teenage pregnancy—new tools to support local health campaigns

8
Teenage pregnancyNew tools to support local health campaigns Jakob Petersen a,b , Philip Atkinson b , Sarah Petrie b , Maurizio Gibin a,b , David Ashby c , Paul Longley a, a Centre for Advanced Spatial Analysis, Department of Geography, University College London, Gower Street, London WC1E 6BT, UK b Southwark Primary Care Trust (NHS), Public Health, Woodmill, Neckinger, London SE16 3QN, UK c Dr Foster (Research) Ltd., 12 Smithfield Street, London EC1A 9LA, UK article info Article history: Received 26 November 2007 Received in revised form 23 May 2008 Accepted 15 June 2008 Keywords: Teenage pregnancy Conceptions Targeting GIS Recruitment flows abstract Teenage pregnancy has remained high in many inner city areas despite several years of campaigns to reduce numbers and to support young people and their families tackle the problem. In this paper we propose new methods to focus local strategies on high-risk areas as well as ranking secondary schools and GP practices most likely to be in contact with young people at risk. The proposed methods proved successful in engaging local schools in a new campaign and have provided a framework for evaluation of local teenage pregnancy rates in years to come. & 2008 Elsevier Ltd. All rights reserved. Overview and problem definition Teenage pregnancy rates in UK are amongst the highest in Europe (Social Exclusion Unit, 1999). The Government’s Teenage Pregnancy Strategy is designed to address this, aiming to reduce the number of teenage pregnancies and support teenagers more effectively (Social Exclusion Unit, 1999). There have been some signs of improvement; in the period 1998–2004 under-18 conception rates fell nationally by 11.6% from 47 to 42 conceptions per 1000 amongst the 15–17 year-old cohort (Office for National Statistics, 2006). However, figures remained stubbornly high in many inner city deprived areas and the London Borough of Southwark (approx. 250,000 population) remains an area with one of the highest teenage pregnancy rates in UK. Here, the figures fell marginally from 87 to 85 conceptions per 1000, or 2.3% of the age cohort, over the same period, and it is this Borough which is the focus of our study. The Government’s Teenage Pregnancy strategy strongly em- phasises the need for a targeted approach (Social Exclusion Unit, 1999). However, local stakeholders and strategies are currently informed only by conception rates released by the Office for National Statistics (ONS) at Local Authority (average population of 300,000) or Census ward level (average population of 5000). While this may be helpful in revealing a Local Authority such as Southwark to be one of the areas with the highest incidence in the UK, it does not present sufficiently detailed information to inform a targeted strategy at the local level. Ward level information can be said to represent local variation for many policy purposes, but because district boundaries are rarely coterminous with the principal catchment areas, e.g. for schools or GP practices, they have proven less adequate for strategies targeting those organisa- tions (Gibin et al., 2007; Glennerster, 1991). At present Primary Care Trusts and their local partner organisations 1 are finding it difficult to target the most appropriate local areas and organisa- tions in their campaigns. The National Health Service (NHS) is funded by the UK government and as such is central to the political debate not only about taxation, but also in questions of its share of the state budget in competition with other public sectors like education, policing, social security, defence, etc. (Hsiao and Heller, 2007). Within the health care system itself, there are complicated trade- offs between equity and efficiency objectives with respect to improving population health, reducing risks and inequalities, and a need to ration services balanced with maintaining a certain level of user satisfaction (Musgrove, 2003). There is consequently increasing pressure to reform health care systems and the NHS is undergoing reforms to make its organisation more cost-effective and to attract private enterprise (Pollock et al., 2007; Talbot-Smith and Pollock, 2006). More resources are now directed towards local health authorities with emphasis on primary and community ARTICLE IN PRESS Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/healthplace Health & Place 1353-8292/$ - see front matter & 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.healthplace.2008.06.003 Corresponding author. Tel.: +44 207679 0500. E-mail address: [email protected] (P. Longley). 1 Local schools, colleges, youth clubs, GP practices, sexual health clinics and a broad range of community groups and non-government organisations. Health & Place 15 (2009) 300–307

Upload: jakob-petersen

Post on 05-Sep-2016

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

Health & Place 15 (2009) 300–307

Contents lists available at ScienceDirect

Health & Place

1353-82

doi:10.1

� Corr

E-m

journal homepage: www.elsevier.com/locate/healthplace

Teenage pregnancy—New tools to support local health campaigns

Jakob Petersen a,b, Philip Atkinson b, Sarah Petrie b, Maurizio Gibin a,b, David Ashby c, Paul Longley a,�

a Centre for Advanced Spatial Analysis, Department of Geography, University College London, Gower Street, London WC1E 6BT, UKb Southwark Primary Care Trust (NHS), Public Health, Woodmill, Neckinger, London SE16 3QN, UKc Dr Foster (Research) Ltd., 12 Smithfield Street, London EC1A 9LA, UK

a r t i c l e i n f o

Article history:

