Impact of socioeconomic deprivation on rate and cause of death in severe mental illness

Construction and content of the database

The Glasgow Psychosis Clinical Information System (PsyCIS) is a database of all individuals with psychotic illness who have been in contact with secondary care psychiatric services in Glasgow, Scotland, covering a population of approximately 1 million people. Data were initially collected retrospectively by two research nurses. From August 2005 onwards new cases have been added prospectively to the database using existing patient information management systems. All patients have detailed socio-demographic information including International Statistical Classification of Diseases and Related Health Problems (ICD-10) diagnosis, unique Community Health Index (CHI) number, postcode, educational attainment, family history of psychosis, psychiatric admissions data, use of the Mental Health Act, current and previous medications, adverse drug effects, psychosocial interventions received and psychiatric comorbidities. All patients provide annual follow up information and there is a reciprocal relationship between local clinicians and the PsyCIS team which facilitates the return to consultants of clinically relevant information at an individual caseload level.

The dataset is regularly added to and updated and currently holds information on over 7,200 patients with a diagnosis of psychotic disorder, including: schizophrenia (n = 4,787); bipolar disorder (n = 1,784); organic psychosis (n = 67); psychotic depression (n = 452); and substance-induced psychosis (n = 160) [23]. There is a process of annual review to check the accuracy of information, such as diagnosis, postcode and current medication and the database has undergone internal validity checking to determine the diagnostic accuracy of individual ICD10 diagnoses.

Sample

Analyses

Rate of death/10,000 population per 5 years by SIMD quintile for the total deceased SMI population (schizophrenia and bipolar disorder combined) was calculated and compared with rate of death in the local Glasgow and wider Scottish Population over the total study time period, using mortality data available from the General Register Office for Scotland. The “Glasgow” Effect was determined by comparison of rate of death between groups. Both all-cause mortality rate and mortality rate excluding suicide in the SMI population were calculated. The rationale being that as suicide is a major contributor to premature death in SMI it may in itself be influenced by socioeconomic status. Cause of death by SIMD quintile for the total deceased SMI population (schizophrenia and bipolar disorder combined) was calculated and compared with cause of death in the local Glasgow and wider Scottish Population over the total study time period, using mortality data. Rates and patterns in cause of death by deprivation quintile were compared in those with SMI to the local Glasgow and wider Scottish populations using Chi squared tests. All statistical tests were performed using STATA version 12.

Patterns of mortality in those with severe mental illness

Cause of death differed between the three groups: for example, death due to cancer occurred less frequently in those with SMI compared to the local Glasgow and wider Scottish populations (14.3% vs. 27.3%, p = <0.001 and 33.2%, p = <0.001) (Table 2). Alcohol related deaths were also lower (6.5% (SMI) vs. 12.8% (Glasgow), p = 0.005 and 9.7% (Scotland), p = 0.099) (Table 2). However, the proportion of suicides was higher in those with SMI compared to that seen in the local Glasgow and wider Scottish populations (14.8% vs. 7.0%, p = <0.001 and 6.5%, p = <0.001), as were drug related deaths (7.8% vs. 5.1%, p = 0.065 and 3.2%, p = <0.001) (Table 2). The proportion of deaths due to cardiovascular disease were similar across the three groups (14.8% in those with SMI, 13.2% in those living in Glasgow (p = 0.546) and 13.7% in those living in Scotland (p = 0.626).

Patterns of mortality by socioeconomic status

The rate of death per 10,000 population per 5 years for SMI compared to the local Glasgow and wider Scottish populations was higher across all socioeconomic quintiles and persisted even when suicide was excluded as a cause of death (Figure 1). For example, in the most deprived cohort, the rate of death (excluding suicide) in those with SMI was 697.2 per 10,000 population compared to a rate of 252.4 per 10,000 population for Scotland and 274.7 per 10,000 population for Glasgow. When suicide was excluded as a cause of death, the increase in mortality associated with deprivation was larger in those with SMI compared to those living in Glasgow and Scotland.

The proportion of suicide was significantly higher in those with SMI living in the more affluent quintiles relative to the local Glasgow and wider Scottish populations (54.6% vs. 5.8%, p = <0.001 and 5.5%, p = <0.001) (Table 3). Similarly, an increase in proportion of suicide was observed for those with SMI living in the most deprived quintile compared to the local Glasgow and wider Scottish population (12.3% vs. 6.9%, p = 0.05 and 6.9%, p = 0.05) (Table 3).

Differences in the rate, age and cause of death

We found that although individuals with SMI died younger than the Scottish population, the age at death was similar to the local Glasgow population. This is potentially in keeping with the “Glasgow” Effect whereby average age at death in Glasgow is lower than the Scottish average. However despite this, differences in cause and rate of death between the SMI and Glasgow groups were apparent. In particular, and perhaps surprisingly, deaths due to cancer and alcohol were lower in the SMI group compared to the Glasgow group. To date, the literature surrounding cancer deaths in individuals with SMI has been mixed, with some studies reporting elevated rates [31], while others report lower rates of cancer deaths [32]. A recent meta-analysis of incidence of cancer [33], reported that pooled overall rates of cancer incidence were not significantly increased in individuals with schizophrenia compared to controls (Standardised Incidence Rate (SIR) 1.05, 95% CI 0.95-1.15) and although incidence of lung cancer was increased, when adjusted for smoking there was no difference in incidence rates. Indeed it also reported that the incidence of several cancers unrelated to smoking was reduced in individuals with schizophrenia. These mixed findings in cancer rates may in part, help explain the unclear evidence for cancer mortality rates in individuals with SMI and highlight the need for further longitudinal research in this area.

Our finding of lower rates of death due to alcohol are perhaps surprising given the high rates of alcohol dependence reported in individuals with SMI [34],[35]. This may in part be explained by the elevated population levels of alcohol related deaths in Scotland compared to other countries. In 2011 the Scottish male alcohol mortality rate was 28.4/100,000 compared to the UK average of 17.2/100,000 while in women the Scottish alcohol mortality rate was 13.9/100,000 compared to the UK average of 8.3/100,000. [36] Our finding of a trend towards significantly higher proportions of drug related deaths and “other” causes of death (which included a wide range of causes) within the SMI cohort may also explain this finding. Given that only the primary cause of death was obtained from the GRO, rates of alcohol as a secondary cause of death within our cohort are unknown and may also explain our lower rates of death due to alcohol.

We also found that the proportion of cardiovascular disease deaths in those with SMI was only marginally higher compared to Glasgow and Scotland. This is in keeping with other studies which have found similar rates of cardiovascular deaths in patients with SMI compared to their local population [29].

The role of deprivation

We found that, as with the local Glasgow and wider Scottish populations, rate of death in those with SMI increased as deprivation increased. The magnitude of the elevation in mortality associated with deprivation was similar; however, when suicide was excluded as a cause of death, the magnitude of increase in mortality associated with deprivation was greatest for the SMI cohort. This association between deprivation and mortality in those with SMI appear similar to that reported in the general population. Of note, was our finding of a significantly higher proportion of suicide in those with SMI in both the least and most deprived quintiles. While a negative correlation between suicide and socioeconomic status has been found in Australia [37] and the US [38], our finding of a more complex relationship has been identified by others in the UK [39],[40]. In particular, Osborn and colleagues found that although suicide rates were increased across all age groups of individuals with SMI, those who were least socially deprived were particularly at risk [41].

Despite the negative impact of deprivation on health outcomes (for example, those living in the most deprived areas are more likely to have unplanned admissions to hospital than those living in the most affluent areas) [42] as well as the premature onset of multimorbidity [42], studies investigating the relationship between socioeconomic status and mortality in individuals with SMI are limited. While there are studies looking at the impact of socioeconomic status on parasuicide [43], suicide [40],[44] and cause of death in major depressive disorder [45],[46], generalised anxiety disorder [47] and alcohol misuse [48], we found only one study which specifically focused on the impact of socioeconomic status on mortality in bipolar disorder [49] and no studies which specifically focused on the impact of socioeconomic status on mortality in schizophrenia. The study which explored the impact of socioeconomic factors on cause and rate of death in affective disorders was limited due its small sample size of only 30 deaths.

The “Scottish” and the “Glasgow” effects

Although life expectancy and health behaviours have improved in Glasgow over the past 50 years, the rates of improvement have differed depending on the individuals’ socioeconomic status. As Scotland’s ranking in terms of European mortality rates have worsened over time, so too has the gap in life expectancy between the most affluent and most deprived areas within Scotland [50],[51]. This health inequality associated with socioeconomic status has been mirrored by an inequality in life expectancy associated with SMI [10]-[14].

Although the impact of deprivation on health and mortality is marked, the differences in life expectancy in Scotland compared to the rest of Europe, and in Glasgow compared to the rest of Scotland, cannot be fully explained by deprivation alone. These gaps in mortality which cannot be explained by deprivation have been termed the “Scottish” and glasgow Effects” [52],[53]. This glasgow Effect” was reflected in our study, where mortality rates for those with SMI were even higher than both the Scottish and Glasgow averages. The differences in mortality between those with SMI and those without also appeared largest for those living in the most deprived quintiles.

Our study of all-cause mortality in those with SMI in Glasgow has identified a trend towards increasing mortality associated with worsening deprivation, but more work is required to investigate this relationship further.

Strengths & limitations

Our findings of a gradient in level of deprivation and magnitude of increase in mortality in those with SMI are novel in that we were able to describe the rate and cause of death of individuals with SMI in detail and compare them to both the local Glasgow and wider Scottish populations. To our knowledge, this is the first study to investigate the impact of deprivation on rate and cause of death of those with schizophrenia and bipolar disorder and as such may help inform the development of population-based strategies to improve the life expectancy of individuals with SMI. Additional strengths of this study include the comprehensive nature of the PsyCIS database, which is representative of a large clinical population accessing mental health services in a defined urban area. Regular checks of data accuracy are carried out by the senior medical practitioners involved in case management, thereby maintaining the reliability of diagnoses, clinical and sociodemographic data.

Limitations of this work include the relatively small number of deaths which occurred during our limited study timeframe. Furthermore, exclusion of a small number of patients with psychosis who are managed exclusively in primary care also occurred. As PsyCIS is a secondary care based database, inclusion of all people with psychosis in the geographical area would require linkage of the database to primary care records, which is not currently possible. Although some patients with psychosis live independently in the community without input from secondary services, further study is required to ascertain the numbers of such patients in Glasgow. PsyCIS also excludes individuals who are under 16, over 65 years old and who are not managed by general adult community services, such as those whose psychotic illness is managed exclusively in addictions psychiatry, old age psychiatry, or learning disability services. These numbers are likely to be relatively small because the majority of patients of working age with psychotic disorder are managed by adult general psychiatric services. SIMD is a relative measure of deprivation and given the impact of socioeconomic drift in the PsyCIS cohort, wider Glasgow and Scottish population during our five year timeframe, its use has a number of limitations. A further limitation was that of the impact of age on rates and causes of death- although adjustment for age profile of the populations were made, full standardisation was not possible due to small sample sizes.