Female urinary incontinence (UI) is a condition with severe economic and psychosocial impact. I have studied the epidemiology of UI in the general female population and the prevailing management of UI by Norwegian general practitioners (GPs). Those who are interested in more details, may take a look at the synopsis below. The main conclusions that may be drawn from these studies are that UI is very prevalent among adult women, mostly in the form of stress incontinence. At least 6% may be regarded as potential patients. Although complicated by methodological difficulties, studies of GPs' management indicate that available therapeutic tools are not used to their full potential. A literature review demonstrates that the history is the most important diagnostic tool in general practice, and that the value of clinical investigations is poorly documented.The general purpose of the present investigations was to describe the epidemiology of UI in the adult female population, and to describe the management offered these patients by Norwegian GPs. Secondly, the validity of some of the employed methods was investigated. Studies were undertaken with the following specific aims:
The study described in Paper I was motivated by the extreme variation found between earlier studies on the prevalence of female urinary incontinence. Our working hypothesis was that the differences could be ascribed to the use of different severity criteria or "thresholds" for labelling the respondents as incontinent. Therefore, we intended to describe the general prevalence of female UI by using validated, graded criteria of severity.
The study comprises data gathered at the outpatient clinic of the Department of Obstetrics and Gynaecology, Trondheim University Hospital and in the rural community of Rissa. Based on data from 116 women referred to the clinic an index score describing the severity of incontinence was validated against a 48-hour pad weighing test. The score was further categorized into slight (5% of the patients), moderate (27%), and severe (68%). Typically, slight incontinence denotes leakage of drops a few times a month, moderate incontinence daily leakage of drops, and severe incontinence larger amounts at least once a week.
Thereafter, an anonymous postal survey was performed among all adult women in Rissa (including those living in institutions). Of the 1 820 women who were included (response rate 77%), 535 (29.4%) reported urinary incontinence of any severity. Highest prevalence was found around the menopause and in old age (80+). In this unselected population 46% were classified as having slight, 27% moderate, and 27% severe UI. There was a shift towards increasing severity with advancing age (Figure).
The psychosocial impact of incontinence, measured by a single question, closely parallelled the severity of leakage. It is concluded that many adult women may experience some degree of urinary leakage from time to time, but most of them are not bothered by it, and they should not be labelled potential patients. Based on recorded severity and impact, it is suggested that approximately 6% of all adult women (96 000 Norwegian women) are potential patients because of incontinence.
In Paper II we tried to bridge the gap between symptoms and diagnoses by correcting the epidemiological survey for lack of validity in this respect. Data from the same settings as in Paper I were used. We included 250 incontinent women at the clinic and all 535 women who reported incontinence in Rissa. At the clinic diagnostic questions about stress and urge incontinence were validated against a final diagnosis made by the gynaecologist after urodynamic evaluation. The sensitivity for stress incontinence was 0.66, specificity 0.88. The corresponding values for urge incontinence were 0.56 and 0.96, and for mixed incontinence 0.84 and 0.66. Stress incontinence was only diagnosed in the absence of urge and vice versa.
Similar diagnostic questions were later used in the epidemiological survey in Rissa, and the indices of validity (sensitivity and specificity) were used as corrective measures for the diagnostic distribution found in the epidemiological survey. The percentage of stress incontinence thereby increased from 51% to 77%, while mixed incontinence was reduced from 39% to 11%. Pure urge incontinence remained stable at 10% - 12%.
There are objections to the method employed here. The crucial question is whether it is justified to claim that sensitivity and specificity will be the same in the clinic and in the community. By comparing the indices of validity in two different groups of severity at the clinic, we were able to substantiate our claim, but further investigations should be done into this field of experimental clinical epidemiology.
The problem of response bias is the subject of Paper III. Known differences between responders and non-responders may be compensated during analysis, but there is also a more severe problem of unknown response bias, such as the possibility of different response rates between continent and incontinent women. In most epidemiological surveys only fragmentary bits of information are known about the non-responders.
This study describes the response bias found when recruiting incontinent women for an interview survey. We identified a total of 3 091 women who had received prescriptions for incontinence aids or drugs. Information gathered at the local Insurance Offices in Bergen, Os and Fjell enabled us to calculate total pad use for each woman.
Of 401 incontinent women randomly selected, 202 (50%) responded, and 153 were willing to participate in an interview survey. This response rate is lower than what is reported in epidemiological surveys in the general population. The willingness to participate was markedly higher among middle-aged women than among the older and younger ones. Furthermore, while middle-aged participants were perfectly representative with regard to pad use, the young and old participants were biased towards lower pad use (probably reflecting incontinence of lesser severity).
Due to the low response rate in this selected population, it is concluded that the prevalence of severe incontinence may be underestimated in epidemiological surveys. Furthermore, the prevalence peak in mid-life often found in such surveys, may be due to response bias related to age and severity of incontinence.
In the study published as Paper IV we intended to describe the management of female UI by GPs in western Norway. The method chosen was vignettes, questionnaires containing six typical case histories. We wanted to study the doctors' performance in day-to-day practice, not the competence at their theoretical best, and this was stressed in the introductory letter. Of 191 randomly selected GPs 139 (73%) responded.
Gynaecological examination and microscopy of the urine were the most frequently investigations stated (76% and 84%). A leakage provocation test was performed in 22%, neurological examination in 15%, frequency-volume chart in 15%, and measurement of residual urine in 8% of the cases. The GPs were less active in the investigation of older women than the younger ones.
Young women were more often instructed in pelvic floor exercises than old women (83% v. 26%), while the opposite tendency was found for prescribing pads (12% v. 80%). Anticholinergic drugs were used for 38% of women with urge/mixed incontinence, oestrogen for 25% of the postmenopausal women. Fourteen per cent were instructed in bladder training, 25% were referred. Adequate treatment was defined as relevant drugs and exercises in combination with pads, or referral. According to this definition 46% of the patients were treated adequately, 28% by the GPs themselves.
A discussion of the validity of the vignette method is included in this paper. One should be cautious in drawing conclusions with regard to daily practice. However, it is argued that although a bias towards over-investigation and over-treatment may have occurred, inter-group comparisons may still be valid if the bias is unidirectional.
Paper V is based on the same material as Paper IV. We attempted to explain the great variation in management strategies found among the physicians. The variables investigated were physician's sex, years since graduation, GP specialisation, and location of practice. The variation within the subgroups was greater than that between groups, and only minor differences were related to the variables examined. However, GP specialists treated fewer patients adequately than non-specialists. Experienced female doctors were most active in teaching pelvic floor exercises, prescribing oestrogens, and in referring patients to specialist care. Similar findings have later been reported by others. Women are more likely to undergo gynaecological procedures if they see female rather than male physicians.
In the study described in Paper VI we expanded on the subject of management by using additional methods for measuring performance. Taken together the Papers IV & VI thus represent a triangulation of methods: vignettes, patient interviews, and review of medical records. Paper VI is based on data from the incontinent women described in Paper III. Of the 153 incontinent women willing to participate, 150 were interviewed, 118 named a GP whom they considered to be their personal doctor, and 104 gave us permission to collect supplementary information from their medical records. Eventually, complete data were obtained for 82 patients, mainly focusing on what examinations and treatments they had received. We also measured the strength of agreement between the two data sources (interviews and medical records) by Kappa statistics.
The GPs had done a gynaecological examination in 54% and a urinalysis in 73% of the patients. They had performed a provocation test in 12%, neurological examination in 10%, frequency-volume chart in 1%, and measured residual urine in 5%. Thirty-two per cent had been taught pelvic floor exercises and 13% bladder training. Forty-eight per cent of the older women had received oestrogens, and 29% of those with urge/mixed incontinence had received anticholinergics. Overall Kappa between the two data sources was a weak 0.37.
With the exception of prescribing oestrogen, it seems that the use of vignettes (Paper IV) overestimates the performance of physicians. However, the patients and the GPs were different in Papers IV & VI, and the time lag between the studies (five years) may explain different management strategies. Nevertheless, it is concluded that GPs' incontinence management can be improved. Taken separately, retrospective chart data and patient interviews are unreliable data sources. A combination is recommended to increase validity.
Paper VII describes a study that was nested inside the previous one (Paper VI). In order to examine the criterion validity of the vignette method used in Papers IV & V a new vignette study was interposed between the interviews and the collecting of information from the medical records (see Paper VI). The methodological problem of different case-mix between vignettes and real life was eliminated by tailoring the vignettes according to the information given by the interviewed patients. The participating GPs were unaware that one of the vignettes resembled a real patient of theirs. It was stressed that they were to answer according to daily practice, not as they would do in a competency exam. A total of 32 GPs and 32 patients were included.
The GPs claimed more actions with the "paper patient" than they had carried out with the real one. However, this tendency towards overreporting was only evident when the GPs chose their actions from an option list (cueing effect). An open-ended question about drug treatment did not entail such overreporting. The individual fit of actions performed with vignettes and real patients was measured by likelihood ratios and Kappa statistics. The fit was somewhat better with the open-ended question than with the cued questions, but caution should be exercised when interpreting this result. Possible confounding could have arisen from the fact that non-cued and cued questions differed in content.
The activity levels with vignettes and real patients correlated significantly (R=0.65, p<0.001). This finding supports the claim put forward in Paper IV that although a bias towards over-investigation and over-treatment may have occurred, inter-vignette comparisons may still be valid.
I Sandvik H, Hunskaar S, Seim A, Hermstad R, Vanvik A, Bratt H. Validation of a severity index in female urinary incontinence and its implementation in an epidemiological survey. J Epidemiol Community Health 1993; 47: 497-9.
II Sandvik H, Hunskaar S, Vanvik A, Bratt H, Seim A, Hermstad R. Diagnostic classification of female urinary incontinence. An epidemiological survey corrected for validity. J Clin Epidemiol 1995; 48: 339-43.
III Sandvik H, Hunskaar S. Incontinence in women: different response rates may introduce bias in community studies of pad consumption. J Epidemiol Community Health 1994; 48: 419.
IV Sandvik H, Hunskaar S, Eriksen BC. Management of urinary incontinence in women in general practice: actions taken at the first consultation. Scand J Prim Health Care 1990; 8: 3-8.
V Sandvik H, Hunskaar S. Doctors' characteristics and practice patterns in general practice: an analysis based on management of urinary incontinence. Scand J Prim Health Care 1990; 8: 179-82.
VI Sandvik H, Hunskaar S. General practitioners' management of female urinary incontinence. Medical records do not reflect patients' recall. Scand J Prim Health Care 1995; 13: 168-74.
VII Sandvik H. Criterion validity of responses to patient vignettes. An analysis based on management of female urinary incontinence. Fam Med 1995; 27: 388-92.
Home Page Department
Home Page
University
Department of Public Health and Primary Health Care, last updated 17.04.97
Hogne.Sandvik@isf.uib.no