VLV data have been harvested in five cantons representing the linguistic, political, and socioeconomic diversity of Switzerland. The objective was to question 3600 individuals aged 65 and over living in private households or in nursing homes (in Switzerland, “établissement médico-social,” or EMS).
The individuals were randomly selected from the official lists of inhabitants, stratified by sex and according to six age groupsFootnote 2 in each concerned canton in order to ensure the representativeness of our sample within each stratum, but also to have sufficient numbers for each sub-group. Table 2 shows the distributions of the completed interviews.
VLV has been a homemade product not only in its conception but also in its concrete realization. The research team recruited interviewers, organized their training, and supervised the fieldwork from the headquarters in Geneva, with two or three young researchers delegated to each of the five cantons to be in charge of the daily management of a little enterprise, attributing the sample units, controlling the quality of the work done by the employees, saving and transferring the various collected data, etc. This decision not to outsource to a private company was rooted in a tradition and justified with three reasons. Indeed, the 1979 and 1994 surveys were done the same way with approximately the same justifications: transparency, costs, and ideology. First, the idea was to keep complete control over the entire process of data collection from the first to the final stages. The statistics shown in this chapter are byproducts of this will. Second, it is likely that outsourcing would have been more costly, as will be illustrated through the insistence strategy discussed below. Third, from 1979 until now it has always been the fundamental choice of the teams to refuse the distinction between the “noble” intellectual stages of research and the “lowest” ones, those stages concerning the hands-on technical and logistic aspects of carrying out a survey (see Bétemps et al. 1997; Nicolet and Oris (forthcoming). Simply said, managing an in-house survey is carried by the belief that researchers should not limit themselves to analyzing and theorizing about social science data; they should also go out and get their hands dirty while collecting it.
In the following, we discuss the survey procedures and contact strategy, the coverage errors, the refusals and their causes, and to what extent the procedures and their adaptations were efficient to “capture” vulnerable persons in the sample of participants, and we conclude with a comparison of the prevalence of socioeconomic and health vulnerability in VLV and in other data sources.
3.1 Procedures and Contact Strategy
To ensure data quality, a clear approach procedure that could be translated into the three languages was put together. The procedure took into consideration a number of situations that the interviewers would encounter:
EgoFootnote 3 is apt (has the capacity to answer) and lives at home;
Ego is apt and lives in a nursing home;
Ego is unapt (at home or in a nursing home);
Ego does not speak French, German, or Italian.
Before the procedure was launched, the sample members had to be allocated to the interviewers. For that purpose, the interviewers were given contact sheets that contained confidential information concerning ego (surname, forename, address, and phone number). The interviewers had to indicate each contact attempt, the date, whether there had been a visit or phone call, with whom the interviewer had spoken, the result, and in the case of a refusal, the reason(s) (Fig. 1).
The most concrete aspect of the approach procedure was making contact. The two first stages were common to all procedures, with small variations. First, ego received a leaflet presenting the study and a personalized contact letter announcing that an interviewer would phone. In the event that the respondent lived in a nursing home, the first letter was sent to the nursing home direction to inform that one of their residents had been selected to take part in the VLV survey and that an interviewer would contact the management before any other step was taken.
The question arose of which was the best solution for the first contact with ego: to phone or to make a visit at home? For the entire survey with its five regional fields, our team decided on an initial phone contact to ensure the best comparability with the 1994 survey and for other very pragmatic reasons. In some cantons, the distances to be covered for each visit, especially if ego was living in a mountain village, could be long. This would have led to a substantial increase in the costs and would also have increased the time devoted to the survey by the interviewer who would have been reluctant to make such an effort with uncertain rewards.
A home visit was done in only two cases: firstly, when ego had no phone number and, secondly, when ego could not be contacted by phone for 2 weeks. It emerged that 5–26 % of elderly people do not have a phone number indicated in public phone directories, the proportion reaching its maximum in villages in the mountains of Valais, Ticino, and Bern Oberland. In Switzerland, for a population of eight million inhabitants, some four million phone solicitations are made each year for surveys, very often for marketing purposes. Combined with aggressive selling and the increased use of mobile phones, this creates an increasingly difficult environment for scientific surveys (Joye et al. 2012). In this respect, the initial letter was crucial in allaying certain fears.
