- Email: [email protected]
The CRIB score
99
of temperature recordings were obtained from the anaesthetised children (first group). The upper and lower limits of agreement between tympanic and rectal recordings were + 1.39°C and - 1.10°C, respectively, ie, measurements with the tympanic thermometer could be up to 1 39°C above or 1 10°C below simultaneous rectal recordings. In the second (ward) group of children, bilateral tympanic measurements were compared with oral and bilateral axillary temperatures taken with a standard electronic thermometer (IVAC model P-850, IVAC Corporation, San Diego, USA). 68 sets of readings were obtained. The limits of agreement were found to be + 1 06°C and - 1 10°C between tympanic and oral measurements, and +2-49°C and —074°C between tympanic and axillary readings. This may, of course, have been due to deficiencies in the standard methods and equipment. However, the agreement between right and left tympanic recordings was found to be poor with limits of + 0 84°C and -1.32°C, a range that is surely far in excess of the normal physiological variation between the two ears. The repeatability coefficient (agreement between repeated measurements from the same ear) was found to be 060°C. The corresponding coefficient for repeated axillary readings by the standard electronic thermometer was better at only 0-48°C. Thus, the performance of the tympanic thermometer in our study was similar to that of other non-invasive thermometers. Although it is expensive (about 310), the technology it uses offers a new approach to a very old problem. sets
Paul Brogan, Charmaine Childs, Barbara M Phillips Booth Hall Children’s
Hospital,
Manchester
Christopher Moulton Department of Accident and Emergency Medicine, Manchester Royal Infirmary, Manchester M13 9WL, UK
1 Abbey JC, Anderson AS, Close EL. How long is that thermometer accurate? Am J Nurs 1976; 76: 1375. 2 Kenney RD, Fortenberry JD, Surratt SS, et al. Evaluation of an infrared tympanic membrane thermometer in pediatric patients. Pediatrics 1990; 85: 854-58. Terndrup TE, Milewski A. The performance of two tympanic thermometers in a pediatric emergency department. Clin Pediatr 1991; 30 (supply): 18-23. 4 Shenep JL, Adair JR, Hughes WT, et al. Infrared, thermistor and glass-mercury thermometry for measurement of body temperature in children with cancer. Clin Pediatr 1991; 30 (suppl): 36-41. 5 Milewski A, Ferguson KL, Terndrup TE. Comparison of pulmonary artery, rectal and tympanic membrane temperatures in adult intensive care unit patients. Clin Pediatr 1991; 30 (suppl): 13-16. 6 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; i: 307-10. 3
The CRIB
score
SIR-Sepkowitz’ response (Oct 9, p 938) to our finding that mortality was twice as high after neonatal care in four non-tertiary centres as in nine tertiary centres after adjustment for risk with the clinical risk index for babies (CRIB), (July 24, p 193) were that CRIB is no substitute for population-based studies and that the tertiary centres’ results might reflect selection bias and fewer early deaths. His first criticism overlooks a simple but important distinction. Populationbased studies are essential when comparing mortality rates in communities, to avoid selection bias arising from referrals between hospitals. However, risk-adjusted hospital mortality rates should be used in comparisons of the performance of hospitals,’ since hospitals cannot be judged by the outcomes of patients they do not treat. His other points are not supported by further data. The table shows no differences in initial risk between infants born and treated in non-tertiary (group 1)
tertiary (group 2) hospitals, but increased initial risk in those transferred from non-tertiary to tertiary hospitals (group 3). Hence selection bias favoured the non-tertiary hospitals.
versus
*No of infants with valid CRIB data, median (quartiles). Groups compared by Kruskal-Wallis one-way analysis of variance: tp=0 0 001, tp < 0 001. Risk variables between infants in groups 1 and 2 did not differ; mfants in group 3 were at greater risk than those in groups 1 and 2, by all variables. Total of 1547 infants is one less than in our original report, because 1 infant was subsequently judged to have lethal congenital anomaly and has been excluded from group 3; this does not alter our conclusions. After adjustment for nsk with CRIB, infants in group 1 were twice as likely to die as those in group 2 (odds ratio 2 13, 95% Cl 1 39-3 29, p< 0 001), but infants in group 3 were not (105, CI 0 67-1 65).
