Longitudinal lab test analysis confirms pre-existing anemia as a severe risk factor for post-viral clearance hospitalization in COVID-19 patients

As the number of new and recovering cases continues to rise, it is increasingly important to understand the long-term impacts of COVID-19 beyond the time of active SARS-CoV-2 infection. We previously used augmented curation methods to analyze physician notes from a large EHR system and identified anemia and acute kidney injury as risk factors for admission to the hospital after PCR confirmation of viral clearance in COVID-19 patients. Here, we analyzed longitudinal lab testing data from this same patient cohort to determine whether these measurements corroborate our notes-derived findings. Indeed we found that COVID-19 patients hospitalized after confirmed viral clearance tended to have lower hemoglobin and hematocrit measurements both during their SARS-CoV-2 positive intervals and during the one year prior to COVID-19 diagnosis compared to patients who have not been rehospitalized after viral clearance. Further, outright laboratory-based diagnosis of moderate or severe anemia was strongly enriched in the hospitalized cohort, suggesting that anemia pre-dating or concurrent with SARS-CoV-2 infection may predispose patients to long-term complications of COVID-19. Interventions which may mitigate anemia did not reduce the risk of post clearance hospitalization, although the efficacy of and patient compliance with these interventions could not be established. This study demonstrates the value of integrated large-scale EHR analyses and highlights the need for further research to determine whether the prevention or mitigation of anemia during SARS-CoV-2 infection may reduce the risk of suffering long-term complications of COVID-19.


Introduction
Since the first diagnosed case of COVID-19 in December 2019, over 50 million people have been infected with SARS-CoV-2 worldwide resulting in over 1.25 million deaths (ref). While significant progress has been made in understanding the pathogenesis of COVID-19, including the rapid development and clinical testing of multiple vaccine candidates (ref, ref, ref, ref, ref, ref) along with detailed characterizations of the SARS-CoV-2 entry receptor ACE2 (ref, ref, ref, ref, ref), there are still few options available for effective treatment of patients with severe COVID-19. Further, as the pandemic has progressed, there have been reports of long-lasting effects of COVID-19 even in patients who did not experience a severe disease course during their active period of infection (ref, ref, ref). However, the clinical, molecular, and/or demographic biomarkers characterizing patients who are more likely to experience these lasting effects after clearing SARS-CoV-2 are not yet known.
The need to answer such questions during the rapidly evolving COVID-19 pandemic has emphasized the requirement for tools facilitating real-time analysis of patient data as it is obtained and stored in large electronic health records (EHR) systems. Specifically, clinical research efforts to understand the features defining COVID-19 patients, or subsets thereof, fundamentally require reliable systems that enable (1) conversion of unstructured information (e.g., patients notes written by healthcare professionals) into structured formats suitable for downstream analysis and (2) temporal alignment and integration of such unstructured data with the already structured information available in EHR databases (e.g., lab test results, disease diagnosis codes).
With these requirements in mind, we have previously reported the development of augmented curation methods that enable the rapid creation and comparison of defined cohorts of COVID-19 patients within a large EHR system (ref, ref, ref, ref, ref). Here we expand on our prior textual sentiment-based analysis (ref) to understand the clinical features of patients likely to experience lasting effects of COVID-19, and we find that lab tests corroborate our previous results suggesting that anemia and kidney malfunction during active SARS-CoV-2 infection may serve as biomarkers of patients who are more likely to be subsequently hospitalized after PCRconfirmed viral clearance.

