Cervical cancer screening

Cervical cancer screening

Public Health (1999) 113, 111±115 ß R.I.P.H.H. 1999 http://www.stockton-press.co.uk/ph Cervical cancer screening: spatial associations of outcome and...

128KB Sizes 15 Downloads 153 Views

Public Health (1999) 113, 111±115 ß R.I.P.H.H. 1999 http://www.stockton-press.co.uk/ph

Cervical cancer screening: spatial associations of outcome and risk factors in Rotterdam FAF Kreuger1,*, HAM van Oers1,2 and HGT Nijs1,3 Department of Health Promotion, Municipal Health Services Rotterdam Area, The Netherlands; 2Addiction Research Institute, Erasmus University Rotterdam, The Netherlands; and 3Department of Health Policy and Management, Erasmus University Rotterdam, The Netherlands 1

Objective: Obtaining insight into the geographic distribution of attendance and smear test results at the cervical cancer screening program in Rotterdam neighbourhoods, associated with socio-economic status, marital status and the percentage migrants. Design: Ecological analysis was carried out on data on cervical cancer screening outcome and population ®gures, provided by the Rotterdam Local Health Information System, in which health information is collected at neighbourhood level. Setting: The cervical cancer screening program in the city of Rotterdam. Participants: Fifty-three neighbourhoods, with overall 569 105 inhabitants, of whom 70 621 women between 1992 and 1994 were invited for the screening program. Main results: Between neighbourhoods a large difference in attendance rate and the percentage positive smears exists. A high socio-economic level of a neighbourhood, and a low percentage migrants, single or divorced women correspond with high attendance. A high socio-economic status of a neighbourhood and a low percentage migrants correspond with a low percentage smear test Pap 3B or higher. Socio-economic status, percentage migrants and marital status are highly interrelated on neighbourhood level. Multivariate analysis showed a negative correlation between the attendance rate and the percentage of single and divorced women, and a positive correlation between the percentage migrants and the percentage of positive smears (Pap 3B or higher). Conclusion: Various risk groups, showing low attendance or a high percentage of positive smears, are clustered in neighbourhoods and can be identi®ed by socio-economic status, marital status and nationality. Activities to improve attendance can be focused towards these neighbourhoods. Keywords: cervical cancer; screening outcome; risk-groups; ecological study

Introduction Attendance is an important factor contributing to the health effect of screening programs. In the case of cervical cancer, on a national scale most cases of invasive cancer occur in unscreened or poorly screened women.1 In The Netherlands attendance to cervical screening varies considerably in different regions.2 Previous studies have found the attendance rate to be dependent on marital status, age, socio-economic status, degree of urbanisation, and nationality.3,4 With regard to health policy, these outcomes can be used in the design of intervention programs, by which these risk-groups are distinctly approached. However, in order to carry out such intervention programs in an ef®cient and effective way, it is worthwhile to investigate whether riskgroups showing low attendance and a high percentage of abnormalities, can be identi®ed geographically, so efforts can be directed at certain areas. In the area of Rotterdam, screening has taken place since 1976. During the period 1976 ± 1984, a screening project was carried out on an experimental basis in three urban regions: Nijmegen, Utrecht and Rotterdam. On basis of this project, it was concluded that a centrally organized screening program could be implemented in the entire Netherlands, which then was started in 1988.5 On this *Correspondence: Mrs FAF Kreuger, Department of Health Promotion, Municipal Health Service of Rotterdam, PO Box 70032, 3000 LP Rotterdam, The Netherlands. Accepted 21 October 1998

occasion the organisation partly changed; instead of being carried out by specially trained women in community centres, as was the case in the experimental program, the smear is now taken by a general practitioner. Also, women are now invited by the coordinating institute to make the appointment for the smear by themselves, while during the experimental program, women were invited on a ®xed date and time. The target population remained women aged 35 ± 54 y, with a screening interval of 3 y. From 1976 until now, the Municipal Health Service Rotterdam Area functions as the coordinating institute for this cancer screening program in the Rotterdam area. The city of Rotterdam is divided into 84 neighbourhoods. In order to investigate whether high-risk areas can be distinguished, the geographic distribution of attendance rate and the results of the smear test at the cervical cancer screening program over these neighbourhoods are presented, and spatial associations with data on socioeconomic status, marital status and the percentage migrants are studied. Methods The number of women invited to the cervical screening program is provided by the Department of Health Promotion of the Municipal Health Service. The number of attended women and the results of the smear test (Papanicolaou-classes) are provided by four laboratories in the surrounding district. Population ®gures, ®gures on

