Mortality patterns and risk among older men and women with intellectual disability: a Swedish national retrospective cohort study

This study used a retrospective cohort design, in which we constructed two cohorts: older men and women aged 55 years and older with ID from two national registers, and a matched control population drawn from the population register. The cohorts were followed up during 2002–2015 to observe subsequent mortality as recorded in the death register.

Information about individuals with ID in Sweden is available from two national registers: (i) the NPR and (ii) the LSS register based on the Swedish Act concerning Support and Service for Persons with Certain Functional Impairments [13]. The Swedish National Board of Health and Welfare maintains both registers.

The NPR includes all inpatient hospital admissions since 1987 and outpatient care since 1992. It is based on mandatory registration for all 21 county councils in Sweden. In this study, we identified all older adults aged 55 years or older who were recorded in the NPR from 2002 to 2015 as having a diagnosis of ID. ID diagnoses were based on the International Classification of Diseases, Tenth Revision (ICD-10), and included the diagnosis categories: mental retardation (F70–73, F78–79), disorders of psychological development (F84, F88–89) and DS (Q90). We excluded diagnosis codes for F84.1 (atypical autism) and F84.5 (Asperger syndrome) to ensure a well-defined study group. Only the first event of hospitalisation was extracted for each individual. We recorded information on the year of hospitalisation, birth year, and sex.

The LSS register started in 2004 and is based on mandatory yearly registration from all 290 municipalities in Sweden. In 2007, the register started recording three different types of disability. In this study, we included all individuals with ID, autism and conditions resembling autism who had received any LSS services. We focused on all adults aged 55 or older who received LSS services from 2007 to 2015. As with the NPR, we only identified the first time an LSS service was used, and extracted information on the year of first access to LSS services, birth year, and sex.

Figure 1 shows how the two extracted datasets were merged into single dataset, resulting in 15,289 observations for unique individuals. Statistics Sweden matched each individual with ID in the dataset with five controls with the same sex, birth year, and year of first hospitalisation/access to LSS services, identified from the Swedish population register. All individuals were alive on January 1, 2002, the date chosen as the starting point of the observation. We calculated the age for all individuals in the dataset in 2002. The youngest individuals were aged 42 years in 2002 (they were either hospitalised or accessed LSS services when they were 55 years old in the last year of follow-up; i.e. 2015).

The dataset was linked with data from the Swedish National Cause of Death register, maintained by the National Board of Health and Welfare, to identify deaths among individuals included in the study from 2002 to 2015. Data extracted from this register and merged with the datasets described above were: death date, age at death, underlying/main cause of death, contributing causes of death, marital status, and country of birth (Swedish- or foreign-born). Marital status was coded as married or not married (including those who were single, divorced or widowed).

In total, 13,102 deaths (4738 among individuals with ID and 8364 in the control population) were identified for the 91,734 individuals included in the study (cohort with ID and control population). We recoded detailed cause-of-death data to obtain causes of death at an ICD-10-chapter level. For 604 deaths, the underlying/main cause of death was recorded as one of the ICD-10 diagnosis codes for ID: F70–73/F78–79, mental retardation (n = 180); F84–89, disorders of psychological development (n = 4); and Q90, DS (n = 420). As we did not regard ID as a cause-of-death, we replaced the underlying/main cause of death for these 604 individuals with diagnosis codes for contributing causes of death that were not the ID diagnosis codes specified above. No other contributing causes of death were available for 60 of the 604 individuals, and the cause-of-death data were recoded as an ‘unclassified diagnosis’ for these individuals.

The final dataset included 374 deaths (2.9% of all deaths) with an unclassified cause of death (166 deaths among individuals with ID and 208 deaths in the control population). DS was recorded as the cause of 942 deaths among individuals with ID (604 underlying as described above and 338 contributing cause of death). The remaining 7422 deaths among individuals with ID were attributed to other causes.

We defined the follow-up period as from 1 January 2002 to either 31 December 2015 or the date of death for those who died during the study period (2002–2015). In total, 389 individuals moved out of Sweden during the study period and were censored. For 18 individuals without a specific date or month of death, we imputed the date of death as 1 July of the respective year. For 65 individuals without a specific date of death, we imputed the date of death as 15th of the respective month of death.

