Assessing the appropriateness of paediatric hospital admissions in the United Kingdom

Transcription

1 Journal of Public Health Medicine Vol. 22, No. 2, pp Printed in Great Britain Assessing the appropriateness of paediatric hospital admissions in the United Kingdom Aneez Esmail, Julie Ann Quayle and Chris Roberts Abstract Background In order to assess whether the documented rise in paediatric admissions was due to inappropriate admissions, an objective measure of the appropriateness of paediatric admissions, modified for use in the United Kingdom, was used to measure the level of inappropriate admissions. The relationship of appropriateness of admissions to age, gender, time of admission and source of referral was investigated. Methods A retrospective review was carried out of a sample of paediatric records in 13 NHS district general hospitals in South Eastern England between April 1990 and March 1991 using the Paediatric Appropriateness Evaluation Protocol (PAEP) modified for use in the United Kingdom. Results A total of 3324 paediatric admissions in 13 hospitals were assessed. Eight per cent of the sampled admissions were inappropriate. Age [odds ratio (OR) ¼ 0.87], gender (OR ¼ 1.39) and weekend admissions (OR ¼ 1.42) were associated with inappropriate admissions. After controlling for these factors, there was no significant variation between hospitals. Conclusion The low level of inappropriate admissions may be a reflection of well-developed primary care services in the United Kingdom. Alternatives to hospital admission for the assessment of minor self-limiting illness in young children may have a role in reducing inappropriate admissions. Keywords: appropriateness, admissions, paediatrics Introduction A decline in infant mortality and deaths from infectious diseases together with evidence that children today are born healthier than ever before 1,2 creates a paradox in which, despite significant advances in services for children and child health, more children are being admitted to hospital than ever before. Several researchers have documented large rises in paediatric hospital admissions in the United Kingdom, 1,3 8 and in a landmark study Hill 9 reported an 88 per cent increase in paediatric medical admissions between 1975 and Although there are no recent studies which have attempted to assess whether these secular changes are persisting in the 1990s, anecdotal evidence together with continued pressure on emergency admissions 10 suggest that paediatric admissions in the United Kingdom are continuing to increase. The reasons for the documented rise in admissions have been attributed to several causes. These include an increase in early diagnoses, 11 treatment of previously untreatable problems, risk of legal claims, 1 and poor social circumstances. 4,12 14 Other factors including changing morbidity, 4,9,15,16 shorter lengths of stay and decreasing threshold of admission 7,17 20 have also been suggested, but the relative importance of these factors is not clear. The role of inappropriate admissions and variations in care as an explanation for increasing hospital admissions has been extensively investigated, especially in North America. A large literature on the variations in appropriateness of care has been published but little work has been carried out in relation to paediatrics, especially in the United Kingdom. Werneke and MacFaul 28 have provided a useful summary of the rationale behind the assessment of appropriateness in paediatric admissions for the United Kingdom. However, they suggest that the American Paediatric Appropriateness Evaluation Protocol (PAEP), which is widely used in the United States as a utilization review instrument, is unsuitable for use in the UK setting, 29 unless it can be modified so that it is more culturally suited to British paediatric practice. This paper describes the results of a study using a modified PAEP for assessing inappropriate hospital use in a health region in the United Kingdom between 1990 and Methods The methods used in the development and modification of the PAEP for use in the United Kingdom closely followed that used by Kemper and Kreger in the United States. 30,31 Consensus School of Primary Care, Faculty of Medicine, University of Manchester, Rusholme Health Centre, Walmer St, Manchester M14 5NP, UK. Aneez Esmail, Senior Lecturer in General Practice National Primary Care Research and Development Centre, Faculty of Medicine, University of Manchester, Rusholme Health Centre, Walmer St, Manchester M14 5NP, UK Julie Ann Quayle, Statistical Assistant Chris Roberts, Senior Research Fellow in Statistics Address correspondence to Aneez Esmail. aesmail@man.ac.uk Faculty of Public Health Medicine 2000

