Stratifying stroke severity: towards a personalised medicine application for primary care

Akyea, Ralph Kwame (2022) Stratifying stroke severity: towards a personalised medicine application for primary care. PhD thesis, University of Nottingham.

[img] PDF (Thesis with corrections - for internal assessor's approval) (Thesis - as examined) - Repository staff only until 30 June 2024. Subsequently available to Repository staff only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Available under Licence Creative Commons Attribution.
Download (23MB)


Stroke remains a major cause of death and disability worldwide, despite advances

in prevention and treatment. Improvements in acute care have led to many

surviving after an incident stroke event. However, the prognosis after surviving

remains compromised. This is due to the high risk of recurrent adverse

cardiovascular events, greatest during the first year but persisting over one’s

lifetime. Reducing long-term residual cardiovascular risk and improving quality of

life are primary goals for clinical practice and research. Identifying patients at the

greatest risk of subsequent major adverse cardiovascular events (MACE) could

help clinicians and policymakers determine which patients need to be prioritised.

This thesis research aimed to identify clinical phenotypes (that is, patient

characteristics and distinct patient clusters) that correlate with subsequent MACE

outcomes (defined as a diagnosis of either CHD, recurrent stroke, PVD, heart

failure, or CVD-related mortality) in adults with an incident stroke diagnosis.

Firstly, a systematic review was completed to identify and summarise the available

evidence on prognostic models and assess their accuracy for predicting MACE

outcomes in an adult with established stroke. Forty (40) full-text articles with 23

distinct prognostic models for predicting MACE outcomes in adults with established stroke were identified by the systematic review. There were 11 prognostic model developments and 77 external validations of models reported. Among the 23 models, the most frequently used predictors were age, sex, history of transient ischaemic attack, hypertension (blood pressure), and diabetes. Critical appraisal identified methodological limitations, in particular: inadequate sample size, improper handling of missing data, and incomplete evaluation of model


The Clinical Practice Research Datalink (CPRD GOLD), a longitudinal database of

anonymised electronic health records (UK primary care data) linked to Hospital

Episode Statistics (HES APC), national death registry, and social deprivation data

was then used to undertake a series of data-related studies. Four cohort studies

were completed using patients aged ³18 years with an incident stroke diagnosis

between 1 January 1998 and 31 December 2017, and no prior history of either

CHD, PVD or heart failure, to assess the risk of subsequent cardiovascular

morbidity and mortality outcomes.

In the analysis of 9,997,376 individual records in CPRD GOLD database, there

were 82,774 non-fatal incident stroke events recorded in either primary care or

hospital data – a stroke incident rate of 109.20 per 100,000 person-years (95%

CI: 108.46 – 109.95). Of the 82,774 patients, 13,879 (16.8%) patients had a

prior history of major adverse outcomes (CHD, PVD, and heart failure) and were

excluded. Subsequent MACE was recorded in 47,500 (69.0%) of the remaining

68,877 patients. In the UK, the incidence of stroke and subsequent major adverse

cardiovascular morbidity and mortality outcomes were higher in women, older

populations, and people living in socially deprived areas.

After excluding patients with stroke not-otherwise specified (n=36,551) and

adjusting for potential confounders, patients with incident haemorrhagic stroke

(n=6,535, 20.4%) had no significantly different risk of subsequent cardiovascular

morbidity, compared with patients with incident ischaemic stroke (n=25,556,

79.6%) – CHD [HR 0.86, 95% CI 0.56 – 1.32], recurrent stroke [HR 0.92, 95%

CI 0.83 – 1.02], PVD [HR 1.15, 95% CI 0.56 – 2.38], or heart failure [HR 1.03,

95% CI 0.61 – 1.74]. However, patients with incident haemorrhagic stroke had a

significantly higher risk of subsequent CVD-related mortality [HR 2.35, 95% CI

2.04 – 2.72] and all-cause mortality [HR 2.16, 95% CI 1.94 – 2.41]. Propensityscore

matched analysis of 1,039 patients with haemorrhagic stroke and 1,039 with

ischaemic stroke showed similar risk in subsequent cardiovascular morbidity

outcomes – CHD, recurrent stroke, PVD and heart failure.

