BMC Public Health BioMed Central

Research article Open Access

Less healthy, but more active: Opposing selection biases when
recruiting older people to a physical activity study through primary
care
Tess J Harris*1,2, Christina R Victor3, Iain M Carey1, Rika Adams2 and
Derek G Cook1

Address: 1Division of Community Health Sciences, St George's, University of London, Cranmer Terrace, Tooting, London, SW17 ORE, UK,
2Sonning Common Health Centre, Wood Lane, Sonning Common, Oxfordshire, RG4 9SW, UK and 3School of Health and Social Care, Reading

University, Whiteknights Lane, Reading, Berkshire, UK

Email: Tess J Harris* - [email protected]; Christina R Victor - [email protected]; Iain M Carey - [email protected];
Rika Adams - [email protected]; Derek G Cook - [email protected]
* Corresponding author

Published: 27 May 2008 Received: 11 January 2008

Accepted: 27 May 2008
BMC Public Health 2008, 8:182 doi:10.1186/1471-2458-8-182
This article is available from: http://www.biomedcentral.com/1471-2458/8/182
© 2008 Harris et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract
Background: Physical activity studies in older people experience poor recruitment. We wished
to assess the influence of activity levels and health status on recruitment to a physical activity study
in older people.
Methods: Comparison of participants and non-participants to a physical activity study using
accelerometers in patients aged ≥ 65 years registered with a UK primary care centre. Logistic
regression was used to calculate odds ratios (OR) of participants in the accelerometer study with
various adjustments. Analyses were initially adjusted for age, sex and household clustering; the
health variables were then adjusted for physical activity levels and vice versa to look for
independent effects.
Results: 43%(240/560) participated in the physical activity study. Age had no effect but males were
more likely to participate than females OR 1.4(1.1–1.8). 46% (76/164) of non-participants sent the
questionnaire returned it. The 240 participants reported greater physical activity than the 76 non-
participants on all measures, eg faster walking OR 3.2(1.4–7.7), or 10.4(3.2–33.3) after adjustment
for health variables. Participants reported more health problems; this effect became statistically
significant after controlling for physical activity, eg disability OR 2.4(1.1–5.1).
Conclusion: Physical activity studies on older primary care patients may experience both a strong
bias towards participants being more active and a weaker bias towards participants having more
health problems and therefore primary care contact. The latter bias could be advantageous for
physical activity intervention studies, where those with health problems need targeting.

Background ticipation, both for surveys (46% [1], 57% [2]) and more
Physical activity studies on older people often recruit markedly for intervention studies (26% [3], 32% [4], 35%
through primary care, such studies usually report low par- [2]). Information about non-participants is lacking, but

Page 1 of 6
(page number not for citation purposes)
BMC Public Health 2008, 8:182 http://www.biomedcentral.com/1471-2458/8/182

