Abstract
Background
Psychiatric illnesses are known risk factors for self-harm but associations between self-harm and physical illnesses are less well established. We aimed to stratify selected chronic physical and psychiatric illnesses according to their relative risk of self-harm.
Design
Retrospective cohort studies using a linked dataset of Hospital Episode Statistics (HES) for 1999–2011.
Participants
Individuals with selected psychiatric or physical conditions were compared with a reference cohort constructed from patients admitted for a variety of other conditions and procedures.
Setting
All admissions and day cases in National Health Service (NHS) hospitals in England.
Main outcome measures
Hospital episodes of self-harm. Rate ratios (RRs) were derived by comparing admission for self-harm between cohorts.
Results
The psychiatric illnesses studied (depression, bipolar disorder, alcohol abuse, anxiety disorders, eating disorders, schizophrenia and substance abuse) all had very high RRs (> 5) for self-harm. Of the physical illnesses studied, an increased risk of self-harm was associated with epilepsy (RR = 2.9, 95% confidence interval [CI] 2.8–2.9), asthma (1.8, 1.8–1.9), migraine (1.8, 1.7–1.8), psoriasis (1.6, 1.5–1.7), diabetes mellitus (1.6, 1.5–1.6), eczema (1.4, 1.3–1.5) and inflammatory polyarthropathies (1.4, 1.3–1.4). RRs were significantly low for cancers (0.95, 0.93–0.97), congenital heart disease (0.9, 0.8–0.9), ulcerative colitis (0.8, 0.7–0.8), sickle cell anaemia (0.7, 0.6–0.8) and Down's syndrome (0.1, 0.1–0.2).
Conclusions
Psychiatric illnesses carry a greatly increased risk of self-harm as well as of suicide. Many chronic physical illnesses are also associated with an increased risk of both self-harm and suicide. Identifying those at risk will allow provision of appropriate monitoring and support.
Keywords
Introduction
Self-harm is an important cause of admission to hospital, with over 200,000 admissions per year in England. 1 Furthermore, people who self-harm have a substantially increased risk of suicide. 2 The new National Suicide Prevention Strategy for England 3 includes several priority groups including people with a history of self-harm, young people, and those living with long-term physical health conditions. The peak age for self-harm is 15–24 years, 4 and suicide is the third leading cause of death in this age group. 5
Psychiatric illnesses, especially depression, anxiety and alcohol misuse disorders, are well-known risk factors for self-harm. 6 Associations between physical illness and self-harm are less well established, particularly in young people. Previous studies of adults have shown associations between suicidal behaviour and chronic physical illness, such as asthma, type I diabetes mellitus, epilepsy and cancer.7–10
We have investigated associations between specific psychiatric and physical illnesses and self-harm. We used hospital data on the English population to stratify long-term conditions by their relative risk of self-harm.
Method
Data sources and study population
We used a linked dataset of English Hospital Episode Statistics (HES) and mortality data for 1999–2011. HES is a nationwide dataset of routine statistical records of every day case and hospital admission in National Health Service (NHS) care in England (population 52 million in 2001, 56 million in 2011), supplied by the Health and Social Care Information Centre. This does not include A&E patients, unless admitted, or patient contacts in general practice. The mortality data are derived from hospital records and death registrations that were supplied by the Office for National Statistics. The linked dataset brings together all successive records for each individual in a time-sequenced data file.
Study design
Retrospective cohort studies were undertaken in the same way for each psychiatric and physical condition studied. The methods are described using depression as the example. A cohort was constructed of people who had been seen as a day case or admitted as an inpatient with a diagnosis of depression. For comparison with the depression cohort, a reference cohort was constructed of other people who had been either day cases or inpatients with a wide range of other, mainly minor, surgical and medical conditions and injuries (see Table 2 footnotes). The depression and reference cohorts were followed to identify subsequent day care for, hospital inpatient admission for, or death from, self-harm.
Selection of cases
Number of people admitted to hospital with each condition (percentage who were female).
