Background

It has been well-established that older age, male sex, and the presence of comorbidities are associated with higher mortality risks following diagnosis with the coronavirus disease of 2019 (COVID-19) [1,2,3,4,5,6,7,8,9,10,11,12]. Given COVID-19 incidence and mortality, and the precarious supply of COVID-19 vaccines globally [13,14,15,16,17,18], prioritization schemes will continue to be needed to triage vaccine delivery in early stages to people at the highest risk of death. When COVID-19 vaccines were first introduced in the United States, the Centre for Disease Control (CDC) Vaccine Rollout Recommendations [19] placed people aged 75 years and older in the Phase 1b stage, while Phase 1c extended eligibility to people aged 65–74 years and younger individuals with “underlying medical conditions which increase the risk of serious, life-threatening complications from COVID-19”.

However, this CDC approach treats “adults of any age” as being at increased risk of severe illness [20]. Thus, it does not account for differential risk between conditions, sex, or the multiplicative impact of older age on adverse outcomes among people with underlying medical conditions [21,22,23,24,25,26]. Other investigators have developed sophisticated risk prediction algorithms for mortality COVID-19 diagnosis [9, 27,28,29,30,31,32], but these do not lend themselves to simple implementation on a large scale by jurisdictions for vaccine prioritization. Furthermore, vaccine hesitancy remains an important stumbling block for vaccination in jurisdictions with adequate vaccine supplies. Unfortunately, the sociodemographic risk factors for chronic disease in the young overlap substantially with predictors of vaccine hesitancy [33,34,35,36,37,38,39,40]. Communication of risk for younger individuals can be hindered by lower absolute event rates, while relative risks can be harder to appreciate [41,42,43,44,45]. This has galvanized the development of alternate approaches for communication of risk for preventative intervention in younger patients [46,47,48,49].

Given these shortcomings, it would be desirable to provide relatively simple means to communicate how the risk of dying after being diagnosed with COVID-19 varies by age and sex for unvaccinated people living with comorbidity. Accordingly, we conducted a population-based cohort study of community-dwelling adults who developed COVID-19 before the availability of vaccines to quantify the incremental risk for death associated with underlying medical conditions as a function of age and sex. We specifically focused on cardiovascular disease, diabetes, hypertension, cancer, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD) and frailty. We hypothesized that chronic medical conditions could elevate the risk of some, but not all, younger individuals to equal that of individuals aged > 65 years.

Methods

Study design and population

Residents of Ontario (Canada’s most populous province) receive universal coverage for essential physician services and hospital-based care through the Ontario Health Insurance Plan (OHIP). This facilitates the conduct of population-based cohort studies using administrative health datasets that are linked using unique encoded identifiers and are analyzed at ICES (formerly Institute for Clinical Evaluative Sciences). Multiple algorithms have been validated to ascertain medical diagnoses using these administrative databases [50]. ICES is an independent, non-profit research institute funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC), and a prescribed entity under Ontario’s Personal Health Information Protection Act (PHIPA), Sect. 45 of PHIPA. As such, the use of the data in this project is authorized under Sect. 45, approved by ICES’ Privacy and Legal Office, exempt from Research Ethics Board review, and does not require patient consent. All methods were carried out in accordance with locally relevant guidelines and regulations.

The Ontario Laboratories Information System (OLIS) was used to identify individuals aged ≥ 18 years with a positive reverse-transcription SARS-CoV-2 polymerase chain reaction (RT-PCR) test in Ontario between January 1 and October 31st, 2020, prior to the availability of vaccines. For people with more than one positive test, we retained the first positive test. The index date was that of collection of the qualifying SARS-CoV-2 swab. We excluded people with missing/invalid key data (age, sex, OHIP number), non-Ontario residents, OHIP coverage < 1 year before the SARS-CoV-2 test (for ascertainment of medical history), or an index positive SARS-CoV-2 test that was collected on a date when the individual was hospitalized (to limit our study to outpatients). We also excluded 5740 long-term care (LTC) residents (minimum age 20 years; maximum age 107 years) since they are already prioritized in the highest risk category globally (e.g., Phase 1a of the CDC framework). The remaining patients constituted our cohort of community-dwelling outpatients with COVID-19.