2 232 JOURNAL OF PUBLIC HEALTH MEDICINE development techniques using the views of UK paediatricians and general practitioners (GPs) were used to alter aspects of the original PAEP and checks were then made to assess reliability of the modified instrument. The modified PAEP achieved high reliability (k ¼ 0.85). The modified PAEP was subsequently used in a field study to assess the appropriateness of paediatric admissions in a large health region in Southern England. In order to detect a minimum variation of at least 30 per cent in inappropriate admissions between any two units (assuming a power of 90 per cent and significance of 5 per cent, and assuming that there were a minimum of 900 admissions per annum in each hospital), we calculated that we would have to sample a minimum of 255 cases per unit. The sampling frame of the study was the Hospital Episode Summary admission lists of all hospitals for the financial year This list was provided by the Regional Health Authority as a computer file of all admissions under 16 years of age to every acute provider unit in the region. A random number program, written for use with SAS 32 statistical software, was used to generate a list of approximately 300 cases for each acute hospital in the region. Admissions to intensive care and burns units, elective admissions for surgery and children admitted to adult and psychiatric wards were excluded from the study. Raters who were trained in the use of the PAEP visited the medical records departments for the hospitals being studied and, with the co-operation of medical records staff selected approximately 300 cases per hospital. Each case was identified by its hospital number. If a particular record was not found, then the rater selected the next record on the list. The rater was also given the day of admission to assess in cases where there were multiple admissions for individuals. Notes were reviewed in the medical records department using a proforma to extract relevant details. The Admission criteria were applied to the time of admission. Data collection for each admission also included patient, demographic and illness episode characteristics. In order to assess the consistency of the trained researchers, they were asked at two points in the study to assess a random selection of hospital records together in order to assess concordance between the raters. Results were analysed and tabulated using the SAS statistical analysis package. Comparisons between groups were made using the x 2 statistic. The relationship between possible explanatory variables was examined using a multivariate analysis as this examines the effect of each variable adjusted for the effect of others. The explanatory variables examined included age, gender, source of referral, time and day of admission. The characteristics and practices of a particular hospital s paediatric department are likely to result in variation between hospitals in the risk of inappropriate admission. A valid statistical analysis needs to take into account such variation between clinical units. 33 The proportion of inappropriate admissions was therefore modelled using a multilevel logistic model, a generalization of logistic regression to incorporate random variation between clinical units as well as variation between individuals. Possible predictors of inappropriate admission that were considered in the model were age, gender, time and source of referral. We have not included length of stay as a covariate as this can only be a predictor of appropriateness as a proxy for severity, partly because the PAEP also includes measures of severity in its assessment. Inclusion of length of stay would therefore be unhelpful as it might explain variation due to other causal variables. Results Thirteen hospitals were sampled, with a total of 3324 records being assessed by the reviewers. Overall we achieved a 1 in 10 sample of paediatric admissions. Sixy-four per cent of admissions sampled were under 5 years of age; 60 per cent were males. The gender distribution was nearly identical across all the hospitals sampled. There was a large variation in the age distribution of admissions across the 13 sampled hospitals with the mean age varying from 2.5 to 5.8 years (F 12,3166 ¼ 11.95, p < ) with an overall mean of 4.4 years. Nearly 47 per cent of cases assessed were admitted for less than 2 days. There was a large variation in length of stay in the sampled admissions between hospitals. In order to assess time of admission and its relationship to appropriateness, time was categorized in out of hours ( h) to identify the time when admissions were most likely to be dealt with by on-call staff. Daytime was classified as occurring between 0901 and 1659 h. Nearly 58 per cent of sampled admissions were admitted during on-call hours. About a quarter of admissions sampled took place over the weekend (Saturday and Sunday). The consistency of the researchers was confirmed by the assessment of concordance at two points in the study by jointly assessing 50 records. On the first occasion, 2 months after the field work started, there was no disagreement between the reviewers. Results from the second occasion towards the end of the field work showed an overall agreement rate of 85.7 per cent [k 0.828, SE 0.168, confidence interval (CI) ¼ ]. An assessment of the null hypothesis that there was a difference between the reviewers classification of an admission using McNemar s test was not proved (x 2 ¼ 0.33, p ¼ 0.53). The researchers therefore maintained their level of reliability throughout the length of the study. Factors associated with inappropriate admissions Table 1 shows the assessment of appropriateness of the sampled admissions. Overall, 8 per cent of the sampled admissions were classified as inappropriate by the assessors, with a range of 3 14 per cent.

