Association of COVID-19 with risks of all-cause and cause-specific mortality post-infection: A UK Biobank cohort study

Abstract

Background

While SARS-CoV-2 causes multi-organ complications, comprehensive assessments of long-term mortality across organ systems remain limited. This study systematically evaluated COVID-19's impact on all-cause and cause-specific mortality.

Methods

This cohort study followed 467,522 UK Biobank participants (Jan2020-Dec2022). COVID-19 exposure (representing clinically apparent infection) was classified as overall, hospitalized, and non-hospitalized, and compared with reference cohort without documented SARS-CoV-2 records. Post-acute mortality (>30 days post-infection) was assessed using landmark analyses; overall-mortality (including acute-phase deaths) was evaluated as secondary analyses and presented in Supplementary Materials. Adjusted Cox models estimated risks for 12 organ systems and 47 diseases, with subgroup analyses by key comorbidities and demographics.

Results

Post-acute all-cause mortality was elevated in overall (hazard ratio [HR]: 1.50) and hospitalized (HR: 3.06) COVID-19 cohorts, but not non-hospitalized group. COVID-19 infection increased post-acute mortality from circulatory, digestive, genitourinary, neurological, respiratory, and external causes, as well as neoplasms (though elevated cancer mortality without prior diagnoses may reflect detection bias). Hospitalized cases showed elevated risks across 11/11 organ systems and 27/37 diseases; non-hospitalized cases showed increased risks for external-cause and neurological outcomes. Advanced age, atrial fibrillation, chronic kidney disease, and hypertension exacerbated post-acute all-cause mortality; atrial fibrillation also amplified respiratory and neurological risks.

Conclusion

Clinically apparent COVID-19 was associated with elevated post-acute mortality across multiple systems, with hospitalized cases exhibiting the broadest risk spectrum. As controls may include unrecorded infections, these estimates are likely conservative. Sustained monitoring is warranted, particularly for older survivors and those with high-risk comorbidities.

Highlights

• •We tracked long-term post-COVID mortality across all systems in a UK Biobank cohort.
• •Hospitalization-requiring COVID-19 raises post-infection multi-organ mortality.
• •We identified elevated novel mortality risks from digestive and genitourinary causes.
• •Our findings suggest a critical need for sustained monitoring of COVID-19 survivors.

1. Introduction

The COVID-19 pandemic, caused by coronavirus SARS-CoV-2, has posed an unprecedented challenge to global health. By 5-Jan-2025, the World Health Organization reported over 777 million confirmed-cases and 7 million deaths worldwide attributed to COVID-19 [1]. Emerging evidence indicates that SARS-CoV-2 infection can trigger complications in respiratory, cardiovascular, neurological, renal, gastrointestinal, and other systems [2]. While the increased morbidity/mortality during the acute phase of COVID-19 are well-characterized [3], with recent data suggesting acute mortality still causing a huge burden in early-2025 [4], the post-acute risks across various organs remain incompletely understood. Evaluating the long-term impacts of COVID-19 sequelae is essential for guiding patient care strategies and healthcare resources allocation.

Previous research on post-COVID-19 disorders mainly focused on the prevalence of persistent symptoms and hospitalizations. It was reported that COVID-19, especially hospitalization-required case, confers a greater risk of downstream symptoms/hospitalizations compared to COVID-19-negative individuals [5,6]. In parallel, population-level analyses documented substantial excess pandemic mortality based on historical trends [7,8]. However, by relying on aggregated data, these excess-death estimates cannot fully adjust for individual-level confounders nor accurately attribute deaths to SARS-CoV-2 infection. Although one work [5] quantified mortality risks in COVID-19 patients, they primarily focused on overall mortality rather than specific death causes. While one study explored post-COVID-19 mortality for respiratory and cardiovascular outcomes [9], the associations between COVID-19 and mortality spanning a broader spectrum of sequelae remain unclear. Therefore, a systematic and comprehensive analysis of the impact of COVID-19 on mortality across all body systems is urgently needed.

To address this gap, we conducted a longitudinal analysis by following a prospective UK Biobank (UKBB) cohort from 31-Jan-2020 to 19-Dec-2022. We leveraged Clinical-Classifications-Software-Refined (CCSR) to facilitate an inclusive grouping of ICD-10 diagnoses, thereby classifying causes of post-COVID-19 deaths into 12 composite organ-system and 47 single-disease outcomes. We primarily estimated the associations between COVID-19 (overall, hospitalized, and non-hospitalized) and post-acute mortality (>30 days post-infection) due to these disorders. Overall-mortality analyses (all mortalities post-infection, including acute and post-acute phases) were additionally conducted as secondary analyses and are presented in Supplementary Materials for reference. For each single outcome, deaths were further stratified by prior history of that condition. Additionally, we performed subgroup analyses stratified by various characteristics to explore potential differences in the mortality risks of different sequelae across subgroups.

This study makes several novel contributions to the understanding of post-COVID-19 mortality. To our knowledge, it is the first to comprehensively examine post-acute mortality across a wide range of diseases spanning all organ systems, identifying novel associations between COVID-19 and elevated risk of death from digestive, genitourinary, and external causes. Furthermore, our work is the first to stratify COVID-19 mortality by patients’ clinical characteristics, revealing how pre-existing conditions amplify specific risks. For example, we found that pre-existing atrial fibrillation significantly increases post-acute mortality from neurological and respiratory diseases. These associations were not previously reported.

2. Methods

2.1. Study design and setting

This prospective cohort study used data from UKBB (project #28732), which tracks the electronic health records of ∼500,000 participants aged 50-87 years [10]. To evaluate the hazard ratios (HRs) for post-COVID-19 deaths, UKBB sample was followed from 31-Jan-2020 (date of the first UK-confirmed COVID-19 case) to 19-Dec-2022 (last date for available records during analysis). Totally 467,522 individuals alive at the start of follow-up were included for subsequent analysis.

2.2. Data sources, outcomes and covariates

Participants’ baseline characteristics, comorbidities, SARS-CoV-2 infection status, and mortality records were ascertained from UKBB-linked electronic health records and national death registries. Our primary study outcomes were post-acute (>30 days after SARS-CoV-2 infection) all-cause and cause-specific mortality. We focused solely on the primary cause of death to avoid confounding from secondary conditions. The ICD-10-coded causes of mortality were categorized using CCSR [11]. Overall mortality across the entire follow-up (including days 0-30) was also assessed as supplementary analysis using the same outcome definitions. By restricting outcomes with ≥5 deaths in both COVID-19 exposure and reference cohorts, up to 12 composite (organ-system) outcomes and 47 single (CCSR-defined-disease) categories were included (Table_S1). For each single outcome, we also performed analysis stratified by whether patients had a pre-existing diagnosis of the same condition. To adjust for baseline differences between cohorts, we selected a set of covariates [6] including sociodemographic, lifestyle, biochemical and clinical factors, as well as pre-existing comorbidities and vaccination status (see Supplementary_methods for full list). Imputation of missing values was conducted by missRanger (v2.6.1) [12].

