Journal of the New Zealand Medical Association, 30-October-2009, Vol 122 No 1305
Monitoring the performance of New Zealand’s National Cervical Screening Programme through data linkage
Hazel Lewis, Li-Chia Yeh, Bobby Almendral, Harold Neal
The introduction of an organised cervical screening programme in New Zealand in 1991 has had a major impact on the incidence of cervical cancer.1 However the effectiveness of cervical screening depends on regular participation in the National Cervical Screening Programme by all women between the ages of 20 and 70 years who have ever been sexually active, as well as robust quality assurance systems for each of the Programme’s components (health promotion, smear taking, colposcopy, treatment, and laboratory services).
The New Zealand Cervical Cancer Audit2, published in November 2004, was carried out following the Ministerial Inquiry into the Under-reporting of Cervical Smear Abnormalities in the Gisborne Region.3 One of the recommendations following the Gisborne Inquiry was that the National Cervical Screening Programme (NCSP) should implement a process for ongoing review of the screening histories of women who develop cervical cancer. Such case reviews are conducted by several countries with organised cervical screening programmes,4–10 and should be distinguished from periodic full scale programme audits.
Evaluation of the performance of the NCSP currently involves independent monitoring of a range of performance indicators against agreed targets; regular independent audits of specific Programme components; three yearly reviews of the Programme as a whole in accordance with the Health (National Cervical Screening Programme) Amendment Act 2004; ongoing monitoring of smear takers, laboratories and colposcopy services against the Programme’s own quality standards; investigation of complaints; and monitoring trends in Programme outcomes—cervical cancer incidence and mortality.11
Ongoing review of the screening histories of women diagnosed with cervical cancer provides a complementary approach to auditing the overall performance of the NCSP. It could also potentially lead to the identification of specific problems requiring more detailed investigation and possible corrective action. The Health (National Cervical Screening Programme) Amendment Act 2004 (s112T–s112Z) provides the legal basis for these reviews.
Women enrolled in the NCSP have their smear histories, laboratory cytology and histology (biopsy) results recorded in a centralised database, the National Cervical Screening Programme Register (NCSP-R). All tissue diagnoses of cancer (including cervical cancers by stage and histological type) are notified to the New Zealand Cancer Registry (NZCR) under the Cancer Registry Act 1993.12 Additionally, both databases identify individuals by their National Health Index (NHI), a unique personal identifier assigned to all persons at first contact with the health system.
It is thus possible to link a woman’s cervical cancer diagnosis (recorded on the NZCR) to her smear history (recorded on the NCSP-R) via her NHI. Since 2000, the NHI completion rate has been very high (close to 100% on both databases). While there have been problems experienced earlier (prior to 1997) of multiple NHIs being assigned to the same woman, this has been greatly reduced due to more active detection and resolution of duplicates on the NHI database.
Ethical concerns regarding privacy and confidentiality have to be considered in undertaking any record linkage process in addition to technical challenges. These concerns were dealt with by special legislation, the Health (National Cervical Screening Programme) Amendment Act 2004. This provides the legal basis for the record linkage and ensures that the linked dataset is maintained securely and is queried only by authorized NCSP staff.
The objectives of this paper are therefore to:
The monitoring process will eventually include both examination of screening histories obtained from the NCSP-R (Phase 1) and review of clinical records as well as rereading of cervical smears and biopsy histologies, where this is indicated (Phase 2). This paper covers only Phase 1 of the case review process—i.e. the examination of linked cancer registrations and screening histories.
Data sources and matching—Every quarter, notifications of cervical cancer on the NZCR were matched against the NCSP-R, with a 6-month delay to allow for processing of notifications by the NZCR. The period covered for this review was 4 years, from 1 January 2003 to 31 December 2006.
For each notification of cervical cancer, data extracted from the NZCR included the NHI; demographics (date of birth, domicile, ethnicity and date of death, if relevant); cancer information including date of registration, histological type, morphology, and stage.
Staging involved two classifications: histological staging of cancers as microinvasive versus invasive; and clinical staging using FIGO (International Federation of Gynecology and Obstetrics).
Each cancer notification was linked to its corresponding NCSP-R record by NHI. Data extracted from the matching NCSP-R record included demographics (as above, except ethnicity—NHI ethnicity used in preference), smear history and histology results (if any). Smear history included the dates of all smears and the cytology result for each. Histology results (and dates recorded) from biopsies taken at colposcopy were extracted and follow-up noted. Both smear histories and histology results were extracted for the full period for which each woman was enrolled on the NCSP-R—i.e. for a maximum of 16 years.
Data analysis—For each quarter that the data were analysed, the same flow chart was used (Figure 1).
