Prognostic impact of heart failure, chronic kidney disease and atherosclerotic disease alone or combined in an elderly population: a retrospective cohort study based on electronic health record data

Abstract

Objectives

Heart failure (HF), chronic kidney disease (CKD) and atherosclerotic cardiovascular disease (ASCVD) are highly prevalent conditions that often coexist. Using electronic health records (EHRs), we evaluated the 1-year risk of all-cause death, major cardiovascular and kidney events in patients with HF, CKD, ASCVD and with combinations of these conditions, compared with an unselected control population aged ≥75 years.

Design

Retrospective cohort study based on EHR data.

Setting

Integrated primary and secondary health unit located in the North of Portugal. Eligible adult patients were identified using EHRs from 2008 to June 2022.

Participants

Eight cohorts were defined: (1) control: patients with ≥75 years; (2) ASCVD alone; (3) HF alone; (4) CKD alone; (5) cardiorenal syndrome (CRS): combined HF+CKD; (6) atherosclerotic HF: combined ASCVD+HF without CKD; (7) atherosclerotic CKD: combined ASCVD+CKD without HF and (8) combined ASCVD+CRS. The risk of these conditions was compared with controls using propensity score age-sex matching. We identified 19 129 patients with ASCVD alone, 13 640 patients with HF alone, 40 545 with CKD alone and 10 499 with CRS. The control group comprised 36 532 patients aged 75 years or older.

Primary and secondary outcome measures

The primary outcome was all-cause mortality. The main secondary outcomes were cardiovascular death, HF hospitalisations and end-stage renal disease.

Results

The 1-year mortality rate was 0.65% in the control cohort, 5.6% for patients with ASCVD alone, 6.05% for patients with HF alone and 3.53% for patients with CKD alone. Adjusted risk of all-cause death was significantly increased in the ASCVD-alone (HR: 8.42, 95% CI 7.12 to 9.95), HF-alone (HR: 9.19, 95% CI 7.75 to 10.9) and CKD-alone (HR: 5.35, 95% CI 4.62 to 6.19) cohorts, compared with control population; however, patients with the combination of all three conditions (ie, ASCVD+CRS) had the highest mortality risk (HR: 14.18, 95% CI 11.62 to 17.3). A similar association pattern was observed for cardiovascular death, HF events and end-stage renal disease.

Conclusions

Our results support the concept of an atherosclerotic cardiorenal phenotype, with a very high risk of mortality, cardiovascular and renal adverse events. Implementation strategies are required to target these conditions simultaneously.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study uses electronic health records (EHRs) to assess the overlap between cardiovascular, kidney and metabolic conditions.

• The prognostic impact of specific disease combinations is analysed.

• The main limitations are the retrospective nature of this study and the use of EHRs that may not allow for fully capturing the details of the populations being studied.

Introduction

Heart failure (HF), chronic kidney disease (CKD) and atherosclerotic cardiovascular disease (ASCVD) are highly prevalent conditions that often coexist. It is expected that the prevalence of these conditions will increase over time along with population ageing and rapid global growth of cardiovascular risk factors such as obesity and type 2 diabetes mellitus (T2D).¹

The coexistence of HF, CKD and ASCVD increases the complexity of the management of patients who suffer from a combination of these conditions. Moreover, patients with more than one of these conditions have an increased risk of death compared with patients with one condition alone.² Previous meta-analysis including data on ~1 million HF patients reported a prevalence of CKD of 32%, and patients with both HF and CKD had a twofold higher risk of death compared with patients with HF without CKD.³ A similar pattern of increased mortality risk was observed among patients with concomitant HF and ASCVD, and an even higher risk was observed in patients with HF, CKD and ASCVD.^{4 5}

Most patients with HF, CKD and ASCVD have concomitant T2D and/or obesity, which are also conditions that require adequate management.⁶,⁸ Recent randomised controlled trials showed that targeting obesity leads to a better health status in HF and improves cardiovascular outcomes in T2D with ASCVD.^{9 10}