2.3. Cohort

In our primary analysis, the COVID-19 exposure cohort comprised individuals with a single episode of clinically apparent SARS-CoV-2 infection (defined as a positive PCR test result or hospital admission coded with U07.1, documented in linked electronic health records or national registries, irrespective of symptom severity or healthcare-seeking behavior), where reinfection was censored at the second infection date. Further sensitivity analysis was also performed for reinfections (see below). Based on initial infection severity, the exposure was divided into hospitalized and non-hospitalized groups. Follow-up for COVID-19-infected individuals commenced on the first positive SARS-CoV-2 test date (T₀) and continued until death, the first reinfection, or 19-Dec-2022, whichever occurred first. For the primary post-acute mortality outcome, a landmark analysis [13] was performed. Individuals who died within 30 days after T₀ were excluded, and follow-up began at T₀+30 days (T_landmark). Deaths from causes other than the outcome of interest were censored.

The reference cohort comprised subjects without documented history of SARS-CoV-2. As this cohort was not defined by confirmed seronegative status, it may include individuals with unrecorded mild/asymptomatic infections. To mitigate potential survival bias from imbalanced follow-up length and ensure comparability, each control participant was assigned a pseudo-index date (T′₀) drawn randomly from the T₀ distribution observed in the exposure cohort, following the prescription time distribution matching method as described in Ref. [14]. Reference individuals were followed from their assigned T′₀ until death or 19-Dec-2022. For the post-acute analysis, the same 30-day landmark was applied to reference (T'_landmark = T′₀+30 days).

2.4. Statistical analyses

Proportional hazards Cox regression was employed to model time to the target death event. Primary analyses focused on post-acute mortality using follow-up starting at T_landmark or T'_landmark. Overall mortality (including days 0-30) was additionally evaluated as supplementary analyses; because this metric includes acute-phase deaths, it does not specifically reflect sequelae risk and results are reported in Supplementary Materials. To mitigate potential convergence issues, we performed variable selection for covariates included in the Cox model (see Supplementary_methods) [15]. False discovery rate (FDR) was used to control for multiple testing, with FDR-adjusted p-values <0.05 considered statistically significant. Overall analytic workflow is illustrated in Fig. S1.

HR for each post-acute mortality outcome was estimated by comparing the exposure (overall, hospitalized, and non-hospitalized) to the reference cohort; as the reference may include unrecorded infections, these estimates reflect mortality risk associated with clinically apparent COVID-19 relative to the general population without known infection. Acute-phase mortality (≤30 days post-infection) was additionally calculated for composite organ-system outcomes for comparison. For individual post-acute outcomes, HRs were also separately calculated for individuals with and without a prior history of the corresponding sequela. For subgroup analyses, we assessed risk heterogeneity by computing the ratio of hazard ratios (RHR) across subgroups stratified by multiple demographic/comorbidity characteristics. Parallel overall-mortality analyses are described in Supplementary_methods.

2.5. Additional analyses

Additional analyses evaluated the robustness of primary post-acute organ-system findings for overall COVID-19 exposure, addressing three potential concerns: (1) contamination of undocumented infections, through quantitative bias analysis (QBA) under scenarios informed by external estimates of cumulative UK COVID-19 prevalence [16], and use of a reference group restricted to those tested negative for COVID-19 infection; (2) effects of vaccination status on mortalities, through stratified analyses and testing for interaction with COVID-19 infection; (3) reinfection-based censoring, by extending follow-up beyond reinfection, and stratifying the exposure cohort by number of infections, with follow-up beginning at the first infection. Further methodological details are provided in Supplementary_methods.

3. Results

For the post-acute analysis, after excluding participants (COVID-19: 446; reference: 345) who died within the 30-day landmark, the COVID-19 cohort included 112,611 participants (median follow-up: 247 days), comprising 16,260 hospitalized (median 292 days) and 96,351 non-hospitalized (median 241 days) individuals; 4942 (4.4%) experienced reinfection during follow-up. The reference cohort included 354,120 participants (median 335 days), of whom 169,400 (47.8%) received SARS-CoV-2 testing but never tested positive (Table_S2). Detailed cohort baseline characteristics are presented in Table_S3.

As summarized in Table_S4a, post-acute deaths totaled 2888 (2.56%) in the overall COVID-19-infected cohort, 2382 (14.65%) in hospitalized, 506 (0.53%) in non-hospitalized, and 5317 (1.50%) in reference participants. Neoplasms, circulatory, neurological, and respiratory disorders were the most common mortality causes across all cohorts. Corresponding overall-mortality distributions are provided in Table_S4b.

3.1. Mortality risks of composite organ-system outcomes

Compared to the reference (general population without documented infection), post-acute all-cause mortality was significantly elevated in overall COVID-19-infected (HR: 1.50, 95% confidence interval (CI): 1.42-1.58) and hospitalized (3.06, 2.87-3.26) groups, but not in the non-hospitalized (1.05, 0.98-1.14) group (Fig. 1a, Table_1). By comparison, acute-phase all-cause mortality was significantly elevated across all groups with higher HRs, indicating substantial attenuation of risk from the acute to the post-acute period (Fig. 1a–Table_1).

Fig. 1.

Associations Between All-Cause and Cause-Specific Post-acute and Acute Mortality Risks in Different Organ Systems and Overall/Hospitalized/Non-hospitalized COVID-19

Legend: The vertical red dashed line represents the line of no effect (hazard ratio (HR) = 1). X-axis indicates the HR of mortality after COVID-19 infection during the post-acute (>30 days) and acute (0-30 days) periods, respectively. Y-axis indicates each COVID-19 disease severity. Confidence intervals are also shown in the figure. Solid bars indicate significance, and transparent bars indicate non-significant associations. We only present the results if the number of events was at least 5 for both COVID-19 exposure and reference cohort. Note that the X-axis scales differ between the post-acute and acute panels to accommodate the wider range of effect sizes observed during the acute period.

Table 1.

Associations between all-cause and cause-specific mortality risks in different organ systems and overall/hospitalized/non-hospitalized COVID-19.