112

Outcome-risk factors Ca Cervix Uteri FAF Kreuger et al

marital status, nationality and the indicator for socioeconomic status are provided by the Municipal Centre for Statistics. The indicator for the socio-economic status of a neighbourhood is composed from several variables such as educational level, unemployment, income, percentage migrants and percentage receiving social bene®ts, by means of a principal component analysis. In this study the indicator for socio-economic status for 1991 was used.6 The other data were available over the years 1992 ± 1994. This period corresponds with one full screening round, in which all women in the Rotterdam area aged 35 ± 53 y were invited to the screening program once. The attendance percentage per neighbourhood is de®ned as the number of women per neighbourhood attending the screening after receiving an invitation in 1992 ± 1994, divided by the number of women who were invited in 1992 ± 1994. For the population ®gures, mean values per neighbourhood were calculated over the years 1992 ± 1994. Two different de®nitions for positive smears were used: 1. Pap 3B or higher. In the screening guidelines, a smear test result of Pap 3B or higher (severe dysplasia or worse) should be followed by a histological examination by a gynaecologist. 2. Pap 3A or higher. In the screening guidelines, women with a smear test result of Pap 3A (mild or moderate dysplasia) are requested for a second smear. Neighbourhoods having 2000 inhabitants or less (which are mainly harbours) are excluded from the analysis to avoid large ¯uctuations in the data due to outliers, caused by small numbers. In total 53 neighbourhoods remained in the analysis, containing 96% of the Rotterdam population. The data on nationality show an asymmetric distribution. A

square-root transformation, as developed by Tukey, leads to a symmetrical distribution. The indicator for socioeconomic status, which is a continuous variable, is divided into four categories; (1) low; (2) middle lower; (3) middle higher; and (4) high, corresponding to the four quartiles. The distribution of the attendance rate, the percentage positive smears, the percentage migrants and the percentage of groups of different marital status over the 53 selected neighbourhoods are presented as interquartile ranges. For spatial associations, bivariate relationships are investigated. As the attendance rate, the percentage positive smears, the percentage migrants and marital status are continuous variables, Pearson correlation coef®cients are used as the measure of association. For relationships with socio-economic status, which is a categorical variable, oneway analysis is used. Multivariate relationships between the risk factors and the attendance rate or the percentage positive smears were studied with regression analysis. For the regression analysis, socio-economic status was reconstructed into three dummy variables. Results Table 1 shows that the restriction to neighbourhoods with more than 2000 inhabitants does not affect the used variables signi®cantly. Between neighbourhoods large differences in the attendance rate and the percentage abnormal smears exist. Table 2 shows a range of over 20% in the attendance rate, which varies from 36 ± 58% per neighbourhood. The percentage Pap 3A or higher varies from 0.0 ± 2.2%, and the percentage Pap 3B or higher varies from 0.0 ± 1.3% per neighbourhood. Also the percentages per neighbourhood of

Table 1 Key ®gures for Rotterdam: entire city and selected neighbourhoods, 1992 ± 1994 Overall Rotterdam

After selection

Percentage remaining

84 594 831 38.6 77 165 266 817 236 933 47 357 43 722 73 860 33 898 46.9% 388 1.05% 128 0.34%

53 569 105 37.4 74 872 255 669 226 199 45 592 41 643 70 621 32 422 45.0% 369 1.21% 123 0.41%

Ð 95.7 Ð 97.0 95.8 95.5 96.3 95.2 95.6 95.6 Ð 95.1 Ð 96.1

Neighbourhoods Inhabitants (male ‡ female) Mean age (y) Migrants Single Married Divorced Widowed Women invited for cervical cancer screening Women attending the screening program Attendance percentage Positive result smear test (  Pap 3A) Mean percentage positive smear test (  Pap 3A) Positive result smear test (  Pap 3B) Mean percentage positive smear test (  Pap 3B)

