The association between the value of a statistical life and frailty in Burkina Faso

1. Introduction

To ensure that resources invested into services are commensurate with their benefit, economists have utilised various methods to assess the value of life. As populations age, it is important that the performance of these methods is understood in older people. This is especially the case for older populations living in low- and middle-income countries (LMICs), given that by 2030, the majority of the world’s older people are expected to be living in LMICs [1]. The Value of Statistical Life Years (VSLY) approach has consistently calculated human life as being far more valuable than alternative methods of cost-benefit analyses (CBAs) such as ‘value of lost earnings’ or the World Health Organization’s (WHO) cost-effectiveness threshold of multiplying national GDP per capita by three [2, 3]. The economic theory underlying the higher VSL measurements raises the possibility that previous methods have undervalued human life, hence its interest to health economists.

VSLY is a commonly used tool specific to CBAs where risk of death is considered as part of evaluation. It is guided by the real-world or theoretical monetary value that individuals place on accepting or reducing a small risk of death, either over their lifetime or annually [3]. The generated values essentially behave as a monetary threshold beyond which additional annual expenditure is considered excessive for the marginal reduced risk of death it provides. Whilst discussion continues on the benefits of using it, the VSLY approach continues to be viewed as beneficial, at least in high income countries [4, 5]. The few studies which have used the approach in sub-Saharan Africa or in LMICs in general have also considered it to be worthwhile [6]. But, given the growing ageing population in LMICs, understanding the impact of ageing on VSLY is needed [7], particularly if this is to be considered an applicable method for LMICs.

There is reasonable evidence from high-income countries showing the relationship between VSL values and increasing age [8]. However, there is a paucity of research exploring whether these findings are consistent in lower-income countries where life expectancy is typically lower and the role of older people in their communities is often more prominent [9, 10]. There is also little evidence on the effect of frailty on VSLY as people age. Frailty is a condition describing an individual’s vulnerability to greatly impaired health status following relatively minor stressor events [11]. Moreover, frailty occurs more frequently as people age and is associated with limiting independence and reduced quality of life; it is prevalent in older populations in LMICs [12–14].

Given the increase in older populations in LMICs, the high prevalence of frailty and its adverse consequences in these populations, and that VSLY is an attractive method to understand the value of life in low resourced settings, it is important to understand the relationship between frailty and VSLY [9, 15, 16]. Our hypothesis was that frailty would be associated with a lower VSLY in participants in rural Burkina Faso. This study aims to test our hypothesis, when controlling for factors found in other studies likely to affect VSLY [17–19], such as, socio-demographics, multimorbidity, quality of life, and disability.

2. Methods

3. Results

We identified 3,998 individuals aged >40 and living in the HDSS site; of those, 3,027 (75.7%) were located and agreed to participate and completed the baseline interview. Of those, 2761 individuals were included in this analysis after excluding participants with missing variables (Figure 1). Overall, 696 (25.2%), 679 (24.6%), 690 (25.0%), and 696 (25.2%) participants were available for the WTP-small, WTP-large, WTA-small, and WTA-large, respectively (Figure 1). Participants’ characteristics for variables included in the analyses are presented in Table 1 and Table 2.

• Download figure
• Open in new tab

Figure 1 CONSORT flow diagram of participants included in final analyses. WTA, what to accept; WTP, what to pay

Significant, negative associations were seen between VSLY and frailty when considering each VSLY arm separately as a continuous variable (See Table 3), with the median willingness to pay value being around double for people in the non-frail compared with the frail or pre frail group.

In bivariate analyses, there were significant negative associations between the combined VSLY quintiles for the whole sample and frailty; those pre-frail/frail participants were less likely than non-frail ones to report higher VSLY (See Figure 2). Being male, married, and educated were significantly positively associated with increasing VSLY quintiles. Increasing age, disability, and quality of life, were negatively associated with increasing VSLY quintile. We saw no association between wealth and VSLY (Table 4).

• Download figure
• Open in new tab

Figure 2 Association of frailty status and VSLY quintiles. Q, Quintile

Multicollinearity between variables was excluded. Sequential binary logistic regression models showed a significantly greater odds having a high VSLY in those who were non-frail compared to frail or pre-frail. This became non-significant for the lower VSLY quintiles (Table 5). The inverse linear relationship between frailty and VSLY wasn’t substantially affected by adding socio-demographic or economic variables, or quality of life, or multimorbidity, or disability to the model (Table 5). Forest plots showing the relationship between frailty and VSLY for each set of quintiles comparisons in each sequential model are shown in Appendix Fig. 2, Supplementary Information 3. Frailty ORs from the fully adjusted sequential model (Model 5) in each comparison of VSLY quintiles are shown in Figure 3.

• Download figure
• Open in new tab

Figure 3 Forest plot representing the Odd Ratios for frailty in each comparison of VSLY quintiles, controlled for all covariates

In addition to frailty, being female or single were significantly associated with a lower VSLY whereas being older, educated, or wealthy had no statistically significant association. These associations were not substantially modified by adding quality of life, multimorbidity, disability, or all together to the model. Higher quality of life was significantly associated with a lower VSLY, greater disability was associated with lower VSLY, whereas there was no significant association between VSLY and multimorbidity (see Table 5 and Appendix Table 2, Supplementary Information 3).