Analysis	Outcome	Post-acute Mortalitya		Acute Mortalityb		Organ System Description
		HR (95% CI)	p.adjc	HR (95% CI)	p.adjc
Overall COVID-19 vs. reference	All-cause mortality	1.50 (1.42–1.58)	2.83E-48	15.56 (13.78–17.56)	< 2.23E-308
	CIR	1.22 (1.05–1.43)	2.17E-02	4.02 (2.76–5.85)	8.08E-13	Diseases of the circulatory system
	DIG	1.50 (1.04–2.16)	4.26E-02	5.87 (2.84–12.13)	3.16E-06	Diseases of the digestive system
	END	1.61 (0.75 - 3.46)	2.88E-01	N/Ad		Endocrine nutritional and metabolic diseases
	EXT	3.75 (1.98–7.09)	1.45E-04	N/A		External cause codes
	GEN	2.19 (1.29–3.71)	8.51E-03	N/A		Diseases of the genitourinary system
	INF	1.25 (0.79 - 1.98)	3.77E-01	3.43 (1.22–9.59)	1.91E-02	Certain infectious and parasitic diseases
	INJ	1.05 (0.52 - 2.10)	8.94E-01	N/A		Injury poisoning and certain other consequences of external causes
	MUS	1.61 (0.73 - 3.58)	2.88E-01	N/A		Diseases of the musculoskeletal system and connective tissue
	NEO	1.38 (1.26–1.51)	6.22E-12	3.37 (2.67–4.27)	1.18E-23	Neoplasms
	NVS	2.13 (1.78–2.56)	2.66E-15	3.41 (1.97–5.91)	1.49E-05	Diseases of the nervous system
	RSP	1.21 (1.04–1.41)	2.62E-02	4.28 (2.23–8.22)	1.49E-05	Diseases of the respiratory system
Hospitalized COVID-19 vs. reference	All-cause mortality	3.06 (2.87–3.26)	1.25E-258	62.83 (55.49–71.14)	< 2.23E-308
	CIR	2.37 (1.96–2.86)	2.20E-18	13.23 (8.84–19.80)	9.32E-36	Diseases of the circulatory system
	DIG	3.07 (1.98–4.76)	7.42E-07	28.70 (13.51–60.96)	4.26E-18	Diseases of the digestive system
	END	4.35 (1.70–11.10)	2.10E-03	N/A		Endocrine nutritional and metabolic diseases
	EXT	11.08 (5.35–22.96)	2.29E-10	N/A		External cause codes
	GEN	5.10 (2.88–9.06)	4.28E-08	N/A		Diseases of the genitourinary system
	INF	4.88 (2.83–8.43)	2.65E-08	12.89 (4.40–37.80)	3.21E-06	Certain infectious and parasitic diseases
	INJ	4.99 (2.39–10.40)	1.94E-05	N/A		Injury poisoning and certain other consequences of external causes
	MUS	8.57 (3.50–20.99)	3.52E-06	N/A		Diseases of the musculoskeletal system and connective tissue
	NEOe	3.10 (2.79–3.45)	3.67E-96	14.44 (11.24–18.54)	1.94E-96	Neoplasms
	NVS	5.06 (4.03–6.35)	7.86E-44	11.75 (6.46–21.40)	1.04E-15	Diseases of the nervous system
	RSP	1.87 (1.41–2.47)	1.44E-05	12.42 (6.17–24.99)	1.96E-12	Diseases of the respiratory system
Non-hospitalized COVID-19 vs. reference	All-cause mortality	1.05 (0.98 - 1.14)	4.70E-01	1.36 (1.06–1.74)	4.78E-02
	CIR	0.74 (0.37 - 1.49)	5.61E-01	1.30 (0.52 - 3.28)	6.10E-01	Diseases of the circulatory system
	DIG	0.90 (0.53 - 1.53)	7.07E-01	N/A		Diseases of the digestive system
	EXT	2.77 (1.29–5.96)	4.63E-02	N/A		External cause codes
	INF	0.65 (0.34 - 1.23)	4.70E-01	N/A		Certain infectious and parasitic diseases
	MUS	0.67 (0.21 - 2.18)	5.61E-01	N/A		Diseases of the musculoskeletal system and connective tissue
	NEO	0.76 (0.38 - 1.52)	5.61E-01	1.26 (0.52 - 3.08)	6.10E-01	Neoplasms
	NVS	1.37 (1.09–1.74)	4.63E-02	N/A		Diseases of the nervous system

HR: hazard ratio; 95% CI represents the lower and upper 95% confidence interval of HR; p.adj³: FDR-adjusted p-value.

^aPost-acute Mortality measures the risk of death by setting the follow-up start date as 30 days after the date of first positive SARS-CoV-2 test.

^bAcute Mortality measures the risk of death within the acute phase (≤30 days after the date of first positive SARS-CoV-2 test); Corresponding overall-mortality results (including both acute- and post-acute deaths) are provided in Supplementary Table 5.

^cFDR correction was separately performed for each COVID-19 exposure cohort; Statistically significant associations (p.adj <0.05) were labelled in bold face.

^dN/A indicates that the analysis for target outcome cannot proceed due to insufficient number (<5) of events in either the COVID-19 exposure or reference cohort.

^eElevated neoplasm mortality, especially among hospitalized patients without prior cancer history, should be interpreted with caution, as this finding is more likely attributable to detection bias (unmasking of pre-existing undiagnosed malignancies) rather than representing a direct post-infection sequela of COVID-19. See Discussion for detailed interpretation.

For organ-system outcomes, a wide range of significant associations were found, demonstrating increased mortality risks post infection (Fig. 1b–Table_1). Hospitalized COVID-19 exhibited significantly increased post-acute mortality in all (11/11) examined systems. Overall COVID-19 infection was associated with significantly elevated post-acute mortality in 7 of 11 systems (circulatory, digestive, external-cause, genitourinary, neoplasms, nervous, and respiratory). For the non-hospitalized cohort, only external-cause and nervous-system post-acute mortality (2/7) reached significance. Acute organ-system mortality was significant for all estimable outcomes in the overall and hospitalized groups with higher HRs; for the non-hospitalized cohort, no composite outcome reached significance during the acute phase (Fig. 1b–Table_1). Supplementary overall-mortality analyses yielded broadly consistent patterns with generally larger effect estimates (Fig. S2, Table_S5).

3.2. Mortality risks of individual outcomes for overall COVID-19 exposure

Among 42 individual disorders investigated, clinically apparent COVID-19 infection significantly increased post-acute mortality for 8 (19.0%) conditions (Table_2a, Fig. S3), including (ranked from the largest to smallest HR per system).

• •Circulatory: Peripheral/visceral vascular disease.
• •External-cause injuries: Accidental/unintentional intent of injury, and subsequent encounter of external cause codes.
• •Genitourinary: Urinary tract infection.
• •Neoplasm: Bile duct cancer, and unspecified malignant neoplasm.
• •Neurological: Neurocognitive disorders, and Parkinson's disease.

Table 2.

Associations between cause-specific mortality risks in single CCSR disorders and COVID-19.

