A retrospective cohort study examining secondary prevention post stroke in primary care in an Asian setting

The healthcare system in Singapore comprises of public, private, and non-profit healthcare institutions delivering a range of services including inpatient and emergency services, intermediate and long-term care (ILTC) services and outpatient services comprising of both specialist and primary care services [19]. The public sector is not only the major provider in acute care setting, covering 80% of tertiary hospitals in Singapore, but it is also central to providing primary care services to individuals with chronic ailments. Following discharge from an inpatient setting, individuals with stroke are followed-up in the community. The public primary care system in Singapore comprises of regional clusters of polyclinics which are “one-stop primary care clinics” providing a range of services including clinical, health education, diagnostic and pharmacy services [20]. The data originates from a time when there were 18 polyclinics across the island which were part of two regional clusters, and we extracted data from one of these regional clusters comprising 10 polyclinics. Given Singapore’s small size and relatively homogeneous distribution of the population, we do not expect the profile of individuals following stroke to substantially vary geographically.

Family physicians in primary care setting play a crucial role in the continuity of care and coordination of services as stroke survivors transition from an inpatient to a community setting. They bring together care professionals and providers to meet the individual’s needs and ensure that the care is integrated across various settings [21]. Continuity and coordination of care following stroke includes management of risk factors like hypertension, hyperlipidaemia and diabetes to prevent stroke recurrence and management of post-stroke mood and behaviour problems. Management of clinical risk factors (i.e., hypertension, hyperlipidaemia and diabetes) is done in accordance with the guidance provided by our Ministry of Health’s clinical practice guidelines [22], which includes initiation and titration of medications, scheduling regular follow-up appointments and monitoring laboratory test results to assess control of clinical indicators. They also give lifestyle modification advise with regards to dietary changes, smoking cessation, weight reduction and physical activity.

This was a retrospective cohort study reviewing data from the institutional electronic healthcare record database. The institutional shared electronic healthcare record facilitates provision of integrated care for patients visiting polyclinics and aids efficient monitoring of patient outcomes. This institutional electronic healthcare record database connects patients’ clinical and administrative data within a cluster of polyclinics providing summary care record including physician visits and laboratory records. Our study included individuals following stroke visiting public primary care setting in Singapore, also known as polyclinics, for follow-up with a family physician. Participants were aged 21 years and above with a past diagnosis of stroke (ICD-9 Classification of Diseases, Ninth Revision) and who received care in any of the National Healthcare Group Polyclinics (NHGP) between 1^st January 2012 to 31^st December 2016. Individuals with ischemic and haemorrhagic strokes were included as the coding of cerebrovascular accidents did not differentiate the cause, however we were able to exclude transient ischemic attacks. For participants meeting the above eligibility criteria, we retrospectively extracted information on exposure variables and outcome variables.

The outcome variables were extracted at the end of 1-year of follow-up from the baseline visit for each participant within the observation period. Our outcome variables of secondary prevention at 1-year post stroke were based on the American Heart Association (AHA) Stroke 2014 guidelines [23]. We operationalized overall control as individuals having a systolic blood pressure < 140 mmHg, and a diastolic blood pressure < 90 mmHg, LDL-cholesterol < 2.6 mmol/L and, if the individuals also had diabetes mellitus, a haemoglobin A1c (HbA1c) < 7.0%. For all individuals meeting the study eligibility criteria, we extracted the systolic and diastolic blood pressure, LDL cholesterol level and HbA1c as continuous variables. Based on the above cut-offs, these parameters were converted into binary categorical variables of ‘yes’ or ‘no’. Similarly, the recommended cut-offs for the parameters were converted to binary variables indicating whether secondary prevention targets were achieved or not. Clinical relevance and information availability in the institutional electronic healthcare record database guided the selection of outcome variables for this study and since the database did not have information on diet, physical activity and alcohol intake, these variables were not included.

For all individuals meeting the study eligibility criteria, we extracted socio-demographic and clinical variables at the baseline visit to the polyclinic. Sociodemographic data included age (continuous variable in years), sex (categorical variable with two categories of ‘male’ and ‘female’), ethnicity (four categories of ‘Chinese’, ‘Malay’, ‘Indian’ and ‘Others’). Clinical data included body mass index (BMI) (continuous variable converted to categorical variable based on the Asian cut-off with three categories of ‘low’ (less than 23 kg/m²), ‘moderate’ (23 to 27.4 kg/m²) and ‘high’ (more than 27.4 kg/m²) risk) and smoking status (categorical variable with two categories of ‘yes’ or ‘no’). In addition, we also captured the presence of the following comorbid clinical conditions (as binary categorical variables of ‘yes’ or ‘no’): hyperlipidemia, hypertension, diabetes, chronic kidney disease (CKD), ischemic heart disease (IHD), and atrial fibrillation (AF). Clinical relevance and information availability in the institutional electronic healthcare record database guided the selection of above mentioned cardiometabolic comorbid conditions.

We used de-identified data to conduct our analysis. We used descriptive analysis to summarize our categorical variables using proportions and frequencies. Complete case analysis was used to study the implementation of treatment goals for secondary prevention in individuals with a prior history of stroke. To describe the factors associated with the achievement of overall control in individuals following stroke, we performed multivariable logistic regression analysis since our outcome variable was binary (overall control achieved versus not), and we controlled for the following pre-decided medically relevant variables: age, sex, ethnicity, BMI, smoking status, presence of IHD, presence of AF and presence of CKD. All the analysis was performed using STATA version14.2, and the statistical significance level was set at 5%.

Published studies tend to focus on singular risk factor reduction rather than the overall control achieved. Comparing our results with single risk factor studies, we found examples of some study populations doing better and others worse. For example, comparing the proportion of individuals following stroke having an optimal level of blood pressure, two studies reported a higher level of control at 89.9% [16], and 86% [26], as compared to our estimate of 71.4%. The second of these studies was a prospective study from Canada in which stroke prevention clinics at a tertiary care setting recruited 119 individuals who were referred to them from primary care for secondary stroke prevention. The relatively higher proportion of individuals with blood pressure control could be due to their study setting being a specialized stroke prevention clinic (as compared to primary care setting in our study) and only a small proportion of all participants in this study had a recent history of stroke (as compared to all participants in our study) [26]. The possible explanation for a higher proportion of participants achieving blood pressure control in the other study may be due to the longer 10-year follow-up period compared to our 1-year follow-up. Moreover, all the participants with stroke in this study had a history of diabetes [16]. Those studies reporting a lower level of blood pressure control as compared to our estimate reported values ranging from 23.8% to 62.4% [10, 18, 27‐29]. Possible explanation for differences could be related to different healthcare systems, care settings and patient characteristics.

Comparing the level of lipid control across different studies, all the reviewed studies reported a lower level of lipid control as compared to our estimate of 66.6%, with reported estimates ranging from 13.9% to 49.0% [16, 18, 26, 28, 29]. We found 52.9% of individuals with diabetes following stroke achieving the target level of glycemic control. Compared to existing literature, two studies reported a higher proportion of glycaemic control [10, 16], and another two reported a lower proportion of glycaemic control [18, 26]. The possible explanation for this difference could be related to the cut-off value used for HbA1c which was 7% in our study (as compared to 7.5% in this compared study) [16]. As for the other study, it included participants from a national survey who self-reported history of stroke [10] and these may be systematically different from the participants in our retrospective record-based study.