Figure 1 Data analysis summary flowchart
Firstly, nonsquamous cancers were excluded (i.e. adenocarcinomas and other histological types of cervical cancer), as cervical screening has limited effectiveness in preventing these cancers. Therefore, analysis was restricted to squamous and adenosquamous cancers only.
Secondly, women aged >80 years at diagnosis were excluded as they were too old to have benefited from the Programme (which began in 1991 and screens only women <70 years).
Thirdly, ‘eligible’ women (i.e. women diagnosed at age <80 years with squamous or adenosquamous cervical cancer) were classified as ‘enrolled’ or ‘not enrolled’ in the NCSP based on their smear history. ‘Enrolment’ was defined as having one or more smears recorded 6 months or more prior to the date of diagnosis or notification of cervical cancer. (A smear taken within 6 months of diagnosis was considered to be part of the diagnostic process i.e. the cancer was already present).2
Finally, eligible enrolled women were classified as ‘irregularly’ or ‘regularly’ screened, again based on their smear history. To be classified as ‘regularly screened’ a woman must have undergone her first smear before a specified age and then have had at least one smear every 5 years thereafter to the date of cancer diagnosis or notification.
Five rather than 3 years was selected to allow a degree of flexibility in screening. This is consistent with many international programmes and the natural progression of cervical pre-cancer to microinvasive and finally fully invasive disease.14 The specified age was determined by the birthdate of the woman in relation to the date of commencement of the Programme  (i.e. specified age <30 years if born >1960, <40 years if born 1950-59, <50 years if born 1940-49, <60 years if born 1930–39). A woman not meeting these criteria was considered to have had only irregular or insufficient smears in her lifetime.
The data were summarised by period (quarter or year), demographics (age group, ethnic group), cancer type (squamous, adenosquamous), histological stage (microinvasive, invasive), and smear history (enrolment status, regularity of screening if enrolled).
Comparison of case reviews with audit results—Key results from the case reviews for the period 2003–2006 were compared, where possible, with results extracted from the report of the New Zealand Cervical Cancer Audit (published in 2004), covering the period 2000–2002.
Differences in methods between the case reviews and NZCCA should be noted, including differences in the period covered, review of women’s full medical records and independent reread of slides (both of which were included in the NZCCA but not in phase 1 of the case reviews).
From 1 January 2003 to 31 December 2006 a total of 625 cases of cervical cancer were notified to the NZCR. Of these, 430 (69.0%) were squamous cancers, 31 (5.0%) were adenosquamous, 116 (18.6%) were adenocarcinoma and 47 (7.5%) were a variety of other histological types.
Of the 461 squamous plus adenosquamous cancers, 23 were diagnosed in women aged >80 years, leaving 438 cases eligible for linkage. These comprised 348 invasive and 90 microinvasive cancers. FIGO stage (a clinical staging system that is distinct from the histological staging of cancers as ‘invasive’ or ‘microinvasive’) was missing for 199 (45%) cancers.
For 225 cases (48% of all eligible cases) a definite linkage was made to a screening record in the NCSP-R (i.e. identical NHI plus consistent demographic details in both records). These cases were therefore considered to have occurred in enrolled women.
For a further 202 cases (43.8% of all eligible cases) there was no matching record on the NCSP-R, or the only smear history recorded related to a period <6 months prior to the date of cancer diagnosis or notification. These cases were therefore considered to have occurred in women who had never enrolled (participated) in the Programme. For the remaining 11 cases (2.4% of all eligible cases), the possibility that the woman might have previously been enrolled but had subsequently chosen to withdraw from the Register, or that an error had occurred in recording either the NHI or the demographic details in one or both of the two databases, could not be excluded. These 11 women were not able to be assigned an enrolment status.
Analysis of the screening histories of the 225 women who were classified as ‘eligible enrolled’ women (i.e. had been diagnosed at <80 years of age with either squamous or adenosquamous cervical cancer and were matched to a record on the NCSP-R extending back more than 6 months prior to cancer diagnosis) found that 137 (60.9% of the 225) had only been irregularly or insufficiently screened; 85 (37.8% of the 225) had been diagnosed with cervical cancer despite being regularly screened; while for 3 women (1.3% of the 225) no assessment could be made because recording of the screening history was inadequate. Therefore 31.3% of ‘eligible’ cancers (137/438) occurred in women who were enrolled in the Programme but had been smeared insufficiently often, or too irregularly, to have derived full benefit from screening.
In total therefore, 73.5% of eligible cancers (137 enrolled and 202 not enrolled) (339/438) occurred in women who were unscreened or insufficiently screened. However, 18.4% (85/438) of eligible cancers did occur in women who had experienced regular screening (5 yearly or more frequently, beginning at specified age depending on birth date in relation to NCSP commencement date).