Definition of COVID-19 cohorts based on hospitalization after PCR-confirmed SARS-CoV-2 clearance
Using NLP-based extraction of phenotypes from a large EHR system, we previously reported that COVID-19 patients who are hospitalized after viral clearance (as assessed by RT-PCR) were more likely to experience anemia and acute kidney injury (AKI) in the year prior to their diagnosis and during their PCR-positive phase of COVID-19 compared to patients who were not re-hospitalized after clearance of SARS- . Here, we sought to assess whether diagnostic lab tests for AKI and anemia corroborate these phenotypic associations. As was previously described, we split the cohort of COVID-19 patients with confirmed viral clearance into two groups: (1) post-clearance hospitalized ("PCH"; n = 93) and (2) post-clearance nonhospitalized ("PCNH"; n = 173), where viral clearance was defined as two consecutive negative SARS-CoV-2 PCR tests following a positive test. A demographic summarization of these two cohorts is provided in Table 1.
To reduce imbalances between cohorts based on initial infection severity, our cohort inclusion criteria required patients in the PCNH cohort (n = 173) to have been admitted to the hospital during their index infection (i.e. while positive for SARS-CoV-2), while it did not require the same for patients in the PCH cohort (only 49 of 93 were hospitalized during index infection). To assess whether the severity of the index infection differed between these groups, we used ICU admission rates as a surrogate of clinical severity. When considering all patients, the rate of ICU admission was similar in the two cohorts (26/93 [28%] vs. 57/173 [33%]; p = 0.49; Figure  S1A), but when considering only patients who were hospitalized during index infection, the PCH cohort had a significantly higher rate of ICU admission (26/49 [53%] vs. 57/173 [33%]; p = 0.01; Figure S1B). Since the patients who were hospitalized during index infection will inherently provide more lab testing data, it will be important in our subsequent analyses to evaluate whether lab test features distinguishing these cohorts are independent of this difference in ICU admission rate.

Lab tests indicate that patients hospitalized after viral clearance are more likely to experience anemia and kidney dysfunction before and during their SARS-CoV-2 infection
After creating these two cohorts, we then compared a set of selected lab test results during two time windows: (1) the year prior to COVID-19 diagnosis and (2) the time during which each patient was SARS-CoV-2 positive according to their RT-PCR results. Given our previous EHRbased findings, we considered both anemia-related and kidney function lab tests including hemoglobin, hematocrit, estimated glomerular filtration rate (eGFR), serum creatinine, and serum blood urea nitrogen (BUN) levels (Figure 1).
For each patient, we first considered the median values of the given lab test over the designated interval. Histograms showing the number of measurements per patient in each time period for the selected tests are shown in Figures S2-S3. Consistent with our previous findings, we found that patients in the PCH cohort showed significantly lower median hemoglobin and hematocrit levels in both the pre-COVID and the SARS-CoV-2 positive phases ( Table 2, Figures  2A-D). In the SARS-CoV-2 positive phase, median eGFR was lower and median BUN was higher in the PCH cohort, but there were no significant differences in renal function test results between the cohorts in the pre-COVID phase (Table 2, Figure S4).
We also tested whether extreme (i.e. minimum or maximum) values of a given lab test over the designated periods varied between hospitalized and non-hospitalized patients, as a measure of central tendency (e.g., median) may fail to capture a single occurrence of phenotypes such as anemia or AKI. Specifically, we compared the patient-level minimum values of hemoglobin, hematocrit, and eGFR, and maximum values of serum creatinine and BUN in each time period. Interestingly, we found that the PCH cohort tended to have lower minimum values of hemoglobin, hematocrit, and eGFR in both the pre-COVID and SARS-CoV-2 positive phases, while also showing higher maximum serum BUN and creatinine during the SARS-CoV-2 positive phase (Tables 3-4, Figure 3 and Figure S5).
To consider a simple explanation, we asked whether these reductions in hemoglobin and hematocrit during the SARS-CoV-2 positive phase could be attributed to more frequent blood draws or higher ICU admission rates in PCH patients compared to their PCNH counterparts, but this was not the case. Specifically, there was no significant difference in the number of blood draws experienced by these cohorts during this interval ( Figure S6A). And while patients admitted to the ICU during SARS-CoV-2 index infection tended to have slightly lower hemoglobin values compared to non-ICU admitted patients (cohen's D = -0.283, p = 0.022; Figure S7), the difference was much stronger when comparing the PCH and PCNH cohorts (cohen's D = -0.70, p = 7.51x10 -6 ; Table 2). This indicates that the observed reduction in hemoglobin among PCH patients during the SARS-CoV-2 positive interval cannot be explained simply by their higher ICU admission rate during index infection ( Figure S1B). Taken together, these analyses corroborate our prior textual . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 sentiment-based EHR findings, suggesting that patients who are hospitalized after SARS-CoV-2 clearance are more likely to experience anemia and kidney malfunction both prior to and during SARS-CoV-2 infection.

