Cardiovascular outcomes in patients with chronic kidney disease and COVID-19: a multi-regional data-linkage study.

BACKGROUND: Data describing cardiovascular outcomes in patients with COVID-19 and chronic kidney disease (CKD) are lacking. We compared cardiovascular outcomes of patients with and without COVID-19, stratified by CKD status.
METHODS: This retrospective, multi-regional data-linkage study utilised individual patient-level data from two Scottish cohorts. All patients tested for SARS-CoV-2 in Cohort 1 between 01/02/2020 and 31/03/2021, and in Cohort 2 between 28/02/2020 and 08/02/2021, were included.
RESULTS: Overall, 86 964 patients were tested for SARS-CoV-2. There were 36 904 patients (61±21 years, 58.1% women, 15.9% CKD, 10.1% COVID-19 positive) in Cohort 1 and 50 060 patients (63±20 years, 62.0% women, 16.4% CKD, 9.1% COVID-19 positive) in Cohort 2. In CKD patients, COVID-19 increased the risk of cardiovascular death by more than two-fold within 30 days (cause-specific hazard ratio [csHR] meta-estimate 2.34, 95% confidence interval [CI] 1.83-2.99), and by 57% at the end of follow-up (csHR meta-estimate 1.57, 95% CI 1.31-1.89). Similarly, the risk of all-cause death in COVID-19 positive versus negative CKD patients was greatest within 30 days (HR 4.53, 95% CI 3.97-5.16). Compared to patients without CKD, those with CKD had a higher risk of testing positive (11.5% versus 9.3%). Following a positive test, CKD patients had higher rates of cardiovascular death (11.1% versus 2.7%), cardiovascular complications, and cardiovascular hospitalisations (7.1% versus 3.3%) than those without CKD.
CONCLUSIONS: COVID-19 increases the risk of cardiovascular and all-cause death in CKD patients, especially in the short-term. CKD patients with COVID-19 are also at a disproportionate risk of cardiovascular complications than those without CKD.

Introduction

Coronavirus disease 2019 (COVID-19), the disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1], has had an unprecedented public health, societal and economic impact. Disease severity in patients with COVID-19 can vary markedly, from no symptoms or a mild respiratory illness, to life-threatening pulmonary and extra-pulmonary complications and death [2]. Several factors have been identified that increase disease severity. These include older age, male sex, social deprivation, obesity, and comorbidities such as cardiovascular disease and chronic kidney disease (CKD) [3-8]. The risk of critical illness and death following COVID-19 increases as kidney function declines, such that patients with the most advanced CKD have the poorest outcomes [4, 9–11].

The commonest complication of CKD is cardiovascular disease [12]. As estimated glomerular filtration rate (eGFR) declines, the risk of cardiovascular disease and major adverse cardiovascular events increases [13]. Whilst pre-existing cardiovascular disease is a risk factor for COVID-19 severity, several studies have also suggested that myocardial injury and cardiovascular complications are common following COVID-19 and associated with worse outcomes for these patients [14-16]. However, data describing the nature and frequency of cardiovascular outcomes in patients with CKD and COVID-19 are lacking. Specifically, it is unclear how COVID-19 modifies the existing cardiovascular risk in patients with CKD in the short- and medium-term. In this unique multi-regional data-linkage study, we evaluated the clinical characteristics and cardiovascular outcomes of patients with and without COVID-19, stratified by the presence or absence of CKD.

Methods

Patient population and study design

We conducted a retrospective, multi-regional study utilising linked individual patient-level data from two cohorts in Scotland, United Kingdom. All SARS-CoV-2 polymerase chain reaction tests performed during the relevant time periods described for each cohort were included, regardless of whether they were from a hospital or community setting, and irrespective of their indication (e.g. clinical, screening, or research). Each patient's index COVID-19 status was defined as follows: 1) COVID-19 positive – where a patient first recorded a positive SARS-CoV-2 test result, they were included as an index positive episode occurring on that date. In this sub-population, prior or subsequent negative SARS-CoV-2 test results within the study period were excluded, as were subsequent positive SARS-CoV-2 test results. 2) COVID-19 negative – where a patient first recorded a negative SARS-CoV-2 test result, but in the absence of any prior or subsequent positive test result, they were included as an index negative episode occurring on that date. In this sub-population, subsequent negative SARS-CoV-2 test results were also excluded. The date of each index positive or negative episode was assigned as the index date.

