Aims To address the absence from the public health ophthalmology literature of age- and sex-specific prevalence and related resource use for diabetic macular oedema (DMO) in England, UK.
Methods Calculation of age- and sex-specific rates from primary source clinical data, and application to the demographic structure of England to estimate the number of cases affected by DMO. A public health commissioner and provider of social care perspective was adopted in a standard cost of illness study.
Results The number of people with diabetes in England in 2010 was estimated at 2 342 951 of which 2 334 550 were aged ≥12 years. An estimated 166 325 (7.12%) had DMO in one or both eyes, and of these, 64 725 individuals had clinically significant DMO reducing the visual acuity to poorer than 6/6 in at least one eye. The overall health and social care costs in 2010, on the pathway from screening to rehabilitation and care in the home, are estimated at £116 296 038.
Conclusions The outcomes of this study should alert public health commissioners and clinical providers to the burden of DMO. The methods employed should also encourage the use of clinical ophthalmic data at the interface between local population and hospital-based recording systems.
- Diabetic macular oedema
- public health
- cost of illness
- age-specific prevalence
- public health
Statistics from Altmetric.com
- Diabetic macular oedema
- public health
- cost of illness
- age-specific prevalence
- public health
Within ophthalmology diabetic macular oedema (DMO) is recognised as an important sight-threatening manifestation of diabetic eye disease requiring swift referral for specialist care. Recent NHS research calls and position papers consider the treatment of DMO and indicate that data on the prevalence of this disease and the size of the likely present burden on the health and social care services are not available to form the basis of discussion or aid decision-making.1 2 Such information is now required within the NHS (Commissioning Support Appraisal Service, 2010 (www.sph.nhs.uk/appraisals)).
Key articles within diabetic retinopathy literature give useful broad prevalence figures for DMO within the context of trials and outcomes of diabetic retinopathy (DR) screening, but not in such detail as to be a tool for ophthalmic public health application to national or regional populations.3–5 As a response to the needs of three sectors—NHS, voluntary agencies and industry—unpublished prevalence data have been combined here with economic service charge inputs to provide a burden of disease estimate with a more detailed focus on clinically significant DMO (CSDMO).
In the move from a local to a standardised national screening programme, the Diabetic Retinopathy Screening Service for Wales (DRSSW) completed the screening and reporting on its first cohort of new patients in 2004/2005. This cohort provided the database for analyses to estimate the age- and sex-specific prevalence proportions for DMO among patients with diabetes. Attendance was 85% of the persons called (as indicated on the DRSSW website, http://www.wales.nhs.uk/sites3/page.cfm?orgid=562&pid=25081). In addition to age and sex, ethics committee approval allowed extraction of a dataset limited to the following clinical information: DR and DMO grades; visual acuity (logarithm of the minimum angle of resolution (logMAR)); scars caused by laser treatment; non-diabetic lesions; and type of diabetes treatment (grading protocol available at www.wales.nhs.uk/sites3/page.cfm?orgid=562&pid=25081). The dataset included all the 27 178 individuals newly screened. Our broad analyses of the dataset for levels of visual impairment and on prevalence of DR have been reported at the European Association for the Study of Diabetes Eye Complications Study Group (EASDEC) 2007 (www.easdec.org/pages/default.asp?id=2&sID=34&cP=175) and on the UK Vision Strategy website6 (www.vision2020uk.org.uk/ukvisionstrategy/page.asp?section=74).
Stage 1 (analyses of the Welsh data)
Persons were classified into mutually exclusive case categories (table 1) according to combinations of: (1) visual acuity level; (2) DMO status; and (3) presence of non-diabetic lesions that may cause sight loss. Age-specific prevalence proportions for the defined case categories were estimated for men and for women using the STATA statistical software package (STATA Corp, Texas, USA).
Stage 2 (estimating numbers affected in England)
The total number of people with diabetes in England and the number screened were obtained from the Department of Health Vital Signs Monitoring Return—Commissioner Based—2010 Q4 report(http://www.dh.gov.uk/en/Publicationsandstatistics/Statistics/Performancedataandstatistics/Vitalsignsmonitoring/DH_112536). The number of people <12 years and those aged 12–15 years were obtained from the Royal College of Paediatrics and Child Health: Growing up with Diabetes: Children and Young People with Diabetes in England—Project Report March 2009 (http://rcpch.hosting.opendev.net/Research/Research-Activity/Completed-Projects/National-Diabetes-Survey). Age/sex splits of adults aged ≥16 years were based on relative frequency distributions reported by the Joint Health Survey Unit (2008; http://www.ic.nhs.uk/pubs/hse08trends).