Obesity, a risk factor for stroke and is also a risk factor for hypertension and

diabetes (known risk factors for CVD), is commonly measured using body mass

index (BMI). In a multivariable analysis of a cohort of 30,702 patients with incident

stroke and BMI record, individuals in higher BMI categories were associated with

lower risk of subsequent:

• MACE [overweight (BMI: 25.0-29.9 kg/m2): HR 0.96, 95% CI 0.93 – 0.99)],

• PVD [overweight: HR 0.65, 95% CI 0.49 – 0.85; obesity class III (BMI: ³40

kg/m2): HR 0.19, 95% CI 0.50 – 0.77],

• CVD-related mortality [overweight: HR 0.80, 95% CI 0.74 – 0.86; obesity

class I (BMI: 30.0-34.9 kg/m2): HR 0.79, 95% 0.71 – 0.88; class II (BMI:

35.0-39.9 kg/m2): HR 0.80, 95% CI 0.67 – 0.96]; and

• all-cause mortality [overweight: HR 0.75, 95% CI 0.71 – 0.79; obesity class

I: HR 0.75, 95% CI 0.70 – 0.81; class II: HR 0.77, 95% CI 0.68 – 0.86]

when compared with those with normal BMI. The results were similar irrespective

of sex, smoking status, history of diabetes mellitus or cancer at the time of

incident stroke.

Using a combination of data-driven feature selection approaches and clinical

expert opinion, 39 out of 336 characteristics (clinical features including

sociodemographic, biochemical, comorbid conditions, and prescribed medications

related to stroke or CVD) at the time of incident stroke were selected. An

unsupervised machine learning approach [clustering algorithm for mixed (both

categorical and continuous) data] was used to identify 4 phenotypic clusters for a

cohort of 48,114 patients with incident stroke and subsequent outcomes occurring

30 days after incident stroke. Cluster 1 (n=5,201, 10.8%) was a cohort with high

prevalence of CHD-related risk factors and prescribed medications; cluster 2

(n=18,655, 38.8%) a cohort with low prevalence of multiple long-term conditions

(MLTC); cluster 3 (n=10,244, 21.3%) a cohort with high prevalence of MLTC; and

cluster 4 (n=14,014, 29.1%), the oldest population cohort and predominantly

female. The phenotypic clusters had different incidences and risks for subsequent

cardiovascular morbidity and mortality outcomes. For instance, the incidence of

the composite outcome of recurrent stroke and CVD-related mortality was lowest

in cluster 1 and highest in cluster 4 (15.13 and 23.17 per 100 person-years,

respectively). The risk of subsequent recurrent stroke + CVD-related mortality

was significantly increased in cluster 2 (HR 1.07, 95% CI 1.02 – 1.12); cluster 3

(HR 1.20, 95% CI 1.14 – 1.26), and cluster 4 (HR 1.29, 95% CI 1.26 – 1.33),

when compared with cluster 1.

Findings from this thesis research indicate patients with incident stroke experience

considerable heterogeneity in subsequent clinical outcomes. In particular, women,

older patients, and those living in socially deprived areas are at greater risk of

subsequent major adverse outcomes. Additionally, age at incident stroke, blood

pressure, LDL cholesterol level, a diagnosis of hypertension and potency of

prescribed statin were identified as key indicators of patients’ phenotypic clusters

and associated risk for subsequent clinical outcomes. The studies add to growing

and wider evidence to identify those who may most benefit from, and be least

likely to be harmed by, preventive treatment. Stratifying patients with stroke

early, could lower the burden of subsequent adverse clinical outcomes, improve

patients’ long-term outcomes, and reduce the associated economic burden. This

should, therefore, be a continuing research and public health priority.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Qureshi, Nadeem
Kai, Joe
Asselbergs, Folkert W.
Keywords: primary care, stroke, secondary prevention, electronic health records, risk stratification
Subjects: W Medicine and related subjects (NLM Classification) > WL Nervous system
Faculties/Schools: UK Campuses > Faculty of Medicine and Health Sciences > School of Medicine
Item ID: 69235
Depositing User: Akyea, Dr Ralph Kwame
Date Deposited: 02 Jul 2022 04:40
Last Modified: 01 Aug 2022 04:40

Actions (Archive Staff Only)

Edit View Edit View