Background

Though well-being robustly predicts health and mortality, it has often been argued that it has little predictive power outside of how it reflects psychological ill-being in reverse.

Aims

This meta-analysis examines prospective longitudinal studies of mortality that measure well-being and its subconstructs and control for ill-being in some manner, assiduously assessing for publication bias. I explore covariates and possible mediators that might explain the effect. I also compare these adjusted measurements to unadjusted measurements in similar longitudinal studies to estimate how much of the predictive effect of well-being is shared by ill-being.

Method

A systematic records search uncovered 88 adjusted and 64 unadjusted hazard ratios from longitudinal mortality studies. I used random-effects meta-analyses to pool the hazard ratios of well-being measurements that control for ill-being, overall and for specific subconstructs. I performed this meta-analysis again after removing outliers and correcting for estimated publication bias using the trim-and-fill method. I also estimated true effects using robust Bayesian meta-analysis (RoBMA). I also performed a meta-analysis of the unadjusted measurements and compared them to the adjusted measurements using Z-tests. Finally, I performed a meta-regression to examine significant covariates of well-being’s protective effect.

Results

Adjusted well-being’s protective hazard ratio was significant (HR=.875 [.854-.897]). Adjusted measurements of optimism, life satisfaction, positive affect, and general well-being were also significantly protective, but perceived social support was not. These effects are robust to the removal of outliers and estimated publication bias. Though many measurements had moderate to high risk of bias, measurements with low risk of bias were also significantly protective. Little predictive power was shared between well- and ill-being. Few covariates emerged from the meta-regression.

Conclusion

These results suggest that early death is independently predicted by a lack of positive qualities as well as the presence of negative qualities. The two measure different parts of human experience. Alongside traditional psychological interventions, positive psychological interventions will likely help people have longer, better lives.