Healthcare visits for mental disorders and use of psychotropic medications before and after self-harm in a cohort aged 75+

Abstract Objectives Non-fatal self-harm (SH) is a major risk factor for late-life suicide. A better knowledge of the clinical management of older adults who self-harm is needed to establish where improvements could be made for the implementation of effective suicide prevention interventions. We therefore assessed contacts with primary and specialised care for mental disorders and psychotropic drug use during the year before and after a late-life non-fatal SH episode. Method Longitudinal population-based study in adults aged ≥75 years with SH episode between 2007 and 2015 retrieved from the regional database VEGA. Healthcare contacts for mental disorders and psychotropic use were assessed during the year before and after the index SH episode. Results There were 659 older adults who self-harmed. During the year before SH, 33.7% had primary care contacts with a mental disorder, 27.8% had such contacts in specialised care. Use of specialised care increased sharply after the SH, reaching a maximum of 68.9%, but this figure dropped to 19.5% by the end of the year. Use of antidepressants increased from 41% before to 60% after the SH episode. Use of hypnotics was extensive before and after SH (60%). Psychotherapy was rare in both primary and specialised care. Conclusion The use of specialised care for mental disorders and antidepressant prescribing increased after SH. The drop in long-term healthcare visits should be further explored to align primary and specialised healthcare to the needs of older adults who self-harmed. The psychosocial support of older adults with common mental disorders needs to be strengthened.


Introduction
Older adults have the highest suicide rates of all age groups in most countries (World Health Organization, 2021).Scandinavian population register-based studies have made important contributions to the current knowledge about late-life suicidal behaviour (Erlangsen et al., 2003;Erlangsen et al., 2015;Erlangsen et al., 2008;Hedna et al., 2021a;Hedna et al., 2020a;Hedna et al., 2020b).However, these national registers do not include primary care data.This is a serious limitation as milder conditions, such as minor depression, have been found to associate with both fatal and non-fatal suicidal behavior in older adults (Waern et al., 2002;Wiktorsson et al., 2010), and these will be missed by hospital registers.
In Sweden, primary care is responsible for the treatment of mild to moderate forms of psychiatric morbidity (Swedish National Board of Health & Welfare, 2021a).General practitioners may play a role in the management of older adults who selfharm.Clinical guidelines recommend that older adults are assessed by specialised psychiatric care after a self-harm (SH) episode and monitored until the situation is stabilized (Västra Götaland Region, 2021a).Research investigating healthcare use in older adults who self-harm is relatively sparse.Existing international research either reported psychiatric services before or at time of SH, or whether those who self-harmed received a psychiatric assessment immediately after the SH episode with no further follow-up (Cheung et al., 2017;Crosby et al., 2007;De Beer et al., 2015;Gheshlaghi & Salehi, 2012;Kim et al., 2011;Lebret et al., 2006;Murphy et al., 2012;Pillans et al., 2017;Wiktorsson et al., 2022).A few examined use of primary care before the SH episode (Dennis et al., 2007;Morgan et al., 2018;Wiktorsson et al., 2022).However, little is known on mental care received in primary and specialised services after a SH episode.
The few studies that assessed the use of psychotropic medications in older adults with non-fatal SH only described the use of antidepressants at time of SH but did not consider other psychotropic medications (Troya et al., 2019).These studies tended to include small and selective samples (De Beer et al., 2015;De Leo et al., 2002;Lamprecht et al., 2005;Lebret et al., 2006;Wiktorsson et al., 2011), or covered a wide age range including individuals in their 60's (Hawton & Harriss, 2006;Murphy et al., 2012).Extrapolating results from the "younger old" group to the oldest segment of the population may not be appropriate, as clinical management including medication use patterns may differ with age, in part due to differential comorbidity involving both mental and medical conditions (Skoog, 2011).A better knowledge of the clinical management of older adults who self-harm is needed to establish where improvements in clinical care could be made for the implementation of effective prevention interventions (Fässberg et al., 2016).We therefore aimed to investigate healthcare contacts for mental disorders in primary and specialised care and use of psychotropic medications during the year before and the year after an episode of non-fatal SH in a regional cohort of Swedish adults aged 75 years and above.