Our study filled an important gap in current literature as a limited number of current studies address the level of secondary prevention attained. Among the studies reviewed, only two attempted to report a composite or combined estimate of control of risk factors achieved, with one reporting 3.3% of individuals with ischemic stroke achieving control of all risk factors [18]. Another study reported the proportion of individuals following stroke achieving control of both blood pressure and lipids to be about 19.4% [29]. While our study reported more encouraging levels of secondary prevention achieved, as compared to the above two studies, there remains considerable room for improvement. The reasons for not achieving the optimal level of control of risk factors could be multiple. These could be at the patient level, the physician level, the health system level, or a combination of these. A study reported that in spite of 90% of individuals following stroke being on specific drug regimens, only about one-fourth of them achieved the recommended risk factor control [18]. This highlights the complexity of addressing secondary control following stroke. For example, the individual’s adherence to medication, compliance to lifestyle factors or healthcare system factors such as difficulty in accessing services or financial barriers, can all play a part. Another study based in Canada reported poor control of risk factors after either a coronary artery disease (CAD) or cerebrovascular disease (CVD) [29]. Those with CVD had worse control of risk factors as compared to those with CAD, with the former group having 46.0%, 40.5% and 19.4% of post-individuals with CVD meeting target levels of blood pressure, LDL and both respectively. In spite of good adherence to secondary prevention guidelines (medication rates ranging from 76.5% to 91.3%), it did not translate to the achievement of risk factor control suggesting the importance of other elements like patient factors.

We reported lower overall control in individuals with diabetes following stroke as compared to individuals without diabetes following stroke. There are literature that support poorer control of other risk factors in individuals following stroke who have diabetes. Individuals with diabetes were associated with lower odds (OR = 0.16; 95% CI: 0.14, 0.19) of achieving the target blood pressure level compared to individuals with a previous cerebrovascular event [29]. Another study reported lower levels of control of lipids in individuals with diabetes [16]. Another possible explanation could be the difficulty managing co-occurrence of multiple chronic conditions experienced by both healthcare providers and patients.

Our finding was in agreement with other studies that showed, a significant association between sex and achievement of target levels of risk factors [16, 29]. One such study reported the largest difference across males and females in the achievement of the target level of serum LDL levels, with 46.1% of men and 38.3% of women achieving the target [16]. It is important to further study these sub-groups of individuals following stroke to intervene in an evidence-based manner and promote optimal secondary prevention since diabetes itself is an independent predictor of recurrent stroke with about 9.1% of stroke cases being attributable to it [30‐32].

Focusing on the modifiable factors found to be significantly associated with overall control post-stroke, following are some implications for practice arising from our current work. We recommend that efforts and resources should be directed towards developing adequate weight management services and smoking cessation services within primary care setting. The family physicians should adopt a holistic approach to secondary prevention post-stroke with incorporation of both pharmacological and non-pharmacological management in their care provision. There should be relatively more focus on lifestyle modification advice and guidance with regards to dietary changes, weight reduction and physical activity. At the health system level, provision of financial subsidies aimed at weight management and smoking cessation services will increase the delivery and uptake of these services. At provider level, directing adequate resources to develop capacity for delivering such services within primary care setting would be beneficial.

Our study has several strengths including the large sample size from 10 polyclinics over a period of 5 years. We captured major risk factors associated with recurrent stroke with a large database. Compared to observational study design which includes self-reported data, our study has the advantage of physician recorded data from electronic health records. This was one of the few studies to provide estimates of the overall control of risk factors post-stroke in an Asian setting, and we have added new knowledge to the existing literature on the prevalence of control of each singular risk factor.

The study also has several limitations. The database could not provide the causation of stroke (ischemic versus haemorrhagic) experienced by each individual, which may influence treatment recommendations by clinicians. Moreover, the database did not have information on other relevant variables such as the functional status of the individuals following stroke, education level, employment status, diet, physical activity, alcohol intake and available psychosocial support. Another limitation was related to missing data, for which we opted to conduct complete case analysis. Another shortcoming was that we could only assess the proportion of individuals meeting or not meeting the treatment goals but could not elicit the reasons why. Qualitative research exploring the experiences of individuals following stroke and their caregivers engaging in secondary prevention related behaviours will be needed. Our focus was on the overall and singular risk factor control at 1-year for individuals following stroke in the primary care setting. We did not explore the relative difference between baseline and 1 year following a stroke.