Lagström et al explore whether the transition to retirement is associated with changes in dietary intake, specifically fish, red meat and fruits and vegetables.1 The authors tap into a shared experience, retirement, common to individuals in countries with retirement programmes.

Drawing on terminology from life course epidemiology,2 the transition to retirement can be viewed as a sensitive period. Changes in health-related behaviours during this time, such as diet, can (re-)set a trajectory for health as people age. The retirement transition represents a pivotal opportunity to support healthy life expectancy, a fundamental component for optimising extended life spans.3

Lagström et al state that the ‘transition to retirement is a significant turning point in life’.1 This turning point can involve both positive and negative experiences. Retiring may disrupt routines and social connections, but it offers more leisure, engagement opportunities and relief from work-related stress.4 Indeed, the authors find that the transition to retirement among study participants was associated with both favourable increases in fruit and vegetable intake and unfavourable increases in red meat consumption. Results from other empirical work examining outcomes such as cardiovascular disease,5 physical activity,6 mental health and cognitive skills7 among others, have been inconclusive. This lack of consensus points to heterogeneity in the effects of and complexity inherent in studying retirement. Here, we outline some considerations for interpreting and researching retirement’s health impact.

Clarify the question and the population

A clear and specific research question is essential for effective research and accurate interpretation. Lagström et al ask how the transition to retirement is associated with changes in dietary intake.1 They do not ask about differences in dietary intake between retirees and age-matched individuals who remain employed. These are both important but distinct questions that offer insights into various aspects of retirement and health-related behaviours.

It is equally important to clarify to whom this research question applies. Having a clearly defined target population is essential for this. Lagström et al use the Whitehall II study and note that the generalisability of their findings is restricted to the urban British population, the majority of whom were male white-collar workers.1 However, they also restrict their analyses to only those who retire for non-health reasons. This may introduce selection bias and requires redefining the target population, further limiting the applicability of findings to a healthy population.

The clarity of the question is fundamentally tied to the study-specific definition of retirement. Lagström et al define retirement as those who self-report moving from full-time or part-time employment to retirement or from employment to unemployment/other and then to retirement.1 In numerous studies, including this one, retirement is viewed as an absorbing state, meaning individuals are classified as retired from the initial report regardless of subsequent responses.

The reality is far more complex. Retirement is influenced by state policies and often involves more than simply transitioning from full-time employment to complete inactivity.8 Financial incentives can also drive these transitions.8 When using observational data, we may be limited in how we operationalise complex transitions. It is unlikely that all types of transitions result in the same changes in dietary intake, or other health outcomes.

The impact of retirement may extend beyond the transition point, so it is crucial to consider its long-term effect on health. Lagström et al use data spanning over 30 years—16 years preretirement and 16 years postretirement.1 Their models provide information on linear change over time in each of these time periods. Others have observed immediate but not long-term effects of retirement on health.5 9 Therefore, studying potential non-linear effects may offer further insights into the effects of retirement on health over time.

Consider causality

Determining if retirement causes health or health-related issues is challenging. By the time individuals retire, they have likely been exposed to various influences that can affect both their decision to retire and their health. Lagström et al adjust their models for sex, occupational status, marital status and financial hardship, but other influencing variables, such as the highest level of education, the type of occupation, partner employment status and income, among others, could also affect the relationship.1 These confounders are difficult to account for if not measured precisely in observational studies. The bidirectional relationship between retirement and health is another challenge in determining causality. Lagström et al address this issue by excluding those who retire for health reasons.1 However, as mentioned, this may have inadvertently introduced selection bias, potentially distorting the true effect of retirement on diet in the original target population.

All methods applied to observational data have their strengths and limitations. Using directed acyclic graphs can help identify and clarify potential sources of bias, although they are not frequently used in retirement studies. Lagström et al use generalised estimating equations and include an interaction term to test whether the changes in diet are different between preretirement and postretirement.1 This approach has the potential to identify whether people anticipate retirement by changing their health or health-related behaviours. This anticipatory effect has been observed in other studies,9 where mental health improves 2 years before retirement.

Studies of retirement and health published in the economics literature tend to use fixed effects and instrumental variable analysis to address issues of confounding and reverse causality. Fixed effects can account for time-invariant individual characteristics and estimate within-person changes in health as people transition to retirement. In instrumental variable analysis, sex and country-specific eligibility age for statutory pension are used as the instrument. Because the instrument is not related to confounding factors nor subject to reverse causality, it is better placed to estimate the causal effect. However, the results only apply to individuals who retire because they reach statutory retirement age. When studying retirement and health, it is essential to identify potential confounding factors, address reverse causality and use multiple methods.

Acknowledge time and place

Lagström et al use data from a UK population collected in the first instance from 1985 to 1988.1 The timeline of this study coincides with a UK outbreak of bovine spongiform encephalopathy, which may have led to the population reducing their consumption of red meat.10 The period effect makes it hard to determine whether the postretirement increase in red meat consumption was due to retirement transition or feeling safer due to the subsiding outbreak. Similarly, other studies have found that the effect of retirement on mental health is stronger during times of recession.11 Considering when in time the study is important, as Lagström et al succinctly note, their ‘findings do not necessarily reflect the eating habits of today’s 60-year-old UK population’.1

Examine inequality and question policy implications

The pursuit of identifying a single causal effect of retirement on a specific health outcome within a specific time is likely to be unsuccessful and unhelpful. For example, Lagström et al find that the postretirement increase in red meat consumption was particularly evident among women, single participants and those in a more disadvantaged occupational status.1 These heterogeneous subgroup effects have been seen in studies with other health outcomes10 12 and should be considered in future studies to avoid exacerbating health inequalities.

It is also important to be mindful of the type of policy interventions that may result from research findings. For example, if empirical evidence suggests that retirement may have detrimental effects on health, one potential policy intervention could be to extend working lives to promote healthy longevity. However, beyond addressing the impact of retirement policies on health, it is essential to explore why these negative changes occur and what can be done—before, during and after the transition—to improve population health.

Conclusion

Understanding how retirement can affect health and health-related outcomes is essential to our goals of healthy longevity and becomes increasingly important as statutory retirement ages rise. The study by Lagström et al 1 provides a useful step forward in understanding this relationship. We hope that the considerations laid out here will guide future interpretation and conduct of such studies.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.

Acknowledgments

We are grateful to Praveetha Patalay for providing comments on a draft of this paper.