19, as rates of infection are higher in countries at higher latitudes^{35– 37}. While there are many other explanations for this latitude effect (e.g. rates of testing, global patterns of spread), here, we show using individual patient data that vitamin D deficiency is associated with increased COVID-19 susceptibility risk. A recent preprint showed that low serum 25-hydroxycholecalcifoerol was more prevalent in ICU compared to non-ICU hospitalized patients⁶⁸. We confirmed the association of vitamin D insufficiency with COVID-19 susceptibility using 25-hydroxyvitamin D laboratory levels when correcting for age and sex, but not when correcting for known risk factors. We did not evaluate whether vitamin D supplementation would be protective.

Less well-established inpatient admission prior risk factors included renal transplant and disorders involving the immune mechanism. Studies from Italy, New York, and China have reported the clinical course of hospitalized renal transplant patients with COVID-19 suggesting some may have a more rapid clinical progression^38–40. Our study also shows that renal transplant patients have an increased risk of inpatient admission with COVID-19.

While there is controversy over whether immunosuppressant medication increases risk of COVID-19 severity^50–54, we found that individuals with the “disorders involving the immune mechanism” phecode, which maps to ICD codes including immunodeficiency, were at increased risk of inpatient admission. Further studies are warranted to determine if other clinical characteristics of this patient group increase the risk of COVID-19 severity and if there are mechanistic links.

In our study cohort, older age was not a risk factor for all stages of the COVID-19 disease spectrum. Individuals older than 65 years old were more likely to be admitted, but made up a lower percentage of individuals experiencing a severe outcome. As a corollary, young age was not completely protective, as those in the intermediate age group between 26 and 55 years old were more likely to have a severe outcome once admitted. Studies from the CDC and New York showed intensive care unit admission or ventilation was lower in intermediate age groups compared to older age groups^69–71, in contrast to our study. In the UCLA Health System, it is possible clinicians have a lower threshold of inpatient admission for older individuals, preferring to monitor their COVID-19 clinical course. However, less of these admitted older individuals progress to severe disease. Clinicians may have a higher threshold for admitting intermediate age group individuals where those admitted may have more significant COVID-19 symptoms. Therefore, once admitted intermediate age group individuals are more likely to progress to severe disease. Risk factors other than age and co-morbidities likely also contribute to COVID-19 severe outcomes. These risk factors may be environmental such as air pollution⁷², socioeconomic such as access to health care^26,73, and genetic^43,74.

The increased COVID-19 disease susceptibility and severity risk in individuals with self-identified Hispanic or Latino ethnicity further highlights health disparities in COVID-19. We found a similar association in individuals with self-identified race other than White or Caucasian and self-identified Black or African American race. The Los Angeles County of Public Health has confirmed these disparities; the age-adjusted rate of COVID-19 cases is over 100 per 100,000 individuals for these minority groups but only 78 for those who self-identify as white, non-hispanic²². Similar findings have been found by the New York City Health Department and Chicago Department of Public health where the rates of COVID-19 cases and severe outcomes are roughly double in Black and Hispanic or Latino individuals compared to White individuals^21,23. Other studies from Connecticut and a 14 state public health COVID-19 surveillance network (COVID-NET) including California predominantly found risk of COVID-19 susceptibility and hospitalization in Black individuals^70,75. As we are unable to appropriately assess socioeconomic status using the available EHR data, ethnic discrepancy may be due to this and other associated variables⁷⁶. For example, shelter-in-place guidelines are most burdensome to Hispanic or Latino individuals in Los Angeles County as many live in neighborhoods with densely populated communities and have limited access to both open spaces and nearby supermarkets²⁶.

There has been concern that some individuals with COVID-19 have an exaggerated immune response resulting in cytokine storm or secondary haemophagocytic lymphohistiocytosis, and studies have suggested that short-term immunosuppression is warranted^51,54,77. Previous studies investigated small numbers of COVID-19 positive patients (<4) on immunosuppression prior to COVID-19 diagnosis, finding that these patients in general did well and did not have a severe outcome^50,52,53. Here we analyzed over 80 patients on either oral steroids or other immunosuppressants. Individuals on oral steroids had a decreased COVID-19 positive testing rate, but an increased inpatient admission risk. The explanation for this may be biological, behavioral, or partially due to clinician bias. It is possible immunosuppression results in asymptomatic COVID-19, thus lowering numbers of immunosuppressed individuals who seek testing and therefore test positive. However, if positive, these individuals have severe symptoms requiring admission. Another explanation is that clinicians lower their threshold for inpatient admission for patients on immunosuppression. We did not assess whether immunosuppressive medication was stopped, continued or changed on admission, and therefore, we are unable to draw conclusions on whether to adjust immunosuppression in COVID-19 hospitalized patients.

This study identifies important and less established risk factors for susceptibility and severity of outcomes in COVID-19. These risk factors should spur future work in understanding the mechanistic underpinnings of these observations and implementing strategies to mitigate both biological and socioeconomic risk factors.

Funding/Support

This work was supported by grants R25NS065723 from the Neurological Disorders and Stroke of the National Institutes of Health and the UCLA David Geffen School of Medicine - Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research Award Program.