Table: Details of infants in groups 1-3*
35/379 (9-2%) of all deaths in infants with CRIB data took place in the first 12 hours of life, of which 5/57 (8-8%) were in non-tertiary hospitals and 30/322 (9-3%) in tertiary hospitals. When these early deaths were excluded the risk-adjusted odds of death with CRIB were still greater in non-tertiary versus tertiary hospitals (odds ratio 2-02, 95% CI 132-311, p=00013). When early deaths and outborn babies were excluded the corresponding odds of death in non-tertiary hospitals increased to 232 (1-47-3-64, p=00001). It is noteworthy that differences similar to these were seen between tertiary and non-tertiary hospitals after adjustment for birthweight, sex, race, and gestation in the New York City study cited by Sepkowitz. We agree with Hughes-Davies (Oct 9, p 938) that infants treated late for streptococcal septicaemia would have higher CRIB scores than those treated early, so their poorer care and outcome might be masked after adjusting for disease severity with CRIB. We discussed this important limitation of physiological scoring systems in our original report. Inappropriate early treatment may bias risk-adjusted comparisons of outcome in favour of hospitals that give poor care, especially if their care improves after initial disease severity has been measured. If such "inappropriate early treatment bias" occurred in our study, the true difference in risk-adjusted mortality between the tertiary and non-tertiary hospitals was probably even greater than that reported. We agree with Zullini and Bonati (Oct 30, p 1115) that prognostic scoring systems should be studied in much larger samples of neonatal units. Unfortunately their comparison of the predictive power of CRIB versus birthweight in 54 infants was too small to allow any reliable inference, which underscores the need for multicentre collaboration. As Rowan et al have shown risk-adjusted comparisons of nationally representative samples will need great care. They are likely to be more reliable when based on scoring systems that have been developed and validated in the countries in which they are applied. However, validation is an iterative process. If variations in risk-adjusted outcome can be further explained by differences in intuitively reasonable, predefined variables such as staffing, clinical policy, or resources, this will in turn corroborate the validity of the scoring system. This approach could allow us to go beyond the present rather crude distinction between tertiary and non-tertiary hospitals, by identifying key aspects of organisation and policy that are associated with
optimum outcome. William
Tarnow-Mordi, Gareth Parry
International Neonatal Network Co-ordinating Centre, Department of Child Health, Ninewells Hospital, Dundee DD1 9SY, UK
1 2
Kiely JL, Susser M. Age at death used to assess the effect of interhospital transfer of newborns. Pediatrics 1984; 73: 854-61. Rowan KM, Kerr JH, Major E, McPherson K, Short A, Vessey MP. Intensive Care Society’s APACHE II study in Britain and Ireland-II: outcome comparisons of intensive care units after adjustment for case mix by the American APACHE II method. BMJ 1993; 307: 977-81.
Paneth N,
1365
of temperature recordings were obtained from the anaesthetised children (first group). The upper and lower limits of agreement between tympanic and rectal recordings were + 1.39°C and - 1.10°C, respectively, ie, measurements with the tympanic thermometer could be up to 1 39°C above or 1 10°C below simultaneous rectal recordings. In the second (ward) group of children, bilateral tympanic measurements were compared with oral and bilateral axillary temperatures taken with a standard electronic thermometer (IVAC model P-850, IVAC Corporation, San Diego, USA). 68 sets of readings were obtained. The limits of agreement were found to be + 1 06°C and - 1 10°C between tympanic and oral measurements, and +2-49°C and —074°C between tympanic and axillary readings. This may, of course, have been due to deficiencies in the standard methods and equipment. However, the agreement between right and left tympanic recordings was found to be poor with limits of + 0 84°C and -1.32°C, a range that is surely far in excess of the normal physiological variation between the two ears. The repeatability coefficient (agreement between repeated measurements from the same ear) was found to be 060°C. The corresponding coefficient for repeated axillary readings by the standard electronic thermometer was better at only 0-48°C. Thus, the performance of the tympanic thermometer in our study was similar to that of other non-invasive thermometers. Although it is expensive (about 310), the technology it uses offers a new approach to a very old problem. sets
Paul Brogan, Charmaine Childs, Barbara M Phillips Booth Hall Children’s
Hospital,
Manchester
Christopher Moulton Department of Accident and Emergency Medicine, Manchester Royal Infirmary, Manchester M13 9WL, UK
1 Abbey JC, Anderson AS, Close EL. How long is that thermometer accurate? Am J Nurs 1976; 76: 1375. 2 Kenney RD, Fortenberry JD, Surratt SS, et al. Evaluation of an infrared tympanic membrane thermometer in pediatric patients. Pediatrics 1990; 85: 854-58. Terndrup TE, Milewski A. The performance of two tympanic thermometers in a pediatric emergency department. Clin Pediatr 1991; 30 (supply): 18-23. 4 Shenep JL, Adair JR, Hughes WT, et al. Infrared, thermistor and glass-mercury thermometry for measurement of body temperature in children with cancer. Clin Pediatr 1991; 30 (suppl): 36-41. 5 Milewski A, Ferguson KL, Terndrup TE. Comparison of pulmonary artery, rectal and tympanic membrane temperatures in adult intensive care unit patients. Clin Pediatr 1991; 30 (suppl): 13-16. 6 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; i: 307-10. 3
The CRIB
score
SIR-Sepkowitz’ response (Oct 9, p 938) to our finding that mortality was twice as high after neonatal care in four non-tertiary centres as in nine tertiary centres after adjustment for risk with the clinical risk index for babies (CRIB), (July 24, p 193) were that CRIB is no substitute for population-based studies and that the tertiary centres’ results might reflect selection bias and fewer early deaths. His first criticism overlooks a simple but important distinction. Populationbased studies are essential when comparing mortality rates in communities, to avoid selection bias arising from referrals between hospitals. However, risk-adjusted hospital mortality rates should be used in comparisons of the performance of hospitals,’ since hospitals cannot be judged by the outcomes of patients they do not treat. His other points are not supported by further data. The table shows no differences in initial risk between infants born and treated in non-tertiary (group 1)
tertiary (group 2) hospitals, but increased initial risk in those transferred from non-tertiary to tertiary hospitals (group 3). Hence selection bias favoured the non-tertiary hospitals.
