Suicide Mortality Following Nursing Home Discharge in the Department of Veterans Affairs Health System

Abstract

Objectives. We assessed suicide rates up to 6 months following discharge from US Department of Veterans Affairs (VA) nursing homes.

Methods. In VA Minimum Data Set (MDS) records, we identified 281 066 live discharges from the 137 VA nursing homes during fiscal years 2002 to 2008. We used MDS and administrative data to assess resident age, gender, behaviors, pain, and indications of psychoses, bipolar disorder, dementia, and depression. We identified vital status and suicide mortality within 6 months of discharge through National Death Index searches.

Results. Suicide rates within 6 months of discharge were 88.0 per 100 000 person-years for men and 89.4 overall. Standardized mortality ratios relative to age- and gender-matched individuals in the VA patient population were 2.3 for men (95% confidence interval [CI] = 1.9, 2.8) and 2.4 overall (95% CI = 2.0, 2.9). In multivariable proportional hazards regression analyses, resident characteristics, diagnoses, behaviors, and pain were not significantly associated with suicide risk.

Conclusions. Suicide risk was elevated following nursing home discharge. This underscores the importance of ongoing VA efforts to enhance discharge planning and timely postdischarge follow-up.

In 2009, 15 700 nursing homes in the United States provided services to 1.4 million individuals.¹ In 2008, 36 035 individuals died by suicide in the United States, and suicide was the 10th leading cause of death.² Men have higher suicide rates than women, and national suicide rates in 2007 were highest among men aged 75 years and older.³ Despite concerns regarding self-destructive behavior among nursing home residents,^4,5 few studies have examined suicide mortality related to nursing home stays.^6,7 Furthermore, although the literature documents elevated suicide risk following discharge from inpatient psychiatric settings,^8,9 we are not aware of any studies that have examined suicide risk following discharge from nursing homes. Assessment of suicide risk following nursing home discharge provides an important opportunity to identify high-risk periods for suicide and to inform discharge planning, outreach, and care coordination activities.

Since passage of federal nursing home reform legislation in 1987,¹⁰ improving quality of care in US nursing homes has been a national focus. Initiatives target community facilities that are eligible for Medicare and Medicaid reimbursement as well nursing homes in the US Department of Veterans Affairs (VA) health system, which are today called Community Living Centers (CLCs). Central to this effort are quality and performance measures derived from information from the national nursing home resident assessment instrument, the Minimum Data Set (MDS). MDS assessments are mandated for all nursing home residents shortly after admission, at periodic intervals, and when there is a change in health status. The MDS includes indicators specific to quality of life within nursing homes as well as quality of care, focusing on medical, rehabilitative, and mental health issues.

The national focus on improving the quality of services provided in nursing homes has not addressed continuity of care after discharge. The VA has begun to evaluate this issue. Previous work has characterized resident risk factors for all-cause mortality following discharge from VA nursing homes.¹¹

We evaluated suicide rates following discharge from VA nursing homes. Because previous research suggests that psychopathology and challenging behaviors are associated with increased suicidal ideation and behavior,^4,12–15 we evaluated measures of serious mental illness, depression, dementia, behavior problems, and pain as predictors of suicide after discharge.

METHODS

The study population for estimating rates of suicide comprised all VA nursing home discharges, excluding discharges on the date of death, for fiscal years 2002 to 2008 (October 1, 2001–September 30, 2008). We excluded discharges on the date of death because it was uncertain which event occurred first. We identified nursing home discharges from VA health system resident assessment instrument MDS records. The MDS had 281 066 live discharge records in this period, for 161 164 unique residents. We used all live discharge records to calculate suicide rates. We calculated rates for all discharges, for all live discharges, and separately for men, overall and by age group. The number of female residents in our data was relatively small, representing only 3.2% of observed risk time following discharge, making the reliability of rates specific to female residents questionable, so we did not report their overall or age-specific suicide rates.

In the evaluation of potential suicide risk factors, we excluded 43 640 discharge records (15.5%) from analyses because of missing data for at least 1 measure. Excluded records were more likely than retained records to be for individuals who were older (average age = 70.3 vs 70.1 years) and male (97.4% vs 96.9%). Among the excluded records were 16 discharges that were followed by suicide within 6 months; however, suicide mortality did not differ between excluded and retained discharge records. For the multivariable assessment of predictors of postdischarge suicide, the analytic data set thus comprised 237 426 live discharges. The unit of analysis was the nursing home discharge. We assessed each discharge as an independent observation. Some individuals had more than 1 discharge in the study period and therefore had more than 1 discharge that occurred up to 6 months prior to a suicide death.

