**3. Results**

The EFA indicated a latent structure of four factors and the bifactor model ( χ2 = 4039.14, df = 273, *p*-value < 0.00l) exhibited good fit (CFI: 0.985; RMSEA: 0.040, 90%CI: 0.039–0.041). The psychometrically informative bifactor-derived statistics indicated that a strong percentage of the total score variance can be attributed to the general factor ( ω H = 0.85). Hence, we proceeded by considering the score of the general factor only. More information about the EFA and CFA is provided in Supplementary File S2.

In our study there were 10,900 participants in the baseline assessment. In the follow-up survey 1648 were healthy agers, 5594 normal agers, 1734 died and 1924 were either untraceable or refused to be re-interviewed. We performed analyses per country and we checked if participants included in our analyses differ from those not available in the follow-up wave (*t*-test for continuous variables, χ2 test for categorical). We found that participants included in our analyses in general did not differ with those excluded (Supplementary File S3). There were some statistically significant (i.e., *p*-value < 0.05) differences in Cuba and Peru regarding the dietary habits between those included in the follow-up wave and those excluded. A higher percentage of participants available in the follow-up wave endorsed daily consumption of fruits and vegetables compared to those not available (i.e., 75.4% versus 66.3% in Cuba; 82.5% versus 76.2% in Peru).

Table 1 presents the characteristics of participants characterised as healthy agers and those characterised as normal agers in the follow-up. Participants in the healthy ageing group, compared with the normal ageing group, were younger at baseline, had completed at least primary level education, were mostly physically active and endorsed daily consumption of fruits and vegetables.

The associations of each lifestyle behaviour with healthy ageing and survival at the end of the follow-up period are presented in Table 2. Compared with non-physically active participants, those who were physically active at baseline had 2.59 (95% CI: 2.20–3.03) times greater odds of being healthy agers and 2.23 (95% CI: 1.98–2.52) times greater odds of survival. Compared with people consuming two or less servings of fruits and vegetables in the last three days, people consuming three or more were more likely to experience healthy ageing (OR: 1.24, 95% CI: 1.06–1.44) and to survive at the end of the follow-up (OR: 1.13, 95% CI: 0.99–1.28). Non-smoking, compared to former or current smoking, was not associated with healthy ageing (OR: 0.95, 95% CI: 0.82–1.10) but it was positively associated with survival as never smokers had 1.17 (95% CI: 1.02–1.34) times greater odds of remaining alive in the follow-up assessment. Moderate drinking, compared to high or never drinking, was not associated with healthy ageing (OR: 1.04, 95% CI: 0.82–1.30) or survival (OR: 1.11, 95% CI: 0.90–1.37). Heterogeneity was not statistically significant in most cases (*p*-value < 0.05) indicating that the variation among countries was limited. The only significant heterogeneity was observed in the associations of physical activity with healthy ageing (I2: 61.3%; *p*-value: 0.035). Further investigation concluded that Puerto Rico was the country exhibiting the highest point estimate (i.e., OR: 6.30; 95% CI: 3.43–11.56). We also reproduced the pooled estimates by excluding this country and the new pooled estimates did not change the direction or strength of associations between physical activity and healthy ageing (i.e., OR: 2.42; 95% CI: 2.05–2.86; I2: 0.0%, *p*-value: 0.693).

Participants who engaged in two-four healthy lifestyle behaviours at baseline, compared with those who engaged in none or one, exhibited increased odds of healthy ageing (OR: 1.72, 95% CI: 1.45–2.05) (Figure 1a). Additionally, the association between number of healthy behaviours and odds of healthy ageing was positive linear. More specifically, compared with participants engaging in none or one healthy behaviour, those engaging in two had 1.46 times (95% CI: 1.22–1.76) the odds of healthy ageing, those engaging in three had 2.00 times (95% CI: 1.65–2.42) the odds and those in four had 2.46 times (95% CI: 1.54–3.92) the odds. A similar beneficial association was also observed in surviving (Figure 1b). Participants engaging in two or more healthy behaviours had 1.83 times (95% CI: 1.61–2.08) the odds of surviving compared with participants that engaged in none or one healthy behaviour. A positive linear relationship was also observed; two healthy behaviours OR: 1.60, 95% CI: 1.40–1.84; three healthy behaviours OR: 2.29, 95% CI: 1.94–2.69; four healthy behaviours OR: 2.64, 95% CI: 1.53–4.54 (Table 3). Heterogeneity was not statistically significant (*p*-value < 0.05) indicating that the variability among countries was limited.



