To introduce an ‘opt out’ referral pathway for smoking cessation in pregnancy and to compare different methods for identifying pregnant smokers in maternity care.
To introduce an ‘opt out’ referral pathway for smoking cessation in pregnancy and to compare different methods for identifying pregnant smokers in maternity care.
Pilot study that analysed routine data from maternity and smoking cessation services with biochemical validation of smoking status.
Dudley and South Birmingham, England.
A total of 3712 women who entered the referral pathway—1498 in Dudley and 2214 in South Birmingham.
Routine monitoring data on smoking at maternity booking, referral to smoking cessation services, number of women who set quit dates set and short-term (4-week) self-report smoking status. Comparison of self-report, carbon monoxide (CO)-validated and urinary cotinine-validated smoking status for a subsample (n = 1492) of women at maternity booking.
In Dudley 27% of women who entered the opt out referral pathway were identified as smokers following CO testing. Of those referred to the smoking cessation services, 19% reported stopping smoking at 4-week follow-up. In South Birmingham 17% were smokers at booking, with 5% of those referred recorded as non-smokers at 4 weeks. The number of women quitting did not increase during the study when compared with the previous year, despite higher referral rates in both areas. An optimum cut-off CO measurement of 4 parts per million (p.p.m.) was identified for sensitivity and specificity.
The introduction of an opt out referral pathway between maternity and stop smoking services resulted in more women being referred for support to quit but not higher numbers of quitters, suggesting that automatic referral may include women who are not motivated to stop and who may not engage with services. Routine carbon monoxide monitoring introduced as part of a referral pathway should involve a cut-off of 4 p.p.m. to identify smoking in pregnancy.
Reducing smoking in pregnancy is a health policy priority in the United Kingdom, as in many other developed countries . Maternal smoking in pregnancy causes substantial harm, increasing the risk of miscarriage, stillbirth, prematurity, low birth weight, perinatal morbidity and mortality, neonatal or sudden infant death, asthma [2-5], attention deficit hyperactive disorder, learning difficulties , obesity and diabetes .
While many women in the United Kingdom stop smoking before becoming pregnant or soon after, one in four women smoke for part of pregnancy and one in eight smoke throughout . Compared to women in advantaged circumstances, women in disadvantaged circumstances are four times more likely to smoke prior to pregnancy and are half as likely to quit in pregnancy; disadvantaged women are also more likely to resume smoking after birth .
However, there are interventions known to be effective in promoting cessation in pregnancy. The most recent Cochrane Review of smoking cessation interventions pooled results from 72 trials and concluded that cessation interventions in early pregnancy can reduce smoking in later pregnancy by approximately 6% and reduce the incidence of low birth weight and preterm births . In the United Kingdom, there is a wide range of information and support available to pregnant women who want to stop smoking. This ranges from self-help materials available online and from health professionals, to brief advice from general practitioners (GPs) or midwives, to specialist support offered by National Health Service (NHS) Stop Smoking Services (SSS). These services were developed as part of England's 1998 tobacco control strategy, informed by evidence from effective interventions and rolled out across the United Kingdom from 2000. The services provide behavioural support delivered by trained advisers along with access to smoking cessation medications for smokers motivated to quit. From 2001, services in all parts of the United Kingdom began to provide tailored support for pregnant women in clinic settings, in the home and by telephone. Results from national monitoring and recent research suggest that these interventions are successful in pregnancy, with between one-third and one-half of service clients stopping smoking, at least in the short term . However, fewer than one in seven pregnant smokers access these services ; the majority of those making quit attempts therefore do so without professional assistance, and most will not succeed .
