readings summary

E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4

. s c i e n c e d i r e c t . c o m

a v a i l a b l e a t w w w

j o u r n a l h o m e p a g e : w w w . e j c o n l i n e . c o m

Travel times to health care and survival from cancers in Northern England

A.P. Jonesa, R. Haynesa,*, V. Sauerzapf a, S.M. Crawfordb, H. Zhaoa, D. Formanc

aSchool of Environmental Sciences, University of East Anglia, Norwich, Norfolk NR4 7TJ, UK bAiredale General Hospital, Skipton Road, Steeton, Keighley, West Yorkshire BD20 6TD, UK cCentre for Epidemiology and Biostatistics, University of Leeds and Northern and Yorkshire Cancer Registry & Information Service,

Arthington House, Cookridge Hospital, Leeds LS16 6QB, UK

A R T I C L E I N F O

Article history:

Received 29 May 2007

Received in revised form 27 July 2007

Accepted 31 July 2007

Available online 20 September 2007

Keywords:

Survival

Neoplasm staging

Health services accessibility

Rural health

Geographic information systems

0959-8049/$ - see front matter � 2007 Elsevi doi:10.1016/j.ejca.2007.07.028

* Corresponding author: Tel.: +44 1603 592554 E-mail address: [email protected] (R. H

A B S T R A C T

The aim was to assess the effect of geographical accessibility on the stage of cancer at diag-

nosis and survival. Records of 117,097 cases of breast, colorectal, lung, ovary and prostate

cancer diagnosed in Northern England between 1994 and 2002 were supplemented with

estimates of travel times to the patients’ general practitioners (GPs) and hospitals attended,

together with measures of access to public transport. Logistic regression and Cox propor-

tional hazards models were used, adjusting for age, sex, whether the first hospital visited

was a cancer centre and deprivation of area of residence. Late stage at diagnosis was

associated with increasing travel time to GP for breast and colorectal cancers and risk of

death was associated with travel time to GP for prostate cancer. Travel times to hospital

and other accessibility measures showed no consistent associations with stage at diagnosis

or survival, so travel to GP was the only influential factor.

� 2007 Elsevier Ltd. All rights reserved.

1. Introduction

The Calman-Hine Report advocated a redistribution of cancer

services in England away from smaller hospitals and towards

the larger centres, to improve the quality of patient care.1 In

the years following the implementation of the report, there

has been some concern that the concentration of services in

larger, generally urban, hospitals might disadvantage patients

living in more distant, rural localities.2 Some studies have

suggested that people living further from health services are

less likely to use them.3,4 If the costs, time and inconvenience

of travel cause cancer patients living remote from services to

delay seeking medical advice, their tumours might be more

advanced at the time of diagnosis and might have a poorer

prognosis.

Survival from most cancers is improved when diagnosis is

made and treatment commenced at an early stage of the

er Ltd. All rights reserved

; fax: +44 1603 591327. aynes).

disease.5 Several previous studies undertaken in France, the

US and Scotland have shown that cancer patients living re-

mote from specialist centres present with later stage disease

and have shorter survival than patients closer to cancer cen-

tres,6–9 though others have not found such an association.10

In Scotland, for example, Campbell and colleagues found that

increasing distance from a cancer centre was associated with

greater chance of the patient being recorded as a ‘death certif-

icate only’ case for stomach, breast and colorectal cancers

and poorer survival after diagnosis for prostate and lung

cancers.8

No similar study has been reported for England, although

transport disadvantage is known to exist in English rural

areas. While over 90% of rural households own a car, one-

third of adults do not have personal access to a vehicle.

Out of the main towns public transport is infrequent and

expensive. In sparsely populated rural areas 28% of the

.

270 E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4

population are over 60 years of age and this proportion is

increasing.11

Most previous studies have relied on straight line distance

from the patient’s home to the nearest appropriate hospital or

urban centre as a measure of patient travel effort. This has

several shortcomings. The road network, governed by topog-

raphy and other constraints, rarely runs directly from point

A to point B, average travel speeds may vary on different

sections of the road network and public transport may be

available on some routes but not others. Individual patients

may not attend the nearest hospital. Geographical access to

the patients’ general practitioner (GP), who is the gatekeeper

to hospital services in the UK, has rarely been considered.