Both male and female rehospitalized COVID-19 patients have lower hemoglobin and hematocrit measurements compared to patients who are not rehospitalized
Given that males and females have different normal ranges of hemoglobin and hematocrit, it was important to consider whether our observation was related to an imbalance in sex distributions between the PCH and PCNH cohorts. As seen in Table 1 To address this concern, we repeated our analysis of anemia lab tests split by sex. We found that patient-level median values of both hemoglobin and hematocrit during the SARS-CoV-2 positive phase were still significantly lower in both the male and female PCH cohorts versus their PCNH counterparts ( Table 5, Figures 4A-D). Further, the pre-COVID median measurements of hematocrit and hemoglobin were lower in the female PCH cohort, while hemoglobin (but not hematocrit) was significantly lower in the male PCH cohort ( Table 5, Figures  4E-H). Similarly, in our analysis of extreme values, both male and female PCH patients showed lower minimum measurements of hemoglobin and hematocrit during the SARS-CoV-2 positive phase ( Table 6, Figures 4I-L), while only females showed lower measurements in the pre-COVID phase ( Table 6, Figures 4M-P). We again confirmed that the reductions observed during the SARS-CoV-2 positive phase were not attributable to more frequent blood draws, as there was no significant difference in the number of blood draws per patient between these cohorts when split by sex (Figures S2B-C).

Rehospitalized patients are more likely to have experienced moderate or severe anemia prior to COVID-19 diagnosis and during active COVID-19 infection
The previous analyses have relied simply on the comparison of continuous lab test measurements between two cohorts without clinical interpretations of those measurements. However, given that anemia can be defined based on the lab tests we have considered, we next sought to determine whether the outright diagnosis of anemia occurs more frequently in the PCH cohort than the PCNH cohort. To do so, we identified every case of clinical anemia (defined as hemoglobin < 13.5 g/dL or hematocrit < 38.3% for males, and hemoglobin < 12.0 g/dL or hematocrit < 35.5% for females [ref]) for each patient in the pre-COVID and SARS-CoV-2 positive phases. Patients were then classified in a binary fashion for each time window based on whether the median of their pre-COVID phase measurements or the minimum of their SARS-CoV-2 positive phase measurements met the above criteria for laboratory-diagnosed anemia (defined as "pre-COVID anemia" and "SARS-CoV-2 positive anemia", respectively) ( Figure 5A). We found that anemia was observed more frequently in the PCH cohort during both the pre-COVID phase (27/42 [64%] vs. 23/65 [35%]; OR = 1.82; p = 0.005) and the SARS-CoV-2 positive phase (53/62 [85%] vs. 123/167 [74%]; OR = 1.16; p = 0.077), although the latter finding did not reach statistical significance (Figures 5B-C).
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10. 1101/2020 Anemia can be further characterized as mild, moderate, or severe based on hemoglobin measurements, where a hemoglobin level <10 g/dL is consistent with moderate or severe disease for both males and females. We thus performed a similar analysis to the one described above, but classified patients in a binary fashion for each time window based on whether the median of their pre-COVID phase hemoglobin measurements or the minimum of their SARS-CoV-2 positive phase measurements was <10g/dL. Here we found that moderate/severe anemia was observed more frequently in the PCH cohort during both the pre-COVID phase ( (Figures 5D-E).
Again it was important to assess whether this strong enrichment for moderate/severe anemia in PCH patients during the SARS-CoV-2 positive phase was related to or independent of index infection severity as estimated by ICU admission status. We found that among patients admitted to the ICU during index infection, the prevalence of moderate or severe anemia in the PCH cohort was 77% (20/26) compared to 30% (16/53) in the PCNH cohort (OR = 2.55; p = 1.11x10 -4 ; Figure S8A); among those not admitted to the ICU during index infection, the prevalence of moderate or severe anemia was 36% (13/36) in the PCH cohort versus 18% (20/114) in the PCNH cohort (OR = 2.06; p = 0.036 Figure S8B). Thus, it seems that the occurrence of moderate or severe anemia during the SARS-CoV-2 positive interval is strongly associated with post-clearance hospitalization status regardless of index infection severity.
Given these findings, we wondered whether therapeutic interventions to treat moderate or severe anemia prior to or during COVID-19 infection provided any protective effect against post clearance hospitalization. To answer this, we classified all patients with moderate or severe anemia on the basis of whether they received any anemia-targeted intervention during the respective time interval. Anemia-targeted interventions included vitamin or mineral supplementation (iron, vitamin B12, multivitamins), recombinant erythropoietin, and red blood cell transfusions. We found that patients who were administered one or more of these interventions were hospitalized after viral clearance at similar rates to patients who did not receive them in both the pre-COVID phase (5/7 [71%] vs. 2/2 [100%]; OR = 0.71; p = 1.0) and SARS-CoV-2 positive phase (26/52 [50%] vs. 7/17 [41%]; OR = 1.091; p = 0.59) (Figures 5F-G). It is important to note that both of these analyses (especially that of the pre-COVID phase) were limited due to the small total number of patients with moderate or severe anemia in these time periods (n = 9 for pre-COVID, 69 for SARS-CoV-2 positive).
Taken together, we conclude the following about COVID-19 patients who are hospitalized after viral clearance (PCH patients) compared to those who are not (PCNH patients): (1) they tend to have lower hemoglobin and hematocrit levels in the year prior to COVID-19 diagnosis and during active COVID-19 infection, (2) they are more likely to be diagnosed with moderate or severe anemia during both intervals; and (3) anemia-mitigating interventions in the hospital setting do not appear to reduce the risk of post clearance hospitalization.