Under-reporting of smoking status during pregnancy can contribute to low levels of uptake of support to stop smoking, as women who do not identify themselves as smokers or who are not asked about their smoking status may not be referred to services that can help them to quit. Previous studies have identified under-reporting as an issue in pregnancy . In response to this, and to improve rates of smoking cessation in pregnancy, the National Institute of Health and Clinical Excellence (NICE) in the United Kingdom issued guidance in 2010. This recommended that all pregnant women be asked about their smoking status at maternity booking, at regular intervals during their pregnancy and post-partum . It also stated that maternity services should use a simple form of biochemical validation of smoking status—a carbon monoxide (CO) breath test—to identify pregnant smokers and introduce ‘opt out’ referral pathways that would involve passing on the details of all identified smokers, without the need for direct consent for referral, to the local smoking cessation service. This recommendation represented a move away from the existing system that involved most (but not all) women being asked about their smoking status and a reliance on self-report to identify smokers. The previous system also involved smokers being asked if they wished to be referred to smoking cessation services, with many declining this referral.
This study aimed to examine whether an opt out referral pathway for smoking cessation in pregnancy changed the number of women referred for assistance in stopping smoking. We also compared different methods for identifying pregnant smokers in routine maternity care.
In two NHS areas in the West Midlands of England—Dudley and South Birmingham—we aimed to introduce referral pathways and investigate whether more women were referred and whether more quit once this process had been completed. Both areas had a higher prevalence of smoking in pregnancy than the national average, and presented two differing models of maternity care—one (Dudley) largely community-based and the other (South Birmingham) organized around a large women's hospital.
In Dudley the opt out referral pathway was implemented from August 2010 to March 2011. In South Birmingham the pathway was implemented from October 2010 to March 2011. In both sites the study involved the collection and analysis of routine data, in the form of acceptance of opt out referral to the local SSS, measured levels of CO, cotinine levels collected through routine urine sampling and follow-up of referred women by the local SSS.
The referral pathway differed slightly between the two sites; in Dudley the pathway was implemented by midwives and in South Birmingham by maternity assistants. The first stage of the opt out referral occurred in Dudley at the 6–8-week appointment at the booking appointment and in South Birmingham at the 12-week scan appointment. Almost all women attending the 12-week scan appointment in South Birmingham will have met with a midwife previously, who may or may not have discussed smoking. This timing followed current practice. The midwife or maternity assistant, respectively, asked the woman for her self-reported smoking status (current smoker, ex-smoker or never smoker). CO was then recorded. If a woman was a smoker, had a CO reading of above 4 parts per million (p.p.m.) or if she had recently quit smoking (since conception), then the midwife or maternity assistant told her that it was routine practice to refer to the local SSS and the referral was made, unless she refused (opt out). The second stage of the opt out referral pathway occurred between 21 and 24 weeks in Dudley and at 28 weeks in South Birmingham. When the woman had her appointment with the midwife or maternity assistant, respectively, she was asked again for her self-reported smoking status, a CO reading was recorded and, if appropriate, referral was made, unless she opted out.
An additional aspect of the pilot was to compare methods for identifying pregnant smokers in maternity care. A combination of asking women and using their reported smoking status with CO testing is the routine method used in England; other methods which could be used are the measurement of cotinine levels in the plasma (blood), urine or saliva. Cotinine is a metabolite of nicotine and is used as a biomarker for exposure to tobacco. After introduction of the referral pathway, the testing of pregnant women's urine for cotinine was implemented and this enabled comparison of self-report smoking status versus CO versus cotinine. Cotinine analysis was conducted for a 3-month period, instead of during the entire data collection period, due to resource constraints.
There were three sources of data for the study: at maternity booking; from the SSS; and biochemical samples that were requested at the same maternity booking appointment during which smoking status was ascertained initially. Box 1 outlines which data were collected for each of the three sources.
At maternity booking (6–8 weeks pregnancy in Dudley, 12 weeks in South Birmingham):
From National Health Service (NHS) Stop Smoking Services (SSS):
Biochemical samples (collected over 3-month period only):
Comparison of ‘smoker or NRT’ versus ‘self report smoker’ versus ‘CO smoker (>4 p.p.m.)’
The data were restricted to all cases with booking dates within the study period. Six cases (0.4%) in Dudley were missing the booking date, but a date was entered based on estimated date of delivery (birth of baby) and number of weeks pregnant. Two cases (0.1%) in South Birmingham were missing the booking date and further information was not available, so these cases were removed. All variables are reported as recorded in the data set, and presented in tabular form; any missing cases were omitted from totals and percentages. Analysis was carried out in the statistical programming package R version 2.12.1 .