This study is, we believe, the first investigation of the

effects of geographical accessibility on survival from a num-

ber of common cancers in a large region of England. It uses

a more sophisticated methodology than previously adopted

to estimate car travel times to primary care and to the hospi-

tal of first referral and takes into account the availability of

public and community transport. The aim was to assess the

effect of geographical accessibility on the stage of cancer at

diagnosis and on subsequent survival. This was made possi-

ble by adding estimates of geographical accessibility to cancer

registry records.

2. Methods

2.1. Setting

The study was located in northern England, the area covered

by the Northern and Yorkshire Cancer Registry & Information

Service (NYCRIS). The population covered is around 6.7

million. The area is very diverse, both geographically and

demographically, encompassing large rural tracts together

with major urban areas.

2.2. Subjects

NYCRIS supplied anonymous records of patients registered

with breast, colorectal, lung, ovarian and prostate cancer

from 1994 to 2002. Age at diagnosis, sex, stage at diagnosis,

hospitals attended, general practitioner (GP) and outcome

were included. The sites of cancer were chosen because they

provided large numbers for analysis and were sites for which

earlier studies had suggested access to services may be

important.7,8 NYCRIS was formed in 1998 following a merger

of two cancer registries. Due to issues of data completeness,

only records from the former Yorkshire Registry area were

used for the first four years of the study. Patients with atypical

pathology types, males with breast cancer, breast ductal car-

cinoma in situ and death certificate only cases were omitted

from analysis, as were those not resident in England.

2.3. Derivation of access and other measures

We added the estimated travel time from the patient’s home

to their GP and to the hospital of first referral and the distance

to the closest cancer centre to each record. Distances to

services and travel times by car were estimated using a

geographical information system (GIS). Travel times were

calculated from average car travel speeds assigned to each

section of the road network dependent on road class, whether

it ran through a rural or urban location, and whether the road

was a dual or single carriageway. Although the method is un-

able to account for variations in individual journey times that

may be associated with, for example, traffic volumes, unusual

delays or adverse weather conditions, it has been shown to

produce estimates of travel times that are closely related

to patients’ own recollections of actual journey times to

hospital.12

We also determined the availability of public bus, rail and

community transport. Bus timetables from the various oper-

ating companies serving northern England were used to iden-

tify sections of the rural road network with a bus service

running at least every hour in the daytime from Monday to

Saturday. The GIS software identified the patients who lived

within 800 m, a generally acceptable walking distance, of

these routes.13 Urban areas were all counted as having an

hourly bus service available. A list of areas containing com-

munity transport that could be used by cancer patients to

get to hospital was collected for every relevant local authority.

Whether the patient lived in an urban or a rural ward14 and

the average deprivation of the area of residence were also

added to the cancer register information. Deprivation was

measured for the appropriate lower level census super output

area using the overall Index of Multiple Deprivation 2004.15

We used the total index minus the access to services domain

contribution to avoid duplication of our own access

measures.

2.4. Outcomes

The primary outcome measured was whether the patient was

alive or dead on 31st March 2005, the last date before abstrac-

tion at which records were considered to be complete. For

each patient the time (days) from diagnosis until death or

31st March 2005, whichever was sooner, was calculated.

A second analysis investigated the associations between

geographical accessibility and late stage at diagnosis. Stage

data were only available for 1.1% of the lung and 0.5% of the

prostate patients, so only the breast, colorectal and ovary can-

cer records were used for this part of the study. Tumour stages

were coded in categories from one (localised tumour) to four

(with distant metastases). Patients with tumours categorised

stage one or stage two were classified as early patients; those

with tumours staged three or four were classified as late. Re-

cords with no stage information were treated as missing data.

