S. Rice et al., The effect of nonscreening smears on screening smear results: a statistical analysis with its implications for the NHS cervical screening programme, CYTOPATHOLO, 11(3), 2000, pp. 158-165
This is a statistical analysis of individual NHS Cervical Screening Program
me laboratories screening smear 'pick-up' rates (defined here as percentage
low grade and percentage high grade of total adequate smears for ages 20-6
4 years derived from general practitioners and community clinics) in relati
on to their nonscreening smear workload proportion (here defined as percent
age nonscreening smears of total laboratory workload from all sources for a
ll ages). This was achieved by the use of three one way ANOVA models in ord
er to receive a complete overview of the results. The models were applied t
o the following: (1) Laboratories with a total workload of less than 15 000
general practitioner (CP) and community clinic smears; (2) laboratories wi
th a total workload of greater than 15 000 CP and community smears and; (3)
all laboratories (i.e. a combination of 1 and 2). The 'test' groups within
each of these models comprised three subgroups based on the percentage of
laboratory workload that consisted of smears from a nonscreening source. Th
is figure ranged from 2.8% to 82.6%. The subgroups were divided so as to co
ntain approximately the same number of laboratories in each one (172 labora
tories in total, 42 with workload < 15 000 and 130 with > 15 000). The resu
lts show that laboratories high and low grade pick-up rates have a positive
ly correlated but variable relationship with the proportion of their worklo
ad that consists of nonscreening smears. The results show significance over
all at the 5% level for high grade and the 10% level for low grade (high gr
ade at P = 0.045 and low grade at P = 0.071). Significance for laboratories
viewing less than 15000 screening smears at the 1% level thigh grade P = 0
.006 and low grade P = 0.005, They show no significance, however, for labor
atories viewing more than 15 000 screening smears (high grade P = 0.457 and
low grade P = 0.622). There is an intriguing possibility that a greater ex
posure to abnormal smears results in a greater tendency to detect them. The
current data provides no evidence to support the NHS Executive's use of 15
000 as a designated figure when quality monitoring and service provision b
ecomes a specific issue. The closure or amalgamation of laboratories with w
orkloads less than this would appear to have no scientific evidence base.