Data sources
We linked data from several national and regional registers using the Swedish personal identity number, unique to all individuals living in Sweden (Ludvigsson et al., 2009).The regional administrative healthcare database VEGA includes all healthcare contacts in the region of Västra Götaland with approximately 1.7 million inhabitants (2020, 17% of the total Swedish population).The region is considered representative for Sweden with regard to healthcare seeking and sociodemographics (Statistics Sweden, 2021;Swedish National Board of Health & Welfare, 2021c).This database encompasses publicly funded and private healthcare (Västra Götaland Region, 2021b).Weekly deliveries of data are compulsory from all healthcare sectors in the region.Diagnoses are coded according to the International Classification of Diseases of the World Health Organization-10 th version (ICD-10) and are entered in the electronic patient chart along with the healthcare contact.The VEGA database was used to retrieve data from primary and outpatient specialised care.Information on hospitalizations was extracted from the National Patient Register which has coverage of inpatient discharges exceeding 99% (Ludvigsson et al., 2011).For this study, specialised outpatient and inpatient care contacts for mental disorder were those where ICD 10 codes F00-F99 were recorded as the main diagnosis of the visit.Such hierarchy is not used in primary care contacts, so we considered all contacts where a mental disorder was recorded.We also retrieved data on type of healthcare professionals who provided primary or specialized outpatient care.
Data on dispensed prescription medications were extracted from the National Prescribed Drug Register (Wettermark et al., 2007).The register has a full coverage in outpatient care and in long-term care facilities (LTCF).Medications are defined by their Anatomical Therapeutic Classification codes of the World Health Organization (ATC code).All relevant ICD-10 and ATC codes are presented in Appendices 1a and 1b.Demographic data were collected from the longitudinal integration database for health insurance and labour market studies (Swedish acronym LISA), administered by Statistics Sweden.Persons residing in LTCF were identified from the National Register of Care and Social Services for Older Adults and Persons with impairment held by the National Board of Health and Welfare.Data on causes of death, including suicide (ICD 10 codes: X60-X84 and Y10-Y32) were collected from the National Cause of Death Register (Brooke et al., 2017).We excluded Y33-Y34 as they included many events not related to self-harm.

Study design and population
Older adults aged 75 years and over who self-harmed between 2007 and 2015 were identified and followed until 31 st December 2016 or censored at migration or death.They were retrieved by ICD-10 codes: Intentional SH (X60-X84) or harm of undetermined intent (Y10-Y32).Healthcare contacts and use of psychotropic medications were assessed in the year before and the year after the first SH episode registered during the observation period (index episode).

Statistical analyses
We first described sociodemographic characteristics in the year before index SH and healthcare contacts for common medical disorders previously found to be associated with SH (chronic pain, cancer, cardiovascular and cerebrovascular conditions).
We then compared the healthcare contacts for mental disorder and use of psychotropic medications in the year before and after the index SH episode using McNemar's test, testing for consistency in responses across two variables before and after index SH.We further followed in 3-month time intervals the healthcare contacts for mental disorder and the use of different psychotropic medications, adjusting the denominator for the number of persons who were alive at each time point after the SH episode.We conducted a sensitivity analysis excluding those with SH of undetermined intent.We further investigated whether health care contacts and treatments during follow-up differed between those who used violent methods compared to those who used poisoning.The same approach was applied to compare those who repeated SH with the rest of the cohort.Data analyses were performed by SAS version 9.4 (SAS Institute.NC, USA).

Ethics approval statement
The study was approved by the Regional Ethics Committee in Gothenburg .Approval was also granted by the register holders.No consent was required from patients as the study was based solely on national register data.All data were coded and anonymised by Statistics Sweden prior to analyses

Results
We identified 659 persons (352 women and 307 men) who had a non-fatal SH episode over the ten-year study period.More than three-quarters (77.6%) were hospitalized following the SH; a further 21.9% used outpatient specialized care in connection with the SH, and 0.5% were identified in primary care.Sociodemographic and selected medical conditions are shown in Table 1.During the year preceding the non-fatal SH episode, almost one tenth resided in LTCF.About one fifth had primary or specialized healthcare contacts for cardiovascular diseases and a similar proportion had care contacts for cancer.Furthermore, about one tenth had a diagnosis indicating a chronic pain.SH methods are detailed in Appendix 2. Poisoning was employed in nearly three-quarters of the index SH episodes.Within one year after the index SH episode, 58 (8.8%) repeated SH, among them 17 (2.6%)died by suicide.

