Time trends over five decades, and recent geographical variation, in rates of childhood squint surgery in England
- Munazzah R Chou1,
- Aeesha N J Malik2,
- Mehrunisha Suleman3,
- Muir Gray3,
- David Yeates4,
- Michael J Goldacre4
- 1East Surrey Hospital, Redhill, UK
- 2Department of Glaucoma, Moorfields Hospital, London, UK
- 3Department of Health QIPP Right Care, Oxford, UK
- 4Unit of Health-Care Epidemiology, University of Oxford, Oxford, UK
- Correspondence to Dr Munazzah R Chou, Department of Emergency Medicine, East Surrey Hospital, Redhill, UK;
- Received 24 November 2012
- Revised 22 January 2013
- Accepted 30 January 2013
- Published Online First 10 April 2013
Aims To study trends in rates of childhood squint surgery in England over five decades, and to study recent geographical variation in England.
Methods Use of routine hospital statistics to analyse trends in squint surgery in the Oxford record linkage study area 1963–2010, and England 1968–2010; analysis of geographical variation in England 1999–2010. All rates, numerators and population denominators were restricted to people aged under 15 years.
Results The study included 519 089 admissions for operations on squint. Annual admission rates for squint surgery in England fell from 188.8 episodes per 100 000 population (95% CI 180.9 to 196.8) in 1968 to 64.1 (62.4 to 65.7) episodes per 100 000 population in 2010. A similar decline was seen in the Oxford region, from 213.2 (181.3 to 245.2) episodes per 100 000 population in 1963 to 61.3 (54.8 to 67.9) episodes in 2010. There was wide variation across local authorities in annual rates of squint surgery from 28.2 (95% CI 22.7 to 34.8) admissions per 100 000 population to 138.6 (123.0 to 155.7) admissions per 100 000, a 4.9-fold difference between areas with the highest and lowest rates.
Conclusions Squint surgery rates have decreased substantially over time. The current wide geographical variation in rates raises questions about whether this scale of variation is clinically warranted, whether it reflects variation in needs for surgery and patient/parental choice, whether it is a result of inequalities in the availability of ophthalmic services, or whether it results from variation between clinicians in clinical decision making about the likely benefits of squint surgery.
Strabismus is one of the most common ocular problems in children, with a reported prevalence ranging from 2.3% to 5.3%.1 ,2 A family history of strabismus has been shown to be a predictor of the condition in children, and a number of genetic susceptibility loci have been proposed.3–6 Low birth weight, prematurity, increasing maternal age and smoking during pregnancy are identified risk factors,7–10 and there is a trend of increasing prevalence with neurodevelopmental problems, multiple disabilities and decreasing social class.1 ,7
Strabismus surgery can improve alignment of the visual axes, stereopsis and dynamic visual fields; expand the binocular field; reduce ocular strain; and prevent development of abnormal head posture. Studies have shown that adults with strabismus are judged more negatively in terms of personality and employment capability.11 Improved confidence and self-esteem have been demonstrated, as well as reduced social anxiety and depression, following squint surgery.12 ,13
Identification of geographical variation in the use of services is potentially important in eliminating inefficiencies in the delivery of ophthalmological services, and in highlighting any inequity in the quality and provision of healthcare services across different areas. There are no publications of studies on a national basis, as far as we are aware, that report on geographical variation in squint surgery rates.
Ophthalmic surgery is a dynamic specialty with changes occurring in practice at a rapid pace. Our aim was to study variation in squint surgery rates to gain insights into current practice. This, as well as the time trend analyses, will provide a context from which further research and decision making about service design and provision can be carried out.
Annual rates of squint surgery in people aged under 15 years were analysed using Oxford record linkage study (ORLS) data, available from 1963 to 2010, and English national hospital statistics available from 1968 to 2010.
ORLS consists of data on hospital admissions in the former Oxford Regional Health Authority area. The population covered rose from part of Oxfordshire in 1963 (population about 350 000) to Oxfordshire and West Berkshire from 1966 (850 000 people) to the whole region by 1991 (2.5 million people). Successive admissions are linked for the same person to allow patients to be traced through multiple episodes of care. English hospital statistics from 1968 to 1985 were collected on a one-in-ten sample basis for the whole of England, in the Hospital In-Patient Enquiry (HIPE). HIPE included data on every hospital admission, except day cases, in the National Health Service in England. From 1989, English national data have been collected on a 100% basis on all NHS inpatients and day cases as the Hospital Episode Statistics (HES) dataset. In the early years covered by this study, up to 1989, the ORLS and the English data were almost independent of each other: a 10% sample of the ORLS data was contributed to HIPE. From 1989, the ORLS is essentially a linked version of the Oxford region subset of national HES.
