Author Contributions
Conceptualization, A.Y., A.R., H.T.B., C.S., A.M.H., J.H.B. and U.M.; methodology, A.Y., A.R., H.T.B., P.V., J.H.B., and U.M.; software, A.Y., A.R., H.T.B., P.V., and U.M.; validation, A.Y., A.R., H.T.B., P.V., and U.M.; formal analysis, A.Y., A.R., H.T.B., and U.M.; investigation, A.Y., A.R., H.T.B., C.S., P.V., J.H.B., and U.M.; resources, A.Y., A.R., H.T.B., P.V., U.M., and J.H.B.; writing—original draft preparation, A.Y., A.R., A.M.H., J.H.B., and U.M.; writing—review and editing, A.Y., A.R., H.T.B., C.S., P.V., A.M.H., J.H.B., and U.M.; visualization, A.Y., A.R., H.T.B., A.M.H., J.H.B., and U.M.; supervision, U.M., H.T.B. and J.H.B.; project administration, U.M. and J.H.B.; funding acquisition, A.R., U.M., and J.H.B. All authors have read and agreed to the published version of the manuscript.
Acknowledgments
We would like to thank a number of team members for assistance with various aspects of the study: Gina Marchesini (IRB approval), Andrew Pickles (SARS CoV-2 data retrieval), Tracey Rose, Adam Ortega, and Ian Kopp (SARS CoV-2 assays), and Amanda Norseth (data extraction). We would also like to thank members of the BRI Translational Research Core: Kassidy Benoscek, Sylvia Posso; and Data Management and Software Development Core: Anna Bjork, Adarsh Manjunath, Charlie Quinn, Alex Walker.
Figure 1.
Patient groups based on clinical outcome. Of the 180 adult patients, 153 completed home recovery without complications (Home Recovery No Complications group), 27 experienced subsequent worsening or new symptoms after discharge with 14 of these requiring hospitalization (Subsequently Hospitalized group), and the remaining 13 completing home recovery (Home Recovery With Complications group). The Combined Home Recovery group consisted of both Home Recovery No Complications and Home Recovery With Complications (n = 166).
Figure 1.
Patient groups based on clinical outcome. Of the 180 adult patients, 153 completed home recovery without complications (Home Recovery No Complications group), 27 experienced subsequent worsening or new symptoms after discharge with 14 of these requiring hospitalization (Subsequently Hospitalized group), and the remaining 13 completing home recovery (Home Recovery With Complications group). The Combined Home Recovery group consisted of both Home Recovery No Complications and Home Recovery With Complications (n = 166).
Figure 2.
Association of lymphocyte counts and virus cycle threshold with clinical outcome. (A) Lymphocyte counts were lower in the Subsequently Hospitalized group compared to the Home Recovery group. Further, within the Home Recovery group, 50% (3/6) of those with lymphopenia had complications compared to 19% (4/21) among those with normal counts. (B) Low lymphocyte counts (<1.0 × 109 cells/L) were associated with increased risk for subsequent hospitalization. Fifty percent (6/12) of participants with low lymphocyte counts were subsequently hospitalized compared to only 4.5% (1/22) of participants with normal counts. (C) Cycle threshold data, a surrogate for viral load, was not significantly associated with outcomes. Lower cycle threshold values indicate higher levels of viral nucleic acid. (D) Duration of illness at presentation was inversely associated with the level of virus.
Figure 2.
Association of lymphocyte counts and virus cycle threshold with clinical outcome. (A) Lymphocyte counts were lower in the Subsequently Hospitalized group compared to the Home Recovery group. Further, within the Home Recovery group, 50% (3/6) of those with lymphopenia had complications compared to 19% (4/21) among those with normal counts. (B) Low lymphocyte counts (<1.0 × 109 cells/L) were associated with increased risk for subsequent hospitalization. Fifty percent (6/12) of participants with low lymphocyte counts were subsequently hospitalized compared to only 4.5% (1/22) of participants with normal counts. (C) Cycle threshold data, a surrogate for viral load, was not significantly associated with outcomes. Lower cycle threshold values indicate higher levels of viral nucleic acid. (D) Duration of illness at presentation was inversely associated with the level of virus.
Figure 3.
Symptom clusters using hierarchal and variable clustering. (A) 2-way hierarchical clustering of symptoms shows three patient clusters/phenotypes and four clusters of co-occurring symptoms: symptom cluster A: flu-like with fevers, chills, cough, myalgia, diarrhea; symptom cluster B: upper respiratory, with nasal congestion, sinus pressure, sore throat, and headache; symptom cluster C: pneumonia-like, with shortness of breath and chest pain; and symptom cluster D: anorexia/nausea, fatigue, loss of smell/taste. (B) Patient cluster 2 is most associated with hospitalization, while cluster 3 had the lowest hospitalization rate. (C) Overall, cluster 1 patients had high prevalence of symptom cluster A (flu-like) and B (upper respiratory) symptoms. Cluster 2 patients experienced high prevalence of symptom cluster A (flu-like) and C (pneumonia-like) symptoms, including the highest prevalence of SOB among all patient clusters. Cluster 3 patients experienced few symptoms with the exception of cough. (D) Association of symptom clusters with outcome. Thirteen of 14 admitted patients had symptoms from Cluster C. Each data point represents a single patient and the marker symbols indicate types of outcomes. Four outpatients were asymptomatic. Odds ratio with 95% CI for the association with subsequent hospitalization for each symptom cluster are annotated at the top of the Venn diagram.
Figure 3.
Symptom clusters using hierarchal and variable clustering. (A) 2-way hierarchical clustering of symptoms shows three patient clusters/phenotypes and four clusters of co-occurring symptoms: symptom cluster A: flu-like with fevers, chills, cough, myalgia, diarrhea; symptom cluster B: upper respiratory, with nasal congestion, sinus pressure, sore throat, and headache; symptom cluster C: pneumonia-like, with shortness of breath and chest pain; and symptom cluster D: anorexia/nausea, fatigue, loss of smell/taste. (B) Patient cluster 2 is most associated with hospitalization, while cluster 3 had the lowest hospitalization rate. (C) Overall, cluster 1 patients had high prevalence of symptom cluster A (flu-like) and B (upper respiratory) symptoms. Cluster 2 patients experienced high prevalence of symptom cluster A (flu-like) and C (pneumonia-like) symptoms, including the highest prevalence of SOB among all patient clusters. Cluster 3 patients experienced few symptoms with the exception of cough. (D) Association of symptom clusters with outcome. Thirteen of 14 admitted patients had symptoms from Cluster C. Each data point represents a single patient and the marker symbols indicate types of outcomes. Four outpatients were asymptomatic. Odds ratio with 95% CI for the association with subsequent hospitalization for each symptom cluster are annotated at the top of the Venn diagram.

Figure 4.
Comorbidities are of high prognostic value. (A) Proportion of patients with individual comorbidities in the Home Recovery and the Subsequently Hospitalized groups. Number of patients with indicated comorbidity are annotated along the top of the figure and top of bars are annotated with proportion of Home Recovery or Subsequently Hospitalized group. BMI was unknown for 31 Home Recovery patients. (B) Increasing numbers of risk factors corresponded with increasing risk of adverse outcome. Probability of home recovery declined steadily with increasing comorbidities.
Figure 4.
Comorbidities are of high prognostic value. (A) Proportion of patients with individual comorbidities in the Home Recovery and the Subsequently Hospitalized groups. Number of patients with indicated comorbidity are annotated along the top of the figure and top of bars are annotated with proportion of Home Recovery or Subsequently Hospitalized group. BMI was unknown for 31 Home Recovery patients. (B) Increasing numbers of risk factors corresponded with increasing risk of adverse outcome. Probability of home recovery declined steadily with increasing comorbidities.
Figure 5.
Outcome decision tree analyses. Partitioning of data, where patient outcome is a binary outcome (Home Recovery or Subsequently Hospitalized). Variables included in this analysis were all symptoms, presence of any known comorbidities, lymphocytes, cycle threshold, age, BMI, and sex. Partitioning identifies SOB (p < 0.001), lymphocyte counts (p = 0.005), and comorbidities (p = 0.03) to be most associated with need for hospitalization. Observations with missing data are randomly and iteratively assigned to one of two sides of the split to utilize the entire cohort.
Figure 5.
Outcome decision tree analyses. Partitioning of data, where patient outcome is a binary outcome (Home Recovery or Subsequently Hospitalized). Variables included in this analysis were all symptoms, presence of any known comorbidities, lymphocytes, cycle threshold, age, BMI, and sex. Partitioning identifies SOB (p < 0.001), lymphocyte counts (p = 0.005), and comorbidities (p = 0.03) to be most associated with need for hospitalization. Observations with missing data are randomly and iteratively assigned to one of two sides of the split to utilize the entire cohort.
Table 1.
Clinical Characteristics and Demographics of Patient Population based on Outcome.
Table 1.
Clinical Characteristics and Demographics of Patient Population based on Outcome.
Characteristics | Home Recovery No Complications (N = 153) | Home Recovery with Complications (N = 13) | Subsequently Hospitalized (N = 14) | p Value |
---|
Age | | | | 0.05 1 |
N | 153 | 13 | 14 | |
Median | 52.6 | 56.5 | 56 | |
Range | (18.2–81.8) | (43.5–76.6) | (41.0–83.5) | |
Female, N (%) | 77 (50.3%) | 4 (30.8%) | 3 (21.4%) | 0.06 2 |
Duration from Symptom Onset (days) | | | | 0.54 1 |
N | 147 | 13 | 14 | |
Median | 5 | 4 | 4 | |
Range | (1.0–60.0) | (1.0–14.0) | (1.0–12.0) | |
Comorbidity, N (%) | 61 (39.9%) | 10 (76.9%) | 12 (85.7%) | 0.0003 2 |
Race, N (%) | | | | 0.62 2 |
Am. Indian, AK Nat., Nat. Hawaiian | 5 (3.3%) | 0 (0.0%) | 0 (0.0%) | |
Asian | 20 (13.1%) | 3 (23.1%) | 2 (14.3%) | |
Black, African American | 9 (5.9%) | 0 (0.0%) | 0 (0.0%) | |
Hispanic | 19 (12.4%) | 1 (7.7%) | 3 (21.4%) | |
Other | 25 (16.3%) | 1 (7.7%) | 0 (0.0%) | |
White | 75 (49.0%) | 8 (61.5%) | 9 (64.3%) | |
BMI Category, N (%) | | | | 0.63 3 |
Obese | 48 (39.0%) | 6 (50.0%) | 7 (50.0%) | |
Overweight | 40 (32.5%) | 5 (41.7%) | 5 (35.7%) | |
Normal | 34 (27.6%) | 1 (8.3%) | 2 (14.3%) | |
Underweight | 1 (0.8%) | 0 (0.0%) | 0 (0.0%) | |
Lymphocyte × 109/L | | | | 0.01 1 |
N | 20 | 7 | 7 | |
Median | 1.6 | 1.1 | 0.3 | |
Range | (0.6–2.4) | (0.7–3.5) | (0.3–1.9) | |
Neutrophil × 109/L | | | | 0.87 1 |
N | 20 | 7 | 7 | |
Median | 3.5 | 3.9 | 6.2 | |
Range | (1.5–8.9) | (1.8–5.1) | (0.8–11.5) | |
Cycle Threshold | | | | 0.46 1 |
N | 79 | 10 | 7 | |
Median | 15.1 | 16.8 | 12.6 | |
Range | (3.2–31.2) | (11.5–28.5) | (4.6–24.0) | |