Healthcare contacts for mental disorders
During the year before the index SH episode, 33.7% had a primary care contact with a registered mental disorder diagnosis (Table 2).Among them, 38.5% had also contacts with specialized care for mental disorder.Contact with specialized outpatient and inpatient services for mental disorder was noted for 27.8%.For almost half of these (47.8%) a mental disorder was recorded also in connection with primary care during the year that preceded the index episode.
Table 2 shows further that during the year after the index SH, use of specialized care increased sharply to 63.3% and exceeded use of primary care for mental disorders.This increase was found in both genders.Use of primary healthcare did not change after SH in those aged 80+.The use of specialized care for mental disorder increased in all age groups and for both community dwellers and persons residing in LTCF.
During the three months before SH, about one fifth had specialized healthcare contacts for mental disorder (Figure 1 and Appendix 3).After the index SH, use of specialized care for mental disorder increased sharply to 69% up to six months after the SH episode but dropped down to 20% by the end of the year.The proportion of those who used primary care for mental disorder increased slightly during the year after the SH episode.Sole use of primary care for mental disorder before and after SH was rare.Within 9-12 months after SH episode, 27% had contacts with either primary or specialized care for mental disorder.

Use of psychotropic medications
The overall use of psychotropic medications was high; about eight out of ten used at least one psychotropic medication before the index episode, and this proportion remained unchanged during the year that followed the SH episode in both genders and in all age groups (Table 3).About nine out of ten of those residing in LTCF were on at least one type of psychotropic medication.
During the 3 months that preceded the SH episode, the most frequently prescribed type of psychotropic medication was hypnotics, used by about six of ten older adults who subsequently self-harmed (Figure 2 and Appendix 4).Antidepressants were used by four out of ten before SH and the proportion increased after the SH episode, peaking at 60%.Use of antipsychotics also increased after SH episodes.Less than 5% of those who self-harmed used medications to treat substance/alcohol use disorders.The same was the case for both anti-dementia drugs and mood stabilisers (including lithium).

Sensitivity analyses
The exclusion of non-fatal SH of undetermined intent (number included = 456) yielded patterns similar to the entire cohort regarding both healthcare contacts and use of psychotropic medications (results not shown).Sensitivity analyses considering the method of index SH episode showed that use of hypnotics was high in those who poisoned while use of antipsychotics was higher in those who used violent methods.Those who used non-violent methods used more specialised psychiatric care during the 6-month period that followed the index episode (Appendix 5).Older adults who repeated SH had a higher use of specialized psychiatric services compared to those who did not repeat (Appendix 6).Further, hypnotics and antipsychotics were more often prescribed to those who repeated SH.

Contacts with healthcare providers
Contacts with psychologists/social workers were rare in both primary care and specialized outpatient services (Appendix 7), with little change after SH.Within specialized services, contacts with registered nurses nearly doubled to almost 20% in the period that followed SH.It is likely that this mainly reflects contacts of supportive character including monitoring of patient safety, symptoms, medication side effects, etc.An important caveat is that type of health care professional was not registered in the VEGA register for approximately a third of the primary care contacts and a fifth of the specialized outpatient contacts.

