Using radiology records to improve epidemiological information in paediatric fractures

Public Health (1998) 112, 243±247 ß R.I.P.H.H. 1998 http://www.stockton-press.co.uk/ph

Using radiology records to improve epidemiological information in paediatric fractures: a feasibility study A-M O'Byrne1, P Edwards2, S Jarvis2 and N Shabde1 North Tyneside General Hospital, North Tyneside Health Care NHS Trust, Rake Lane, North Shields, Tyne and Wear and Community Child Health, Department of Child Health, University of Newcastle upon Tyne, 13 Walker Terrace, Gateshead, NE8 1EB

1

2

Objectives: To assess the feasibility of using routine computerised radiology records for community injury surveillance data using fractures in the child population as an example. Design: Radiology and in-patient computerised ®les were accessed to extract information concerning type of fracture, age, sex, and home address. Diagnostic coding of radiological reports was carried out using the ICD-9 classi®cation. Children were assigned to local authority wards using home postcodes derived from home addresses. Ward fracture rates were calculated using 1991 census data. The association between ward fracture rates and deprivation was explored using Townsend scores. Setting: North Tyneside General Hospital. Subjects: Children aged 10 ±14 y receiving care as in-patients or out-patients for long-bone fractures. Results: Between April 1991 and March 1996 a total of 497 long-bone fractures were identi®ed. Fractures in boys exceeded those in girls by a ratio of 2 : 1. The most common fracture identi®ed was of the radius and ulna. There was no evidence of an ecological association between long-bone fracture rates in children aged 10 ±14 y and social deprivation. Conclusions: Computerised radiological records may be used to improve epidemiological information concerning fractures. However, at present, considerable time and effort is required to access the information, to identify and to classify, long-bone fractures. Such data could be used to assist in the audit of clinical care and long-term outcomes, and to inform effective local planning and evaluation of injury prevention initiatives. Keywords: computerised radiology records; long bone fractures; epidemiology; social deprivation; local planning; injury prevention initiatives

Introduction Unintentional injury is the single commonest cause of death or medical attendance for children over the age of one.1,2 For unintentional injury among children aged under 15 y, the Health of the Nation3 target is a reduction in death rates by at least 33% by the year 2005. In an average health authority this would amount to one or two fewer deaths annually. Therefore local attention will need to focus on less severe injuries. However there is little accurate data on the frequency of non-fatal injuries in childhood.4 Hospital in- and out-patient data are often used as measures of the community frequency of illness. However these estimates can suffer from selection bias4±6 because the decision by a patient to attend a hospital and the decision by a doctor to admit a patient are both in¯uenced by factors extraneous of the condition itself.7,8 This study focuses on long-bone fractures because they are easily de®ned, they represent over 80% of the total burden of severe injury receiving medical attention (Injury Severity Score 4),4,9 and are unlikely to be managed in the UK without hospitalbased radiological con®rmation. Out-patient diagnostic data are not routinely available but it is known that most people with fractures are treated on an out-patient basis.10 An accurate account of fractures in the child population would enhance our understanding of injury frequency and

Correspondence: Prof S Jarvis, Community Child Health, Department of Child Health, University of Newcastle upon Tyne, 13 Walker Terrace, Gateshead, NE8 1EB Accepted 26 February 1998

distribution and therefore our efforts to reduce unintentional injury in childhood. Our objective was to establish such a data source by focusing on computerised radiology records. Methods The study area was North Tyneside district which had a paediatric population (aged 0±15 y) of 37 438. Most residents are served by a single Accident and Emergency (A&E) department based at North Tyneside General Hospital (NTGH). Approximately 10 miles from NTGH are other hospitals serving the adjacent city of Newcastleupon-Tyne. Those North Tyneside residents who live closer to Newcastle may either self-refer or be referred directly to Newcastle hospitals. The 10 ±14 y age band was chosen to pilot the method as previous work has shown a higher fracture incidence in this age group.11±15 Children who were outside NTGHs catchment area were excluded from the study. Fractures with Abbreviated Injury Scale (AIS)11 scores of 2 or more involving the humerus, radius, ulna, femur, tibia and ®bula were selected. Slipped femoral epiphyses were excluded as the preceding event was not known in all cases and may have been non-traumatic. The A&E department at NTGH maintains close links with the orthopaedic service, particularly the fracture clinic, which is held daily. The records kept in the A&E department and fracture clinics are not computerised. A computerised list of paediatric X-rays requested by fracture clinics between 1 April 1991 and 31 March 1996 was produced using the NTGH radiology information