Prior diagnosis of acute kidney injury is not associated with rehospitalization status in COVID-19 patients
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10. 1101/2020 Given that AKI is not defined by static lab measurements, but rather by changes in measurements (e.g. serum creatinine) over time, we next classified patients on the basis of whether they experienced laboratory-confirmed AKI during the pre-COVID and/or SARS-CoV-2 positive intervals. Specifically, we referred to the creatinine-related components of the KDIGO (Kidney Disease: Improving Global Outcomes) criteria for diagnosis and staging of AKI in adults: stage 1 AKI is characterized by an increase in serum creatinine by ≥0.3 mg/dL within 48 hours or an increase in serum creatinine to ≥1.5x the baseline value (known or assumed to have occurred within the last 7 days); stage 2 AKI is characterized by an increase in serum creatinine to ≥2x the baseline value; and stage 3 AKI is characterized by an increase in serum creatinine to ≥3x the baseline value or to ≥4 mg/dL ( Figure S9A).
We then used our longitudinal lab testing data to identify all instances meeting these criteria for each patient, and we compared the number of patients in the PCH and PCNH cohorts who did and did not experience AKI during the pre-COVID phase and during the SARS-CoV-2 positive phase. This analysis showed that similar fractions of patients in the two cohorts experienced any stage AKI during the pre-COVID phase (2/28 [7%] vs 6/53 [11%]; OR = 0.63; p = 0.71) ( Figure S9B). In the SARS-CoV-2 positive phase, the rate of AKI was slightly higher in the PCH cohort, but this finding was not statistically significant (15/38 [39%] vs. 23/92 [25%]; OR = 1.58; p = 0.14) ( Figure S9C).
Taken together, these results suggest that laboratory-based diagnosis of AKI in the year prior to COVID-19 diagnosis is not associated with post viral clearance hospitalization, while the diagnosis of stage 2+ AKI during the SARS-CoV-2 positive phase may be associated with subsequent hospitalization in male patients. However, the validity of this association must be further tested in larger patient cohorts.

