BRITISH JOURNAL Clinical audit
in
LONDON, SATURDAY 20 AUGUST 1977
diagnostic radiology
Forty million units (a report on a chest radiograph equals one unit) of radiological investigation were undertaken in the NHS in 1974, 1 and in diagnostic radiology demands are growing at the rate of 5-10tio per annum.2 Two-thirds of all new casualty attenders have an x-ray examination,3 as do 2300 of all antenatal cases4-in spite of a rapid increase in the number of pregnant women who are also referred for diagnostic ultrasound.5 In 1970 two-thirds of the total population of the United States underwent medical radiographic examination.6 Much diagnostic radiology has come to be seen as a routine request rather than the consequence of a clinical decision, a change which poses profound clinical and medicolegal dilemmas. Yet we know very little about how diagnostic radiological procedures influence the subsequent management of patients. Valid data are needed to answer this question and so help decide which procedures from the wide range of established practice are useful. What form will these data take, how will they be collected, and how can they be expected to influence individual clinical decisions ? The collective process of fact finding and interpretation is often referred to as clinical audit-defined as a formal surveillance of any action with the intention of maximising its effect on and minimising its harm to the individual and assessing its value to society as a whole. Audit may be observational or experimental. The first aim of observational audit is to examine the use of the procedure under surveillance and its consistency from one centre to another and from one user to another. One early example was the study of inpatient management in Scottish hospitals,7 which showed a substantial variation between clinicians and between hospitals in the length of stay after myocardial infarction. Recognition of this inconsistency encouraged a more rational approach to this aspect of treatment and led to controlled trials of early mobilisation. 8 A second aim of observational audit is to help determine how useful any procedure is. The yield of abnormalities detected by a radiological investigation can be determined for various categories of recipient, forming the basis of guidelines for more efficient use. For example, a recent study of preoperative chest radiographs in surgery (excluding cardiopulmonary operations) found no abnormalities in patients under the age of 30; the conclusion was that routine chest films should not be requested for such patients.9 The results of C BRITISH MEDICAL JOURNAL 1977. All reproduction rights reserved.
follow-up studies of patients may also be profitable. The value both of x-ray films of the skull in the management of head injury1° 11 and of epilepsy12 and of films of the lumbar spine in patients with a recent history of pain or trauma'3 has been questioned, since follow-up studies of patients with radiographic abnormalities failed to show any differences in their subsequent clinical management compared with patients with the same clinical history but normal radiographs. Nevertheless, evidence from this kind of audit will not always enable us to judge that a given procedure is being used to the right extent; nor will it always show, with any degree of confidence, that the results of the procedure necessarily improve the patient's management. More formal evidence of usefulness may be needed, and usually this may be acquired only by experimental audit-a modification of the controlled clinical trial of treatment. A recent example in diagnostic radiology was a prospective randomised study of endoscopy and radiology in acute upper gastrointestinal bleeding by Dronfield and his colleagues.'4 No detectable difference in the subsequent management of the two groups of patients was observed, and the finding cast doubt on the need to provide emergency endoscopy facilities for such patients where radiological services are already adequate. For established radiological procedures clinical audit will inevitably be progressive and prolonged, with sometimes as many as three distinct phases. Phase one is identifying procedures which may be of doubtful value, unsafe, or overused. Such procedures may be identified informally at scientific meetings or more formally through reports in journals or the recommendations of specialist advisory bodies. Phase two is observational audit, to establish consistency of use, followed by yield and follow-up studies to help to separate the useful from the not so useful procedures, and to identify those categories of patients most likely to benefit. If uncertainties still persist phase three-experimental audit-may be needed, despite the difficulties this may pose. For newly established diagnostic radiological procedures the evaluation need not be so protracted: ideally, all of these-including computerised axial tomography and diagnostic ultrasound-should be exposed to experimental audit before being made generally available. Finally, the processes of clinical audit must remain in the hands of doctors themselves. We welcome the recognition of NO 6085 PAGE 479
480
BRITISH MEDICAL JOURNAL
this principle by the Royal College of Radiologists, which has set up a working party to look into the problems of audit as they affect the specialty. I
Office of Health Economics, Information Sheet No 28. London, OHE, 1976. Evans, K T, British Journal of Radiology, 1977, 50, 299. 3Galasko, C S B, and Monahan, P R W, British Medical,Journal, 1971, 1, 2
643.
4Carmichael, J H E, and Berry, R J, Lancet, 1976, 1, 351. 5 Davies, A C, Chalmers, I, and Fahmy, D R, British Medical J'ournal, 1977, 1, 443. 6 US Public Health Service, Population Exposure to X-Rays, US, 1970. Department of Health, Education and Welfare, Washington. Publication (HSM) 73-8047.
Heasman, M A, and Carstairs, V, British Medical3Journal, 1971, 1, 495. 8 Medical Division, Royal Infirmary, Glasgow, Lancet, 1973, 2, 346. 9 Rees, A M, et al, British Medical_Journal, 1976, 1, 1333. 50 Bell, R S, and Loop, J W, New England Journal of Medicine, 1971, 284, 236. Roberts, F, and Shopfner, C E, American Journal of Roentgenology, Radium Therapy, and Nuclear Medicine, 1972, 114, 230. 12 Bull, J W D, and Zilkha, K J, British Medical J'ournal, 1968, 4, 569. 13 Radiological Health Services Education Project, Contract No PHS 86-67210. (Unpublished) Bureau of Radiological Health, Public Health Service, Rockville, Maryland. 14 Dronfield, M W, et al, Lancet, 1977, 1, 1167. 1
Familial risks of childhood cancer
affected child, and such twins should be examined regularly. This small familial influence in the distribution of childhood cancers could be due either to sharing of environmental factors (before or after birth) or to shared genes, or to a combination of the two. A genetic influence in- leukaemia is suggested by the predisposition in conditions such as Fanconi's anaemia, which may entail aberrations of chromosomes,7 and by the observation that in Japan familial cases are often associated with parental consanguinity.8 Subclinical immunodeficiency may underlie some familial clusters of lymphomas.9 Clusters of other tumours may also sometimes be due to inherited premalignant syndromes in subclinical form: families with only minimal signs of neurofibromatosis, for example, may have a high incidence of brain tumours.2 Estimates of the risk in families with childhood cancer have often been based on flimsy evidence, and some estimates have been too gloomy. One textbook on genetic counselling'1 puts the risk of Wilms's tumour in sibs of an affected child at about 1 in 20, whereas Draper et a16 found only one case among 2800 sibs of 1200 patients, and that was in a twin. Hence while familial risks of childhood cancer are of great interest to researchers, parents will be relieved to know that in most cases the risk is tiny. ISorsby, A, British 2
3
4 5
Cancer in childhood is uncommon and leaves parents of afflicted children shocked and bewildered. Instinctively, parents seek explanations for the calamity, and their concern is often mingled with anxiety that their other children may also be at risk. Doctors will want to allay these fears, and we now have reliable information about the familial recurrence of childhood cancers that will generally allow them to do so. Certain tumours have long been known to run in families. For example, retinoblastoma can be inherited as an autosomal dominant trait.' Children may inherit conditions such as neurofibromatosis or xeroderma pigmentosum, which themselves carry a high risk of malignancy.2 Such cases, however, account for only a small fraction of childhood cancers. Failure to find environmental causes for the bulk of cancers in children has enlivened the search for genetic factors.3 Striking clusters of tumours in families have been reported. Li et a14 recently described six families in which three or more sibs were affected. Of course, occasional clusters would be expected by chance, but these families seemed to be at special risk. Surveys of death certificates in the United States5 have also provided evidence for familial factors. Since 1953 in Britain most cancers occurring in children have been included in the survey started by Dr Alice Stewart. Information about most of the 20 000 patients has been culled from medical records, and the parents of 15 000 have been interviewed. Draper et at6 have recently used this massive study to estimate the risk of cancer for the sib of an affected child. Two or more children were affected in just over 100 families. Excluding patients with retinoblastoma or tumours known to be associated with premalignant syndromes, they found that the risk for a sib was about 1 in 300, compared with a risk in the general population of about 1 in 600. This small increase in risk existed whether the first child had leukaemia, lymphoma, or another malignancy. The second tumour was often, but not always, of the same type as the first. The risk appeared to be much higher in a twin of the same sex as an
20 AUGUST 1977
6
8 9 10
Medical3Journal,
1972, 2, 580. Lynch, H T, et al, Cancer, 1977, 39, 1867. Miller, R W, in Tumours in Children, eds H B Marsden and J K Steward, 2nd edn. Berlin, Springer-Verlag, 1976. Li, F P, Tucker, M A, and Fraumeni, J F, Journal of Pediatrics, 1976, 88, 419. Miller, R W, Journal of the National Cancer Institute, 1971, 46, 203. Draper, G J, Heaf, M M, and Kinnier Wilson, L M, Journal of Medical Genetics, 1977, 14, 81. Doll, R, The Epidemiology of Leukaemia, p 25. London, Leukaemia Research Fund, 1972. Kurita, S, Kamei, Y, and Ota, K, Cancer, 1974, 34, 1098. Grundy, G W, Creagan, E T, and Fraumeni, J F, J7ournal of the National Cancer Institute, 1973, 51, 767. Stevenson, A C, and Davison, B C C, Genetic Counselling, 2nd edn, p 283. London, Heinemann, 1976.
Pulmonary eosinophilia Eosinophilia, an important response to immune reactions, may be initiated by "eosinophilic chemotactic factor" released from tissue mast cells.' On contact with this factor, the eosinophils are deactivated2 and so are held at the site of the allergic process, where they can ingest and metabolise antigen-antibody complexes; they can also ingest the granules from mast cells and degrade the histamine contained within them.3 Blood eosinophilia is said to be present when there are more than 0-5 x 109 eosinophils per litre, but it is not necessarily accompanied by an increased total white cell count. Since the blood is a transient compartment for eosinophils, with considerable diurnal variation, there may be an excess of eosinophils in tissues or secretions when the blood count is normal,4 and tissue eosinophilia cannot be excluded on the basis of an isolated blood count. Loffler described two clinical conditions associated with blood eosinophilia: endocarditis5 and pulmonary infiltrates.6 Heart disease and eosinophilia were the subject of a recent review in the British Heart3Journal.7 Pulmonary eosinophilia, which is characterised by blood eosinophilia with pulmonary infiltrates, has been classified by Crofton and others into five kinds, of which only the first corresponds to Loffler's original
in
LONDON, SATURDAY 20 AUGUST 1977
diagnostic radiology
Forty million units (a report on a chest radiograph equals one unit) of radiological investigation were undertaken in the NHS in 1974, 1 and in diagnostic radiology demands are growing at the rate of 5-10tio per annum.2 Two-thirds of all new casualty attenders have an x-ray examination,3 as do 2300 of all antenatal cases4-in spite of a rapid increase in the number of pregnant women who are also referred for diagnostic ultrasound.5 In 1970 two-thirds of the total population of the United States underwent medical radiographic examination.6 Much diagnostic radiology has come to be seen as a routine request rather than the consequence of a clinical decision, a change which poses profound clinical and medicolegal dilemmas. Yet we know very little about how diagnostic radiological procedures influence the subsequent management of patients. Valid data are needed to answer this question and so help decide which procedures from the wide range of established practice are useful. What form will these data take, how will they be collected, and how can they be expected to influence individual clinical decisions ? The collective process of fact finding and interpretation is often referred to as clinical audit-defined as a formal surveillance of any action with the intention of maximising its effect on and minimising its harm to the individual and assessing its value to society as a whole. Audit may be observational or experimental. The first aim of observational audit is to examine the use of the procedure under surveillance and its consistency from one centre to another and from one user to another. One early example was the study of inpatient management in Scottish hospitals,7 which showed a substantial variation between clinicians and between hospitals in the length of stay after myocardial infarction. Recognition of this inconsistency encouraged a more rational approach to this aspect of treatment and led to controlled trials of early mobilisation. 8 A second aim of observational audit is to help determine how useful any procedure is. The yield of abnormalities detected by a radiological investigation can be determined for various categories of recipient, forming the basis of guidelines for more efficient use. For example, a recent study of preoperative chest radiographs in surgery (excluding cardiopulmonary operations) found no abnormalities in patients under the age of 30; the conclusion was that routine chest films should not be requested for such patients.9 The results of C BRITISH MEDICAL JOURNAL 1977. All reproduction rights reserved.
follow-up studies of patients may also be profitable. The value both of x-ray films of the skull in the management of head injury1° 11 and of epilepsy12 and of films of the lumbar spine in patients with a recent history of pain or trauma'3 has been questioned, since follow-up studies of patients with radiographic abnormalities failed to show any differences in their subsequent clinical management compared with patients with the same clinical history but normal radiographs. Nevertheless, evidence from this kind of audit will not always enable us to judge that a given procedure is being used to the right extent; nor will it always show, with any degree of confidence, that the results of the procedure necessarily improve the patient's management. More formal evidence of usefulness may be needed, and usually this may be acquired only by experimental audit-a modification of the controlled clinical trial of treatment. A recent example in diagnostic radiology was a prospective randomised study of endoscopy and radiology in acute upper gastrointestinal bleeding by Dronfield and his colleagues.'4 No detectable difference in the subsequent management of the two groups of patients was observed, and the finding cast doubt on the need to provide emergency endoscopy facilities for such patients where radiological services are already adequate. For established radiological procedures clinical audit will inevitably be progressive and prolonged, with sometimes as many as three distinct phases. Phase one is identifying procedures which may be of doubtful value, unsafe, or overused. Such procedures may be identified informally at scientific meetings or more formally through reports in journals or the recommendations of specialist advisory bodies. Phase two is observational audit, to establish consistency of use, followed by yield and follow-up studies to help to separate the useful from the not so useful procedures, and to identify those categories of patients most likely to benefit. If uncertainties still persist phase three-experimental audit-may be needed, despite the difficulties this may pose. For newly established diagnostic radiological procedures the evaluation need not be so protracted: ideally, all of these-including computerised axial tomography and diagnostic ultrasound-should be exposed to experimental audit before being made generally available. Finally, the processes of clinical audit must remain in the hands of doctors themselves. We welcome the recognition of NO 6085 PAGE 479
480
BRITISH MEDICAL JOURNAL
this principle by the Royal College of Radiologists, which has set up a working party to look into the problems of audit as they affect the specialty. I
Office of Health Economics, Information Sheet No 28. London, OHE, 1976. Evans, K T, British Journal of Radiology, 1977, 50, 299. 3Galasko, C S B, and Monahan, P R W, British Medical,Journal, 1971, 1, 2
643.
4Carmichael, J H E, and Berry, R J, Lancet, 1976, 1, 351. 5 Davies, A C, Chalmers, I, and Fahmy, D R, British Medical J'ournal, 1977, 1, 443. 6 US Public Health Service, Population Exposure to X-Rays, US, 1970. Department of Health, Education and Welfare, Washington. Publication (HSM) 73-8047.
Heasman, M A, and Carstairs, V, British Medical3Journal, 1971, 1, 495. 8 Medical Division, Royal Infirmary, Glasgow, Lancet, 1973, 2, 346. 9 Rees, A M, et al, British Medical_Journal, 1976, 1, 1333. 50 Bell, R S, and Loop, J W, New England Journal of Medicine, 1971, 284, 236. Roberts, F, and Shopfner, C E, American Journal of Roentgenology, Radium Therapy, and Nuclear Medicine, 1972, 114, 230. 12 Bull, J W D, and Zilkha, K J, British Medical J'ournal, 1968, 4, 569. 13 Radiological Health Services Education Project, Contract No PHS 86-67210. (Unpublished) Bureau of Radiological Health, Public Health Service, Rockville, Maryland. 14 Dronfield, M W, et al, Lancet, 1977, 1, 1167. 1
Familial risks of childhood cancer
affected child, and such twins should be examined regularly. This small familial influence in the distribution of childhood cancers could be due either to sharing of environmental factors (before or after birth) or to shared genes, or to a combination of the two. A genetic influence in- leukaemia is suggested by the predisposition in conditions such as Fanconi's anaemia, which may entail aberrations of chromosomes,7 and by the observation that in Japan familial cases are often associated with parental consanguinity.8 Subclinical immunodeficiency may underlie some familial clusters of lymphomas.9 Clusters of other tumours may also sometimes be due to inherited premalignant syndromes in subclinical form: families with only minimal signs of neurofibromatosis, for example, may have a high incidence of brain tumours.2 Estimates of the risk in families with childhood cancer have often been based on flimsy evidence, and some estimates have been too gloomy. One textbook on genetic counselling'1 puts the risk of Wilms's tumour in sibs of an affected child at about 1 in 20, whereas Draper et a16 found only one case among 2800 sibs of 1200 patients, and that was in a twin. Hence while familial risks of childhood cancer are of great interest to researchers, parents will be relieved to know that in most cases the risk is tiny. ISorsby, A, British 2
