We have found an association between preferential status and the likelihood of getting two out five chronic conditions – COPD and diabetes, but not dementia, hip fracture and Parkinson’s disease – and also with the probability of dying. We also found that preferential status is strongly related with home care use. For residential care the relationship is weak for men and non-existent for women. For death and home care use, the association with preferential status declines with age, such that (within the population studied) it is strongest for those aged 65, and near zero for those aged around 90 and older.
As explained in the methods section, we interpret (initial) preferential status, which is conditional on low income, as a measure of socio-economic status. The observed effects of preferential status on COPD, diabetes and death can then be interpreted as instances or consequences of socio-economic differences in morbidity and mortality. A discussion of the possible mechanisms which could be responsible for these differences is beyond the scope of this paper (see for example  for a review). For COPD and diabetes, it is plausible that smoking, unhealthy food and other life-style factors could be involved. The fact that preferential status is a dichotomy is an important limitation of our study, as it makes it impossible to find a gradient in its association with chronic conditions and long-term care use. Another limitation is that the presence of chronic conditions is not observed directly, but imputed on the basis of medicines or medical care use. Some medicines or treatments might be cheaper for patients with preferential status than for others.
An interesting finding is that the effect of preferential status on mortality is not mediated by the five chronic conditions that could be identified in the data, even though most of those conditions are shown to be important predictors of death. This suggests that other health problems play a role here, with heart problems being a prime candidate. Unfortunately, the data that would allow us to check this hypothesis are lacking. In a similar vein, we interpret the observed effect of preferential status on home care use (and for men on use of residential care) as a consequence of the worse health of persons with low incomes, given age, sex, living situation and province of residence. Yet, this supposed worse health is captured to only a limited extent by the five chronic conditions mentioned.
We have found that the effect of preferential status declines with increasing age, both for death and for home care use. One must be careful with the interpretation of such interaction effects in logistic models, since they can be an artefact of the functional form chosen . If a linear specification (without interaction terms) would in fact be correct, then interaction terms might well be significant if the model is estimated using a logistic equation. However, other analyses not shown here indicate that the effect of preferential status on death and home care is indeed fairly substantial at ages 65–75, and not only not significant, but also near zero at ages over 85. This is true when this effect is measured in terms of odds-ratio’s (which we use implicitly when applying logistic regression) and also when we look at simple differences between rates. One possible interpretation of this finding is in terms of survivalship bias, or selective mortality. Suppose that the population is in fact composed of two groups, one at high risk of death (say, because of heart problems), and another one at low risk, but that membership of these groups is not observed. Among persons with preferential status, the high-risk group would represent a higher proportion. As persons age, the high-risk group falls more often prey to mortality, and only the low-risk group is left. At that stage, no effect of preferential status on the risk of death would be measured. Such a mechanism could also explain why the effect of preferential status on use of residential care is much smaller than the association with home care. Persons enter residential care generally at age 85 or older, while first use of home care is registered for many older persons below that age. In other words, the reason that we find that persons with preferential status are not more likely to move into care homes than those without that status (and are also not more likely to get dementia) is that the former tend to die before they attain the age at which those events commonly occur. This would be an instance of what in survival analysis terminology is called ‘informative censoring’ : conditional on observed variables, those persons whose observation periods are censored by dying would have been more likely to experience the event of interest (entering residential care) if they had continued to live, compared to those who do not die. It is important to stress that such an interpretation, if correct, does not change the evaluation of health inequalities in a life-course perspective. If differences in the likelihood of starting to experience health problems by socio-economic status are larger at younger than at older ages, this does not change anything for a birth cohort that will pass through all those ages.
Alternatively, one might interpret the effect of preferential status on the use of home care in terms of prices. For persons enjoying preferential status, co-payments for this kind of care are reduced, and this might induce them to use it more frequently, or at lower levels of need. The difference in prices is not negligable, about 4 € per day for standard packages of home care . On the other hand, many persons receiving home care do not have to pay co-payments, irrespective of preferential status, as the nurses do not always charge these, or because those persons are covered by the system of maximum billing (which puts a ceiling on the total amount of co-payments during a calendar year). There are no co-payments for care in residential settings, so in this respect the limited effect of preferential status on residential care use is in agreement with the economic interpretation in terms of prices. In addition, there might be an income effect, as persons have to pay from their own resources the substantial costs for bed and board in care homes. Older people with low incomes might be less inclined to enter residential care for this reason, especially if they are unwilling to relinquish their own home at the same time. However, such an interpretation requires an additional explanation for why this supposed price effect would be much smaller, or non-existent, for the very old than for the not so old. Also, differential prices cannot explain why persons enjoying preferential status die at younger ages than older persons without that status. So the principle of scientific parsimony would favor the health interpretation of the effect of preferential status.
Moreover, these rival explanations have a number of different implications which can be tested. For instance, if lower co-payments would induce persons with preferential status to use home care at lower levels of need, compared to other persons, then persons with preferential status should be more likely to use home care at a low level of intensity than others, since the provider decides on the level of home care provided (subject to periodical checks by the insurer). In a logistic regression with the level of home care as the dependent variable, conditional on receiving home care, preferential status had no significant effect, however. Also, if the interpretation in terms of prices of the effect of preferential status would be correct, then within the group of persons receiving home care at a low level those having preferential status would be less likely than those without that status to make the transition to either home care at a high level, or to death. Again, in analyses of these transitions, preferential status had no significant effect (results available on request). One must keep in mind, though, that due to the much smaller sample sizes the power of the significance tests was lower than for the analyses reported in the body of the paper. Of course it is also true that these interpretations are not mutually exclusive, and both may operate in the real world.
We have also seen that the effect of preferential status is consistently smaller for women than for men. A possible reason for this finding is that preferential status is a better indicator of socio-economic status for men than for women. Almost all men in this age group have worked for most of their active lives, so a low income in old age is an indication of low earnings during that period, and therefore of less favorable occupations and educational levels. On the other hand, many women may have been housewives for a large part of their former lives, irrespective of their own education and occupation, or those of their husband. A low income in old age may be less correlated for this reason with those other indicators of socio-economic status.