Article Text

Download PDFPDF

Subarachnoid haemorrhage in Sweden 1987–2002: regional incidence and case fatality rates
  1. H Koffijberg1,
  2. E Buskens1,2,
  3. F Granath3,
  4. J Adami3,4,
  5. A Ekbom3,
  6. G J E Rinkel5,
  7. P Blomqvist3
  1. 1
    Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, Utrecht, The Netherlands
  2. 2
    Department of Epidemiology, University Medical Centre Groningen, Groningen, The Netherlands
  3. 3
    Department of Medicine at Karolinska Solna, Clinical Epidemiology Unit, Karolinska Institutet, 171 76 Stockholm, Sweden
  4. 4
    Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, 171 77 Stockholm, Sweden
  5. 5
    Department of Neurology, Rudolf Magnus Institute of Neuroscience, University Medical Centre Utrecht, Utrecht, The Netherlands
  1. H Koffijberg, MSc, University Medical Centre Utrecht, Julius Centre for Health Sciences and Primary Care, STR 6.131/7.113, P.O. Box 85500, 3508 GA Utrecht, The Netherlands; h.koffijberg{at}


Background: Incidence estimates of subarachnoid haemorrhage (SAH) in Sweden vary, which may be caused by regional variations. Reliable estimates of age-specific case fatality rates are lacking. We analysed regional incidence rates and case fatality rates of SAH in Sweden.

Methods: The Swedish Hospital Discharge and Cause of Death Registries from 1987 to 2002 yielded data on 18 443 patients with SAH. Incidence and case fatality rates by age, gender, region and time period were calculated by Poisson regression.

Results: The incidence rate was 12.4 per 100 000 person-years (95% CI 12.2 to 12.6) and increased with age, from 6.4/100 000 person-years in patients who were 30–39 years old to 25.8/100 000 person-years in patients who were older than 80 years. Incidence was higher for women (14.4 (95% CI 14.2 to 14.7)) than for men (10.3 (95% CI 10.3 to 10.6)), and higher in the north than in the south (RR 1.31 (95% CI 1.25 to 1.37)). This geographical gradient was more evident in women (RR 1.41 (95% CI 1.33 to1.49)) than in men (RR 1.23 (95% CI 1.15 to 1.33)). The 28-day case fatality rate was 31.7% (95% CI 31.0 to 32.3). It increased with age from 18.1% (95% CI 16.0 to 20.3) in patients who were 30–39 years old to 57.6% (95% CI 55.2 to 59.9) in patients over 80 years, then levelling off. Over time (1995–2002 compared with 1987–1994), the incidence rate decreased (RR 0.93 (95% CI 0.90 to 0.96)) and case fatality rate decreased (RR 0.89 (95% CI 0.85 to 0.93)).

Conclusions: SAH incidence rates in Sweden increase from south to north, more in women than in men. Octogenarians have a quadrupled incidence and a tripled case fatality compared with young adults. During 16 years, both incidence and case fatality have decreased.

  • subarachnoid haemorrhage
  • incidence
  • case fatality
  • Sweden

Statistics from

The worldwide incidence rate of subarachnoid haemorrhage (SAH) approximates 8 per 100 000 person-years,1 but varies between countries, with much higher rates in Finland and Japan.2 3 Geographical differences within countries have been reported for incidence rates of stroke in general and coronary heart disease,4 5 but detailed analyses of within-country incidence rate differences are lacking for SAH.

Previous studies on SAH incidence in Sweden have reported rates varying from 6.4/100 000 person-years for men and 13.5/100 000 person-years for women in southern Sweden,6 up to 13.3/100 000 person-years for men and 24.4/100 000 person-years for women in northern Sweden (age-adjusted to the world population, among patients 25 years and older).7 Approximately 75% higher SAH attack rates were found in northern Sweden, compared with central Sweden.8 This suggests that incidence rates vary by geographical region.

Currently, the proportion of patients admitted with SAH over 75 years of age is around 10%.9 As SAH incidence rates have been thought to reach their maximum around the fifth decade of life, relatively little attention has been paid to SAH in the elderly, and studies of SAH did sometimes not include patients older than 65 or 75 years at all.1

Recent studies of SAH incidence and case fatality rates were typically based on relatively small populations with limited observation time.1012 Therefore, little detailed information is available on elderly patients with SAH and it is often based on few cases. As the number of elderly patients with SAH increases because of an ageing population, and more accurate reporting, an accurate evaluation of prognosis and case fatality for elderly patients becomes increasingly relevant.6 1315

At present, case fatality rates for SAH are often reported on population level. If case fatality rates are provided per age category, the numbers are usually small, especially in the older age groups, thus rendering uncertain estimates.1012

We performed a detailed analysis of SAH incidence by geographical region, for men and women separately, and of age-specific case fatality using national health statistics from Sweden covering the entire population during the period 1987–2002.