Study cohorts

Cohort 1

All patients who had a SARS-CoV-2 test in the NHS Lothian Health Board between 02/01/2020 and 03/31/2021 were identified. Positive and negative COVID-19 episodes were linked with regional electronic patient and biochemistry records and national hospitalisation, dispensed community prescription and death records within the DataLoch Repository and Safe Haven (University of Edinburgh/NHS Lothian, Scotland) (Appendix).

Cohort 2

All patients who had had a measure of serum creatinine in the NHS Fife or Tayside Health Boards since 04/01/2009, and subsequently had a SARS-CoV-2 test between 02/28/2020, and 02/08/2021, were identified. Positive and negative COVID-19 episodes were defined as for Cohort 1 before being linked with national hospitalisation, dispensed community prescription, and death records within the Health Informatics Centre Safe Haven (University of Dundee/NHS Fife and Tayside, Scotland) (Appendix) [17].

Determination of patient demographics, CKD status, comorbidities and causes of death

Patient age, sex and socio-economic status were determined from linked hospitalisation records. Socio-economic status was defined according to the Scottish Index of Multiple Deprivation – a validated measure of social deprivation determined by factors related to residential address (zip code) (Appendix) [18]. CKD status was determined at the time of index SARS-CoV-2 test utilising the same, previously validated criteria for both study cohorts [19]. eGFRs were calculated for all serum creatinine results using the Chronic Kidney Disease Epidemiology (CKD-EPI) Collaboration equation [20]. CKD was defined when a patient's most recent eGFR was <60 mL·min−1/1.73 m2 and at least one value obtained >90 days prior was also <60 mL·min−1/1.73 m2. Using the eGFR value closest to the index date for each patient, CKD stage was classified according to “Kidney Disease: Improving Global Outcomes” guidelines [21]. Patients with kidney failure requiring kidney replacement therapy (i.e. haemodialysis, peritoneal dialysis or kidney transplantation) were identified from record linkage with regional or national renal registries (Cohort 1 – VitalData; Cohort 2 – Scottish Renal Registry). Those patients with only a single measure of eGFR<60 mL·min−1/1.73 m2 prior to their index test were excluded (Supplementary Figure 1).

Patient comorbidities (i.e. angina, atrial fibrillation, cancer, chronic liver disease, chronic lower respiratory disease, heart failure, myocardial infarction, and stroke) were defined from International Classification of Diseases (ICD) codes associated with hospitalisations during a 5-year “lookback” period prior to the index SARS-CoV-2 test (Appendix). Diabetes status was obtained via record linkage with a national diabetes registry (Scottish Care Information – Diabetes Collaboration) [22]. History of prescribed medications was determined from the community prescribing records of individual patients during the 6 months preceding their index SARS-CoV-2 test. Causes of death were determined following the identification of relevant ICD codes in linked National Records of Scotland death records (Appendix).

Study follow-up and outcomes

Patients were followed-up from the index date until either their date of death or April 30, 2021 (Cohort 1) or May 20, 2021 (Cohort 2), whichever came first. For both study cohorts, information relating to primary and secondary causes of death was obtained via record linkage with the National Records of Scotland death registry (Appendix). Primary outcomes included cardiovascular, COVID-19-related, and all-cause death. Secondary outcomes included subsequent fatal and non-fatal myocardial infarction, heart failure or stroke, hospitalisation for cardiovascular diagnoses, hospitalisation for any reason and length of hospital stay. For each cohort, outcomes were reported at 30 days, 90 days and to the end of follow-up.

Statistical analysis

Summary statistics

Baseline clinical characteristics and crude outcomes for all index SARS-CoV-2 tests included in each cohort were summarised according to CKD and COVID-19 status. Continuous variables were presented as medians and interquartile ranges and categorical variables were presented as percentages. Where appropriate, groupwise comparisons were performed using Chi-square tests.