*Conditions used in reference cohort, with office of population, censuses and surveys (OPCS) code edition 4 for operations and ICD10 code for diagnosis (with equivalent codes used for other coding editions): appendectomy (OPCS4 H01–H03), adenoidectomy (E20), dilation and curettage (Q10–Q11), hip replacement (W37–W39), knee replacement (W40–W42), squint (ICD10 H49–H51), cataract (H25), otitis (H60–H67), sebaceous cyst (L72.1) upper respiratory tract infections (J00–J06), varicose veins (I83), haemorrhoids (I84), deflected septum, nasal polyp (J33 + J34.2), impacted tooth and other disorders of teeth (K00–K03), inguinal hernia (K40), head injury (S06), in-growing nail, toenail and other diseases of nail (L60), contraceptive management (Z30), internal derangement of knee (M23), bunion (727.1), dislocations, sprains and strains (S03, S13, S23, S33, S43, S53, S63, S73, S83, S93), selected limb fractures (S42, S52, S62, S82, S92).
†Adjusted for age in five-year bands, time period in single calendar years, region of residence and deprivation score associated with patients' area of residence, in quintiles.
‡At two decimal places, the RR was 0.95 (0.93–0.97).
Data analysis
The rate of self-harm in the depression cohort was compared with that in the reference cohort, and expressed as a rate ratio (RR). The calculation was based on person-time (see Appendix [online supplementary material] for more details). In all analyses, the data were initially stratified by age (five-year groups), sex, year of admission (single years), region of residence (nine in England) and quintile of the index of multiple deprivation (a standard method of scoring socioeconomic status in England). The indirect method of standardisation was used taking the combined depression and reference cohorts as the standard population. All calculations were initially done within each stratum. For example, within each five-year age stratum, the rate in the combined depression–reference cohort was applied to the number of people in the same age stratum in the depression cohort, and, separately, applied to the number of people in the same age stratum in the reference cohort. This gave an ‘expected’ number of people with self-harm in the depression cohort, and an ‘expected’ number in the reference cohort. The observed and expected numbers in each stratum were then summed to give results in broader age groups and an all-ages-combined set of observed (O) and expected (E) numbers. The final RR is given by the formula (Odepr/Edepr) divided by (Oref/Eref), where O and E are the observed and expected numbers in the depression and reference cohorts. The RRs and their 95% confidence intervals (CIs) were calculated using standard methods. 11
We further subdivided the results by whether the first episode for self-harm was within one year of admission to hospital or after a longer interval.
We undertook a separate analysis in which the end-point was death from suicide as the certified cause of death. This was identified from the relevant ICD codes on the death record (ICD-9 E950–E959 and ICD-10 X60–X84; we were not permitted access to data from coroners' verdicts as such). In this analysis, all deaths with these codes were included – those outside as well as those inside hospital.
Further details on methods is given in the Appendix (online supplementary material).
We ranked the psychiatric and physical conditions according to the size of the RR for self-harm. We defined levels of risk as very high or high empirically after examination of the results.
Results
Table 1 shows the number of people in the study by age group. In Table 2, the illnesses studied are ranked by their risk of admission for self-harm. Based on these RRs we arbitrarily defined categories of risk: those with very high risks of self-harm (RR >5), those with elevated risk of self-harm (RR <5, lower 95% CI >1), those with a neither high nor low risk (CIs that cross 1) and those with a reduced risk of self-harm (upper 95% CI <1). Figure 1(a) and (b) and the Supplementary Table show these data for males and females separately.
(a) Risk of self-harm in chronic illness in males ranked by rate ratios. Error bars represent 95% CIs (those with high or very high RRs had error bars that are too small to show meaningfully, see Table 2). Note the log scale on the horizontal axis. (b) Risk of self-harm in chronic illness in females ranked by rate ratios. Error bars represent 95% CIs (those with high or very high RRs had error bars that are too small to show meaningfully, see Table 2). Note the log scale on the horizontal axis.
Main findings
*,†See footnotes of Table 2.
§Data not shown for fewer than five observed cases.
Of the physical illnesses studied (Table 2), epilepsy, asthma, migraine, psoriasis, diabetes mellitus, eczema and inflammatory polyarthropathies were associated with an increased risk of self-harm. Cystic fibrosis, coeliac disease, Crohn's disease and spina bifida were associated with a neither high nor low RR. The association between cancers and self-harm was borderline significantly low (Table 2, footnote). The physical illnesses and conditions associated with a reduced risk of self-harm were congenital heart disease, ulcerative colitis, sickle cell anaemia and Down's syndrome.
The RRs for self-harm were highest in the year following admission for all of the psychiatric illnesses (Table 2). However, the subsequent RRs, though lower, were nonetheless also strikingly high. The physical disorders showed less of a difference between RRs in the first year and subsequent periods (Table 2). Cancers were associated with a borderline significant but very small increase in rate of self-harm in the first year after diagnosis (RR = 1.05, 95% CI 1.00–1.09), but a reduced incidence of self-harm after the first year (RR = 0.91, 95% CI 0.86–0.94).
