An online intervention for improving stroke survivors’ health-related quality of life: study protocol for a randomised controlled trial

Individuals will not be eligible to participate in the study if they have documented evidence of a previous stroke (second to the indexed stroke event) or experience disability at a level that may limit their use of the programme (determined using a score of ≤ 3 on the modified Rankin Scale) [26].

If a participant wishes to withdraw consent, they will be able to at any point in time. They will be given the option to remove all their previously collected data or just remove consent for further data collection.

The main reportable outcomes for the cost-effectiveness analysis of the intervention will be (1) an average cost-effectiveness ratio, and (2) an incremental cost-effectiveness ratio, per Quality-adjusted Life Year (QALY). A modified version of the Client Service Receipt Inventory (CSRI) will be used in conjunction with the MBS and PBS data to determine use of health services (frequency, length and cost of physician consultations; allied health care; inpatient stays; outpatient episodes; medication use) and potentially other economic impacts (such as time off work due to disability) [34].

At baseline, a standard and previously validated item will assess smoking status ‘Do you currently smoke any tobacco products?’ (Daily, At least once a week, Less often than once a week, Not at all) [35]. At follow-up, 7-day point-prevalence abstinence (PPA) will be used to assess quit success using recommended items to determine the proportion of participants who have not smoked any tobacco in the preceding 7 days ‘Have you smoked at least part of a cigarette in the last 7 days?’ Seven-day PPA has high concurrent validity and reliability [36].

The brief Alcohol Use Disorders Identification Test – Consumption (AUDIT-C), [37] will be used. The AUDIT-C is a three-item version of the more extensive AUDIT and has been found to be a valid and reliable screening tool for heavy drinking and alcohol abuse or dependence. Scoring for the AUDIT-C ranges from 0 to 12 with cut-offs of 3 for women (sensitivity 66–73%, specificity 91–94) [38, 39] and 4 for men are used to indicate heavy drinking (sensitivity 86%, specificity 72–89%) [39].

The Australian Recommended Food Score (ARFS) [40] will be used to assess fruit and vegetable consumption, a measure which has been validated in adults. The ARFS is strongly correlated with food frequency questionnaire nutrient intakes (fibre, vitamin A, beta-carotene, vitamin C, and minerals). To determine salt intake, participants will be asked ‘How often do you add salt to your cooking?’ and ‘How often do you use salt at the table?’ (Never/rarely, Sometimes, Usually, Always) [41].

Physical activity will be measured using the Godin Leisure Time Exercise Questionnaire (GLTEQ), [42] which measures the number of occurrences in a week in which a participant engages in periods of mild, moderate or vigorous exercise for 10 min or longer. Weights are then applied to the number of weekly occurrences of each type of exercise (mild, moderate, vigorous) and summed to calculate the number of metabolic equivalent or MET-minutes of exercise per week; it has acceptable reliability and validity.

Depression and anxiety levels will be assessed using The Patient Health Questionnaire four-item (PHQ-4) [43]. The PHQ-4 is a validated and ultra-brief screening measure for both anxiety and depression. Scores range from 0 to 12, with categories of psychological distress being: none (0–2); mild (3–5); moderate (6–8); and severe (9–12). Subscale categories for anxiety and depression both range from 0 to 6 with a score of 3 or more considered positive.

Physical functioning and independent living will be measured using the Barthel Index (BI) [44, 45] and Instrumental Activities of Daily Living (IADL) scales [46]. The BI measures activities of daily life on a 10-point scale related to self-care and mobility. The IADL questions attempt to capture the ability to live independently in the community.

Other measures include hospital admission for recurrent stroke, myocardial infarction and all other causes, and data from medical health records. These data will be collected from national health care databases and will be used to inform resource-use assumptions in the cost-effectiveness analysis for patients utilising hospitals in health districts without access to an equivalent intervention.

Will be conducted according to the intention-to-treat (ITT) principle, comparing the change in HRQoL between groups from baseline to 6-month follow-up. The comparison will be done using a linear regression model, where the outcome variable will be the EQ-5D-5 L HRQoL score, and the main predictors of interest will be the baseline value of the outcome variable, and treatment group. The two stratifying variables, state of residence within Australia and stroke type at baseline, will be included in the model as covariates. The primary method of dealing with missing data will be through multiple imputation, this will provide a sensitivity analysis exploring how robust the results are when considering missing values. The missing data will be explored to test the underlying assumptions of multiple imputations. Adjusting for potentially imbalances in important prognostic variables and pattern mixture models will be used to correct for violations in the missing at random assumption.

Generalised linear mixed models will be used to examine between-group differences in changes in secondary outcomes between baseline and follow-up; appropriate distributional families and link functions will be used according to outcome type. The models will include fixed effects for time, group and the interaction between time and group, as well as stratification variables. Random subject-level intercepts will be included to account for serial correlations. Associations between programme usage and outcome will be examined where greater adherence to the programme will be expected to be associated with improved outcomes.

In this study, a cost-utility analysis will be undertaken from the perspective of health care providers to identify whether the resources consumed to achieve the outcomes represent a cost-effective improvement over usual stroke follow-up care. The analysis will adhere to guidelines for conducting and reporting health economic evaluations [47, 48]. Given that the majority of resource savings are anticipated to arise following conclusion of the intervention, a patient lifetime time-horizon will be modelled founded on the trial outcomes. The primary measure of effectiveness for the economic evaluation will comprise HRQoL, as measured by EQ-5D-5 L, from which QALYs will be calculated using Australian utility weights. Resource utilisation for the respective trial arms will include: intervention development costs, health care costs derived from the AuSCR hospital International Classification of Disease (ICD) data, the CSRI data, and the MBS/PBS data. The ICD data will be translated to an estimate for cost differences between arms using allocations to diagnostic-related groups and according cost weights. It is implicitly assumed that differences between trial arms for non-stroke utilisation of inpatient services will be random. This assumption will be tested with the CSRI inpatient data and admitted patient data from a subset of health districts. Analysis of the QALYs and resource-use estimates will derive average and incremental cost-effectiveness ratios. Appropriate sensitivity and uncertainty analysis will be conducted.

A full statistical analysis plan will be written prior to analysis.