2.5. Analysis

Each tumour site was analysed separately. For the survival

analysis, Cox proportional hazards models were fitted. The

predictors were age, gender, whether initial treatment was gi-

ven at a cancer centre, area deprivation score and several

measures of geographical accessibility. Bowel sub-site (colon

and rectum) and specific lung pathology (small and non-small

cell) were also included as covariates in the models. Esti-

mated travel times to the hospital of first referral and to the

patient’s GP were included together in the first run of the

model. Subsequent runs omitted the travel time variables

E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4 271

and substituted alternative accessibility measures in turn. For

the analysis of late stage diagnoses, logistic regression mod-

els were fitted using the same predictors.

3. Results

Characteristics of patients in the analyses are summarised in

Table 1. The majority of breast and colorectal patients were

diagnosed with early stage disease, while most ovarian can-

cer patients had advanced cancer at the time of diagnosis.

Survival was most favourable for breast cancer patients and

worst for lung cancers. Stage information was generally not

well recorded. The patients in our study lived an average of

17 min travel time (range 0.5–254 min) from the first hospital,

and 7 min (range 0.5–48 min) from the GP surgery.

Table 2 gives the results from the Cox proportional hazards

model, with hazard coefficients indicating the risk of death

associated with a unit increase in each predictor. Table 2

shows that each increasing year of age was significantly asso-

ciated with increasing risk of death for all cancer sites. There

was a 1.5% increase with each year for lung cancer and a 7%

Table 1 – Characteristics of patients in the study

Breast

Total cases 28,002

No. (%) of males –

Mean age (SD) 62.37 (14.6)

No. (%) of early stage tumours 6177 (22.1)

No. (%) of late stage tumours 1154 (4.1)

No. (%) of patients with no stage information 20,671 (73.8)

Cases died on or before 31.3.05 8,722

No. (%) of males –

Mean age (SD) 69.66 (15.5)

Table 2 – Associations between risk of death and travel times t gender, deprivation, tumour site/type and type of hospital wh confidence intervals

Breast Colorectal

Age at diagnosis

(years)

1.050** (1.048–1.051) 1.036** (1.035–1.038)

Male sex N/A 1.112** (1.077–1.148)

Area deprivation 1.008** (1.007–1.009) 1.005** (1.004 – 1.006)

Tumour sited in

colon, caecum or

appendix

N/A 1.052** (1.019 – 1.087)

Non-small cell

lung cancer

N/A N/A

Small cell lung

cancer

N/A N/A

First treated at

Cancer Centre

0.915** (0.872–0.961) 0.949** (0.915–0.985)

Travel time to first

hospital (min)

0.995** (0.993–0.997) 1.000 (0.998–1.001)

Travel time to GP

surgery (min)

1.002 (0.999–1.006) 1.002 (0.999–1.005)

* p < 0.05.

** p < 0.01.

increase per year for prostate cancer, for example. Male sex

was detrimental for survival for colorectal and lung cancer.

The deprivation level of the area of residence was consis-

tently and significantly related to survival for all sites, with

worst survival in more deprived areas. Colorectal patients

with tumours in the colon, caecum or appendix had a higher

risk of death than those with tumours in the rectum or recto-

sigmoid junction. Likewise, patients with either non-small or

small cell lung cancers had a greater risk of death than pa-

tients with lung tumours that were not specifically coded as

small cell or non-small cell. Being treated first at a cancer cen-

tre rather than at another hospital was significantly associ-

ated with better survival for all sites except prostate cancer.

The access variables showed weaker and less consistent ef-

fects. Estimated travel time to the first hospital was signifi-

cantly associated with the risk of death for cancers of the

breast and lung, but in the opposite direction to that antici-

pated. In these cases, patients further from hospital had a

better chance of survival. Prostate cancer patients living fur-

ther from their GP had a worse chance of survival than those

living closer (the rate was 0.4% increase in risk per minute of

Colorectal Lung Ovary Prostate

28,256 34,923 5228 20,688

15,556 (55.1) 20,983 (60.1) – 20,688 (100)

71.43 (11.6) 70.86 (10.1) 64.33 (14.5) 73.44 (8.8)

11,055 (39.1) – 128 (2.4) –

7166 (25.4) – 994 (19.0) –

10,035 (35.5) – 4107 (78.6) –

17,543 33,018 3,465 10,318

9705 (55.3) 19,926 (60.3) – 10,318 (100)