Differences in demographic characteristics and death-related data between the two groups were compared in a bivariate analysis using chi-square tests for proportions. Differences in the mean age of death and follow-up period (years) between the two groups were assessed using two-sample t-tests. We conducted a sub-sample analysis to assess differences in marital status and nationality between individuals who died in the two groups. We calculated the proportionate mortality (%) and cause-specific mortality rate (per 100,000 population) in each group and among individuals with DS. We estimated the cause-specific mortality rate ratio between individuals with ID and the control population using conditional fixed-effect Poisson regression. We conducted conditional Cox proportional hazards regression with a robust variance estimator to assess if overall and cause-specific mortality differed between the two groups. A p-value below 0.05 was considered statistically significant. All analyses were conducted using Stata Version 13 [28].

The crude mortality rate was 2483 per 100,000 people among individuals with ID and 810 in the control population (Table 1). This yielded an unadjusted incidence rate ratio of 3.07 (95% confidence interval [CI] 2.96–3.18), indicating the mortality rate among individuals with ID was three times higher than in the control population. Individuals with DS had an 11-fold higher mortality risk than the control population (incidence rate ratio 10.6, 95% CI 9.86–11.5). They also died earlier compared with individuals with other ID diagnoses and the control population, with the mean age at death being 63.5 years in those with DS, 72.1 years in those with other ID diagnoses and 76.2 years in the control population. Only 5.7% of individuals with DS were aged 65 years and older compared with about 16% of the control population.

Table 2 shows the demographic characteristics of individuals who died in each group. Of all deaths that occurred among individuals with ID (n = 4738), about 87% were younger than 75 years, 96% were not married and only 2.8% were foreign-born. Almost all the sub-sample with DS were younger than 75 years and not married. Of deaths that occurred in the control population, three-quarters were younger than 75 years and 52% were married; 58% were men and 11% were foreign-born.

Table 3 presents the cause-specific proportionate mortality (%) and cause-specific mortality rates for individuals with ID and the control population, and the mortality rate ratio (95% CI) for each cause. The leading cause of death in both groups was diseases related to the circulatory system (about 35% of all deaths in both groups). However, the mortality rate for diseases of the circulatory system was about three times higher among individuals with ID compared with the control population, with a mortality rate ratio of 2.71 (95% CI 2.55–2.87). Diseases of the respiratory system were the second most common cause of death among individuals with ID (17% of all deaths), with a mortality rate ratio of 7.61 (95% CI 6.82–8.49) compared with the control population. Although neoplasms accounted for 33% of all deaths and was the second leading cause of death in the control population, we observed a lower mortality rate for neoplasms in the control population compared with individuals with ID. The mortality rate ratios for all cause-specific mortality rates were significantly higher for individuals with ID than the control population, except for diseases related to the skin and subcutaneous tissue (p > 0.05). When analysed specifically for individuals with DS (n = 513) (Table 4), the three leading causes of death were related to diseases of the respiratory system, diseases of the circulatory system and mental and behavioural disorders.

We further investigated the three leading causes of death within each selected major cause of death based on 3-digit ICD-10 chapters (Table 5). The analysis for neoplasms was stratified by sex to account for sex-specific reproductive tumours. Several causes of death, including heart failure, pneumonitis due to solids and liquids, Alzheimer’s disease and epilepsy, were more common among individuals with DS compared with the control population. Epilepsy was the underlying cause of death for 21.6% of all deaths related to the nervous system among individuals with ID (DS excluded), and accounted for 37% of deaths related to the nervous system among individuals with DS. The most common cause of death related to the nervous system in the control population was Alzheimer’s disease (44%), compared with 10% of individuals with ID (DS excluded) and 54% of individuals with DS. Overall, individuals with DS had higher cause-specific mortality rates compared with individuals with ID or the control population, particularly for deaths-related to heart failure, pneumonitis due to solids and liquids, Alzheimer’s disease and epilepsy (Additional file 1). The cause-specific mortality rate for Alzheimer’s disease was 215 per 100,000 population for individuals with DS, compared with 12 per 100,000 for individuals with ID and 14 per 100,000 for the control population.