Relatively novel treatments such as sodium glucose cotransporter 2 inhibitors, glucagon-like peptide-1 receptor agonists and non-steroidal mineralocorticoid receptor antagonists have been shown to improve cardiovascular and renal outcomes in patients with HF, CKD, T2D and obesity with ASCVD.¹¹,¹⁴ The use of combinations of these agents may greatly reduce the risk of adverse events among patients with concomitant HF, CKD and ASCVD.¹⁵ Therefore, diagnosing patients with these conditions, alone or in combination, as well as recognising their characteristics and associated risks, is crucial to identify a target population for such therapies.

Our aim in this analysis is to ascertain the 1-year risk of death, major cardiovascular and kidney events in patients with initial presentation of HF, CKD and ASCVD, when compared with an unselected population aged 75 years and older, followed in an integrated healthcare unit (with both primary and secondary care).

Methods

Study design

This study was conducted by performing a clinical database analysis of an integrated healthcare institution, Unidade Local de Saúde de Matosinhos (ULSM). ULSM is a large healthcare institution that includes 14 primary care centres supported by one hospital that provides secondary care services in the region of Matosinhos, in northern Portugal.¹⁶

Eligible adult patients were identified using electronic health records (EHRs) from 2008 to June 2022.

Patient cohorts

We defined eight incident cohorts as follows: (1) controls: a control cohort of people aged ≥75 years; (2) ASCVD alone: ASCVD without prior HF or CKD; (3) HF alone: HF without prior CKD or ASCVD; (4) CKD alone: CKD without prior HF or ASCVD; (5) cardiorenal syndrome (CRS): combined HF and CKD; (6) atherosclerotic HF: combined ASCVD and HF without CKD; (7) atherosclerotic CKD: combined ASCVD and CKD without HF; and (8) ASCVD+CRS: combined ASCVD and HF and CKD.

The ULSM EHR stores patient information, including demographics and clinical diagnosis according to International Classification of Disease, 9th Revision (ICD-9), International Classification of Disease, 10th Revision (ICD-10), International Classification of Primary Care (ICPC-2), laboratory and imaging test results. All definitions were constructed to include all patients with either ICPC-2, ICD-9 or ICD-10 codes corresponding to the relevant pathologies or with imaging or laboratory criteria that would correspond to the relevant diagnosis.

We defined ASCVD as the presence of at least one of the following: unstable angina, peripheral artery disease, history of myocardial infarction (MI) or ischaemic stroke, as identified by relevant ICPC-2, ICD-9 or ICD-10 codes.

HF was defined either by a relevant ICPC-2, ICD-9 or ICD-10 code and/or an ejection fraction (EF) ≤40% and NT-proBNP ≥200 pg/mL (≥600 pg/mL if atrial fibrillation (AF)) or BNP ≥100 pg/mL (≥125 pg/mL if AF); or EF >40% in the presence of structural cardiac abnormalities, according to the European Society of Cardiology (ESC) guidelines.¹⁷ Still, the use of natriuretic peptide thresholds was slightly higher than those recommended by the ESC guidelines because we are dealing with an elderly population with prevalent CKD.^{18 19} CKD was defined by the presence of a relevant ICD-9 or ICD-10 code and/or eGFR ≤60 mL/min/1.73 m² using the CKD-EPI 2009 formula and/or a urinary albumin-to-creatinine ratio ≥30 mg/g (≥3 mg/mmol).

Patients were indexed at the date of the diagnosis that qualified them for the cohort and followed for 1 year after the diagnosis (ie, HF, CKD and/or ASCVD).

Clinical outcomes

For each cohort, death and major cardiovascular and kidney events were searched for during the one-year period following the index date (ie, the date in which any of HF, CKD or ASCVD was recorded).