The results for the multinomial regression done for the sensitivity analyses were not substantially different to the binary logistic regression models (See Appendix Table 3 and Appendix Fig. 3, Supplementary Information 3).

4. Discussion

We found a strong and significant association between frailty and VSLY in this population in Burkina Faso with people who are frail assigning lower values to prevent a mortality risk than those who are non-frail (robust). This association persisted after adjustment for other variables found in other studies likely to impact upon the relationship, such as quality of life, multimorbidity, and age [17–19].

Given the global and widespread use of VSLY approaches to indicate benefits and calls to use this or similar approaches to inform investment cases for health, it is important to understand how conditions which affect health of populations in whom VSLY might be calculated impact upon it [5, 33]. Our findings have relevance to the interpretation of VSLY values in older people in low-income settings and will become increasingly important to recognise as ageing populations increase in these settings. As far as we are aware, this is the first study to assess implications of frailty on VSLY in any setting, hence our results have applicability for many countries where VSLY is being considered for use in an older population.

When frailty was included in the model, the effect of age, wealth, or education on VSLY – which were strongly associated with VSLY on bivariate analyses, became non-significant. Age and wealth have been shown to be associated with VSLY in other studies but when controlling for other variables in these studies these relationships were attenuated [9], as they did in our study. In fact, that frailty has such a strong association with VSLY—over and above that of other variables traditionally thought of as being associated with VSLY—suggests that it may be necessary to assess frailty status when estimating VSLY. This is particularly important given that frailty is a highly dynamic state, thus VSLY captured in frail individuals may change if that person changes frailty status [34].

Although not our main variables of interest, we saw associations between VSLY and other covariates that are worth discussing. Similar to our study, previous studies have shown a relationship between gender and VSLY that persists despite controlling for covariates. Women consistently value their lives lower than do men [35]. Our finding that being married or cohabitating is associated with odds of a higher VSLY is novel, but unsurprising if it is hypothesised that having family responsibilities may increase the willingness to pay to avoid risk. Another study showed that married respondents have significantly higher WTP to take care of a child/other adult than to protect themselves [18]. The relationship between greater disability and lower willingness to pay was small, but significant. Again, this finding is not surprising, given the influence of disability on people’s perceived value [36–38]. However, our finding that increasing quality of life is associated, both in bivariate and fully adjusted analyses, with lower VSLY is puzzling. We hypothesise that this might be due to older people with a higher quality of life also being comfortable with their status quo, but this hypothesis requires further investigation in qualitative studies. That multimorbidity was associated with lower VSLY in bivariate analyses but not the fully adjusted model suggests that the impact of MM on VSLY may be mostly mediated through frailty and somewhat through disability.

Our study has several limitations. Given the cross-sectional nature of the study, we can only determine associations and lack the longitudinal data needed for determining causation. We grouped frail and pre-frail individuals into one category, losing the nuance of showing effects by three categories. However, this has been done in to aid in the interpretation of results. We have also shown (personal communication) that these categories behave in a similar way when considering future outcomes of this population, so grouping them together is reasonable. The number of conditions that we used to derive multimorbidity was limited by questions asked in the survey, but they have been used in previous publications to classify multimorbidity [15]. Limited sample sizes for the individual VSLY methods necessitated combining the data using quintiles; however, reassuringly, we saw similar relationships between frailty and VSLY when the analysis was done by individual VSLY methodologies. If the inverse U shape of the association with age found in other studies, was found in our study, our adjustment for age may have introduced some errors. However, in our sample, we found a linear association between VSLY and age. We used the method of Patenaude [6] to estimate VSLY, as it has been found suitable for use in other similar settings [39]. However, we acknowledge that there are other methods that could have been applied. Finally, our study lacks the contextual grounding which would be afforded by doing mixed methods analysis and qualitative exploration to allow a deeper understanding of the context. Future studies should explore what is important to people in the Nouna region of Burkina Faso in terms of what they value in life, and how they relate their values to money.

Our analysis also has several strengths. There have not been many former attempts to study associations with VSLY in LMICs, and none that we are aware of globally which have ascertained associations with frailty in an ageing population. We used multiple ways of assessing VSLY in a fairly large sample, and included a broad range of covariates into the analysis. We used a standard way of evaluating frailty, including use of objective measures of physical performance; standard QOL and disability tools, facilitating replication in other settings.

In conclusion, ageing is a neglected phenomenon in many LMICs, where research has tended to focus on diseases of poverty and infectious diseases more common in younger people. This is especially the case in low income countries (LICs), which are typically thought of as being less far along the epidemiological transition [7]. However, our recent work has shown that populations in LICs suffer a high burden of multimorbidity, which is associated with frailty [40, 41]. Whilst our findings suggest that people who are frail place a lower value on their lives, and given that frailty status is dynamic, we recommend that frailty status is taken into account when calculating VSLY to avoid older people being “assigned” lower values than they should be.

Declarations