Article Text
PDF
Abstract
Background For evaluation of breast cancer screening and informed prioritisation, it is important to examine the downstream healthcare use associated to participation. The objective of this study is to determine the healthcare use among breast cancer screening participants compared with screening-naïve controls.
Methods The study is a register-based cohort study with 14 years of follow-up. We compare healthcare use among women who participated in the initial phase of the stepwise breast cancer screening implementation in Denmark (stratified on screening result: normal, false positive and breast cancer) compared with those invited in subsequent phases.
Results Screening participants, especially those with false-positive results, tended to use primary healthcare services more than the screening-naïve group. Women with breast cancer and false positives received more breast imaging compared with the screening-naïve group. False positives consistently had the highest use of drugs compared with the control group. All screening groups had significantly higher use of outpatient clinic visits in the year of and following screening compared with the screening-naïve group. Screening groups were more likely to receive additional diagnoses in the years following screening than the screening-naïve group. There were no significant differences in medical procedures and days of hospitalisation.
Conclusions The study highlights differences in primary healthcare use among screening groups compared with the screening-naïve group. Since use of primary care services is at the discretion of the women, this implies increased worries about health. Thus, these results indicate increased healthcare-seeking behaviour, especially among women with false-positive results.
- SCREENING
- COHORT STUDIES
- EPIDEMIOLOGY
- HEALTH SERVICES
- PUBLIC HEALTH
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOW ON THIS TOPIC
Receiving a false-positive result in breast cancer screening alters healthcare-seeking behaviour through medicalisation.
WHAT THIS STUDY ADDS
In this register-based cohort study, mammography screening was associated with significantly more contacts to the general practitioner and outpatient clinics, higher drug use, and more diagnoses.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Attending additional healthcare services can be considered an economic consequence of screening and procedures might cause patient-relevant harms.
Introduction
While breast cancer screening is a widely employed tool for breast cancer detection, it is not without unintended consequences. For instance, a woman faces the risk of receiving a false-positive test result. This risk accumulates to 20%–60% for those having annually or biennially screening mammographies between the ages of 40–69.1 Prior research established that false-positive screening outcomes are linked to psychosocial repercussions lasting up to 3 years.2 Moreover, a recent study suggests that some of these consequences may persist for up to 14 years.3
Receiving a false-positive result in a screening programme for a potentially life-threatening disease has been shown to alter an individual’s healthcare-seeking behaviour; studies demonstrate that women who receive false-positive breast cancer screening tests increase their use of healthcare services in the first year after the test.4 5 Qualitative studies attribute these changes to medicalisation, wherein individuals are concerned for their health and seek medical advice to confirm their well-being.6 7 This trend is also observed among heavy smokers who had a false-positive test in CT screening for lung cancer.8 9 However, no studies have assessed the healthcare use effects of a false-positive screening results beyond a year after the screening, and therefore we do not know for how long such an effect is present.10 11 It is important to examine the direct and indirect changes in healthcare use as to inform prioritisation and the benefit-harm ratio of screening.12 13
Our research aims to address this gap by examining whether breast cancer screening induces changes in healthcare-seeking behaviour over a 14-year period. Through an analysis of healthcare data, we will examine usage patterns, differentiating them based on screening results; cancer diagnosis, normal findings and false positives and compare these to a sample comprising screening-naive women.
Method and materials
We hypothesised that:
Women who received false positives increase their use of healthcare services compared with women having normal results and screening-naïve women.
Women with normal results would have a use similar or marginally increased compared with screening naïve, and less than women with false positives and breast cancer.
Women with false positives would have a similar or marginally lower use than women diagnosed with breast cancer for services not directly related to breast cancer treatment.
Differences between groups will subside over time as controls are being invited to screening and the immediate effects of screening will fade.
Study design
This is a matched cohort study based on patient registers. We present data from Danish administrative health and social registers from 2000 to 2018. A protocol is registered at Open Science Forum.14 This study is reported in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) checklist for cohort studies.15
We accessed registers with information about healthcare services. The quality of such register data is high because it is used to generate claims to facilitate reimbursement by the state, to communicate across sections of the healthcare system, and the data will become available to the patients. We also accessed registers on sociodemographic factors as potential confounders. Individual social security numbers link these registers.
Sampling
We included cases who had participated in the initial phase of the stepwise breast cancer screening implemented in Denmark and matched these with controls who were invited in subsequent phases; in other words, controls were screening naïve at study baseline. Denmark introduced breast cancer screening in a stepwise manner based on geographical location, and in 2007, a national breast cancer screening programme was implemented throughout Denmark.16 By the end of 2010, all women living in Denmark aged 50–69 had been invited to the screening.
The screening cohort was sampled from a survey study; 1318 women who underwent screening in 2004–2005 participated in a survey on psychosocial consequences of breast cancer screening.3 17 For the survey, all women with an abnormal screening (later proved breast cancer or false positive) were invited and each woman included was matched with two women with normal results. These women were 50–69 years old in 2004–2005, that is, screening eligible. Each of these women was matched with 10 random screening-naïve women sampled from national registers. We included 10 screening-naïve controls per case to detect differences in rare outcomes, such as hospitalisation. The matching was based on age and population density areas. The screening-naïve controls were not invited for screening until 2007–2010. We followed both screened cases and their matched controls until 31 December 2018, death or emigration whichever occurred first. We traced the women for 3 years before the screening, to account for baseline differences and identify long-term trends. Baseline measurement is the year of screening for the screening groups. Among women with false-positive results, 65.8% had cancer ruled out the same day of recall visit, and only 4% had to wait more than 120 days.18
Outcomes
We included measures of healthcare use from both primary and secondary healthcare (table 1). Outcomes from primary care included: contact to the general practitioner (GP) in-of-hours and out-of-hours, and out-of-hospital specialist contacts with and without out-of-pocket costs. Out-of-hospital specialists with out-of-pocket payments included psychologist, chiropractor, physiotherapist and dentist visits. These primary care outcomes were considered services initiated by the individual woman and thereby measures of healthcare-seeking behaviour. Out-of-hours contacts were non-comparable after 2007 due to organisational restructuring across Danish regions. Secondary care outcomes included hospital admissions, outpatient visits, hospital visits and surgical and non-surgical procedures. These were considered services initiated by healthcare professionals because the GPs in Denmark work as gatekeepers to secondary care and certain primary healthcare specialists. We also included a specific outcome on the use of breast imaging as a proxy for screening-related care cascades. For all these measures, we count the number of services provided or of contacts in a given calendar year. If subjects were registered with more than one hospital admission, visit or surgery on a given day, these are counted as one unless registered in different settings, for example, one visit in general practice and one surgery. We counted all prescription drugs prescribed and redeemed in the given calendar year. These did not include drugs administered at the hospital. We group these outcomes as secondary care as these are not measures of healthcare-seeking behaviour.
Outcomes
Statistical analysis
We employed a two-part model to analyse the healthcare use data. The first part assesses the probability of using a specific healthcare service during a 1-year period modelling a relative risk in reference to the screening-naïve group. The second part models the intensity of use among those who used the particular service within the year, as well in reference to the screening-naïve control group.
The first part was done using a Poisson regression model, estimating the relative risk of using a specific service for cases compared with controls.19 The second part was done using a gamma regression estimating rate ratios, indicating the factor by which cases use the service more or less than their comparators. For both parts, we employed generalised estimating equations (GEE), ensuring an appropriate covariance matrix evaluating a sequence of periods for the same woman and an offset to adjust for incomplete years. The product of the two estimates provides a comprehensive measure of the overall increase in healthcare service use in this group, hereafter coined the group use.
To enhance precision, each analysis was adjusted for the 3 years of healthcare use prior to the screening date, tentatively accounting for the individual’s general level of healthcare use. Additionally, adjustments were made for health status at screening using Charlson’s Comorbidity Index (CCI).20 21 We also included age, cohabitation, degree of urbanisation, working status, country of origin, education, and income as potential confounders. Data exclusion and handling of missing data involved excluding subjects with missing values on ≥90% of background variables (n=15).
To perform the analyses, we used SAS, V.9.4 (SAS Institute) and the r software package, V.3.4.1.22 A significance level of 5% was employed.
Results
We included 14 252 women and their baseline characteristics are summarised in table 2.
Baseline characteristics
The screening groups were similar to screening-naïve controls with regard to age, occupational status, country of origin, educational level, income and CCI (table 2). Women with false positives were more likely to have fewer assets.
The analyses on which the figures are based are presented in online supplemental appendix table 1 and 2.