Death and major cardiovascular and kidney events were defined by the following outcomes as the main diagnosis: all-cause death; cardiovascular death; hospitalisation for ASCVD; MI; ischaemic stroke; peripheral artery disease; HF composite outcome (a composite of HF hospitalisation or cardiovascular death); end-stage renal disease (ESRD) (defined as having an eGFR <15 mL/min/1.73 m² using the CKD-EPI 2009 formula); acute kidney injury; and a 50% reduction of eGFR. Full definitions of each of these events with the respective ICPC-2, ICD-9 or ICD-10 codes can be found in online supplemental appendix table 5.

Outcomes were migrated (from national registries) to ULSM even if they had occurred outside the ULSM.

Statistical analyses

Continuous variables are described using mean and standard deviation. Categorical variables are described by the percentage of patients in each category. The number of patients with missing data for each variable was reported. Only available data are summarised.

Patients in each of the seven test cohorts were matched 1:1 with the control cohort based on age and sex using a nearest neighbour approach. Each of the seven test cohorts was individually matched to the control cohort, that is, in total, seven different matching processes occurred. Propensity scores were calculated using a generalised linear model. A nearest neighbour (without a specified calliper distance), optimal process strategy was used. The matching process was conducted using a function matching software (MatchIt; R software). Standardised differences of post-matched patient characteristics were used to assess the adequacy of propensity score matching, where<10% standardised difference between the two groups after propensity score matching was observed in each of the seven matching processes (online supplemental appendix figure 21).

HRs were estimated using Cox regression models, adjusted for age and sex, with 95% CIs. Patients were censored at death, on leaving the database, or after 1 year. Patients without registered follow-up did not enter the matched cohorts; therefore, deaths occurring at index date were not considered.

Sensitivity analyses were conducted for subgroups of patients presenting with T2D and/or obesity. As the database could not be sure to capture the correct index dates for T2D and obesity, building cohorts with these patients could introduce bias. Therefore, a sensitivity analysis was performed to assess whether these two metabolic diseases played a role in the risk of developing the studied clinical outcomes.

T2D was defined either by a HbA1c equal to or superior to 6.5 or any blood glucose test superior to 200 mg/dL or prescription of glucose lowering drugs (ATC code) (in patients without a code for type 1 diabetes mellitus). Obesity was defined by a BMI >30 kg/m².

The data available for this study included the full records of public primary care centres in the region as well as the full records of the region’s secondary care centre. In ULSM, data are fully linked between primary and secondary care.

Analysis was conducted through local execution of analytical programmes developed using VERO technology that were compiled for ULSM target data infrastructure built on Apache Spark Framework version 3.2.1 and source data harmonised according to the OMOP CDM 5.4 standard. The package implements a complete data engineering pipeline to transform the source data into the final dataset as well as execution of statistical analysis and generation of results reports.^{16 20}

Data were fully processed at the study site, adhering to all local and regional laws on data protection and privacy. A study report was then automatically generated containing non-identifiable aggregate summary results and shared with the investigators. The output files did not contain any data that allowed identification of subjects included in the study, guaranteeing full anonymisation.

Patient and public involvement

Patients were not involved in the design of this study. The results will be disseminated in primary care and in plain language to patients with similar characteristics to those reported herein.

Results

Cohort characterisation

Before matching, we identified 19 129 patients with ASCVD alone, 13 640 patients with HF alone, 40 545 with CKD alone, 10 499 with CRS, 7 004 with atherosclerotic HF, 11 987 with atherosclerotic and CKD and 5 571 with ASCVD and CRS, according to the definitions described above. Individuals aged 75 years and older (n=36 532) were used as a control group for matching and outcome comparison. There were more women than men in all cohorts except the ASCVD cohort. Participant characteristics before and after matching are depicted in table 1.

Table 1. Characterisation of unmatched and matched cohorts.

