Main

The accompanying paper (Allgar and Neal, submitted) summarises the important literature describing diagnostic delays in six cancers (breast, colorectal, lung, ovarian, prostate, or non-Hodgkin's lymphoma (NHL)), and discusses the importance of diagnostic delay. This paper explores the relationship between sociodemographic factors and the components of diagnostic delay (total, patient and primary care, referral, secondary care) for these six cancers, about which there is a small body of literature in breast and colorectal cancer, but not in prostate, ovarian, or lung cancer or NHL.

For breast cancer, there are conflicting findings with respect to age. No associations have been reported with patient delays (Nosarti et al, 2000; Meechan et al, 2002) or physician delays (Tartter et al, 1999). Longer delays have been associated with older age (Arndt et al, 2002), but faster times to treatment have also been associated with increasing age (Robertson et al, 2004). Other positive findings from the literature include: African-American women having longer delays than white women (Gwyn et al, 2004), and unmarried women having longer patient delays than married women (Thongsuksai et al, 2000). Other negative findings include: no other socioeconomic factors being important in patient delays (Thongsuksai et al, 2000); no sociodemographic factors being important in patient delay (Meechan et al, 2002); and socioeconomic status and ethnicity not being contributory to referral delays (Nosarti et al, 2000). Similarly, there are conflicting findings from the colorectal literature, although this is more limited. One paper has reported faster time to treatment in patients aged 50–74 years (Robertson et al, 2004), another has reported that age and gender were not associated with differences in delays (Gonzalez-Hermoso et al, 2004); and another that marital status is one of several multifactorial reasons for delay (Langenbach et al, 2003).

This paper aims to explore the relationship between sociodemographic factors and the components of diagnostic delay (total, patient and primary care, referral, secondary care) for these six cancers, using patient-reported data from the National Survey of NHS patients: Cancer (DoH, 2002). If associations exist between sociodemographic factors and diagnostic delays, this should influence the design of interventions aimed at reducing diagnostic delays with the aim of improving morbidity, mortality, and psychological outcomes through earlier stage diagnosis.

Materials and methods

Data source and calculating delays

The accompanying paper contains details regarding from The National Survey of NHS Patients: Cancer (DoH, 2002) and our analysis of data to calculate delays therefrom (Allgar and Neal, submitted). In summary, the survey collected data from 65 192 patients with one of six types of cancer (female breast, colorectal, prostate, NHL, lung, and ovarian) from NHS Trusts in England. Various components of delays (patient and primary care delays, referral delays, secondary delays, and total delays) were calculated from answers to questions about their cancer journey. Owing to different diagnostic pathways and ways in which the survey questions were asked, delays were calculated differently for patients who reported seeing their GP prior to diagnosis than for those that did not (diagnosed by screening, direct hospital admission, or interspecialty referral).

Sociodemographic factors

The survey collected demographic data relating to age, sex, social class, marital status, and ethnic group. Age was calculated by subtracting date of birth from the date that the patient first saw a hospital doctor for their cancer, and was then categorised into seven groups (<25, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+ years) for the univariate comparisons. Marital status was classified as ‘married/living with partner’, ‘divorced/separated’, ‘widowed’, or ‘single’. Social class was derived from occupation using the Registrar General categorisation ‘professional’, ‘managerial/technical’, ‘skilled nonmanual’, ‘skilled manual’, ‘partly skilled’, ‘unskilled’, ‘armed forces’, and ‘never worked’. Ethnic group was further categorised to ensure there were adequate numbers in each category: White; Black (Black-Caribbean, Black-African, and Black–other); South Asian (Indian, Pakistani, and Bangladeshi); and Other (Chinese and ‘other’). There was some missing data for the sociodemographic factors, which accounts for the individual category totals in Table 1 sometimes not equalling the base number for each group.

Table 1 Mean delays (s.d.) for each component of delay for all cancers for each sociodemographic variable

Statistics

Initially T-tests and ANOVA were used for each cancer group to compare mean delay between the categorical sociodemographic factors: age categories, sex, marital status, ethnic group, and social class. However, the univariate analysis makes no allowance for confounding factors (e.g. age, sex, marital status, ethnic group, and social class). Generalised Linear Modelling (GLM) was therefore used to investigate which were the most important factors associated with variation in delay, while controlling for the potentially confounding factors. This was undertaken for each cancer group (age was included as a continuous variable, rather than using the age categories). Generalised Linear Modelling provides regression analysis and analysis of variance for one dependent variable (components of delay) by one or more factors. It allows testing of the null hypotheses about the effects of other factors on the means of various groupings of a single dependent variable. For regression analysis, the independent (predictor) variables are specified as covariates (age, sex, marital status, ethnic group, and social class). A P-value of <0.05 was used to indicate statistical significance. All analyses were performed on SPSS (Version 11).

Results

The main results are presented in Tables 1,2 and 3. Table 1 shows the mean delay and standard deviation for each of the component delays, for each of the cancers, and for each of the sociodemographic factors. Table 2 shows the results of the univariate analysis, and Table 3 the results of the GLM. The direction of the trends where there were differences between groups were all in the same direction. These were as follows: sex – female subjects had longer delays than males; age – younger people had longer delays than older people; marital status – single and separated/divorced people had longer delays than married people; social class – lower social class groups had longer delays than higher social class groups; and ethnic group – Black and south Asian people had longer delays than white people.

Table 2 Summary of significant findings from univariate analysis
Table 3 Summary of significant findings from Generalised Linear Modelling (GLM)

Total delay

Individual sociodemographic factors

There was a significant difference in delay and age group for colorectal, lung, NHL, and breast. There was a significant difference in delay and marital status for colorectal, lung, and breast cancer. There was a significant difference in delay and ethnic group for breast cancer.