When contact was made with ego, the interviewer could have been confronted with two situations: the person was apt, i.e. able to answer, and could therefore decide whether he or she agreed or refused to participate in the study, or ego was non-apt. During the phone call, the interviewer could already be able to detect whether ego had cognitive problems by asking simple spatial orientation or temporal questions:
“Could you suggest a date for the appointment?”
“Is there an entry code for where you live?”
“Could you explain how to get to where you live?”
“Could you remind me of your address?”
If ego appeared not to have the capacity to answer, the interviewer had to activate the so-called proxy procedure, asking the assistance of a close friend, relative, or caregiver to answer a short questionnaire of some 60 questions. Similarly, if ego was living in a nursing home, when the interviewer contacted the management, he or she would obtain information on ego’s health, that is to say his or her ability to participate in the survey. If the reply was affirmative, a first contact letter was sent to ego to provide the person, just as the other sample members, with information on VLV and to allow him or her to express informed acceptance or refusal to take part. Otherwise, the proxy procedure was applied.
If ego was no able to answer, it was rare that the person could be directly contacted, but contact was made with a spouse, family member, guardian, or the nursing home. In such a case, the interviewer would suggest to one of the aforesaid to participate in the survey by replying to a limited number of questions on ego. This “proxy” procedure, as indicated above, was initially the only procedure to diverge from the standard one. It was essential to avoid the trap, still too frequent in gerontology surveys, of not gathering data on individuals who are in very bad physical health and/or suffering from cognitive problems. It was all the more crucial to ensure correct comparability with the 1994 survey, since the weight of diseases such as Parkinson’s or Alzheimer’s has constantly increased in the causes of death of the very old (80 years and more), in Switzerland as elsewhere (Berrut and Junker 2008). This procedure allowed us to interview 555 close friends, relatives, or caregivers, which represents somewhat more than 15 % of all the interviews collected (Table 3). Without this special procedure and its adapted questionnaire, an entire segment of the population would have been excluded and the sample of respondents seriously biased.
In the framework of the standard procedure, once contact had been made with ego or a proxy, several situations could arise:
Ego had died. The file was then closed for reason of death;
Ego was no longer living in the same place, with three possible situations:
Ego had moved outside the territory concerned by the survey; his or her file was closed under the category “no reply” (NR),
Ego had moved but had remained in the area of the survey; the first letter was forwarded to the new address,
Ego had moved into a nursing home; the contact letter was sent to the director of the nursing home,
Ego had not received the first letter, so it therefore had to be sent again and ego had to be called a few days later;
Ego had received the letter.
When the interviewer was finally in situation 4, he/she was faced by one of three situations:
Ego refused. The file was closed due to refusal;
Ego hesitated and wished to see the questionnaire before deciding;
Ego accepted and an appointment was made with the interviewer.
In cases b and c, a second letter was sent to ego. It contained:
the self-administered questionnaire (SAQ)
the personalized life history calendar (LHC)
the information leaflet explaining in greater detail the survey and the confidentiality of personal data
If an appointment was made, a letter of confirmation was sent.
Following the second letter, ego may:
refuse by calling the interviewer or the cantonal office;
in situation b, accept and set up an appointment with the interviewer.
3.2 Coverage Errors and Mistakes in the Samples
Figure 2 shows the concrete results of the procedures described above. We will discuss the refusals in the next section. Here, we deal with the coverage errors that affected no <13 % of the original sample.
People who cannot be found at the address provided by the cantonal and federal population or statistical offices reflect discrepancies between the target population and the sampling frame. These problems can be explained by the length of the VLV field and consequently a growing temporal duration between the samples extracted from the lists of inhabitants, but also because the latter are theoretically updated every 3 months, and obviously less often in some municipalities. Other discrepancies are due to the very nature of ageing, such as risks of dying or moving to an institution between the date of the population enumeration and the date of a contact attempt by a VLV collaborator.