73.61 (11.3) 71.15 (10.1) 68.39 (13.0) 76.49 (8.5)

o hospital and GP surgery, controlling for age at diagnosis, ere first treated: proportional hazard ratios and 95%

Lung Ovary Prostate

1.015** (1.013–1.016) 1.047** (1.044–1.050) 1.072** (1.069–1.075)

1.107** (1.082–1.134) N/A N/A

1.001** (1.000–1.002) 1.003** (1.001–1.005) 1.006** (1.005–1.007)

N/A N/A N/A

0.568** (0.553–0.583) N/A N/A

0.798** (0.770–0.827) N/A N/A

0.926** (0.903–0.950) 0.909* (0.838–0.987) 0.965 (0.919–1.014)

0.998** (0.998–0.999) 0.999 (0.996–1.002) 0.999 (0.998–1.001)

0.999 (0.997–1.002) 0.999 (0.993–1.005) 1.004* (1.000–1.007)

Table 3 – Associations between risk of death and the secondary access variables, controlling for age at diagnosis, gender, deprivation, tumour site/type and type of hospital where first treated: proportional hazard ratios and 95% confidence intervals

Breast Colorectal Lung Ovary Prostate

Straight line distance

to nearest cancer

centre (km)

0.998**(0.997–1.000) 1.000 (0.999–1.001) 1.001 (1.000–1.001) 1.000 (0.998–1.002) 1.003** (1.002–1.004)

Car journey to closest

railway station (min)

1.002 (0.999–1.005) 1.002 (1.000–1.004) 0.999 (0.998–1.001) 1.005* (1.001–1.010) 1.001 (0.999–1.004)

Close to hourly bus

service

0.958 (0.886–1.036) 0.999 (0.946–1.055) 0.970 (0.926–1.016) 0.939 (0.835–1.056) 0.898** (0.840–0.960)

Ward with community

transport

1.030 (0.986–1.076) 0.978 (0.949–1.009) 1.047** (1.024–1.070) 1.000 (0.933–1.072) 1.026 (0.985–1.068)

Rural ward 1.016 (0.965–1.070) 1.019 (0.982–1.057) 0.995 (0.967–1.024) 0.975 (0.900–1.057) 1.062* (1.013–1.113)

* p < 0.05.

** p < 0.01.

272 E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4

travel), but coefficients for other sites were not significantly

different from 1.00.

Replacing the travel time to hospital and GP variables with

alternative access variables, and continuing to control for age,

deprivation, site in bowel, lung pathology type, first hospital

and sex, where appropriate, did not identify strong effects

(Table 3). While prostate cancer patients who lived greater

distances from the nearest cancer centre appeared to be at

a disadvantage compared with those nearer, breast cancer pa-

tients appeared to be at a relative advantage. Measures of ac-

cess to public transport and rurality were also not

consistently associated with the risk of death.

Table 4 identifies the variables which were significant

predictors of late stage disease at diagnosis for the three

sites with staging information. Diagnosis at a late stage

was significantly associated with patient age for all three,

but while age increased the chances of late stage diagnosis

for breast cancer and ovarian cancer, it was younger pa-

tients with colorectal cancer who tended to present with

late stage disease. For patients suffering from cancer of

the ovary, age was the only significant predictor, perhaps

indicating a lack of power associated with low numbers

(only 128 patients were diagnosed at an early stage). The

other associations were consistent across breast and colo-

rectal cancer patients. Being treated in a hospital that

was not a cancer centre and living in an area with high

Table 4 – Associations between late stage tumour at diagnosis age at diagnosis, gender, tumour site, deprivation and type of 95% confidence intervals

Breast

Age at diagnosis (years) 1.049** (1.043–1.

Male sex N/A

Area deprivation 1.018**(1.014–1.0

Tumour sited in colon, caecum or appendix N/A

First treated at Cancer Centre 0.661**(0.570–0.7

Travel time to first hospital (min) 0.996 (0.991–1.0

Travel time to GP surgery (min) 1.011* (1.000–1.0

* p < 0.05.

