Dear Editor,

We thank Dr Gabrielle Laing for her comments regarding our study.[1] Specificity is indeed valuable, but due to the way we defined our infant cohorts the denominator (total live births minus those infants with abnormalities) could only be approximated by using the obstetric data from each study hospital during the period of the study. The sensitivity and positive predictive values were presented in our paper as this data was collected directly. However in view of her comments the calculated specificities in the cardiac group were 2,760/2,765 (99.82%) for paediatric trainees, and 2,233/2,234 (99.96%) for advanced neonatal nurse practitioners (ANNPs)(p <_0.05. in="in" the="the" eye="eye" group="group" paediatric="paediatric" trainees="trainees" displayed="displayed" a="a" calculated="calculated" specificity="specificity" of="of" _2770="_2770" _2774="_2774" _99.86="_99.86" with="with" _2234="_2234" _2238="_2238" _99.82="_99.82" for="for" annps="annps" p="p">0.05). In the hip group calculated specificity was 2,600/2,756 (94.3%) for paediatric trainees and 2,012/2,218 (90.7%) for ANNPs (p <_0.05. thus="thus" it="it" would="would" appear="appear" that="that" annps="annps" are="are" more="more" specific="specific" than="than" paediatric="paediatric" trainees="trainees" when="when" detecting="detecting" cardiac="cardiac" abnormalities="abnormalities" similar="similar" eye="eye" and="and" less="less" for="for" hip="hip" abnormalities.="abnormalities." this="this" latter="latter" result="result" was="was" mainly="mainly" to="to" be="be" accounted="accounted" by="by" the="the" referring="referring" _70="_70" clicky="clicky" hips="hips" compared="compared" _40="_40" referred="referred" _-="_-" is="is" difficult="difficult" know="know" whether="whether" due="due" over-referral="over-referral" or="or" under-referral="under-referral" of="of" a="a" high-risk="high-risk" group="group" trainees.="trainees." p="p"> We agree with Dr Laing that likelihood ratios are helpful and thank her for the table. In view of the fact that a high likelihood ratio for a positive test and a low value for a negative test indicates a more effective screening test,[2] her table demonstrates a clear trend that ANNP screening appears more effective than that performed by paediatric trainees in five out of six likelihood ratios.

We disagree that a rate of 1.05% for hip abnormalities is ten-times the rate seen previously. Using similar criteria other groups have quoted rates of 1.1%[3] and 1.2%[4].

We endorse Dr Laing's statement that both over-referral and under- referral have consequences. In this regard, our data suggests that ANNPs are as effective as trainee paediatricians at detecting hip, eye, and cardiac abnormalities during the neonatal examination, and may in fact be more effective.

Tim Lee
St James's University Hospital

Ruth Skelton
The General Infirmary at Leeds

Caryl Skene Hull and East Yorkshire Hospitals NHS Trust

References

(1) Lee TWR, Skelton RE, Skene C. Routine neonatal examination - Effectiveness of trainee paediatrician compared to advanced neonatal nurse practitioner. Arch Dis Child Fetal Neonatal Ed 2001; 85: F100-F104.
(2) Gilbert R, Logan S. Assessing diagnostic and screening tests. In: Evidenced based paediatrics and child health, edited by Moyer VA, Elliot EJ, Davis RL et al. BMJ Books 2000; P24-36.
(3) Falliner A. Hahne HJ. Hassenpflug J. Sonographic early hip screening and early management of developmental dysplasia of the hip. Journal of Paediatric Orthopaedics 1999. Part B. 8(2): 112-7.
(4) Committee on Quality Improvement, American Academy of Pediatrics - Clinical Practice Guidelines: Early detection of developmental dysplasia of the hip. Pediatrics 2000; 105: 896-905.

Dear Editor

Lees et al report that advanced nurse practitioners (ANNPs) are significantly more effective than trainee paediatricians in detecting abnormalities on neonatal examination. This is an important finding but the paper raises some methodological and clinical issues.

The authors report the sensitivity of the examinations. Given the intrinsic trade off between sensitivity and specificity, it would be helpful if both parameters were presented, together with likelihood ratios. The table gives the likelihood ratios for "positive" and "negative" screening examinations for each test, based on the assumption that all children in each hospital were screened. All 95% confidence intervals for ANNPs and trainee paediatricians overlap.