Data describing the outcomes of implementing the referral process in terms of women referred and engaged into NHS SSS are reported as simple percentages. A descriptive comparison of the proportions obtained from the routine data set during the pilot period with the year before was then undertaken. We used an exhaled CO reading 4 p.p.m. as a cut-point above which women were considered to be smokers, irrespective of their self-reported smoking status. This was the cut-off used by services in both areas at the time of the study, although national guidance suggests that a cut-off of up to 7 p.p.m. can be used in pregnancy . For urinary cotinine, below 5 ng/ml was classified as a non-smoker, any value between 10 and 50 ng/ml indicated exposure to passive smoking and above 49.7 ng/ml indicated active smoking or nicotine replacement therapy (NRT) use. We present the proportion of women smokers identified by self-report, exhaled CO and urinary cotinine estimation and then also present the percentage difference in smokers identified when CO or urinary cotinine or both were used in conjunction with self-report.
In Dudley there were 2915 antenatal bookings during the data collection period and a subset of 1498 women were invited for CO testing. Only a subset were invited for a range of reasons, including miscarriage, and factors which meant that the pilot protocol was not followed by all midwives, discussed further below. Of these 1498 women, 73% reported or were CO-validated as being non-smokers. Of the 27% of women who were identified as smokers, 217 (54%) of these were referred to the SSS, 159 (39%) declined referral and an additional 24 (6%) and were already in contact with the service.
In South Birmingham, there were 3160 antenatal bookings during the data collection period. Of these, 2214 women entered the pathway and had a CO test administered. A higher proportion of eligible women in this study site entered the pathway. Among the women entering the pathway in South Birmingham, 83% were identified as non-smokers. Of the remaining 17%, 79 (21%) were referred to the SSS, 153 (41%) declined referral and an additional 135 (36%) were already in contact with the service. A higher proportion of women were already in contact with SSS in South Birmingham, which may have been because the first appointment occurred slightly later in pregnancy, and therefore some smokers may have already been referred by midwives or their GP.
Overall, therefore, a total of 3712 entered the referral pathway when the sample from both study sites was combined. For Dudley, 212 women were referred, 146 of whom were then contacted successfully by the SSS (69% of those referred), 96 had a first appointment (just under half of those referred), 62 set a quit date and 40 had quit successfully after 4 weeks (19% of those referred or 64% of those who set a quit date) (Fig. 1). Note that these 212 were based on SSS data rather than data taken from the maternity appointment, and so includes only some of those 217 identified in the previous section as being referred, plus some additional women not referred at the antenatal appointment.
In South Birmingham, of the women referred (n = 112) or already in contact with the SSS, all were recorded as having been followed-up by the SSS but, of these women, just 33 (29%) had a first visit, 18 (16%) set a quit date and 11 had quit by 4 weeks (61% of those who set a quit date, and just 5% of those referred).
In both study sites, the introduction of the new pathway resulted in an increase in the number of pregnant smokers referred to the local SSS when compared with the previous year. As Table 1 shows, in the 3 months when the pilot was implemented fully in both study sites (quarter 4), referrals increased by 34% in Dudley when compared with the same quarter the year before, and 20% in South Birmingham.
|Quarter||Site||2009/10||2010/11||% change from 2009/10 to 2010/11|
|Q3 (October–December)||South Birmingham||117||169||44.4|
|Q4 (January–March)||South Birmingham||169||203||20.1|
However, as Table 2 shows, the number of women quitting did not increase during the study period when compared with the previous year, despite higher referral rates to both local stop smoking services.
|n||% of those referred||n||% of those referred|
|Q3 (October–December)||South Birmingham||17||14.5||13||7.7|
|Q4 (January–March)||South Birmingham||17||10.1||9||4.4|
Six hundred and one women in Dudley and 882 women in South Birmingham provided self-report, CO and urinary cotinine measures of smoking status (1492 women in total). In Dudley, 21% of women were classified as smokers by self-reported smoking, 18% were classified as smokers by CO reading and 26% by cotinine. In South Birmingham, 14% were classified as smokers by self-reported smoking, 15% by CO readings and 14% by cotinine. Figure 2 shows a Venn diagram for the three measures of smoking, which illustrates the extent to which the measures agree; so, for example, 303 women are identified as smokers by at least one method; 198 of these are identified by all three, with an additional 46 by both self-report and cotinine sample. Thus, the majority of disagreements between the measures are where women are defined as smokers based on their cotinine samples, but non-smokers according to their CO measurement. This implies that the CO test may not be picking up some women who are smokers.
