COVID-19 Symptoms and Deaths Among Healthcare Workers, United States

Shao Lin; Xinlei Deng; Ian Ryan; Kai Zhang; Wangjian Zhang; Ese Oghaghare; DeeDee Bennett Gayle; Benjamin Shaw


Emerging Infectious Diseases. 2022;28(8):1624-1641. 

In This Article


We found that HCWs who died of COVID-19 in the United States were disproportionately older (≥50 years of age), male, and Black or Asian. Consistent with our findings, previous research found that older age groups are more vulnerable to COVID-19 infection and death, likely because of their lower immunity against viral infections and multiple preexisting medical conditions, both known to exacerbate COVID-19–related deaths.[5,20,21] Of interest, we found that the highest risk for death among HCWs occurs in a relatively younger group (50–59 years of age) than the general population of hospitalized patients (≥65 years of age). A possible explanation is that HCWs are generally a younger working population compared with retirees in the general population who suffer a higher COVID-19 burden. Unfortunately, there was no available literature in this area to confirm our findings.

Although this study found that female deaths were higher among HCWs than non-HCWs, female deaths were significantly lower than for HCW-controls, implying no significant increase in deaths among female HCWs after controlling for occupation. This finding could be attributable to the confounding factor of HCWs composition; that is, women are more likely to work in healthcare occupations.[22] Therefore, using non-HCWs as a control group may lead to biases due to occupational confounders. On the other hand, worldwide, men were more likely to be infected by COVID-19 and have severe symptoms than were women, which is consistent with our findings of a 2-fold increase in death risk for male HCWs compared with male HCW nondeaths. Furthermore, several studies suggest that sex differences in the susceptibility to COVID-19 may be because of differences in immune response.[23,24] These studies found higher plasma levels of innate immune cytokines, such as interleukin 8 and 18, among male patients but more robust T-cell activation among female patients infected by COVID-19.

Our findings regarding the increased risk for COVID-19 death among ethnic minority populations (Black and Asian) agree with several studies on the general population. For instance, Rogers et al.[12] and Kirby[15] reported an increased risk for COVID-19 deaths among non-Hispanic Black and Hispanic minority populations.[13,15] In addition, two thirds of HCWs in the United Kingdom who died of COVID-19 identified as an ethnic minority.[25] Another study found that COVID-19 infection and death were strongly linked with overcrowded neighborhoods, higher body mass index, and low incomes, all categories that are overrepresented in Hispanic and Black communities.[26] However, we did not detect an increased risk for COVID-19 deaths among Hispanic HCWs, which could be caused by the small sample size (n = 271) or missing ethnicity information.

Our study also reported a significantly higher risk for COVID-19 deaths (3- to 5-fold) among Asian HCWs than in all 3 control groups. Sze et al.[27] reviewed 50 studies, including 18,728,893 COVID patients from the United States and the United Kingdom, and found that Black and Asian persons were at an increased risk for COVID-19 infection compared with White persons. Pooled adjusted OR for Black persons was 2.02 (95% CI 1.67–2.44), and for Asian persons, 1.50 (95% CI 1.24–1.83).[27] Nevertheless, few studies have reported that Asian persons are also at higher risk. Coronary heart disease, a high-risk comorbidity of COVID-19 death, is more common among Black persons, Asian persons, and persons of other ethnic minorities.[28] Overcrowded households may be another risk factor associated with the spread of COVID-19 among Asian persons.[29] Our results contradicted the belief we noted among Asian populations that they are at a lower risk of contracting COVID-19 because of genetic protection. Therefore, these findings are of critical public health importance in terms of communicating accurate COVID-19 information to particularly at-risk populations.

We found that almost one third of the HCW death cases in the United States occurred in June, and >40% occurred in the summer months (June–August) of 2020. The initial surge in COVID-19 cases led to a profound increase in HCWs' exposure to the virus. This surge is likely the result of inadequate intensive care units and hospital beds, insufficient PPE supply, inadequate training and experience among HCWs, and heavy workloads due to a large and rapid influx of patients. Another study reported that HCWs with inadequate access to PPE had increased SARS-CoV-2 infection compared with those with adequate PPE access.[30]

We found that the fatality rate in the US general population (2.48%) was more than 7-fold higher than that among HCWs (0.33%). This finding is consistent with that of Sahu et al.,[9] who also found that the mortality rate among HCWs was 7 times lower than that among all cases.[10] However, almost all severe indicators and symptoms were higher among HCW deaths than in the 3 control groups in this study, which may be because of HCWs' proximity to and longer duration of exposure to COVID-19 patients. Alternatively, HCWs may have better access to healthcare and treatment, which prevented deaths in this population. Unfortunately, there is a paucity of literature regarding the severity indicators for COVID-19 to compare with our study.

We found that underlying conditions were the most important predictor of COVID-19 deaths among HCWs, followed by shortness of breath, fever, cough, nausea/vomiting, and diarrhea. Multiple studies found that chronic conditions were the most critical COVID-19 severity and death indicators in different countries.[6,7] Other studies have shown that shortness of breath is another important symptom of COVID infection.[31,32] Our findings that fever >100.4°F, even subjective fever (felt feverish), was consistently more common for HCW deaths than for the 3 control groups, indicating that fever may be an early indicator of disease severity. Another finding is that HCW death cases reported significantly higher gastrointestinal symptoms (diarrhea, nausea, vomiting, abdominal pain) than all 3 control groups. Consistent with our findings, Wiersinga et al.[31] reported that initial COVID-19 symptoms might include shortness of breath, fever, cough, nausea/vomiting, or diarrhea. The general public expects respiratory symptoms but may be unaware of gastrointestinal symptoms related to COVID-19. Therefore, this study may provide valuable insight for public education and severity prediction.

