Critical Periods for Increased Mortality After Discharge From Inpatient Mental Health Units: Opportunities for Prevention

Study analyses were undertaken as part of the ongoing surveillance, program planning, and evaluation activities conducted as part of VA operations. As such, institutional review board review was not required.

VA clinical records were scanned to identify discharges from inpatient mental health units in 2013 and 2014. Discharge dates, patient characteristics, and inpatient diagnoses were extracted from VA patient records. Information regarding postdischarge mortality, including indicators of vital status and cause of death, were drawn from National Death Index data included in the joint VA–Department of Defense Suicide Data Repository, a resource that includes comprehensive information on mortality. For the primary, hypothesis-testing analyses, deaths were classified as due to suicide, other external causes (accidental or undetermined), or natural causes. For exploratory, follow-up analyses, deaths from natural causes were categorized as being due to either a broad category of circulatory diseases and risk factors (cardiovascular, cerebrovascular, peripheral vascular, hypertension, hyperlipidemia, and diabetes), pulmonary diseases, cancer, infectious causes, or dementia. Given the findings related to dementia as a cause of death (see below), supplemental analyses evaluated mortality among patients with clinical diagnoses of dementia or related neurodegenerative diseases. ICD-10 codes for causes of death occurring in this population and ICD-9-CM codes for dementia diagnoses are provided in Table 1.

Analysis utilized Stata, version 15.1. The primary questions about time-limited increases in mortality were addressed through life tables and log-rank tests for differences in survival functions for the first 30 and 90 days after discharge, compared with the time from 91 to 365 days.

When the initial hypothesis-testing analyses demonstrated critical periods after discharge for mortality from nonsuicide external causes, questions arose about whether there were meaningful distinctions between suicide and other external causes as identified through death certificate diagnoses. These questions were addressed through analyses based on data on suicide attempts (ICD-9-CM codes E950–959 from admission or discharge diagnoses, from other electronic health records from the period within 7 days of admission, or from other administrative data), and admission or discharge diagnoses of suicidal ideation (ICD-9-CM code V62.84), with testing of associations between these known suicide risk factors across causes of mortality with Cox models.

When the initial hypothesis-testing analyses demonstrated critical periods for natural-cause mortality and for dementia as a cause of death, exploratory analyses were conducted to determine whether there were critical periods for all-cause and natural-cause mortality specifically among patients with clinical diagnoses of dementia and to estimate the proportion of excess natural-cause mortality among older patients that could be attributed to those with dementia. As above, the significance of the increased rates during the critical periods was evaluated with log-rank tests for differences in survival functions. To estimate the proportion of the excess natural deaths attributable to dementia, the excess deaths during the critical periods were calculated as the number of observed deaths minus the number of expected deaths, where the number of expected deaths was estimated by applying the rates observed during days 91 to 365 to the number of patient-years of exposure within the first 30 and 90 days.

The population discharged from VA inpatient mental health units during 2013−2014 included 106,430 unique patients: women, 10.2% (N=10,817); mean±SD age=49.9±14.7; ages 18–64, 85.7% (N=91,219); and ages ≥65, 14.2% (N=15,124); data on age were missing for 0.1% (N=87). In terms of race, 47.4% (N=50,487) of the patients were white, 17.0% (N=18,083) were African American, 0.8% (N=824) were Asian, 0.5% (N=500) were Native American, and 34.3% (N=36,536) declined to report a race or reported two or more; 3.4% (N=3,568) were Hispanic. Suicide attempts within 1 week of hospitalization were documented for 6.1% (N=6,475) of patients, and 36.8% (N=39,187) had an admission or discharge diagnosis of suicidal ideation. A diagnosis of dementia was observed for 4.1% (N=4,398) (23.5% [N=3,328] of those ages ≥65). Of the 106,430 patients, 3.6% (N=3,829) died within 1 year of discharge. One-year mortality ranged from 1.0% for those ages ≤30 to 2.6% for those ages 30–64 and 11.2% for those ages ≥65. For comparison, 1-year mortality was .14%, 1.0%, and 5.6% for all VA patients, respectively, in these age groups and .21%, 1.3%, and 6.6%, respectively, in these age groups for VA patients with a psychiatric diagnosis.