Received 26 November 2007

Received in revised form

23 May 2008

Accepted 15 June 2008

Keywords:

Teenage pregnancy

Conceptions

Targeting

GIS

Recruitment flows

92/$ - see front matter & 2008 Elsevier Ltd. A

016/j.healthplace.2008.06.003

esponding author. Tel.: +44 207 679 0500.

ail address: [email protected] (P. Longl

a b s t r a c t

Teenage pregnancy has remained high in many inner city areas despite several years of campaigns to

reduce numbers and to support young people and their families tackle the problem. In this paper we

propose new methods to focus local strategies on high-risk areas as well as ranking secondary schools

and GP practices most likely to be in contact with young people at risk. The proposed methods proved

successful in engaging local schools in a new campaign and have provided a framework for evaluation of

local teenage pregnancy rates in years to come.

& 2008 Elsevier Ltd. All rights reserved.

Overview and problem definition

Teenage pregnancy rates in UK are amongst the highest inEurope (Social Exclusion Unit, 1999). The Government’s TeenagePregnancy Strategy is designed to address this, aiming to reducethe number of teenage pregnancies and support teenagers moreeffectively (Social Exclusion Unit, 1999). There have been somesigns of improvement; in the period 1998–2004 under-18conception rates fell nationally by 11.6% from 47 to 42 conceptionsper 1000 amongst the 15–17 year-old cohort (Office for NationalStatistics, 2006). However, figures remained stubbornly high inmany inner city deprived areas and the London Borough ofSouthwark (approx. 250,000 population) remains an area withone of the highest teenage pregnancy rates in UK. Here, the figuresfell marginally from 87 to 85 conceptions per 1000, or 2.3% of theage cohort, over the same period, and it is this Borough which isthe focus of our study.

The Government’s Teenage Pregnancy strategy strongly em-phasises the need for a targeted approach (Social Exclusion Unit,1999). However, local stakeholders and strategies are currentlyinformed only by conception rates released by the Office forNational Statistics (ONS) at Local Authority (average population of300,000) or Census ward level (average population of 5000).While this may be helpful in revealing a Local Authority such asSouthwark to be one of the areas with the highest incidence in the

ll rights reserved.

ey).

UK, it does not present sufficiently detailed information to informa targeted strategy at the local level. Ward level information canbe said to represent local variation for many policy purposes, butbecause district boundaries are rarely coterminous with theprincipal catchment areas, e.g. for schools or GP practices, theyhave proven less adequate for strategies targeting those organisa-tions (Gibin et al., 2007; Glennerster, 1991). At present PrimaryCare Trusts and their local partner organisations1 are finding itdifficult to target the most appropriate local areas and organisa-tions in their campaigns.

The National Health Service (NHS) is funded by the UKgovernment and as such is central to the political debate notonly about taxation, but also in questions of its share of the statebudget in competition with other public sectors like education,policing, social security, defence, etc. (Hsiao and Heller, 2007).Within the health care system itself, there are complicated trade-offs between equity and efficiency objectives with respect toimproving population health, reducing risks and inequalities, anda need to ration services balanced with maintaining a certain levelof user satisfaction (Musgrove, 2003). There is consequentlyincreasing pressure to reform health care systems and the NHS isundergoing reforms to make its organisation more cost-effectiveand to attract private enterprise (Pollock et al., 2007; Talbot-Smithand Pollock, 2006). More resources are now directed towards localhealth authorities with emphasis on primary and community

1 Local schools, colleges, youth clubs, GP practices, sexual health clinics and a

broad range of community groups and non-government organisations.

Page 2: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

J. Petersen et al. / Health & Place 15 (2009) 300–307 301

health care. Additionally, GP practices are rewarded for matchingservices to local needs in their recently renewed contracts withthe Department of Health (Department of Health, 2006). The urgefor public health decisions to be based on evidence andprovide value-for-money is a ubiquitous demand and central tothe health care reforms (Department of Health, 2000, 2006;Muir Gray, 1996). While it seems clear that economic gains shouldaccrue from different forms of optimisation and resource target-ing in sexual health campaigns, this does also raise some criticalissues concerning individual rights to privacy and data protection.