Discussion
In this population study, we could link a large clinical database including both primary and specialized care contacts to numerous national registers to examine healthcare contacts and psychotropic medication use longitudinally both before and after a SH episode.Such an approach is rare in the previous literature of late-life SH.We found that during the year that preceded the SH episode, about one third of older adults aged 75 and over who self-harmed had contacts for a mental disorder in primary care and a rather similar proportion had specialized care for a mental disorder.Use of specialized care increased after the SH episode but dropped back to the initial levels within a year while proportions using primary care were rather similar before and after the SH episode.Use of antidepressants increased from 40% before to 60% after the SH episode.Use of hypnotics was consistently high over the entire observation period.Contact with psychologists/social workers was marginal both before and after SH.
The proportion with specialist care for a mental disorder before SH in our study (28%) was considerably higher than that observed in the general 75+ population according to the regional statistics based on ICD 10 codes F00-F99 (Swedish National Board of Health & Welfare, 2021b).The use of specialist care was lower than that reported in a recent systematic review where combined results of included studies found that 41% of older adults who self-harmed had attended specialized psychiatric services prior to the SH episode (Troya et al., 2019).However, that review included persons in their 60's and we could identify no study reporting specialized care contacts for a mental disorder in persons aged 75 and above.Our study expands on those involving younger older adults, highlighting the need for specific attention to mental health issues in the oldest segment of the population.
After SH, about two-thirds used specialized care for mental disorder.This suggests a relatively high adherence to clinical guidelines (Västra Götaland Region, 2021a, 2021b) in the short term.This finding contrasts with a large British primary-care-based study with linkage to hospital records that identified a low referral rate to psychiatric services after SH (12%), despite national guidelines (Morgan et al., 2018).The high rate of hospitalizations during 3-6 months after the SH episode observed in our study may indicate that there were rehospitalizations for some persons; perhaps these were not fully recovered at discharge after the initial episode.Another explanation might be that some non-psychiatric hospitalizations for comorbid conditions were captured by our definition.
There may be several explanations for the drop in healthcare contacts after nine months.Older adults who self-harm are a heterogeneous group.Some may have recovered and may no longer require continuous contacts with specialized services.Others may decline further contacts or may have missed their appointments, perceiving them no longer a necessity.The finding that almost three quarters of the older adults who self-harmed had no contact for mental disorder during 9-12 months after the index SH episode may also signal a lack of continuation in care which might be detrimental to the patient.Our group previously showed that, among persons aged 70 and above, depression and anxiety scores, high suicide intent and low Sense of Coherence at index episode were associated with non-remission of major depression within a year after a suicide attempt (Wiktorsson et al., 2011).Older adults with such profiles at SH may be in need of more intense treatment.Effective long-term monitoring by specialist services is needed for those whose conditions do not stabilize in the short term, and stabilized conditions must be followed-up in primary care to detect signs of relapse.
While the amount of missing data was considerable regarding the type of healthcare professional involved in outpatient and primary care contacts, available data showed very few contacts with psychologists/psychotherapists.We could identify no previous study on this topic for comparison.A recent review, focusing on adults of mixed age, showed that access to appropriate psychological services for adults who self-harm is limited with little or no specialist aftercare (House & Owens, 2020).Even when psychological services are available, they may not be tailored to people who self-harm (Saunders & Smith, 2016).Further, services that provide evidence-based psychological therapies to people with anxiety disorders and depression may be reluctant to treat people with elevated suicide risk (House & Owens, 2020).
We could identify no published study to compare our findings on use of primary care with a mental disorder diagnosis before SH.The previously cited systematic review found that 62% had been in contact with primary care during the month before the SH episode but information was lacking regarding psychiatric diagnoses recorded at the healthcare visit (Troya et al., 2019).We anticipated that we might observe an increase in the proportion with primary care contacts with psychiatric diagnoses during the end of the observation period, when older adults whose condition might have stabilized would be referred to primary care for continuation treatment.The fact that this was not the case might partially be explained by the accessibility of specialized psychiatric services for persons who self-harmed in the region where the study was conducted.The authors of a recent systematic review concluded that general practitioners have an important role in SH management, which encompasses frontline assessment and treatment, ongoing management in primary care, and referral to specialist care (Mughal et al., 2020).However, an Australian qualitative study reported that general practitioners believed they had no significant role to play in the care of older adults (aged ≥80 years) who self-harm; they felt helpless when dealing with complex medical and social needs, and cited a lack of treatment options (Wand et al., 2018a(Wand et al., , 2019)).
Before SH, about 40% were on antidepressants.That is a larger proportion than for the general Swedish population aged 75+ (25%) (Hedna et al., 2020b), but similar to the 42% reported by the above cited systematic review involving also younger older adults (Troya et al., 2019).We noted an increased use of antidepressants after SH.This may be explained by healthcare contacts for SH which provided an opportunity to initiate treatment for people with untreated mental disorders such as depression.This is in line with the findings of our recently published study in the total 75+ Swedish population, where 25% of older adults 75+ who self-harmed initiated an antidepressant in the year following a SH episode (Hedna et al., 2022).However, we did not notice an increase in the proportion of users of antidepressants in the month before SH.This finding is contrary to that of a US study conducted in a mixed age group with mainly young adults (Katz et al., 2020).This may be explained by potential differences in help seeking behavior in the oldest population segment compared to younger age groups.It may also be explained by a suggested less antisuicidal effect of antidepressants in younger age groups (Gibbons et al., 2012).Pharmacological treatment of depression in older adults can be challenging, considering, for instance, the reported increased risk of falls (Jung et al., 2022), and the questioned effect (Mallery et al., 2019), especially in those who are frail (Brown et al., 2022).We report a high and stable proportion with hypnotics' use before and after the SH episode.Hypnotics are known to be associated with increased risk of SH in the older population (Hedna et al., 2018), and our results underline the importance of careful evaluation of prescriptions and use of hypnotics in older adults (Tseng et al., 2018).However, it must also be taken into consideration that sleep disturbance itself may be a risk factor for suicide in older adults (Bernert et al., 2014).Noteworthy, the use of antipsychotics increased after SH and may be explained by a diagnosis of psychosis, an attempt to increase efficacy of antidepressants in those with serious or bipolar depression not responding to antidepressants alone or off label use for a large spectrum of behavioral symptoms associated with SH (Carton et al., 2015;McKean & Monasterio, 2015).Careful attention should be paid to the risk-benefit balance of antipsychotics in this high-risk suicidal population (Kasckow et al., 2011;Kheirbek et al., 2019).Prescription of mood stabilisers (including lithium) remained low after SH.With the purported anti-suicidal properties of the mood stabiliser lithium in the treatment of mood disorders (Lewitzka et al., 2015), it was somewhat unexpected to note that this was a rare therapy in the aftermath of SH.This likely reflects a lack of comfort in initiating this medication, the need for established follow-up, as well as the concern of its potential lethality in overdose.
The increase in healthcare contacts for mental disorder observed during the 3 months before SH may suggest a higher level of distress or symptomatology that clinicians were attempting to manage and treat.Furthermore, while healthcare contacts dropped within 9 months after SH, the use of psychotropic medications remained stable and high.This raises concerns about the long-term follow-up in healthcare of older adults who have self-harmed.A prescription renewal without healthcare visits might suggest suboptimal management after SH.Since self-poisoning was a main method of self-harm, our findings advocate the need for medication reviews for older adults with depression and/or suicidal behavior, for the consideration of the benefit-risk balance as well as safer medication alternatives when possible.Community pharmacies have been recognized as a key, yet underused, component in suicide prevention (Murphy et al., 2015).Suicide prevention training for pharmacists has been implemented in some countries (Scottish National Suicide Prevention Leadership Group, 2018; Washington State Legislature, 2016), and the integration of pharmacists as members of pluridisciplinary mental health care teams needs to be explored (El-den et al., 2021).
The very low proportion of healthcare contacts with psychologists and social workers before and after SH, taken together with what we already know about the salient role of social factors in late life suicidal behavior (Fässberg et al., 2012), highlights the need to enhance psychotherapeutic and psychosocial treatments in addition to psychopharmacological therapies.