Covariate-balanced propensity scoring and regression modelling

With the aim of obtaining unconfounded estimates, we utilised a “doubly-robust” estimator with bootstrapped standard errors and 95% confidence intervals in our evaluation of the primary outcomes (Appendix) [23]. This approach combines a multivariable outcome regression model with weighting by the covariate-balanced propensity score (CBPS). For the primary analysis, Cox regression was used to explore the association between COVID-19 status (the primary exposure) and cardiovascular and all-cause death (the primary outcomes). For the secondary analysis, Cox regression was used to explore the association between CKD status (the primary exposure) and cardiovascular, COVID-19 and all-cause death (the primary outcomes). Confounders were specified a priori and included age, sex, socio-economic status, comorbidities (i.e. angina, myocardial infarction, heart failure, stroke, diabetes, cancer, chronic respiratory disease, chronic liver disease) and selected current medication (ACE-inhibitors/ARBs, immunosuppressant therapy) based on their potential relevance to COVID-19 outcomes [24, 25]. For each primary outcome, hazard ratios derived from CBPS-weighted, adjusted multivariable models in individual cohorts were pooled using a fixed-effects model to obtain an overall meta-estimate (Appendix). Finally, a sensitivity analysis was also performed to evaluate the effect of restricting the cohorts to those patients not hospitalised either in the week before or in the 2 weeks following their index COVID-19 test. All data were analysed using the R statistical programming language (Version 3.6.2, Vienna, Austria).

Ethical approval

The study was performed with the approvals of local Research Ethics Committees and delegated Caldicott Guardians for the NHS Fife, Lothian, and Tayside Health Boards, in accordance with the Declaration of Helsinki (Appendix). Data provision and linkage were carried out by the DataLoch (University of Edinburgh/NHS Lothian, https://www.dataloch.org/) and University of Dundee Health Informatics Centre (HIC, https://www.dundee.ac.uk/hic) within ISO27001 and Scottish Government accredited secure Safe Havens for Cohorts 1 and 2, respectively. Patient consent was not sought as the study utilised fully-anonymised data. Our analysis code is publicly available here.

Results

In total, 86 964 of 102,894 (84.5%) patients tested for SARS-CoV-2 were included in the study (Supplementary Figure 1). There were 36 904 patients (61±21 years; 58.1% women) in Cohort 1 and 50 060 patients (63±20 years; 62.0% women) in Cohort 2. Overall, the distribution of patient demographics and clinical characteristics between each cohort was similar. CKD was present in 15.9% (5853/36 904) of patients in Cohort 1 and 16.4% (8201/50 060) of patients in Cohort 2 (Supplementary Table 1), whilst a positive SARS-CoV-2 test was recorded in 10.1% (3731/36 904) and 9.1% (4556/50 060) of patients in Cohorts 1 and 2, respectively (Supplementary Table 2). In both cohorts, CKD was commoner in patients with COVID-19 than in those without (Cohort 1: 19.7% versus 15.4%; Cohort 2: 19.0% versus 16.1%).

Baseline characteristics of patients with CKD according to COVID-19 status

In patients with CKD, those with COVID-19 were older and more socially deprived than those without COVID-19 (table 1). Although SARS-CoV-2 testing was performed more frequently in women than in men with CKD (Cohort 1: 53.8% versus 46.2%; Cohort 2: 59.5% versus 40.5%) (Supplementary Table 3), the proportion of women testing positive and negative was similar (Cohort 1: 52.3% versus 54.0%; Cohort 2: 58.0% versus 59.7%) (table 1). Patients with CKD who tested positive had higher rates of cardiovascular comorbidity but similar eGFR compared to patients who tested negative. Despite this, across both cohorts, patients with CKD who tested positive were less likely to be prescribed an ACE-inhibitor or ARB at the time of their index SARS-CoV-2 test than those who tested negative (Cohort 1: 35.8% versus 41.5%; Cohort 2: 25.2% versus 36.7%) (table 1). The baseline characteristics of patients without CKD summarised according to COVID-19 status (table 1), and of patients with and without COVID-19 summarised according to CKD status (Supplementary Table 4), are described in the Supplementary Results.

TABLE 1

Clinical characteristics of patients included in Cohorts 1 and 2, grouped according to CKD and COVID-19 status