Generalised Linear Modelling

For colorectal cancer, the significant factors were marital status and age. For lung cancer, none of the factors were significant. For ovarian cancer, none of the factors were significant. For prostate cancer, the only significant factor was social class. For NHL, the only significant factor was age. For breast cancer, the significant factors were marital status and ethnic group.

Pre-hospital delay

Individual sociodemographic factors

There was a significant difference in delay and age group for lung, NHL, and breast. There was a significant difference in delay and marital status for colorectal and breast cancer. There was a significant difference between delay and ethnic group for breast cancer.

Generalised Linear Modelling

For colorectal cancer, the only significant factor was marital status. For lung cancer, the only significant factor was age. For ovarian cancer, none of the factors were significant. For prostate cancer, none of the factors were significant. For NHL, the only significant factor was age. For breast cancer, the significant factors were marital status and ethnic group.

Referral delay

Individual sociodemographic factors

There was a significant difference in delay and age group for all six cancers. There was a significant difference between male and female subjects for colorectal and NHL. There was a significant difference in delay and marital status for colorectal and breast cancer. There was a significant difference in delay and social class for colorectal cancer. There was a significant difference in delay and ethnic group for colorectal, prostate, and breast cancer.

Generalised Linear Modelling

For colorectal cancer, the significant factors were sex, ethnic group, and age. For lung cancer, the only significant factor was age. For ovarian cancer, none of the factors were significant. For prostate cancer, the only significant factor was age. For NHL, the only significant factor was age. For breast cancer, the significant factors were marital status and age.

Secondary care delay

Individual sociodemographic factors

There was a significant difference in delay and age group for colorectal, lung, prostate, NHL, and breast cancer. There was a significant difference between sex and delay for colorectal and lung. There was a significant difference in delay and marital status for colorectal, prostate, NHL, and breast cancer. There was a significant difference in delay and social class for colorectal, ovarian, prostate, and breast cancer. There was a significant difference in delay and ethnic group for lung cancer.

Generalised Linear Modelling

For colorectal cancer, the significant factors were sex, marital status, social class, and age. For lung cancer, the only significant factor was ethnic group and age. For ovarian cancer, the only significant factor was social class. For prostate cancer, the significant factors were marital status, social class, and age. For NHL, the significant factors were sex, marital status, and age. For breast cancer, the significant factors were marital status, social class, and age.

Discussion

The main findings of this study show significant associations between some of the sociodemographic variables and some of component delays in these six cancers. These findings have significant implications for further research and for policy development.

The GLM showed that the significant factors varied by cancer type. Looking at total delay, for colorectal, age, and marital status were the key factors in explaining the variation in delays; this strengthens the limited evidence base to date (Langenbach et al, 2003; Gonzalez-Hermoso et al, 2004; Robertson et al, 2004). For lung and ovarian cancer, none of the factors stood out as being important. For prostate, social class was an important factor. For NHL, age was an important factor. For breast cancer, marital status and ethnic group were important factors, again strengthening the current evidence base (Tartter et al, 1999; Nosarti et al, 2000; Thongsuksai et al, 2000; Arndt et al, 2002; Meechan et al, 2002; Gwyn et al, 2004; Robertson et al, 2004). The trends in the associations using both statistical approaches were all in the same direction for each of the six cancers. The findings for each of the component delays demonstrate the importance of the sociodemographic factors on that stage in the cancer diagnostic journey. The findings for pre-hospital delay most closely mirror the findings for total delays since this is the part of the process where the majority of the delay occurs (Allgar and Neal, submitted). The small, but statistically significant findings for referral and secondary care delay may be of less clinical significance.

Where gender differences existed, female subjects had longer delays than male subjects; this was an unexpected finding, and the reasons for it are unclear and warrant further investigation. Where age differences existed, younger people had longer delays than older people. This may be because cancer is rarer in younger people, so is more likely to go unnoticed by both patients and their health professionals. Where marital status differences existed, single and separated/divorced people had longer delays than married people. The presence of a partner may facilitate earlier diagnosis by noticing symptoms, discussing the meaning of symptoms, and encouraging their presentation to a health professional. Where social class differences existed, lower social class groups had longer delays than higher social class groups. This may be as a result of lower levels of knowledge regarding significant symptoms, and as a result of poorer access to services. Where ethnic group differences existed, Black and south Asian people had longer delays than white people. This may be a result of primary care being slow to provide accessible care appropriate to the needs of minority ethnic populations, and the health care needs of South Asian patients being either ignored or, if they are recognised, subject to various stereotypes and myths (Atkin, 2004).

Strengths and limitations

The strengths and limitations of the data analysed in this paper are discussed in full in the accompanying paper (Allgar and Neal, submitted). In summary, the analysis was based on a large, high-quality data set. Limitations of the data set include the number of patients who had died prior to receiving the survey, recall bias due to time from diagnosis, and lack of data relating to diagnostic stage and comorbidity. In our analysis, various assumptions concerning the data had to be made in the calculation of delays, and, despite the large numbers overall, the numbers of patients in younger age groups for some cancers was small. As a result, our findings must be interpreted with some caution, and may need replicating with other data.

Implications for further research and policy development

Interventions intended to reduce delay (e.g. the urgent suspected cancer referral guidance) need to be appropriate for the population. Research is needed to develop and evaluate interventions aimed at specific groups in order to reduce diagnostic delays with the aim of improving morbidity, mortality, and psychological outcomes through earlier stage diagnosis. The findings of this work will inform the process of who those interventions should be aimed at, and at what stage of the cancer diagnostic journey they are likely to impact upon.