METHODOLOGY
4.1. URBAN RURAL CLASSIFICATION (BRITISH SAMPLE)
The NDNS Dental Survey used 80 randomly selected postcode sectors as sampling units. The participants of the NDNS survey were classified as living in an urban or rural location, according to a two-stage classification. The main criterion referred to the population density of the postcode sector they were sampled from. The second criterion was the geographic location of the postcode sector.
Population density is calculated by dividing the population that lives in an area with the acreage that this area covers. 73 of the postcode sectors sampled in NDNS were in England and Wales and 7 in Scotland. Data on the acreage of each postcode sector in England and Wales were provided by the Office for National Statistics (1997), while the respective population figures were taken from the 1991 census (Office of Population Censuses and Surveys, 1994). Data on population and acreage of the postcode sectors in Scotland referred also to the 1991 census (General Register Office for Scotland, 1997). An illustration of the population density of the selected postal sectors is in Appendix 15.
The average population density in Great Britain in 1991 was 0.99 persons per acre. The average density of Greater London was 17.66 persons per acre (ranging between an average of 13.70 in Outer London and 33.23 in Inner London), while the respective figure for the remaining metropolitan areas was 6.01 persons per acre (ranging from an average of 9.67 in the principal cities to 5.08 in the other districts) and the average for the non-metropolitan cities was 5.52 persons per acre (Champion et al., 1996: p. 2). In accordance with those figures, the standard of 5 persons per acre was chosen as the cut off point for the classification of the different postcode sectors into urban or rural. The sectors with population density greater than 5 persons per acre were classified as urban, while the ones with less than that were classified as rural.
The geographic location of the postcode sectors, according to the United Kingdom atlas (Ordnance Survey, 1995), was also taken into account. The classification according to population density was, in general terms, satisfactory, as all sectors classified as urban
were part of big cities. Nevertheless, there were 3 sectors with population density lower than 5 persons per acre, but situated in urban areas. For these sectors, the geographic location criterion prevailed and their population was classified as urban. This two-stage process resulted in the classification of 50 postal sectors (47 from the first and 3 from the second stage) as urban, while the remaining 30 were classified as rural. In terms of participants in the study sample, 444 people were living in urban and 299 in rural areas.
The testing procedure investigated whether the key outcome variable of the study, namely the score of the index of Oral Impacts on Daily Performance (OIDP), was affected by the classification of the British subjects into urban or rural. Apart from that, a similar procedure was carried out for the main explanatory variables (Appendix 16). The relationship between OIDP score and urban-rural classification was investigated for the total British sample, as well as for the dentate and edentulous subsamples (Table 4.1.1). As the distribution of OIDP scores did not follow the pattern of normality, non- parametric tests (Mann-Whitney) had to be used. In all three occasions, the difference in the mean ranks of OIDP scores between urban and rural participants was minimal and statistically not significant (p=0.73 for the total sample, 0.65 for dentate and 0.88 for edentulous subjects). The OIDP scores were not different between urban and rural subjects in neither the total British sample nor its dentate and edentulous subsamples. Apart from the OIDP score, the relationship between OIDP binary (yes/no) variable and urban-rural classification was also investigated for both the dentate and the edentulous British samples (Table 4.1.2). For that purpose, the simple logistic regression analysis, with OIDP binary as the outcome and urban-rural classification as the explanatory variable, was used. The odds ratio was close to 1, thus implying no difference in the OIDP binary variable between urban and rural subjects, and statistical significance was far from reached in both samples.
Based on these results, the classification of the British subjects into urban and rural did not directly relate neither to OIDP score nor to the dichotomous OIDP variable. Furthermore, there were no significant differences on the relationship of the binary OIDP variable with main sociodemographic and clinical variables between subjects classified as living in an urban or a rural area (Tables A 16.1 and A 16.2^). AH those findings were valid for both dentate and edentulous subjects. These results provide evidence that the inclusion of all the subjects in the British sample would not bias its comparability with the Greek sample. Consequently, the total - and not only the urban- British sample was used in the analysis.
^ Throughout this thesis, tables and figures numbered with in the beginning (e.g. A20.1) appear in Appendices. The number immediately after the letter A indicates the num ber o f the relevant Appendix.
Table 4.1.1. Non-parametric test (Mann-Whitney) for the comparison o f OIDP scores between urban and rural British subjects in the w h o le (N=743), dentate (N =399), and edentulous (N=344) sample: number o f subjects, mean ranks, and p value corrected for ties.
Sam ple Urban/Rural
C lassification Number of subjects Mean Rank (OIDP scores) P
All sample Urban 444 370
(N = 7 4 3 ) Rural 299 373 0.734
Dentate sample Urban 233 198
(N = 3 9 9 ) Rural 166 201 0.654
Edentulous sample Urban 211 172
(N = 3 4 4 ) Rural 133 173 0.889
Table 4.1.2. Simple Logistic Regression Analysis for the estimation o f difference in OIDP binary (Yes/No) classification between urban and rural British dentate (N=399) and edentulous (N=344) subjects: frequencies and unadjusted Odds Ratios (95% Confidence Intervals).
Sam ple Urban/Rural OIDP binary U nadjusted
C lassification N 0 Y es O R. (95% c.i.)
Dentate sample Urban 206 27 0.85 (0.47, 1.56)
(N = 3 9 9 ) Rural 144 22
Edentulous sample Urban 177 34 0.96 (0.53, 1.74)