	COHORT 1				COHORT 2
	CKD		No CKD		CKD		No CKD
	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative
Number of patients, n	734	5119	2997	28 054	865	7336	3691	38 168
Age, years	81 (11)	79 (12)	59 (21)	57 (20)	84 (10)	82 (10)	59 (20)	59 (19)
Sex
Women	384 (52.3)	2764 (54.0)	1726 (57.6)	16 584 (59.1)	502 (58.0)	4377 (59.7)	2453 (66.5)	23 696 (62.1)
Men	350 (47.7)	2355 (46.0)	1271 (42.4)	11 470 (40.9)	363 (42.0)	2959 (40.3)	1238 (33.5)	14 472 (37.9)
SIMD quintile
1 (most deprived)	103 (14.0)	616 (12.0)	476 (15.9)	4263 (15.2)	162 (18.7)	1160 (15.8)	831 (22.5)	7522 (19.7)
2	207 (28.2)	1280 (25.0)	795 (26.5)	6872 (24.5)	185 (21.4)	1391 (19.0)	747 (20.2)	7264 (19.0)
3	113 (15.4)	920 (18.0)	532 (17.8)	4945 (17.6)	182 (21.0)	1491 (20.3)	665 (18.0)	7187 (18.8)
4	112 (15.3)	889 (17.4)	519 (17.3)	4873 (17.4)	216 (25.0)	1977 (26.9)	905 (24.5)	9932 (26.0)
5 (least deprived)	198 (27.0)	1404 (27.4)	675 (22.0)	6883 (24.5)	120 (13.9)	1317 (18.0)	543 (14.7)	6262 (16.4)
Co-existing medical conditions
Angina	91 (12.4)	571 (11.2)	125 (4.2)	1130 (4.0)	69 (8.0)	533 (7.3)	98 (2.7)	960 (2.5)
Atrial fibrillation#	–	–	–	–	207 (23.9)	1444 (19.7)	215 (5.8)	2003 (5.2)
Myocardial infarction	124 (16.9)	807 (15.8)	178 (5.9)	1765 (6.3)	91 (10.5)	744 (10.1)	96 (2.6)	1239 (3.2)
Heart failure	196 (26.7)	1152 (22.5)	135 (4.5)	1268 (4.5)	133 (15.4)	847 (11.5)	84 (2.3)	740 (1.9)
Stroke	127 (17.3)	704 (13.8)	257 (8.6)	1729 (6.2)	123 (14.2)	671 (9.1)	237 (6.4)	1454 (3.8)
Diabetes	275 (37.5)	1819 (35.5)	491 (16.4)	3592 (12.8)	277 (32.0)	2388 (32.6)	589 (16.0)	5831 (15.3)
Cancer	159 (21.7)	1372 (26.8)	384 (12.8)	5373 (19.2)	98 (11.3)	858 (11.7)	154 (4.2)	2415 (6.3)
Chronic lower respiratory disease	223 (30.4)	1418 (27.7)	809 (27.0)	7411 (26.4)	232 (26.8)	1877 (25.6)	634 (17.2)	7797 (20.4)
Chronic liver disease	27 (3.7)	213 (4.2)	77 (2.6)	984 (3.5)	21 (2.4)	171 (2.3)	40 (1.1)	590 (1.5)
Renal history
Kidney failure	35 (4.8)	302 (5.9)	–	–	29 (3.4)	211 (2.9)	–	–
Baseline eGFR	40 (14)	42 (14)	90 (19)	91 (18)	43 (12)	44 (12)	93 (18)	94 (19)
Baseline eGFR category
≥90	–	–	1396 (46.6)	14 090 (50.2)	–	–	2020 (54.7)	20 764 (54.4)
60–89	–	–	1601 (53.4)	13 964 (49.8)	–	–	1671 (45.3)	17 404 (45.6)
45–59	314 (42.8)	2394 (46.8)	–	–	436 (50.4)	3828 (52.2)	–	–
30–44	236 (32.2)	1632 (31.9)	–	–	270 (31.2)	2279 (31.1)	–	–
15–29	118 (16.1)	669 (13.1)	–	–	105 (12.1)	809 (11.0)	–	–
≤15	66 (9.0)	424 (8.3)	–	–	54 (6.2)	420 (5.7)	–	–
Current medication
ACE-inhibitor or ARB	263 (35.8)	2126 (41.5)	674 (22.5)	5959 (21.2)	218 (25.2)	2695 (36.7)	589 (16.0)	7171 (18.8)
Aspirin#	–	–	–	–	157 (18.1)	1599 (21.8)	299 (8.1)	3437 (9.0)
Other antiplatelet agent#	–	–	–	–	128 (14.8)	861 (11.7)	247 (6.7)	2099 (5.5)
Beta-blockers#	–	–	–	–	300 (34.7)	2655 (36.2)	497 (13.5)	5862 (15.4)
Immunosuppressants	27 (3.7)	201 (3.9)	66 (2.2)	566 (2.0)	11 (1.3)	151 (2.1)	13 (0.4)	246 (0.6)
Loop diuretic#	–	–	–	–	303 (35.0)	2312 (31.5)	264 (7.1)	2553 (6.7)
Mineralocorticoid receptor antagonist#	–	–	–	–	66 (7.6)	558 (7.6)	64 (1.7)	644 (1.7)
Novel oral anticoagulant#	–	–	–	–	142 (16.4)	1136 (15.5)	154 (4.2)	2018 (5.3)
Warfarin#	–	–	–	–	43 (5.0)	499 (6.8)	45 (1.2)	651 (1.7)

Values are mean±sd, n (%), or median [interquartile range]. Abbreviations: ACE: angiotensin converting enzyme; ARB: angiotensin receptor blocker; CKD: chronic kidney disease (please refer to Methods section for definition of CKD status); eGFR: estimated glomerular filtration rate (mL·min−1/1.73 m2); SIMD: Scottish index of multiple deprivation.