higher activity levels have been associated with increased the cluster option in STATA 9 [16]. For age and sex com-
recruitment [4,5], leading to potential selection bias and parisons the analyses were based on all 520 subjects. For
difficulties in generalizing results [6]. Unfortunately, analyses examining health and physical activity from the
these studies did not report on non-participants' health brief questionnaire, the comparison was based on all par-
status. Whilst older people with higher activity levels tend ticipants to the accelerometer study and those non-partic-
to be healthier [7], recruiting through primary care could ipants who received a questionnaire and returned it.
encourage those with more illnesses and primary care Analyses were adjusted for age, sex and household cluster-
contact to respond, leading to bias in an opposing direc- ing. Health measures were adjusted additionally for the
tion. Our objective was to compare the self-reported effect of physical activity (walking pace and number of
health status and physical activity levels of participants hours walked in the last week) and physical activity meas-
and non-participants in a primary care based physical ures for the effect of health (limiting long-standing illness,
activity study. number of chronic diseases, disability, falls and chronic
pain) in order to see whether any effects were independ-
Methods ent. To check that it was reasonable to combine postal
As part of a randomized controlled trial of different questionnaires on participants with those completed at
recruitment strategies to a physical activity study, 560 baseline assessment, a comparison of participants and
patients ≥ 65 years registered with a primary health care non-participants as above was repeated, restricted to those
centre (general practice) in Oxfordshire, UK were ran- randomised to postal questionnaires.
domly selected by household [8]. Those living in care
homes, those with dementia, terminal illness, poorly con- Results
trolled cardiac failure or unstable angina and those house- Recruitment rate to the physical activity study was
bound due to disability were first excluded by computer 43%(240/560). Comparison overall of participants (n =
record search and by general (family) practitioner and dis- 240) and non-participants (n = 320), adjusted for age, sex
trict (community) nurse examination of registered patient and household clustering, showed that males were more
lists. All 560 patients were invited to take part in a study likely to participate OR 1.4(95% C.I. 1.1–1.8). There was
measuring customary physical activity levels objectively no statistically significant association between age and
for a 7-day period using motion sensors (accelerometers participating, baseline age 65–69 OR = 1, age 70–79 OR
and pedometers). A random half (280) also received a 12- 0.8(95% C.I. 0.5–1.1), age = 80 OR 0.8(95% C.I. 0.5–
page questionnaire with their study information, asking 1.4).
details about physical health (general health, limiting
longstanding illness [9], disability [10], pain [11], chronic Of the 280 people sent a postal questionnaire, 116 were
disease [12], smoking status, weight and height), depres- recruited to the study and 164 were not, of these non-par-
sive symptoms [13], self-reported physical activity levels ticipants 76/164(46%) completed the questionnaire (see
[14,15] and attitudes towards physical activity [16] (See Figure 1). Table 1 shows the comparison of participants (n
Additional file 1). Subjects were encouraged to return the = 240) (116 from the postal questionnaire group, 124
questionnaire, whether or not they participated in the who completed the questionnaire at baseline assessment)
physical activity study, thus allowing a comparison of par- and non-participants (n = 76) who returned the question-
ticipants and non-participants. Those participating who naire. The results for age and sex are very similar to those
had not been randomized to receive a questionnaire, for participants and non-participants overall.
completed one at their baseline assessment.
560 randomly selected
Ethical approval & informed consent older people aged • 65
Ethics committee approval for the study was given by
Oxfordshire REC A (reference no. 06/Q1604/94). The 280 randomly sent 280 randomly not
questionnaires sent questionnaires
patient information sheets sent to all 560 individuals
explained the study in detail, including the use of ques-
164 not 124 recruited 156 not
tionnaire information provided by those returning ques- 116 recruited
recruited into into PA study recruited into
into PA study
tionnaires, but not wanting to participate further. Full PA study (questionnaires completed
at recruitment)
PA study

informed written consent was obtained from the 240 par-

ticipating in the physical activity study. 76 returned
88 did not
return
questionnaires
questionnaires
Analysis
Logistic regression was used to estimate odds ratios for Figure 1 flow through the study
Participant
participating in the accelerometer study, adjusting for age Participant flow through the study.
and sex as appropriate and household clustering, using

Page 2 of 6
(page number not for citation purposes)
BMC Public Health 2008, 8:182 http://www.biomedcentral.com/1471-2458/8/182

Table 1: Comparison of non-participants & participants of physical activity study amongst questionnaire responders

Non-participants N Participants N = Crude OR for OR (95% CI) adj for

= 76 n (%) 240 n (%) participating (95% age, sex, household
CI) clustering1

DEMOGRAPHIC
Sex
Female 45 (59.2) 115 (47.9) 1 1
Male 31 (40.8) 125 (52.1) 1.6 (0.9–2.7) 1.6 (1.1–2.4)
Age
65–69 23 (30.3) 87 (36.3) 1 1
70–79 40 (52.6) 111 (46.3) 0.7 (0.4–1.3) 0.8 (0.4–1.4)
80 or more 13 (17.1) 42 (17.5) 0.9 (0.4–1.9) 0.8 (0.3–2.0)

SELF-REPORTED OR (95% CI) adj for

HEALTH age, sex, household
clustering & self-
reported activity2

General Health
Very good/Good 58 (84.1) 200 (85.1) 1 1 1
Fair/bad 11 (15.9) 35 (14.9) 0.9 (0.7–1.4) 1.1 (0.7–1.6) 1.3 (0.9–2.0)
Limiting
longstanding illness
No 53(77.9) 176 (73.6) 1 1 1
Yes 15 (22.1) 65 (28.6) 1.4 (0.7–2.7) 1.6 (0.8–3.3) 2.9 (1.2–7.1)
Disability
No 40 (53.3) 125 (52.7) 1 1 1
Yes 35 (46.7 112 (47.3) 1.0 (0.6–1.7) 1.3 (0.7–2.5) 2.4 (1.1–5.1)
Chronic pain
No 46 (66.7) 146 (63.5) 1 1 1
Yes 23 (33.3) 84 (36.5) 1.2 (0.7–2.0) 1.4 (0.7–2.6) 2.1 (1.0–4.5)
Chronic disease
No 25 (32.9) 55 (22.9) 1 1 1
Yes 51 (67.1) 185 (77.1) 1.6 (0.9–2.9) 1.7 (0.9–3.0) 1.8 (1.0–3.2)
Use a walking aid
No 66 (89.2) 217 (91.9) 1 1 1
Yes 8 (10.8) 19 (8.1) 0.7 (0.3–1.7) 0.8 (0.3–2.1) 1.4 (0.5–3.8)
Fallen in last year
No 56 (76.7) 169 (71.9) 1 1 1
Yes 17 (23.3) 66 (28.1) 1.3 (0.7–2.4) 1.6 (0.8–2.9) 2.0 (1.0–4.0)
Current smoker
No 72 (96.0) 221 (94.0) 1 1 1
Yes 3 (4.0) 14 (6.0) 1.5 (0.4–5.4) 1.3 (0.3–5.1) 1.5 (0.4–5.1)
Body Mass Index
Normal weight 28 (46.7) 97 (43.1) 1 1 1
Overweight or obese 32 (53.3) 128 (56.9) 1.2 (0.7–2.0) 1.2 (0.7–2.4) 1.4 (0.7–2.6)
Geriatric
Depression Score3
<4 67 (88.2)) 213 (89.5) 1 1 1
4 or more 9 (11.8) 25 (10.5) 0.9 (0.4–2.0) 0.9 (0.4–2.0) 1.1 (0.5–2.5)

SELF-REPORTED OR (95% CI) adj for

ACTIVITY age, sex, household
clustering & self-
reported health4

Walking pace
compared to
others
Slower/much slower 16 (21.3) 22 (9.4) 1 1 1
About the same 34 (45.3) 100 (42.5) 2.1 (1.1–4.5) 1.9 (1.0–4.2) 4.6 (1.7–12.5)
Faster/much faster 25 (33.3) 113 (48.1) 3.3 (1.5–7.1) 3.2 (1.4–7.7) 10.4 (3.2–33.3)

Page 3 of 6
(page number not for citation purposes)
BMC Public Health 2008, 8:182 http://www.biomedcentral.com/1471-2458/8/182

Table 1: Comparison of non-participants & participants of physical activity study amongst questionnaire responders (Continued)
Hours walked in
last week?
None 24 (31.6) 37 (15.4) 1 1 1
Up to 2 hours 25 (32.9) 87 (36.3) 2.3 (1.1–4.5) 2.3 (1.1–4.7) 2.5 (1.1–5.6)
More than 2 hours 27 (35.5) 116 (48.3) 2.8 (1.4–5.4) 2.7 (1.3–5.5) 2.8 (1.3–6.1)
Average hours
gardening weekly
None 19 (25.3) 23 (10.0) 1 1 1
Up to 3.5 hrs/week 30 (40.0) 102 (44.4) 2.8 (1.4–5.8) 2.7 (1.3–5.6) 2.7 (1.1–6.9)
>3.5 hrs/week 26 (34.7) 105 (45.7) 3.3 (1.6–7.0) 2.8 (1.3–6.3) 3.3 (1.2–8.7)
Activity compared
to others
Less active or the 27 (35.5) 49 (20.6) 1 1 1
same
More active 33 (43.4) 116 (48.7) 1.9 (1.1–3.6) 1.9 (1.0–3.6) 2.2 (1.0–4.5)
Far more active 16 (21.1) 73 (30.7) 2.5 (1.2–5.1) 2.3 (1.1–5.2) 3.1 (1.3–7.0)
Do you cycle?
No 63 (87.5) 187 (78.9) 1 1 1
Yes 9 (12.5) 50 (21.1) 1.9 (0.9–4.0) 1.7 (0.7–4.1) 2.0 (0.8–5.1)
Do you walk a dog?
No 67 (88.2) 184 (78.6) 1 1 1
Yes 9 (11.8) 50 (21.4) 2.0 (0.9–4.3) 2.0 (0.8–5.2) 2.0 (0.7–5.2)
Do you do heavy
housework?
No 23 (34.3) 50 (22.7) 1 1 1
Yes 44 (65.7) 170 (77.3) 1.8 (1.0–3.2) 1.7 (1.0–3.2) 2.3 (1.2–4.5)
Positive attitudes
towards activity?
Low 35 (53.9) 73 (31.6) 1 1 1
High 30 (46.2) 158 (68.4) 2.5 (1.4–4.4) 2.5 (1.4–4.5) 3.4 (1.8–6.6)

1Household clustering – adjusted for the clustering by 237 households.

2Self-reported physical activity- adjusted for effect of walking pace and hours walked in last week
3Geriatric Depression Score 15 items, using cut-off <4/≥4, which gives a sensitivity of 91% and specificity of 72% for detecting major depression
[13].
4Self-reported health – adjusted for effect of limiting longstanding illness, chronic disease, disability, falls & pain.

After adjusting for age, sex and household clustering, par- Analyses restricted to participants (116) and non-partici-
ticipants tended to report more health problems than pants (76) who returned the postal questionnaire showed
non-participants for most variables, (but no differences very similar findings in terms of effect estimates to those
were statistically significant at p = 0.05). The exceptions presented, based on all questionnaire completers.
were: use of a walking aid which showed a non-significant Although the confidence intervals were wider due to
effect in the opposite direction; and depression score smaller numbers, several associations with participation
which was unrelated to participation. Adjusting for self- reached statistical significance at p = 0.05 for both health,
reported physical activity levels strengthened the associa- e.g. disability, adjusted OR 2.4 (05% C.I. 1.1–5.5) and
tions between poorer health and participating such that physical activity variables, e.g. higher activity levels than
those reporting limiting longstanding illness, disability, others, adjusted OR 3.1 (95% C.I. 1.1–8.7).
chronic pain, chronic disease and a fall in the last year
were more likely to participate. Discussion
Recruitment to our motion sensor activity study was 43%,
After adjusting for age, sex and household clustering, par- higher than for physical activity intervention studies in
ticipants reported more physical activity than non-partic- this age group [2-4], but lower than for surveys [1,2]. With
ipants for all measures and more positive attitudes this recruitment level, estimation of potential non-
towards physical activity. Apart from cycling and dog- response bias is important.
walking, where numbers were small, these differences
were all statistically significant and several showed dose- Our findings suggest two separate issues leading to poten-
response effects. Adjusting these estimates for self- tial bias. Firstly, participants were more physically active
reported health strengthened these effects still further. than non-participants (consistent with men being more

Page 4 of 6
(page number not for citation purposes)
BMC Public Health 2008, 8:182 http://www.biomedcentral.com/1471-2458/8/182

likely to participate). This has been reported in other Authors' contributions

physical activity studies in older people [3-5]. Secondly, TJH, CRV and DGC designed the study. TJH and RA col-
participants reported more health problems than non- lected data. TJH, IMC and DGC were involved in analysis
participants. This effect opposes the physical activity and interpretation of data. All authors were involved in
effect, and was much clearer after controlling for self- drafting the manuscript and revising it critically for impor-
reported physical activity. At first, these findings seem tant intellectual content. All authors have given final
counter-intuitive. Other studies in older people have approval of the version to be published.
reported that participants have better health than non-
participants [18-20], suggesting a "healthy volunteer" Additional material
effect, although Ives et al found participants were more
likely to have a disease history and use health services
more [21]. However, it seems plausible that recruiting
Additional file 1
supplementary questionnaire harris2008. Questionnaire on health and
older people to studies through primary care may lead to self-report physical activity levels.
those with increased contact with primary care (ie more Click here for file
illnesses) to be more likely to take part. This fits with other [http://www.biomedcentral.com/content/supplementary/1471-
work showing that older primary care patients with 2458-8-182-S1.doc]
poorer physical and psychological health and greater pri-
mary care service use were more likely to give consent for
their health records to be accessed for research [22].
Unfortunately, the other primary care based studies show- Acknowledgements
ing differences in physical activity levels between partici- We are grateful to all the partners, staff and patients of Sonning Common
Health Centre, Oxfordshire, UK for their support with this study. Funding
pants and non-participants did not report on health or
for the study was provided by the Thames Valley Primary Care Research
functional ability [3-5].
Partnership (WCRM03). The sponsors played no role in the design, execu-
tion, analysis & interpretation of data or in the writing of the manuscript or
An important weakness of our study was that self-reported the decision to submit the manuscript for publication.
health and physical activity details were only available on
those non-participants who completed the questionnaire. References
The similarities seen in age and sex comparisons between 1. Crombie IK, Irvine L, Williams B, et al.: Why older people do not
participants and non-participants overall and in those participate in leisure time physical activity: a survey of activ-
ity levels, beliefs and deterrents. Age Ageing 2004, 33:287-92.
responding to the questionnaire is reassuring and suggests 2. Stevens W, Hillsdon M, Thorogood M, McArdle D: Cost-effective-
that the non-participant questionnaire responders are rep- ness of a primary care based physical activity intervention in
45–74 year old men & women: a randomised controlled trial.
resentative of non-participants, at least in terms of age and Br J Sports Med 1998, 32:236-41.
sex, although they could still differ in other important 3. Munro JF, Nicholl JP, Brazier JE, Davey R, Cochrane T: Cost effec-
ways (such as activity level or health) that we lack infor- tiveness of a community based exercise programme in over
65 year olds: cluster randomised trial. J Epidemiol Community
mation on. The similarities found when restricting analy- Health 2004, 58:1004-10.
ses to only those sent postal questionnaires, confirms that 4. Halbert JA, Silagy CA, Finucane P, Withers RT, Hamdorf PA:
Recruitment of older adults for a randomized, controlled
it was reasonable to include participants in the analysis trial of exercise advice in a general practice setting. J Am Ger-
who completed the questionnaire at their baseline assess- iatr Soc 1999, 47:477-81.
ment. 5. Crombie IK, McMurdo ME, Irvine L, Williams B: Overcoming bar-
riers to recruitment in health research: concerns of poten-
tial participants need to be dealt with. BMJ 2006, 333:398.
Conclusion 6. Tai SS, Gould M, Iliffe S: Promoting healthy exercise among
Physical activity studies on older people recruited from older people in general practice: issues in designing and eval-
uating therapeutic interventions. British Journal of General Prac-
primary care settings may be biased by two opposing tice 1997, 47:119-22.
issues which need consideration when generalizing the 7. Ettinger WH Jr: Physical activity and older people: a walk a day
keeps the doctor away. J Am Geriatr Soc 1996, 44:207-8.
results: a strong bias towards participants being more 8. Harris TJ, Victor CR, Carey IM, Adams R, Cook DG: Optimising
physically active, and a weaker bias towards participants recruitment into a study of physical activity in older people:
having more health problems and therefore likely primary a randomised controlled trial of different approaches. Age &
Ageing in press. 2008
care contact. This latter bias could be used to advantage 9. Joint Health Surveys Unit: Health Survey for England 1998. In
when considering interventions to increase physical activ- Methodology and Documentation Volume 2. London, The Stationery
ity, where those who are least active and those with more Office; 1999.
10. McGee MA, Johnson A, Kay D: The Medical Research Council
physical health problems need targeting most. Cognitive Functioning and Ageing Study (MRC CFAS). The
description of activities of daily living in five centres in Eng-
land and Wales. Age & Ageing 1998, 27:605-13.
Competing interests 11. Ware JE, Sherbourne CD: The MOS 36-item short form health
The authors declare that they have no competing interests. survey: conceptual framework and item selection. Med Care
1992, 30:473-83.

Page 5 of 6
(page number not for citation purposes)
BMC Public Health 2008, 8:182 http://www.biomedcentral.com/1471-2458/8/182

12. Roberts RE, Kaplan GA, Shema SJ, Strawbridge WJ: Prevalence and
correlates of depression in an aging cohort: the Alameda
County Study. J Gerontol B Psychol Sci Soc Sci 1997, 52:S252-S258.
13. D'Ath P, Katona P, Mullan E, Evans S, Katona C: Screening, detec-
tion and management of depression in elderly primary care
attenders. I: The acceptability and performance of the 15
item Geriatric Depression Scale (GDS15) & the develop-
ment of short versions. Fam Pract 1994, 11:260-6.
14. Caspersen CJ, Bloemberg BP, Saris WH, Merritt RK, Kromhout D:
The prevalence of selected physical activities and their rela-
tion with coronary heart disease risk factors in elderly men:
the Zutphen Study, 1985. Am J Epidemiol 1991, 133:1078-92.
15. Washburn RA, Smith KW, Jette AM, Janney CA: The Physical
Activity Scale for the Elderly (PASE): development and eval-
uation. J Clin Epidemiol 1993, 46:153-62.
16. Jette AM, Rooks D, Lachman M, Lin TH, Levenson C, Heislein D, et
al.: Home-based resistance training: predictors of participa-
tion and adherence. Gerontologist 1998, 38:412-21.
17. STATA 9.0 Statistics/Data Analysis. College Station, Texas,
USA, Statacorp; 2005.
18. van Heuvelen MJ, Hochstenbach JB, Brouwer WH, de Greef MH, Zijl-
stra GA, van Jaarsveld E, et al.: Differences between participants
and non-participants in an RCT on physical activity and psy-
chological interventions for older persons. Aging Clin Exp Res
2005, 17:236-45.
19. Hebert R, Bravo G, Korner-Bitensky N, Voyer L: Refusal and infor-
mation bias associated with postal questionnaires and face-
to-face interviews in very elderly subjects. J Clin Epidemiol 1996,
49:373-81.
20. Wagner EH, Grothaus LC, Hecht JA, LaCroix AZ: Factors associ-
ated with participation in a senior health promotion pro-
gram. Gerontologist 1991, 31:598-602.
21. Ives DG, Traven ND, Kuller LH, Schulz R: Selection bias and non-
response to health promotion in older adults. Epidemiology
1994, 5:456-61.
22. Harris T, Cook DG, Victor C, Beighton C, DeWilde S, Carey I: Link-
ing questionnaires to primary care records: factors affecting
consent in older people. J Epidemiol Community Health 2005,
59:336-8.

Pre-publication history
The pre-publication history for this paper can be accessed
here:

http://www.biomedcentral.com/1471-2458/8/182/pre
pub

Publish with Bio Med Central and every

scientist can read your work free of charge
"BioMed Central will be the most significant development for
disseminating the results of biomedical researc h in our lifetime."
Sir Paul Nurse, Cancer Research UK

Your research papers will be:

available free of charge to the entire biomedical community
peer reviewed and published immediately upon acceptance
cited in PubMed and archived on PubMed Central
yours — you keep the copyright

Submit your manuscript here: BioMedcentral

http://www.biomedcentral.com/info/publishing_adv.asp

Page 6 of 6
(page number not for citation purposes)