Notes: SD: standard deviation; †: *p*-value of the *t* test or the χ2 < 0.05. \* healthy ageing group includes those participants who in the follow-up assessment were in the two highest fifths of the healthy ageing score distribution (i.e., better health level) and normal ageing group those who were in the three lowest fifths; percentages in the parentheses are calculated in the non-missing cases in the covariates.



Notes: CI: confidence interval; OR: Odds ratio; I2: heterogeneity. \*: healthy ageing compared to normal ageing or death; survival (healthy or normal ageing) compared to death; number of participants per country: Cuba: 2557, Dominican Republic: 1608, Mexico: 1633, Peru: 1392, Puerto Rico: 1554. †: models are adjusted for age, gender, education level and all other lifestyle behaviour variables.

**Figure 1.** Associations of (**a**) healthy ageing and (**b**) survival with the number of healthy behaviours. Notes: ORs: Odds Ratios (pooled effects of the fixed-effects meta-analysis); models are adjusted for sex, age and education level and for all behaviours at baseline. Associations are given per individual number of healthy behaviours (two, three, four) and for two-four compared to no or one healthy behaviour (reference category). Error bars indicate 95% confidence intervals.

Further sensitivity analysis showed that when we excluded deaths from our sample, the associations of lifestyle behaviour with healthy ageing did not substantially change (i.e., fairly or very physically active OR: 2.28, 95% CI: 1.93–2.69; never smoking OR: 0.95, 95% CI: 0.81–1.10; moderate drinking OR: 1.05, 95% CI:0.83–1.32; daily consumption of fruits and vegetables OR: 1.22, 95% CI: 1.04–1.42). In addition, supplementary analysis showed that the higher the level of physical activity the more increased the ORs of healthy ageing compared to normal ageing or death (very physically active OR: 9.64, 95% CI: 5.62–16.55; fairly OR: 8.20, 95% CI: 4.77–14.12; not very OR: 4.20, 95% CI: 2.42–7.31; not at all: reference category). The same trend was also observed for the vegetable and fruits consumption (more than six servings OR: 1.51, 95% CI: 1.23–1.85; three to six servings OR: 1.16, 95% CI: 0.99–1.36; less than two servings: reference category). Smoking and alcohol consumption had no significant associations.

To examine whether our findings differed between men and women, we also performed our analyses separately for these two subpopulations. The observed relationships between lifestyle behaviours and healthy ageing compared to normal ageing or death did not change (Supplementary File S4). In addition, to examine whether our conclusions were influenced by the categorisation of participants as healthy agers based on the two highest quintiles of the healthy ageing score, we also performed our analyses by using a different categorisation. We performed our analyses by considering participants scoring in the three highest quintiles of the healthy ageing score distribution as healthy agers (i.e., 59.70–100 scores) and compared them against normal agers (i.e., 0–59.69 scores) or dead participants. Our conclusions regarding the protective effect of physical activity and daily consumption of fruits and vegetables did not alter (Supplementary File S5).


**Table 3.** Associations between number of healthy behaviours, healthy ageing and survival.

Notes: CI: confidence interval; OR: Odds ratio; I2: heterogeneity. \*: healthy ageing compared to normal ageing or death; survival (healthy or normal ageing) compared to death; number of participants per country: Cuba: 2557, Dominican Republic: 1608, Mexico: 1633, Peru: 1392, Puerto Rico: 1554. †: models are adjusted for age, gender, education level and all other lifestyle behaviour variables.