On the ground, it was decided that a home visit would be a last resort when unable to contact ego by telephone. Therefore, if there was no phone number and no reply after 2 weeks, the interviewer contacted the relevant commune (municipality) to find out whether ego was still living in the same place, had moved or had died. The communes and post offices (sometimes the priests also, and neighbors) proved to be very helpful throughout the data collection stage. Another source of coverage error appeared during the fieldwork. In Geneva, it happens that people have a postal box in the canton but in reality live in France, the former being their formal address, explaining why they were on the lists of inhabitants and were included in our sample. This comes from a law regarding civil servants that was abrogated some years ago, but the practice survived. In Valais, several persons have their formal residence in a chalet in the mountains while it is in fact a secondary residence. Usually this is for fiscal purposes. At the end, in 14.1 % of the 4105 files opened in Geneva and Valais, it was impossible to make contact with the person. For the five cantons altogether, 10 % of the files were in this situation, which is a bit less than the average proportion observed in European countries in the 2004 round of the Survey on Health, Ageing and Retirement (De Luca and Peracchi 2005, 94).
Those cases were costly in terms of management time. From the attribution of the contact sheet to the interviewer to the transfer of the complete interview to the headquarters, it took on average some 40 days for a full participation in the survey; the length was 60 days in cases of non-contact (NR in Fig. 2), to harvest nothing at the end. It was consequently also very negative for the motivation of the interviewers who had to deal with those cases.
3.3 Many Refusals
Figure 2 shows that we ended with 54 % refusals against 33 % acceptances. For Groves and Couper (1998), individuals who refuse to take part in a survey are more likely to be uninterested in the topics of the survey; not have the time; or find it difficult to understand the language of the questionnaire, which would indicate a low level of education, or different origins and partial learning of the host society’s language, or both. Analysis of the reasons for refusals bears out their first judgments (see Fig. 3, which reports the percentage of men and women citing each reason). For Geneva and Valais in 2011, the refusals “without reason” dominated, but among those who gave some justification, a lack of interest was top of the list, which justified more than one refusal in five. This was followed by health problems, being “too tired,” or being “too old,” with around 14 %; and inversely, being too busy with 8/9 %. Personal and family reasons, which often expressed a desire to protect one’s intimacy, varied between 6 and 8 %; and a clear rejection of surveys (“I don’t want to be a guinea pig”) was manifested by 7/8 %. Men were somewhat more likely to refuse without explanation, while women more frequently indicated their absence of interest or desire. Globally, however, an obvious gender pattern does not emerge.
As we can see in Table 4, the two main causes are not particularly affected by age, but among the oldest old, 31 % of the nonagenarians refused without explanation. As we could expect, health problems or feeling tired or too old is a reason that increases with age. Personal and family reasons show the same pattern, being a marginal factor among those aged 65–69, then growing in importance. Protecting intimacy is more of a concern for the oldest old. Interestingly, the rejection of the survey appears as more “modern”; it is more of a reason in the recent cohorts that in the older ones. Refusing “to be a guinea pig” and similar expressions reached 11 % among the 65–69 cohort and then continuously fall to 4 % among those aged 90 and more. As we could expect, being busy with work, holidays, or other activities was also important among those who had just retired and strongly decreased, becoming marginal from the age of 80.
This issue is crucial, since the credibility—and consequently, the survival—of surveys as a tool for the social sciences seems to be engaged when so many potential participants refuse to contribute. However, it is difficult to go deeper into this issue right now, because we know very little about these persons, aside from the little data provided by the administrative files when the samples were delivered. Some logistic regressions on refusals and acceptances, respectively, are not shown because they only brought limited additional information: higher participation of men and people living in urban areas, and those aged 80–84 showing the lowest propensity to contribute. In terms of interactions, the sex of the interviewer (whatever the sex of the sample member) and his or her age had no impact. While an initial recruitment condition to work for VLV was to have a bachelor’s degree in social sciences (in a broad sense), it appears that the interviewer’s level of education had no effect. Only the interviewer’s accumulated experience increased the probability of obtaining an acceptance from the sample members, but this is a tautological result, since those with few successes were discouraged and gave up.
3.4 Adaptations and the Inclusion of Vulnerable Populations
Faced with so many refusals, the VLV team decided against using standard ‘refusal conversion’ efforts (see Groves and Lyberg 2010, 872) for ethical (ideological) reasons. A “no” has to be respected. However, to reduce the number of negative answers as much as possible, we chose to have some interviewers specialize; they were requested to make the first phone contact, not only for themselves but for the entire team. The crucial nature of this first interaction has been studied for some 15 years (see Snijkers et al. 1999; Durrant et al. 2010) and now appears to be an explanation for national variations, as well as being likely to affect the comparability of results (Blom et al. 2011). In our case, certain interviewers clearly proved to be more effective than others in obtaining the cooperation of potential participants. Figure 4 shows the success rates per interviewer in Valais. A similar distribution was observed in the other four cantons. With a large majority of collaborators showing an average level of efficiency, with approximately one in every four being clearly less efficient, several champions emerge. In each field, they became famous as heroes of the survey journey. By distributing the appointments obtained by the more persuasive collaborators throughout the entire team, we managed to prevent any increase in interviewer effect. This is a typical illustration of our initial compromise: combining maximum flexibility to obtain as many acceptances to participate as possible, but applying the same questionnaires and interview mode to all of the participants, apart from those unable to answer, as rigorously as possible.
A second strategy was to show endurance and accept large extensions in the duration of the fieldwork. It was our wish to never force an individual to take part in the survey, but also to insist as much as possible until having a clear reply directly from ego. This obstinacy was costly, both in time and money. When the fieldwork was launched, it was estimated that we would need around 3 months per canton to gather the data—in the end, it took us 8 months per canton. Table 5 explains why. In Geneva and Valais, 4105 people were contacted to obtain 1428 acceptances (including both proxies and “normal” interviews). Roughly 25 % of the latter were obtained following five or more attempts to contact ego by phone or by a visit to the home to request his or her agreement. Generally, in surveys repeated calls is a well-known strategy to reduce noncontacts, while home visits are barely part of the procedures. Indeed, Groves and Lyberg (2010, 872) rightly insist on the costs of obstinacy. Take the example of a car drive through a mountain village with uncertain results. However, without those home visits and knocking directly on the door of ego, we would have lost almost 8 % of the final participants, and without calling five times and more (until 23), we would have missed 17 %. The potential related biases are assessed below.
This question of bias is important and less obvious that it seems at first glance. Indeed, if certain sub-populations reply less than others and efforts to increase the general response rate are made without taking this into account, such efforts would potentially result in an increase in the selective bias (Peytchev et al. 2009, 786; Roberts et al. 2014). However, when walking the tightrope between ethical evidence of respect for refusals and the wish to obtain a reply directly from ego, whatever this reply may be, could we consider that our procedures worked? Can we conclude that insistence and recourse to proxies allowed more vulnerable individuals to be included in the survey? Of course, we assume that the risk of capturing too many vulnerable people is illusionary.
To provide initial replies to this question, Figs. 5 and 6 represent the distribution of interviews obtained by age group, for men and for women. By differentiating the different procedures according to age and gender, we are able to conduct a more detailed analysis on the impact of our perseverance. We also take into account the proxies and individuals who were contacted by the standard procedure. We therefore have five categories: 1 or 2 calls to obtain participation, 3 or 4 calls, 5 calls and more, a visit (to ego’s home), and proxy. In addition, Table 6 shows the statuses of functional health and each of the respective depression categories, which in both cases are distributed according to the type of procedure that allowed information to be obtained.
Among the participants, women aged 75–79 and men aged 75–84 were the most difficult to contact by phone. The 75–84 cohort required the heaviest procedure of a home visit most often. An initial explanation is linked to the availability of a phone number. In Valais, we had no phone numbers to contact roughly 22 % of the women in this stratum. Both men and women aged 75–84 were also the groups in which the rates of refusal to participate in the survey were the highest, with reasons that mixed those of the oldest old and the young old (Table 4). Taking into account the evidence collected in another study (Duvoisin et al. 2012), we put forward the hypothesis that a fair proportion of individuals aged 75–84 tend to live their experience of ageing negatively. They are affected by biological changes, realize that their losses exceed their gains, and suffer from this evolution, which leads them to refuse useful offers from associations working for the elderly; similarly, they are also more likely to refuse to participate in a survey “on the old” (Duvoisin et al. 2012). This interpretation requires more research, of course, but all elements at this stage point in this direction. In any event, the complicated and costly procedure of home visits allowed the survey to include relatively isolated individuals in borderline age groups between what is widely referred to as the 3rd and 4th ages, that is to say, between 75 and 84.
In the same order of ideas, repeated phone calls—from 5 to as many as 23—mainly appeared useful for contacting and convincing the “young old,” exactly those whose refusals were largely explained by a lack of time, holidays, or work. This is all coherent, and we must also give credit to obstinacy. Whilst this observation is valid for both sexes, the greatest effort had to be made to reach the men. Participation was obtained after five or more phone calls in around 15 % of the cases, versus a little <9 % among women for this type of recruitment. This originates from two phenomena. One is mechanical, since the number of proxies is lower on the male side (16.5 %, against 21.3 % for women). This confirms the well-known health–gender paradox, according to which women live longer than men but age with worse health (see Van Oyen et al. 2013). Another explanation has more of a psychosocial nature, since experience in the field showed that wives often blocked phone calls, feeling the need to protect their husband (who were generally older than they were) from aggressive phone canvassing or suspected risks of invasion of privacy. This does not emerge very clearly on Fig. 3, for which we wished to avoid overly subjective coding.
The proxy procedure obviously centers on the oldest individuals. Initially, it may appear less heavy than a standard procedure (shorter questionnaire, fewer letters to be sent), but the Valais figure for the time spent on each case shows that this approach required great perseverance to obtain the desired result, since it required just as much time, if not more, depending on the strata. An average of 43 days was needed to close a proxy file for the men, whereas 38 days were needed for a normal procedure. It could take time to discover that ego was non-apt (and/or moved into a nursing home), and contacting a close friend, relative, or caregiver of ego and making an appointment was more complicated than with the retired sample members, since at least 50 % of the proxies had a job.
With Table 6, we touch even more directly the central question: the inclusion of vulnerable people. Whilst we could have hoped that our insistence on contacting the sampled individuals by repeated phone calls would allow us to capture more vulnerable individuals, the results on functional health, in fact, point in the other direction: when we repeatedly insisted on participation by calling repeatedly, there were more independent individuals than those among the files obtained easily (after 1–4 calls), as many people in difficulty, and fewer dependents. This can be explained by what Figs. 5 and 6 showed—that is to say, the relative youth of the respondents recruited in this way. In their case, by seeking out dynamic seniors in this way, we tended to avoid a negative bias that would have underestimated their global good health and numerous activities.
Home visits, on the other hand, allowed significantly more people to be recruited whose functional status was described as “in difficulty,” which can be considered a latent state of vulnerability. But the proxy procedure—although requiring great insistence—turned out to be even more important in accounting for the vulnerable aged than we expected. As a reminder, somewhat less than one file out of six was obtained in this way, but no <70 % of the individuals included in this way were in a situation of dependency.
When we look at the second panel of Table 6, which describes categories of mental health, we find confirmation that repeated calls did not help to capture more vulnerable people, when compared with those whose participation was obtained easily, which already included some 15 % of persons who were depressed but participated anyway. Home visits were once again—and in fact, much more clearly than for functional health—useful for obtaining the contributions of those in a latent state of vulnerability—the “worried”—to the survey. Additionally, the proxy procedure confirmed its usefulness for including people who were obviously vulnerable, with 4 or more symptoms of depression. However, Table 6 also shows a trade-off: in 32 % of the cases, we were not able to establish the category of mental health because of item non-response. Indeed, the questions that constitute the Wang test of psychic health were the only ones about ego’s feelings in the proxy questionnaire, due to the importance of this dimension in any assessment of well-being among the elderly. But we can understand that approximately one in three proxies refused or were unable to guess ego’s pleasure, sadness, etc.
At this stage, it is fair to note that the inclusive approach based on recourse to a proxy was recently criticized. In this critical perspective, an individual whose cognitive capacities have deteriorated should be excluded from the process; indeed, such an individual cannot really give his or her informed consent and someone who ego has not necessarily designated replies to questions concerning ego. Depending on how close this person is to ego, the social situation, socioeconomic status, and, a fortiori, past life of ego will be documented with some degree of uncertainty. It is this “silence by proxy” (Fillit et al. 2010) that is thus being denounced. An answer could be that the absence of impaired elderly in the survey, the scientific results, and ultimately the citizen debates about the social and political management of ageing would be another form of highly negative silence. At the same time, the ethical question, i.e. the point at which the individual’s consent was informed, cannot just be rejected.Footnote 4 We face here a real tension.
