Initial approval was given by Newcastle and North Tyneside Local Research Ethics Committee 1 on the 24^th February 2006. The REC reference number: The Newcastle 85+ Study: Biological, Clinical and Psychosocial factors associated with healthy ageing is 06/Q0905/2. Written informed consent was obtained from participants; where individuals lacked capacity to consent, for example because of cognitive impairment, a formal written opinion was sought from a relative or carer [9].

The Newcastle 85+ Study approached all people turning 85 in 2006 and permanently registered with participating general practices (83% 53/64) in Newcastle or North Tyneside. Participating practices were similar to those which did not participate across key practice variables [7]. Individuals consented to health assessments (HA) and/or a review of their general practice medical records (GPRR). HA was conducted at baseline and then at 3 follow-up phases (at 18 months [Phase 2], 3 years [Phase 3] and 5 years [Phase 4]) by a research nurse in the participant's usual residence. GPRR was conducted at baseline, and at phases 3 and 4, in order to obtain information on disease status, medication and use of general practice GP services (Figure 1).

Reasons for non-participation at baseline and post inclusion (post-baseline) attrition -due to death and -due to withdrawal were recorded, with the latter being sub-categorised into two groups: i) health - the participant's own health status was given as a major factor, and ii) non-health (included ‘no reason given’ as participants were advised they did not have to provide a reason).

Maintaining contact and managing up-to-date contact information plays a vital role in participant retention [11], [17]. Information such as time of day, day of week, and date of any contact attempt, even if failed, was fully recorded alongside address, telephone number, and consultee or family representative details. Participants were asked to nominate an alternative contact person in case their health should deteriorate. Salient points and outcomes of conversations with participants or their families/carers were also recorded, including caring responsibilities of the participant, family dynamics and health, planned holidays and appointments, significant relationships with neighbours and key holders.

The nurse team used such information to assess whether participants, or where appropriate their family/carers, might benefit from additional brief personal visits, usually whilst the research nurse was in the area or visiting another participant in a particular care home. Balancing the maintenance of contact, as opposed to causing annoyance or distress, was founded on our actively managed awareness of participants' individual circumstances and their changing needs. All such details were held securely on an electronic database (Microsoft Access), and regularly updated and crosschecked, with the nurse team taking ownership of this task.

Additionally, strategies to keep the study ‘alive’ for participants, both between research visits within a phase and between research phases were employed, including actions recommended by others [11], [17]–[19] such as reminder letters, newsletters, telephone calls, Christmas cards, and gifts of everyday items for instance a mug and tea-towel, all displaying the study logo.

The most ambitious efforts to maintain contact and keep the study ‘alive’ included holding a celebratory event for participants and their families during phase 3 of the study. Lunch and refreshments were provided, along with entertainment from a drama group and a choir. Transport to and from the venue was made available to those who needed this. This event provided an opportunity to share preliminary research findings and for open discussion between participants and the study team. Also, the publication of a book telling the ‘story’ of the Newcastle 85+ study from participant and research team perspectives; how the study came into being, what had been discovered so far (including ‘lay’ synopsis of publications), and why it mattered [20]. A copy of the book was given to each participant along with a signed letter of thanks. In the event of participant death books were given to significant others.

To determine if particular socio-demographic and health characteristics were associated with post-inclusion attrition due to withdrawal, multinomial logistic regression was used to compare three groups across all the phase-to-phase transition points: those who withdrew for health reasons, those who withdrew for non-health reasons, and those who remained in the study. The same socio-demographic and health characteristics - years of education, National Statistics Socio-economic Classification (NS-SEC [21]), area deprivation (Index of Multiple Deprivation IMD [22]), marital status, living arrangements, housing type, self-rated health (SRH), depression, disability, cognition, disease burden (disease count – Table 1. Box 1), and health service use (comprising: GP consultations and hospital inpatient in past year) - were examined at each analysis point; these were time-varying if measured at subsequent phases and time-constant otherwise. These characteristics were selected upon existing evidence linking them to post-inclusion attrition due to withdrawal in longitudinal cohorts of older adults [23]–[30]. Initially analysing variables individually, then constructing a parsimonious model using all variables by means of a back step approach. No models were retained in the final longitudinal model (baseline to phase 4). We therefore repeated analysis for each phase-to-phase transition separately. Models were also adjusted for sex since others have shown this to impact on attrition. [31], [32] and we have already shown that very old women have significantly greater morbidity and disability [7]. Analyses were carried out in Stata 12.1 (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP).