	Control	ASCVD	HF	ASCVD+HF	CKD	ASCVD+CKD	CRS	ASCVD+CRS
Characterisation of unmatched cohorts
Patient N	36 532	19 129	13 640	7004	40 545	11 987	10 499	5571
Age, mean (SD)	75.00 (1.41)	71.01 (20.59)	77.40 (15.38)	78.05 (15.37)	70.91 (17.85)	76.97 (14.86)	79.57 (13.00)	79.81 (13.09)
Sex (% males)	40.35	52.65	44.39	49.51	44.28	49.77	41.76	46.65
Type 2 Diabetes (%)	43.79	67.07	74.82	81.45	65.25	76.83	78.75	83.68
Obesity (%)	13.62	17.56	23.66	23.14	21.14	21.81	25.50	24.47
Prior MI (%)	1.90	17.36	14.55	32.84	3.49	20.26	14.35	31.29
Prior Ischaemic Stroke (%)	4.37	49.28	16.80	46.03	6.99	51.62	19.91	49.11
Prior PAD (%)	0.60	6.52	3.99	11.01	1.06	8.79	4.76	12.30
Characterisation of matched cohorts
Patient N	35 619	5 453	4 629	2 252	14 669	3 906	3 457	1 821
Age, mean (SD)	76.66 (3.55)	76.16 (4.29)	76.42 (4.20)	76.27 (3.87)	76.63 (4.83)	75.76 (3.56)	76.19 (3.77)	76.41 (3.82)
Sex (% males)	40.29	43.57	37.33	42.50	38.93	45.14	37.32	41.13

ASCVD, atherosclerotic cardiovascular disease; CKD, chronic kidney disease; CRS, cardiorenal syndrome; HF, heart failure; PAD, peripheral artery disease.

Mortality rates

The 1-year mortality rate was 0.65% in the control cohort, 5.6% in patients with ASCVD alone, 6.05% among patients with HF alone and 3.53% in patients with CKD alone. Adjusted risk of all-cause death was significantly increased in the ASCVD-alone (HR: 8.42, 95% CI 7.12 to 9.95), HF-alone (HR: 9.19, 95% CI 7.75 to 10.9) and CKD-alone (HR: 5.35, 95% CI 4.62 to 6.19) cohorts, compared with control population. However, the combination of all three conditions (ie, ASCVD+CRS) was associated with the highest mortality risk (HR: 14.18, 95% CI 11.62 to 17.3) (figure 1).

Figure 1. . Event risk across cardiovascular and kidney conditions alone and combined. One year all-cause death, cardiovascular death, HF composite outcome and end-stage renal disease hazard. ASCVD, atherosclerotic cardiovascular disease; CKD, chronic kidney disease; CRS, cardiorenal syndrome; HF, heart failure.

Cardiovascular death was also significantly higher in HF-alone, ASCVD-alone and CKD-alone cohorts (the adjusted HRs were 10.09; 95% CI 8.19 to 12.41 for HF, 9.12; 95% CI 7.41 to 11.21 for ASCVD, and 4.44; 95% CI 3.7 to 5.34 for CKD), compared with the control cohort (figure 1). The risk was higher for those with CRS, with an HR of 12.52 (95% CI 10.1 to 15.51) compared with controls and worsened for those with ASCVD+CRS: HR 20.35 (95% CI 16.29 to 25.41).

Major cardiovascular events

The risk of hospitalisation for major adverse cardiovascular events (MACEs) was higher in those with ASCVD alone and CRS compared with those with CKD alone and the control cohort.

The risk of hospitalisation for ASCVD was particularly high in cohorts with ASCVD, either alone or with associated comorbidities.

MI hospitalisation rate in patients with atherosclerotic HF and ASCVD+CRS was particularly high (5.65% and 5.95%, respectively) when compared with CRS without ASCVD (3.38%). The risk of MI in the HF-alone cohort was twofold that observed for patients with CKD alone (HR: 2.00, 95% CI 1.64 to 2.44 for CKD; HR: 5.86, 95% CI 4.75 to 7.24 for HF).

