Differences across countries in baseline scores

Differences in the preintervention scores across countries are shown in Table 2. In the healthy aging sample, the Mexican sample had higher scores in the LSI-A scores (p = 0.040, Hedge’s g =  − 0.89) and in the AMT negative specific memories recall (p = 0.006, Hedge’s g =  − 1.07), compared to the Spanish sample.

Table 2 Comparison of pretest scores across countriesCross-cultural differences in the effect of the reminiscence program on LSI-A scores

To test the effects of the REMPOS program on LSI-A scores, we estimated the significance of the interaction between time and condition in the mixed model for each type of aging and country as explained in the “Methods” section above.

For LSI-A scores in the Alzheimer’s sample, the triple interaction between time, condition, and country was significant in the pooled model (t =  − 2.82, p = 0.007, and see unstandardized coefficients in Table 3), indicating that the REMPOS program had a differential effect in each country. Furthermore, the interaction between time and condition was significant only in the Mexican sample. Simple main effects analysis found that neither of the experimental and control groups in the Spanish sample showed pre–postdifferences in LSI-A scores. On the contrary, in the Mexican sample, the experimental group showed pre–postintervention differences, but not the control group (see Supplementary Table 1). In summary, the reminiscence intervention groups reported higher LSI-A scores after the intervention, the effect was stronger in the Mexican sample, and only there the effect reached statistical significance. Overall, the reminiscence intervention showed stronger effects than the control condition in the Alzheimer’s sample of both countries, with stronger effects in the Mexican sample.

Table 3 Unstandardized coefficients of linear mixed models predicting LSI-A scores and autobiographical memory test (AMT) scores for positive (EPOS) and negative (ENEG) specific memories to test the effect of the REMPOS program. The columns for Spain and Mexico show separate models for each sample, where the interaction between time and condition test for the differential effect of the REMPOS program. The column “Both” tests a model pooling all participants from both countries and it uses the interaction between time, condition, and country to test cross-cultural effects

For the LSI-A scores in the MCI sample, the triple interaction in the pooled model between predictors time, condition, and country was not significant (t =  − 1.37, p = 0.18, see Table 3), but the interaction between time and condition was significant (t = 2.49, p = 0.016). Unexpectedly, neither sample of each country showed a significant interaction between time and condition, though the coefficient was in the direction of larger LSI-A scores after the intervention. In the simple main effects analyses, neither the experimental nor the control group in the Spanish sample showed significant differences after intervention. On the contrary, in the Mexican sample, the experimental group showed differences between pre- and postintervention scores, but not the control group (see Supplementary Table 1). Although the overall pattern was for the experimental group to have higher LSI-A scores after the intervention, it only reached statistical significance for the Mexican sample, which showed a stronger effect size. In summary, in the MCI sample, the reminiscence intervention had a significant positive effect in the Mexican sample, with higher scores after intervention, compared to the control condition; but not in the Spanish sample, where there was the same trend, but did not reach statistical significance.

For the LSI-A scores in the healthy aging sample, neither the triple interaction between time, condition, and country (t =  − 0.93, p = 0.35) nor the double interaction between time and condition (t = 1.8, p = 0.076) were significant in the pooled model (see Table 3). In the models for each country, neither showed a significant interaction between time and condition, indicating that the REMPOS intervention did not differ from the control condition in neither sample. Although the hypotheses were that significant differences would appear in all subgroups, none of the tests in the simple main effects analyses reached statistical significance in either the experimental or control group of either country (see Supplementary Table 1). In summary, the overall trend of the interventions in the healthy aging sample was that there were higher LSI-A scores after intervention (i.e., the marginal mean for post intervention scores was higher than preintervention scores), but there was insufficient statistical power to separate the effects in each subgroup Figure 1.

Fig. 1

LSIA scores for each type of aging, condition, and country

Overall, for LSI-A scores, the Alzheimer’s and MCI Mexican samples showed greater benefits from the reminiscence intervention, compared to the control condition. While the Spanish sample showed some trends of these kinds of benefits, it failed to reach statistical significance. In the healthy aging Spanish and Mexican samples, the overall trend was for better scores after intervention, for both experimental and control conditions, but none reached statistical significance.

Cross-cultural differences in the effect of the reminiscence program positive specific memories

Figure 2 shows the scores of the AMT, both positive and negative specific memories recall. We report the effects of the REMPOS intervention on the AMT-positive specific memories scores with the same structure as the LSI-A scores: the coefficient of interest is the interaction between time and condition, and we report separate models for each country and type of aging.

Fig. 2

AMT specific positive and negative memories scores for each type of aging, condition, and country

For the positive specific memories in the Alzheimer’s sample, the triple interaction between predictors time, condition, and country (t =  − 3.66, p < 0.001) and the double interaction between time and condition (t = 8.55, p < 0.001) were significant in the pooled model (see Table 3). The double interaction between time and condition was also significant in each country, where the Mexican sample showed the larger coefficient for this interaction. This pattern of results suggests that the experimental group differed from the control group on the effectiveness of the intervention in both countries, but the effect size was larger in the Mexican sample. In the simple main effects analyses, both experimental groups in the Spanish and Mexican samples showed differences between pre- and postintervention scores, where scores increased after intervention. Neither of the control groups showed significant differences (see Supplementary Table 2). In summary, the reminiscence program increased the recall of positive specific memories after intervention, while the control intervention had no effect on this, in both Spanish and Mexican participants with Alzheimer’s disease.