The data were used to calculate annual, age-truncated episode-based admission rates (ie, in the calculation of the rates, all numerators and all population denominators were restricted to people aged under 15 years). They were also used to calculate person-based admission rates in the ORLS from 1963, and in English HES from 1999 (when linkable data became available). Episodes denote a single hospital admission and count each person more than once if admitted more than once. The data for the person-based analysis take account of repeat admissions for the same individual by selecting the person once only regardless of number of admissions per person.
HES was also analysed to produce average annual rates from 1999 to 2010 for each local authority (LA) area. Local authorities are standard units of local government in England.
In the analysis of trends, and of variation, we used population denominators from the Office for National Statistics for England in each year and for each local authority area. CIs for the rates were calculated.
The coefficient of variation (CV) was calculated using the formula CV=σ/µ, where σ is the standard deviation of the number of procedures carried out in each LA, and µ is the mean number of procedures in each LA. The CV provides a measure of the extent of variation, and is used to assess whether the extent of variation found in one set of data is higher or lower than that in another. Variation in squint surgery rates was compared with variation in tonsillectomy rates. This was done because tonsillectomy is an operation that is considered to show appreciable geographical variation in operation rates.
To investigate whether the use of NHS squint surgery was associated with socioeconomic status, the person-based admission rate for each LA was plotted against the Index of Multiple Deprivation (IMD) score for each LA. The IMD is a standard multidomain measure used in England as a measure of socioeconomic status. Numerators and denominators were restricted to people aged under 15 years. In addition, the area rates were age-standardised within the under 15 years age group using the age groups <5, 5–9 and 10–14 years, the indirect method of standardisation, and the total <15 England population as the standard.
The operative procedure codes used to identify records of squint surgery in the datasets were C31.1–C37.9 in the Office of Population, Censuses and Surveys Classification of Surgical Operations and Procedures, 4th Revision (OPCS4), and equivalent codes in earlier versions of the coding manual.
The study included 519 089 admissions for operations on squint. From 1968 to 2010, there were 18 719 admissions for squint surgery in people aged under 15 years in the ORLS. In English HIPE from 1968 to 1985, there were 328 670 admissions, scaled up from the 10% sample to 100%, with an average annual number nationally of 19 334 admissions per year. In English HES from 1990 to 2010, there were 171 700 admissions at a national average of 8176 admissions per year.
Annual rates of admission for squint surgery in the Oxford region showed a substantial decline from 213.2 episodes (95% CI 181.3 to 245.2) per 100 000 population in 1963 to 61.3 (54.8 to 67.9) episodes in 2010, more than a three-fold drop. National data from 1968 to 2010 showed a similar three-fold fall in rates of admission for squint surgery from 188.8 (180.9 to 196.8) per 100 000 population to 64.1 (62.4 to 65.7). The trends are shown in figure 1. The average annual rate of surgery in 5-year periods are shown in table 1. It is evident that the greater part of the fall occurred in the 1970s–1990s. Trends for squint operations within the individual age groups <5, 5–9, and 10–14 are shown in the online supplementary table.
Figure 2 shows the average annual age-standardised person-based squint surgery rate for each LA arranged in descending order. Each point represents the rate for each LA and the vertical line represents its 95% CI. It shows substantial variation in squint surgery rates across England that is statistically significant (the lower CIs of the higher rates do not overlap the upper CIs of the lower rates). The spread of rates shows that there is a graduated distribution of rates, and that the variation across the country cannot be attributed to a small number of atypically high or low outliers.