** p < 0.01.

deprivation were significantly related to late stage presenta-

tions. For colorectal cancer patients, male sex and a tumour

in the colon, caecum or appendix were additional signifi-

cant factors associated with late stage disease. Travel time

to the hospital was not significantly related to stage of

the disease at diagnosis, but patients with longer travel

times to their GP had a significantly increased risk of being

diagnosed at a late stage compared with patients with short

journeys. The risk coefficients were 1.1% and 0.8% per min-

ute of travel for breast and colorectal cancer patients,

respectively. The coefficient for ovary cancer patients (0.7%

per minute) was comparable in magnitude but not statisti-

cally significant.

Table 5 shows the results when travel times to hospital

and GP were omitted from the analysis and alternative access

variables were added, each time controlling for the other fac-

tors, as in Table 4. No consistent associations were observed

across the three sites. Colorectal cancer patients who lived

further from their nearest cancer centre were significantly

more likely to be diagnosed at a late stage compared with

those closer. For breast and colorectal cancers, the availability

of community transport was associated with a reduced risk of

late stage diagnosis. However, the associations with time to a

railway station and rural location, for ovary and colorectal

cancers, respectively, were in the opposite direction to that

hypothesised.

and travel times to hospital and GP surgery, controlling for hospital where first treated: proportional hazard ratios and

Colorectal Ovary

054) 0.991** (0.988–0.993) 1.026** (1.011–1.042)

1.108** (1.039–1.181) N/A

21) 1.004** (1.002–1.006) 0.998 (0.987–1.009)

1.130** (1.059–1.207) N/A

66) 0.925* (0.859–0.996) 0.793 (0.526–1.196)

02) 1.000 (0.997–1.003) 0.988 (0.975–1.001)

22) 1.008** (1.003–1.013) 1.007 (0.975–1.041)

Table 5 – Associations between late stage tumour at diagnosis and the secondary access variables, controlling for age at diagnosis, gender, deprivation, tumour site and type of hospital where first treated: proportional hazard ratios and 95% confidence intervals

Breast Colorectal Ovary

Straight line distance to nearest cancer centre (km) 1.001 (0.998–1.005) 1.003**(1.001–1.004) 1.002 (0.991–1.013)

Car journey to closest railway station (min) 1.002 (0.994–1.010) 1.000 (0.996–1.004) 0.972** (0.955–0.989)

Close to hourly bus service 1.038 (0.812–1.326) 1.040 (0.933–1.159) 1.164 (0.639–2.122)

Ward with community transport 0.738**(0.646–0.844) 0.836**(0.787–0.888) 1.046 (0.715–1.532)

Rural ward 0.967 (0.822–1.138) 0.919* (0.854–0.988) 0.773 (0.497–1.201)

* p < 0.05.

** p < 0.01.

E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4 273

4. Discussion

The principal new finding of this work is that geographical

accessibility to primary care appears to be more important

for early diagnosis and survival than access to hospital. We

found that travel time to hospital was not a significant factor

influencing stage at diagnosis, suggesting that patients, once

advised of the possibility of a life threatening illness, make

every effort to attend appointments. There was no indication

either that long journeys to hospital were detrimental to sur-

vival, but we did find some evidence of an adverse influence

of travel to the GP who acts as the gatekeeper to secondary

care in the UK. For breast and colorectal patients, those living

at a greater distance from their GP surgery were significantly

more likely to present with late stage disease after controlling

for patient age, sex and deprivation. The probability of late

stage detection increased by about one percent for every min-

ute of car travel time. Survival analysis indicated very slightly

raised hazard ratios with each additional minute of travel

time to the GP for breast and colorectal cancer patients, but

these were not statistically significant and could have been

due to chance. Prostate cancer patients living further from

their GP were, however, significantly less likely to survive

than those living closer. This study was unable to identify

the precise link between travel time and outcome, but it

may be, for example, that people with a long journey to their

doctor were less likely to seek an appointment for signs and

symptoms that did not appear urgent compared with people

who lived closer.