For the positive specific memories in the MCI sample, the triple interaction between predictors time, condition, and country (t =  − 5.97, p < 0.001) and the double interaction between time and condition (t = 7.32, p < 0.001) were significant in the pooled model (see Table 3), meaning that the interaction between time and condition, and therefore the differential effect of both types of interventions, varied across countries. In the models for each country, the double interaction between time and condition was only significant in the Mexican sample, and not in the Spanish sample. Simple main effects analyses showed that in the Spanish sample, the experimental and control conditions showed no significant differences between pre- and postintervention scores. In the Mexican sample, the experimental group showed significant differences, but the control group did not (see Supplementary Table 2). In summary, in the MCI sample, the reminiscence program had positive effects increasing recall of specific positive memories in the Mexican, but not in the Spanish sample. The control intervention did not have any effect on the recall of positive specific memories.

For the positive specific memories in the healthy aging sample, the triple interaction between predictors condition, time, and country (t =  − 5.41, p < 0.001) and the double interaction between time and condition (t = 7.46, p < 0.001) were significant in the pooled model (see Table 3). Similar to what happened in the MCI sample, the double interaction between time and condition was only significant in the Mexican sample. In the Spanish sample, there was a significant effect of time, indicating that both experimental and control groups improved after the intervention. Simple main effects analyses showed that both experimental and control groups had significantly higher scores after intervention in the Spanish sample. On the other hand, in the Mexican sample, only the experimental group showed significant differences, also with higher post intervention scores (see Supplementary Table 2). In summary, for the healthy aging sample, both interventions had similar positive effects in the Spanish sample, increasing the recall of positive specific memories after intervention, while the reminiscence program had a greater effect than the control condition in the Mexican sample.

Overall, the effect of the reminiscence program compared to the control condition varied across countries in the recall of positive specific memories. In general, the reminiscence program had a positive effect in increasing the recall of positive specific memories in all three types of aging, in both countries, except for the Spanish MCI sample, in which it had no effect. The control condition improved the recall of positive specific memories only in the healthy aging Spanish sample but had no effect in the remaining groups.

Cross-cultural differences in the effect of the reminiscence program on negative specific memories

For the negative specific memories in the Alzheimer’s sample, neither the triple interaction between condition, time, and country (t = 1.46, p = 0.148) nor the double interaction between time and condition (t = 0.93, p = 0.356) were significant in the pooled model (see Table 3). In the models for each country, neither the double interaction between time and condition nor the effects of time were significant. In summary, neither of the reminiscence program and control intervention had any significant effect on recall of negative specific memories in the Alzheimer’s sample.

For the negative specific memories in the MCI sample, the triple interaction between predictors condition, time, and country (t =  − 2.24, p = 0.028) and the double interaction between time and condition (t = 2.84, p = 0.006) were significant in the pooled model (see Table 3). Nevertheless, in the models for each country, neither the interaction between time and condition nor the main effects of time were significant. To further analyze this, we ran simple main effects of the predictor time within each condition in each country, but neither of the experimental or control groups in either Spanish or Mexican sample showed significant differences between pre- and postintervention scores (see Supplementary Table 3). Nevertheless, the experimental group in the Mexican sample showed a higher effect size, even when the adjusted p value did not reach significance. In summary, the effect of the reminiscence program on the recall of negative specific memories had a small effect in both countries, only detectable in the model that pooled participants from both countries, therefore, with higher statistical power, but not in simple main effects analyses which used adjusted p values for multiple comparisons.

For the negative specific memories in healthy aging, the triple interaction between time, condition, and country (t =  − 2.74, p = 0.008) and the double interaction between time and condition (t = 4.23, p < 0.001) were significant in the pooled model, indicating that the differences in the effects of the experimental and control conditions varied across countries. When analyzing the models for each country, the double interaction between time and condition was significant only in the Mexican sample. In the simple main effects analyses, both experimental and control conditions showed significant effects in the Spanish sample, with higher recall of negative specific memories after intervention. Interestingly, in the Mexican sample, the control condition showed significant differences between pre- and postintervention scores, where the recall of negative specific memories was lower after intervention, while the experimental condition did not show statistical differences (see Supplementary Table 3). In summary, in healthy aging, the reminiscence program increased recall of negative specific memories in the Spanish sample, but not in the Mexican sample. In the Spanish sample, the control condition also increased recall of negative specific memories, but in the Mexican sample, it showed the opposite effect, decreasing them.

In the case of recall of negative specific memories, the differential effects of reminiscence intervention compared to the control condition varied across countries and types of aging. In the Alzheimer’s and MCI samples, the effects of both interventions were negligible, especially in the Alzheimer’s groups. Interestingly, in the healthy aging sample, there was an unexpected difference across countries: both interventions increased the recall of specific negative memories in the Spanish sample, while the Mexican control condition decreased the recall of those types of memories.