UCLA Precision Health Data Discovery Repository Working Group Authors

Anna Liza Malazarte Antonio, DrPH; Maryam Ariannejad; Angela Marquessa Badillo, R.N.; Brunilda Balliu, Ph.D.; Yael Berkovich, B.A.; Michael Broudy, B.A.; Tony Dang; Chris Denny, M.D.; Eleazar Eskin, Ph.D.; Eran Halperin, Ph.D.; Brian L. Hill, M.S.; Ankur Jain, M.S.; Vivek Katakwar, B.Sc.; Clara Lajonchere, Ph.D.; Clara Magyar, Ph.D.; Sheila Minton; Ghouse Mohammed, B.Tech; Ariff Muhamed; Pabba Pavan; Michael A. Pfeffer, M.D.; Nadav Rakocz, B.Sc; Akos Rudas, M.Sc; Rey Salonga; Timothy J Sanders, B.A.; Paul Tung, B.S.; Vu Vu, B.Sc.; Ailsa Zheng, B.A.

Affiliations of the UCLA Precision Health Data Discovery Repository Working Group Authors

UCLA Health, David Geffen School of Medicine, University of California, Los Angeles, California (Antonio, Badillo, Berkovich, Broudy, Dang, Jain, Katakwar, Minton, Mohammed, Muhamed, Pavan, Salonga, Sanders, Tung, Vu, Zheng); UCLA Institute for Precision Health, Los Angeles, USA (Ariannejad, Lajonchere); Department of Computational Medicine, David Geffen School of Medicine, University of California, Los Angeles, California (Balliu, Eskin, Halperin, Rakocz, Rudas); Division of Hematology/Oncology, Department of Pediatrics, Gwynne Hazen Cherry Memorial Laboratories, University of California, Los Angeles, California (Denny); Department of Computer Science, University of California, Los Angeles, California (Hill); Department of Pathology and Laboratory Medicine, University of California, Los Angeles, California (Magyar)

Author Contributions

Conceptualization, T.S.C, Y.D., M.K.F., R.J., T.S., J.M.Y., C.H., A.W., D.H.G., M.J.B., and B.P.; Methodology, T.S.C, Y.D., M.K.F., J.M.Y., C.H., M.J.B., and B.P.; Software, T.S.C, Y.D., M.K.F., R.J., and T.S.; Validation, T.S.C, Y.D., M.K.F., R.J., and T.S.; Formal Analysis, T.S.C, Y.D., M.K.F., R.J., and T.S.; Investigation, T.S.C, Y.D., M.K.F., R.J., and T.S.; Resources, D.H.G., M.J.B., B.P., and UCLA Precision Health Data Discovery Repository Working Group; Data Curation, T.S.C, Y.D., M.K.F., R.J., T.S., and UCLA Precision Health Data Discovery Repository Working Group; Writing – Original Draft, T.S.C, M.K.F., D.H.G., M.J.B., and B.P.; Writing – Review & Editing, T.S.C, Y.D., M.K.F., R.J., T.S., J.M.Y., C.H., D.H.G., M.J.B., and B.P.; Visualization, T.S.C, Y.D., and M.K.F.; Supervision, T.S.C., M.J.B., and B.P.; Funding Acquisition, D.H.G., M.J.B., and B.P.

Declaration of Interests

The authors declare no competing interests

Methods

Supplemental Information

Table S1. Known risk factor diagnoses and assigned phecodes

Table S3. Bonferroni corrected significant prior phecode risk factors for COVID-19 susceptibility correcting for age and sex (Model 1). Orange shaded cells indicate Bonferroni corrected significant phecodes where nominal p-value < Bonferroni critical p-value. Red shaded cells indicate phecodes with Bonferroni critical p-value < nominal p-value < 0.05. Unshaded cells indicate phecodes with nominal p-value > 0.05. Nominal p-values are reported. Phecodes not assigned to any patients or resulting in a model error were excluded.

Table S4. Bonferroni corrected significant prior phecode risk factors for COVID-19 susceptibility correcting for age, sex, and known risk factors (Model 2). Orange shaded cells indicate Bonferroni corrected significant phecodes where nominal p-value < Bonferroni critical p-value. Red shaded cells indicate phecodes with Bonferroni critical p-value < nominal p-value < 0.05. Unshaded cells indicate phecodes with nominal p-value > 0.05. Nominal p-values are reported. Phecodes not assigned to any patients or resulting in a model error were excluded.

Table S5. Bonferroni corrected significant prior phecode risk factors for COVID-19 inpatient admission correcting for age and sex (Model 1). Orange shaded cells indicate Bonferroni corrected significant phecodes where nominal p-value < Bonferroni critical p-value. Red shaded cells indicate phecodes with Bonferroni critical p-value < nominal p-value < 0.05. Unshaded cells indicate phecodes with nominal p-value > 0.05. Nominal p-values are reported. Phecodes not assigned to any patients or resulting in a model error were excluded.

Table S6. Bonferroni corrected significant prior phecode risk factors for COVID-19 inpatient admission correcting for age, sex, and known risk factors (Model 2). Orange shaded cells indicate Bonferroni corrected significant phecodes where nominal p-value < Bonferroni critical p-value. Red shaded cells indicate phecodes with Bonferroni critical p-value < nominal p-value < 0.05. Unshaded cells indicate phecodes with nominal p-value > 0.05. Nominal p-values are reported. Phecodes not assigned to any patients or resulting in a model error were excluded.

Table S8. Medication class and medication. ACE inhibitor = angiotensin converting enzyme inhibitors, ARBs = angiotensin receptor blockers.