The risk of hospitalisation for ischaemic stroke was, similarly to the risk of hospitalisation for ASCVD, over eight times higher in patients with ASCVD either associated or not with another comorbidity, when compared with the control cohort, whereas the risk of this same outcome in patients with HF-alone or CRS was only 3.8 to 4.5 times higher than the control cohort.

The risk of hospitalisation for peripheral artery disease was highest in the atherosclerotic HF and ASCVD+CRS cohorts.

As for the composite outcome of CV death or HF hospitalisation, the 1-year event rate/100PY in atherosclerotic HF and ASCVD+CRS was 6.5 and 7.97, respectively, while the control cohort was 0.16. This resulted in an HR of 40.75, 95% CI 36.5 to 65.91 for atherosclerotic HF and 49.05, 95% CI 30.44 to 54.55 for ASCVD+CRS.

Forest plots regarding cardiovascular events hazard ratios are available in online supplemental appendix figures 17–20.

Renal outcomes

Patients with CRS, atherosclerotic HF and ASCVD+CRS were more likely to progress to ESRD, with HRs of 13.23 (95% CI 11.38 to 15.39), 13.38 (95% CI 11.28 to 15.88) and 16.96 (95% CI 14.3 to 20.12) respectively, higher than that of patients with only one condition (figure 1). The same pattern was observed for the hazard risk of acute kidney failure, where patients with HF alone (HR: 6.78 95% CI 5.62 to 8.19) were also more likely to present this outcome than those with CKD (HR: 3.31, 95% CI 2.82 to 3.89).

Sensitivity analysis of patients with T2D and obesity was conducted to evaluate the impact of these two metabolic conditions on the risk of death and major cardiovascular and renal outcomes 1 year after index date. Characterisation of sensitivity cohorts is described in online supplemental appendix tables 1–4. Both T2D and obesity seem to have an attenuated risk for both all-cause death and cardiovascular death across all cohorts, as demonstrated in online supplemental appendix figures 1–16. HRs for all-cause death in ASCVD+CRS patients were 9.29, 95% CI 7.08 to 12.18 in the T2D group and 20.48, 95% CI 13.28 to 31.57 in the non-T2D group; 4.98, 95% CI 2.57 to 9.63 in the obesity group and 11.07, 95% CI 7.86 to 15.6 in the non-obese group. A similar pattern was observed for both HF composite outcome and end-stage kidney disease. HRs for the composite of CV death or hospitalisation for HF in ASCVD+CRS patients were 27.27, 95% CI 19.39 to 38.35 in the T2D group and 62.23, 95% CI 26.12 to 148.23 in the non-T2D group; 20.19, 95% CI 10.71 to 38.05 in the obesity group and 31.38, 95% CI 19.66 to 50.07 in the non-obese group.

Discussion

In this study, we examined the EHRs of patients followed at a healthcare unit comprising both primary and secondary medical care, to estimate the risk of death, major cardiovascular events and major renal outcomes within a 1-year follow-up in populations presenting relevant comorbidities such as ASCVD, HF, CKD and their combinations, when compared with a control group of patients 75 years or older. In short, atherosclerotic HF and ASCVD+CRS were associated with the highest risks of HF hospitalisation, ESRD, all-cause and cardiovascular death, while patients with ASCVD had the highest risk of MI and stroke. Regarding kidney outcomes, the presence of CRS seems to be a major risk factor to the development of ESRD and acute kidney injury, and the risk increases for patients who present concomitant ASCVD and CRS. Cohorts with isolated CKD and isolated ASCVD presented the lowest hazard risks of adverse kidney outcomes, which may as well be related to the broad definition (eGFR <60 mL/min) of CKD used in the study. Together, these findings suggest that the coexistence of these conditions (HF, CKD, ASCVD) greatly increases the risk of major adverse cardiovascular outcomes and that strategies are needed to mitigate the cardiovascular risk in these cardio-renal-metabolic patients.^4,5 Furthermore, by comparing the risk of adverse outcomes with an elderly population aged 75 years and older, we are able to show that tailored treatment approaches may be necessary to decrease the risk of poor cardiovascular and renal outcomes among elderly patients with HF, ASCVD and CKD, particularly when two or more of these conditions present concomitantly. Such strategies may not be only related to a better use of pharmacological treatments (such as SGLT2 inhibitors, aldosterone antagonists and/or GLP1 receptor agonists) but may also require a profound re-organisation of care, which should be more multidisciplinary and patient-centred rather than focused on the multiplication of subspecialty appointments. Of course, the present work will not solve the current issues in the management of these elderly patients with CKM comorbidities, but it may serve as a base on which one can benchmark the risk associated with these conditions and then compare the impact of interventions, such as medical therapies and/or re-organisation of care.²¹,²⁴