We explored the effect of varying the CO cut-off value, and whether an alternative value would provide a better indication of smoking status. For this analysis, we combined all women from Dudley and South Birmingham who had measurements for CO, self-reported status and cotinine; there were 1492 women in total.
Figure 3 shows the receiver operating characteristic (ROC) curve for CO value based on the combined sample of 1492 women from both sites. The ‘gold standard’ used is a combination of self-reported status and cotinine: all those identified as smokers via cotinine analysis, plus anyone else who self-identified as a current smoker. This is used because it is unlikely that women would self-report as a smoker when they are not. The ROC curve shows the trade-off between sensitivity and specificity for different choices of cut-points for the CO measurement; the closer the curve to the top left corner, the higher the accuracy. The dot marks the existing cut-off of 4 p.p.m.; this has a sensitivity of 0.79 and a specificity of 0.94. The optimum cut-off is chosen to maximize both sensitivity and specificity; that is, the point closest to the top left corner. It is clear from the figure that the 4 p.p.m. cut-off is optimum for these data. Note that this is lower than the usual cut-off figure for CO measurements of 7 p.p.m. recommended for referrals in national (NICE) guidance in England. A cut-off of 7 p.p.m. for these data has a sensitivity of 0.60 and a specificity of 0.99; the lower sensitivity indicates that fewer smokers are identified correctly when this cut-off is used.
This pragmatic pilot found that it is possible to implement routine CO monitoring of pregnant women using maternity services and that opt out referral pathways can be put into place between maternity care and services to support women to stop smoking. The pilot found that implementation of these pathways resulted in more women being referred to smoking cessation services.
However, the configuration of antenatal services and existing referral procedures had a significant impact on the extent to which the pathway was fully implemented. In both sites not all eligible women entered the pathway, sometimes for legitimate reasons, such as women miscarrying between their booking visit and data entry or referral; yet service factors had the biggest impact on who was referred. The pilot was designed to change the existing system of reliance on self-report of smoking status and smokers then agreeing explicitly to a referral to smoking cessation services. The researchers worked with maternity services in both sites to train midwives and midwifery assistants to change their practice (through face-to-face briefings and documents about the pilot), so that routine CO testing was introduced for all and automatic referral introduced for identified smokers. However, as the results show, more eligible women entered the pilot pathway in the South Birmingham study site than in Dudley. This was due at least in part to the fact that some midwives in Dudley were unable or unwilling to change practice to include routine CO monitoring, a point corroborated by findings from interviews with staff involved in the pilot, reported elsewhere . In contrast, in South Birmingham this task was devolved to midwifery assistants and as a result most eligible women in this site received a CO test. Other research has found that adding additional activities to the already busy booking visit with midwives may be challenging, and that midwifery assistants can play a valuable role in delivering screening tests, including routine CO testing . There may also be alternative explanations for the difference in uptake between the two sites, such as factors relating to the population served and organizational issues in maternity services that the study team could not influence.
Despite the opt out referral system resulting in increased referrals to the SSS in both study sites, this did not result in more pregnant women quitting. Although our comparisons were limited to the equivalent period the year before and the numbers were not sufficient to detect significant differences in quit rates, this was unexpected. As SSS deliver interventions that can increase the likelihood that a woman will stop smoking in pregnancy, such as behavioural support to quit , it was anticipated that increased referrals would result in more women using the services and more quitting. However, the findings suggest that higher referrals did not necessarily translate to more women attending an appointment with a smoking in pregnancy specialist or going on to set a quit date. This may be because the opt out system resulted in women who were less motivated to quit being referred, and these women may have refused to engage with services when attempts were made to contact them. Alternatively, the smoking cessation services may not have had sufficient capacity to try repeatedly to reach women who were referred but did not respond to an invitation for an appointment, or may not have tried to re-contact women who made an appointment but did not attend. This suggests that if opt out referral pathways are to be introduced into routine practice, information to motivate women to think about smoking cessation should also be provided and local services need to have the capacity to respond to increased client numbers.