#data not available for Cohort 1.

Outcomes of patients with CKD according to COVID-19 status

In patients with CKD, the crude rate of cardiovascular death at 30 days in those with COVID-19 was double that of patients without COVID-19 (Cohort 1: 7.8% versus 3.4%; Cohort 2: 7.2% versus 3.5%) (table 2; figure 1a; Supplementary Table 5). After balancing differences in covariates between positive and negative patients (Supplementary Figure 2) – and following adjustment for confounders – this increase in short-term cardiovascular risk persisted and was more than two-fold higher in positive than in negative patients at 30 days (cause-specific hazard ratio [csHR] meta-estimate 2.34, 95% confidence interval [CI] 1.83 to 2.99) (figure 2a). By the end of study follow-up, the difference in cardiovascular risk between positive and negative patients with CKD had narrowed (csHR meta-estimate 1.57, 95% CI 1.31 to 1.89) (figures 1a and 2a).

TABLE 2

Outcomes of patients included in Cohorts 1 and 2, grouped according to CKD and COVID-19 status

	COHORT 1				COHORT 2
	CKD		No CKD		CKD		No CKD
	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative	COVID-19 positive	COVID-19 negative
Number of patients, n	734	5119	2997	28 054	865	7336	3691	38 168
Primary outcomes
Cardiovascular death
30 days	57 (7.8)	172 (3.4)	57 (1.9)	254 (0.9)	62 (7.2)	254 (3.5)	55 (1.5)	316 (0.8)
90 days	72 (9.8)	290 (5.7)	68 (2.3)	377 (1.3)	72 (8.3)	453 (6.2)	70 (1.9)	515 (1.3)
End of study follow-up	86 (11.7)	426 (8.3)	86 (2.9)	542 (1.9)	92 (10.6)	730 (10.0)	94 (2.5)	840 (2.2)
COVID-19-related death
30 days	250 (34.1)	–	377 (12.6)	–	295 (34.1)	–	397 (10.8)	–
90 days	267 (36.4)	–	404 (13.5)	–	313 (36.2)	–	423 (11.5)	–
End of study follow-up	270 (36.8)	–	407 (13.6)	–	318 (36.8)	–	427 (11.6)	–
All-cause death
30 days	260 (35.4)	442 (8.6)	399 (13.3)	974 (3.5)	302 (34.9)	642 (8.8)	405 (11.0)	1025 (2.7)
90 days	304 (41.4)	803 (15.7)	473 (15.8)	1702 (6.1)	342 (39.5)	1157 (15.8)	466 (12.6)	1782 (4.7)
End of study follow-up	346 (47.1)	1246 (24.3)	541 (18.1)	2688 (9.6)	397 (45.9)	1880 (25.6)	547 (14.8)	3007 (7.9)
Secondary outcomes
Fatal/non-fatal myocardial infarction
30 days	7 (1.0)	132 (2.6)	8 (0.3)	254 (0.9)	9 (1.0)	154 (2.1)	NA	386 (1.0)
90 days	8 (1.1)	155 (3.0)	9 (0.3)	293 (1.0)	9 (1.0)	193 (2.6)	9 (0.2)	428 (1.1)
End of study follow-up	11 (1.5)	190 (3.7)	15 (0.5)	351 (1.3)	13 (1.5)	248 (3.4)	11 (0.3)	505 (1.3)
Fatal myocardial infarction
End of study follow-up	NA	62 (1.2)	NA	74 (0.3)	7 (0.8)	113 (1.5)	10 (0.3)	142 (0.4)
Fatal/non-fatal heart failure
30 days	16 (2.2)	117 (2.3)	NA	133 (0.5)	21 (2.4)	348 (4.7)	6 (0.2)	362 (0.9)
90 days	21 (2.9)	168 (3.3)	8 (0.3)	166 (0.6)	23 (2.7)	423 (5.8)	8 (0.2)	414 (1.1)