a. Compared to reference cohort (general population without documented infection), associations of overall COVID-19 with mortality due to various disorders (i.e. Overall COVID-19 vs. reference)
Outcome	Post-acute Mortalitya,b		CCSR Category Description
	HR (95% CI)	p.adjc
CIR001	N/Ad		Chronic rheumatic heart disease
CIR003	1.13 (0.48 - 2.65)	8.44E-01	Nonrheumatic and unspecified valve disorders
CIR004	1.81 (0.64 - 5.16)	4.12E-01	Endocarditis and endocardial disease
CIR005	0.68 (0.29 - 1.62)	5.10E-01	Myocarditis and cardiomyopathy
CIR008	1.38 (0.72 - 2.66)	4.50E-01	Hypertension with complications and secondary hypertension
CIR009	0.78 (0.55 - 1.11)	2.85E-01	Acute myocardial infarction
CIR011	1.34 (0.97 - 1.85)	2.18E-01	Coronary atherosclerosis and other heart disease
CIR019	1.93 (0.97 - 3.85)	2.00E-01	Heart failure
CIR020	N/A		Cerebral infarction
CIR021	1.19 (0.75 - 1.88)	5.42E-01	Acute hemorrhagic cerebrovascular disease
CIR024	1.69 (0.63 - 4.52)	4.41E-01	Other and ill-defined cerebrovascular disease
CIR026	4.11 (2.13–7.94)	2.76E-04	Peripheral and visceral vascular disease
CIR029	N/A		Aortic; peripheral; and visceral artery aneurysms
DIG006	N/A		Gastrointestinal and biliary perforation
DIG012	1.13 (0.35 - 3.63)	8.78E-01	Intestinal obstruction and ileus
DIG017	2.17 (0.77 - 6.11)	2.70E-01	Biliary tract disease
DIG019	1.42 (0.57 - 3.51)	5.42E-01	Other specified and unspecified liver disease
DIG021	2.58 (0.81 - 8.20)	2.59E-01	Gastrointestinal hemorrhage
DIG025	2.12 (0.92 - 4.87)	2.18E-01	Other specified and unspecified gastrointestinal disorders
END016	2.07 (0.81 - 5.33)	2.70E-01	Other specified and unspecified nutritional and metabolic disorders
EXT020	4.58 (2.30–9.14)	2.11E-04	External cause codes: intent of injury, accidental/unintentional
EXT029	4.00 (2.05–7.82)	4.23E-04	External cause codes: subsequent encounter
GEN003	1.95 (0.70 - 5.39)	3.22E-01	Chronic kidney disease
GEN004	3.11 (1.52–6.39)	1.18E-02	Urinary tract infections
INF002	2.19 (0.77 - 6.21)	2.70E-01	Septicemia
INF003	2.07 (0.89 - 4.85)	2.42E-01	Bacterial infections
NEO012	0.86 (0.52 - 1.41)	6.20E-01	Gastrointestinal cancers - esophagus
NEO013	1.40 (0.72 - 2.76)	4.50E-01	Gastrointestinal cancers - stomach
NEO015	1.37 (1.02 - 1.84)	1.62E-01	Gastrointestinal cancers - colorectal
NEO017	0.95 (0.43 - 2.08)	8.99E-01	Gastrointestinal cancers - liver
NEO018	2.30 (1.24–4.25)	4.10E-02	Gastrointestinal cancers - bile duct
NEO024	0.92 (0.28 - 2.99)	8.99E-01	Sarcoma
NEO025	1.90 (1.02 - 3.55)	1.79E-01	Skin cancers - melanoma
NEO043	1.23 (0.71 - 2.15)	5.42E-01	Urinary system cancers - bladder
NEO048	1.27 (0.79 - 2.05)	4.50E-01	Nervous system cancers - brain
NEO051	0.91 (0.62 - 1.31)	6.64E-01	Endocrine system cancers - pancreas
NEO067	2.20 (0.79 - 6.16)	2.70E-01	Mesothelioma
NEO068	1.89 (0.73 - 4.91)	3.21E-01	Myelodysplastic syndrome (MDS)
NEO071	2.13 (1.35–3.37)	8.36E-03	Malignant neoplasm, unspecified
NVS004	2.29 (1.64–3.19)	2.39E-05	Parkinson's disease
NVS005	1.91 (0.78 - 4.65)	2.85E-01	Multiple sclerosis
NVS006	1.79 (0.87 - 3.66)	2.59E-01	Other specified hereditary and degenerative nervous system conditions
NVS011	2.32 (1.81–2.97)	1.18E-09	Neurocognitive disorders
RSP002	1.62 (0.98 - 2.68)	2.00E-01	Pneumonia (except that caused by tuberculosis)
RSP008	1.36 (1.00 - 1.85)	1.90E-01	Chronic obstructive pulmonary disease and bronchiectasis
RSP010	N/A		Aspiration pneumonitis
RSP016	0.76 (0.39 - 1.49)	5.32E-01	Other specified and unspecified lower respiratory disease

b. Compared to reference cohort (general population without documented infection), associations of hospitalized COVID-19 with mortality due to various disorders (i.e. Hospitalized COVID-19 vs. reference)
Outcome	Post-acute Mortalitya,b		CCSR Category Description
	HR (95% CI)	p.adjc
CIR003	1.39 (0.46 - 4.17)	5.90E-01	Nonrheumatic and unspecified valve disorders
CIR004	N/Ad		Endocarditis and endocardial disease
CIR005	N/A		Myocarditis and cardiomyopathy
CIR008	1.99 (0.81 - 4.88)	1.53E-01	Hypertension with complications and secondary hypertension
CIR009	1.54 (0.99 - 2.39)	6.95E-02	Acute myocardial infarction
CIR011	2.71 (1.81–4.04)	3.64E-06	Coronary atherosclerosis and other heart disease
CIR019	4.53 (2.06–9.93)	4.14E-04	Heart failure
CIR020	N/A		Cerebral infarction
CIR021	1.57 (0.75 - 3.28)	2.47E-01	Acute hemorrhagic cerebrovascular disease
CIR024	2.93 (1.01 - 8.54)	6.39E-02	Other and ill-defined cerebrovascular disease
CIR026	8.61 (4.21–17.58)	1.76E-08	Peripheral and visceral vascular disease
DIG012	1.47 (0.36 - 5.93)	6.08E-01	Intestinal obstruction and ileus
DIG017	5.99 (1.82–19.71)	5.90E-03	Biliary tract disease
DIG019	3.96 (1.40–11.21)	1.36E-02	Other specified and unspecified liver disease
DIG025	2.86 (0.99 - 8.27)	6.75E-02	Other specified and unspecified gastrointestinal disorders
END016	4.76 (1.51–14.97)	1.13E-02	Other specified and unspecified nutritional and metabolic disorders
EXT020	12.01 (5.42–26.60)	5.51E-09	External cause codes: intent of injury, accidental/unintentional
EXT029	11.91 (5.52–25.66)	1.91E-09	External cause codes: subsequent encounter
GEN003	4.21 (1.35–13.13)	1.79E-02	Chronic kidney disease
GEN004	7.20 (3.35–15.47)	1.74E-06	Urinary tract infections
INF002	N/A		Septicemia
INF003	4.45 (1.49–13.30)	1.13E-02	Bacterial infections
NEO012e	2.33 (1.31–4.16)	6.80E-03	Gastrointestinal cancers - esophagus
NEO013e	4.37 (2.03–9.42)	4.14E-04	Gastrointestinal cancers - stomach
NEO015e	3.40 (2.34–4.95)	1.50E-09	Gastrointestinal cancers - colorectal
NEO017	1.20 (0.38 - 3.84)	7.58E-01	Gastrointestinal cancers - liver
NEO018e	7.52 (3.84–14.72)	1.76E-08	Gastrointestinal cancers - bile duct
NEO024	2.45 (0.62 - 9.72)	2.28E-01	Sarcoma
NEO025e	4.26 (1.93–9.43)	7.98E-04	Skin cancers - melanoma
NEO043e	2.56 (1.36–4.83)	6.30E-03	Urinary system cancers - bladder
NEO048e	2.59 (1.46–4.59)	2.42E-03	Nervous system cancers - brain
NEO051e	1.94 (1.20–3.13)	1.13E-02	Endocrine system cancers - pancreas
NEO067	N/A		Mesothelioma
NEO068e	4.93 (1.82–13.36)	3.27E-03	Myelodysplastic syndrome (MDS)
NEO071e	11.36 (6.49–19.86)	2.91E-16	Malignant neoplasm, unspecified
NVS004	6.74 (4.53–10.04)	2.31E-19	Parkinson's disease
NVS005	7.32 (2.72–19.66)	2.26E-04	Multiple sclerosis
NVS006	4.10 (1.79–9.38)	1.84E-03	Other specified hereditary and degenerative nervous system conditions
NVS011	3.47 (2.49–4.82)	1.79E-12	Neurocognitive disorders
RSP002	4.61 (2.52–8.44)	2.50E-06	Pneumonia (except that caused by tuberculosis)
RSP008	2.20 (1.55–3.12)	2.87E-05	Chronic obstructive pulmonary disease and bronchiectasis
RSP016	0.35 (0.10 - 1.31)	1.42E-01	Other specified and unspecified lower respiratory disease

c. Compared to reference cohort (general population without documented infection), associations of non-hospitalized COVID-19 with mortality due to various disorders (i.e. Non-hospitalized COVID-19 vs. reference)
Outcome	Post-acute Mortalitya,b		CCSR Category Description
	HR (95% CI)	p.adjc
CIR003	N/Ad		Nonrheumatic and unspecified valve disorders
CIR004	N/A		Endocarditis and endocardial disease
CIR005	N/A		Myocarditis and cardiomyopathy
CIR008	1.00 (0.41 - 2.45)	9.96E-01	Hypertension with complications and secondary hypertension
CIR009	0.48 (0.17 - 1.32)	4.49E-01	Acute myocardial infarction
CIR011	0.73 (0.45 - 1.18)	4.49E-01	Coronary atherosclerosis and other heart disease
CIR019	N/A		Heart failure
CIR021	1.22 (0.73 - 2.05)	5.85E-01	Acute hemorrhagic cerebrovascular disease
CIR026	2.19 (0.85 - 5.65)	4.16E-01	Peripheral and visceral vascular disease
DIG017	N/A		Biliary tract disease
DIG025	2.35 (0.82 - 6.80)	4.16E-01	Other specified and unspecified gastrointestinal disorders
EXT020	4.04 (1.80–9.06)	7.80E-03	External cause codes: intent of injury, accidental/unintentional
EXT029	3.44 (1.56–7.59)	1.58E-02	External cause codes: subsequent encounter
INF003	1.96 (0.74 - 5.17)	4.49E-01	Bacterial infections
NEO012	0.34 (0.06 - 1.80)	4.49E-01	Gastrointestinal cancers - esophagus
NEO013	N/A		Gastrointestinal cancers - stomach
NEO015	0.78 (0.52 - 1.19)	5.03E-01	Gastrointestinal cancers - colorectal
NEO017	0.97 (0.37 - 2.54)	9.96E-01	Gastrointestinal cancers - liver
NEO025	1.15 (0.50 - 2.67)	8.54E-01	Skin cancers - melanoma
NEO043	N/A		Urinary system cancers - bladder
NEO048	0.73 (0.33 - 1.62)	5.85E-01	Nervous system cancers - brain
NEO051	0.57 (0.34 - 0.94)	1.55E-01	Endocrine system cancers - pancreas
NEO071	1.23 (0.67 - 2.26)	6.24E-01	Malignant neoplasm, unspecified
NVS004	1.25 (0.79 - 1.99)	5.30E-01	Parkinson's disease
NVS006	0.99 (0.36 - 2.73)	9.96E-01	Other specified hereditary and degenerative nervous system conditions
NVS011	1.86 (1.37–2.50)	1.21E-03	Neurocognitive disorders
RSP002	0.63 (0.27 - 1.48)	5.30E-01	Pneumonia (except that caused by tuberculosis)
RSP008	0.76 (0.41 - 1.39)	5.39E-01	Chronic obstructive pulmonary disease and bronchiectasis
RSP016	1.42 (0.70 - 2.88)	5.30E-01	Other specified and unspecified lower respiratory disease

HR: hazard ratio; 95% CI represents the lower and upper 95% confidence interval of HR; p.adj³: FDR-adjusted p-value.

CCSR: Clinical Classifications Software Refined, a method to classify disease categories.

^aPost-acute Mortality measures the risk of death by setting the follow-up start date as 30 days after the date of first positive SARS-CoV-2 test.

^bCorresponding overall-mortality results (including both acute and post-acute deaths) are provided in Supplementary Table 6.

^cFDR correction was separately performed for each COVID-19 exposure cohort; Statistically significant associations (p.adj <0.05) were labelled in bold face.

^dN/A indicates that the analysis for target outcome cannot proceed due to insufficient number (<5) of events in either the COVID-19 exposure or reference cohort.

^eElevated neoplasm mortality, especially among hospitalized patients without prior cancer history, should be interpreted with caution, as this finding is more likely attributable to detection bias (unmasking of pre-existing undiagnosed malignancies) rather than representing a direct post-infection sequela of COVID-19. See Discussion for detailed interpretation.

The proportion of significant associations was 42.9% (12/28), and 10.7% (3/28) for post-acute death outcomes with, and without a prior history, respectively (Fig. S3). Supplementary overall-mortality analysis (Fig. S4) showed a consistent but broader pattern of significance. Detailed results are elaborated in Supplementary_text, Table_S6a and Table_S7.

3.3. Mortality risks of individual outcomes for hospitalized COVID-19

Compared to the reference population, hospitalized COVID-19 significantly elevated post-acute mortality for 73.0% (27/37) of outcomes encompassing multiple systems (Table_2b). Among post-acute deaths with and without a pre-existing diagnosis, 91.7% (22/24) and 71.4% (10/14) of disorders exhibited greater mortality risk (Table_S7). However, we cautioned that mortality without a prior history, particularly for neoplasms, might reflect detection bias (i.e., unmasking of pre-existing subclinical malignancies during hospitalization) rather than truly new-onset diseases. These associations should not be interpreted as COVID-19-driven sequelae and are mechanistically distinct from other outcomes such as circulatory or respiratory mortality. Supplementary overall-mortality analyses showed higher proportions of significant associations (see Supplementary_text, Table_S6b and Table_S7).

3.4. Mortality risks of individual outcomes for non-hospitalized COVID-19

Two external-cause outcomes and neurocognitive disorders were significantly associated with increased post-acute mortality (3/22) for non-hospitalized COVID-19 (Table_2c). The same significant associations were observed for deaths with prior history (3/9; Table_S7).

For deaths without pre-existing diagnoses, only neurocognitive disorders remained significant for post-acute mortality. Detailed overall-mortality findings are provided in Supplementary_text, Table_S6c and Table_S7.

3.5. Subgroup analysis

Significant heterogeneity in post-acute mortality was observed across subgroups (Table_3, Fig. 2). All-cause mortality risk was significantly greater in individuals with advanced age (>65 years), atrial fibrillation (AF), chronic kidney disease (CKD), or hypertension (HTN) than in their risk-factor-absent counterparts.

Table 3.

Significant differences in associations between post-acute mortality risks in organ systems and overall COVID-19 across subgroups.