Discussion
Almost one year after the first confirmed case, the COVID-19 pandemic continues to ravage communities across the globe. While efforts early in the pandemic rightly focused on the acute lung inflammation caused by SARS-CoV-2, the subsequent realization that COVID-19 may have more lasting effects has mandated a better understanding of factors that predispose patients to experience long-term COVID-19 related complications. We have previously sought to address . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint this knowledge gap using state-of-the-art NLP models deployed on a massive EHR system, and here we have expanded this effort to include the longitudinal analysis of laboratory measurements both prior to COVID-19 diagnosis and during active SARS-CoV-2 infection.
It is first important to note that this study has several limitations. First, this analysis considers only patients within one EHR system; while this system does contain patient data from multiple sites of clinical care in distinct geographic locations (Minnesota, Arizona, Florida), there are still likely underlying biases in important factors such as patient demographics and tendencies around the ordering of laboratory tests by clinicians. Such biases could prevent the studied cohort and their associated data points from serving as true representative samples of all COVID-19 patients. Second, as was previously mentioned, the analyzed cohort was relatively small, with even smaller patient counts available for particular lab tests of interest. Finally, we defined a SARS-CoV-2 positive window based on the presence of two negative RT-PCR tests, but the likely true date of clearance would precede the first negative test by an unknown amount of time.
Consistent with our previous conclusions, this lab test analysis suggests that both anemia and renal function in the pre-COVID and SARS-CoV-2 positive phases are associated with the risk of post viral clearance hospitalization. While the pathophysiologic basis for these associations are not yet clear, the findings do merit consideration in the context of clinical care of COVID-19 patients. Indeed, pre-existing conditions are already integrated in the clinical decision-making algorithms around COVID-19 as the Center for Diseases Control (CDC) has designated various chronic conditions as risk factors for severe COVID-19 infection (e.g. cancer, chronic kidney disease, chronic obstructive pulmonary disease, and cardiovascular diseases such as heart disease, obesity, and diabetes). However, there is much less known regarding factors or conditions that place people at risk for subsequent complications such as rehospitalization after viral clearance. Once identified, such factors and conditions should similarly be incorporated into the clinical decision-making process when treating COVID-19 patients.
Our finding that lower hemoglobin and hematocrit levels, and the outright diagnosis of moderate or severe anemia, prior to or during active SARS-CoV-2 infection is associated with post viral clearance hospitalization has not been previously reported. And while sickle cell disease is considered a risk factor for severe COVID-19, anemia itself is not considered to be such a risk factor. While we did not find evidence that administration of potential anemia-mitigating interventions was associated with lower risks of hospitalization after viral clearance, this does not rule out a role for anemia in long-term COVID-19 complications. It is important to note that we were not able to account for patient compliance in our analysis, and it is likely that patients who were moderately or severely anemic during their COVID-19 hospitalization were variably compliant with supplementation after their initial discharge. Further, even if mitigation of anemia does not impact subsequent hospitalization, these strong associations between anemia and COVID-19 are interesting in light of several previous lines of research.
First, several groups have reported an association between blood groups and susceptibility to and/or severity of ref,ref,ref), suggesting that individuals with type O blood may be at lower risk for contracting COVID-19 or experiencing respiratory failure in the context of COVID-19. Whether this association reflects a direct or indirect interaction between SARS-CoV-2 and erythrocytes is not known, but it could certainly be relevant to pursue . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint whether blood type is also associated with the occurrence or severity of anemia in the setting of COVID-19.
Second, fatigue has been commonly reported as both an acute symptom and a lasting effect of ref,ref), but the mechanisms underlying this phenotype have not been established. It is worth noting that 179 of the 229 (78%) COVID-19 patients in this study had at least a mild anemia during their SARS-CoV-2 positive phase, and 103 of the 229 (45%) patients had a moderate or severe anemia (defined as hemoglobin < 11 g/dL) during this interval. It would be worthwhile to perform a longitudinal follow-up on these patients to determine whether they continue to experience anemia in the months following SARS-CoV-2 clearance, and whether the presence of such a post-COVID anemia is associated with reports of fatigue.
Our findings regarding renal function tests and acute kidney injury may also be of clinical interest. Indeed, chronic kidney disease (CKD) has been recognized as a risk factor for severe COVID-19 infection. The fact that both median and extreme lab measurements suggest poorer renal function in the post clearance hospitalized cohort is consistent with this established risk factor, and suggests that the severity of one's CKD may have implications for their likelihood of hospitalization after viral clearance. Our finding that moderate/severe AKI during the SARS-CoV-2 positive phase was more frequently observed in the male PCH cohort compared to the male PCNH patients is interesting, but the small sample size available for analysis here would require further validation of this association.
Along with our previous analysis (ref), this study illustrates the value of deploying sophisticated platforms across EHR systems that enable the integrated analysis of diverse data types including sentiment-laden text and laboratory test measurements. Taken together, these studies provide the first example of leveraging augmented curation methods to first identify phenotypes that distinguish defined clinical cohorts and to then cross check these phenotypic associations through a hypothesis-driven analysis of the most relevant lab tests. This framework can be effectively scaled for other clinical research efforts not only in COVID-19 but also in any disease areas of interest.