End of study follow-up	24 (3.3)	226 (4.4)	10 (0.3)	212 (0.8)	28 (3.2)	496 (6.8)	14 (0.4)	478 (1.3)
Fatal heart failure
End of study follow-up	18 (2.5)	88 (1.7)	NA	60 (0.2)	20 (2.3)	182 (2.5)	11 (0.3)	93 (0.2)
Fatal/non-fatal stroke
30 days	10 (1.4)	96 (1.9)	18 (0.6)	348 (1.2)	15 (1.7)	249 (3.4)	31 (0.8)	574 (1.5)
90 days	13 (1.8)	121 (2.4)	23 (0.8)	415 (1.5)	19 (2.2)	294 (4.0)	38 (1.0)	648 (1.7)
End of study follow-up	16 (2.2)	158 (3.1)	28 (0.9)	487 (1.7)	25 (2.9)	348 (4.7)	44 (1.2)	744 (1.9)
Fatal stroke
End of study follow-up	6 (0.8)	57 (1.1)	14 (0.5)	135 (0.5)	NA	263 (3.6)	NA	410 (1.1)
Fatal/non-fatal pulmonary embolism
End of study follow-up	NA	60 (1.2)	15 (0.5)	286 (1.0)	7 (0.8)	87 (1.2)	17 (0.5)	370 (1.0)
Fatal pulmonary embolism
End of study follow-up	NA	11 (0.2)	NA	20 (0.1)	NA	11 (0.1)	NA	41 (0.1)
Atrial fibrillation hospitalisations
30 days	NA	189 (3.7)	7 (0.2)	461 (1.6)	7 (0.8)	259 (3.5)	10 (0.3)	626 (1.6)
90 days	12 (1.6)	225 (4.4)	13 (0.4)	557 (2.0)	12 (1.4)	303 (4.1)	13 (0.3)	714 (1.9)
End of study follow-up	15 (2.0)	292 (5.7)	18 (0.6)	676 (2.4)	24 (2.8)	435 (5.9)	24 (0.6)	939 (2.5)
Cardiovascular hospitalisations
30 days	28 (3.8)	630 (12.3)	53 (1.8)	1847 (6.6)	41 (4.7)	1311 (17.9)	80 (2.2)	3076 (8.1)
90 days	41 (5.6)	765 (14.9)	82 (2.7)	2175 (7.8)	49 (5.7)	1483 (20.2)	100 (2.7)	3381 (8.9)
End of follow-up	54 (7.4)	958 (18.7)	105 (3.5)	2549 (9.1)	60 (6.9)	1636 (22.3)	119 (3.2)	3685 (9.7)
All hospitalisations
30 days	274 (37.3)	2928 (57.2)	896 (29.9)	13 361 (47.6)	359 (41.5)	5063 (69.0)	899 (24.4)	18 562 (48.6)
90 days	335 (45.6)	3408 (66.6)	1128 (37.6)	15 767 (56.2)	388 (44.9)	5267 (71.8)	970 (26.3)	19 242 (50.4)
End of follow-up	386 (52.6)	3887 (75.9)	1311 (43.7)	17 952 (64.0)	411 (47.5)	5429 (74.0)	1043 (28.3)	19 911 (52.2)
Length of stay	12 [5, 24]	5 [2, 12]	7 [3, 18]	3 [1, 7]	9 [4, 17]	6 [2, 16]	5 [2, 15]	2 [0, 7]

Values are median [interquartile range] or n (%). NAs represent potentially identifiable or count data <5, redacted in order to protect patient confidentiality.

FIGURE 1

Survival curves for cardiovascular (a) and all-cause (b) death according to COVID-19 status (i.e. positive versus negative) for patients with CKD (Panel a: Cohort 1; Panel c: Cohort 2) and patients without CKD (Panel b: Cohort 1; Panel d: Cohort 2). Note: y-axis scales are different for cardiovascular and all-cause death plots.

FIGURE 2

Forest plot summarising adjusted hazard ratios (HR) from Cohorts 1 and 2 and associated pooled meta-estimates for cardiovascular (a) and all-cause death (b) according to COVID-19 status (i.e. positive versus negative) for patients with CKD (red) and patients without CKD (blue) at 30 days (top panel), 90 days (middle panel) and to the end of study follow-up (bottom panel).