Radiology recordsÐpaediatric fractures A-M O'Byrne et al

244

systemÐ`KRIS'. The list provided name, date of birth, sex, date of radiological investigation, anatomical site X-rayed and a unique patient identi®cation number. However in some cases only the initial X-ray in A&E had been reported and so those ®lms requested by A&E were also incorporated in the list. The data were then sorted by name of child and each child's X-rays were ordered chronologically to allow elimination of repeat X-rays for the same fracture and to identify children who had sustained more than one fracture over the time period. Each fracture was regarded as a discrete episode (namely, the ®nal list was event-rather than person-based). It was necessary to call up each patient's record on screen in order to determine their home address and radiological diagnosis, as the present radiological system cannot download these ®elds. This process was time consuming and labour intensive. Diagnostic coding of the radiological reports was carried out by one of the authors (AMOB) using the International Classi®cation of Diseases12 (ICD-9), and addresses were postcoded. Though all X-rays performed were logged into the information system, not all had a radiology report attached. This was particularly true of check X-rays and those cases where the ®lms were returned late for reporting. In cases where it was not possible to decide from computerised records whether the child had sustained a fracture or not, clinical notes were accessed with the permission of the appropriate consultant. Where a child had a series of X-rays performed, the ®rst of which was reported as normal, clinical notes were obtained to ®nd out if the child had had a fracture diagnosed following the initial X-ray. For validation purposes a search of computerised inpatient records for the same ®ve-year period was made using ICD-9 diagnoses 812, 813, 820, 821, 823 and 824, between 1 April 1991 and 31 March 1995, and the corresponding ICD-1013 diagnoses from 1 April 1995 to 31 March 1996. For analytic purposes ICD-9 diagnoses were then used throughout. To calculate fracture rates, an estimate of the population of children aged 10±14 y for the mid-point of the study period (1 October 1993) was required. The number of children aged 10±14 y at the start of the study period was obtained from 1991 census data (11 599 children) and an estimate of the number of children aged 10±14 y at the end of the study period (31 March 1996) was made by using the numbers of children who were aged 5±9 y in April 1991 (11 717 children). The mean of these two ®gures (11 658) was used as the denominator. Adjustment to this ®gure was necessary as certain North Tyneside children living closer to Newcastle either self-refer or are referred directly to Newcastle hospitals. To do this, the proportion of children resident in North Tyneside admitted to NTGH with a Table 1

fracture between April 1991 and March 1996 was used to estimate NTGHs catchment population. The impatient data were supplied by Newcastle and North Tyneside Health Authority. For example 90% of children from Battle Hill electoral ward who required inpatient treatment for a fracture sustained in the study period were admitted to NTGH. Since the number of children aged 10±14 y in Battle Hill is 945, the number of children served by NTGH is 850. Postcodes were used to assign children to local authority ward, and ward-based fracture rates were then calculated. Townsend deprivation scores,14 derived from ward-level 1991 population census data, were used as an index of socio-economic deprivation. Results Approximately 2800 X-rays were requested for children aged 10±14 y over the ®ve-year period. Clinical notes were required in 38 cases where it had not been clear from computerised records whether the patient had sustained a fracture or not. In this group, discrepancies between radiological and clinical opinion were noted in six cases. Four of these were scaphoid fractures and were subsequently excluded from the analysis. Discrepancies with in-patient diagnoses were also found in 63 cases where the in-patient diagnoses were less speci®c (for example `unspeci®ed part') than the radiological diagnoses (for example `lower end'). A total of 497 long-bone fractures was identi®ed, representing an incidence rate of 5.5% overall (1.1% per annum). Overall, 241 long-bone fractures (48.5%) were treated as outpatients. Fractures of the radius and ulna were commonest (64%) followed by tibia and ®bula (12%), humerus (11%), ankle (9%), shaft of femur (3%) and neck of femur (1%) (Table 1). There were twice the number of fractures in boys as there were in girls (330 compared to 167). Fractures of the radius and ulna were most frequent for both sexes, peaking at the age 13 for boys and age 11 for girls. There were more leg fractures in older boys than girls (Figure 1). On assessing seasonal variation, a bimodal distribution was evident with peaks in May and August (Figure 2). Fracture rates at ward level over the ®ve-year period ranged from 3.1% (0.6% per annum) to 9.3% (1.8% per annum). These rates are mapped in Figure 3, excluding three wards where NTGH serves less than 60% of the population. The relationship between ward fracture rates and ward Townsend scores was examined using regression analysis (Figure 4). No evidence of a linear relationship was found (b ˆ 0.0014, 95% con®dence interval 70.0007± 0.0035).