In our observational study, obesity and T2D (conditions that often coexist) were associated with a lower risk of events. This may be explained by the observational bias called ‘obesity paradox’ whereby obese patients have better survival compared with lean, and more often frail and cachectic patients.²⁵,²⁷ In this context, the ‘lean’ patient group may be disproportionately burdened with comorbidities that lead to both weight loss and a higher risk of adverse events, making obesity appear protective by comparison. Additionally, it is possible that patients with obesity and T2D have a closer follow-up, monitoring and treatment, further accentuating these observational biases. Still, it is important to stress that pharmacological weight loss was associated with improved outcomes among patients with obesity and a high cardiovascular risk.¹¹ Future studies with granular data on comorbidities, medication use and measures of frailty are needed to further explore these complex relationships.

Beyond the frequent co-existence of HF, CKD, ASCVD, T2D and obesity, currently, treatments for these conditions largely overlap. For example, patients with albuminuric CKD and T2D should initiate an SGLT2 inhibitor and a non-steroidal mineralocorticoid receptor antagonist (MRA). Additionally, a glucagon-like peptide receptor agonist (GLP1ra) can be considered in patients with obesity, DM, ASCVD or those with HF and preserved EF in whom obesity is contributing to an impaired health-related quality of life. Furthermore, patients with HF may benefit from an SGLT2 inhibitor, an MRA, a GLP1ra and ACEi/ARBs which are transversal to all these conditions.²² Still, as discussed above, such a holistic approach may require a profound re-organisation of care.

Limitations

This study presents some limitations, notably when it comes to extrapolating results to other populations. ULSM has a patient population of 177 000 individuals in northern Portugal; however, the results of this study may not be generalisable to other populations as characteristics regarding ethnicity and socioeconomic factors may vary. External replication of our findings using EHRs is required as a means to extrapolate these findings to other regions and populations. Additionally, diagnosis codes may be associated with some misclassification. Furthermore, the specific aetiology of diseases (such as HF) and respective treatments cannot be assessed based on the EHRs. The subclassification of HF by ejection fraction subtypes cannot be correctly performed using EHRs because many patients with HF may not have an echocardiogram with ejection fraction reported in the system. The definition of CKD relied on available data, including eGFR, urinary albumin-to-creatinine ratio and ICD codes. Since the availability of albuminuria measurements was not universal across all patients, there is a risk of underestimating CKD in patients without recorded albuminuria, particularly those in GFR stages G1 and G2, leading to a potential underestimation of the overall prevalence of CKD within the source population. However, by including ICD codes, we aimed to capture clinically recognised CKD that might have been missed by laboratory criteria alone. It should also be highlighted that because albuminuria was not available in most patients classified as having CKD, we could not use KDIGO classification for kidney disease.

Conclusions

Our results support the concept of an atherosclerotic cardiorenal phenotype, with a high associated risk of mortality, and poor cardiovascular and renal outcomes. Strategies targeting these conditions are available and largely overlap. Implementation strategies are needed to improve the outcomes of patients with overlapping atherosclerotic cardiorenal conditions.

Data availability statement

Data are available upon reasonable request to the corresponding author.