These results are summarised in Table 1. This table also shows the distribution of squamous cancers by histological stage (microinvasive or invasive).
Table 1. Number of cervical cancer cases, 2003–2006
Table 2. Contribution of likely causes
* Annual average number of eligible squamous and adenosquamous cases, 2003–2006.
Table 2 shows that, of 110 eligible cases occurring per year on average, almost half (49%) occur in women who are not enrolled in the Programme and so are likely to have never been screened. Just under one-third (31%) occur in women who are enrolled in the Programme but have been screened only infrequently and/or irregularly. Therefore, combining these two categories, exactly 80% of potentially preventable cases are estimated to probably result from inadequate coverage.
The remaining 20% of cases (85 over 4 years or 21 per year on average) occur despite regular screening (Table 3). These cases represent a mix of true interval cancers (cancers that develop within 3 years of a confirmed negative smear) and Programme quality problems (which could have involved any or all of smear taker, laboratory or colposcopy/treatment issues).
Table 3. Characteristics of eligible women who developed squamous cervical cancer between 2003 and 2006, despite regular screening
Note: Number in brackets is percent of total.
Table 3 shows that cancers among regularly screened women are more likely than cancers among all eligible women to occur at young adult ages (27% <30 years compared with 11 %, p=<0.001) Regularly screened women who develop cervical cancer are also more likely to be diagnosed at a microinvasive stage than all women eligible for screening (31% microinvasive compared with 21%, p=<0.001).
Comparison with audit—Table 4 compares key findings from the case reviews for 2003-06 with those reported by the NZ Cervical Cancer Audit (NZCCA) for 2000–2002.
The results from the two studies show strong agreement, despite differences in the period included and in methodology. In both the NZCCA and case reviews, approximately 50% of cancers occurred in unscreened, 30% in irregularly screened and 20% in screened women. There were some differences in type distribution, with a higher proportion of squamous and a lower proportion of adenocarcinomas in the NZCCA. The NZCCA found FIGO stage missing in only approximately 26% of cases, compared with approximately 50% in the case reviews. The most likely explanation is that the NZCCA obtained staging data directly from the records of women with cervical cancer.
Table 4. Comparison of case review 2003–2006 with Audit 2000–2002
Smear histories of women diagnosed at <25 years of age—The findings of the case reviews of this age group are of particular interest. These young women will have had only a few smears, so will be more dependent than older women on the quality of each smear. Cancers in young women may be more aggressive and fast growing, 27 so interval cancers are more likely in this age group. Trends in this age group will influence whether the age of screening initiation can be shifted to age 25 (or even 30 years), as recommended by WHO and several European countries.14
There were 10 women, (approximately 3 per year over the 4 years studied) with squamous or adenosquamous carcinoma, who were aged under 25 years at diagnosis. Of these 10 cancers, 4 were diagnosed as microinvasive, 4 as invasive, and 2 had missing histological stage.
The purpose of this paper is to report on one of several quality assurance processes used by the NCSP to monitor the Programme’s performance.
This paper demonstrates that the NHI (the National Health Index, a unique personal identifier) can be used to successfully link cervical cancer registrations to women’s cervical screening histories as recorded on the NCSP Register. The linked dataset enables further investigation of the ‘cause’ of cervical cancer to be carried out by an analysis of the smear histories of women, with relatively little delay (6 months post registration). Despite differences in method, it was also possible to compare trends in cervical cancer ‘causation’ with those found in a slightly earlier time period, through the NZCCA, with highly consistent results.
A key finding from the first 4 years of ongoing screening history reviews (and from the Audit) is that 80% of potentially preventable cervical cancers involve women who are not enrolled in the Programme or who have been only infrequently and irregularly screened. This confirms that improving coverage remains a priority for the Programme, which was also the main conclusion of the NZCCA.
It is disappointing that the proportion of cancers attributable to insufficient coverage has not declined since the NZCCA. However, a recent OECD report13 suggests that improving coverage may not be an easy task: at about 72% overall, New Zealand’s coverage is already among the highest achieved anywhere in the developed world, being only slightly lower than that of some Nordic countries. However, coverage is lower for Māori, Pacific, and Asian women, so scope for improvement in coverage still exists.
At the same time, the review of women’s screening histories has revealed that each year an average of 21 women will be diagnosed with this potentially preventable cancer despite having participated fully in the screening programme. The young age distribution of these cancers and their higher likelihood of diagnosis at a microinvasive stage (seen in Table 3) suggest that many could be interval cancers. These women clearly need further, more detailed investigation to determine which are in fact interval cancers and which result from quality problems in the screening pathway.
