• nonresponse bias;
  • urinary incontinence;
  • postal survey


  1. Top of page
  2. Abstract


To investigate nonresponse bias in a postal survey on urinary symptoms in people aged ≥ 40 years.


Nonresponders to a postal survey on incontinence and other urinary symptoms were studied. A random sample of 1050 nonresponders (stratified for age and sex) was traced by a team of interviewers. Eligible nonresponders were asked several questions from the postal questionnaire, and their reason for not participating in the postal survey.


Only 1% of those not responding were not traced in person or accounted for, and 12% were identified as not eligible to be in the survey sample (moved from address, deceased, residential home). Half of the eligible nonresponders (51%) did not answer the interviewer's questions, the main reason being general unwillingness or disinterest. The number in whom poor health was the reason increased with age. Comparing nonresponders who answered the interviewer's questions with a random sample of responders from the postal survey showed little difference in the reporting of urinary symptoms, although there were differences in general health and long-term health problems. Separate analyses by age showed greater reporting of some urinary symptoms and of poorer general health in the older nonresponders (≥ 70 years).


Overall, for people aged ≥ 40 years there was no evidence of a nonresponse bias in the reporting of urinary symptoms, providing confidence in such prevalence rates. However, poorer general health and greater reporting of some urinary symptoms by the older nonresponders (≥ 70 years) suggests prevalence rates in this age group may be underestimated.


  1. Top of page
  2. Abstract

A systematic review of prevalence studies on urinary incontinence and storage symptoms in the UK since 1960 identified 25 publications from 21 studies [1]; response rates were 53–98%. Analysis of response bias was limited and inconsistent, with five studies reporting an under-representation by elderly women and young groups, and five reporting no difference in relation to age and sex. Increasing awareness of data protection issues is likely to reduce response rates further, making insights into nonresponse bias increasingly important.

Nonresponse bias may be a particular problem in surveys on sensitive subject areas such as urinary incontinence. The potentially embarrassing nature of the questions may discourage participation by some people, although the anonymity of the postal questionnaire may encourage participation by others. Questionnaires which contain sensitive questions are less likely to be returned [2], but the evidence that nonresponse bias occurs is conflicting [3,4].

The question of nonresponse bias has been investigated in depth in the context of general health and several specific diseases and conditions, but little has been done in the area of incontinence and urinary symptoms. The pattern of symptom reporting over time taken to return a questionnaire [5] and differences between subsequent mailings [6] have been used as indicators of nonresponse bias. Another approach is to follow-up a sample of nonresponders until they are all traced, and then assess whether and how they differ from the responders [7]. The aim of the present study was to identify the nature and extent of nonresponse bias in a postal survey on urinary symptoms, by following a sample of nonresponders.


  1. Top of page
  2. Abstract


The postal survey was part of the Leicestershire MRC Incontinence Study on the prevalence and incidence of incontinence and other LUTS [8]. The sampling frame was men and women aged ≥ 40 years on the Leicestershire Health Authority's lists of 108 general practices (71% of the practices in Leicestershire and Rutland). People living in residential or nursing homes were excluded. Two random samples were selected by household (39 603 and 20 012) and the larger sample subsequently formed a cohort for longitudinal follow-up to investigate the natural history and aetiology of symptoms. The two samples were mailed similar questionnaires between October 1998 and July 1999, which included questions on demographic characteristics, health variables, physical disabilities, lifestyle, and a series of questions on urinary and bowel symptoms developed for the study [8]. Two reminder questionnaires were sent at monthly intervals, making three mailings in all. As the time taken to return a postal questionnaire may be an indicator of nonresponse bias, the ‘early’ responders (replied to the first two mailings) were compared with the ‘late’ responders (replied to the third mailing).

A random sample of 1050 was selected from those nonresponders on whom no outcome information was obtained within 4 weeks of the third mailing of the questionnaire. The sample was stratified by age and sex with 105 in each of following age groups (years) for men and women: 40–49, 50–59, 60–69, 70–79 and ≥ 80 years. Stratification ensured sufficient numbers in the top age groups, where there were fewer subjects and the risk of bias occurring greater because of poorer response rates.