Greater travel times to hospital were associated with bet-

ter chances of survival for breast and lung cancers. This

was contrary to what might have been expected, but was per-

haps in line with another study in part of the same region

which found some evidence that breast cancer patients living

over 30 min car journey from their radiotherapy unit had sig-

nificantly shorter delays before radiotherapy was

commenced.16

A number of alternative measures of geographical accessi-

bility were included. Straight line distance to the nearest can-

cer centre was investigated to permit comparison with the

results of Campbell and colleagues in Scotland.8 Like Camp-

bell, we found associations between distance to the nearest

centre and both survival and stage of diagnosis for some can-

cer sites. The apparent contradiction with the results for tra-

vel to hospital is explained by the relatively low proportion of

patients in our study (26.17% overall) who actually attended a

cancer centre as their first hospital. Those who attended an-

other type of hospital (probably those who lived further from

a cancer centre) may well have had slightly reduced chances

of early detection and survival.

Associations with the other access measures were not

consistent across cancer sites. A parallel survey of cancer pa-

tients attending clinics in various hospitals across the same

study area found that 87% had travelled to hospital by car.

Only 5% had used a bus, and less than 1% had used commu-

nity transport or travelled by train,12 so car travel times are

more likely to influence patient behaviour than the availabil-

ity of other means of transport.

Overall, the possible risks associated with travel time to

primary care were small compared with the other variables

controlled, which were age, sex and tumour location (where

appropriate), whether or not the first hospital was a cancer

centre, and deprivation. The strong and consistent associa-

tions observed between area deprivation, late stage diagnosis

and survival are another contribution of this research, con-

firming some earlier studies.17,18 Living in a deprived area

was associated with worse survival for all the cancers stud-

ied, and analyses of breast and colorectal cancers demon-

strated that diagnosis at a later stage could account for

some of this effect.

Deprivation of the area of residence was used as a surro-

gate for individual measures of socio-economic disadvantage,

which were not available. The ecological fallacy is a possible

pitfall: we do not know that cancer patients living in deprived

areas were themselves deprived. The observed positive asso-

ciation between distance to hospital and better survival might

have been an artefact of imperfect control of the effects of

deprivation or something similar, since inner city populations

tend to be more deprived and closer to hospitals than subur-

ban or rural populations in the UK.

There were other limitations in this study. Access to health

services depends on a wider range of factors than those asso-

ciated with transport, such as the local ratio of doctors to pa-

tients and the availability of appointments.19 Furthermore,

presentation with advanced disease does not always imply

delay but can reflect the aggressive nature of some tumours.20

We had no information on waiting times and other such bar-

riers, so these issues were not investigated. Our analysis of

the ovary cancer records was restricted by low numbers,

and the absence of reliable staging information for lung and

prostate cancers meant that questions relating to the stage

of disease at diagnosis could be followed up only for breast,

274 E U R O P E A N J O U R N A L O F C A N C E R 4 4 ( 2 0 0 8 ) 2 6 9 – 2 7 4

colorectal and ovary cancer patients. No information was

available regarding patient co-morbidity, ethnicity or tumour

grade. We were also not able to distinguish cancer-specific

deaths, so the survival analysis could not avoid including

some deaths due to other causes. These shortcomings are

not unusual in studies of this type and are difficult to avoid.

This study has used more direct measures of patient travel

effort than previous work and has investigated the potential

effects of access to primary care as well as to hospital ser-

vices. We found no evidence of detrimental effects of long

car journeys to hospital on cancer survival in Northern Eng-

land. This provides some reassurance that access to hospital

cancer services is not seriously affected by transport difficul-

ties in rural areas. There was some evidence of longer car

journeys to GP surgeries being associated with detection of

breast and colorectal cancer at a late stage and poorer sur-

vival from prostate cancer, so access to primary care might

be an influential factor. The mechanisms which might ex-

plain why difficulties in travel to primary care are linked to

outcome require further investigation.

Conflict of interest statement

None declared.