Here we propose a new approach for a targeted strategy at alocal level, using fine-scale conception data from both localhospitals and private abortion clinics. We describe and demon-strate some simple methods that make it possible to estimaterisks for areas, schools and GP practices, while remainingcognisant of confidentiality strictures and the practical require-ments of ethical approval procedures. Local health authoritieswill, using the proposed methods, be able to target their limitedresources more effectively, based on local evidence and in linewith the sector reforms described above. In what follows, wepropose a method for ascribing information on unwantedpregnancies to localities using visual communication that isreadily intelligible to a range of stakeholders. As such, we see ourmethod as providing a valuable ‘first filter’ to preventive healthcare campaigns, that is applicable in the widest range of settingsand which can provide a framework to specific locality studies.The findings are discussed with regard to geographical targeting,data confidentiality, multi-agency strategies and resource alloca-tion in a local health care setting. We do not see this as a completesolution to targeting initiatives, but rather as a contribution to themore effective use of routinely collected local data as a kind of‘first filter’ for locality studies of health care needs. As such, wethink we offer an important contribution to increasing costeffectiveness and improving intersectional collaboration withoutinfringing the privacy rights of any individual or storing sensitivedata material on any identifiable individuals.

Methods

The study area

Southwark Primary Care Trust is situated on the south bank ofRiver Thames in Central London. Historically Southwark wascentrally located for London’s port and associated industries.Today Southwark is among the most deprived local authorities inEngland; ranking 18th out of 325 local authorities on the incomedeprivation and 25th on the employment deprivation scale (Officefor National Statistics, 2008). The majority of patients (65%) live inareas ranked as the 20% most deprived areas in England. Morethan 50% of patients live in publicly rented accommodation;typically social housing apartment blocks built in the 1960s onWorld War II bomb sites across the northern half of the Borough.Another 28% live in pre-war terraced housing of which some ispublicly owned. Southwark has a very fluid, diverse and multi-cultural population with a large African and Caribbean commu-nities. Southwark scores low on a number of health indicatorsincluding infant mortality, low birth weight, low male lifeexpectancy at birth, low disability-free life expectancy and highteenage conception rates.

Data analysis and mapping

In order to address the problems of disaggregating officialrecords in a way that could inform local strategies to reduce

teenage pregnancy, Southwark Primary Care Trust collecteddetailed teenage conception data including all recorded maternityand legal abortions from all NHS service providers and privateproviders commissioned by the trust during the period2002–2005 (Butt et al., 2006). This data assembly exerciseidentified 885 conceptions locally coded with residential unitpostcodes, which equates to 86% of the conceptions reported bythe Office for National Statistics (ONS) over the same period. It isdifficult to identify the exact causes of the apparent shortfall, butone possibility is likely to be users paying a private providerthemselves. In these cases ONS would be notified, but no claim forpayment would be made to the users’ PCT. Under-18 yearsdenominators for unit postcodes were derived from the Mosaic(Experian Ltd., Nottingham, UK) directory. Other population dataobtained included anonymous postcoded records of resident11–17 year-old girls attending either a state secondary school orregistered with a GP in the trust. The former were obtained withthe consent of the Local Educational Authority. Ethical approval toanalyse the data sets containing patients’ postcodes was obtainedin accordance with NHS ethical guidelines from the Local ResearchEthical Committee (Department of Health, 2003). Data analysiswas carried out using Stata 9.2 (StataCorp, College Station, USA),ArcGIS 9.1 (ESRI, Redlands, USA) with the Hawth’s Tool extension(Beyer, 2004), and the DCluster disease cluster package for R(Gomez-Rubio et al., 2005).

Results

Persistent inequalities associated with place have causesrooted in a matrix of compositional and contextual factors, withpotentially self-enforcing ties between the two (Cummins et al.,2007). For purposes of local policy, identification of spacesexacerbating ‘risk’ is therefore important. However, the relationalview of place in health inequality also leads us to think about howthe knowledge of flows from residence to schools or GP practicescould be utilised to inform local strategies. For purposes of ourempirical analysis, we therefore approach ‘risk’ using threedifferent indicators: (a) areas with the highest densities of teenageconceptions; (b) areas with the highest densities adjusted for theunderlying population; and (c) school or GP practices whererisk is particularly high. By the term ‘risk’ we here take afrequentist approach (Waller and Gotway, 2004, p. 9), where riskis evaluated numerically as the number of cases relative to eitherspace (as in (a) above), population aged under 18 years (as in (b)above) or gender-specific school populations (as in (c)). In eachcase mapping and data analysis can be used to inform publichealth campaigns about the most appropriate areas, populationsand organisations to engage in a range of targeted strategies—

including the location of advertising boards, routing of the publichealth information bus and the involvement of GP practices,schools and youth clubs.