Measures

Independent variables.

We identified resident gender and age at discharge from indicators in MDS and National Patient Care Database records. We categorized age by decade (< 20, 20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and ≥ 80 years). We used MDS indicators as well as diagnosis records in the National Patient Care Database during the period of the nursing home stay and in the preceding 3 years to create indicators of serious mental illness (schizophrenia, bipolar disorder, and other psychoses), dementia, and depression.¹⁶ We assessed behavior by the sum of 3 MDS measures of behavioral symptoms (physically abusive, verbally abusive, and socially inappropriate or disruptive). We assessed each of these component elements on a scale from zero to 3, indicating on how many of the past 7 days the behavior was exhibited (0 = no days; 1 = 1–3 days; 2 = 4–6 days; 3 = daily). The summary behavior score could range from zero to 9, with higher scores indicating greater behavior challenges. We calculated pain scores from MDS items about the frequency and intensity of pain recorded in the most recent MDS assessment for the resident. We categorized pain frequency and intensity as no pain, less than daily pain, daily pain that was mild or moderate, and daily pain that was excruciating.

Outcome.

The primary study outcome was death from suicide within 6 months of live discharge. Although alternative follow-up periods could be examined, we chose a 6-month follow-up period, which is consistent with nursing home quality assurance assessments examining mortality.¹⁷ With regard to all-cause mortality, a 6-month follow-up period was recommended by the Institute of Medicine.¹⁸ We ascertained date and cause of death from National Death Index (NDI) search results. We conducted NDI searches for the VA health system patient population, as described elsewhere.¹⁹ The NDI is considered a gold standard for mortality assessments that use secondary data.²⁰ NDI data include information regarding date and cause of death, derived from death certificate records filed in state vital statistics offices. When NDI searches identified multiple records as potential matches, we determined true matches with established procedures.²¹ As in previous work, we categorized deaths as suicide if the cause of death was International Classification of Diseases, 10th Revision code X60–X84, Y87.0, or U03.^19,22 This approach does not include open-verdict deaths of undetermined intent.

Analyses

We assessed the number of discharges with suicide within 6 months. In bivariate analyses, we compared independent variables for live discharges by whether they were followed by suicide death within 6 months. We calculated suicide rates and standardized mortality ratios for all discharges and separately for men, overall and by age stratum. Although in previous studies of suicide among VA patients the general US population was used as the reference population,¹⁹ our reference population was the VA patient population, and we calculated standardized mortality ratios relative to the average suicide rates from fiscal year 2001 to fiscal year 2008 among VA all-user cohorts, by stratum.

We used multivariable proportional hazards regression to assess associations between risk of suicide and resident age at discharge, gender, psychiatric diagnoses, behavioral summary score, and pain score. Risk time for an observed suicide death began on the day after the live discharge and continued until 6 months after discharge or death, whichever came first. We censored risk time at 6 months for individuals who survived the follow-up period or at date of death for deaths from other causes during the 6-month follow-up period. We used robust covariate sandwich estimators to adjust for the nested nature of the data, with individual discharges clustered by facility.²³ In exploratory analyses, we assessed suicide risk by week in the 6-month follow-up period. In supplemental analyses, we characterized VA nursing home discharges by admission source, length of stay, and discharge status, as indicated in the MDS records. We conducted all analyses with SAS version 9.2 (SAS Institute Inc, Cary, NC).

RESULTS

Between fiscal years 2002 and 2008, VA nursing homes discharged 281 066 live residents. For 53 304 of these discharges, death from any cause occurred within 6 months, including 106 discharges that were followed by suicide deaths. Because unique individuals could contribute risk time following multiple live discharges, these represented 83 unique suicides among 161 164 unique individuals. Table 1 presents suicide rates both in the general VA patient population and in the 6 months following nursing home discharges. These are reported overall and separately for men, and rates for men are also reported by age group. Rates following nursing home discharges are compared with those in the general VA patient population. In the 6 months following live nursing home discharges, crude suicide rates per 100 000 person-years were 89.4 overall and 88.0 for men, and 36.3 and 38.8, respectively, in the general VA patient population. The standardized mortality ratio for suicide rates following nursing home discharge was 2.4 overall (95% confidence interval [CI] = 2.0, 2.9) and 2.3 for men (95% CI = 1.9, 2.8). Age-specific standardized mortality ratios were significantly elevated (95% CI > 1.0) following nursing home discharge for discharges of men older than 40 years and for each age category for those older than 40 years.