versus
*No of infants with valid CRIB data, median (quartiles). Groups compared by Kruskal-Wallis one-way analysis of variance: tp=0 0 001, tp < 0 001. Risk variables between infants in groups 1 and 2 did not differ; mfants in group 3 were at greater risk than those in groups 1 and 2, by all variables. Total of 1547 infants is one less than in our original report, because 1 infant was subsequently judged to have lethal congenital anomaly and has been excluded from group 3; this does not alter our conclusions. After adjustment for nsk with CRIB, infants in group 1 were twice as likely to die as those in group 2 (odds ratio 2 13, 95% Cl 1 39-3 29, p< 0 001), but infants in group 3 were not (105, CI 0 67-1 65).
Table: Details of infants in groups 1-3*
35/379 (9-2%) of all deaths in infants with CRIB data took place in the first 12 hours of life, of which 5/57 (8-8%) were in non-tertiary hospitals and 30/322 (9-3%) in tertiary hospitals. When these early deaths were excluded the risk-adjusted odds of death with CRIB were still greater in non-tertiary versus tertiary hospitals (odds ratio 2-02, 95% CI 132-311, p=00013). When early deaths and outborn babies were excluded the corresponding odds of death in non-tertiary hospitals increased to 232 (1-47-3-64, p=00001). It is noteworthy that differences similar to these were seen between tertiary and non-tertiary hospitals after adjustment for birthweight, sex, race, and gestation in the New York City study cited by Sepkowitz. We agree with Hughes-Davies (Oct 9, p 938) that infants treated late for streptococcal septicaemia would have higher CRIB scores than those treated early, so their poorer care and outcome might be masked after adjusting for disease severity with CRIB. We discussed this important limitation of physiological scoring systems in our original report. Inappropriate early treatment may bias risk-adjusted comparisons of outcome in favour of hospitals that give poor care, especially if their care improves after initial disease severity has been measured. If such "inappropriate early treatment bias" occurred in our study, the true difference in risk-adjusted mortality between the tertiary and non-tertiary hospitals was probably even greater than that reported. We agree with Zullini and Bonati (Oct 30, p 1115) that prognostic scoring systems should be studied in much larger samples of neonatal units. Unfortunately their comparison of the predictive power of CRIB versus birthweight in 54 infants was too small to allow any reliable inference, which underscores the need for multicentre collaboration. As Rowan et al have shown risk-adjusted comparisons of nationally representative samples will need great care. They are likely to be more reliable when based on scoring systems that have been developed and validated in the countries in which they are applied. However, validation is an iterative process. If variations in risk-adjusted outcome can be further explained by differences in intuitively reasonable, predefined variables such as staffing, clinical policy, or resources, this will in turn corroborate the validity of the scoring system. This approach could allow us to go beyond the present rather crude distinction between tertiary and non-tertiary hospitals, by identifying key aspects of organisation and policy that are associated with
optimum outcome. William
Tarnow-Mordi, Gareth Parry
International Neonatal Network Co-ordinating Centre, Department of Child Health, Ninewells Hospital, Dundee DD1 9SY, UK
1 2
Kiely JL, Susser M. Age at death used to assess the effect of interhospital transfer of newborns. Pediatrics 1984; 73: 854-61. Rowan KM, Kerr JH, Major E, McPherson K, Short A, Vessey MP. Intensive Care Society’s APACHE II study in Britain and Ireland-II: outcome comparisons of intensive care units after adjustment for case mix by the American APACHE II method. BMJ 1993; 307: 977-81.
Paneth N,
1365