In patients with CKD, the risk of all-cause death at 30 days in those with COVID-19 was substantially higher than in those without COVID-19 (Cohort 1: 35.4% versus 8.6%; Cohort 2: 34.9% versus 8.8%) (table 2; figure 1b; Supplementary Table 5). In the fully adjusted models, the risk of all-cause death in patients with CKD and COVID-19 was increased more than four-fold at 30 days (HR meta-estimate 4.53, 95% CI 3.97 to 5.16), and by more than two-fold overall (HR meta-estimate 2.41, 95% CI 2.17 to 2.64) compared to those with CKD testing negative. In contrast, cardiovascular complications and subsequent hospitalisations were lower in positive than in negative patients with CKD at 30 days, 90 days and to the end of study follow-up (table 2).

In a sensitivity analysis restricted to patients with CKD not hospitalised either in the week before or in the 2 weeks following their index COVID-19 test, overall rates of cardiovascular and all-cause death were lower than those reported in the primary analysis (Supplementary Table 6). However, COVID-19 was associated with a significantly increased risk of cardiovascular and all-cause death at all time points, especially in the short-term – a pattern which was comparable to the primary analysis (Supplementary Table 7).

Outcomes of patients with COVID-19 according to CKD status and eGFR

In patients with COVID-19, those with CKD had a higher risk of cardiovascular death than those without CKD (csHR meta-estimate 1.64, 95% CI 1.29 to 2.10) (Supplementary Table 8; Supplementary Figures 3 & 4). Similarly, CKD was associated with a significantly increased risk of all-cause death in patients testing positive (csHR meta-estimate 1.25, 95% CI 1.12 to 1.41) (Supplementary Table 8; Supplementary Figures 3 & 4). When eGFR was analysed as a continuous variable, the risk of both cardiovascular and all-cause death increased as kidney function declined (Supplementary Figure 5).

In patients with COVID-19, CKD was associated with an increased risk of COVID-19-related death (csHR meta-estimate 1.27, 95% CI 1.12 to 1.43) (Supplementary Table 8 & Supplementary Figure 6). Again, the risk of COVID-19-related death increased significantly as eGFR declined, even after adjustment for confounders (Supplementary Figure 7). Rates of cardiovascular complications and subsequent hospitalisations were also higher in COVID-19 positive patients with CKD than without (Supplementary Table 8).

The outcomes of patients without CKD summarised according to COVID-19 status (table 2; figures 1a and 2a), and of patients without COVID-19 summarised according to CKD status and eGFR (Supplementary Table 8; Supplementary Figures 3 & 4), are described in the Supplementary Results.

Discussion

In this multi-regional data-linkage study, we utilised a robust statistical approach combining multivariable outcome regression with propensity score weighting to evaluate the outcomes of ∼87 000 patients with and without CKD tested for SARS-CoV-2. Overall, one-in-ten patients had a positive SARS-CoV-2 test. In patients with CKD, those with COVID-19 had a higher risk of cardiovascular and all-cause death than those without COVID-19 throughout follow-up, but especially within 30 days of SARS-CoV-2 testing. During this early period, patients with CKD and COVID-19 had a more than two-fold increased risk of cardiovascular death – and a more than four-fold increased risk of all-cause death – than CKD patients testing negative. Compared to patients without CKD, those with CKD were also more likely to test positive. Following a positive test, CKD patients had higher rates of cardiovascular complications, including hospitalisations, and cardiovascular death than those without CKD. Moreover, the risks of cardiovascular, COVID-19-related, and all-cause death increased as kidney function declined.

Our study has several strengths. First, its multi-regional design combined high-fidelity, high-quality Scottish linked healthcare data from patients undergoing community and hospital-based SARS-CoV-2 testing in three large NHS Health Boards (which together provide care for ∼1.7 million people), irrespective of age, sex, socio-economic, kidney function, or hospitalisation status. Thus, the influence of case selection bias on our patient cohorts was minimised. Moreover, the accuracy and completion rates of the data sources used in this study were recently reported as 96% [26] and 99% [27], respectively. Second, we utilised routinely-collected biochemistry data and criteria previously validated in electronic health records [19] to determine baseline kidney function, reducing the potential for misclassification of CKD status. Third, our inclusion of control populations – COVID-19 negative patients and patients without CKD – alongside our use of a “doubly-robust” estimator (concomitant multivariable outcome regression and weighting by the propensity score), limited the influence of confounding bias in our analyses [23, 28].