For estimation of the numbers affected by DMO in England, the age- and sex-specific prevalence proportions obtained from analysis of the DRSSW dataset were applied to the corresponding age and sex strata of the diabetic population of England (2010). The estimates for England were thus adjusted for differences between the diabetic populations of Wales and England with regard to age and sex distribution. Further adjustments for differences in the ethnic mix were also made assuming a possibly real ethnic effect: risk of DMO 1.61 times higher in ‘Black+Asian’ compared with ‘White+other’ ethnic groups. This RR was derived from the report of a large study by Gulliford et al (2010) in south London.7 Other necessary assumptions involved taking the proportions ‘Black+Asian’ as 8.0% for England (from the Yorkshire & Humber Public Health Observatory Model Phase 3) and a much smaller proportion for Wales at 1.40% (based on relativities from the Office of National Statistics (ONS) population projections split by ethnicity). Under these assumptions, the calculations showed that the ethnicity-adjusted DMO prevalence figures would be only 1.04 times higher than the unadjusted figures.
Stage 3 (methods for economic cost of illness study)
The task in hand was the construction of a health and social care direct cost of illness (COI) study for DMO for the year 2010, that is, estimates of the annual cost of DMO from the perspective of the health and social care providers. Given the paucity of published costing literature on this topic, the necessary resource use and related costing data for health were derived mainly from government and voluntary sector websites. For social care, the published literature for eye disease provided mainly rates of usage of social care, the cost of which were updated from unit costs of health and social care.8
General and disease-specific health costs from these key sources were collected. The related items of resource input were identified where possible and regrouped, and the monetary value corrected for inflation to 2010. Re-grouping took the form of categories most relevant to the DMO care pathway, that is: screening/detection; outpatient diagnostic and treatment inputs; and social care excess costs related to visual impairment from this disease.
No recommended method was found in the screening literature for a multi-disease costing. Screening detection costs were therefore calculated by multiplying the total number screened by a fraction of the cost per head screened (21.72% of £32),9 the fraction being the estimated proportion with DMO among those with DR and/or DMO.
For the sight test at the general ophthalmic service using the clinical time allocation method from the UK Vision strategy pilot method,6 the cost attributable to DMO was 0.468% of the total sight test cost of £285 428 057 for the relevant population age strata.
For the hospital diagnostic and treatment costs, the charges based on resource inputs estimated in the National Screening Handbook and updated in the Health Technology Assessment (HTA) report9 10 were further adjusted for NHS inflation to 2010. Assuming that NHS estimates are averaged across routine (70%) and complicated (30%) laser outcomes,11 the costing of the latter required an adjustment to the average cost to reflect 1.5 times higher resource input. Assumptions on the ‘false +ve’ rate (10% of all referrals) and on the proportion requiring confirmation (5% of non-CSDMO cases) were less robust in source.12 From the NHS site an outpatient appointment was taken as £73 and integrated with diagnostic/monitoring costs to give £206 per head for ‘false +ve’ and clinical confirmation cases (http://data.gov.uk/dataset/nhs-reference-costs, 2008–2009). Laser treatment and related diagnostics were costed per finished consultant episode at £844,9 10 and with the addition of travel (four visits) and one outpatient appointment, amounted to £966.60.
Travel and travel time for the patient and accompanying person were allocated an average cost of £12.40 per visit based on updated costs for glaucoma treatment.13 While this item could be seen as outside a charge to health and social services, given the age structure of the population it is included here bearing in mind government schemes for free travel.
Hospital-based low vision (LV) services and local authority LV and rehabilitation provision per partially sighted or blind person incur different levels of charge. The former service includes a hospital general ophthalmic appointment plus an LV consultation, totalling £134.40. The basis here is the fee reported by Ryan for the LV practitioner in 2005 Service Manual—Welsh LV Scheme (http://www.eyecare.nhs.uk/uploads/Recommendedstandardsforlowvisionservices.doc). The local authority service cost was set at £304.22 using the costing approach and updated rates from earlier literature.14
Estimated excess of routine health and social care service usage attributable to sight loss in general is applied to partially sighted and blind persons as enumerated in the section on epidemiology. Considerable attention has been paid to these calculations in the UK Vision Strategy Report of 2009. This work uses a combination of key sources to provide specific usage estimates for the absence or presence of age-related macular disease.6 15–18 Direct non-ophthalmic medical-related costs (per person) were £19.75 per annum.6 15 16 Average cost per blind or partially sighted person for excess residential care was £246 per annum. This included disaggregated costs from a European study, which included the UK.6 15 17 For excess paid care the average cost per annum per blind person was £9711.56 and per partially sighted person £1981.95.6 15 18
Capital costs were calculated as 2% of direct care costs, and deadweight loss as 12 pence in the pound (£) for each £1.00 of overall health and social care expenditure (method from Access Economics).6 15
Stage 1 (Welsh data analysis results)
The overall prevalence proportions for DMO and CSDMO observed in the 27 178 screened patients in the DRSSW dataset are shown in table 2. Breakdown of the prevalence proportions by age and sex are presented in table 3.
Stage 2 (numbers affected in England)
The number of persons with diabetes in England in 2010 was estimated at 2 342 951, of which 2 334 550 were aged ≥12 years (for the splits by age and sex see Table Suppl-01 in the supplemental material). The number of patients screened in 2010 was 1 634 763. The estimated number of DMO and CSDMO cases in England in 2010 are shown in table 4. Of the estimated 62 083 CSDMO-VI (corrected visual acuity of <6/6 to >6/60) cases potentially eligible for treatment, some 2385 would be permanently excluded from the screening/treatment process (because of no light perception in either eye, terminal illness, or physical/mental disability preventing either screening or treatment), leaving 59 698 referable to hospital. The number of patients referred for confirmation (section on economics method) were calculated as 3557 (5% of 71 145 non-clinically significant DMO cases). The number of ‘false +ve’ cases, estimated at 6633, was based on the assumption mentioned in the section on economics.
Stage 3 (health and social care direct costs of DMO for 2010 in England)
The estimated health and social care direct costs of DMO for 2010 in England are shown in table 5. If adjustment is made for a possible ethnic effect (DMO RR of 1.61 for the ‘Black+Asian’ group), as outlined in the Methods section, the total COI would be increased by 3.15%, from £116 296 038 to £119 959 086. Assuming a much larger RR of 2.00, the increase in COI would be 5.14%.
The objective of this study was to improve the basis for public health decision-making on healthcare provision for DMO. In applying the Welsh prevalence proportions to the population of patients with diabetes in England, our method adjusted for any differences in the age/sex distributions between the two populations with diabetes.
We also report the modest amount of increase in COI if a real effect of ethnicity is assumed. Is it necessary to make this adjustment? In our view, there is as yet no definitive epidemiological evidence for a significant effect of ethnicity that is relevant to the UK situation in respect of diabetes care and equity of access to such care in the UK. A study by Sedgwick et al found that access to the UK medical services for diabetes was generally equitable across ethnic groups.19 Increased risk of CSDMO in black and Hispanic people has been observed in studies in the US and in the Caribbean. The excess risks, however, may have been partly or largely related to the relatively poor access to (or uptake of) diabetes care services by these ethnic groups, rather than being entirely due to the ethnicity per se. In the US study,20 the observed excess risk of CSDMO was not independent of differentials in obesity, duration of diabetes and treatment. For the situation in England, there is considerable uncertainty about excess risk of DMO in African and other ‘Black’ or Asian ethnic groups. At the time of writing, the most recent pertinent study involving 31 484 subjects attending screening for DR in south London found some significant ethnic effects.7 The authors, however, considered the Asian ethnic effect as ‘weak’, and also pointed out the following: “we caution that the finding of increased maculopathy in people of African origins could be a chance finding that requires confirmation in future studies. There is also potential bias if rates of non-attendance for screening differ among ethnic groups and according to retinopathy status.”
The economic COI is presented but does not address societal costs such as informal care and lost productivity, though this is addressed for DR in the Vision Strategy document.6 15 The detailed innovative method for general ophthalmic service allocation is also addressed there. Concepts such as dead weight loss and capital (administration) costs are added here because they are now considered best practice in costs of studies on blindness internationally.6 15 21 The excess care costs relating to those persons with sight loss come mainly from those with age-related macular disease, but we have no reason to suppose that the loss of well-being is less in those whose vision is impaired from CSDMO.
For those who consider the assumption of one-third of patients treated for CSDMO having outcomes that would require additional laser inputs as not up-to-date or uniform, this costing can be treated differently. As an alternative we took the suggested usage rates from a report on the north-east of England2 and adopted the lower end of that projection for laser treatment sessions (ie, two at £561 each) and allowed travel costs for four related visits. This would provide higher costs in this project (£69 942 177 vs £65 594 792), but is an alternative measure of average charges without specifying differential outcomes.
We hope that these epidemiological and costing adaptations will encourage decision-makers in ophthalmology services to use our calculations and to take the work presented here further, allowing for local circumstances.
Funding Novartis UK provided funding and additional economist support for the cost of illness aspect of this study.
Ethics approval The epidemiological analysis of the initial DRSSW dataset was carried out under the auspices of the University of Cardiff Diabetes Research Unit, with ethical committee approval.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.