TABLE 1—

Suicide Mortality in 6 Months Following US Department of Veterans Affairs Nursing Home Discharges: Fiscal Years 2002–2008

		VA Nursing Home Discharges
Variable	VA General Patient Population,a Suicides/100 000	Discharges Followed by Suicide, No.	Person-Years at Risk	Suicides/100 000, PYAR	Predicted Suicides, No.	Standardized Mortality Ratio, Point Estimate (95% CI)
All	36.3	106	118 571.1	89.4	43.9	2.4 (2.0, 2.9)
Men	38.8	101	114 814.6	88.0	43.5	2.3 (1.9, 2.8)
Age group,b y
< 20	36.2	0	21.9	0.0	0.0	0.0
20–29	36.2	0	284.2	0.0	0.1	0.0
30–39	46.3	0	707.4	0.0	0.3	0.0
40–49	49.7	7	4908.4	142.6	2.4	2.9 (1.2, 5.4)
50–59	42.6	18	22 611.1	79.6	9.6	1.9 (1.1, 2.8)
60–69	31.9	21	23 122.6	90.8	7.4	2.9 (1.8, 4.2)
70–79	33.4	22	33 404.0	65.9	11.2	2.0 (1.2, 2.9)
≥ 80	41.9	33	29 754.9	110.9	12.5	2.7 (1.8, 3.6)

Note. CI = confiddence interval; PYAR = person-years at risk; VA = Department of Veterans Affairs. The sample size was n = 281 066.

^aVA Office of Mental Health Operations.

^bAmong men only; too few women were in the data set to analyze separately.

In sensitivity analyses that excluded discharges with missing information on the measures used in the evaluation of risk factors, the results were not substantially different: the standardized mortality ratio for men was 2.3 (95% CI = 1.8, 2.8). Although we did not present rates for female residents because fewer than 3800 person-years of risk days were available for analysis, the observed trend also indicated elevated rates among women.

For multivariable proportional hazards regression analysis, we restricted the sample to the 237 426 discharges with data on all covariates. Among these discharges, 42 220 were followed by a death from any cause, of which 90 were suicide deaths. Because some individuals had multiple discharges in the study period, these represented 82 unique suicides and 149 962 individuals. In this subsample, we observed no significant differences in any of the relevant variables between discharged patients who died from suicide and those who did not, as evaluated through univariate (Table 2) or multivariate (Table 3) analyses. Both analyses, however, suggested trends toward decreased rates of suicide in individuals diagnosed with dementia.

TABLE 2—

Characteristics of Persons Discharged Alive From US Department of Veterans Affairs Nursing Homes: Fiscal Years 2002–2008

	Overall (n = 237 426), No. (%) or Mean ±SD	No Suicide Within 6 Months (n = 237 336), No. (%) or Mean ±SD	Suicide Within 6 Months (n = 90),a No. (%) or Mean ±SD	χ2 or t (df)	P
Age, y	70.1 ±12.4	70.1 ±12.4	68.5 ±12.5	1.2 (237 424)	.229
Age group, y				4.4 (7)	.734b
< 20	44 (0.0)	44 (0.0)	0
20–29	572 (0.2)	572 (0.2)	0
30–39	1344 (0.6)	1344 (0.6)	0
40–49	9695 (4.1)	9689 (4.1)	6 (6.7)
50–59	44 895 (18.9)	44 877 (18.9)	18 (20.0)
60–69	47 667 (20.1)	47 647 (20.1)	20 (22.2)
70–79	67 654 (28.5)	67 635 (28.5)	19 (21.1)
≥ 80	65 555 (27.6)	65 528 (27.6)	27 (30.0)
Gender				1.8 (1)	.181b
Male	230 046 (96.9)	229 961 (96.9)	85 (94.4)
Female	7380 (3.1)	7375 (3.1)	5 (5.6)
Psychological diagnoses
Serious mental illness	62 677 (26.4)	62 655 (26.4)	22 (24.4)	0.2 (1)	.674
Dementia	85 990 (36.2)	85 966 (36.2)	24 (26.7)	3.6 (1)	.059
Depression	92 399 (38.9)	92 362 (38.9)	37 (41.1)	0.2 (1)	.669
Behavioral scorec	0.19 ±0.81	0.19 ±0.81	0.11 ±0.46	1.6 (89.2)	.116
Pain scaled				2.2 (3)	.532
None	94 529 (39.8)	94 495 (39.8)	34 (37.8)
< daily	53 723 (22.6)	53 707 (22.6)	16 (17.8)
Daily mild/moderate	65 878 (27.8)	65 848 (27.7)	30 (33.3)
Daily excruciating	23 296 (9.8)	23 286 (9.8)	10 (11.1)

Note. Records with missing data were excluded.

^aSome individuals were discharged more than once; number of suicide deaths was 82.

^bCell sizes too small for reliable assessment.

^cSum of 3 measures of behavioral symptoms (physically abusive, verbally abusive, and socially inappropriate or disruptive; total score, range = 0–9).

^dCombined score for frequency and intensity of pain.

TABLE 3—

Hazard Ratios for Suicide Among Patients Discharged From US Department of Veterans Affairs Nursing Homes, by Patient Characteristics: Fiscal Years 2002–2008

Predictor	HR (95% CI)
Gender
Female	1.63 (0.69, 3.87)
Male (Ref)	1.00
Age group, y
< 20	0.00
20–29	0.00
30–39	0.00
40–49	1.44 (0.53, 3.88)
50–59 (Ref)	1.00
60–69	1.13 (0.59, 2.19)
70–79	0.84 (0.42, 1.65)
≥ 80	1.36 (0.63, 2.94)
Psychological diagnosis
Serious mental illness	1.02 (0.60, 1.73)
Dementia	0.66 (0.37, 1.19)
Depression	1.12 (0.73, 1.71)
Behavioral scorea	0.90 (0.67, 1.22)
Pain scaleb
None (Ref)	1.00
< daily	0.78 (0.44, 1.38)
Daily mild/moderate	1.13 (0.63, 2.02)
Daily excruciating	1.06 (0.49, 2.29)

Note. CI = confidence interval; HR = hazard ratio. The sample size was n = 237 426 after exclusion of records with missing data. Results adjusted for clustering of individuals within facilities.

^aSum of 3 measures of behavioral symptoms (physically abusive, verbally abusive, and socially inappropriate or disruptive; total score, range = 0–9).

^bCombined score for frequency and intensity of pain.

We examined characteristics of the VA nursing home stays in the multivariable analysis. The majority were admitted from nonpsychiatric hospital settings (63.0%), followed by home (29.2%), other nursing homes (2.5%), psychiatric hospital settings (1.9%), group homes (1.7%), and other settings (1.7%). The median length of stay was 30 days, and the mean length of stay was 99.2 days (SD = 297.1). Residents were most often discharged to home without home health services (42.6%); 13.9% were discharged to home with home health services, 3.2% to group home settings, 7.6% to another nursing facility, 29.2% to an acute care hospital, and 3.6% to other settings.

Figure 1 shows the unadjusted hazard function of risk of suicide in the 6 months following nursing home discharge. Post hoc exploratory analyses suggested that, although suicide risk was elevated in the entire 6-month period following nursing home discharge, risk was greater in the first 3 weeks following discharge than in the remaining postdischarge period. However, we derived the cutpoint for this assessment empirically; it may not generalize to other VA nursing home cohorts.

FIGURE 1—

Estimated hazard function for suicide over the first 6 months following discharge from US Department of Veterans Affairs nursing homes: fiscal years 2002–2008.

Note. The sample size was n = 237 426.

DISCUSSION

We found that suicide risk in the 6 months following discharge from a VA nursing home was substantially greater than for age- and gender-matched individuals receiving care in the entire VA system. Suicide risk was 2.4 times as high overall and 2.3 times as high for men as for our comparison group. Because of the high prevalence of psychiatric and dementia conditions in nursing home populations and concerns regarding individuals with behavioral problems and pain, we evaluated associations between these conditions and suicide risk in the 6 months following nursing home discharge. These analyses did not identify statistically significant associations between suicide risk and residents’ demographic, clinical, behavioral, and pain characteristics.

The increased rate of suicide among individuals discharged from nursing homes could reflect the broad range of nursing home patients, a group characterized by chronic illness, disability, and care needs that exceed the capacity of their support system. It could also be a reflection of the process of moving from a highly structured and supportive environment to living at home or in other community settings, and this transition may be characterized by discomfort, disability, hopelessness, and dependence.²⁴ Post hoc analyses suggested that risk may be particularly elevated in the first 3 weeks following discharge. Recent studies have examined the effectiveness of follow-up interventions such as postcards, letters, and other forms of contact.²⁵ Although the evidence is mixed,^26,27 development and implementation of effective suicide prevention efforts are particularly important during periods of known elevated risk of suicide. Our results indicate that nursing home discharge begins such a period.