Outcome	HR (CI)present	p.adjpresent	Sigpresent	HR (CI)absent	p.adjabsent	Sigabsent	RHR	p.adjRHR*	Comparison	Organ System Description
All-cause mortality	1.61 (1.52 - 1.70)	3.84E-59	***	1.15 (0.98 - 1.36)	1.94E-01		1.40	1.18E-03	Advanced agea vs. Middle ageb
All-cause mortality	1.69 (1.50 - 1.89)	2.69E-18	***	1.43 (1.34 - 1.52)	2.35E-29	***	1.18	3.55E-02	AF vs. AF-free
All-cause mortality	1.80 (1.63 - 1.97)	4.86E-33	***	1.34 (1.26 - 1.43)	2.13E-17	***	1.34	6.31E-06	CKD vs. CKD-free
All-cause mortality	1.59 (1.48 - 1.69)	8.92E-42	***	1.31 (1.19 - 1.44)	2.87E-07	***	1.21	7.52E-03	HTN vs. HTN-free
CIR	1.37 (1.16 - 1.61)	2.39E-04	***	0.51 (0.28 - 0.92)	1.50E-01		2.67	4.61E-03	Advanced agea vs. Middle ageb	Diseases of the circulatory system
NVS	2.49 (2.04 - 3.03)	8.44E-19	***	1.12 (0.67 - 1.87)	6.73E-01		2.23	2.33E-02	AF vs. AF-free	Diseases of the nervous system
RSP	1.89 (1.26 - 2.85)	5.29E-03	**	0.94 (0.70 - 1.25)	6.73E-01		2.02	2.33E-02	AF vs. AF-free	Diseases of the respiratory system

HR(CI)_present: Hazard ratio and 95% confidence interval for the risk factor present subgroup with the examined characteristics; p.adj_present: FDR-adjusted p-value for the risk factor present subgroup; HR(CI)_absent: Hazard ratio and 95% confidence interval for the risk factor absent subgroup without the examined characteristics; p.adj_absent: FDR-adjusted p-value for the risk factor absent subgroup; Sig_present and Sig_absent: * indicates an FDR-adjusted p-value between 0.01 and 0.05, ** indicates between 0.001 and 0.01, and *** indicates a value smaller than 0.001 for the risk factor present and absent subgroups, respectively.RHR: Ratio of hazard ratios for comparisons between the risk factor present and absent subgroups.

p.adj_RHR*: FDR-adjusted p-value for comparisons between the risk factor present and absent subgroups, where a p.adj_RHR < 0.05 indicates significant differences. Only results with significant differences (p.adj_RHR < 0.05) are shown.

^aAdvanced age: >65 years old.

^bMiddle age: 50-65 years old.

Fig. 2.

Significant Differences in Associations between Post-acute Mortality Risks in Organ Systems and Overall COVID-19 Across Subgroups

Legend: Comparisons showing significant differences (pval.adj_RHR < 0.05) in hazard ratios (HR) for mortality with (risk factor present) and without (risk factor absent) specific characteristics are presented. The red dashed line represents the line of no effect (hazard ratio (HR) = 1). X-axis indicates the HR of post-acute mortality after COVID-19 infection. Y-axis indicates each CCSR disease category. Red and blue color respectively represent association results in the risk factor present and absent subgroups. Solid bars indicate significance, and transparent bars indicate non-significant associations for each subgroup. We only present the results if the number of events was at least 5 for COVID-19 and uninfected individuals in both subgroups. Advanced age represents >65 years old, and Middle age refers to 50-65 years old.

Post-acute circulatory mortality differed significantly between age subgroups, with higher risk in older adults (>65 years), whereas no statistically significant change was observed among middle-aged participants after FDR correction (within-group HR: 0.51, 0.28-0.92; adjusted p = 0.15). AF patients experienced higher post-acute mortality from nervous and respiratory diseases than AF-free individuals.

No significant subgroup differences were found for individual CCSR outcomes after FDR correction. Detailed results are provided in Tables_S8–S19.

3.6. Additional analyses

• •Undocumented infections: both QBA and the negative-test restricted reference analyses yielded generally higher HRs than the primary analysis (Tables_S20-S21), indicating that primary estimates quantify the mortality risk of clinically apparent COVID-19 relative to the general UK population rather than relative to truly uninfected individuals. Undocumented infections in the reference cohort likely attenuated effect estimates, though the overall pattern of significant associations was preserved.
• •Vaccination status: Elevated all-cause mortality was observed in both vaccinated and unvaccinated cases, with greater risk estimates in the unvaccinated group but no significant interaction with vaccination detected (Tables_S22-S23).
• •Reinfection: Estimates were similar when follow-up was extended beyond reinfection (Table_S24), suggesting the reinfection-censoring strategy had minimal impact on the results. Stratified analyses by infection number showed higher risks of post-acute all-cause, neoplastic, neurological and respiratory mortality among reinfected individuals compared to those with a single infection (Table_S25). Further details are presented in Supplementary_text.

4. Discussion

This UKBB cohort study examined the risks of all-cause and cause-specific mortality across 12 organ systems and 47 individual disease categories for up to 35 months following COVID-19. SARS-CoV-2 infection was associated with increased post-acute mortality due to circulatory, digestive, genitourinary, neurological, and respiratory disorders, as well as external causes and neoplasms. The risk was particularly pronounced in individuals who had been hospitalized for COVID-19. Furthermore, analyses stratified by patient history and comorbidities revealed that these mortality risks are significantly influenced by pre-existing conditions and demographic factors.

An important interpretive consideration is that the reference cohort, defined by the absence of documented SARS-CoV-2 infection, likely included individuals with unrecorded infections given the high cumulative UK prevalence during the study period [16]. Our findings therefore reflect the mortality risk of clinically apparent COVID-19 relative to the general UK population during the pandemic, rather than relative to strictly uninfected individuals. Both QBA and negative-test restricted reference analyses confirmed that this contamination biases estimates toward the null, indicating that true infection-attributable mortality risks are likely larger than reported.

Previous studies [7,8] reported excess mortality after the emergence of SARS-CoV-2, most prevalently involving respiratory, cardiovascular, neurological disorders, and cancers. However, since these estimates were based on population-level aggregates without adjustment for individual confounders, they can be sensitive to the choice of pre-pandemic reference period and may reflect healthcare overloads or socioeconomic disruptions rather than COVID-19's unique effects [17]. In contrast, this study directly assessed mortality following SARS-CoV-2 infection using individual-level data and Cox regression, with a primary focus on the post-acute phase (beyond 30 days). We reported increased mortality for previously identified conditions, but also uncovered new associations with other disorders such as digestive, genitourinary diseases, and external injuries.

Consistent with existing literature [5], we observed elevated post-acute all-cause mortality in COVID-19 survivors, with the highest HR among those requiring hospitalization, though this association did not reach significance in the non-hospitalized cohort. Cause-specific analyses further showed that hospitalized COVID-19 individuals experienced increased post-acute mortality across a broader spectrum of organ systems and CCSR disorders than the non-hospitalized group, demonstrating a clear linkage between initial COVID-19 severity and subsequent mortality burden.

Our work addresses a critical gap in prior research, which primarily investigated symptoms or hospitalizations (but not mortality) as COVID-19 sequelae [5,6]. While Mendelian randomization studies support potential causal associations between hospitalized COVID-19 and post-COVID-19 syndromes across multiple systems [[18], [19], [20]], this method is less suited to examine post-COVID-19 mortality due to methodological constraints, for example the difficulty in obtaining genetic instruments for mortality outcomes and the inability to capture the full temporal evolution of post-infection risk. Many previous epidemiology or MR studies [21,22] are also limited to single systems or a restricted number of pre-specified conditions. In contrast, our study provides a systematic and comprehensive assessment of post-infection mortality using CCSR classification, avoiding any arbitrary exclusions or preferential selections and thus enabling the identification of a wider array of high-risk outcomes. Additionally, subgroup analyses were conducted to explore differences in post-acute effects across subpopulations.