Teenage conception risk mapping

A rich literature exists on the visualisation of spatiallyreferenced incidence data, including methods to adjust local ratesfor variations in the underlying populations (Cressie, 1991; deSmith et al., 2008; Haining, 2003; Waller and Gotway, 2004). Oneof the most basic methods of exploring spatial incidence data is2D kernel density estimation (Atkinson and Unwin, 2002; Silver-man, 1986). This produces an unadjusted risk mapping that makesno assumption about the underlying population. It does, however,tell us something about the geographical concentration of teenageconceptions that in our case coincides with the locations of thebiggest social housing estates in Southwark (Fig. 1: an estimated

Page 3: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

J. Petersen et al. / Health & Place 15 (2009) 300–307302

58% of under-18 year-olds in Southwark reside in social housing).To retain the information about the precise geographical locationof the conceptions, while at the same time maintaining con-fidentiality in the visual display, we have used percent volumecontours (PVC). A PVC is the isoline which bounds a givenpercentage of the phenomenon in question. In this case we haveadded the 50% isoline to the conception density surface. Compar-ing this to the 50% isoline of the under-18 year-olds populationreveals a generalised indication of where conceptions are under-or over-represented. Fig. 1 shows that an area to the south east ofthe main clusters appears to have more conceptions than wouldbe expected from the underlying under-18 year-old populationfigures.

In order to evaluate the apparent concentration of conceptions,we have further deployed a cluster detecting algorithm, theGeographical Analysis Machine or GAM (Gomez-Rubio et al.,2005; Openshaw et al., 1987). Using this method, both thenumbers of cases and denominators are used to identify localrates within a specified search radius of the nodes of a fine gridoverlain onto a base map. The resulting clusters are thus adjustedfor the variation in the underlying population of under-18 yearolds. In this case the results confirmed the main ward levelclusters evident from the official ONS data, but furthermorehighlighted several smaller clusters that would have been over-looked by the ONS mapping (see Fig. 2). This shortcoming arisesbecause of the scale effects of aggregating data to population unitsas large as Census wards, and provides an instance of what is alsoknown as the ecological fallacy (Openshaw, 1991). One muchvaunted claim of the private sector geodemographics industry(Sleight, 2004) is that recourse to finer levels of granularity suchas unit postcodes can reveal heterogeneity that is concealed atward scales of analysis. We concur with this general view but notnecessarily that private sector geodemographic systems presentthe best solutions to public sector problems such as healthprofiling (Longley, 2005; Petersen et al., 2008).

Kernel density estimation, PVCs and GAM are all usefulexploratory tools for summarising high-resolution spatial datawithout compromising data confidentiality of human subjects.They both can help to identify general or coarse patterning, andGAM can be used to highlight more localised clusters that mightbe missed in maps constrained to official administrative areas,whilst maintaining confidentiality.

Risk estimates for schools and GP practices

Clusters based on residence alone may not necessarily be themost effective for targeted messages or strategies aiming tosupport young people. A different strategy is to use the mostappropriate organisations in order to target individuals. Second-ary schools and GP practices are often used in teenage pregnancyprevention campaigns and activities. This raises a resourceallocation/optimisation problem: which schools and GP practicesshould be targeted in an evidence-based local strategy? In order toaddress this problem we have taken the postcode level conceptiondataset (unit postcode) and ‘allocated’ each conception to schoolsand GP practices, respectively, based upon the underlyingrecruitment flow between residential postcodes in Southwarkand secondary schools and GP practices. By recruitment flows wemean actual at-risk counts per postcode of individuals attendingspecific local schools and GP practices, calculated by

ci ¼X

cj

rijPrj

(1)

where c is conceptions and r is recruits (e.g. girls attending aparticular school), i is the service unit in question (e.g. a school ora GP practice), and j is the residential unit postcode. If for example

three conceptions, c, occurred in a given postcode, j, and 10 girls ofsecondary school age (11–17 year olds) are recruited to school A,rAj, and 20 girls to school B, rBj, then one conception is attributedto school A and two to school B, and so forth for all conceptionsacross all postcodes. The relative risk, RR, was estimated asfollows:

RRi ¼ci=P

c

ri=P

r100 (2)

In this way it was possible to rank and estimate relative risk for allsecondary schools and GP practices (see Fig. 3, Tables 1 and 2).For public health purposes, the numerators (i.e. conceptions) maybe just as important as relative risk measures (i.e. conceptions pergirl-at-risk). We have therefore provided both. In order to addressat least 50% of all relevant conceptions, the results show that acampaign would need to include the seven schools or 18 GPpractices with the highest conceptions figures (see Fig. 4).