Previous evaluations document that depression, serious mental illness, and pain are significant risk factors for suicide in the VA population.¹³ In one of the few studies to examine postdischarge suicide risk, Lin et al. found that among cancer patients discharged from non-VA inpatient settings, depression was not a significant predictor of suicide in the initial 3 months after discharge.²⁸ The lack of evidence in our study for an impact of these risk factors in nursing home discharges could be related to difficulties with assessment in the nursing home or to the emergence of problems, symptoms, or conditions after individuals were discharged to community settings. Although our analysis did not identify specific resident-level risk factors, it is important that clinicians carefully evaluate potential indicators of suicide risk, such as depression and other mental health conditions and suicidal thoughts, plans, or previous attempts, in nursing home residents.

Suicide prevention is a high priority in the VA health care system, and VA suicide prevention initiatives include mandated suicide risk assessments following a positive depression screen, an innovative 24-hour crisis hotline, and suicide prevention coordinators at all medical centers. Recent VA initiatives include suicide risk assessment and suicide prevention training for staff, with specific resources that focus on assessments and safety planning for older veterans. The VA has responded to our findings with an evaluation of its policies and practices for the management of veterans discharged from its CLCs to the community. The VA is initiating a discharge-planning process, which incorporates an evaluation of caregiving resources and other supports, early in residents’ stays in CLCs, facilitating handoffs from the CLC to providers who will work with veterans after discharge and promoting outreach and follow-up to veterans after discharge, in the clinic or in the home. Of note, the VA has developed a national initiative to integrate full-time, doctoral-level mental health providers into each of its more than 130 Home-Based Primary Care teams, which provide comprehensive, interdisciplinary primary care services in the homes of veterans with complex and chronic disabling disease.

Limitations

Although we examined the most recent available data, important recent developments in VA nursing home care may affect suicide risk following discharge. At the end of the study period, in late 2008, the VA began a cultural transformation of its nursing homes.²⁹ This involved an official name change for nursing home care units, to CLCs; implementation of a person-centered, recovery-oriented model of care; and enhancements to the nursing home environment. Furthermore, the time frame for our data (fiscal years 2002–2008) preceded national VA initiatives to transform its mental health care delivery system to an evidence-based and recovery-oriented model of care.³⁰ One such national initiative and policy development involved the integration of full-time mental health providers into CLCs.³¹ Further study is needed to assess the impact of these changes on suicide risk for CLC residents.

Our findings were derived from VA data and apply only to veterans who receive care from the VA health system. It may be possible to generalize findings from the VA to draw inferences about other nursing homes, but this needs to be confirmed. It is possible that the increased risk may be unique to veterans. From another perspective, it is important to recognize that the VA health care system already supports continuity by incorporating both nursing home and subsequent care in a single electronic medical record and by facilitating follow-up for both mental health and geriatric services in both ambulatory care and home-based primary care. Consequently, it is possible that suicide risk may be greater outside of the VA health care setting. Finally, the number of women discharged from VA nursing homes was relatively low. Further analysis is needed to assess trends in suicide mortality following nursing home discharge, for both VA and non-VA nursing homes.

Conclusions

Although further research is needed to determine whether our findings apply beyond the VA, it is important to recognize the significant opportunities for prevention in both VA and community-based nursing homes. The current version of the MDS (3.0), required as of May 2010 for nursing homes receiving Medicare or Medicaid reimbursement, is an important new resource. It includes a validated rating scale for depression, the 9-item Patient Health Questionnaire, which includes a question about suicidal ideation, “thoughts that you would be better off dead, or of hurting yourself in some way.” It also includes questions related to pain. We did not identify associations between suicide and either pain or depression, but the MDS 3.0 presents new methods for assessment.

In our data, individuals discharged from VA nursing homes were at increased risk for suicide. This suggests the importance of active outreach efforts and attention to residents’ adjustment to living in community settings after discharge. It will be important to assess suicide risk in recent years for both VA and non-VA nursing home facilities. Suicide risk factors should be assessed in care planning both within the nursing home and for postdischarge follow-up care. Targeted suicide prevention efforts should consider the special needs and concerns of individuals following nursing home discharge.