The mechanisms driving the excess mortalities are likely multifactorial. The observed increase in respiratory-related post-acute mortality aligns with previous studies suggesting non-COVID pneumonia as a common COVID-19 sequela [23]. The elevated post-acute cardiovascular mortality may be attributable to direct infiltration of SARS-CoV-2, autoimmune dysregulation, and inflammatory response within cardiac tissues [24]. A recent study [9] also revealed increased 12-month risks of all-cause, respiratory, and cardiovascular mortality in hospitalized COVID-19 patients, though respiratory/cardiovascular conditions accounted for only 20.5% of deaths. This corroborates our observation that SARS-CoV-2's impact extends beyond these two systems. For instance, we found increased post-acute mortality for neurological disorders, consistent with research demonstrating an elevated 12-month risk of incident neurological sequelae following COVID-19 [25].

We also identified increased mortality from conditions less frequently associated with COVID-19. For example, we observed elevated post-acute genitourinary mortality, which could be mediated through coronavirus binding to ACE2 receptors in the urinary tract, inducing cellular damage and increasing susceptibility to infection [26]. Besides, renal complications were reported among COVID-19 survivors [27], potentially resulting from direct viral invasion, persistent tubular/microvascular injury from systemic inflammation and coagulopathy, and podocyte damage. Our finding of elevated post-acute digestive mortality aligns with emerging evidence of viral RNA persistence in gut tissues and sustained systemic inflammation, contributing to gastrointestinal/hepatobiliary manifestations of long COVID [28]. Furthermore, the increased post-acute mortality from external causes, primarily accidental injuries, highlights significant secondary consequences of COVID-19. Post-infection muscle weakness, fatigue, and cognitive dysfunction may impair physical coordination and risk perception [23]. COVID-19 or its clinical treatment (e.g., glucocorticoids) may also accelerate bone loss and osteoporosis progression, raising the likelihood of fractures and fatal injuries [29].

Post-acute mortality from neoplasms was elevated; however, interpretation differs by prior cancer status. Among patients with a prior cancer history, the elevated mortality may represent a true sequela, mediated by COVID-19-induced immune dysregulation that exacerbates pre-existing malignancies. Viral infection may trigger excessive inflammatory cytokine release, fostering a pro-tumorigenic environment [30]. Moreover, SARS-CoV-2 may deplete critical cancer-fighting immune cells such as CD4+/CD8+ T cells, impairing immune surveillance and worsening prognosis in cancer patients following infection.

In contrast, elevated cancer mortality in patients without prior diagnoses, observed exclusively in hospitalized cases, is more plausibly attributable to detection bias. Unlike other conditions (e.g., circulatory and respiratory) that can manifest acutely, neoplasms are uniquely susceptible to this bias because cancers often develop and progress asymptomatically over prolonged periods, remaining clinically silent until incidentally discovered through extensive imaging or laboratory testing during hospitalization. Additionally, the physiological stress of severe infection may have accelerated the clinical presentation of subclinical cancers [31]. Given the biological implausibility of de novo carcinogenesis within the study timeframe, these findings should be distinguished from deaths attributable to post-acute sequelae of COVID-19.

While mortality risks appeared less pronounced in individuals without prior history of the examined outcomes, several non-neoplastic post-acute associations remained significant and are more biologically plausible as direct consequences of SARS-CoV-2-mediated pathology. For example, peripheral/visceral vascular disease may arise from SARS-CoV-2 invasion of vascular epithelial cells via ACE2 receptors, triggering intravascular thrombotic events and endothelial dysfunction [32]. Unlike the neoplasm findings discussed above, these associations may reflect established post-infection mechanisms and support a genuine sequela interpretation.

Comparing acute and post-acute composite organ system mortality provided key insights. Where estimable, acute-phase HRs were consistently and substantially higher than their corresponding post-acute estimates, suggesting that mortality risks are primarily concentrated within the first 30 days after infection. Conversely, the persistence of significantly elevated risks across multiple organ systems beyond the 30-day landmark indicates that SARS-CoV-2's impact on mortality extends well beyond the acute phase. However, the limited number of death events within the 30-day window reduced statistical power and produced large standard errors, precluding acute-mortality analysis of individual CCSR diseases. For completeness, overall-mortality analyses combining both phases, which reflect total health burden with maximum statistical power though not specifically isolating sequelae risk, are provided in the Supplementary Materials.

Our subgroup analysis indicated that older adults with chronic conditions (AF, CKD, or HTN) were more vulnerable to deleterious consequences of SARS-CoV-2 with elevated all-cause and circulatory mortality. Moreover, AF patients demonstrated increased post-acute mortality for both neurological and respiratory diseases, potentially suggesting a synergistic prothrombotic effect, as both AF and COVID-19 may induce thromboembolic complications like ischemic stroke and pulmonary embolism, contributing to the corresponding organ dysfunction and observed mortality patterns [33]. Importantly, although at a lower magnitude than their risk-factor-present counterparts, elevated post-acute all-cause mortality was observed in certain risk-factor-absent groups, highlighting the multifactorial nature of post-COVID-19 mortality.

Although unvaccinated cases showed a broader risk spectrum than vaccinated individuals, our data did not provide sufficient evidence that vaccination attenuates post-acute sequelae, as no significant interaction between COVID-19 infection and vaccination status was detected. However, incomplete vaccination records in the UKBB may have limited our ability to detect the true effect. Mortality risks were also higher among reinfected individuals, though this warrants cautious interpretation given limited statistical power due to sparse events, and the possibility that reinfected groups represent a more immunologically vulnerable population. As our study was not designed to evaluate the effects of vaccination or reinfection on post-COVID-19 mortality, further investigations are needed.

5. Strengths and limitations

This study possesses unique strengths. First, it leveraged the breadth and depth of UKBB electronic health record system to establish a large cohort with sufficient sample size and follow-up. Instead of focusing on symptoms/hospitalizations, we analyzed mortality as the most severe post-infection consequence. Secondly, our primary 30-day landmark design helped distinguish post-acute effects from acute-phase mortality, with supplementary overall-mortality analyses provided for reference. Also, the contribution of SARS-CoV-2 to deaths was assessed via regression using individual-level data rather than excess-mortality estimates, and causes of post-COVID mortality were systematically categorized into a comprehensive range of 12 organ systems and 47 individual disorders using CCSR. In addition, analyses stratified by infection severity, prior history, and demographic/comorbidity factors further revealed heterogeneity in post-acute mortality risks.

Our work has several limitations. First, the UKBB cohort comprises primarily older, white, healthier individuals with higher educational attainment [34], which may limit generalizability of the observed mortality to other populations. However, the pathophysiological mechanisms underlying post-acute sequelae of COVID-19 are likely broadly applicable across diverse populations. Therefore, the fundamental insights derived from our findings remain highly relevant beyond the study cohort. Second, as an observational study, causality cannot be established between COVID-19 and post-acute mortality risks. While we adjusted for a comprehensive set of covariates, unmeasured confounding cannot be completely excluded. Moreover, although we identified neoplasms as particularly susceptible to detection bias due to their prolonged asymptomatic development, this bias may also partially affect other disease categories, where intensive clinical evaluation during COVID-19 may have uncovered pre-existing but undiagnosed conditions, inflating some observed associations. Thirdly, our results represent average mortality risks observed during the study period (31-Jan-2020 to 19-Dec-2022), and longer-term effects beyond this timeframe remain uncertain. Fourth, cause-of-death classification based on ICD-10 codes may be subject to misclassification, and the small number of events for certain outcomes may result in limited statistical power. Fifth, given the high cumulative UK COVID-19 prevalence during the study period [16], the reference cohort likely included individuals with undocumented infections. As discussed above, this shifts the comparative baseline such that our estimates reflect clinically apparent COVID-19 versus the general population rather than strictly uninfected individuals. Both QBA and negative-test restricted reference-group analyses confirmed that true infection-attributable risks are likely larger, while the overall pattern of significant associations remained unchanged. Lastly, due to restricted data availability during analysis, UKBB vaccination records including booster doses were incomplete. Hence, vaccination status was treated only as a binary covariate and its effects on post-COVID mortality could not be examined in detail.

6. Clinical implications

This study demonstrates that clinically apparent COVID-19 is associated with elevated post-acute all-cause and cause-specific mortality across multiple organ systems relative to the general population without known infection, with hospitalized infections showing significantly increased risks across a broader spectrum of outcomes. However, elevated cancer mortality among patients without prior diagnoses may reflect detection bias rather than direct post-infection pathology, and should be interpreted distinctly from other sequelae-driven mortality. These findings underscore the critical need for preventive strategies to mitigate COVID-19 severity through early treatment. Notably, even non-hospitalized infection is linked to increased post-acute deaths from neurological disorders and external-cause injuries. Given the vast number of infections worldwide and that our estimates might be conservative relative to truly uninfected individuals, the absolute mortality burden imposed by COVID-19 is likely substantial.

In the post-pandemic era, public health policies should prioritize the management of post-COVID-19 complications by allocating healthcare resources for long-term follow-up and risk assessment of COVID-19 survivors, especially older adults and those with pre-existing high-risk comorbidities. Future research is crucial to investigate the underlying mechanisms, identify effective interventions to ameliorate the mortality burden, and examine the long-term trajectory of mortality risks tied to post-COVID-19 outcomes.

Consent to publish

Not applicable.

Ethics approval statement

The UK Biobank study has received ethical approval from the NHS National Research Ethics Service North West (16/NW/0274). Individual consents were obtained by the UK Biobank. The current study was conducted under the project number 28732. Only de-identified data was accessed, and no attempts were made to identify any individual participants in this study.

Availability of data and materials

UK Biobank (UKBB) data was used for the current study. Access to UKBB data is restricted to researchers who have made prior applications. Detailed instructions on how to apply for access to the database can be found on the UK Biobank homepage (https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access). All supplementary tables are available at the journal's website and at https://drive.google.com/drive/folders/1Jce5fXCnW-SyEp2Z5mOceKV4T64NFCdG?usp=sharing.

Funding source

This work was supported by the National Natural Science Foundation of China (Grant-81971706), the Lo-Kwee-Seong Biomedical Research Fund, and the Joint Laboratory of Bioresources and Molecular Research of Common Diseases of the Kunming Institute of Zoology and The Chinese Univeristy of Hong Kong, China.

CRediT authorship contribution statement

Ruoyu Zhang: Formal analysis, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Yong Xiang: Conceptualization, Formal analysis, Software, Validation, Visualization, Writing – review & editing. Jinghong Qiu: Formal analysis, Validation, Visualization, Writing – review & editing. Hon-Cheong So: Conceptualization, Data curation, Funding acquisition, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

Supplementary Fig. 1.

An Overview of Analytic Workflow

Legend: Cohort establishment: UKBB participants were classified into overall COVID-19 exposure and reference cohorts. This overall exposure cohort was further divided into hospitalized and non-hospitalized exposure cohorts. For our primary analysis, samples who died within the 30-days acute period after COVID-19 infection were excluded for all the cohorts.

Outcome definition: The primary (composite) outcomes were: (1) all-cause mortality and (2) cause-specific mortality grouped by 12 organ systems; Secondary (single) outcome included cause-specific mortality grouped by 47 individual CCSR disorders, which were further stratified by the presence or absence of a prior history of the same condition.

Statistical analysis: In our primary analysis, adjusted Cox regression models were employed to assess post-acute mortality risks of overall COVID-19 exposure compared to the reference. Additional analyses included (1) stratification of the exposure cohort into hospitalized and non-hospitalized cohorts; (2) stratification of single CCSR outcomes by prior history of the same condition; and (3) subgroup comparisons based on the presence or absence of specific risk factors. For further details, please refer to the Method sections in the main text.

Supplementary Fig. 2.

Associations Between All-Cause and Cause-Specific Overall Mortality Risks in Different Organ Systems and Overall/Hospitalized/Non-hospitalized COVID-19

Legend: The vertical red dashed line represents the line of no effect (hazard ratio (HR) = 1). X-axis indicates the HR of overall mortality after COVID-19 infection. Y-axis indicates each COVID-19 disease severity. Confidence intervals are also shown in the figure. Solid bars indicate significance, and transparent bars indicate non-significant associations. We only present the results if the number of events was at least 5 for both COVID-19 exposure and reference cohort.

Note: Overall mortality includes the acute phase (0-30 days) and therefore does not specifically reflect sequelae risk.

Supplementary Fig. 3.

Associations Between Cause-Specific Post-acute Mortality Risks in Individual Disorders (Stratified by Prior History) and Overall COVID-19

Legend: The 1st, 2nd and 3rd column respectively represent the results for any death records, death with a prior history, and death without a prior history of the same condition. The red dashed line represents the line of no effect (hazard ratio (HR) = 1). X-axis indicates the HR of post-acute mortality after COVID-19 infection. Y-axis indicates each CCSR disease category. Solid bars indicate significance, and transparent bars indicate non-significant associations. We only present the results if the number of events was at least 5 for both COVID-19 exposure and reference cohort. Individual CCSR disorders are grouped by organ systems. CIR: circulatory system diseases; DIG: digestive system diseases; END: endocrine, nutritional and metabolic diseases; EXT: external causes of morbidity; GEN: genitourinary system diseases; INF: certain infectious and parasitic diseases; NEO: neoplasms; NVS: nervous system diseases; RSP: respiratory system diseases.

Supplementary Fig. 4.

Associations Between Cause-Specific Overall Mortality Risks in Individual Disorders (Stratified by Prior History) and Overall COVID-19

Legend: The 1st, 2nd and 3rd column respectively represent the results for any death records, death with a prior history, and death without a prior history of the same condition. The red dashed line represents the line of no effect (hazard ratio (HR) = 1). X-axis indicates the HR of overall mortality after COVID-19 infection. Y-axis indicates each CCSR disease category. Solid bars indicate significance, and transparent bars indicate non-significant associations. We only present the results if the number of events was at least 5 for both COVID-19 exposure and reference cohort. Individual CCSR disorders are grouped by organ systems. CIR: circulatory system diseases; DIG: digestive system diseases; END: endocrine, nutritional and metabolic diseases; EXT: external causes of morbidity; GEN: genitourinary system diseases; INF: certain infectious and parasitic diseases; NEO: neoplasms; NVS: nervous system diseases; RSP: respiratory system diseases.

Note: Overall mortality includes the acute phase (0-30 days) and therefore does not specifically reflect sequelae risk.