Our primary exposure was age. We also studied underlying medical conditions that have been shown to increase mortality risk in COVID-19 [1,2,3,4,5,6,7,8,9,10,11,12], affect a substantial proportion of the population, and are objective enough to be implemented by governments in vaccine prioritization policies: (1) major cardiovascular disease, defined as a recent (in past 5 years) myocardial infarction [51] or lifetime history of heart failure [52]; (2) diabetes [53]; (3) hypertension [54]; (4) cancer diagnosed within 5 years [55]; (5) COPD [56]; and (6) Stages 4/ 5 CKD (defined as dialysis-dependence or estimated glomerular filtration rate (eGFR) < 30ml/min/m2) [57]. For the analysis of CKD, we excluded individuals who were not dialysis-dependent and did not have creatinine measurements in the 2 years before contracting COVID-19. We also studied the Johns Hopkins ACG System binary frailty indicator [58] as a marker of global illness that can be applied widely across multiple health systems.

Outcome

The primary outcome was death within 30 days of the positive SARS-Cov-2 test; this follow-up period after a positive PCR test is expected to capture most deaths directly attributable to COVID-19 [59, 60] while decreasing the likelihood of incorporating deaths due to other illnesses (which may be more likely for older patients with comorbidities).

Statistical analysis

We fit univariable logistic regression models using Firth’s penalized likelihood approach to address potential bias in parameter estimates due to small sample and outcomes sizes in some age/condition strata [61, 62]. In these models, age was the only predictor, and modelled using a restricted cubic spline with knots at the 5th, 35th, 65th, and 95th percentiles, as suggested by Harrell to account for the non-linear relationship between age and our outcome [63] . Since separate curves are fit to each segment (i.e., range of ages), the model better reflects the relationship between age and death. The fitted model was used to estimate the risk of death within 30 days for 75- and 65-year-old individuals. These two probabilities were used as benchmarks against which to compare other subjects, since these are the age cut-offs used in Phases 1b and 1c respectively of the CDC’s COVID-19 Vaccine Rollout Recommendations. We then repeated the same logistic regression analyses, this time stratifying the cohort by sex and presence of the underlying medical conditions described above [62]. For each sex/comorbidity stratum, we determined the predicted risk of death within 30 days at all ages and identified the age at which individuals with the medical condition exceed the predicted benchmarks risk in the general population at age 65 years and 75 years. The age at which the benchmark risks were crossed were rounded up to the next integer for ease of presentation. All analyses were performed using SAS Enterprise Guide 7.1 (SAS Institute Inc., Cary, NC).

Results

We studied 52,429 community-dwelling individuals who tested positive for SARS-CoV-2 between January 1 and October 31st, 2020 (Tables 1and Supplemental Fig. 1). Median age of our study population was 42 years [minimum, Q1, Q3, maximum 18, 29, 56, 104] years with 5,962 and 2,596 individuals 65- and 75 years and older, respectively; and 27,535 [52.5%] were women. A total of 1,185 individuals (2.3%) had major cardiovascular disease, 7,336 (14.0%) had been diagnosed with diabetes, 12,275 (23.4%) with hypertension, 966 (1.8%) with recent cancer, and 2,466 (4.7%) with COPD. Of 34,724 people (66.2% of cohort) whose CKD status could be ascertained, 350 (0.7%) were classified as having advanced CKD, of whom 146 (0.3%) were dialysis dependent. Using the Johns Hopkins indicator, 1,794 individuals (3.4%) were classified as frail.