Strengths and limitations
A main strength of this study is the inclusion of primary care data, which has been lacking in previous Swedish studies based on national hospital registers.The addition of primary care data is particularly important considering our focus on an age group (75+) with high use of primary care as well as high rates of suicide.A further strength of VEGA is the high coverage of healthcare contacts with public healthcare providers in the region with a large source population.While there is always a chance that some mental disorder ICD codes might not be registered, this is likely to have a minor impact on results as health care earnings are dependent on the diagnoses (ICD-codes) of each patient.Our results include all unique SH patients who have been registered with one or more of the specified ICD-codes at least once during the study period.The reporting of diagnoses from private providers is not mandatory which is a limitation.Also, some psychologists and social workers may not be in the habit of recording their sessions in the medical records reported to the VEGA database and those contacts may therefore be underestimated.Information on use of modern communication technologies in the care of our population was not available in our data.While a recent report shows that digital health care services are being implemented in Sweden, use is very low in older adults (66+) compared to other age groups (Gabrielsson-Jarhult et al., 2019).
While the study considered codes of mental disorders and common medical conditions previously found to be associated with SH, other non-psychiatric diagnoses were not available in the data set.A further limitation is that we did not have access to VEGA data for the entire population aged 75 and above residing in the region.Also, it is important to note that the current study focuses exclusively on individuals who visited healthcare services.The proportion of individuals with mental disorders and specialist contacts must be interpreted in light of this.Results can therefore not be generalized to all older adults who self-harm.Another consideration is that the availability of specialized geriatric psychiatric care is greater in this region compared to some other regions in the country.Our dataset did not distinguish between behavior with and without suicidal intent.This means that some nonfatal SH episodes that were identified in our study may actually have been non-suicidal SH (Wand et al., 2018b).However, non-suicidal SH is uncommon among older persons in Sweden (Wiktorsson et al., 2022).Further, a Canadian reported a fairly good positive predictive value of ICD10 relative to the clinical assessment (Randall et al., 2017) The Swedish Prescribed Drug Register does not cover medications dispensed during hospitalizations, which means that prescription data are lacking for hospital stays which were primarily observed during the first 6 months after the index SH episode that followed the index SH.We do not consider this a major limitation since psychotropic drug treatment that is initiated in hospitals will, in almost all cases, be continued after discharge and thus captured in our study.We could not determine adherence to psychotropic medications based on refill data due to the high use of multidose prescribing in this age group, which automatically creates high refill adherence.Another consideration is that we could not determine the indication for the specific prescribed medications, nor could we compare potential changes in dosages for various drug types as the number of persons using specific types of psychoactive drugs types was insufficient.This could be of interest in the future but would require a larger population.