Discussion

National and local strategies to reduce teenage pregnanciesfollow a multi-agency and multi-modal design incorporatingschool-based sexual health programmes, community-based edu-cation, contraceptive services, youth development and familyoutreach programmes (Social Exclusion Unit, 1999). This approachhas been endorsed by recent reviews (Bennett and Assefi, 2005;Swann et al., 2003). However, the lack of positive policy outcomesin inner city areas like Southwark suggests a need for a renewedand more focused approach. Where should a local evidence-basedstrategy focus? Which settings are the most effective fordelivering messages about sexual health to young people? Howmight we identify these settings? The mapping of risk across theSouthwark PCT area has highlighted how much teenage concep-tions vary locally. This in itself emphasises the need for a targetedapproach. Our Southwark hotspot maps (Figs. 3 and 4) have theadvantage of being intuitive and easy for public health staff to usein the planning of campaigns, as for example with the routing of acampaign bus with sexual health information. Another advantageof these maps is that they generalise locational information in away that protects the confidentiality of any individual in thedataset.

Selecting hotspot areas is not, however, necessarily the mosteffective way of focusing a local strategy. As pointed out by someauthors, hotspot mapping of apparent sexual health is a sensitiveissue, and may have the negative outcome of adding tostigmatisation of neighbourhoods rather than achieving theintended end of empowering of young people and their families(Arai, 2007). This is clearly one of the dilemmas that face publichealth departments: to intervene without undermining the trustof participants or aggravating their circumstances in an indirectway by identifying their neighbourhood as a ‘bad’ place to live.Arai (2007) also advocate a ‘normality by locality’ view for a moregeneral acceptance of teenage motherhood.

The counter-arguments are that leaving teenage pregnancy asa private matter is a way of respecting the autonomy ofindividuals rather than creating autonomy. Viewed from thisperspective, anti-teenage pregnancy campaigns can be thought ofnot a response from a perfectionist and paternalistic State torestrict personal freedom, but rather as a way of putting personalfreedom in a longer perspective that involves greater choice ofeducation, family and lifestyle (Holland, 2007). In this context,for example, it is important to note that those undergoingpregnancy in their teens are, as a consequence, more likely todrop out of school and experience diminished life chances as aconsequence (Barnet et al., 2004; Social Exclusion Unit, 1999,

Page 4: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

Fig. 1. Percent volume contours of teenage conceptions versus under-18 year-old population. The bandwidth for kernel density estimation was 1000 m. The isolines show

the top 50% percent volume contour.

J. Petersen et al. / Health & Place 15 (2009) 300–307 303

Page 5: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

Fig. 2. Teenage pregnancy hotspot map showing ONS-released rates (2001–2003) and GAM hotspots for the local disaggregated data set. GAM search radius: 250 m, cell

size: 50 m.

J. Petersen et al. / Health & Place 15 (2009) 300–307304

Page 6: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

J. Petersen et al. / Health & Place 15 (2009) 300–307 305

2006; Swann et al., 2003). Our view is that PCTs should providethe type of services and support that young people would likecombined with a greater understanding of peer group influences,and the findings of some research supports this agenda (Crosnoeand McNeely, 2008; DiCenso et al., 2001; Pearson et al., 2006).Consultations of young people about their experiences withsexual health education, for example, have led to a change inthe emphasis of interventions from anatomy and scare-tactics tothe advocacy of negotiation skills in sexual relationships andbetter contraceptive services. There have also been calls for wideradvertising of local services, for example in shopping centres andother public places (DiCenso et al., 2001). This does, however,suggest a role for hotspot mapping when locating many essentialservices.

Teenage pregnancy rates are amongst the new social targetsthat have been suggested for school performance auditing(Department of Children Schools and Families, 2007; Lipsett,2008). If we look more broadly at health-related issues (aggression,alcohol and drug use) in adolescence, then there are evidence to

c

School B

School A

rAj

rBj

cB

cA

Fig. 3. Recruitment flows and service point estimation (see text for explanation).

Table 1Teenage conceptions estimated for (anonymised) state secondary schools (SS)

School Girls-at-

risk (n)

Percentage of all

girls-at-risk among

SS (%)

Estimated

conceptions (n)

Percentag

conceptio

SS (%)

SS1 769 16 103 16

SS2 580 12 85 13

SS3 643 13 80 12

SS4 360 7 62 10

SS5 317 7 48 7

SS6 487 10 45 7

SS7 358 7 44 7

SS8 184 4 35 5

SS9 281 6 32 5

SS10 187 4 27 4

SS11 160 3 26 4

SS12 181 4 22 3

SS13 135 3 20 3

SS14 225 5 19 3

No School – – 238 –

Total 4867 100 885 100

A total of 53% of conceptions could be reached by targeting a minimum of seven secon

suggest that schools do play a role above the potentiallyconfounding factors studied such as prior health status, parentaland neighbourhood effects (West et al., 2004) and some wouldlike schools to work specifically with improving their ethos (Bonellet al., 2007). Others point to peer influences—negative aspositive—within the school classrooms and not just school unitsper se (Johansen et al., 2006). There also seems to be a lot yet to begained from studying social network in relation to teenagepregnancy and effectiveness of interventions using such insight(Crosnoe and McNeely, 2008; Pearson et al., 2006).