The next phase (Phase 2) will involve the clinical review of the patient records of these regularly screened women, including re-reading relevant smears and histology specimens as appropriate. This will need to be done sensitively and securely, so as not to compromise the privacy of these women and to minimise the impact of the review on confidentiality of their personal information. The main objective is to distinguish probable interval cancers from possible ‘errors’.
It must be emphasised that the objective is not to ‘blame’ providers for isolated errors, but to try and identify systemic problems that the Programme can address. No screening programme can ever be 100% effective and a low error rate is inevitable. For some women, no satisfactory explanation for their cancer may ever be found. Phase 2 will also provide an independent check of a number of other quality assurance processes already in place, including the routine ’42-month look back’ carried out by laboratories (in which the slides of every woman diagnosed with cervical cancer are re-read)
The review of screening histories may ultimately be extended to a case control design. This could be achieved by recruiting women enrolled on the Programme who have not developed cancer to act as ‘controls’. Comparison of their screening histories with those of the ‘case’ women will strengthen the conclusions that can be drawn as to the reasons why women continue to develop cervical cancer.
This case review process could also be strengthened in other ways. Most importantly, the high rate of missing FIGO stage data recorded on the NZCR needs to be addressed. Documentation of periods when women may have withdrawn from the Programme (e.g. because of ‘opting off’ or being overseas) is already being improved through the NCSP-R redevelopment.
Finally, this process of reviewing screening histories will also contribute to the evaluation of screening policy regarding the appropriate age to start and stop screening and the screening interval. Of particular relevance is the question of whether to offer screening to women less than 25 years of age.
Our results show that approximately three cases per year in this age group are to be expected. Whether this justifies screening women under 25 years of age is a matter of debate. A review of the literature shows that, as in our study, the main reason for failure to prevent cervical cancer is lack of participation in regular screening.4–10,15–22
A recent study of 225 women with cervical cancer in Canada16 found that 68% of women who developed cervical cancer had not participated in routine screening and 15% had participated only sporadically. Another study of women with cervical cancer in Denmark found that 23% had never had a smear and 61% did not participate regularly in screening.15
However, it is also well recognised that the conventional Pap smear has limitations as a screening test and has a false negative rate of about 20% for high grade lesions.2,26 Yet in spite of these limitations, the NCSP has achieved a 50% reduction in incidence and 60% reduction in mortality since becoming operational.1
In order to encourage women to participate in cervical screening the National Screening Unit launched a communications campaign in September 2007, targeting Maori and Pacific women. This has resulted in a slow but steady increase in participation in the Programme.28 The Programme has also recently reviewed the 1999 guidelines for the management of women with abnormal smears as well as new technologies (such as HPV testing) to improve overall Programme performance. New ways of managing cervical abnormalities in conjunction with new technologies are likely to be needed as we enter the era of HPV vaccination.23-25
In conclusion, our experience over 4 years has demonstrated that ongoing linkage of cancer data to screening data provides a robust approach to monitoring the performance of the NCSP alongside other approaches. This process should continue but needs to be accompanied by a more detailed investigation of the small number of women who develop cancer despite regular screening (Phase 2 investigation). Recruitment of control women should also be included, so that causal conclusions can be more reliably drawn through use of a case control study design.
Note: This paper is published in accordance with The Health (National Cervical Screening Programme) Amendment Act 2004 (section 112S).
Competing interests: None known.
Author information: Hazel Lewis, Clinical Leader, National Cervical Screening Programme, National Screening Unit, Ministry of Health, Wellington; Li-Chia Yeh, Statistician, Public Health Intelligence, Ministry of Health, Wellington; Bobby Almendral, Technical Specialist, National Screening Unit, Ministry of Health, Wellington; Harold Neal, Scientific Advisor, National Cervical Screening Programme, National Screening Unit, Ministry of Health, Christchurch
Acknowledgements: We acknowledge assistance of Jane McEntee (previous NCSP manager), Diane Casey (current NCSP manager), and Lesley Mack, Pauline Fallon, Jennifer Beaulac (previous NCSP policy analysts).
We are grateful to NZHIS staff for data matching; including Tracey Vandenberg, Ian Galley, Rebecca Kay, and Rebecca Hislop.
We also acknowledge peer review by Alistair Woodward and Lynn Sadler (University of Auckland); Philip Castle (NIH, USA); and Eileen Hewer (National Screening Unit) who reviewed this article.
We thank Michelle Hooper for administrative support.
Correspondence: Dr Hazel Lewis, Clinical Leader, National Cervical Screening Programme, National Screening Unit, Ministry of Health, P O Box 5013, Wellington, New Zealand. Phone 04 816 4343, Fax 04 816 4484, email: email@example.com
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals