Methods
The National Health Service (NHS) provides the majority of healthcare in England, including ophthalmology and medical day-case and inpatient care; this is particularly true for the long-term care of chronic conditions such as dementia and POAG.
As in the previous analysis, the complete data set of linked English national hospital episode statistics (HES), from 1999 to 2011, was used for this study. English national HES are data collected by the NHS Information Centre as a statistical database of demographic, medical and administrative information about all admissions to NHS hospitals in England and admissions funded by the NHS for treatment in non-NHS clinical organisations. Record linkage was undertaken by the Oxford record linkage group. Linkage means that data relating to successive episodes of care for each person are brought together, so that data about individuals can be analysed across multiple episodes of care (eg, for POAG and subsequently for AD).
A POAG cohort was constructed by identifying all people in the linked HES data set aged 55 years or more who had a record of an admission or day case care where POAG was recorded (International Classification of Diseases Revision 10 (ICD10) code H40.1). For each person, the earliest known record of POAG was the record used in the analysis. A reference cohort was constructed by identifying individuals without POAG who were admitted to hospital for various other, mainly minor medical and surgical, conditions (listed in Table 2 footnotes). Unlike the POAG diagnosis, which could be recorded in any diagnostic position on the HES abstract, these conditions were required to be recorded in the main diagnostic position on the hospital record for the individual to be admitted to the reference cohort. People were only included in the POAG or reference cohort if they did not have an admission for AD or vascular dementia either before or at the same time as the admission with POAG or the reference condition.
We then searched the database for subsequent NHS inpatient or day case care for AD (ICD10 codes F00, G30) or vascular dementia (ICD10 code F01) in these cohorts. We considered that rates of care for dementia in the reference cohort would approximate those in the general population of the region while allowing for migration in and out of it (data on migration of individuals were not available).
The reverse analysis, that is, POAG hospital diagnoses in people aged 55 years or over with AD or vascular dementia, was performed using similar methods. In this analysis, people were excluded if they had a record of POAG before, or at the same time as, admission for dementia or reference condition. This ensured that no one was counted twice. The statistical methods used were the same for each analysis. We describe these methods as follows, taking AD in people with POAG as the example.
Rates of people with AD in each cohort were calculated based on person-days 'at risk'. Date of cohort entry was the date of earliest recorded admission for POAG or reference condition, subject to eligibility as described above; date of cohort exit was the date of earliest known record of AD, death or the end of the data collection period (31 December 2011), whichever was the earliest. We adjusted for age in 5-year bands, sex, calendar year of admission, region of residence and Index of Multiple Deprivation (IMD) score—a standard English measure of socioeconomic status—stratified into quintiles.
Rate ratios were calculated by taking the standardised rate of occurrence of dementia in the POAG cohort relative to the reference cohort. Expected numbers of people with AD in each cohort were calculated by applying the rates of AD in the combined population of the POAG and reference cohorts to each individual cohort separately. We did this within each age-sex-year-region-IMD stratum and then summed the stratum-specific expected numbers to obtain expected totals for each cohort. The formula (O/E)/(O/E), where O and E are the observed and expected numbers of dementia cases in the POAG and reference cohorts, respectively, is equivalent to the rate ratio. The CI for the rate ratio and χ statistics for its significance were calculated as described elsewhere.