Conclusion
Our research contributes to inform policy makers and healthcare professionals on the clinical management of older adults with non-fatal self-harm, to align healthcare services to the mental health care needs of older adults who self-harm and to enhance preventive measures in both primary and specialized care.The relative lack of psychological treatments, taken together with the observed reduction in care contacts during the ensuing year may signal a need for improved follow-up treatment.The extensive use of hypnotics, known to be associated with increased risk of SH and suicide may be considered, and availability reduced when possible.The psychosocial support of older adults with common mental disorders needs to be strengthened in both primary and specialised care.

Figure 1 .
Figure 1.Proportions with healthcare contacts for mental disorders in older adults who self-harmed in the year before and after index self-harm episode a (N = 659).a A person may use more than one level of care.the mental disorder diagnosis was not required to be primary diagnosis.b Mental disorder recorded as the main diagnosis.last day not included in each period.

Figure 2 .
Figure 2. Proportions of users of psychotropic medications in older adults who self-harmed in the year before and after index self-harm episode a .a last day not included in each period.

Table 1 .
Characteristics of older adults aged 75+ with non-fatal self-harm who sought healthcare a (N = 659).
a Data from the regional register VegA in the year before index date(between  2007-2015).b Marital status was missing for 117 persons.c Characteristics in the year before self-harm.d Primary care and specialized outpatient and inpatient care.

Table 2 .
Healthcare contacts for mental disorders in older adults (75+) during the year before and after the index self-harm episode a (N = 659).
a the mental disorder diagnosis was not required to be primary diagnosis.b inpatient and outpatient care.Mental disorder recorded as the main diagnosis.c Mcnemar's test, testing for consistency in responses across two variables before and after index episode.d Place of residence reassessed after self-harm to account for those who moved to a long-term care facility after the index episode.* P < 0.05.*** P < 0.001.ns: non-significant.

Table 3 .
Psychotropic medication use in older adults 75+ during one year before and one year after index self-harm episode (N = 659).Place of residence reassessed after self-harm to account for those who moved to a log-term care facility after the index episode.