Contact was established with 97% (1409/1459) of those invited to participate in the study and written informed consent was obtained from 74% (1042/1409) of these [9]. 851 individuals consented to undergo both HA and GPRR; 3 gave consent to HA only (declining GPRR); 188 consented to GPRR only (declining multidimensional health assessment); and 358 declined all participation. Women were more likely than men to decline either all participation or participation in HA, however the resulting 85+ cohort (those participating in HA n854) was socio-demographically representative of the local population, and of England and Wales, including the proportion in care homes [7].

Figure 2 presents the phase-to-phase retention profile for the 85+ cohort illustrating overall retention rates of 74% (631/854) baseline to phase 2 (attrition -due to death 16% (135/854), –due to withdraw 10% (88/854)), 57% (484/854) phase 2 to phase 3 (attrition -due to death 11% (95/854), –due to withdraw 6% (51/854)), and 40% (344/854) phase 3 to phase 4 (attrition -due to death 13% (114/854), –due to withdraw 3% (27/854)). Ultimately across the five years of the study 40% of participants (344/854) were lost to death. Of the survivors, two-thirds were retained (344/510 (or 40% (344/854) of 85+ cohort)) and one-third lost to withdrawal (166/510 (or 19% (166/854) of 85+ cohort)).

Figure 2 also presents results for men and women separately, where for both sexes at successive follow-up phases, we observed a decrease in post-inclusion attrition due to withdrawal and an increase due to death. Slightly more women than men withdrew at each follow-up phase ((baseline to phase 2 f:m = 12%∶8%); (phase 2 to phase 3 f:m = 9%∶7%) and (phase 3 to phase 4 f:m = 7%∶3%)) but fewer women died ((baseline to phase 1 f:m = 13%∶20%); (phase 2 to phase 3 f:m = 14%∶18%) and (phase 3 to phase 4 f:m = 20%∶30%)), though gender differences were not significant (p = 0.1325).

Table 2 presents the reasons given by the 85+ cohort (for men and women separately), or their families where indicated, for withdrawal at each follow-up phase. At transition from baseline to phase 2, health reasons accounted for 40% (35/88) of all withdrawals with ‘too ill’ being the most frequently reported reason for both men and women (83%; 29/35). For non-health reasons, accounting for 60% (53/88) of all withdrawals, ‘too much hassle’ was most frequently reported (28%; 15/53), driven by women (f:m = 13∶2). At transition from phase 2 to phase 3, health reasons increased to account for 69% of all withdrawals (35/51) with ‘too ill’ again the most frequent reason for both sexes (83%; 29/35). Non-health reasons reduced considerably to account for 31% (16/51) of all withdrawals, again driven by women (f:m = 14∶2). Of the non-health category, ‘no reason’ was the most frequently recorded response (38%; 6/16), with a f:m ratio of 5∶1, and ‘lost interest’ a close second (31%; 5/16) populated entirely by women. At transition from phase 3 to phase 4, although the proportion of withdrawals due to health reasons decreased, they still accounted for the majority of withdrawals (56%; 15/27), with ‘too ill’ being the most frequently reported (52%; 14/15), and led by women (f:m = 13∶1). Non-health reasons accounted for 44% of all withdrawals (12/27), with ‘no reason’ most frequently reported (58%; 7/12), led by women (f:m = 5∶2).