We also found that cough is among the top 3 reported symptoms of COVID-19 infections and deaths in HCWs and the general population (62.77%–78.15%). However, although headache (59.68%–66.47%) and myalgia (51.95%–59.10%) were the other 2 top symptoms among COVID-19 infection cases, preexisting conditions (88.31%–93.01%) and shortness of breath (66.5%–70%) were the commonly reported symptoms for COVID-19 deaths. In addition, we found that runny nose, sore throat, and headache symptoms were notably lower in HCW deaths than in nondeath controls, implying that these symptoms may not be essential predictors of COVID-19 death. Unfortunately, we found no relevant literature with which to compare our results.

Temporal patterns of infection and death in our study showed that, whereas COVID-19 deaths in the general US population experienced 3 distinct peaks, deaths among HCWs only peaked during the first surge. Deaths among HCWs went down after April 2020 and remained low. This finding is consistent with CDC reports and other studies that show a large initial surge and subsequent decline. The similar peaks in both HCWs and non-HCWs illustrate how quickly the COVID-19 pandemic spread from the general population to HCWs who took care of the deadliest cases.[2,22,33] The flatter COVID-19 death trend after the first surge among HCWs could be attributed to their early and high immunization rate, improved PPE, access to healthcare facilities, and early detection and treatment for HCWs compared with the general population.

A strength of our study is that we used multiple reference groups to minimize different biases. We further validated our findings using HCW controls with similar exposure opportunities and socioeconomic backgrounds. In addition, using 2 general-population reference groups helped us examine how demographics and symptoms differed between HCWs and the general population. We also used dynamic, nationwide CDC surveillance data. These objective data reduce reporting bias, which is typically a major concern in studies relying on media reports or questionnaires. Finally, we controlled for several known risk factors for COVID-19, including demographics, different dominant SARS-CoV-2 variants, and the start of vaccinations.

Our findings illustrate how timely CDC surveillance data, reported every 2 weeks, can be used to monitor the temporal trend of infections and deaths among different populations. In addition, unique findings, such as HCW deaths increasing in younger age groups and in Asian persons, could be used in targeted interventions. Furthermore, clinical agencies could use the severe clinical indicators and symptoms we identified to predict deaths and plan hospital beds.

A limitation of our study is that, because the COVID-19 case surveillance system is passive, our data may underestimate the number of cases, although reporting cases to the CDC is federally mandated. In addition, the availability of diagnostic testing, resources, and the priorities of health officials may influence the completeness of reporting. To address this issue, we included total cases (both laboratory-confirmed and probable cases); most COVID-19 deaths (>99%) were laboratory confirmed. Although the case report form captures severity indicators, these data may be inaccurate or underreported because some outcomes were unknown at the time of reporting. We repeated the analyses multiple times using different lengths of the cohort after the initial analysis by adding updated data, and the results are robust. Missing values, especially demographic information for death cases in the general population, are also important limitations, suggesting that control 2 would not be a good control group. Furthermore, we examined race and ethnicity variables separately and could not combine these 2 variables because of a lack of personal identifiers and the arbitrary coding in the data; only 1 variable (race or ethnicity) was coded. We used race/ethnicity as a surrogate for sociodemographic status; however, household income and deprivation indices, which correlate with COVID infection, were not available. Although multiple testing may cause false positives, our results are robust after Bonferroni correction, a conservative test used to protect from type I error.

Another challenge we faced was determining whether a reported death was caused by COVID-19 or other diseases. We believe that COVID-19 was the primary cause of death for the cases we used because the death-related question from the CDC questionnaire is specific to COVID-19 (i.e., "Did the patient die as a result of this illness [COVID-19]?"). In addition, similar to other infectious diseases under mandatory reporting, all COVID-19 cases and causes of death were confirmed and validated by hospitals or health departments. Finally, we could not separate unique infections from reinfections because no personal identifier is available. However, this problem may not have a substantial effect for several reasons. We aimed to examine fatality (deaths per infection) rather than infection. Although death was almost certainly reported only once, infections could be reported multiple times because of reinfection. Therefore, we may have underestimated fatalities because reinfections increased the denominator. This underestimation bias is likely similar between the HCW deaths and HCW controls because of identical occupations and similar sociodemographic status, which would be nondifferential and toward the null. When comparing HCW deaths to non-HCW controls, the denominator, including reinfection numbers, is likely larger among HCWs than the general population because of HCWs' frequency and duration of exposure to patients. Therefore, the fatality ratio of these 2 groups may have been underestimated. Finally, the effect of such bias may be minimal because most COVID-19 reinfections occurred when the Omicron variant was dominant (86.9%) after December 31, 2021,[34] which occurred after our study period (January 1, 2020–October 12, 2021). For instance, the COVID-19 reinfection rate in New York through December 31, 2021, was 0.56%, but as of January 1, 2022, it was 3.74%.[34]

In conclusion we found that HCWs who were ≥50 years of age, male, Black, or Asian experienced higher deaths from COVID-19. In addition, HCW COVID-19 patients experienced fewer deaths but significantly higher risks for the most severe indicators than the 3 reference groups. We also found that underlying conditions, shortness of breath, fever, cough, nausea/vomiting, and diarrhea were the most relevant indicators for COVID-19–related deaths among HCWs. Conversely, runny nose, sore throat, and headache may not be critical indicators for COVID-19 death in this population.