Conditional Cox regression analyses showed that individuals with ID had a 4-fold higher mortality risk compared with the control population. When analysed for cause-specific mortality (Table 6), the mortality risk among individuals with ID ranged from 1.8 times higher for neoplasms to 12.5 times higher for diseases of the respiratory system compared with the control population. When the analyses were repeated to assess the hazard risk for men and women with DS, individuals with DS had much higher risk of death from respiratory- and nervous system-related diseases (note the large CIs for these two mortality groups). Figure 2 shows a lower survival rate among individuals with ID, with a median survival time of 68.8 years for men and 69.8 years for women with ID, compared with 75.2 years for men and 78.4 years for women in the control population.

In this study, we observed a 4-fold higher mortality rate and a shorter median survival time among older adults with ID compared with the general older adult population. Similar results have been reported in the US [29, 30], Canada [31], the UK [32], Finland [33] and Ireland [34]. Women in the control population and women with ID had higher life expectancies than men. The gender gap in median survival was relatively small among Swedish older adults with ID compared with the control population. Our finding of better survival among women with ID confirm those reported in the US [29] and Taiwan [35]. However, these findings differ from results reported in the UK [4, 36, 37], Finland [33] and Canada [31], which showed higher mortality rates among women with ID compared with their male counterparts. As our study focused on older adults (aged 55+ years), we might have missed individuals with more severe and profound ID, such as Rett syndrome, Williams syndrome and tuberous sclerosis, among whom life expectancies are reported with a maximum age of 50–59 years [1].

Our study confirmed observations from other countries that some causes of death were more common among individuals with ID [38]. Deaths from respiratory diseases (mainly pneumonia and pneumonitis due to solids/fluid) were more common among individuals with DS compared with the control population. A similar observation was reported by Englund et al., who identified pneumonia as the main cause of death among individuals with DS [3]. Among individuals with ID (DS excluded), cardiovascular deaths dominated, followed by neoplasm deaths. A national population-based study in the UK showed that adults with ID were more prone to die from congenital malformation and diseases in different body systems, including the nervous, respiratory, digestive, genitourinary and circulatory systems. However, their comparative risks of dying from respiratory and digestive organ neoplasms were lower compared with the general population [24]. Individuals with ID are more prone to cardiovascular disease risk factors, morbidity and mortality than the general population. A study in the Netherlands showed that hypertension, diabetes, hypercholesterolemia and metabolic syndrome were common health problems among adults with ID [39]. Cardiovascular disease health promotion and prevention programmes should be tailored for individuals with ID, to prevent or delay cardiovascular diseases and premature mortality in this vulnerable group.

Our study showed that deaths related to diseases of the nervous system (mainly epilepsy) were more prevalent among individuals with ID compared with the control population. Recent reviews showed a higher risk of sudden unexpected death in epilepsy among individuals with ID [40, 41]. A Swedish study using a national cohort of individuals with DS who died between 1969 and 2003 reported that death from dementia was common among individuals with DS [3]. A recent systematic review showed that pneumonia-related mortality increased more than 2-fold among individuals with dementia [42], which might explain the patterns of morbidity and mortality observed among individuals with DS in Sweden [3].

As the population ages, more people are living with dementia or Alzheimer’s disease. Our study verified existing evidence that dementia is a common cause of death among individuals with DS as well as in the general population. Dementia is an age-related degenerative disease and its prevalence increases sharply between ages 40–60 years among individuals with DS [43]. To date, no cure for dementia is available; however, treatments are available to slow the progression of dementia to more severe stages [44]. Early diagnosis of dementia will decrease the need for long-term care and improve the quality of life among people with dementia. Understanding how current health and social care systems provide dementia care for individuals with DS is necessary to develop best practice strategies and programmes [16].