The pilot also compared three measures of smoking status—self-report, CO-validated and cotinine-validated smoking status. Overall, the three measures produced more similar estimates than those found in other studies, where a significant discrepancy has been noted between self-reported smoking status in pregnancy and either cotinine or CO [12, 17]. This may suggest that women were more willing to report their smoking status accurately in this study, or simply that the manner in which the question was asked by midwives and midwifery assistants ensured that women felt comfortable in indicating whether or not they were a smoker. Comparison of these three measures also allowed assessment of the optimum cut-off for CO monitoring in pregnancy. In most studies with the general adult population a cut-off of 10 p.p.m. is used . There is general acknowledgement in the literature and from CO monitor manufacturers that a lower cut-off in pregnancy is required , although this is often identified as 7 p.p.m. Our findings suggest that 4 p.p.m. is the optimum cut-off and should be used to validate the smoking status of pregnant women.
This study faced a number of limitations. The first is that it was conducted in two areas of England with maternity SSS that may not be representative of other parts of the United Kingdom or in other countries. We did, however, select two pilot areas that had contrasting models of service delivery, with one providing maternity booking and some follow-up services almost entirely in the community, with the other including services that were organized primarily around one large hospital. This variation should provide some assurance of the relevance of the pilot for other areas.
In addition, the new method of identifying and referring smokers was not implemented consistently in both areas, with particular problems with take-up in Dudley, as outlined above. This means that not all eligible women were invited to participate in the study, and therefore we cannot be entirely confident that those included were representative of all pregnant smokers in the pilot sites at the time of the study. This is both a limitation and a key learning point from this study, suggesting that future introduction of opt out referral pathways needs careful preparation and concerted efforts to secure commitment from all relevant staff before implementation.
An additional potential limitation relates to assessing smoking status. The fact that we did not identify much difference between self-reported smoking status and biochemically validated smoking status may be related in part to the fact that women may have been told that they were going to have a CO test before being asked about their smoking status. This may have encouraged more truthful disclosure than some other studies have found. In addition, we were not able to introduce a question about NRT use into the booking visit, and this may have confounded the cotinine results for a small number of women.
Finally, comparisons between outcomes achieved before and after the pilot period were limited. This was due to the fact that local services collected limited information routinely about throughput and quit rates. Our comparison was limited to the equivalent months in the year before the study and was limited to short-term (4 weeks), self-report quit rates. Future studies should explore in more detail, with more comprehensive and biochemically validated data, what the impact of the introduction of opt out referral pathways can be on supporting more women to stop smoking during pregnancy.
This research was carried out as part of a programme of inequalities projects implemented by the UK Centre for Tobacco Control Studies and funded by the Department of Health. The views expressed are those of the author and not necessarily those of the Department of Health. L.B., L.H., T.C. and J.F. are members of the UK Centre for Tobacco Control Studies (UKCTCS), a UK Centre for Public Health Excellence.
The authors would like to thank the members of the project advisory group who provided valuable input into the design and delivery of the project. Members were Professor David Tappin, Professor Marcus Munafo, Dr Andy McEwen, Professor Ann McNeill, Yvonne Hermon, Ruth Olding, Heather Gray, Elizabeth Punter, Jenny Henry, Jane Beech and Anne Fitchett. Additional thanks to ABS Laboratories for the processing and interpretation of the cotinine samples and to Carol Anne Greenan for her assistance with producing the paper. Funding from the British Heart Foundation, Cancer Research UK, the Economic and Social Research Council, the Medical Research Council and the National Institute of Health Research, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged.