*2.1. Study Population*

The 10/66 Dementia Research Group (10/66 DRG) population-based studies of ageing and dementia have been conducted in geographically defined catchment areas in LMICs (Cuba, Dominican Republic, Peru, Venezuela, Mexico, China, India and Puerto Rico) with a sample size between 1000 and 3000 (generally 2000). In this study, we used data from the baseline survey (2003–2008), including people 65 years old and over, and the follow-up assessments 3 to 5 years later (2008–2010) from Cuba (n = 3000), Dominican Republic, Peru, Mexico and Puerto Rico (n = 2000). Detailed description of the study settings is available elsewhere [19,20]. Access to the dataset can be obtained via the official website www.alz.co.uk/1066. Local ethical committees and the ethical committee of the Institute of Psychiatry of King's College London approved the studies: Cuba-Finlay Albarran Medical Faculty of Havana Medical University Ethical Committee (January 2003, September 2006); Dominican Republic-Consejo Nacional de Bioética y Salud (CONABIOS); Mexico-Instituto Nacional de Neurología y Neurocirugía Ethics Committee (Ref: 96/07; 78/07); Peru-El Instituto de la Memoria, Depresión y Enfermedades de Riesgo (IMEDER) Ethics Committee (August 2004, May 2007); Puerto Rico-Oficina para la Protección de Participantes Humanos en Investigación (OPPHI) (Ref: FWA00005561); King's College London-Institute of Psychiatry/SLAM NHS Trust Ethical Committee Research (ECR) (2004, Ref: 076/03), King's College Research Ethics Committee (KREC) (February 2007, Ref: CREC/06/07/38).

## *2.2. Baseline Measures*

Lifestyle behaviours were assessed based on participants' responses to specific questionnaires. Physical activity was categorised as: "very physically active", "fairly", "not very", and "not at all". Smoking status was categorised as "current", "former" and "never smoker". Alcohol consumption was assessed by asking the units of alcohol drunk per week and then categorised as "no alcohol consumption", "moderate alcohol consumption" (1–14 units per week for women; 1–21 units per week for men) and "heavy alcohol consumption" (≥15 units per week for women and ≥21 units per week for men) [21]. We assessed dietary habits by asking the number of fruit and vegetable servings in the last 3

days. In our analyses, we compared physically active participants ("very physically active" or "fairly") with non-physically active ("not very" or "not at all"); non-smokers with ever-smokers ("current" or "former"); moderate drinkers with abstainers (no alcohol consumption) or heavy drinkers; participants with "daily consumption" of fruits and vegetables (3 or more servings the last 3 days) with "non-daily consumption" (0–2 servings the last 3 days). We also adjusted our results for socioeconomic factors: age, sex and level of education (none, some, did not complete primary, completed primary, completed secondary, tertiary).

#### *2.3. Outcome Assessment*

In accordance with the WHO definition of healthy ageing, we conceptualised healthy ageing within the functional ability framework [8]. We created a healthy ageing index by employing information from 26 health related question-indicators which were either self-reported or were provided by key informants. Items included information on daily disabilities and difficulties, on pain and sleep problems and on cognition abilities among others (Supplementary File S1).

To create a measurement model for healthy ageing, we employed the framework of Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) [22]. We initially identified the number of factors needed to reproduce the latent (unobserved) construct of healthy ageing (EFA) and then we used a bifactor model to represent a single general construct of it (CFA) [23]. To establish model fit, we examined the comparative fit index (CFI) and root mean square error of approximation (RMSEA) with 90% confidence intervals (CI). We considered a model to have an acceptable fit when CFI ≥ 0.90 and RMSEA values were close or less than 0.06 [24]. We also calculated omega hierarchical ( ω H) reliability coefficient of the bifactor structure; a high ω H > 0.80 indicates that the general factor is the dominant source of systematic variance [25]. Finally, we calculated healthy ageing (factor) scores by using the regression method [26].

We transformed scores in a scale of 0–100 with higher values indicating better health. We examined 3 categories for the outcome of interest: healthy ageing, normal ageing and death at follow-up. To create these categories, we calculated the quintiles of the baseline healthy ageing score distribution and participants belonging in the three lowest fifths (i.e., 0–67.92 scores) were characterised as "normal agers", whereas participants belonging in the two highest fifths (i.e., 67.93–100 scores) were characterised as "healthy agers".

We performed two separate logistic regression analyses: one to estimate the odds ratios (ORs) of healthy ageing (with normal ageing and death as the non-event) and another one to estimate the ORs for survival (in which death was the non-event). We did all analyses per country and we then provided a pooled result by performing fixed-effects meta-analysis. We also computed Higgin's I2 to estimate the proportion of variability across countries that accounted mostly for heterogeneity rather than sampling error [27]. Heterogeneity up to 40% is conventionally considered negligible, whereas heterogeneity more than 50% is considered moderate to substantial [28].

We also performed sensitivity analyses to assess the robustness of our findings. Firstly, we examined whether the associations of lifestyle behaviours with healthy ageing were influenced by deaths. To do this, we excluded deaths from our sample. Secondly, we performed our analyses by including the non-dichotomised lifestyle variables. Finally, we investigated whether our findings were influenced by the way we categorised participants as healthy or normal agers and if there were differences between men and women. Our analyses were performed in STATA 14.1 and Mplus 7.4 [26].