Overall, 74.5% of the 3,829 deaths within 1 year after discharge were from natural causes and 25.5% from external causes (suicide, 8.9%; and other external causes, 16.6%) (Table 2). These proportions differed by age groups. For younger and middle-aged patients (ages 18–64), natural and external causes accounted for 61.1% and 38.9% of deaths, respectively; suicide and other external causes accounted for 14.6% and 24.3%, respectively. Deaths among patients ages ≥65 were primarily due to natural causes (91.4%), with 8.6% from external causes (suicide, 4.1%; and other external causes, 4.5%).

All-cause mortality (calculated as deaths per 100,000 patient years) was elevated in the first 30 and 90 days, compared with the rest of the year (Table 3). Compared with the rest of the year, suicide rates were higher in the first 30 and 90 days, and rates for external causes of mortality other than suicide were higher in the first 30 days. Rates of natural-cause mortality did not differ significantly across these time frames in the population as a whole; however, offsetting effects were noted, with elevated rates in both periods among older patients and decreased rates in the first 90 days after discharge among younger and middle-aged patients.

Among younger and middle-aged patients, significantly higher rates of suicide were noted in the first 30 and 90 days, compared with the rest of the year, and higher rates of nonsuicide external-cause mortality were noted in the first 30 days, similar to rates in the population as a whole. By contrast, among younger and middle-aged patients, natural-cause mortality was lower in the first 90 days after discharge, compared with the remainder of the year. Compared with the remainder of the year, higher rates of all-cause and natural-cause mortality were noted in the first 30 and 90 days among older patients, as were higher rates of other external-cause mortality in the first 90 days.

Analyses conducted to attribute the increased natural-cause mortality among older patients to more specific causes showed significant increases in mortality due to dementia but not to a broad category of circulatory conditions, pulmonary disease, infectious disease, or cancer. The increased rate of mortality from dementia was also significant in analyses that considered patients of all ages

The finding that the time after discharge represented a critical period for increased rates of death from both suicide and other external causes (accidental or undetermined), together with recurring questions about the reliability of information on cause of death from death certificates, raised questions about whether there are meaningful distinctions between the deaths attributed to these causes. Findings from Cox proportional hazards models controlling for age and sex (Table 4) demonstrated that medical record indicators of suicidal ideation and a recent suicide attempt were associated with death certificate diagnoses of suicide but not with death certificate diagnoses of deaths from other external causes, providing a kind of validation of the distinction between these causes of death. Further insight may come from review of the specific causes of nonsuicide external-cause mortality. In the first 30 days, 51 of the 77 deaths from nonsuicide external causes were classified as accidental or undetermined overdoses. Other causes observed at lower frequencies included motor vehicle and other transportation collisions or accidents, falls, firearm injuries, drowning, burns or smoke inhalation, inhalation of objects, and assaults.

Because diagnoses of dementia, rather than coexisting conditions or complications, as a cause of death may be unreliable, other analyses evaluated mortality after hospital discharge among patients with admitting or discharge diagnoses of dementia. All-cause mortality for patients with dementia was 3.0% at 30 days, 7.5% at 90 days, and 20.2% at 1 year. For those ages ≥65, all-cause mortality was 3.6%, 8.9%, and 24.0%, respectively. Rates of both all- and natural-cause mortality were higher in the first 30 and 90 days than in the remainder of the year for patients with dementia in the population as a whole and for older patients. In contrast, other analyses (not shown) demonstrated that there were no significant increases in all-cause and natural-cause mortality among older patients without clinical diagnoses of dementia. Based on calculations that applied rates from days 91 to 365 to earlier time periods, among older patients there were 54 excess deaths from natural causes (30.3% of all deaths from natural causes) during the first 30 days and 120 excess deaths from natural causes (24.6%) for the first 90 days after discharge. The numbers of excess deaths for older patients with clinical diagnoses of dementia were 55 and 107, respectively. Therefore, the greater natural-cause mortality among older patients during the critical periods can be attributed almost entirely to the subgroup with clinical diagnoses of dementia.

These analyses demonstrated that the first 30–90 days after discharge from inpatient mental health units was a critical period marked by higher rates of all-cause mortality, compared with the remainder of the year following discharge. The increased mortality was not specific to suicide. It included elevated mortality for external causes other than suicide, for which there were significantly higher rates in the first 30 days for the population as a whole and for young or middle-aged patients, as well as higher rates in the first 90 days for older patients, all compared with the remaining 9 months of the year.