Data sources

The Swedish hospital discharge register (SHDR)

The SHDR was established in 1964, and attained full national coverage in 1987. It comprises all public inpatient care. For each hospital discharge, date of admission and discharge, up to 10 surgical procedure codes, a hospital identification code, patient gender, date of birth and county of residence are recorded. One primary (main) diagnosis, and a set of secondary diagnoses indicating co-morbidities, are recorded.16

We identified all records in the SHDR between 1987 and 2002, with a diagnosis of SAH using ICD codes 430 (ICD-9 from 1979 to 1996) and I60 (ICD-10 from 1997 onwards). Every Swedish resident has a unique personal identification number that was used to retrieve each person’s first SAH discharge.

Due to incomplete coverage of the SHDR before 1987, the first primary SAH diagnosis found may, in fact, in some counties with late implementation of the SHDR, represent a recurrent SAH event or sequelae thereof. To minimise the impact of this potential selection bias, we restricted the study period to patients who were admitted with SAH as a primary diagnosis between 1 January 1987 and 31 December 2002. All SAH records had the personal identifiers but 14 records had no county information and were excluded. The SHDR study base consisted of 16 329 patients with SAH.

The Swedish cause of death register (SCDR)

The SCDR registers underlying and contributing causes of death, date of death, gender, date of birth and county of residence for all deaths among Swedish citizens.17 The same personal identifiers as in the SHDR are used in the SCDR. We selected cases with SAH as underlying cause of death by using the same ICD codes as in the SHDR. In 134 records, either the month or day of death or both were missing. In these, the day 15 and the month 6 were imputed.

The SCDR study base comprised 5 583 patients with SAH as main cause of death between 1 January 1987 and 31 December 2002. Of these, 3 469 persons (62.1%) were also registered in the SHDR. The remaining 2 114 (37.9%) persons were found only in the SCDR.

The SAH cohort

After linking data from the two registries, we retrieved data on 18 443 patients with a discharge diagnosis or underlying cause of death of SAH, or both. No inconsistencies in data regarding gender, date of birth or county were found. The mean age of the patients at the time of the haemorrhage was 58.9 years (SD 16.3 years), and 10 858 (58.9%) were women.

Data analysis

To analyse geographical differences, we aggregated the patients by domicile counties into three regions: southern, central and northern. Age was categorised as: 0–29 years, 30–39 years, 40–49 years, 50–59 years, 60–69 years, 70–79 years, 80 years or older. The 16-year observation period was divided into two 8-year periods.

Calculation and comparisons of incidence rates

The source population consisted of all Swedish citizens during 1987–2002 with 148.8 million person-years at risk, compiled from annual census data by gender, age category and region. Incidence rates were compared after direct standardisation to the years 1987–2002—that is, the complete observation period. The standardised incidence rate was modelled by applying a Poisson regression, and relative incidence rates were presented together with 95% confidence intervals (CIs). Confidence intervals for incidence rates were calculated based on the assumption of Poisson distributed number of cases. The 2 114 SAH events for persons recorded only in the SCDR were included in the incidence rates.

Calculation of case fatality rates

The time at risk of dying was calculated as the difference between the day of the haemorrhage and the date of death, and case fatality rates were computed at 0, 2, 7 and 28 days. Patients retrieved only in the SCDR, and those in the SHDR with registered causes of death and identical days of admission and discharge, were considered to have survived zero days.

To enable a more detailed analysis of elderly patients, a separate analysis was performed for age categories 70–74 years, 75–79 years, 80–84 years, 85–89 years, and 90 years or older.


Incidence rates

The crude incidence rate was 12.4/100 000 person-years (95% CI 12.2 to 12.6), and was higher in women (14.4/100 000 person-years; 95% CI 14.2 to 14.7) than in men (10.3/100 000 person-years; 95% CI 12.2 to 12.6) (RRadj 1.30 (95% CI 1.26 to 1.34)) (table 1).