Table 2 Review of data used in the analysis; interquartile range and mean. Values are given per neighbourhood, the data are on the selected 53 neighbourhoods

Percentage migrants Percentage single Percentage married Percentage divorced Percentage widowed Percentage attendance Percentage positive smears (  Pap 3A) Percentage positive smears (  Pap 3B)

1st quartile

2nd quartile

3rd quartile

4th quartile

Mean

1.6 ± 4.0% 30.8 ± 37.8% 28.1 ± 32.6% 3.4 ± 6.6% 1.6 ± 4.1% 36.0 ± 41.7% 0.0 ± 0.9% 0.0 ± 0.2%

4.1 ± 8.7% 38.2 ± 46.2% 32.9 ± 37.0% 7.3 ± 8.4% 4.1 ± 5.7 41.7 ± 44.2% 0.9 ± 1.1% 0.2 ± 0.4%

9.4 ± 23.8% 47.4 ± 53.4% 37.5 ± 43.8% 8.5 ± 9.3% 5.7 ± 8.7% 44.4 ± 48.0% 1.1 ± 1.5% 0.4 ± 0.6%

24.5 ± 39.2% 53.4 ± 57.7% 43.8 ± 51.3% 9.4 ± 12.3% 9.4 ± 14.0% 48.2 ± 58.0% 1.5 ± 2.2% 0.6 ± 1.3%

13.5% 46.3% 38.7% 8.2% 6.8% 45.0% 1.2% 0.4%

Outcome-risk factors Ca Cervix Uteri FAF Kreuger et al

the different risk groups vary signi®cantly. The percentage migrants per neighbourhood shows a range of over 35%, varying from 1.6 ± 39%, and the percentages per neighbourhood of women being single, married, widowed or divorced show ranges from over 25% to under 10%. Bivariate analysis A high percentage of both married and widowed women correspond with a high percentage women attending the screening program. Vice versa, a high percentage of both single and divorced women and a high percentage of migrants correspond with a low attendance percentage (Table 3). Furthermore, the mean attendance rate is signi®cantly higher in neighbourhoods with a high socioeconomic status compared to neighbourhoods with a low or middle lower socio-economic status (P < 0.05). The percentage positive smears of Pap 3B or higher shows a positive correlation with the percentage migrants. In the four categories of neighbourhoods with increasing socio-economic status, the percentage Pap 3B or higher

113

declines accordingly, being signi®cantly higher in neighbourhoods with low socio-economic status, compared to neighbourhoods with a high socio-economic status (P < 0.05). No signi®cant correlation was found between the socio-economic status and the percentage positive smear of Pap 3A or higher. In Figure 1 the mean attendance rate and the mean percentage smear of Pap 3B or higher is presented per groups of neighbourhoods with increasing socio-economic status, percentage single and divorced women, percentage married and widowed women, and percentage migrants. Marital status, the percentage migrants and the socioeconomic status of a neighbourhood appear to be highly interrelated. A high percentage of migrants corresponds with a high percentage of both single and divorced women, and with a low percentage of both married and widowed women (P < 0.001). Also, among groups of neighbourhoods with decreasing socio-economic status, the mean percentage migrants, and the mean percentage of both single and divorced women increase (and the mean percentages of both married and widowed women

Table 3 Correlation between percentage migrants and marital status, to attendance and results of cervical cancer screening. (N ˆ 53 neighbourhoods) Attendance Percentage root) Percentage Percentage Percentage Percentage

migrants single married divorced widowed

(square

70.51 (P < 0.001) 70.62 0.60 70.35 0.43

(P < 0.001) (P < 0.001) (P < 0.01) (P < 0.01)

% positive smear (  Pap 3A) 0.16 (ns) 0.11 70.09 70.06 70.05

(ns) (ns) (ns) (ns)

% positive smear (  Pap 3B) 0.43 (P < 0.01) 0.31 70.32 0.28 70.22

(ns) (ns) (ns) (ns)

Figure 1 Mean attendance and mean percentage of smears  Pap 3B per neighbourhoods with an increasing percentage of the different risk groups (the categorization of low, middle lower, middle higher and high corresponds to the interquartile ranges of the presented risk groups over the 53 neighbourhoods). N ˆ 53 neighbourhoods.