3 UK PAEDIATRIC HOSPITAL ADMISSIONS 233 Table 1 Proportion of inappropriate admissions by hospital Inappropriate... Assessment of admission (No.) % Hospital A 7/264 3 B 17/229 7 C 14/255 5 D 36/ E 38/ F 18/255 6 G 27/ H 25/ I 13/249 5 J 23/255 9 K 14/255 5 L 14/249 6 M 33/ Total 279/ A multilevel logistic model was fitted including age, gender and time and origin of referral. This full model is summarized in Table 2. From this model there was evidence that age (x 2 ¼ 46.5, p < ) and gender (x 2 ¼ 6.04, p ¼ 0.014) of the child influenced the proportion of inappropriate admissions. There was also evidence that the proportion of inappropriate admissions was higher at week-ends than during week-days (x 2 ¼ 5.5, p ¼ 0.019). There was no evidence that the proportion of inappropriate admissions was higher for out of hours admissions (x 2 < 0.001, p ¼ 1.0) or for GP referrals (x 2 ¼ 2.15, p ¼ 0.146). Interaction terms were added to the model to check for any differential relationship between age and gender and between out of hours and weekend, but no significant interaction was found. A simplified model is given in Table 3, retaining only those variables (age, gender and weekend admissions) that were significant at the usual 0.05 significance level. This was used to give estimates of the age, gender and weekend effects and to describe the variation between the clinical units. Age and inappropriate admissions Overall, nearly 14 per cent of admissions under 1 year of age were inappropriate, compared with 9 per cent for 1 4 years Table 2 Factors associated with appropriateness of admissions; summary of the full model, with parameter estimates, standard errors, odds ratios and individual x 2 values Parameter Standard Odds ratio x 2 value Model term estimate error (95% confidence interval) (upon one degree of freedom) p value Fixed effect Constant Age ( ) Gender ( ) Out of hours ( ) Weekend ( ) GP referral ( ) Random effect Hospital Table 3 Factors associated with appropriate admissions; summary of the simplified model, with parameter estimates, standard errors, odds ratios and individual x 2 values Parameter Standard Odds ratio x 2 value Model term estimate error (95% confidence interval) (upon one degree of freedom) p value Fixed effect Constant Age ( ) Gender ( ) Weekend ( ) Random effect Hospital

4 234 JOURNAL OF PUBLIC HEALTH MEDICINE Table 4 Variation in appropriateness of admissions by age stratifying by hospital Under 1 year 1 4 years 5 9 years 10þ years Hospital Inapp (n) % inapp Inapp (n) % inapp Inapp (n) % inapp Inapp (n) % inapp A 1/39 3 4/94 4 1/58 2 1/73 1 B 7/74 9 8/ /49 4 C 3/52 6 9/99 9 2/51 4 D 22/ / / /23 4 E 16/ / /65 8 3/48 6 F 9/ /78 5 4/67 6 1/65 2 G 9/ / / /59 3 H 13/ / /45 7 I 3/69 5 7/ /70 4 J 9/ / / /17 6 K 4/69 6 7/95 7 2/41 5 1/50 2 L 8/ /94 5 1/49 2 M 10/ / /30 17 TOTAL 114/ / / /621 2 and 2 per cent for children more than 10 years (Table 4). A plot of age groups against the observed logits confirmed a linear relation between log odds and age. The log odds of an inappropriate admission decreases as age increased. From the statistical model in Table 3, the adjusted estimate of the odds ratio between adjacent years was 0.87 with 95 per cent CI ( ). From this the odds ratio for inappropriate admission for under 1 year compared with 15 years can be estimated to be 8.1. For a child of under 1 year of age the estimated proportion of inappropriate admission was 10.6 per cent compared with that of 1.5 per cent for a child of 15 years. Admissions under the age of 1 year of age were approximately seven times more likely to be inappropriate than 15 year old admissions. Gender and inappropriate admissions Despite 60 per cent of admissions assessed being males, the observed proportion of males admitted inappropriately was 7 per cent compared with 10 per cent for females. The adjusted odds ratio for an inappropriate admission for females compared with males is 1.39, with an 95 per cent CI ( ). The adjusted estimate of the proportion of inappropriate admissions of female children was 7.7 per cent as compared with 5.7 per cent for male. Female children were more than a third more likely to be inappropriately admitted than male children. Time and inappropriate admissions There was no difference in the assessment of admission as appropriate or inappropriate based on whether the admission took place during daytime or on call. The adjusted odds of inappropriate admission at weekends as compared with during the week was 1.42 (95 per cent CI ). Weekend admissions were estimated to have a greater proportion of inappropriate admissions (8.3 per cent), compared with weekday admissions (6 per cent). Weekend admissions were almost 40 per cent more likely to be inappropriate than weekday admissions. Source of referrals and inappropriate admissions The majority of admissions were referred either by their GP directly or were admitted via the Accident and Emergency Department (A&E). There was no evidence that GP referrals were more likely to be inappropriate. The level of inappropriate admission was marginally lower for GPs (8 per cent 126 out of 1513 admissions) than from A&E (9 per cent 133 out of 1446 admissions), although this was not significant ( p ¼ 0.143). There is no suggestion from the data we analysed that GPs referrals were more inappropriate than those from A&E Departments or consultants. Discharge diagnosis and inappropriate admission Differences in morbidity in the population are unlikely to differ between districts in a single region but there is good evidence that management policies for individual conditions vary. 34 We therefore sought to assess the appropriateness of admissions for selected conditions. The raters were asked to extract the discharge diagnosis from the notes. In all hospitals in the selected region, the assessment of discharge diagnosis is made by trained clerks using the International Classification of Diagnosis (ICD) coding system. This information is entered on the administrative sheet that is summarized with each admission. The completeness of medical diagnosis recording is variable throughout the

5 UK PAEDIATRIC HOSPITAL ADMISSIONS 235 Table 5 Appropriateness of admission by selected diagnoses Inappropriate Frequency Discharge diagnosis (No.) (%) Gastro-intestinal infections (includes infectious and non-infectious causes) 18/ Acute upper respiratory infections (acute pharyngitis, tonsillitis, 38/ laryngitis, and unspecified) Lower respiratory infections (acute bronchitis and pneumonia) 13/141 9 Asthma 9/324 3 Appendicitis 0/96 0 Arthropathies and related conditions 8/91 9 Conditions originating in the perinatal period 13/55 24 Fractures (skull, neck, upper and lower limbs) 2/248 1 Intracranial injuries 1/178 1 Open wounds and foreign bodies 7/68 10 Poisoning 16/64 25 Unspecified viral illness 16/64 25 Symptoms, signs and other unspecified 58/647 9 Not specified 80/ Missing data n ¼ 886. region. We tabulated and analysed conditions that accounted for more than 2 per cent of the admitted sample. Table 5 shows the assessment of appropriateness of admission by discharge diagnoses. The greatest proportion of inappropriate admissions was confined to conditions originating in the perinatal period, unspecified viral illnesses and poisoning. Children with fractures, appendicitis and asthma had high rates of appropriate admissions. Variations between hospitals The rate of inappropriate admission varied from 3 per cent to 14 per cent (Table 1). These differences between hospitals may be explained by imbalances in other factors that may relate to inappropriate admission. The multilevel models in Table 3 give an estimate of the variation (on the scale of log odds) between hospitals after adjusting for the age, gender and weekend effect. There is some evidence of variation between clinical units in rates of inappropriate admission (x 2 ¼ 3.24, p ¼ 0.072). Figure 1 shows the estimated probability of an inappropriate admission for each hospital, with approximate 95 per cent CIs for hospital effects. The logits have been calculated for each hospital, for a child of average age within the sample (4.4 years). After controlling for the observable characteristics, the probability of an inappropriate admission still appears to vary between hospitals. However, no single hospital stands out, as most CIs overlap. Figure 1 Estimated probabilities of an inappropriate admission, with approximate simultaneous 95% CIs for hospital effects.

6 236 JOURNAL OF PUBLIC HEALTH MEDICINE Discussion A review of the literature prior to the beginning of the study had led us to expect a much higher level of inappropriate admissions than the level of 8 per cent that we found. There have been no previous studies of admissions in the United Kingdom using an appropriately modified PAEP. The use of the PAEP in this study closely followed the methods described by Kreger and Kemper, 30,31 who reported an inappropriate admission rate of per cent in the group of hospitals they studied. The large differences in health care systems between the United States and the United Kingdom mean that rates of inappropriate hospital use are not directly comparable but do raise questions about the possible reasons for the differences between the two countries. One of the reasons for the differences between hospitals could be the possibility of systematic bias as a result of the reviewers use of the instrument. An assessment of a difference between the reviewers classification of an admission during the course of the study suggested that there was no bias although it is not possible to entirely discount such an effect. It would have been preferable if each rater had examined an equal proportion of records from each hospital in order to eliminate any potential bias due to the review in the comparison of hospitals although this would have added to the logistical complexity of the study. There is no gold standard of appropriate admissions and it is possible that in some of the units with very low inappropriate admission rates there are some patients who may need to be admitted but are not. An alternative explanation is that these units have the best clinical practice and other units should try and achieve results similar to them. We have no data on the quality or extent of primary care services in the areas that we studied and one area of further study would be to investigate the relationship between appropriate admission rates and the level and quality of primary care services. This latter relationship may be critical to an understanding of why there is a lower inappropriate admission rate in the United Kingdom than in the United States. There is some evidence that higher levels of primary care services are associated with lower hospitalization of children, 35 and our findings that admissions via the GP tend to be more appropriate than admissions via A&E Departments may be a pointer in this direction. Recent work by Bindman et al. 36 has suggested that there is a relationship between better perceived access to health services overall and lower hospitalization rates for conditions preventable by adequate ambulatory care. Our findings show that the younger the age of the child, the more likely it is that the admissions are inappropriate. Physical signs in young children are vague, the history is frequently not very specific and the ability to exclude serious illness is difficult. The question remains as to whether the hospital remains the best place to make this assessment. We had expected more admissions during on-call hours to be classified as inappropriate on the basis that more junior staff were responsible for the decision to admit during these hours. An alternative explanation could be the categorization that we used (out of hours includes h) is too crude to pick up any differences. The reality of hospital work is that many staff are still working late, past our cut-off of 1700 h, and hence able to offer advice to junior colleagues. The significantly higher number of weekend admissions that are inappropriate would suggest that our reasoning is plausible, as fewer medical staff are around to offer advice to junior staff who are responsible for the majority of admissions. Our findings suggest that there was no difference in the assessment of admission and source of referral. Depending on the policy of individual hospitals, admissions to hospitals may have been directed via the A&E Department and classified as such, even though they originated from a GP. There is no means of ascertaining this from the hospital records that we reviewed. What is certain is that admissions via the A&E Department would have been assessed by a paediatrician (albeit a junior doctor) prior to admission to the ward. More diagnostic facilities would be available to reduce clinical uncertainty in decision making. It is therefore not unreasonable to expect that inappropriate admissions via this route would be significantly different compared with