Table 1 Number of SAH episodes and person-years obtained from the registries, by gender, period, age category and region

The incidence rates increased by latitude from 11.4/100 000 person-years (southern region—reference), over 12.8/100 000 person-years (central region—RRadj 1.14 (95% CI 1.10 to 1.18)) to 15.2/100 000 person-years (northern region—RRadj 1.31 (95% CI 1.25 to 1.37)) (fig 1). The south–north gradient was mainly attributable to women. With the southern region as reference, the risk of SAH in the central region was RRadj 1.15 (95% CI 1.11 to 1.20) for women and RRadj 1.07 (95% CI 1.02 to 1.12) for men. In the northern region, the risk was RRadj 1.41 (95% CI 1.33 to 1.49) for women and RRadj 1.23 (95% CI 1.15 to 1.33) for men. Similar north–south gradients were detected within regions when SAH incidence was analysed by county (fig 2) (data not shown).

Figure 1 Crude incidence rates of SAH in Sweden, divided by region, separately for men (A) and women (B). SAH, subarachnoid haemorrhage.
Figure 2 The relative risk of SAH per county in Sweden, adjusted for gender and age categories, with the most southern county Skåne as reference county. Indicated are the three regions: Götaland (south), Svealand (central) and Norrland (north), and for each county the county code and name. SAH, subarachnoid haemorrhage.

The incidence increased with age but levelled off at 60 years of age in men and at 70 years of age in women in all regions (fig 1). These differences were significant in a combined model, considering gender, age category and region (table 2).

Table 2 The relative risk of SAH by region, gender, age category and period, and the (standardised) SAH incidence rates per 100 000 person-years (95% CI in brackets)

The incidence was lower in the last period than in the first period, with a decrease from 12.7/100 000 person-years to 11.3/100 000 person-years (RRadj 0.93 (95% CI 0.90 to 0.96)). This overall decrease with calendar period comprised a decrease in incidence among persons below the age of 70 years and an increase in persons of 70 years or older. The incidence rates below the age of 70 years were 11.0/100 000 person-years in the first period (95% CI 10.7 to 11.2) and 9.7/100 000 person-years in the second period (95% CI 9.5 to 10.0). The incidence rates for 70 years and older were 24.4/100 000 person-years in the first period (95% CI 23.4 to 25.5) and 28.4/100 000 person-years in the second period (95% CI 27.4 to 29.4). These changes with calendar period were consistent for both men and women, all age categories and in the three regions.

Case fatality rates

Of the 18 443 SAH episodes, 2114 (11.5%) episodes were identified based on death certificates only, indicating that death occurred before reaching the hospital. The overall case fatality rate at 28 days was 31.7%, which increased sharply with age (fig 3). It was higher in women (32.5%) than in men (30.5%) (RRadj 1.06 (95% CI 1.02 to 1.11)), and significantly lower in the last time period, with an absolute risk reduction of 3.7% and a RRadj of 0.89 (95% CI 0.85 to 0.93); this was identical across the three regions. In patients below the age of 70 years, this absolute risk reduction was 5.8%, whereas in patients of 70 years or older it was 3.5%.

Figure 3 The case fatality rates of SAH in patients, for each age category. SAH, subarachnoid haemorrhage.

For the subgroup of 5062 patients of 70 years or older, the case fatality rates according to the 5-year age categories are shown in table 3. The case fatality rate at 1 year continued to increase until the age of 80 years, and then levelled off to about 70%.

Table 3 The case fatality rates of SAH in patients over 70 years old


Using nationwide registries, we monitored the Swedish population (9 000 000 persons) during 148 million person-years of follow-up and collected information of more than 18 000 patients with SAH. The incidence increased by latitude from southern, over central to northern Sweden, and similar gradients were found within regions. This gradient was more apparent in women in all age categories. Incidence increased with age until the age of 60 years for men and 70 years for women. The case fatality rate increased with age until the age of 80 years. Incidence increased over time, for the age category 70 years and older, and case fatality rates decreased over time, for all ages, regardless of gender.

Our findings on incidence are consistent with previously reported incidence rates in southern and northern Sweden.6 7 Our overall 28-day case fatality rate of 31.7% is slightly lower than the 30-day case fatality rate of 38.0% that was reported in Denmark18 and the 3-month case fatality rate of 37% that was reported in Japan.12 It is also lower than the case fatality rates of most of the studies discussed in a previous systematic review on SAH.19 In this review, the authors found a decrease in case fatality of 0.9% per year and suggested that improved management of patients with SAH might explain this decrease. This suggestion is confirmed by our data and is an explanation for the relatively low case fatality rate that we found compared with previous studies. The observed 11.5% of all SAH episodes resulting in death of the patient before reaching the hospital is consistent with the estimated risk of sudden death of 12.4% derived in a meta-review study of 18 population-based studies.20