3
4 5
Cancer in childhood is uncommon and leaves parents of afflicted children shocked and bewildered. Instinctively, parents seek explanations for the calamity, and their concern is often mingled with anxiety that their other children may also be at risk. Doctors will want to allay these fears, and we now have reliable information about the familial recurrence of childhood cancers that will generally allow them to do so. Certain tumours have long been known to run in families. For example, retinoblastoma can be inherited as an autosomal dominant trait.' Children may inherit conditions such as neurofibromatosis or xeroderma pigmentosum, which themselves carry a high risk of malignancy.2 Such cases, however, account for only a small fraction of childhood cancers. Failure to find environmental causes for the bulk of cancers in children has enlivened the search for genetic factors.3 Striking clusters of tumours in families have been reported. Li et a14 recently described six families in which three or more sibs were affected. Of course, occasional clusters would be expected by chance, but these families seemed to be at special risk. Surveys of death certificates in the United States5 have also provided evidence for familial factors. Since 1953 in Britain most cancers occurring in children have been included in the survey started by Dr Alice Stewart. Information about most of the 20 000 patients has been culled from medical records, and the parents of 15 000 have been interviewed. Draper et at6 have recently used this massive study to estimate the risk of cancer for the sib of an affected child. Two or more children were affected in just over 100 families. Excluding patients with retinoblastoma or tumours known to be associated with premalignant syndromes, they found that the risk for a sib was about 1 in 300, compared with a risk in the general population of about 1 in 600. This small increase in risk existed whether the first child had leukaemia, lymphoma, or another malignancy. The second tumour was often, but not always, of the same type as the first. The risk appeared to be much higher in a twin of the same sex as an
20 AUGUST 1977
6
8 9 10
Medical3Journal,
1972, 2, 580. Lynch, H T, et al, Cancer, 1977, 39, 1867. Miller, R W, in Tumours in Children, eds H B Marsden and J K Steward, 2nd edn. Berlin, Springer-Verlag, 1976. Li, F P, Tucker, M A, and Fraumeni, J F, Journal of Pediatrics, 1976, 88, 419. Miller, R W, Journal of the National Cancer Institute, 1971, 46, 203. Draper, G J, Heaf, M M, and Kinnier Wilson, L M, Journal of Medical Genetics, 1977, 14, 81. Doll, R, The Epidemiology of Leukaemia, p 25. London, Leukaemia Research Fund, 1972. Kurita, S, Kamei, Y, and Ota, K, Cancer, 1974, 34, 1098. Grundy, G W, Creagan, E T, and Fraumeni, J F, J7ournal of the National Cancer Institute, 1973, 51, 767. Stevenson, A C, and Davison, B C C, Genetic Counselling, 2nd edn, p 283. London, Heinemann, 1976.
Pulmonary eosinophilia Eosinophilia, an important response to immune reactions, may be initiated by "eosinophilic chemotactic factor" released from tissue mast cells.' On contact with this factor, the eosinophils are deactivated2 and so are held at the site of the allergic process, where they can ingest and metabolise antigen-antibody complexes; they can also ingest the granules from mast cells and degrade the histamine contained within them.3 Blood eosinophilia is said to be present when there are more than 0-5 x 109 eosinophils per litre, but it is not necessarily accompanied by an increased total white cell count. Since the blood is a transient compartment for eosinophils, with considerable diurnal variation, there may be an excess of eosinophils in tissues or secretions when the blood count is normal,4 and tissue eosinophilia cannot be excluded on the basis of an isolated blood count. Loffler described two clinical conditions associated with blood eosinophilia: endocarditis5 and pulmonary infiltrates.6 Heart disease and eosinophilia were the subject of a recent review in the British Heart3Journal.7 Pulmonary eosinophilia, which is characterised by blood eosinophilia with pulmonary infiltrates, has been classified by Crofton and others into five kinds, of which only the first corresponds to Loffler's original