Supplemental material
Primary care outcomes
From 2005 to 2011, false positives used their GP 1.1–1.2 times more than the screening-naïve group, significantly the first 4 years after screening, while the two other screening groups did not differ from the screening-naïve group. From 2012 to 2018, the groups converge in GP use (figure 1A). From screening and until 2007, the three screening groups are similar and have no differences in the use of GP out-of-hours compared with the screening-naïve group (figure 1B). In general, there are no statistically significant differences between the four groups concerning out-of-hospital specialist visits but women with breast cancer used out-of-hospital specialist visits with no out-of-pocket payment less than the two other screening groups compared with the screening-naïve group (figure 1C). For specialist visits with out-of-pocket payment, screening groups tend to use specialists with out-of-pocket-payment less than the screening-naïve group, again with no statistically significant differences (figure 1D).
Group use of healthcare services for A–F outcomes for screening groups compared with screening-naïve controls. GP, general practitioner.
Breast imaging
Women with breast cancer had statistically significantly more breast imaging than the screening-naïve group; at times more than 13 times higher. False positives also had increased breast imaging throughout the study period; 10%–154% higher. The group with normal results tended to have a lower use compared with the screening-naïve reference. After 2014, the three screening groups regressed towards the screening-naïve group; yet women with breast cancer and false positives continued to have more breast imaging compared with the screening-naïve group, with the group with normal results even lower (figure 1E).
Secondary care outcomes
The group with false positives had the highest use of drugs compared with the screening-naïve group, converging towards no difference at the end of the follow-up period. Women with breast cancer tend to have a statistically non-significant lower use of drugs (figure 1F). In the years after screening, women diagnosed with breast cancer have more hospitalisation days followed by a dip below the screening-naïve group. False positives are not significantly different from the screening-naïve group with regard to days of hospitalisation (figure 2G). For outpatient clinic visits, all three screening groups have significantly higher use in the year of and following screening compared with the screening-naïve group. Women diagnosed with breast cancer continue to have more visits throughout the study period (figure 2H). From 2004 to 2011, the groups have similar use of emergency room visits. From 2012 to 2018, the three screening groups have similar but higher use compared with the screening-naïve group, ending at about two times higher use (figure 2I).
Group use of healthcare services for G–K outcomes for screening groups compared with screening-naïve controls.
In the first years after screening, women with breast cancer have more surgical and non-surgical procedures performed at the hospital. The two other groups show no difference from the screening-naïve group (figure 2J).
Aside from time point 2016, women with false positives had a higher rate of diagnoses compared with the screening-naïve group. The rate of diagnoses was 3%–67% higher among false positives compared with the screening-naïve group. Similarly, women diagnosed with breast cancer were also significantly more likely to receive diagnoses in the first 6 years after screening; 38%–280% more likely. Generally, women with normal results were more likely to receive diagnoses than the screening-naïve group; 2%–36% higher rate of diagnoses except years 2004, 2012 and 2018 were it was 1%–18% lower. All three screening groups were more likely to be diagnosed with a psychiatric or somatic diagnosis and this statistically non-significant trend persisted for the full period (figure 2K).
Discussion
Generally, the three screening groups had a higher use of healthcare services compared with the screening-naïve group. For most outcomes, there was a gradient in the use of services, with highest use among women with breast cancer followed by false positives, and then by normal results, relative to the screening-naïve group. On most outcomes, women diagnosed with breast cancer were naturally more likely to use a given service, yet on some outcomes, false positives were just as likely. Hospitalisation and surgery and non-surgery procedures showed no differences.
These results indicate that women with false-positive results from breast cancer screening use more healthcare services in the primary sector; in other words, contacts initiated by themselves. However, these trends were less persistent in secondary healthcare outcomes. This implies that women with false positives are likely to have their problems dealt with by their GP and other primary healthcare professionals. We did observe that women with false-positive results were more likely to receive diagnoses, which might be due to Berkson’s bias.23 We have not examined if these additional diagnoses are clinically relevant, related to breast cancer or are overdiagnosis. In the same way, women with false positives continuously had a higher rate of prescription drugs compared with the screening-naïve group. We have no clear explanation to why women with false positives have a higher prescription of drugs. It might be due to Berkson’s bias, where the women with false positives are more likely to receive diagnoses and thus treated for these. Thus, this could be considered a derived effect of increased visits to the GP due to screening result and potentially an indicator of a degree of overdiagnosis. Indirect evidence from low-dose CT screening for lung cancer suggest that people with false positives had a higher prescription of anxiolytics or antidepressants.24 We did not stratify type of drugs, due to the anticipated rarity of this outcome, and cannot conclude if certain classes are driving the differences between groups.