The strengths of this study are the large number of patients included in the analyses, the long follow-up period available and the completeness of the medical data, enabling us to obtain precise estimates even in detailed subgroups of patients. Furthermore, a high proportion of patients with the disease is admitted to hospital because of the serious clinical symptoms of SAH and the (often) young age at which it occurs, which make hospital discharge data a valuable tool for monitoring SAH.21 This holds especially in countries with universal healthcare, such as Sweden. As for the validity and accuracy of the registry data, these have been shown to be acceptable.22 Performing another validation study from among decades of registry data recorded all over the country would require a major effort, and would hardly add to the credibility of our results. Although, in practice, SAH cases might be miscoded occasionally (eg, as spontaneous intracerebral haemorrhage), there is no reason to assume that coding problems differ between regions or time periods. Miscoding will therefore not have influenced the internal validity of our study. However, some aspects of the diagnosis of SAH, and the data in particular, were not taken into account.

First, we included all diagnoses of SAH, not only the instances with confirmed aneurysms. When trauma is excluded, 10–15% of all SAH cases are of non-aneurysmal origin and generally have a better prognosis.23 Traumatic SAH is expected to have had trauma as primary diagnosis, in almost all cases, although in rare instances it may be difficult to disentangle aneurysm and trauma as the cause of the SAH.24 25 Second, SAH may be diagnosed more accurately and frequently in more recent years, as diagnostic tools have become more advanced. Incidence rates may seem to increase as more cases are found, whereas, in fact, this increase is caused by detection bias—that is, the number of incident cases is actually stable over time but more cases have been detected in recent years.1 2 The additional cases found, in general, are probably milder cases of SAH, which could partly explain the decrease in case fatality rates over time. However, more accurate diagnoses may also cause cases that were previously diagnosed incorrectly as SAH, to be differentiated from SAH and diagnosed correctly, which ultimately results in a lower number of cases.1 During the entire period 1987–2002, CT scans were widely available, and there was minimal change in diagnostic work-up;2 therefore, it is unlikely that detection bias has played an important role. Third, using death certificates in our analyses may lead to both underestimation and overestimation of the incidence rates. Underestimation is likely to occur, as the actual cause of death may be hard to determine for persons dying from SAH before reaching a hospital. Overestimation may occur, as patients dying in a nursing home some time after an episode of SAH are still allocated SAH as main cause of death, when the actual cause is not apparent. The net effect of overestimation and underestimation, however, seems to be negligible as our findings corroborate recent existing literature.13 18

Fourth, age was the only risk factor of SAH that we adjusted for, as data on risk factors such as smoking, hypertension and excessive alcohol intake26 were not available. Other studies found a decrease in smoking prevalence, and an increase in alcohol consumption and body mass index in Sweden, in the past two decades,27 28 and relationships between socioeconomic status and stroke incidence rates.8

Fifth, we only used the primary diagnosis from the SHDR and the main cause of death from the SCDR for the selection of SAH episodes. Genuine episodes of SAH may have had SAH as a secondary diagnosis or an additional cause of death only, leading to underestimation. However, given the severity of symptoms of most instances of SAH, this source of underestimation seems unlikely. Furthermore, the incidence rates reported may represent a slight overestimation as some episodes might be recurrent events that have been incorrectly assessed as first events. Given the low absolute risk of SAH and the low risk of recurrence (1.3% in, on average, the first 8 years following the first episode),7 this overestimation will be very small.

Currently, it is unclear what causes the regional differences in incidence rate and the decrease in incidence rate for age categories under 70 years. Additional information on the distribution of risk factors, over individual counties and within specific age categories, may clarify these findings.

The increase in incidence rate for the age category 70 years and older, might be explained by age-dependent detection-bias—that is, a more pronounced tendency to examine even very old patients by CT in the later period compared with the first period. The reduction in case fatality in the later period can have many possible causes, with the introduction of coiling and an improvement of post-SAH care, in general, as the most obvious ones.

Previous studies have reported increasing case fatality with increasing age, which may in part be explained by a less aggressive approach in elderly patients with SAH. However, our data show that, even in very old patients, a substantial proportion survives the haemorrhage. Moreover, one in every three elderly patients who survive the initial weeks after the haemorrhage regain independence in daily life.9 Accordingly, adequate treatment of elderly patients with SAH is essential. As regional differences in case fatality rates are small,19 the results on case fatality of SAH in Sweden presented are likely to be valid also in most other developed countries.



  • Funding: There are no financial or personal relationships with other people or organisations that could inappropriately influence our work. This study was not funded.

  • Competing interests: None.

Request Permissions

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.