Whilst the majority of patients with COVID-19 are considered to have increased cardiovascular risk [29], few studies have examined the nature or extent of this risk in patients with CKD [30]. This is important given the well-recognised association between cardiovascular disease and CKD [12, 13]. Here, in ∼14 000 non-hospitalised and hospitalised patients with CKD tested for SARS-CoV-2, we found that COVID-19 more than doubled the risk of cardiovascular death within 30 days, and by 57% overall. Our secondary analysis showed that patients with COVID-19 and CKD had an increased risk of fatal and non-fatal myocardial infarction, heart failure and stroke compared to patients with COVID-19 but no CKD. In an adjusted model, CKD was also associated with a 64% increased risk of cardiovascular death in patients with COVID-19. In contrast, Rao and colleagues investigated the risk of cardiovascular complications in patients with COVID-19 but found no increase in risk in patients with CKD compared to those without CKD [31]. However, this study excluded non-hospitalised patients and patients who had tested negative for SARS-CoV-2, and relied on manual case note review to determine CKD status.

Our data add to the literature on COVID-19 outcomes in high-risk populations. However, a novel aspect of our approach is the inclusion of patients with CKD who tested negative for SARS-CoV-2. Of the few studies that have included such patients, all have identified COVID-19 as being strongly associated with a poor prognosis. Indeed, a recent meta-analysis found that COVID-19 increased the odds of death approximately six-fold in patients with CKD [32]. This is more in-line with the risk of all-cause death we report in patients with CKD and COVID-19 at 30 days post-index test, consistent with the fact that most studies included in this meta-analysis reported in-hospital mortality only. Thus, our study is unique in reporting both short- and medium-term patient outcomes. Our study is also less biased towards severe COVID-19 as we included both non-hospitalised and hospitalised patients, making our results more accurate, representative, and informative for patients with all severities of CKD and COVID-19.

We found that, compared to patients without CKD, those with CKD were more likely to test positive for COVID-19. Thereafter, patients with COVID-19 and CKD had higher rates of fatal and non-fatal myocardial infarction, heart failure and stroke, cardiovascular hospitalisations, and cardiovascular, COVID-19-related, and all-cause death than patients with COVID-19 but no CKD. Few studies have reported on all these aspects. Our data are consistent with reports of an increasing incidence of COVID-19 as kidney function declines [33, 34]. A number of factors likely contribute to this increased risk of COVID-19 in CKD, including case ascertainment bias (i.e. patients with CKD are more likely to be tested for SARS-CoV-2), greater viral exposure secondary to more frequent healthcare encounters (e.g. in-centre haemodialysis) [11], and an underlying increased predisposition to infection due to altered immune response [35].

We recognise some limitations. First, our inability to account for selected variables (e.g. body mass index, smoking status, type of atrial fibrillation [i.e. paroxysmal versus permanent; non-valvular or valvular]) means that we cannot exclude the potential for residual confounding. In addition, selected data relating to co-existing atrial fibrillation and prescribed cardiovascular medications were not available for patients included in Cohort 1. Second, rates of non-fatal cardiovascular complications, including atrial fibrillation and cardiovascular hospitalisations, may have been under-reported due to the competing risk of death in patients with COVID-19. Given the crude rates of non-fatal cardiovascular complications were generally higher in patients without COVID-19 suggests that this might be the case. We overcame this when evaluating cardiovascular and COVID-19-related death by calculating cause-specific hazard ratios from our regression models. Third, we excluded patients with no record of kidney function and those with only a single eGFR<60 mL·min−1/1.73 m2 during the biochemistry “lookback” period. Consequently, young or less comorbid patients – who are less likely to have had their kidney function tested – may be relatively under-represented. Finally, those patients with CKD who tested negative for SARS-CoV-2 were a relatively “sick” control group; their all-cause mortality was substantially higher than might be expected for the CKD population in general [36]. One explanation for this is that SARS-CoV-2 PCR testing was largely restricted to the in-hospital setting for much of the early phase of the COVID-19 pandemic in the UK [37]. To address this issue – and to illustrate the effect of including a more widely representative control group – we performed a sensitivity analysis restricted to those patients not hospitalised either in the week before or in the 2 weeks following their index COVID-19 test, and demonstrated a similar pattern of increased risk of cardiovascular and all-cause death in patients with COVID-19 compared to those without.

Conclusions

Our unique and comprehensive analysis suggests that COVID-19 significantly increases the risk of cardiovascular complications and death in patients with CKD, especially in the short-term. There is an urgent need to prioritise COVID-19 vaccination and cardiovascular risk reduction strategies in all patients with CKD.