Outcome-risk factors Ca Cervix Uteri FAF Kreuger et al

114

decrease) signi®cantly (P < 0.001). This means that the factors associated with attendance and the percentage abnormal smears, are clustered on neighbourhood level. Multivariate analysis On basis of the similar correlation with the attendance rate, the percentages of single and divorced women per neighbourhood were combined before they were put into the regression model. Regression analysis with attendance rate as the dependent variable shows that of the highly interrelated socio-economic status, marital status and percentage migrants per neighbourhood, only marital status remains as a signi®cant indicator of attendance. This explains 39% of the variance in the attendance rate, whereby a high percentage of single and divorced women corresponds with a low attendance rate. Regression analysis with the percentage of a smear of Pap 3B or higher on one hand, and socio-economic status and the percentage migrants per neighbourhood on the other, explains 26% of the total variance. In this analysis, only the percentage migrants of a neighbourhood remains a signi®cant indicator, showing a positive correlation with the percentage abnormalities at the neighbourhood level. Discussion In order to establish the geographical distribution of risk groups for cervical cancer in the city of Rotterdam, the attendance rate and smear test results on neighbourhood level and associations with known risk-factors were investigated. Neighbourhoods showing low attendance and neighbourhoods showing a high percentage abnormal smears could be distinguished. Furthermore, risk groups for cervical cancer were found to be clustered on a neighbourhood level. The presented correlations, however, do not indicate causal relationships. Furthermore, it must be realised that the attendance rate can be distorted because of factors like migration, previous hysterectomy or opportunistic screening (screening outside the of®cial program). Our ®ndings are consistent with those from a British study in which the percentage of general practice population from ethnic minority groups and variables associated with social deprivation were signi®cantly negatively correlated with general practice attendance rates.7 Also similar correlations were found at the individual level: a low attendance rate among migrants and women of low socio-economic status was found in other regions of The Netherlands8,9 and the relationship between marital status and attendance was reported in a Belgian and an Italian study.10,11 Multivariate analysis of our data showed that after putting all factors together, marital status remained a signi®cant indicator of the attendance rate. It is, however, unlikely that the correlation between attendance rate and the percentage migrants or socio-economic status is caused solely by an underlying relationship with marital status. More likely, the high interrelation of marital status, socioeconomic status and the percentage migrants on neighbourhood level, combines them into one factor, showing how much these three factors together associated with attendance, are clustered on a neighbourhood level. Similarly, although the regression analysis with the percentage migrants and socio-economic status only left the percentage