The sample was traced by a team of trained and experienced interviewers who were working on the research project. The interviewers made every possible attempt to contact in person or account for the nonresponder at the address to which the postal questionnaire had been sent. This was achieved by making telephone calls, home visits and talking to local residents and neighbours. Tracing was not abandoned until at least three home visits had been made at different times of the day, and days of the week. After tracing or accounting for the nonresponder, the interviewer determined their eligibility to be in the study sample (at the time of mailing they lived at the address and this was not a residential home) and then attempted to ask the eligible nonresponders 10 questions from the postal questionnaire. These were questions on different urinary symptoms, the size of any urinary problem experienced and whether they had received treatment for urinary symptoms in the previous year. They were also asked to rate their current general health in one of five categories (excellent, very good, good, fair or poor) and whether they had a long-term health problem which limited their daily activity or work [9]. Finally they were asked why they had not previously returned the postal questionnaire.


The data were analysed in three stages. First, responders and all eligible nonresponders in the postal survey (Fig. 1) were compared by age (Mann–Whitney U-test) and sex (χ2 test), demographic variables available on the whole sample. Ethnic origin was also compared (χ2 test), because of the many South Asians in the study population (5.3% of adults aged ≥ 40 years living in Leicestershire are of South Asian origin [10]). The Nam Pehchan computer software [11] was used to code the surnames of the survey sample as ‘South Asian’ or ‘not South Asian’ (referred to in this paper as all other ethnic groups).


Figure 1. The flow of the sample through the postal survey and nonresponder study.

Download figure to PowerPoint

Second, the nonresponders who were traced and answered the interviewer's questions were compared with a random sample of responders to the postal survey, a group matched for age and sex on a 1 : 4 ratio. Conditional logistic regression analyses were used to take account of the matching. Responses to the urinary symptom questions were categorical and given as frequency or severity of symptom. The responses to these questions were dichotomised at the threshold used to indicate presence of a clinically significant symptom [8].

Third, logistic regression analyses, adjusting for age and sex, were used to compare the urinary symptom questions and health variables of the early and late responders to the postal questionnaire.


  1. Top of page
  2. Abstract


Of the 59 615 people who were mailed questionnaires, 4088 (6.9%) were not eligible at the time of mailing (3779 moved from the address, 195 had died and 114 were admitted to residential care; Fig. 1). Of those who appeared eligible (55 527), there were 35 131 responders and 20 396 nonresponders (158 unable to complete the questionnaire, 2897 refusals and 17 341 on whom no information was obtained after three mailings). The overall response rate (responders as a percentage of the apparently eligible sample) was 63.5%. Responders were older than nonresponders (median age 58 years vs 54 years; P < 0.001) and women were more likely (P < 0.001) to participate than men (response rates of 66% vs 61%). There were significantly fewer South Asians in the responders (P < 0.001) because of their low response rate (39% compared with 66% in all other ethnic groups).


Tracing was very successful and only abandoned in 14 cases (1%); 128 (12.2%) proved to be ineligible for the postal survey at the time of mailing (101 not at the address, 22 died and five in residential care). As a result, the true response rate was higher than initially suggested. The ineligibility rates were higher in men and increased with age, although the rate in the youngest men (40–49 years) was also high (Table 1). The remaining 908 nonresponders were identified as eligible to be in the postal survey and 448 (49%) answered the interviewer's questions.

Table 1.  The outcome of tracing the sample of 1050 nonresponders
Age (years)SampleNot tracedNot eligibleInterviewer’s questions answeredInterviewer's questions not answered
Health problemUnwilling/ too busyOther*
  • *

    recently moved, or away for extended period (38), English not first language (28), other communication problem (seven), domestic problem (12).