Findings in the subgroups by age
We divided the cohorts into four age groups – 10–24 years, 25–44 years, 45–64 years and ≥65 years (Table 3) and compared each disease group with the control cohort within the same age range. The RRs for self-harm were highest for most illnesses in the 45–64 year-old age group. There were, however, some exceptions. Eating disorders had a particularly high risk of self-harm in 25–44-year-olds. Spina bifida showed a reduced RR (0.70) in the 10–24-year-old group and an increased RR (1.90) in the 45–64-year-old group. Coeliac disease and Crohn's disease were associated with an increased risk of self-harm in the 45–64-year-old group but no overall increased risk. Cancers and congenital heart disease were associated with a significantly low risk in people aged under 45 years, but, for cancers, with a high risk in people aged 65 years and over.
Findings in the subgroups by sex
We divided the cohorts by sex, as shown in Figure 1(a) and 1(b) and the Supplementary Table. Overall, the number of cases of admission for self-harm for men and women were broadly similar (101,370 men vs. 112,265 women). When considering individual illnesses, however, there were some marked differences in the number of male and female cases; for example there were many more female admissions with eating disorders (as expected), and many more male admissions with substance abuse (Table 1).
The RRs for self-harm were broadly similar for males and females for most illnesses, with some notable differences. Both Crohn's disease and cancer were associated with low risk ratios in females (RR = 0.91 and 0.87, respectively), but increased risk ratios in males (RR = 1.16 and 1.07, respectively). In general, the risk of admission for self-harm for most illnesses studies was higher in males than females (Supplementary Table).
Risk of suicide
*,†See footnotes of Table 2.
Table 4 also shows the ratio of suicides (the observed numbers in Table 4) to all self-harm events (the observed numbers in Table 2) for each illness. Overall, the ratio of self-harm events to suicides was 23. The ratio was particularly high for eating disorders and low for cancers.
Discussion
We found a strong link, as expected, between all of the psychiatric illnesses and self-harm. There was also an elevated risk for several physical illnesses studied – asthma, epilepsy, diabetes mellitus, migraine, psoriasis, eczema and inflammatory polyarthropathies – but at much lower levels of risk than for psychiatric disorders. Some physical illnesses were associated with a reduced risk of admission for self-harm; namely ulcerative colitis, cancer in young age groups, congenital heart disease, sickle cell anaemia and Down's syndrome.
Strengths and limitations
This study is based on all NHS hospital discharges, and all deaths, for the whole of England. To the best of our knowledge, this is the largest study examining chronic illness and self-harm. A particular strength is the fact that we have compared the risk of self-harm in people with different psychiatric and physical disorders within the same total population. Accordingly, the level of risk in the different disease groups is directly comparable.
The diagnoses of psychiatric and physical illness were made by doctors, as opposed to self-reported diagnosis, which means that the diagnosis is likely to be clinically reliable and free of reporting bias. It has previously been shown that discharge coding for non-fatal self-poisoning is accurate compared with that identified through a self-harm monitoring service. 12
While the use of population data for all hospitals in England provides representative and reliable data, it has several limitations. We were only able to identify cases of self-harm that resulted in hospital admission or death, and that were identified as intentional self-harm. It is likely, for example, that where the attending doctor is uncertain whether an overdose was deliberate or accidental a diagnosis will be recorded and coded as an event of ‘undetermined intent’ (ICD10 codes Y10–Y34). Other possibilities in miscoding/misdiagnosis of self-harm are possible; people with diabetes may self-harm using insulin, for example, which may instead be coded as a hypoglycaemic attack. Our data for suicide were taken from ICD codes on death certificates rather than coroners' verdicts as we did not have access to the latter. Open verdicts, many of which are likely to be suicides, 13 are not included. Self-harm and suicide are, therefore, likely to be under-reported in this study. However, this is the case for the reference cohort as well as the psychiatric and physical illness cohorts and may not impact much on the comparisons made between them.
With regard to physical illness, we were only able to identify cases where people were admitted to hospital for the disease investigated, which may exclude people whose disease is well controlled and do not require admission. Outpatients were not included. Accordingly, our findings may only apply to people whose illnesses were serious enough to warrant hospital admission.