Study design
This was a nested case-control study evaluating hospitalized and non-hospitalized postclearance patients within a cohort of 22,223 patients presenting to the Mayo Clinic Health System, comprised of tertiary medical centers in Minnesota, Arizona, and Florida with at least one positive SARS-CoV-2 PCR test between the start of the COVID-19 pandemic and October 27, 2020. This retrospective research was conducted with approval from the Mayo Clinic Institutional Review Board (IRB 20-003278).

Selection of study participants
Selection of cases and controls in this cohort are described in detail in our previous study (reference post-clearance paper). Briefly, hospitalized post-clearance cases (N=93) were defined as patients who had two documented negative SARS-CoV-2 PCR tests following their last positive test result and were subsequently admitted to the hospital within 90 days of clearance. A non-. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint hospitalized post-clearance control group (N=173) was defined as patients who had two documented negative SARS-CoV-2 PCR tests following their last positive test result and were not hospitalized within 90 days of clearance. As previously described, there were no statistically significant differences between these groups in age, relative cleared date (defined as time to second negative SARS-CoV-2 PCR test after first positive test), race, and ethnicity (reference post-clearance paper). A majority of hospitalized post-clearance cases were female as compared to non-hospitalized controls (61.3% vs. 41.0%, p <.01).

Exposures and outcomes measures
Laboratory results were assessed during the year prior to COVID-19 diagnosis, hitherto referred to as the pre-COVID-phase, and during the period in which SARS-CoV-2 PCR was positive, hitherto referred to as the SARS-CoV-2 positive phase. A diagnosis of COVID-19 was conferred by a positive SARS-CoV-2 RT-PCR test and clearance was defined as two consecutive negative SARS-CoV-2 RT-PCR tests occurring after a positive test.
Our primary outcome measurements were anemia and AKI. Measures related to anemia included hemoglobin and hematocrit, and measures related to acute kidney injury included serum creatinine, serum blood urea nitrogen (BUN), estimated glomerular filtration rate (eGFR), and stage of AKI. For each lab test, we considered the median, maximum, and minimum measurements during the specified time windows (i.e., one year prior to COVID-19 diagnosis, and the PCR-defined SARS-CoV-2 positive phase). Given the directionality of these tests, we were primarily interested in comparing the patient-level minimum values of hemoglobin, hematocrit, and eGFR, and maximum values of serum creatinine and BUN in each time period.
We also classified patients in a binary fashion for each time window based on whether their lab tests were consistent with the clinical diagnosis of anemia or acute kidney injury. Classifications were defined according to the Mayo Clinic reference ranges for anemia (ref, ref) and the KDIGO (Kidney Disease: Improving Global Outcomes) criteria (ref) for AKI (ref) as follows: • Pre-COVID anemia (mild, moderate, or severe): for males, median hemoglobin < 13.5 g/dL or median hematocrit < 38.3% during the one year prior to PCR confirmation of COVID-19 diagnosis. For females, median hemoglobin < 12.0 g/dL or median hematocrit < 35.5% during the one year prior to PCR confirmation of COVID-19 diagnosis.
• Pre-COVID anemia (moderate or severe): for both males and females, median hemoglobin < 10.0 g/dL during the one year prior to PCR confirmation of COVID-19 diagnosis.
• SARS-CoV-2 positive anemia (moderate or severe): for both males and females, minimum hemoglobin < 10.0 g/dL during the SARS-CoV-2 positive interval defined by PCR testing.
• Acute kidney injury (stage 1, 2, or 3): increase in serum creatinine by ≥0.3 mg/dL within 48 hours or an increase in serum creatinine to ≥1.5x the baseline value which is known or . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 assumed to have occurred in the prior 7 days. Here, the baseline was defined as the minimum value among all serum creatinine tests for the given patient in the prior 7 days.
• Acute kidney injury (stage 2 or 3): increase in serum creatinine to ≥2x the baseline value which is known or assumed to have occurred in the prior 7 days, or a serum creatinine value of ≥4 mg/dL. Again, the baseline was defined as the minimum value among all serum creatinine tests for the given patient in the prior 7 days.