In this case study, the need for directed action was importantfor the local strategy implementation. We considered bothsecondary schools and GP practices as appropriate agents. In thiswe moved from simple hotspot mapping to selecting theorganisations most likely to be in the closest contact with thetarget group. Dealing with these organisations compared tohotspots had several advantages. They represent well-establishednetworks of institutions with professional staff in daily contactwith young people. This makes interventions easier and morecost-effective to deliver. In our case the estimation of conceptionsfor each school helped significantly to engage local schools in newsexual health programmes. By contrast, the official Local Authorityand ward level figures made it much harder to provide theevidence about conception rates and thus also more difficult toengage school partners.

Although being the most appropriate organisations for localpolicy implementation there are also indications that this is anarea that needs improvement not only in the targeting, but also inthe substance of the programmes. School-based programmespromoting contraception resulted in higher uptake of contra-ception relative to abstinence-only programmes (Bennett andAssefi, 2005). However, a systematic review of school-basedteenage pregnancy prevention programmes reported no overallsuccess in reduction of pregnancies (DiCenso et al., 2002).

Compared to GP figures there were a higher proportion of casesin which we had to assume that the individuals were not registeredwith a local school (27% vs. 15%: see Tables 1 and 2). This could beinterpreted as an association between school dropout rates andteenage pregnancy occurrences either as an effect (suggested incohort study by Barnet et al., 2004) or a cause. We must stresshowever that, although these are interesting hypotheses, they arebeyond the limit of the study design used here.

e of est.

ns among

Relative risk index

among SS

Estimated

conceptions all

areas (%)

Estimated

conceptions all

areas (Acc. %)

101 12 12

111 10 21

93 9 30

129 7 37

113 5 43

70 5 48

93 5 53

144 4 57

84 4 60

109 3 63

120 3 66

91 2 69

111 2 71

63 2 73

– 27 100

100 100 –

dary schools.

Page 7: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

Table 2Teenage conceptions estimated for (anonymised) GP practices

GP practice

code

Girls-at-risk.

Local

residents (n)

Percentage of all

girls-at-risk among

GP (%)

Estimated

conceptions (n)

Percentage of

conceptions est.

among GP (%)

Relative risk index

among GP

Estimated

conceptions of total

(%)

Estimated

conceptions (Acc.

%)

GP1 304 13 89 12 92 10 10

GP2 84 4 35 5 130 4 14

GP3 102 4 26 3 78 3 17

GP4 114 5 25 3 69 3 20

GP5 88 4 24 3 86 3 23

GP6 84 4 23 3 86 3 25

GP7 55 2 22 3 122 2 28

GP8 54 2 21 3 121 2 30

GP9 69 3 20 3 91 2 32

GP10 80 3 20 3 78 2 35

GP11 51 2 20 3 121 2 37

GP12 71 3 19 3 84 2 39

GP13 33 1 18 2 168 2 41

GP14 37 2 17 2 145 2 43

GP15 57 2 17 2 93 2 45

GP16 46 2 17 2 115 2 47

GP17 35 1 17 2 148 2 49

GP18 50 2 15 2 97 2 50

GP19 37 2 15 2 130 2 52

GP20 48 2 15 2 96 2 54

GP21 38 2 14 2 118 2 55

GP22 35 1 14 2 124 2 57

GP23 48 2 14 2 88 2 58

GP24 52 2 13 2 77 1 60

GP25 41 2 11 1 84 1 61

GP26 24 1 11 1 141 1 62

GP27 36 2 11 1 91 1 64

GP28 35 1 10 1 93 1 65

GP29 41 2 10 1 79 1 66

Other GPs 492 21 167 22 106 19 85

No GP – – 134 – – 15 100

Total 2341 100 885 100 – 100 –

A total of 50% of conceptions could be reached by targeting a minimum of 18 GP practices.

0

20

40

60

80

Cum

. sha

re o

f con

cept

ions

(%)

0 20 40 60 80 100Cum. share of units (%)

GP practices Secondary schools

Fig. 4. Gains chart to attain targets of GP practices and secondary schools.