Finally, HES lack detailed demographic data; while controls were matched by age, gender, socioeconomic status and location of residence, we could not take account of other confounders. There is also the issue of overlap between different diagnoses. We considered the possibility of studying risks in people with diseases in combination, e.g. diabetes with and without depression, but deemed it beyond the scope of the present study because of the large number of possible combinations. While this limits our ability to draw conclusions on inter-related causes, we believe that this does not diminish the aim of this study, namely to compare the risk of self-harm and suicide between different clinical conditions.
Comparisons with previous research
Psychiatric disorders are well known as risk factors for self-harm and suicide, and our findings are consistent with those of other investigators.6,14 Our findings of increased risk of self-harm in people with epilepsy, diabetes mellitus and asthma are consistent with previous findings.7–10 Psoriasis and migraine have also previously been associated with suicidal ideation in small studies.15,16
A recent Danish register-based case-control study showed epilepsy, asthma, cancer and diabetes mellitus were associated with a higher risk of attempted suicide in adolescents, but were not independent risk factors following adjustment for other confounders, such as socioeconomic class. 17 In contrast, our study shows a reduced risk overall of self-harm in patients with cancer aged under 45 years. Previous studies in adults have shown an increased risk of suicide and attempted suicide in patients with cancer, 18 while other studies on children and adolescents have not found an increased risk of suicide. 19
There is a lack of data regarding Down's syndrome and self-harm, although suicide rates in people with learning disorders have been showed to be lower than average. 20
Our data contrast with findings from elsewhere that inflammatory bowel disease is associated with an increased risk of suicide in adults. 21
The ratio of overall self-harm to suicide was 23 in our study, which is slightly lower than reported elsewhere. 22 This is likely to be because our study only including admitted patients. The particularly high ratio in individuals with eating disorders is probably because repetitive self-cutting is common in eating disorders. 23
Mechanisms of association
Physical illness is often associated with mental illness. 24 Also, childhood physical disorders, such as asthma and type I diabetes, have been shown to be associated with subsequent psychiatric disorders.25,26 In adulthood, migraine has been associated with depression and anxiety 16 and dermatological disease with depression. 15
The association between epilepsy and suicidal phenomena is well recognised. The possible mechanisms have been discussed by others 27 and include a high prevalence of co-morbid psychiatric disease and a tendency towards lower socioeconomic status. Antiepileptic drugs have also been suggested to increase the risk of suicide, although the evidence is inconclusive. 28
Depression and suicidal ideation have been shown to be more common in patients with inflammatory bowel disease than the general population in a Canadian study. 29 However, we found that the incidence of self-harm was lower than average in patients with ulcerative colitis and neither high nor low in Crohn's disease overall.
Cancers had a slightly reduced risk of self-harm (RR = 0.95) but an increased risk of suicide (RR = 1.2). While younger age groups of cancer patients had a reduced risk of self-harm (Table 3), individuals aged over 65 years had a slightly increased risk (RR = 1.1). Notwithstanding the modest increased risk of suicide, one possible explanation for the generally low risk of self-harm in younger patients is that many people with cancer receive increased support and help. There was also a significant gender split, with males having a slightly increased risk of self-harm (RR = 1.07) but females having a reduced risk (RR = 0.87). This may be due to differences in the types of cancer in males and females, and support received. Breast cancer patients, for example, may receive greater psychological support than sufferers of other cancers.
Although Down's syndrome is associated with dementia, it appears to be protective against other psychiatric co-morbidity. 30 Self-harm in patients with Down's syndrome was a very rare event, with only 26 cases overall (RR = 0.11). Possible explanations for this include increased supervision and a reduced capacity for self-injury.
There have been fewer large-scale studies on the association between other physical illnesses and suicidal phenomena, and consequently less evidence on possible mechanisms.
Conclusion
We have quantified the risk of admission to hospital for self-harm in patients with psychiatric and physical illnesses. While psychiatric illnesses were associated with a greatly elevated risk of self-harm, a moderately elevated risk was seen with common physical illnesses such as diabetes, epilepsy and asthma. Patients with psychiatric illness had a greatly increased risk of suicide while some physical illnesses such as epilepsy, asthma, eczema and cancers had a moderately increased risk.
The link between psychiatric illnesses and self-harm is well established, but associations between physical illnesses and self-harm are less well known. It is important for physicians to be aware of the physical disorders that are associated with an increased risk of self-harm so that at-risk individuals may be better identified and can be monitored for psychiatric symptoms and mental distress. This illustrates the need for greater integration of medical and mental health services.
Footnotes
Declarations
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