Table 2 shows data for the 10 highest and lowest rates of squint surgery in England. It shows that Easington had the highest rate per 100 000 at 138.6 (95% CI 123.0 to 155.7) people per year. Kensington and Chelsea had the lowest rate at 28.2 (22.7 to 34.8). This shows a 4.9-fold, highly significant difference between the extremes. The CV over the period of study was 76. This figure is given to provide context on the scale of variation: the higher the CV the greater the variation. For comparison with squint at 76, the CV in the same period for tonsillectomy was 84. In other words, variation for squint surgery was almost as great as that for tonsillectomy. Tonsillectomy rates, analysed in the same way as those for squint, showed variation ranging from 122.2 admissions per 100 000 children under 15 years of age in the LA with the lowest rate to 704.4 in the LA with the highest rate. This is a 5.8-fold difference, that is, only slightly greater than that for squint.
Geographical profiles of the annual rate of squint surgery, calculated as quintiles of LAs based on admission rates, are shown in figure 3.
Readmission was uncommon, and the vast majority of LAs showed ratios of numbers of admissions for squint surgery, per person admitted for it, of between 1.0 (ie, no readmissions) and 1.03. A small number of LAs had higher rates. Twenty-one of a total of 352 LAs had ratios of 1.1 or above, and were to be found spread throughout England. The 3 LAs with the highest ratios were Windsor and Maidenhead, Slough and Bracknell Forest with ratios of 1.19, 1.19 and 1.14, respectively. All these areas are within the catchment area of the same hospitals (Royal Berkshire, and Heatherwood and Wrexham Park).
The relationship between the IMD 2004 score and person-based rate of surgery in each LA was very weakly positive, that is, LAs with the highest rates of squint surgery were marginally more likely to be in areas with relatively high levels of deprivation. The coefficient of determination (R2) is 0.053. This signifies that differences in levels of deprivation can explain 5.3% of the difference between LAs in rates of surgery, which is significant (t score=4.43) but very small. In other words, 95% of the difference between LAs in admission rates for squint surgery is unexplained by deprivation.
Annual rates of childhood squint surgery in England have fallen three-fold since the 1960s. A number of different factors may account for this substantial drop. These include the possibility of a change in incidence, and/or a change in severity of strabismus. Carney et al16 showed that the incidence of esotropia reduced by 55%, from 28.3 to 12.8 per 10 000 population, between 1971 and 1991. However, a study by MacEwen et al14 reported that the incidence of esotropia, which accounts for the vast majority of childhood strabismus, had remained stable between 1986 and 1996 in Scotland and Tayside, which suggests that a reduced prevalence cannot fully explain the rate of fall that we found. Reports from the USA also show that the incidence of infantile esotropia remained stable between 1965 and 2009.17
A decline in clinical severity of strabismus has been identified from 1986 to 1996, measured by mean angle at presentation and retinoscopy findings. The mean angle at presentation fell from 30.4 dioptres in 1986 to 25.3 dioptres in 1996, while the age at presentation was unchanged.14
A switch from surgical to non-surgical treatment is likely to be an important factor in the decline in surgical rates. Full hypermetropic correction, with no subtraction for cycloplegia, became more common from the 1990s.18 Prior to this time, full correction was thought to adversely affect emmetropisation in the long term.19 MacEwen et al14 report that the mean undercorrection in 1986 was by 2.02 dioptres, compared with 0.76 dioptres in 1996.
It is also possible that there have been downward trends in the willingness of clinicians, and children and their parents, to consider surgery as the desired treatment for squint.
Another possible contribution to the falling rates of surgery is a reduced detection rate following the abolition of school eye tests. Screening of school children generally included tests at ages 5, 7 and sometimes 11 years. Following recommendations made in 2003 for the introduction of a ‘gold standard’ of testing to be carried out between the ages of 4 and 5 years,20 these school services were replaced by a national programme. There have been serious concerns, however, that the service is failing to screen all children.21
The use of botulinum toxin treatment as an alternative to surgery needs consideration. A Cochrane review in 2012 concluded that botulinum toxin treatment outcomes are comparable with surgery in acute onset esotropia, sixth nerve palsy and infantile esotropia,22 and its use, therefore, seems likely to increase. However, the role of botulinum toxin in the period covered by our data is likely to be extremely small and unlikely to have had any material effect on the overall trend.
We cannot judge the likely scale of contribution of each of the factors discussed above to the overall decline in squint surgery. Acknowledging the trend of falling rates of squint surgery in children, and understanding their causes, if possible, is important as it will contribute to the planning of service provision nationally. There are also implications for the training of ophthalmologists, with reducing exposure and fewer opportunities to achieve proficiency in strabismus surgery. This may require alterations to service design.