Our study confirmed there were fewer deaths due to neoplasms among individuals with ID, especially among those with DS, compared with the control population. Breast cancer was the leading cause of death among women with ID (DS excluded) in our study. Neoplasms in women with ID were diagnosed in more advanced stages compared with women in the general population [45]. Barriers to accessing screening programmes have been reported at individual and health systems levels. A study in Canada highlighted the inequitable access to cancer screening among women with ID, with a lower rate of access to cervical neoplasms and mammography screening compared with the general population [46]. Attention should be directed to measures to reduce the barriers and increase access for individuals with ID, to ensure this population benefits from available screening programmes [46, 47]. Further studies should use longer-term longitudinal data to determine if the mortality risk related to neoplasms among individuals with ID has changed over time, and how access to cancer prevention and treatment programmes may influence mortality risk over time.

Individuals with ID, especially those with more severe forms of ID, are often dependent on others for care and assistance in activities of daily living, and have potential to be excluded from receiving high quality education and healthcare services [5, 48]. Barriers to care for individuals with ID include less interpretable symptoms of illness, more difficulty in performing examinations and communication problems [48]. The quality of care received by older individuals with ID is dependent on the skills of care staff in overcoming barriers in caring for individuals with ID, staff psychological factors that might influence interactions with individuals with ID and the quality of care delivered [49]. Future studies should investigate whether life-threatening diseases among individuals with ID with different living arrangements could have been averted. Early identification of symptoms and signs by trained and qualified carers in environments such as group homes means that seriously ill individuals with ID with life-threatening conditions may be referred to hospitals for adequate advanced care in a timely manner.

In this study, we identified a more representative population with ID by combining the LSS register and the NPR than what we would have achieved using each register independently. Even so, the true prevalence of individuals with ID in Sweden cannot be fully established. The LSS register contains an estimated 50–70% of adults with ID in Sweden, and includes all individuals with ID who live in group homes, work in day care centres or receive any other support from society [50]. To obtain a more reliable estimate of the prevalence of ID in Sweden, future studies should include other registers, such as the primary care register, and use a capture-recapture method [51]. No national-level register for primary health care in Sweden exists yet, as all registers are locally or regionally-based.

This study fills in the gap of knowledge on cause-of-death among older people with intellectual disabilities. Use of the Cause of Death Register allowed us to better identify the leading cause of death among individuals with ID. Disease diagnosis among individuals with ID may be challenging. Merrick et al. reported a lower prevalence of cardiovascular disease with increasing severity in cognitive dysfunction that could be explained by under-diagnosis [52]. The detailed information in the death register allowed us to understand the patterns of death among this population, which may contribute to better healthcare planning and management for individuals with ID. Considering the growing number of evidence on this field, a systematic review and/or meta-analysis on cause-of-death patterns among older people with intellectual disabilities is warranted.

As the Swedish population register does not contain any disease/diagnosis information, it is possible that the matched control population also included individuals with ID who did not access any LSS services or did not have an ID diagnosis recorded in the NPR. The LSS register might also include individuals without ID but with an autism diagnosis who received service and support under the LSS law. It is also possible that DS was not recorded as either the underlying or contributing cause of death. These potential misclassification biases might mean our results are under- or overestimated. However, we anticipated this by using clear inclusion criteria for ID and exclusion of atypical autism criteria.

We did not classify the causes of death by underlying and contributing causes of death. The focus was mainly on the underlying cause of death, which was imputed with the contributing cause of death only recorded for those who had been identified with DS as an underlying or contributing cause of death. There was also no detail regarding different ID diagnoses other than DS, which might have enriched our analyses and interpretation of cause-of-death data.

As the LSS data is only available from 2007 (the register started in 2004 with very limited data) and no earlier register of individuals with ID exists, it is not possible for us to control for prior institutionalisation among the individuals with ID before the closure of care institutions in the 70s. In Sweden, no register covering the adult age for the cohort of older people in our study exists. Such registration was forbidden considering what happened to people with intellectual disability before or during the World War II Period. The establishment of the LSS law and the LSS register in 2004 can therefore be seen as a breakthrough in understanding the health patterns and health care needs among people with intellectual disability in Sweden. Therefore, the results of this study should be interpreted in light of other potential confounding factors such as residential history, prior institutionalization, access to health care, exposure to different factors, as well as life-style behaviours which might influence the outcome of this study – the mortality. Similarly, as these data are not available in our study, we cannot explain whether the differential mortality level observed among individuals with ID and the general population were due to biological differences or the effects of the above-mentioned confounding factors.