2. INTRODUCCIÓN
2.4. FIESTAS Y TRADICIONES
2.4.1. Paseo Procesional Del Chagra
3.2.1 Accuracy and usefulness of STATS19
The process of data collection and validation of the STATS19 data generally involves several stages. The police officer who attends the accident will record the details in his/her report book. This information results in an accident report and an entry in the accident register at the police station. The data on the accident report is coded and entered on to the computer and subsequently validated according to the requirements set out by the Department of Transport. Subsequently, about 80% of Highways Authorities consult the police when checking apparent errors in accident data (Ibrahim and Silcock, 1992).
In her review of STATS19, Haigney (1995) states that STATS19 is not "a definitive or unimpeachable source of data on road accident statistics in Great Britain". She makes the point that there is evidence of cases being omitted from the database that could seriously impair its representativeness, and that there exists inaccuracies in the database which are likely to cause bias. For every five casualty records recorded by the police, four had errors in their socio-demographic data.
Amongst the accident variables, Shikar and colleagues (1983) found that accuracy was greatest for accident location and time. Accuracy decreased for collision type, light conditions, weather conditions and accident severity. These authors emphasised that these inaccuracies mean that safety programmes
evaluated on the basis of whether or not they result in a reduction of accidents reported to the police are of questionable scientific validity.
3.2.2 Accuracy of hospital data
Evidence of accuracy of STATS19 is often based on matching to hospital data. Previous work tends to assume that the hospital data are accurate. Haigney (1995) believes that this is a supportable assumption. This view is also supported by work, which originated from the previous South East Thames Regional Health Authority (SETRHA, 1993), that found no major data quality problems with date of admission, date of discharge, discharge destination, age, date of birth, and district of residence.
3.2.3 Inaccuracies in specific variables
Age
In a review by James (1991), she reported that age was likely to be accurate in hospital data because it was derived from date of birth, but that it might be estimated in STATS19. Austin (undated) reported that age was omitted by the police in 3.6% of records. He found that age differed between police and hospital casualty (A&E) records by 1 year or less in 60% of cases, and within 5 years in 83% of cases. He also found casualties that differed by as much as 35 years. This is similar to our findings.
Sex
Austin (undated) found only 3 cases (0.3%) where the gender differed. In two out of the 3 cases the police classification of sex was inconsistent with the forename. The discrepancy is much greater than this in the current Sussex study.
Place of Occurrence
The inaccuracy of the place of occurrence of the accident as indicated by the grid reference or the plain language description have been found to be the two most frequent problems with
STATS19 data (Ibrahim and Silcock, 1993). Many errors have been found, such as displaced figures, faulty translation of the 100 kilometre square letters to digits, and transposition of the grid reference easting and northing. The work by Austin (1993) concurs with this. In his review of Highways Authorities, 85% stated that accident location included the greatest number of errors.
Injury Severity
STATS20 (Department of Transport, 1991) lays down guidelines for recording the severity of casualties' injuries. A crude method for measuring the severity of the accident/casualty is used, based on only 3 codes: ‘Fatal’; ‘Serious’; and ‘Slight’.
The recording of‘Fatal’ injuries by the police appears to be accurate, the criterion being explicit and unambiguous (ie. A fatal injury comprises only those cases where death occurs in less than 30 days as a result of the accident, but does not include death from natural causes or suicide.)
However, the differentiation between ‘Serious’ and ‘Slight’ injuries is frequently unsuccessful. The instructions for distinguishing between the two have generally been interpreted as implying that any casualty who is detained in hospital should be classified as seriously injured. However, STATS20 indicates that a casualty should be judged seriously injured even if he/she is not detained in hospital, but is judged to have one or more of the following injuries: fracture, internal injury, severe cuts and lacerations, crushing injury, and contusion.
Nicholl (1980) maintains that 'appreciable numbers' of casualties detained in hospital are incorrectly recorded on the STATS19 form as 'Slight' rather than 'Serious' injuries. Stone (1984) also suggests some miscoding of severity on STATS19. Bull and Roberts (1973) reported that, because of misclassification of the severity of the casualty, the number of seriously injured cases should be increased by 13% above the figure recorded by the police.
Amongst non-fatal injured casualties who were admitted to hospital, Austin (undated) reported that on the police files 190 were classified as ‘Serious’, and 65 as ‘Slight’. That is, 25% of inpatients were misclassified as 'Slight'. The level of misclassification estimated by the current study is even larger than this (39% classified to ‘Slight’). Over all casualties, Austin (undated) found that 12 % had severity incorrectly coded (ie. ‘Slight’ to ‘Serious’, or ‘Serious’ to ‘Slight’) on STATS19 based on data from Humberside. He also estimated the net effect of miscoding was that the number of seriously injured casualties should be increased by 35%.