Table 1 Baseline characteristics of study population by sex
Full size table

Within 30 days following their positive SARS-CoV-2 test, 417 (0.8%) people died, with the greatest death rates for both men and women among those with hypertension (78,7% and 85.9%, respectively) or classified as frail (45.8% and 60.4%, respectively) (Table 2). The predicted risk of death in the general population was 1.1% at age 65 years and 3.4% at 75 years. The estimated 30-day mortality risk increased with age (1.7% among 65–74 year olds, and 11.5% among those 75 years and older), and was generally higher in men (Fig. 1). After stratifying by presence of underlying medical conditions, the estimated risk of death was generally higher in those with a comorbidity.

Table 2 Baseline characteristics of study population relative to sex and mortality status at 30 days following the qualifying positive SARS-CoV-2 test
Full size table
Fig. 1
figure 1

Predicted risk of 30-day death by age and sex. The shaded region indicates 95% confidence intervals. The blue curves depict estimated risk in men, while the red curves depict risk in women

Full size image

Figures 2 and 3 illustrate the differences in risk of death at 30 days for men and women with and without underlying medical conditions. Overall, both men and women with a medical condition had higher 30-day death rates than those without, though a few exceptions were observed (i.e., among middle-aged men with versus without COPD and older women with versus without diabetes).

Fig. 2
figure 2

Predicted risk of 30-day mortality for men by presence of underlying medical conditions. Horizontal lines indicate the predicted risk of death in the general population at age 65 years and 75 years. Vertical lines highlight the age at which the risk for men with the medical conditions is equivalent to the general population aged 65- or 75- years

Full size image
Fig. 3
figure 3

Predicted risk of 30-day mortality for women by presence of underlying medical conditions. Horizontal lines indicate the predicted risk of death in the general population at ages 65 years and 75 years. Vertical lines highlight the age at which the risk for women with the medical conditions is equivalent to the general population aged 65- or 75- years

Full size image

Table 3 summarizes the age (rounded up to the next integer) at which men and women with a particular comorbidity had the same predicted risk of death at 30-days as the general population aged 65 years and 75 years of age. The benchmark risks were exceeded at the earliest age by people with CKD, cancer, and frailty. For example, the risk of death at 30 days in men with CKD at age < 40 years and women with CKD at age 45 years equalled the risk of 75-year-olds from the general population. In contrast, the risk of death at 30 days for women aged < 65 years who had diabetes or hypertension was not higher than the 30-day mortality risk of 65-year-olds in the general population.

Table 3 Age at which the risk of death in men and women with an underlying medical condition exceeds the risk in the general population aged 65- and 75-years of age. The presented age has been rounded up to the next integer
Full size table

Discussion

In this population-based cohort study, we determined the estimated age at which a community-dwelling man or woman with underlying medical conditions will exceed the 30-day mortality risk of the typical person included in Phase 1b or 1c of the CDC recommendations based on age alone. While the risk of death after COVID-19 was higher in people with underlying medical conditions, the prognostic implications varied by sex and condition. The risk of 30-day death was generally higher in men than in women. The increase in risk incurred by the presence of CKD and recent cancer was higher than isolated diabetes or hypertension. Thus, it would be inappropriate to treat hypertension and diabetes as being equivalent to the other comorbidities studied when triaging vaccine rollout. The Johns Hopkins ACG System binary frailty indicator was a useful composite measure which can be applied by large health systems using administrative data for identification of individuals at higher risk for death with COVID-19 despite younger age.