J. Petersen et al. / Health & Place 15 (2009) 300–307306

The purpose of the method described here is to provide an easyand cheap method to predict the most likely schools and GPpractices from routinely collected data for targeted interventions.The coupling of the data sets by postcode rather than by anyindividual identifiers (such as name, address, date of birth, NHSnumber, etc.) should have fewer implications for data protectionand individual privacy. In the current practice postcode isconsidered identifiable information and as such subject to ethical

approval from government bodies (Department of Health, 2003).In a recent survey concerning the collection, storage and analysisof cancer registry data, however, only 8% of responders objected tothe use of full postcode, while the proportion of objectors rose to16% on a question of whether to store personal name and fulladdress (Barrett et al., 2006). The debate about patient consentand rights with respect to the accumulation of identifiable data byhealth authorities continues (e.g. McGilchrist et al., 2007) as docalls for greater integration of medical databases for epidemiolo-gical gains (e.g. Mladovsky et al., 2008). With the methoddescribed here, however, we wish to demonstrate that data setscollected for local service purposes do not need to be fully‘identifiable’ but can be effective even if it only contains small areacounts with no other identifiers.

There is at present little evidence to suggest that GP practicesprovide useful foci for teenage pregnancy intervention. It hashowever been argued that some young people may prefer therelative anonymity of attending a general as opposed to aspecialist sexual health clinic (Tripp and Viner, 2005). Moreresources are now directed towards primary and communityhealth care and GP practices are rewarded for matching servicesto local needs in their new contracts with the Department ofHealth (2006). Estimating conceptions for each GP is another wayin which a trust can continue to make this issue ‘visible’ andrelevant for local stakeholders.

Finally, consideration of teenage pregnancy and other adoles-cence health issues are complicated by peer group influences(Crosnoe and McNeely, 2008) that may lie beyond the control offamilies, family doctors and school head teachers. Target setting

Page 8: Teenage pregnancy—New tools to support local health campaigns

ARTICLE IN PRESS

J. Petersen et al. / Health & Place 15 (2009) 300–307 307

(as in performance indicators) for teenage pregnancy rates mayfor this reason not be well-received. Targeted strategies based onlocal evidence, as suggested in this paper, however, can provide anew focus for the many local initiatives to reduce unwantedteenage pregnancy.

Acknowledgement

This research was funded by ESRC Knowledge TransferPartnership no. 666, joint funded by Southwark Primary CareTrust. We are also indebted to Southwark LEA for their permissionto use local school records in this study. Finally, we would like tothank Bromley Local Research Ethical Committee for theirapproval of this project (Ref. no. 06/Q0705/2).

References

Arai, L., 2007. Peer and neighbourhood influences on teenage pregnancy andfertility: qualitative findings from research in English communities. Health andPlace 13 (1), 87–98.

Atkinson, P., Unwin, D.J., 2002. Density and local attribute estimation of aninfectious disease using MapInfo. Computers and Geosciences 28 (11),1095–1105.

Barnet, B., Arroyo, C., Devoe, M., Duggan, A.K., 2004. Reduced school dropout ratesamong adolescent mothers receiving school-based prenatal care. Archives ofPediatric and Adolescent Medicine 158 (3), 262–268.

Barrett, G., Cassell, J.A., Peacock, J.L., Coleman, M.P., 2006. National survey of Britishpublic’s views on use of identifiable medical data by the National CancerRegistry. British Medical Journal 332 (7549), 1068–1072.

Bennett, S.E., Assefi, N.P., 2005. School-based teenage pregnancy preventionprograms: a systematic review of randomized controlled trials. Journal ofAdolescent Health 36, 72–81.

Beyer, E.L., 2004. Hawth’s analysis tools for ArcGIS.Bonell, C., Fletcher, A., McCambridge, J., 2007. Improving school ethos may reduce

substance misuse and teenage pregnancy. British Medical Journal 334, 614–616.Butt, D., Jani, N., Petrie, S., 2006. Mapping of hotspots: an analysis of teenage

conceptions in Southwark 2002–04: Public Health, Southwark PCT.Cressie, N., 1991. Statistics for spatial data. Wiley, New York.Crosnoe, R., McNeely, C., 2008. Peer relations, adolescent behavior, and Public

Health research and practice. Family and Community Health 31 (18), S71–S80.Cummins, S., Curtis, S., Diez-Roux, A.V., MacIntyre, S., 2007. Understanding and

representing ‘place’ in health research: a relational approach. Social Scienceand Medicine 65, 1825–1838.

de Smith, M.J., Goodchild, M.F., Longley, P.A., 2008. Geospatial Analysis. Matador,Leicester.

Department of Children Schools and Families, 2007. The Children’s Plan—BuildingBrighter Futures. London.

Department of Health, 2000. The NHS Plan: A Plan for Investment, A Plan forReform. London.

Department of Health, 2003. NHS Confidentiality Code of Practice, Best PracticeGuidance. London.

Department of Health, 2006. Our Health, Our Care, Our Say: A New Direction forCommunity Services. Department of Health, London.