Statistical Analysis
Laboratory values are reported as medians, minima, or maxima with interquartile range. Mann-Whitney U-tests were applied to continuous outcome measures, generating a p-value. Fisher exact tests were applied to categorical outcome measures, generating a p-value and an Odds Ratio. These tests were applied using the SciPy package 18 in Python (version 3.5). p-values were corrected using a Benjamini-Hochberg correction for multiple hypothesis testing.  Both median hemoglobin and median hematocrit are significantly lower in the hospitalized cohort than in the non-hospitalized one. Red shading indicates normal ranges for hemoglobin and hematocrit; as these ranges are lower for females than males, the shaded range here spans from the lower limit of normal for females (12 g/dL hemoglobin, 35.5% hematocrit) to the upper limit of normal for males (17.5 g/dL hemoglobin, 48.6% hematocrit).

Figure Legends
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020.

Figure 3. Comparison of minimum values for anemia-related tests (hemoglobin and hematocrit)
in the pre-COVID (A-B) and SARS-CoV-2 positive (C-D) intervals. In both intervals, the hospitalized cohort tends to have lower measurements of both hemoglobin and hematocrit. Red shading indicates normal ranges for hemoglobin and hematocrit; as these ranges are lower for females than males, the shaded range here spans from the lower limit of normal for females (12 g/dL hemoglobin, 35.5% hematocrit) to the upper limit of normal for males (17.5 g/dL hemoglobin, 48.6% hematocrit).  Comparison of anemia frequency in the hospitalized and non-hospitalized cohorts during the pre-COVID and SARS-CoV-2 positive phases. Contingency tables show the counts of patients with and without anemia in the hospitalized and non-hospitalized cohorts. Below each contingency table, the associated odds ratio and Fisher Exact test p-value is shown. (B-C) Mild, moderate, or severe anemia is defined as hemoglobin < 13.5 g/dL or hematocrit < 38.3% for males, and hemoglobin < 12.0 g/dL or hematocrit < 35.5% for females. (D-E) Moderate or severe anemia is defined as hemoglobin < 10 g/dL for both males and females. (F-G) Among patients with moderate or severe anemia (hemoglobin < 10 g/dL), comparison of the rates of administration of potential anemia-mitigating interventions between the hospitalized and non-hospitalized cohorts in the pre-COVID (F) and SARS-CoV-2 positive (G) intervals.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Figure 1
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Figure 5
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Non-Hospitalized
Post-Clearance  Table 1. Demographics and clinical characteristics of study cohorts, including patients who were and who were not hospitalized after PCR-confirmed clearance of SARS-CoV-2. The "Hospitalized post-clearance" cohort is composed of patients who were admitted or readmitted to the hospital following the estimated clearance date of SARS-CoV-2 infection. The "non-hospitalized postclearance" cohort is composed of patients who were not admitted to the hospital following the estimated clearance date. Each demographic variable or clinical characteristic was tested for difference in proportion with a Fisher Exact test or a difference in magnitude (for continuous variables) using a Mann-Whitney U test, and p-values were corrected for multiple testing using the Benjamini-Hochberg (BH) correction.

BH-adjusted p-value
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 Table 2. Analysis of median values for all selected lab tests in pre-COVID and SARS-CoV-2 positive phases, including both male and female patients. Table 3. Analysis of minimum values for all selected lab tests in pre-COVID and SARS-CoV-2 positive phases, including both male and female patients. Table 4. Analysis of maximum values for all selected lab tests in pre-COVID and SARS-CoV-2 positive phases, including both male and female patients.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 Table 5. Sex-split analysis of median values for all selected lab tests in pre-COVID and SARS-CoV-2 positive phases. Table 6. Sex-split analysis of minimum values for all selected lab tests in pre-COVID and SARS-CoV-2 positive phases.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 Supplementary Material Figure S1. Comparison of rates of admission to the intensive care unit (ICU) during index COVID-19 infection between the post-clearance hospitalized (PCH) and post-clearance non-hospitalized (PCNH) cohorts. (A) When considering all patients in the PCH (n = 93) and PCNH (n = 173) cohorts, the rates of ICU admission during index infection are similar. (B) When considering only the patients who were hospitalized during index infection (n = 49 for PCH cohort, 173 for PCNH cohort), the rate of ICU admission is higher for the PCH cohort.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S2. Histograms depicting the number of measurements per patient for the selected set of anemia-related lab tests in the pre-COVID (left) and SARS-CoV-2 positive (right) intervals.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S3. Histograms depicting the number of measurements per patient for the selected set of renal function lab tests in the pre-COVID (left) and SARS-CoV-2 positive (right) intervals.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S4. Median (baseline) values of (A, D) estimated glomerular filtration rate (eGFR), (B, E) blood urea nitrogen (BUN), and (C, F) serum creatinine during the Pre-COVID phase (A-C) and SARS-CoV-2-positive phase (D-F). In the SARS-CoV-2 positive phase, eGFR is lower and BUN tends to be higher in the hospitalized cohort. Shaded regions correspond to normal ranges for each test. For eGFR, the blue shading (60-90 mL/min/BSA) indicates moderately reduced levels which can be considered normal in older patients, while the green shading (>90 mL/min/BSA) indicates the normal range for younger patients. Normal ranges shown for BUN are serum creatinine are 7-20 mg/dL and 0.84-1.21 mg/dL, respectively.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S5. Comparison of "extreme" values for renal function tests in the pre-COVID (A-C) and SARS-CoV-2 positive (D-F) intervals. For each test, the extreme in the direction indicative of pathology is considered, i.e. minimum for eGFR versus maximum for BUN and creatinine. With the exception of the pre-COVID creatinine measurements, all other comparisons of these extreme values suggest lower renal function in the hospitalized cohort during both the pre-COVID and SARS-CoV-2 positive phases. Shaded regions correspond to normal ranges for each test. For eGFR, the blue shading (60-90 mL/min/BSA) indicates moderately reduced levels which can be considered normal in older patients, while the green shading (>90 mL/min/BSA) indicates the normal range for younger patients. Normal ranges shown for BUN are serum creatinine are 7-20 mg/dL and 0.84-1.21 mg/dL, respectively.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S6. Comparison of number of blood draws per patient during the SARS-CoV-2 positive interval between the hospitalized and non-hospitalized cohorts. Boxplots shown include only patients with at least one blood test performed during this interval. Data is shown for (A) all patients, (B) male patients only, and (C) female patients only.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S7. Comparison of hemoglobin levels in patients based on ICU admission status during COVID-19 index infection. Each dot corresponds to the median ("baseline") hemoglobin level of an individual patient during the SARS-CoV-2 positive interval. All 229 patients comprising the post-clearance hospitalized and post-clearance non-hospitalized cohorts were divided based on whether they were admitted to the ICU during their index infection. Red shading indicates normal ranges for hemoglobin and hematocrit; as these ranges are lower for females than males, the shaded range here spans from the lower limit of normal for females (12 g/dL hemoglobin, 35.5% hematocrit) to the upper limit of normal for males (17.5 g/dL hemoglobin, 48.6% hematocrit). Green shading indicates mild anemia, defined as a hemoglobin level greater than 10 g/dL and less than the sex-dependent lower limit of normal.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S8. Comparison of rates of moderate and severe anemia in the post-clearance hospitalized versus non-hospitalized patients who were (A) admitted to the ICU during their index infection or (B) not admitted to the ICU during their index infection. Regardless of ICU admission status, patients who experienced moderate or severe anemia during the SARS-CoV-2 positive phase were more likely to be hospitalized after viral clearance.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.02.20242958 doi: medRxiv preprint Figure S9. Comparison of acute kidney injury (AKI) frequency in the hospitalized and nonhospitalized cohorts during the pre-COVID and SARS-CoV-2 positive phases. (A) Schematic illustrating how patients are classified as having no AKI, stage 1 AKI, or stage 2/3 AKI. Thresholds are derived from the KDIGO (Kidney Disease: Improving Global Outcomes) criteria for the diagnosis and staging of AKI. (B-G) Contingency tables showing the counts of patients with and without acute kidney injury (AKI) in the hospitalized and non-hospitalized cohorts. Below each contingency table, the associated odds ratio and Fisher Exact test p-value is shown. (B-C) Evaluation of the association between Stage 1, 2, or 3 AKI and post viral clearance hospitalization status in male and female patients. (D-E) Evaluation of the association between Stage 2 or 3 AKI and post viral clearance hospitalization status in male and female patients. (F-G) Evaluation of the association between Stage 2 or 3 AKI and post viral clearance hospitalization status in only male patients.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10. 1101/2020