The paradigm clinical condition in studies of clinical variation in the historical literature has been tonsillectomy, which has shown substantial geographical variation.23 We have shown that, currently, variation in squint surgery is only marginally less than variation in tonsillectomy. As further context, the four-fold variation in annual squint surgery rates can be compared with three-fold variation in that for cataract surgery between 1998 and 2003, and a 47-fold variation in intravitreal injection rates recently.24 ,25
The wide geographical variation in squint surgery rates raises questions about whether this scale of variation is clinically warranted. The differences may, in part, reflect variation in need for surgery due to local population factors, for instance, geographical variation in recognised risk factors, such as rates of maternal smoking, maternal age, family history and associated neurodevelopmental problems.
Additionally, inequalities in the availability of ophthalmic services may be an important factor. Our findings showed that the level of deprivation did not significantly affect the rate of surgery. However, without data on the geographical distribution of squint itself, as well as on surgical rates, it is impossible to judge whether operation rates accurately reflect the needs of the population.
Part of the variation may reflect differences between clinicians in clinical decision making about the likely benefits of squint surgery and, perhaps, variation in familiarity with, and proficiency at, administering non-surgical treatments. The most recent Royal College of Ophthalmologists’ guidelines for the management of childhood strabismus sets out broad principles and management options; however, patient or parental preference influences the final management option chosen.
Geographical variation in readmission rates may reflect differences in quality of surgical technique, postoperative care, allied professional input, or a clustering of complex squint cases and increased need for ‘two-stage’ squint surgery.
Of the various explanations for geographical variation, the most important single one may be variation in clinical judgement—by general practitioners in referring patients for surgery, and by ophthalmologists—in the thresholds for recommending surgery. The current absence of good evidence for the treatment of strabismus is illustrated by the Cochrane review which failed to recruit any randomised trials for meta-analysis when assessing the effectiveness and optimal timing of surgical and non-surgical treatment options for infantile esotropia.26 This lack of an evidence base has been a reason for controversies related to strabismus surgery including age at surgery, type of procedure and the effect of surgery on amblyopia. In the absence of objective evidence, variation between clinicians in their exercise of clinical judgment is likely to become more pronounced.
We have shown substantial geographical variation in the rates of squint surgery, and have identified a number of possible causes for this. It is essential to understand whether this variation is clinically warranted to increase the evidence base for the provision of squint care, and to reduce variation if necessary.
Strengths and weaknesses of the study
This is a national, and therefore representative, study of squint operation rates in England. The data shows surgical rates, that is, they are referenced to population denominators and, therefore, take account of demographic differences between geographical areas and changes over time in the size of the child population.
The quality and consistency of HES data for squint surgery, nationally across England, are unknown. However, at a broad level of coding, squint operations are straightforward and gross error seems very unlikely. There may be error and inconsistency at the detailed level of precise operations undertaken, within the range of codes for squint, but we did not use the data at this level of detail. HES are limited by the absence of data on surgery carried out in the private sector. As private healthcare uptake has generally increased over time, the decrease in rates with data confined to the NHS may be overestimated.
Rates of squint surgery have decreased substantially over time. There is wide geographical variation in rates of squint surgery. Future work should address whether this scale of variation is clinically warranted.
Contributors MRC was involved in analysis and interpretation of data, drafted the original article and revised the paper; ANJM was involved in conception of the article, analysis and interpretation of data, revised the paper and gave final approval; MS revised the paper and gave final approval; MG gave final approval; DY wrote the programmes for the analyses of geographical variation and time trends; MJG designed the analyses, oversaw collation of the datasets and revised the paper.
Funding The Unit of Health-Care Epidemiology is funded by the English National Institute for Health Research to analyse the linked data. The views expressed in this paper do not necessarily reflect those of the funding body.
Competing interests None.
Ethics approval The study was done using anonymised data held by the Unit of Health-Care Epidemiology, University of Oxford. The Unit has ethical committee approval for a programme of work using the data for research (from the NHS North Somerset and South Bristol Multi-Centre Research Ethics Committee, reference 04/Q2006/176).
Provenance and peer review Not commissioned; externally peer reviewed.