DiCenso, A., Borthwick, V.W., Busca, C.A., Creatura, C., Holmes, J.A., Kalagian, W.F.,2001. Completing the picture: adolescents talk about what’s missing in sexualhealth services. Canadian Journal of Public Health 92, 35–38.

DiCenso, A., Guyatt, G., Willan, A., Griffith, L., 2002. Interventions to reduceunintended pregnancies among adolescents: systematic review of randomisedcontrolled trials. British Medical Journal 324, 1426–1430.

Gibin, M., Longley, P., Atkinson, P., 2007. Kernel density estimation and percentvoulme contours in General Practice catchment area analysis in urban areas.In: Winstanley, A.C. (Ed.), Geographical Information Science Research UKConference. Maynooth, Ireland, pp. 270–276.

Glennerster, H., 1991. Quasi-markets for education? Economic Journal 101,1268–1276.

Gomez-Rubio, V., Ferrandiz-Ferragud, J., Lopez-Quilez, A., 2005. Detecting clustersof disease with R. Journal of Geographical Systems 7, 189–206.

Haining, R., 2003. Spatial Data Analysis. Cambridge University Press, Cambridge.Holland, S., 2007. Public Health Ethics. Polity Press, Cambridge.Hsiao, W.C., Heller, P.S., 2007. What macroeconomists should know about health

care policy. International Monetary Fund, Washington, DC.Johansen, A., Rasmussen, S., Madsen, M., 2006. Health behaviour among

adolescents in Denmark: influence of school class and individual risk factors.Scandinavian Journal of Public Health 34 (1), 32–40.

Lipsett, A., 2008. They are the indicators that schools can’t be responsible for (30April 2008), Guardian. London.

Longley, P.A., 2005. A renaissance of geodemographics for public service delivery.Progress in Human Geography 29, 57–63.

McGilchrist, M., Sullivan, F., Kalra, D., 2007. Assuring the confidentiality of sharedelectronic health records. British Medical Journal 335 (7632), 1223–1224.

Mladovsky, P., Mossialos, E., McKee, M., 2008. Improving access to research data inEurope. British Medical Journal 336 (7639), 287–288.

Muir Gray, J.A., 1996. Evidence-based healthcare. Churchill Livingstone, Kidlington.Musgrove, P., 2003. Health Economics in Development. World Bank, Washington,

DC.Office for National Statistics, 2006. The 2001 Census: ONS.Office for National Statistics, 2008. Key Population and Vital Statistics. London.Openshaw, S., 1991. Developing appropriate spatial analysis methods for GIS. In:

Maguire, D.J., Goodchild, M.F., Rhind, D.W. (Eds.), Geographical InformationSystems. Longman Scientific & Technical, Harlow, pp. 389–402.

Openshaw, S., Charlton, M., Wymer, C., Craft, A., 1987. A mark 1 GeographicalAnalysis Machine for the automated analysis of point data sets. InternationalJournal of Geographical Information Systems 1, 335–358.

Pearson, M., Sweeting, H., West, P., Young, R., Gordon, J., Turner, K., 2006.Adolescent substance use in different social and peer contexts: a socialnetwork analysis. Drugs: Education, Prevention and Policy 13 (6), 519–536.

Petersen, J., Mateos, P., Longley, P., Atkinson, P., 2008. Inclusion and exclusionproblems in geodemographic targeting, Geographical Information ScienceResearch UK. Manchester Metropolitan University.

Pollock, A.M., Price, D., Viebrock, E., Miller, E., Watt, G., 2007. The market in primarycare. British Medical Journal 335, 475–477.

Silverman, B.W., 1986. Density Estimation for Statistics and Data Analysis.Chapman & Hall, London.

Sleight, P., 2004. Targeting Customers. How to Use Geodemographics and LifestyleData in Your Business. WARC, Henley-on-Thames.

Social Exclusion Unit, 1999. Teenage Pregnancy Report: Social Exclusion Unit,Cabinet Office.

Social Exclusion Unit, 2006. Social Exclusion Unit.Swann, C., Bowe, K., McCormick, G., Kosmin, M., 2003. Teenage Pregnancy and

Parenthood: A Review of Reviews. Health Development Agency, London.Talbot-Smith, A., Pollock, A.M., 2006. The New NHS—A Guide. Routledge, London.Tripp, J., Viner, R., 2005. ABC of adolescence: sexual health, contraception, and

teenage pregnancy. British Medical Journal 330, 590–593.Waller, W., Gotway, C.A., 2004. Applied Spatial Statistics for Public Health Data.

Wiley, New York.West, P., Sweeting, H., Leyland, A., 2004. School effects on pupils’ health

behaviours: evidence in support of the health promoting school. ResearchPapers in Education 19 (3), 261–291.