migrants as a signi®cant indicator of the percentage Pap 3B and higher, it may be incorrect to conclude that the percentage Pap 3B and higher per neighbourhood is only related with the percentage migrants, as the percentage migrants and socio-economic status are highly interrelated. Overall, the attendance rate is low. Koopmanschap and others calculated that the attendance rate at an invitation scheme without a reminder, should be at least 60% in order to make the screening cost-effective. High attendance to a well balanced invitation scheme means a good coverage of the population at risk. This is far more effective than, for example, opportunistic screening by which women with a relative low risk get screened too much, while nonattenders, with a relative high risk for cervical cancer, are still not screened.4 The attendance rate depends for an important part on the organisation of the screening. In the UK, for example, until the late 1980s, screening was organized on an ad hoc basis by general practitioners and by some local health authorities who set up screening services on their own initiatives. Since, however, a national cervical screening policy was designed, district health authorities are made responsible for call and recall systems and general practitioners are only fully paid when 80% of the women on their list are screened in the previous 5.5 y (when 50% of the women are screened 33% of the full amount is paid for). As a result the percentage of general practices reaching the 80% target increased from 53 ± 83% between 1990 and 1993 in England.12 In the Netherlands a national organized screening program already functions. Still, further efforts could lead to an increase in the attendance rate. In a Dutch project in which the general practitioner invited women on a ®xed date and time, instead of an invitation by the coordinating institute to make the appointment themselves, a 14% higher attendance rate was found.13 In some areas general practitioners now of®cially invite women for the screening program. An invitation system by the general practitioner, however, may not work in a large city like Rotterdam as there is a considerable movement of patients between general practices and not all doctors have yet a fully computerized administration. Therefore, for Rotterdam a central organized invitation system reinforced with a reminder and active case-®nding by the general practitioner seems to be the best strategy. Conclusions The attendance rate can also be raised by organising special promotion campaigns directed towards the women residing in those neighbourhoods in which target groups are clustered. In the city of Rotterdam preparations are made for an intervention program by which per neighbourhood migrant women are approached by (female) peer health educators for health education meetings in local centres. In order to reach as many women as possible a community intervention approach is chosen: the women are personally invited through telephone and house-to-house calls and men of the respective migrant groups are informed about the meetings by (male) peer health educators in coffeehouses. Also through neighbourhood health care workers, imams, and articles in local newspapers, women are stimulated to participate in the education groups. During the meetings the migrant women receive information about cervical cancer and the screening program in their native language. Much attention will be paid to speci®c questions of the participants, for example, about the possibility of

Outcome-risk factors Ca Cervix Uteri FAF Kreuger et al

getting a female doctor for the examination. The results of the intervention program will be evaluated and eventually published. Acknowledgements The authors wish to thank the Project Committee Cervical Cancer Screening Rotterdam Area for carefully reading the manuscript. References 1 Ballegooijen M van et al. Preventive pap-smears: balancing costs, risk and bene®ts. Br J Cancer 1992; 65: 930 ± 933. 2 Nijs HGT, Kreuger FAF, Poel MBP van der. The Screening Program for Cervical Cancer in Rotterdam Area. First Round `New Organisation Structure' 1989 ± 1991 (in Dutch). GGD Rotterdam e.o.: Rotterdam, 1993. ISBN 90-5175-084-6. 3 Kreuger FAF, Nijs HGT, Poel MBP van der. Determinants of attendance and Pap smear results for the national screening program for cervical cancer in the area of Rotterdam, The Netherlands, 1989 ± 1991 (in Dutch). Tijdschr Soc Gezondheidsz 1994; 72, 309 ± 313. 4 Koopmanschap MA et al. Cervical-cancer screening: attendance and cost-effectiveness. Int J Cancer 1990; 45: 410 ± 415.

5 Evaluation Committee. Population screening for cervical cancer in The Netherlands. Int J Epidemiol 1989; 18: 775 ± 781. 6 Leijs ALC, Das P. Socio-Economic Indicators of Rotterdam Neighbourhoods In 1991 (in Dutch). Municipal Centre for research and statistics: Rotterdam, 1991. 7 Majeed FA et al. Using patients and general practice characteristics to explain variations in cervical smear update rates. BMJ 1994; 308: 1272 ± 1276. 8 Prins M. Screening for Cervical Cancer: Attendance, Coverage, Cytology, Costs and Effects in the City of Utrecht (in Dutch). GG & GD Utrecht, maart: Utrecht, 1991. 9 Verberk H. General practitioner and cervical cancer screening (in Dutch). Medisch Contact 1988; 43: 815 ± 816. 10 Hal G van, Weyler J, Eylenbosch WJ. Attenders and nonattenders to screening for cervical cancer in Burcht-Zwijndrecht (in Dutch). Tijdschrift Sociale Gezondheidszorg 1987; 65: 740 ± 742. 11 Ciatto S et al. Attendance to a screening programme for cervical cancer in the city of Florence. Tumori 1991; 77: 252 ± 256. 12 Ibbotson T, Wyke S. A review of cervical cancer and cervical screening: implications for nursing practice. J Advanced Nursing 1995; 22: 745 ± 752. 13 Kant AC et al. Screening for cervical cancer: the effectiveness of different call-systems (in Dutch). Medisch Contact 1991; 46: 469 ± 471.

115