40–49  105  2  19  51  2  24  7
50–59  105  3  10  53  1  33  5
60–69  105  1  13  39  4  3810
70–79  105  0  14  45  6  2911
≥  80  105  1  21  3321  20  9
Total men  525  7  772213414442
40–49  105  3    9  52  1  35  5
50–59  105  3    2  51  2  38  9
60–69  105  0    9  46  5  3213
70–79  105  1    9  4111  35  8
≥  80  105  0  22  3722  16  8
Total women  525  7  512274115643
Total sample1050141284487530085


In all, 460 (51%) of the nonresponders traced did not answer the interviewer's questions, the main reason being they were too busy or unwilling (300). In 75 the reason was associated with poor health (35 current ill-health, 12 long-term admission to hospital or residential care and 28 died since mailing the questionnaire). The number unable to speak to the interviewer because of a health problem increased markedly with age (Table 1) and the total number was high because of the stratification of the sample and resultant high number of older people. The nonresponders who answered the interviewer's questions were younger than those who did not (median ages 61 and 65 years, respectively; P < 0.001), but there was no significant difference in gender between the groups (P = 0.65). A higher interview rate in the South Asians than in all other ethnic groups (57% vs 48%) resulted in a difference in the ethnic composition of the two groups (P = 0.05).


Overall, the conditional logistic regression analyses (Table 2) showed a slight tendency for increased reporting of urinary symptoms by the nonresponders, but this was only significant for one symptom, with nonresponders more likely to report a high frequency of micturition (Table 2). There was significantly less reporting of a long-term health problem by nonresponders but a difference in the opposite direction for current general health (poorer in the nonresponders) approached significance (Table 2).

Table 2.  The urinary symptoms and health variables relating to nonresponse [448 nonresponders (NR) vs 1792 responders (R)] and late response in the postal survey (2018 ‘late’ vs 33 113 ‘early’; adjusted for age and sex)
VariableOR (95% CI)
NR vs RLate vs early
Any urinary leakage
Several times a month or more vs less often1.11 (0.84–1.48)0.93 (0.82–1.07)
Stress leakage
Several times a month or more vs less often1.33 (0.95–1.87)1.08 (0.93–1.26)
Urge leakage
Several times a month or more vs less often1.19 (0.88–1.61)1.05 (0.91–1.22)
Frequency of micturition
Hourly or more vs less often1.44 (1.03–2.02)0.94 (0.78–1.12)
3 times a night or more vs less often1.14 (0.81–1.62)0.87 (0.71–1.06)
Frequency of very strong urgency
Several times a month or more vs less often1.09 (0.81–1.42)0.84 (0.67–1.04)
Received treatment in last year
Yes vs no0.88 (0.57–1.36)0.87 (0.71–1.07)
Problem with urinary symptoms
Yes vs no1.23 (0.84–1.80)1.00 (0.83–1.21)
Perceived health
Fair or poor vs good/very good/excellent1.24 (0.98–1.57)1.15 (1.03–1.28)
Long-term health problem
Yes vs no0.75 (0.59–0.96)0.99 (0.89–1.10)

The analysis was repeated separately for those aged < 70 years and ≥ 70 years. In the younger group there were no differences in any of the urinary symptom questions or the two health measures. However, in the older group nonresponders reported more symptoms of urge leakage (OR 1.59, 95% CI 1.04–2.43), urgency (OR 1.53, 95% CI 1.00–2.34) and high frequency of micturition (OR 2.27, 95% CI 1.4–3.67). The older nonresponders also reported fewer long-term health problems (OR 0.69, 95% CI 0.48–0.99), but were more likely to describe their current general health as poor or fair (OR 1.53, 95% CI 1.07–2.20).

The interviewers asked the nonresponders why they had not previously returned the postal questionnaire, and the main reason given was they were generally disinterested, unwilling or too busy (370, 83.3%). Only 27 (6.1%) gave a reason associated with poor health or disability and 21 of these were aged ≥ 70 years.


Women were less likely to be late responders (OR 0.88, 95% CI 0.81–0.96) and the probability of late response decreased with age (OR 0.99, 95% CI 0.98–0.99). South Asians were more likely to be late responders than the white ethnic group (OR 1.92, 95% CI 1.66–2.22) but there was no significant difference between ‘other’ and white ethnic groups (OR 1.01, 95% CI 0.75–1.38).

Comparing the six urinary symptoms and the questions on treatment and extent of any urinary problem showed no significant differences between early and late responders (Table 2). Late responders were more likely to report poor or fair general health but there was no significant difference in reporting of a long-term health problem (Table 2).

When the analysis was repeated separately for the two age groups, the only differences were in the older group (≥ 70 years) where the late responders were less likely to have received treatment for urinary symptoms in the previous year (OR 0.64, 95% CI 0.42–0.99) and were more likely to report poor or fair general health (OR 1.31, 95% CI 1.04–1.64).


  1. Top of page
  2. Abstract

Nonresponders are by definition difficult to trace and interview, and in studies which have used this approach [7,12] many have always remained untraced. In the present large study on 1050 nonresponders, the interviewers were very successful in tracing them and only 1% remained unaccounted for.

At around the time of mailing 6.9% of the survey sample were identified as ineligible. After tracing, 12.2% of the nonresponder sample were identified as ineligible, mainly because they had moved. The high rate of ineligibility in the youngest men (40–49 years) may have been a result of less frequent visits to the GP [9], which is when general practices are usually informed of address changes. The high ineligibility rates in the very elderly men and women (≥ 80 years) are the result of the high morbidity, mortality and rates of admission to residential care in this group. If the sample details of this age group are adjusted for the levels of ineligibility found in the nonresponders, the response rates to the postal questionnaire increase from 63.5% to 67.5% in men, and from 58.5% to 62.5% in women.

Although almost the whole sample was successfully traced, only half (49%) of the eligible nonresponders answered the interviewer's questions, which is not unexpected as they had previously not responded to three mailings of the questionnaire. Overall, the main reason for not speaking to the interviewer was a general unwillingness or lack of interest, reasons commonly given for not participating in surveys [13]. Very few had a hostile attitude to the survey. However, in many the reason was associated in some way with poor health, and the percentage increased markedly with age. The nonresponders who spoke to the interviewer were asked why they had not previously returned the postal questionnaire, and the pattern of reasons given was similar, with the main one being general unwillingness or ‘never got round to it’. Poor health as a reason increased with age. This is consistent with reports of high nonresponse rates caused by illness or frailty in surveys in the very elderly [14].

Between responders and nonresponders overall there was little difference in the reporting of urinary symptoms, and this was further supported by comparing the early and late responders. However, there was greater reporting of some urinary symptoms by the elderly nonresponders and differences in the two measures of general health. Current poor health was more common among elderly nonresponders, but a long-term health problem was more common among responders.

Reports on nonresponse have shown poorer health in responders [15] and nonresponders [16], but many studies have shown small or no differences depending on who and what is being investigated [17–20]. Two studies on prostatism used follow-up questionnaires to collect information on samples of those not participating [21,22]. Both studies reported fewer urinary symptoms in the non-participants, which had sizeable effects on the adjusted prevalence rates. However, the studies compared men who had agreed or not agreed to participate in studies which involved substantial commitment and intrusive clinical procedures, so it is not surprising that men with symptoms were more interested and willing to take part. Moller et al.[23] reported less incontinence and strong urgency when 529 late responders to a postal questionnaire were compared with 2860 early responders, while a smaller study [24] showed no difference in reporting of incontinence in 38 nonresponders who returned the third mailing of a questionnaire. The evidence that withdrawal rates in longitudinal studies are greater in people with no urinary symptoms is conflicting [25–27], and where there were differences these were not large.

The present study compared data from a self-completed postal questionnaire with data collected by an interviewer. In addition, the nonresponders answered the questions later than the responders, and there may have been onset or progression of symptoms. There is reported evidence about the mode of administration of questionnaires on urinary symptoms, with two studies [28,29] showing less reporting of urinary symptoms in a telephone interview than in an earlier self-completed postal questionnaire. A third study, investigating a few ‘visually impaired or illiterate men’, found no difference [30]. The chances of an effect of mode of administration are probably greater in the elderly group. Although instructed not to, the interviewers may have helped some of the very elderly and frail nonresponders to answer the questions.

Concern about the possible sensitive nature of the questionnaire content proved unfounded. The actual refusal rate to the postal questionnaire was not large, and when a reason was given for not taking part, it was usually a general unwillingness to take part in surveys. Only very few complained about the actual contents of the questionnaire. The general acceptance of the questionnaire is indicated by the response rate of 63.3%. Response rates to postal surveys depend among other things upon the nature of the sampling frame. The present large county-wide sample was selected from the lists of over 400 GPs, and the response rate compares well with other large postal surveys on health which have been conducted recently in this country using the same sampling procedure [31].

In conclusion, we found little evidence of nonresponse bias in the reporting of urinary symptoms in the overall sample, which gives confidence in the population prevalence rates for urinary symptoms that we calculated for people aged ≥ 40 years. Although the prevalence rates among the very elderly are possibly underestimated, the proportion of very elderly people within the population is very small.


  1. Top of page
  2. Abstract

The authors acknowledge the interviewers and all the members of the Leicestershire MRC Incontinence Study team. The project was funded by the Medical Research Council.


  1. Top of page
  2. Abstract
  • 1
    McGrother CM, Donaldson MMK, Wagg A et al. Continence. In: StevensA, RafteryJ, MantJ (eds) Health Care Needs Assessment: the epidemiologically based needs assessment review. Abingdon, UK: Radcliffe Medical Press Ltd, 2003; in press
  • 2
    Edwards P, Roberts I, Clarke M et al. Increasing response rates to postal questionnaires: systematic review. Br Med J 2002; 324: 11837
  • 3
    Barker P, Cooper RF. Do sexual health questions alter the public's response to lifestyle questionnaires? J Epidemiol Community Health 1996; 50: 688
  • 4
    Peters TJ, Harvey IM, Bachmann MO, Eachus JI. Does requesting sensitive information on postal questionnaires have an impact on response rates? A randomised controlled trial in the South West of England. J Epidemiol Community Health 1998; 52: 130
  • 5
    Tennant A, Badley EM. Investigating non-response bias in a survey of disablement in the community – implications for survey methodology. J Epidemiol Community Health 1991; 45: 24750
  • 6
    Paganini-Hill A, Hsu G, Chao A, Ross RK. Comparison of early and late respondents to a postal health survey questionnaire. Epidemiology 1993; 4: 3759
  • 7
    Hill A, Roberts J, Ewings P, Gunnell D. Non-response bias in a lifestyle survey. J Public Health Med 1997; 19: 2037
  • 8
    Perry S, Shaw C, Assassa P et al. An epidemiological study to establish the prevalence of urinary symptoms and felt need in the community: the Leicestershire MRC Incontinence Study. J Public Health Med 2000; 22: 42734
  • 9
    Bridgwood A, Lilly R, Thomas M, Bacon J, Sykes W, Morris S. Results from the 1998 General Household Survey. Office for National Statistics. London: The Stationery Office, 2000
  • 10
    Office of Population Censuses and Surveys. 1991 Census. County Monitor, Leicestershire 1992
  • 11
    Cummins C, Winter H, Cheng K, Maric R, Silcocks P, Varghese C. An assessment of the Nam Pehchan Computer Program for the identification of names of South Asian ethnic origin. J Public Health Med 1999; 21: 4016
  • 12
    Hoeymans N, Feskens E, Van Den Bos GAM, Kromhout D. Non-response bias in a study of cardiovascular diseases, functional status and self-rated health among elderly men. Age Ageing 1998; 27: 3540
  • 13
    Kotaniemi JT, Hassi J, Kataja M et al. Does non-responder bias have a significant effect on the results in a postal questionnaire study? Eur J Epidemiol 2001; 17: 80917
  • 14
    Hebert R, Bravo G, Kornerbitensky N, Voyer L. Refusal and information bias associated with postal questionnaires and face-to-face interviews in very elderly subjects. J Clin Epidemiol 1996; 49: 37381
  • 15
    Melton LJ, Dyck PJ, Karnes JL, O'Brien PC, Service FJ. Non-response bias in studies of diabetic complications: The Rochester Diabetic Neuropathy Clinic. J Clin Epidemiol 1993; 46: 3418
  • 16
    Beard CM, Lane AW, O'Fallan WM, Riggs BL, Melton LJ. Comparison of respondents and non-respondents in an osteoporosis study. Ann Epidemiol 1994; 4: 398 403
  • 17
    Jackson R, Chambless LE, Yang K et al. Differences between respondents and non-respondents in a multicenter community-based study vary by gender and ethnicity. J Clin Epidemiol 1996; 49: 14416
  • 18
    Wu AW, Jacobson DL, Berzon RA et al. The effect of mode of administration on medical outcomes study health ratings and EuroQol scores in Aids. Qual Life Res 1997; 6: 310
  • 19
    Perkins J, Sanson-Fisher RW. An examination of self and telephone administered modes of administration for the Australian SF-36. J Clin Epidemiol 1998; 51: 96973
  • 20
    Lyons R, Wareham K, Lucas M, Price D, Williams J, Hutchings HA. SF-36 scores vary by method of administration: implications for study design. J Public Health Med 1999; 21: 415
  • 21
    Panser L, Chute CGH, Larsonkeller JGC, Oesterling J, Lieber M, Jacobsen SJ. The natural-history of prostatism – the effects of non-response bias. Int J Epidemiol 1994; 23: 1198205
  • 22
    Blanker MH, Groeneveld FPA, Bernsen RMD, Bohnen AM, Bosch J. Strong effects of definition and nonresponse bias on prevalence rates of clinical benign prostatic hyperplasia: the Krimpen Study of male urogenital tract problems and general health status. BJU Int 2000; 85: 66571
  • 23
    Moller LA, Lose G, Jorgensen T. The prevalence and bothersomeness of lower urinary tract symptoms in women 40–60 years of age. Acta Obstet Gynecol Scand 2000; 79: 298305
  • 24
    Samuelsson E, Victor A, Tibblin G. A population study of urinary incontinence and nocturia among women aged 20–59 years – prevalence, well-being and wish for treatment. Acta Obstet Gynecol Scand 1997; 76: 7480
  • 25
    Brown M, Lin D, Normolle D, Herzog AR. Models for and the effect of non-response in a survey of the non-institutionalized elderly. Proc Survey Res Section, Am Statist Assoc 1988; 38: 61520
  • 26
    Moller LA, Lose G, Jorgensen T. Incidence and remission rates of lower urinary tract symptoms at one year in women aged 40–60: longitudinal study. BMJ 2000; 320: 142932
  • 27
    Lee AJ, Garraway WM, Simpson RJ, Fisher W, King D. The natural history of untreated lower urinary tract symptoms in middle-aged and elderly men over a period of five years. Eur Urol 1998; 34: 32532
  • 28
    Rhodes T, Girman CJ, Jacobsen SJ et al. Does the mode of questionnaire administration affect the reporting of urinary symptoms? Urology 1995; 46: 3415
  • 29
    Fowler FJ. Mode Effects in a Survey of Medicare Prostate Surgery Patients. American Statistical Association. Proceedings, Survey Methods Section. 1993: 7305
  • 30
    Barry MJ, Fowler FJ, Chang YC, Liss CL, Wilson H, Stek M. The American Urological Association Symptom Index. Does mode of administration affect its psychometric properties? J Urol 1995; 154: 10569
  • 31
    Harrison J, Barrow S, Gask L, Creed F. Social determinants of GHQ score by postal survey. J Public Health Med 1999; 21: 2838

odds ratio.