Women diagnosed with breast cancer are taken care of in secondary care due to their cancer, which is also reflected in our study by high rates of breast imaging and hospitalisation. This might explain why we observe the low rates of primary care outcomes such as GP contacts and drugs in this group, as these are dispensed in secondary care which are not registered in the available data. The 10-year convergence period observed for GP contacts (figure 1A) may indicate the extent of the secondary care embrace for these patients.
The increased use might be driven by long-term psychosocial consequences such as fear of disease.3 6 We observe an increase in primary healthcare services for women who do not have breast cancer (normal and false positives) which are at the discretion of the individual woman. This might indicate an increased healthcare-seeking behaviour induced by psychosocial consequences and medicalisation where women are concerned for their health and seek medical advice to confirm their well-being.6 7 These findings add to the evidence base of the long-term healthcare use increase following screening but also points to overlooked indirect costs of screening programmes. Even though differences in use were relatively small for most outcomes, these might have a large impact as about 80% of invited women participate in breast cancer screening,25 and 20%–60% of these will receive at least one false-positive result.1 Such effects should be included in future cost-effective analysis.
Comparison to literature
Prior to conducting this study, we did a systematic search and most economic and health use analyses of breast cancer screening were cost-effectiveness analyses and did not take the indirect or downstream costs into account.26–29 Other studies quantified the individual direct costs of receiving false positives.30–32 One study estimated the direct costs of diagnostics of mammography recall among commercially insured American women to US$1238.33 Studies on healthcare use also showed increased direct costs.10 11 These studies focus on the costs directly associated to spending time or money on follow-up diagnostics and not how the participation might impact future use of healthcare services. An Australian study estimated that the national total annual flow-on diagnostic costs for breast cancer population screening including costs of screen-detected abnormalities was $A73 876 961.34 Generally, costs are difficult to assess as they are subject to economic fluctuations and developments. Denmark is a welfare state with a publicly funded healthcare system providing free access to all citizens. Thus, costs of services are determined by the political prioritisation, capacity, demand and supply and can vary significantly over a short period of time. Thus, costs do not reflect the actual value of the service and interpretation is not very meaningful. We did therefore not report the specific price or costs. Further, billed charges do not represent costs in a broader sense, as people can have additional costs such as transportation, lost work, and so on.
Two studies demonstrate that women, following a false-positive breast cancer screening, increase their use of healthcare services in the first year after the test.4 5 Our study adds that this increase persist long term for primary healthcare services.
Strengths and limitations
A strength of this study is the use of the Danish National administrative registers. These have few missing data and are based on billed charges of healthcare services. However, the register for drugs only covers prescription drugs sold at pharmacies and not drugs administered at hospitals. This might explain the contraintuitively low drug use among women diagnosed with breast cancer. Prescriptions are a proxy for intake, yet we do not know if the prescriptions are consumed.
We conducted several analyses and had a relatively large sample size, which was needed to identify changes in the rarer outcomes.
People who participate in screening are more likely to have higher socioeconomic status.35–38 However, we did not observe this trend in our sample. We accounted for potential confounders but we did not adjust for ethnicity, which could also be a potential confounder.39 However, the participation rate in Denmark is high and is in most years higher than 80%,25 and we adjusted for country of origin as a proxy. Further, Denmark is a welfare state where most services are available without out-of-pocket costs, and ability to pay should not affect use in this setting.40
Implications for research and practice
These findings expand the evidence on the consequences of screening. There might be economic incentives to implement screenings such as screening for breast cancer; however, this study emphasises that breast cancer screening does not only carry direct costs through healthcare demand but has derived indirect costs through increased healthcare use that persists long term. This study can contribute to policy-making when weighing the benefits and harms of screening.
We did observe an increased use of primary healthcare services but not secondary healthcare services, which indicate that this might be due to psychosocial consequences leading to increased healthcare-seeking behaviour. Such a medicalisation can cause downstream harm such as work-related harms as women take time off work to attend medical services. Attending additional healthcare services might also cause physical harm such as side effects, psychological harm, for example, due to waiting time on test results, or further out-of-pocket expenditures.13
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors Conceptualisation and methodology: EGG, VS and JB. Software, validation and analysis: EGG, VS, DRN. Data curation: EGG and DRN. Writing - original draft: EGG. Writing - review and editing: VS, DRN and JB. Visualisation: EGG and VS. Supervision: VS and JB. Project administration, funding acquisition: EGG. EGG is the guarantor.
Funding This work was funded by Helsefonden grant number 22-B-0502. The funding body was not involved in any part of this work.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.