Investigation of attrition between baseline and phase 2 highlighted associations with three of the socio-demographic and health characteristics examined (Table 3). Participants who belonged to the ‘withdrew for health reasons’ group versus those who continued participation were nearly five times as likely to have held routine/manual occupations during their working life (OR: 5.38; 95% CI [1.59 to 18.19]). A similar pattern was seen for those who had rated their health as fair or poor at baseline (OR: 3.09; 95% CI [1.15 to 8.31]). However, confidence intervals surrounding these odds ratios are wide due to the small numbers of people belonging to these groups. Participants were 51% less likely to have withdrawn for health reasons versus continued to participate if they had difficulty with 1 to 6 activities of daily living ((I)ADL's) compared with those who had difficulty with 7 or more activities (OR: 0.49; 95% CI (0.27 to 0.91)). When examining attrition between phases 2 and 3, participants withdrawing for non-health reasons were more likely to live in areas with higher deprivation (IMD) or in ‘institutional care’ housing type when compared with those continuing participation. However, numbers were low in these groups giving wide margins in the calculated confidence intervals. When examining attrition between phases 3 and 4, those who suffered from severe depression were more likely to withdraw for health reasons than those continuing participation; again a wide confidence interval prevents drawing any definitive conclusion. Constructing a parsimonious model yielded results indicating that only NS-SEC showed an association with post-inclusion attrition between baseline and phase 2 after adjusting for sex, with participants belonging to the ‘withdrew for health reasons’ group being significantly more likely to have held routine/manual occupations (OR 3.41; 95% CI [1.23 to 9.44]) than those who remained in the study. There were no detectable associations with participation status at the phase 2 to 3 or phase 3 to 4 transition points when all variables were considered together.

A particular challenge to maintaining participant contact was the use of a telephone system withholding caller ID, as does the Newcastle University system. This problem was detected when participants, first labelled as ‘unable to establish telephone contact’ and then successfully contacted by alternative approaches, disclosed that they did not answer the telephone, or had been instructed by their family not to do so, if caller ID was withheld. To resolve this, in addition to contact protocols stipulating that telephone calls be made on alternate days and at different times of the day [9], the nurse team attempted at least one or two calls using a mobile phone.

Moving home is commonplace among older populations, for example to live in closer proximity to family or to move into a care home [40]. On average 6% of the Newcastle 85+ Study cohort changed address at each of the 3 follow-up phases (40;49;57/854), with over 50% of these being ‘new’ transitions into care homes from private homes. Clearly, change of address is a challenging issue for researchers attempting to remain in contact with study participants. Other longitudinal studies involving older people report change of address as a major contributory factor of loss to follow-up accounting for 9% of total attrition over 4 years [27]. Across the duration of our study, only 0.5% (5/854) of participants were lost to follow up. This was made possible through effective engagement of various ‘gatekeepers’, often in combination. Researchers are urged to use great sensitivity when participants move into care, as these transitions are often associated with deterioration in health and independence. In such situations, family emotions of anxiety and guilt are not uncommon and notifying researchers of a change of address is not a priority. Maintaining a good relationship with general practice (GP) staff and, where relevant, care home staff proved valuable, as they were supportive in forwarding study contact letters asking participants or relatives to contact research staff. A further resource in re-establishing contact with participants who changed address existed for those who had consented to having their records flagged with the Health and Social Care Information Centre. This enabled us to identify and contact a new registering general practice, again to request the forwarding of a study contact letter. If the new practice was not already participating in, or indeed aware of, the study, we reverted to our GP recruitment protocol [9], guiding surgeries through the aims of the study.

Feedback not only relates to providing participants with ongoing information but also to the collection of information from the participants' perspectives. Where given, we formally captured reasons for withdrawal, as well as the experiences of participants who continued with the study, asking at the end of each health assessment ‘How did you find this interview?’, and recording responses verbatim.

Effective use was made of this information through ‘reflexive practice’ techniques at scheduled weekly meetings of the research nurse team. This involved questioning our own attitudes, thoughts and reactions in relation to others and, if necessary, revising our ways of relating [41], [42]. Although participant withdrawal cannot be eliminated, this exercise increased the confidence of our research nurses in recognising and actively seeking to remove barriers to retention (Table 4).

In agreement with Patel et al [14] that research can be a solitary occupation, we found the continual motivation of staff to be crucial. EJiogu et al [43] offered specific recommendations such as staff incentives, including supporting pursuit of academic goals. We found this incentive successful not only in maintaining staff motivation but also in enhancing team knowledge. Other incentives included target-related bonus schemes, especially to compensate staff nearing the end of a fixed term appointment and postponing the search for alternative employment.