Acknowledgements

We thank the staff at NYCRIS, especially Alison Crawford, for

data matching and abstraction. Dr. Eva Morris and Professor

Bob Haward (NYCRIS) gave advice regarding tumour pathol-

ogy. Chris Dibben (St Andrews University) gave assistance

regarding amendment of IMD 2004 Scores. This research

was funded by HM Treasury, the UK Department for Transport

and the UK Department for Environment, Food and Rural Af-

fairs under the Treasury Evidence Based Policy Fund initiative.

We are grateful to the reviewers for helpful comments.

R E F E R E N C E S

1. Expert Advisory Group on Cancer, A Policy Framework for Commissioning Cancer Services: a Report by the Expert Advisory Group on Cancer to the Chief Medical Officers of England and Wales. London, Department of Health, HM Stationery Office, 1995.

2. Pitchforth E, Russell E, Van der Pol M. Access to specialist cancer care: is it equitable? Br J Cancer 2002;87:1221–6.

3. Nuffield Institute for Health & NHS Centre for Reviews and Dissemination, Effective health care: hospital volume and health care outcomes, costs and patient outcomes. York: University of York, 1996.

4. Jordan H, Roderick P, Martin D. The Index of Deprivation 2000 and accessibility effects on health. J Epidemiol Commun Health 2004;58:250–7.

5. Macdonald S, Macleod U, Campbell NC, Weller D, Mitchell E. Systematic review of factors influencing patient and practitioner delay in diagnosis of upper gastrointestinal cancer. Br J Cancer 2006;94:1272–80.

6. Liff JM, Chow WH, Greenberg RS. Rural–urban differences in stage at diagnosis. Possible relationship to cancer screening. Cancer 1991;67:1454–9.

7. Launoy G, Le Coutour X, Gignoux M, Pottier D, Dugleux G. Influence of rural environment on diagnosis, treatment and prognosis of colorectal cancer. J Epidemiol Commun Health 1992;46:365–7.

8. Campbell NC, Elliott AM, Sharp L, Ritchie LD, Cassidy J, Little J. Rural factors and survival from cancer: analysis of Scottish cancer registrations. Br J Cancer 2000;82:1863–6.

9. Campbell NC, Elliott AM, Sharp L, Ritchie LD, Cassidy J, Little J. Rural and urban differences in stage at diagnosis of colorectal and lung cancers. Br J Cancer 2001;84:910–4.

10. Koka VK, Potti A, Fraiman GN, Hanekom D, Hanley JF. An epidemiological study evaluating the relationship of distance from a tertiary care cancer center to early detection of colorectal cancer. Anticancer Res 2002;22:2481–4.

11. Commission for Rural Communities, Rural disadvantage: priorities for action. London: Countryside Agency; 2006.

12. Haynes R, Jones AP, Sauerzapf V, Zhao H. Validation of travel times to hospital estimated by GIS. Int J Health Geograph 2006;5:40.

13. Lovett A, Haynes R, Sunnenberg G, Gale S. Car travel time and accessibility by bus to general practitioner services: a study using patient registers and GIS. Soc Sci Med 2002;55:97–111.

14. Countryside Agency (2000) Ward Level Definition of Rural Areas. <www.neighbourhood.statistics.gov.uk>.

15. Office of the Deputy Prime Minister. The English indices of deprivation 2004 (revised). Wetherby: ODPM Publications; 2004.

16. Mikeljevic JS, Haward R, Johnston C, et al. Trends in postoperative radiotherapy delay and the effect on survival in breast cancer patients treated with conservation surgery. Br J Cancer 2004;90:1343–8.

17. Ionescu MV, Carey F, Tait IS, Steele RJC. Socioeconomic status and stage at presentation of colorectal cancer. Lancet 1998;352:1439.

18. Macleod U, Ross S, Gillis C, McConnachie A, Twelves C, Watt GCM. Socio-economic deprivation and stage of disease at presentation in women with breast cancer. Ann Oncol 2000;11:105–7.

19. Jarman B, Gault S, Alves B, et al. Explaining differences in English hospital death rates using routinely collected data. BMJ 1999;318:1515–20.

20. McGurk M. Socioeconomic status and bowel cancer. Lancet 1999;353:240.