Risk of subarachnoid haemorrhage in first degree relatives of patients with subarachnoid haemorrhage: follow up study based on national registries in Denmark

Nationwide registries

We performed a follow up study based on information from two nationwide registries: the Danish National Discharge Registry and the Danish Central Person Registry. We identified patients with subarachnoid haemorrhage in the discharge registry and identified their first degree relatives in the central registry. Information on essentially all (99.4%) discharges from Danish hospitals has been recorded in the discharge registry since 1977.9 Recorded data include the dates of admission and discharge, up to 20 discharge diagnoses, hospital and department codes, and the surgical procedures performed. Discharge diagnoses were classified according to the Danish version of the ICD-8 (international classification of diseases, 8th revision) for 1977 to 1993 and the ICD-10 (10th revision) for 1994-5. A 10 digit civil registration number, which is permanent and unique to each Danish citizen, is also recorded, enabling the compilation of patient specific discharge histories. Data were retrieved on all persons listed with a diagnosis of subarachnoid haemorrhage (ICD-8 codes 430; ICD-10 codes I60) in the discharge registry from 1 January 1977 to 31 December 1995. We identified index episodes, defined as the first registered admission to hospital with a diagnosis of subarachnoid haemorrhage within the study period. Coding of the index episode as a recurrent or previous episode of subarachnoid haemorrhage led to exclusion. We also excluded patients with concurrent codes for cerebral arteriovenous malformations (ICD-8 codes 430.01 and 430.91; ICD-10 codes Q28.2 and Q28.3). In Denmark, suspicion of subarachnoid haemorrhage usually leads to rapid transfer of patients to neurosurgery or neurology departments for more advanced diagnostic investigation and treatment. We therefore classified patients according to the most specialised department they had been admitted to or transferred to within 21 days of their first admission with diagnosed subarachnoid haemorrhage: neurosurgery (most specialised), neurology, or “other department” (least specialised) according to the register data.

Since 1 April 1968 all persons residing in Denmark have been registered in the Danish Central Person Registry with their civil registration number, date of birth, residency, date of immigration or emigration, and date of death. For each individual the civil registration numbers of the parents are also recorded. The coverage of information on the parental civil registration numbers, however, increased over time from less than 10% for cohorts born before 1952 to 43% in the 1952 cohort, 96% in the 1959 cohort, and 99% for cohorts born after 1970.10 Adopted children cannot be identified as such in the registry and are recorded under their adopted parents, but the magnitude of this misclassification is small.10

We retrieved data on vital status from the registry for the period 1 April 1968 to 31 December 1995 for members of the subarachnoid haemorrhage inception cohort and their first degree relatives who could be traced in the central registry. For each group of first degree relatives, the person discharged with a diagnosis of subarachnoid haemorrhage on the earliest date, designated the index case, was identified. First degree relatives living in Denmark on the date of admission of the index case were followed up as of that date and until the first admission for subarachnoid haemorrhage, emigration, death, or end of study period, whichever event came first. Incident cases of subarachnoid haemorrhage that occured among first degree relatives during follow up were designated first degree relative cases.

Validity of the diagnosis

We studied the validity of a register diagnosis of subarachnoid haemorrhage in a sample from hospitals in Funen County (465 000 inhabitants; 9% of the Danish population). Only hospitals with more than 10 registered patients in the study period were used—that is, a university hospital with a neurosurgery and a neurology department, and two other hospitals. Discharge records on 210 patients were retrieved for review. Data from discharge abstracts, medical records, or autopsy reports were abstracted to a structured form and evaluated, blinded with respect to family relationship, by an experienced neurologist (IT) according to predefined criteria. A register diagnosis of subarachnoid haemorrhage was accepted when a typical clinical history and at least one of the following were described: computed tomography findings compatible with blood in the subarachnoid space, xanthochromic cerebrospinal fluid, or subarachnoid haemorrhage or aneurysm established by angiography or autopsy. Medical records could be traced in all but one case. All cases marked as recurrent or previous episodes of subarachnoid haemorrhage in the register (18 patients) were also recorded as such in the medical documents. In the remaining 191 cases the positive predictive value of a register diagnosis of subarachnoid haemorrhage was 93% (95% confidence interval 85% to 98%) for patients discharged from the neurosurgery department and 75% (60% to 87%) for patients discharged from the neurology department. For patients discharged from other non-specialty wards, the predictive value of a register diagnosis of subarachnoid haemorrhage was low: 47% (36% to 59%). We also validated the register diagnosis of subarachnoid haemorrhage for patients from families with more than one case of subarachnoid haemorrhage identified in the nationwide material (37 patients). The predictive value of a register diagnosis of subarachnoid haemorrhage in these cases, which were mostly admitted to neurosurgery wards, was 92% (78% to 98%). Overall the misclassification may thus be less than 20%.