Numerous publications have demonstrated that individuals with comorbidities are at higher risk for adverse outcomes following COVID-19 infection [1,2,3,4,5,6,7,8,9,10,11,12]. The comorbidities studied in our analysis have been among the most studied and most consistently linked to higher mortality with COVID-19 [1, 2, 6, 11]. In meta-analyses of studies prior to vaccine availability, reported risk estimates have ranged from 3.07 to 4.90 for CKD, 1.47 to 1.90 for cancer, and 2.25–3.05 for cardiovascular disease [1, 6, 11]. Male sex has also been consistently shown to increase the risk of adverse outcomes following COVID-19 infection [5, 8, 9], which may be related, in part, to the X-linked nature of the SARS-CoV-2 receptor [64]. The combination of multiple comorbidities increase risk even further, and this has been utilized to develop comprehensive models which predict the risk of death following COVID-19 infection with high accuracy [9, 27,28,29,30,31,32]. In a study of Veteran Affairs data in the United States, a model of nine risk factors including age, sex, diabetes, CKD and heart failure demonstrated a discriminative ability of 83.4% compared to 74.0% in a model using age alone, as in the CDC approach [65]. Furthermore, once population vaccination rates reach 50%, vaccine prioritization based on the model was estimated to result in 21.5% fewer deaths than prioritization on the CDC phased approach [65].

However, while these studies have provided valuable information about factors associated with increased COVID-19 mortality, we are not aware of studies reporting their age-equivalent mortality risk. Additionally, the inferences from these studies about the impact of comorbidity and sex on the risk of dying from COVID-19 at different ages are challenging to communicate to the public in a transparent and easily understood manner [41,42,43]. The communication gap is expected to be largest for demographic groups that are at higher risk for COVID-19 and more susceptible to misinformation [66,67,68]. The profound impact of some comorbidities on mortality is important to communicate to younger individuals with vaccine hesitancy, which remains an important issue among people with comorbidity, particularly those living in communities in which medical comorbidity is more likely to emerge at a younger age [33,34,35,36,37,38,39,40]. Provision of visual cues and expressing risk in terms of relative age (e.g., “heart age”) improves communication of cardiovascular risk for younger individuals and is more likely to promote behaviour change than traditional methods of communicating cardiovascular risk [42,43,44,45,46,47,48,49, 69]. We present our data in an analogous approach, which we believe can be helpful for vaccine-hesitant individuals with comorbidities that confer higher risk of adverse outcomes following COVID-19.

Importantly, we showed that regardless of underlying medical condition, the mortality risk rises substantially with age, meaning that age should continue to be a key factor in the triage process. The presence of most underlying medical conditions in individuals under the age of 45 years did not elevate their risk to that of the general population at age 65 years, with the notable exception of Stage 4–5 CKD, and among men, recent cancer and frailty. In other words, the protective effects of younger age persisted despite the presence of underlying medical conditions for most people aged < 45 years.

Several limitations to our study are noted. By design, we adopted an approach prioritizing parsimony and simplicity by focusing on the presence/ absence of comorbidity without accounting for the severity of disease. Although 12.8% of our study population had more than one of the conditions studied, we also did not study the combined impact of multiple comorbidities outside the frailty indicator for the same reasons. As a result, in our comparisons of risk among men and women with specific conditions versus without, it is possible that some differences may be attenuated because those without the condition have other conditions which also increase their risk. Another limitation was the small event counts for sex-specific strata of patients with cancer and CKD, which decreases precision of our estimates for these groups. Our results also do not apply to LTC residents, who should already be considered at highest risk regardless of age or comorbidities. Finally, since our study population was identified during the first eight months of the pandemic and prior to the appearance of variants of concern, such as Delta and Omicron, the absolute risk estimates by age may not be applicable to patients diagnosed with these newer variants; however, the higher mortality risks associated with CKD, cancer and frailty compared with hypertension and diabetes are likely applicable beyond the first wave though the relative risk associated with variant types differs.[70, 71] Additionally, the relative ages reported from our analyses remain valuable for communication of relative risks.

Conclusion

The mortality risk in COVID-19 increases with age and comorbidity but the prognostic implications varied by sex and condition. The risk was generally higher for men, and the increase in risk associated with CKD, cancer, and frailty was higher than what was observed with hypertension and diabetes. We hope to improve communication by presenting the combined impact of patient age, sex, and comorbidity using visual scales and reporting the “equivalent age” at which individuals would have the same mortality risk as the average 65- and 75-year-old with COVID-19. We believe these observations can inform vaccine rollout and support communication efforts in jurisdictions with limited vaccine supplies.