2.2.4. Assessment of Pro-oxidant–Antioxidant Balance

The PAB was indirectly established in the literature based on multiple components and comparable weighting methods (such as equal-weighted, study data-based, literature review-derived, and Bayesian methods) [33,34]. In the current study, the PAB components were defined based on core health behaviors related to stroke, including overweight/obesity, smoking, and drinking with pro-oxidant properties, physical activity, diet quality, fruit intake, and vegetable intake with antioxidant properties.


All components were categorized into three levels and assigned 0, 1, or 2 values, accordingly. Then, the PAB score was calculated by summing all components after weighting by multiplying the values with −1 for pro-oxidant and +1 for antioxidant (Table 1). The PAB score ranged from −6 to 8, with higher PAB scores indicating a beneficial balance shifting toward antioxidant dominance.

### 2.2.5. Assessment of Covariates

Socio-demographic factors (including age, gender, educational attainment, occupation, social status, marital status, and ability to pay for medication) were self-reported.

Stroke classification was categorized based on the ICD-10 into infraction (including cerebral infarction (I63)), hemorrhage stroke (including subarachnoid hemorrhage (I60), intracerebral hemorrhage (I61), other nontraumatic intracranial hemorrhage (I62)), and others (including stroke not specified as hemorrhage or infarction (I64); occlusion and stenosis of precerebral arteries, not resulting in cerebral infarction (I65); occlusion and stenosis of cerebral arteries, not resulting in cerebral infarction (I66); other cerebrovascular diseases (I67); cerebrovascular disorders in diseases classified elsewhere (I68); and sequelae of cerebrovascular disease (I69)).


**Table 1.** Categories and weights of pro-oxidant–antioxidant balance components.

Comorbidity was judged using the 16-item Charlson comorbidity index (CCI), which is validated and used widely in Vietnam [47,48]. In the current study, items of cerebrovascular disease or stroke and dementia were not included in the comorbid conditions due to the study participants and exclusion criteria, respectively. Additionally, the item of depression was not counted as a comorbidity to avoid the duplicated assessment. The remaining items were regrouped into none vs. one/more CCI to simplify the analysis.

Depressive symptoms were assessed using a two-item patient health questionnaire (PHQ-2), which was suggested for use in busy medical settings [49,50]. Patients were asked about the frequency of being affected by positive or depressed moods in the past two weeks and were rated on a three-point scale from 0 (not at all) to 3 (nearly every day). The PHQ-2 score ranged from 0 to 6, and a score ≥ 2 was suggested for clinicians to ensure that the few cases of depression were not overlooked [51].

Health literacy (HL) (including three domains of healthcare, disease prevention, and health promotion) was estimated using a 12-item short-form survey (HLS-SF12). The difficulty of performing each item was rated from 1 (very difficult) to 4 (very easy) based on a 4-point Likert scale. The HL index (the standardized HL indices) was calculated using the following formula:

$$Index = \frac{(mean - 1) \times 50}{3}$$

in which *mean* displays the average of 12 items, while *1*, *3*, and *50* display the minimal possibility of the mean, the range of the mean, and the chosen maximum HL index score, respectively. The HL index varied from 0 (worse HL) to 50 (best HL).
