Results

There were 93 284 records on the laboratory computer belonging to resident infants born in the study period. Of these, 1106 records were marked as repeat samples, and a further 21 records were marked as duplicates. The main matching algorithm identified a further 336 duplicate or repeat records, and 78 more were identified using just two matching variables, leaving 91 743 records.

These records represent 89 347 matched records, 893 unmatched Guthrie cards, and 1503 unmatched CHC records (fig 1). Of the unmatched CHC records, 757 belonged to infants recorded as having been tested elsewhere, moved away, died, or whose parents refused testing. The remaining 746 unmatched records are therefore taken to represent infants who are apparently untested, which is 0.82% of the total 90 850 births notified to CHCs. The proportion apparently untested is highly dependent on district, and considerably lower in non-metropolitan districts (0.14%) than in suburban (0.39%) or inner city districts (1.68%) (table1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Breakdown of records of births in 14 North East Thames districts.

The number of apparently untested infants (746) was similar to the number of unmatched Guthrie cards that could not be attached to a birth record (893). This raised the possibility that some of the unmatched cards belonged in fact to the apparently untested children. The two lists were therefore compared by hand, in order to determine whether names, street addresses, or birth dates had been wrongly entered. It was concluded that only a small proportion, no more than 5% of the apparently untested infants, could be linked to the unmatched Guthrie cards.

Discussion

A 1984–1988 review of phenylketonuria screening showed that only one of 455 cases had been missed through a failure of coverage11 and none of the 728 cases between 1989 and 1998.12 As most untreated phenylketonuria results in profound handicap by 2 years of age,13 this would indicate an approximate 99.9% coverage (95% confidence interval 99.44% to 99.997%). This suggests that the low estimates derived from direct studies of coverage, 97.4%3 and 95.8%4, which led to the National Audit project,6 along with the 95.8% figure from the national study itself, may have severely underascertained the true coverage. Further, the proposed national standard of 99.5% could be seen as insufficiently ambitious, as the data for phenylketonuria suggest that it has already been exceeded.

Detailed consideration of how coverage can be accurately measured has been lacking, and a number of methodological weaknesses can be detected in many of these earlier reports. Because CHC systems receive birth notifications and include fields for the phenylketonuria and congenital hypothyroidism tests, they provide a natural platform for routine coverage estimates. Unfortunately, test results may not always be entered on CHCs, leading to underascertainment of coverage.2 ,6 ,14 In other studies the denominator list may have been overinflated by double counting, as records from multiple sources, such as CHCs, obstetric computers, and laboratory records, appear to have been included.4

These difficulties have, in principle, been avoided in this study, which uses CHC birth notifications as a denominator and is based on a system in which the arrival of each Guthrie card is acknowledged and recorded on the CHC of origin. However, lack of a unique common identifier on CHC records and on Guthrie cards makes the system described here vulnerable to other risk factors for underascertainment, many of which are more prevalent in inner city areas. These include incomplete recording of perinatal death, movement in and out of the region, failure to recognise repeat samples as such, and difficulty in matching incoming Guthrie cards to birth records because of changes of name and/or address.

Although there was no indication that the unmatched Guthrie cards “belonged to” the unmatched birth records, each must belong to someone, possibly to a child who had been resident in another region at birth but who had moved in later. The existence of these cards in large numbers, which has not been reported in previous direct studies of coverage, illustrates the weakness of the existing information systems. This is reinforced by the disagreement between the average numbers of births notified to North East Thames CHCs and the numbers of birth registrations recorded at the Office for National Statistics: the annual average difference for the years 1992 through 1995 was 0.565%,15 comparable with the 0.82% unmatched birth records and 0.98% unmatched Guthrie cards in fig 1.

In spite of this, using the laboratory's routine information system with minimal further processing, we were able to show programme coverage of over 99.9% in three districts outside London, coverage of over 99.5% in seven districts, and an average 98.3% coverage in inner London. The latter figure exceeds the 97.4% average for Merseyside in 1988/893 and the 95.8% reported in two inner London districts in South Thames in 19934 and in the National Audit programme,6 and suggests that earlier reports of low coverage may have been unduly pessimistic.

It is significant that previous studies reporting higher coverage locally or regionally have relied on either careful case by case tracing of those apparently untested or dedicated well coordinated local staff.2 ,3 ,5 ,8 A notable exception is the 99.93% coverage reported in Birmingham as early as 1986, using a routine laboratory information system linked to local CHCs.7 This was the inspiration for the North Thames system, but has the additional advantage of having a unique identifier, the CHC sequence number, printed on to the Guthrie card. Using this system, the West Midlands neonatal laboratory has been able to consistently record coverage far in excess of the 99.5% recommended coverage standard.

Screening coverage does need to be monitored closely for both clinical and public health reasons. Firstly, admission to special care units is known to be associated with a higher risk of failure to screen. Children with congenital hypothyroidism are at risk of perinatal problems, and have been missed or screened late for this reason.16 ,17 This could become a serious issue if screening is extended to include a wider range of metabolic disorders,13 some of which can result in admittance to hospital in the first week of life. Secondly, neonatal screening for sickle cell disease throughout the United Kingdom has been recommended, but accurate local data on coverage are required to determine whether a selective or universal screening strategy would be the most cost effective for each district.18

The new system for assigning NHS numbers, to be implemented before the end of 2002, will create an opportunity to develop a reliable neonatal screening information system on a national scale. The proposal is that maternity computers issue NHS numbers at delivery, and it is also intended that the NHS number will be printed on to Guthrie cards. Accurate reliable local routine monitoring of neonatal screening coverage could then follow, but will require that maternity computers and neonatal laboratory computers are both linked into a national network of CHCs or their future equivalents, to ensure that results are recorded against the correct child's record, wherever the child is resident and wherever the tests are carried out. An operational information system of this sort could measure coverage accurately, could flag up warnings on unscreened children, and could allow assessment of efforts to improve performance, lending itself to continual quality improvement.19 Development of such a system would be more useful than attempts to compare inadequately estimated coverage with arbitrary national standards,6 or attempts to develop “evidence-based” standards.9