3.5 Prevalence of Vulnerabilities in VLV and Other Data Sources
To the instructive but indirect evidence discussed above, this section adds the results of a benchmark approach, comparing the prevalence of forms of vulnerability in the VLV’s final sample of respondents with the numbers that can be found in other data sources.
Being poor is an obvious state of vulnerability, since the poverty line is fixed by a confederation of Swiss institutions of social help and applied all across the country. This threshold is set at a monthly income of 2450 SFr. per person. In VLV, we used 2400 to simplify, and the poverty rate in the sample reached 17.9 % for men and 23.3 % for women.Table 7 offers a comparison with four other studies. Variations can generally be explained with the use of different sources, definitions, and temporal differences. However, between the SILC use of the 60 % median income level and the SKOS/CSIAS (2013) threshold, the difference is marginal (2450 versus 2500 SFr. in 2012). It can also be pointed out that there is no significant evidence that the “Great Recession” from 2008 has affected the economic well-being of the Swiss elderly, as it did in many other European countries (Cavasso and Weber 2014). Globally, we can conclude that VLV faced the challenge of fair inclusion of the poor in the survey with success.
For functional health, the VLV estimates can be compared with two recent reports from the Swiss Federal Statistical Office on the elderly living in private households (Bundesamt für Statistik [OFS/BFS] 2014) and in institutions (Bundesamt für Statistik 2012). According to the findings of the VLV survey, 86.9 % of people aged 65 and more were found to be independent, as opposed to the 91 % claimed by the first OFS/BFS study. Moreover, VLV estimates that 6.9 % of elderly people are in a situation of difficulty, meaning that they are generally able to accomplish all activities of daily living themselves but uneasily for at least one of them. Additionally, an estimated 6.1 % are no longer capable of performing their daily activities independently, meaning that they require external help with at least one activity. Contrasting these results, the BFS/OFS report found 7 % of elderly people in Switzerland with difficultiesFootnote 5 and 2 % being completely dependent. Based on the comparison with this first report, VLV seems to have slightly over-reported the prevalence of dependence in the population. The differences between the VLV data and the aforementioned report, however, are most likely due to differences in their samples: whereas VLV also included people living in nursing-homes, the OFS/BFS study exclusively considered those living in private households.
In the second report that focused exclusively on elderly people living in care institutions (Bundesamt für Statistik 2012) the numbers change drastically: 96 % of individuals were found to have trouble accomplishing at least 1 activity of daily living; thus, only 4 % of individuals were completely independent. The analysis was restricted to a population that was no longer living in their own households due to the importance of their restrictions in managing their daily lives, which explains these very high percentages. Given these two selective and therefore biased sources, however, it seems plausible that VLV generally captured a relatively representative sample of the general population, in terms of functional health, especially thanks to the proxy procedure.
Finally, comparing the mental health findings in VLV with those from other sources was not a straightforward task, since there are many indicators to measure depressive symptoms, different classifications to determine the actual categories and statuses of psychic health or to measure the intensity of depressive disturbances, and generally very little research on the topic of elderly people with depression. A relatively exhaustive report on depression among the Swiss population by the Swiss Health Observatory (OBSAM) estimated the prevalence of depressive symptoms among the elderly population (Baer et al. 2013). The results are summarized and contrasted with VLV figures in Table 8.
Despite the cautiousness needed when comparing these findings, it can be seen that depression was reported more often in the VLV data. Both studies found that roughly 75 % of the population was free from any symptoms of depression in the 65–74 age group. VLV, however, lowered the estimation of persons with minor signs of depression and suggested that there were more than three times as many people suffering from intermediary to strong depressive symptoms in this age group, compared to the OBSAN report. Among those aged 75 or higher, individuals who are not vulnerable or in a latent state of vulnerability were proportionally less represented in the VLV data, where we once again find much higher estimates for people who were obviously vulnerable, from a mental health point of view.
The challenge was to avoid or at least limit the risk of differential rates of answers between vulnerable and secure sample members leading to the exclusion of the former. Although the rate of participation for VLV was low and although it is not really possible to provide a strict, definite answer with the comparisons discussed above, all of the evidence is clearly positive and suggests a fair inclusion of the vulnerable elderly in the survey.