referrals directly from GPs. The fact that our results failed to demonstrate this suggests that the level and quality of primary care services may alter the hospitalization rate. GPs are uniquely placed to know more about the home circumstances of children than hospital physicians and this knowledge may influence the decision to hospitalize a patient. Because GPs are more likely to see minor self-limiting illnesses, they are better able to judge the severity of these illnesses and, more importantly, place them in a social context. The evidence for the quality of primary care services influencing the hospitalization rate of children in the United Kingdom is mixed, 4,37 although it has been postulated as a reason for differences in variations of hospitalizations in parts of the United States. 35 Prevention of hospital admission with the use of paediatric home nursing teams and more intensive support by GPs has been cited by the Audit Commission 19 as one way of reducing unnecessary hospitalization and inappropriate admissions. The Nottingham Paediatric Community Nursing team 38 is cited as an example of an intervention in the United Kingdom that has reduced hospitalization of children. The fact that 21 per cent of appropriate admissions in our study were solely for the purposes of receiving services (mainly nursing) in hospital suggests that many of these services could be provided in the home for example, frequent monitoring by nurses and nebulizer use. It should be borne in mind that increasing care at home may increase the caregiver s burden and therefore the secondary effects of such interventions and their unintended consequences need to be carefully considered. Better nursing support

7 UK PAEDIATRIC HOSPITAL ADMISSIONS 237 may further improve GPs ability to monitor and look after children at home and avoid unnecessary hospitalization. The challenge is to introduce changes that help reduce clinical uncertainty in minor illness and hence reduce the need for hospitalization. There are considerable problems in interpreting the information presented on discharge diagnosis. Nearly 20 per cent of diagnoses were missing and a further 23 per cent were classified under the heading of symptoms, signs and other unspecified conditions. The possibility of systematic bias cannot be excluded when nearly 40 per cent of admissions cannot be classified accurately. However, nearly one-third of classified admissions were for gastro-intestinal infections, acute upper and lower respiratory infections and asthma. Within these four categories, 10 per cent of admissions for gastro-intestinal infections were judged inappropriate, as were 13 per cent of admissions for upper respiratory infections and 9 per cent for lower respiratory infections. In contrast, only 3 per cent of admissions for asthma were classified as inappropriate. Gastro-intestinal infections and upper respiratory infections are self-limiting minor illnesses in developed countries and mortality from these conditions is extremely low. Better management of these conditions in a primary care setting may reduce unnecessary admissions and hospitalizations. Although there has been a suggestion both in published research and in UK government sponsored reports that inappropriate admissions are a problem in paediatrics, our findings do not support this view. The key issue is whether there should be other forms of care apart from hospital admission, which may be cheaper, more efficient, less restrictive and more favourable from the patient s perspective. Key points (1) The use of a modified Paediatric Appropriateness Evaluation Protocol in the UK setting as a utilization review instrument is feasible. (2) The level of inappropriate paediatric admissions identified in the United Kingdom is much lower than that in the United States and probably reflects the greater development of primary care services in the United Kingdom. (3) The level of inappropriate paediatric admissions probably reflects the amount of clinical uncertainty associated with the assessment of acutely ill children. The development of services that mitigate the uncertainty will have a much greater impact on inappropriate admissions to hospital. Acknowledgements The study was funded by the Department of Health Locally Organized Research Scheme. Aneez Esmail was a Harkness Fellow of the Commonwealth Fund of New York at the time of writing this paper. References 1 Forfar JO. Child health in a changing society. Oxford: Oxford University Press, Pharoah POD, Alberman ED. Mortality of low birthweight infants in England and Wales Arch Dis Child 1981; 56: Pharoah POD, Alberman ED. Annual statistical review. Arch Dis Child 1990; 65: Durojaiye LI, Hutchison T, Madeley RJ. Improved primary care does not prevent the admission of children to hospital. Publ Hlth 1989; 103: Golding J, Haslum J. Hospital admissions. In: Butler NR, Golding J, eds. From birth to five. Oxford: Pergamon Press, Anderson HR. Increase in hospitalisation for childhood asthma. Arch Dis Child 1978; 53: Anderson HR. Increase in hospital admissions for childhood asthma: trends in referral, severity, and readmissions from 1970 to 1985 in a health region of the United Kingdom. Thorax 1989; 44: Mitchell EA. International trends in hospital admission rates for asthma. Arch Dis Child 1985; 60: Hill AM. Trends in paediatric medical admissions. Br Med J 1989; 298: Capwell S. The continuing rise in emergency admissions. Br Med J 1996; 312: Forfar JO. Trends in paediatric medical admissions. Br Med J 1989; 298: Dale A. The changing context of childhood. Demographic and economic changes. In: Botting B, ed. The health of our children. Dicennial supplement. London: OPCS, 1995: Benzeval M, Judge K, Whitehead M. Tackling inequalities in health. An agenda for action. London: King s Fund, Wynne J, Hull D. Why are children admitted to hospital? Br Med J 1977; 2: Anderson HR. Trends in the hospital care of acute childhood asthma : a regional study. Br Med J 1980; 281: Khot A, Burn R, Evans N, Lenney C, Lenney W. Seasonal variation and time trends in childhood asthma in England and Wales Br Med J 1984; 289: Logan RFL, Ashley JSA, Klein RE, Robson DM. Dynamics of medical care: the Liverpool study into the use of hospital resources. London: London School of Hygiene and Tropical Medicine (Memoir 14), Ham C. A review of the literature. In: Ham C, ed. Variations in health care: assessing the evidence. London: King s Fund Institute, 1988: The Audit Commission for England and Wales. Children first: a study of hospital services. London: HMSO, Anderson H, Britton J, Esmail A, Hollowell J, Strachan D. Respiratory disease and sudden infant death syndrome. In: Botting B, ed. The health of our children decennial supplement. London: HMSO, 1995: Fink A, Kosecoff J, Chassin MR, Brook RH. Consensus methods: characteristics and guidelines for use. Am J Publ Hlth 1984; 74:

8 238 JOURNAL OF PUBLIC HEALTH MEDICINE 22 Brook RH. Appropriateness: the next frontier. Br Med J 1994; 308: Hicks NR. Some observations on attempts to measure appropriateness of care. Br Med J 1994; 308: Chassin MR, Brook RH, Park RE, et al. Variations in the use of medical and surgical service by the medicare population. N Engl J Med 1986; 314: Park RE, Fink A, Brook RH, et al. Physician ratings of appropriate indications for six medical and surgical procedures. Am J Publ Hlth 1986; 76: Chassin MR, Kosecoff J, Park RE, et al. Does inappropriate use explain geographic variations in the use of health care services. JAMA 1987; 258: Bernstein SJ, McGlynn EA, Siu AL, Roth C, Sherwood MJ, Keesey J. The appropriateness of hysterectomy: a comparison of care in seven health plans. JAMA 1993; 270: Werneke U, MacFaul R. Evaluation of appropriateness of paediatric admission. Arch Dis Child 1996; 74: Werneke U, Smith H, Smith IJ, Taylor J, MacFaul R. Validation of the paediatric appropriateness evaluation protocol in British practice. Arch Dis Child 1997; 77: Kreger BE, Restuccia JD. Assessing the need to hospitalise children: paediatric appropriateness evaluation protocol. Pediatrics 1989; 84: Kemper KJ. Medically inappropriate hospital use in a pediatric population. N Engl J Med 1988; 318: SAS Institute, Inc. SAS/Stat User s Guide, Release 6.03 Edition. Cary, NC: SAS Institute, Inc., Rice N, Leyland A. Multilevel models: applications to health data. J Hlth Serv Res Pol 1996; 1: Anderson HR. Trends and district variations in the hospital care of childhood asthma: results of a regional study Thorax 1990; 45: Perrin JM, Homer CJ, Berwick DM, Woolf AD, Freeman JL, Wennberg JE. Variations in rates of hospitalisation of children in three urban communities. N Engl J Med 1989; 320: Bindman AB, Grumbach K, Osmond D, et al. Preventable hospitalizations and access to health care. JAMA 1995; 274: Rajaratnam G. A study of admissions to paediatric beds. Postgrad Med J 1991; 67: Dryden S. Care in the community: the work of Nottingham paediatric community nursing team. Paediatric Nursing 1989; Accepted on 26 January 2000