Long-bone fractures identi®ed between April 1991 and March 1996 by sex and age Boys

Girls

Fracture

10

11

12

13

14

10±14

10

11

12

13

14

10±14

Humerus Radius=Ulna Neck of Femur Shaft of Femur Tibia=Fibula Ankle All fractures

6 40 0 2 4 2 54

9 38 0 6 3 3 59

9 46 0 1 13 8 77

3 52 0 1 11 6 73

4 39 2 2 14 6 67

31 215 2 12 45 25 330

5 26 1 2 2 1 37

6 32 2 1 2 2 45

6 21 0 1 5 8 41

4 17 0 0 6 5 32

1 7 0 0 1 3 12

22 103 3 4 16 19 167

Children 10±14 53 318 5 16 61 44 497

(11%) (64%) (1%) (3%) (12%) (9%)

Radiology recordsÐpaediatric fractures A-M O'Byrne et al

Discussion Injuries are the leading cause of mortality and morbidity in childhood.15 More than one in ®ve children sustains an injury requiring medical attention each year and 10% of these children are admitted to hospital.4 Hospital in-patient statistics have previously been used as being representative of community frequency of illness because the quality and

Figure 1

Arm and leg fractures identi®ed between April 1991 and March 1996 by sex and age.

quantity of the information held on each case is greater than at any other health service contact point and also because such research can be done cheaply.6 Lyon's work, using a centralised data base of A&E attendances as a tool for injury surveillance, acknowledges the variables that affect hospital attendance and suggests that it is more meaningful to monitor serious childhood injury (namely fractures) to assess progress towards injury reduction targets.16 We consider that radiology records represent a more direct way to compose a register of fractures, using a source that is already fully computerised. Other authors have studied fractures using different methods. A Nottingham study, conducted over a six-month period, identi®ed fractures in children aged under 12 by

Figure 2 Long-bone fractures identi®ed between April 1991 and March 1996 in children aged 10±14 y by month of year.

Figure 3 Map showing long-bone fracture rates in North Tyneside wards (April 1991 to March 1996): children aged 10±14 y. This map excludes Longbenton, Weetslade and Benton wards (see text).

245

246

Radiology recordsÐpaediatric fractures A-M O'Byrne et al

Figure 4 Plot of long-bone fracture rate for ward over 5 y (rate per 100 children aged 10±14 y) against Townsend score.

manually extracting information from hospital activity analysis, fracture clinic and A&E records. Original radiographs were assessed by one of the authors.17 Mann et al,18 performing a 12-year retrospective study of paediatric fractures in Wisconsin, accessed information recorded from initial A&E treatment. This did not include any subsequent radiological input. In determining the incidence of fractures in the Leicestershire Health Authority, Donaldson et al10 constructed a computerised fracture clinic index, based on clinical rather than radiological information, and linked these records to hospital in-patient data. Focusing on the epidemiology of physeal fractures, the Olmstead county study entailed detailed review of all cases and radiographs by one author.19 Cheng's20 Hong Kong study involved retrieval and analysis of the medical records of all children admitted with limb fractures over the study period. It did not address out-patient fractures. We were unable to ®nd any previously published study describing the use of computerised radiology records to create a register of childhood fractures. Our study highlights the importance of using out-patient data in paediatric fractures, as approximately half were identi®ed only from this source. Direct comparison of incidence rates is dif®cult because other studies focused on in-patients alone18±20 and differ in age groups studied. Our annual incidence rate for long bone fractures in 10±14 y olds is 1.1%, which is lower than that described for Nottingham children17 (1.6% in children under 12) and for Swedish children21 (2.1% in children under 16). The sex difference in injury rates, which has been noted in previous studies,17,21,22 is evident here also with boys sustaining twice the number of fractures. The months with the highest incidence of fractures were May (n ˆ 62) and August (n ˆ 60). A similar seasonal variation is seen in Cheng's study20 and the Nottingham study also showed a peak in fracture rates in the early summer (this study did not extend past June). Our ®ndings contrast with Landin's21 where the lowest monthly fracture rate occurred in June and July. This may be accounted for by families leaving the city for summer vacation. We found no ecological association between long-bone fracture rates and social deprivation. This result supports that of Walsh and Jarvis,4 who investigated children with injury severity scores of four or more attending A&E departments, and