The second issue is that 53/5027(1.05%) are classified as having abnormal hips, approximately ten times the risk of established dislocation in an unscreened population. The clinically appropriate definition of "abnormal hip" used here must include many children whose hips would have "normalised" without intervention. This may signficantly bias estimates of test performance.

The consequences associated with both missing true cases and over- referring normal children need to be traded against one another. Childre referred have the potential to overload the diagnostic services and may suffer long term effects from parental anxiety generated. In addition, over referral from screening examinations may lead to over treatment of normal children. "Diagnostic" assessments are themselves test that are seldom 100% specific: the greater the number of children referred for diagnostic assessment, the greater the number who will be unnecessarliy treated.

Presentation only of sensitivity leads the authors to suggest that skill mix at routine neonatal examination is more effective. However, calculation of specificity and likelihood ratios may lead to a less clear cut conclusion.

Dear Editor

I am reassured by Dr Lee et al’s response indicating that the same secondary screeners worked in both hospitals in their study. I am not so convinced that their interpretation of sensitivity is correct but this does not affect their conclusions, as there is no reason to suspect that the secondary screeners acted differently at the two hospitals. With that information I accept that they have indeed proved their point.

Dear Editor,

We thank Dr Cliona Ni Bhrolchain for her interest in our study [1] and for the points she raises, but we disagree with her interpretation.

She is correct in noting that the referrals for eye and cardiac abnormalities were routed through senior paediatricians, but wrong to conclude that this would invalidate the results. The same individuals at each hospital performed this senior review, and in any case such a "filter" would not affect the sensitivity of the screening [2], as this is dependent purely on the rate of abnormalities detected by the primary examiners. As we state in our discussion, the calculated positive predictive values may indeed have been improved for both SHOs and ANNPs, but there is no reason to suspect any selective bias.

In terms of infants failing to attend at the specialist clinics, validation of our data was performed demonstrating that even if it were assumed firstly that all such infants were normal, and then secondly that all such infants had abnormalities, the significance of our findings would not be affected.

We originally hypothesised that ANNPs are as effective as trainee paediatricians at detecting hip, eye, and cardiac abnormalities during the neonatal examination. This we demonstrated, furthermore we found that ANNPs are more effective in detecting hip and eye abnormalities. Thus we conclude that in this regard the case for ANNPs is proven.

Tim Lee
Department of Paediatrics,
St James's University Hospital,
Leeds LS9 7TF, UK

Ruth Skelton
Department of Paediatrics,
The General Infirmary at Leeds,
Leeds LS2 9DE, UK

Caryl Skene
Department of Paediatrics,
Hull and East Yorkshire Hospitals NHS Trust,
Anlaby Road,
Hull HU3 2JZ, UK

References

(1) Lee TWR, Skelton RE, Skene C. Routine neonatal examination - Effectiveness of trainee paediatrician compared to advanced neonatal nurse practitioner. Arch Dis Child Fetal Neonatal Ed 2001; 85: F100-F104.
(2) Gilbert R, Logan S. Assessing diagnostic and screening tests. In: Evidenced based paediatrics and child health, edited by Moyer VA, Elliot EJ, Davis RL et al. BMJ Books 2000; P24-36.

Dear Editor,

Lee et al compare routine neonatal examinations by SHOs and ANNPs. They claim that ANNPs were more sensitive and equally specific at detecting all three conditions examined. The study proves this however only for hip abnormalities, referred directly by the ANNPs and SHOs for specialist assessment.

Referrals for eye and heart abnormalities were routed through senior paediatricians in both hospitals. The sensitivity and specificity measured for these conditions therefore reflects those of the senior paediatricians, not the ANNPs and SHOs. Differences in referral practices between the paediatricians at the two hospital sites and different rates of non-attendance from the two hospitals could also have affected the numbers seen at the specialist clinics, particularly for the smaller arms of the study, and should have been acknowledged in the report.

The study therefore encourages the role of ANNPs in neonatal checks but has not yet proved their competence compared to the existing system.