For natural-cause mortality, significantly elevated rates were noted in the first 30 and 90 days among older patients—rates that could be attributed specifically to patients with clinical diagnoses of dementia—and no significant elevations were noted for other patients. The elevated rates of natural-cause mortality observed in the critical periods among older patients were not found in the patient population as a whole, probably because of the unanticipated decrease in rates of natural-cause mortality during the first 90 days among young and middle-aged patients, compared with the remainder of the year. Further studies are needed to characterize the impact of hospitalization on natural-cause mortality among younger and middle-aged patients and to distinguish between mortality displacement—excess mortality during the hospitalization or the period leading up to it followed by lower mortality among the presumably healthier patients who survived this period and were discharged—versus time-limited benefits from hospitalization and follow-up care.

Earlier reports of increased rates of suicide in the first months after discharge from inpatient mental health units (1, 2) were important because they pointed to opportunities for suicide prevention based on care enhancements during these critical periods. In the years since the initial reports, VA has implemented requirements for follow-up within 1 week of discharge, for additional follow-up encounters within 1 month, and for safety planning for those recognized as being at risk of suicide. The observation that suicide rates after discharge have remained elevated suggests that these enhancements have not been sufficient. Ongoing research is evaluating more intensive interventions (21).

The observation of increased mortality from nonsuicide external causes in the period closest to discharge, compared with the remainder of the year, may be related to the stresses of hospitalization and discharge, to residual symptoms of conditions that led to hospitalization, or to adverse effects of medications started in the hospital. Observations regarding the frequency of overdoses suggest that opportunities for prevention include ensuring that substance use disorders are addressed during hospitalization, including cases in which the admission is related to another condition. Other opportunities include extending the process of safety planning to focus on preventing accidents as well as intentional self-harm. In addition, it is important to ensure that cognitive and psychomotor side effects of medications that will be taken after discharge are recognized and minimized before discharge and that patients (and, where appropriate, families) are counseled about relevant risks.

These analyses demonstrated the existence of critical periods for increased rates of natural-cause mortality only for patients with dementia. These findings raise several issues. Deaths of patients with dementia are not, in general, included in discussions of excess mortality among patients with serious mental illness. Nevertheless, treatment of individuals with dementia represents a significant component of mental health care. In this study, patients with dementia represented 4.1% of all patients discharged from inpatient mental health units, and 22% of those ages ≥65. These patients were admitted for treatment of mental disorders or behavioral symptoms as comorbidities or complications of dementia, rather than for treatment of the dementia itself. For some patients, the behavioral symptoms that led to hospitalization may have been indications that they were near the end of life and that it would be useful to consider transitioning to hospice or comfort care (22). Alternatively, the increased mortality among patients with dementia may have reflected stresses related to hospitalization or changes in caregiving for some patients or to adverse effects of treatment (23). The Food and Drug Administration warning that “elderly patients with dementia-psychosis treated with antipsychotic drugs are at increased risk for death,” as well as related findings (24) and questions about adverse effects of antipsychotic agents (25), raise questions about whether the observed effects could be related to these medications. Additional studies based on reviews of electronic medical records are in progress.

The major strength of this study was the availability of data on vital status, causes of death, and clinical variables for a cohort of more than 100,000 individuals. The data were available because VA has acquired this information for more than a decade to support program planning and evaluation. Limitations included questions about the reliability of the classification of causes of death. Although the analyses reported here confirmed that there was a meaningful distinction between death certificate diagnoses of suicide and other external causes, there are recurring questions about the attribution of death to external versus natural causes. We could not estimate the proportion of deaths from natural causes that may have been related to self-neglect and treatment nonadherence and, therefore, that may have been related to self-harm. Questions also exist about the natural deaths coded as due to mental and substance use disorders (ICD-10 codes F01–F99). Excluding deaths due to dementia, these accounted for 3.1% of all deaths in the sample; these deaths may have included some deaths by suicide or other external factors as proximate causes of death.

VA is disseminating information on mortality after discharge from inpatient mental health units as a matter of transparency and accountability and also because such information may be relevant in other systems. The findings demonstrate that there are critical periods for increased mortality during the first 1 to 3 months after discharge from inpatient mental health units. They provide evidence that the critical period for increased suicide rates demonstrated in earlier VA research has persisted despite enhancements to care provided after discharge. The findings also show that the increased risks during the critical periods were not limited to suicide. The findings include increased rates of nonsuicide external-cause mortality, primarily due to overdoses, in the first month and increased rates of natural-cause mortality among older patients, attributable to those with dementia. The findings regarding deaths from overdoses and nonsuicide external causes are actionable and can be translated into strategies for prevention. The findings about older patients with dementia should alert providers and health systems to the vulnerability of these patients. However, further study is required before more specific recommendations for changes in policy and practice can be made.