that of Lyons16 who investigated all paediatric fractures attending A&E. This contrasts with the strong and increasing association of injury mortality with social class.23 We believe this study has shown that complete epidemiological data describing limb fractures in the community could be produced in a non-labour intensive, repeatable manner through modi®cation of present radiology information systems. Routine health service information, largely used for administrative purposes, can be manipulated to account for repeat X-rays and unreported ®lms, allowing production of useful epidemiological data and adding value to the present systems. The computerised radiology system at NTGH requires modi®cation to achieve these bene®ts routinely as it is presently impossible to download home address and radiology reports alongside patient details. However, as 38 cases (8%) in our study had no radiological diagnosis, the issue of unreported ®lms will need to be considered jointly by A&E, orthopaedic and radiological consultants. Trusts may have their own local reporting arrangements whereby the radiologist is not required to report every ®lm and such agreements have to be taken into account when implementing this system. The ideal information system will cover all ages and be shared jointly by A&E, radiology and fracture clinics. This would allow incorporation of valuable data from A&E concerning injury circumstances, alongside de®nitive radiological reports and details of fracture clinic diagnosis, treatment and outcome. With such a system in place, an essential component to ensure added value is a management group with clinical, radiological and epidemiological participants to develop, analyse and feedback data to potential users. This wider hospital-based injury registration system could serve to facilitate clinical audit, to highlight populations and geographical areas with a high preponderance of fractures, and to thereby inform the planning and evaluation of local injury prevention and control programmes. Key messages  Much clinical information on fractures is recorded but is currently very time consuming to access for aggregated analysis.  It is feasible to use computerised radiology records to improve epidemiological information. They allow comprehensive de®nition of fracture events by person, place and time.  Fractures are frequent and relatively severe injuries. A register could inform local planning and evaluation of injury prevention and control programmes.  More extensive information concerning the antecedents of the injury and clinical outcome could be obtained by linking A&E, radiology and fracture clinic computerised records. Acknowledgements Thanks are due to Ms R Hurst, Mr A Beatty, Mrs D Laidlaw, Mr R Cassidy, Dr R Barton for their help in accessing information; to the consultants for their consent. Thanks are also due to the staff of Newcastle and North Tyneside Health Authority Information Services for assistance with coding and mapping.

Radiology recordsÐpaediatric fractures A-M O'Byrne et al

References 1 Department of Health. The Health of Our Children: A Review in the mid 1990s. Ed. B Botting. OPCS: London, 1995. 2 Sibert J. Accidents to children: the doctor's role. Education or environmental change? Arch Dis Child 1991; 60: 890±893. 3 Department of Health. The Health of the Nation: A Strategy for Health in England. HMSO: London, 1992. 4 Walsh SSM, Jarvis SN, Towner EML, Aynsley-Green A. Annual incidence of unintentional injury among 54000 children. Injury Preven 1996; 2: 16±20. 5 Walsh SS, Jarvis SN. Measuring the frequency of `severe' accidental injury in childhood. J Epidemiol Communities Health 1992; 46: 26±32. 6 Rothman K. Modern Epidemiology. Little, Brown and Co: Boston, 1986, pp 83±84. 7 Whitehead M. The Health Divide: Inequalities in Health in the 1980s. Health Education Council: London, 1987, pp 46±50. 8 Padgett DK, Brodsky B. Psychosocial factors in¯uencing nonurgent use of the emergency room: A review of the literature and recommendations for research and improved service delivery. Soc Sci Med 1992; 9: 1189±1197. 9 Baker S et al. The injury severity score. J Trauma 1974; 14: 187±196. 10 Donaldson L, Cook A, Thomson R. Incidence of fractures in a geographically de®ned population. J Epidemiol Community Health 1990; 44: 241±245. 11 American Association for Automotive Medicine. The Abbreviated Injury Scale. Illinois, USA: American Association for Automotive Medicine, 1985. 12 WHO. Manual of the International Statistical Classi®cation of Diseases, Injuries and Causes of Death. WHO: Geneva, 1977.

13 WHO. International Statistics Classi®cation of Diseases and Related Health Problems. Tenth Revision. Vol 1. Geneva: WHO, 1992. 14 Townsend P, Phillimore P, Beattie A. Health and Deprivation: Inequality and the North. Croom Helm: London, 1987. 15 Smith R, Pless IB. Preventing injuries in childhood. BMJ 1994; 308: 1312±1313. 16 Lyons R, Lo S, Heaven M, Littlepage B. Injury surveillance in childrenÐusefulness of a centralised database of accident and emergency attendances. Injury Preven 1995; 1: 173±176. 17 Worlock P, Stower M. Fracture patterns in Nottingham children. J Pediatr Orthop 1986; 6: 656±660. 18 Mann D, Rajmaira S. Distribution of physeal and nonphyseal fractures in 2650 long-bone fractures in children aged 0±16 years. J Pediatr Orthop 1990; 10: 713±716. 19 Peterson H et al. Physeal fractures: Part 1. Epidemiology in Olmsted County, Minnesota, 1979±1988. J Pediatr Orthop 1994; 14: 423±430. 20 Cheng J, Shen W. Limb fracture pattern in different pediatric age groups: a study of 3350 children. J Orthop Trauma 1993; 7: 15±22. 21 Landin L. Fracture patterns in children: analysis of 8682 fractures with special reference to incidence, etiology and secular changes in a Swedish urban population 1950±1979. Acta Orthop Scand Supple 1983; 54: 1±109. 22 Rivara F et al. Epidemiology of childhood injuries: sex differences in injury rates. Am J Dis Child 1982; 136: 502±506. 23 Roberts I, Power C. Does the decline in child injury mortality vary by social class? A comparison of class speci®c mortality in 1981 and 1991. BMJ 1996; 313: 784±786.

247