Familial subarachnoid haemorrhage is reported to be associated with congenital heart disease11 and several rare hereditary disorders, such as polycystic kidney disease and Ehlers-Danlos syndrome,4 which we wanted to exclude from the study. The retrieved medical records made no mention of such disorders. Furthermore, for the 37 familial cases we scanned all discharges in the register under any diagnosis for these persons and found no discharges with diagnostic codes corresponding to the above mentioned disorders.

Statistical analyses

We calculated age specific and sex specific incidence rates for the entire Danish population for each of the periods 1977-81, 1982-6, 1987-91, and 1992-5. Twenty age categories (5 year intervals) were used. Appropriate population denominators (person years at risk) were derived from annual population counts (1 year age categories for each sex) obtained from Statistics Denmark.12 To study the incidence rate of subarachnoid haemorrhage in first degree relatives the follow up data were aggregated in a multiway person year table. The main table was based on sex, age (5 year intervals), and calendar period (four periods). To study the effect of follow up time (time since admission of index case categorised as less than 1 year, 1-7 years, and more than 7 years), type of relative (parent, sibling, child), and the sex of the index case, three additional person year tables were made based on each of these factors and including age, sex, and calendar period. The table including sex of the index case was restricted to children of the index cases.

Time at risk for first degree relatives was calculated from the date of admission of the index case until the first admission for subarachnoid haemorrhage, emigration, death, or end of study period, whichever event came first. The incidence rate of subarachnoid haemorrhage in first degree relatives of patients with subarachnoid haemorrhage was compared with that of the entire population using a Poisson regression model. The results were summarised by standardised incidence ratios and 95% confidence intervals.13 The main table was used to assess the effect of sex, age, and calendar period. These factors were examined one at a time because the limited number of cases did not allow a comprehensive analysis of interactions. The additional tables were used to evaluate the effects of follow up time, type of relative, and sex of index case. To study the effect of potential misclassification—that is, labelling registered cases (which in fact were prevalent cases) as incident cases, we successively restricted the study period by letting it begin in 1982, 1987, and 1992.

Misclassification of diagnoses in the discharge registry results in inclusion of non-cases in the analyses, which leads to an underestimate of the relative risk. Neurosurgery units had the lowest proportion of misclassification of cases according to our validation study. We therefore repeated all analyses after restricting the sample to patients (including index cases and first degree relative cases) discharged from neurosurgery units. The study was approved by all seven regional ethics committees and the Danish registry board.

Potential limitations of the study

Our study has several potential weaknesses. Prevalent cases, surviving a first admission to hospital for subarachnoid haemorrhage before 1977, could be misclassified as incident cases if admitted for a recurrent episode of subarachnoid haemorrhage during the study period. We addressed this problem by excluding patients with a register code at the index admission to hospital, indicating a recurrent episode. The high case fatality rate and the robustness of our results to progressively left truncating the time window indicate that the extent of the problem is minor.

The coverage of family relations by the central registry increased over time, which could affect our results. However, restricting the material to families identified during the period with essentially full coverage by the central registry had little influence on the estimates (standardised incidence ratio 3.0). The identified first degree relatives were mainly offspring and often did not reach the age at which subarachnoid haemorrhage incidence peaks during follow up. This may result in an underestimate of the relative risk of subarachnoid haemorrhage in first degree relatives, in particular for siblings, which have been reported to be at a higher risk than other first degree relatives.7 Longer follow up of the cohort will disclose whether this is the case. The registration of adopted children under their non-biological parents in our data also tends to dilute the observed association, but the magnitude of this problem is small.10

It is conceivable that relatives of patients with subarachnoid haemorrhage would have a lower threshold for seeking medical help if they experience symptoms that could be attributed to subarachnoid haemorrhage. Review of the discharge records and the high case fatality rate observed in this group preclude this. Our study did not include patients with subarachnoid haemorrhage who died before admission to hospital. This, however, happens only in some 11% to 14% of all patients with subarachnoid haemorrhage.18 Familial cases have not been reported to be more likely to belong to this group.