The Relationship of Post-traumatic Stress Disorder to End-of-life Care Received by Dying Veterans: a Secondary Data Analysis

INTRODUCTION

Despite its inevitability, death is often considered a poor outcome in medicine and in society. Accepting a poor prognosis and navigating the last months of life can be challenging for anyone¹ and is potentially harder for individuals predisposed to depression or anxiety.²

Generally, individuals with psychiatric illness have higher healthcare utilization,^3–6 lower quality of life and functional status,^7–10 and worse medical outcomes^{10, 11} than the general population. Individuals with psychiatric illness may have difficulty coping with life stressors¹² and are at risk for relapse or worsening symptoms during major life changes, such as retirement, divorce, or death.^{13, 14} Individuals with post-traumatic stress disorder (PTSD) may be specifically at risk for relapse or worsening symptoms as their own death approaches because PTSD requires exposure to a trauma, specifically near death or threatened death.¹⁵ Due to the traumatic encoding of memories in PTSD,¹⁶ reminders of one’s own death may consciously or unconsciously trigger reminders of prior trauma. Thus, in terminal illness, any reminders of impending death could be re-traumatizing or exacerbate the symptoms of patients with PTSD. However, the effects of PTSD on symptom burden and healthcare utilization in the last months of life have not been well studied.

Diagnostically, PTSD consists of four clusters, re-experiencing, avoidance, negative cognitions and mood, and increased arousal.¹⁵ PTSD frequently co-exists with other psychiatric illness, such as mood or substance use.^{17, 18} PTSD also negatively affects pain sensation and interpretation,^19–21 causes poor sleep,¹⁵ and can lead to somatization.^{22, 23} PTSD’s course can be chronically symptomatic, resolving over time, remitting and relapsing, or with late onset.^24–29 Trajectories vary depending on trauma type, treatment, and other factors.^{25–27, 30–32} Life stressors, such as retirement or death of loved ones, have been correlated with PTSD relapse or worsening symptoms.^32–36 Specifically, the theory of later-adulthood trauma re-engagement describes later-life worsening of PTSD symptoms as a part of life reminiscence, which is triggered by common older life events, such as physical or cognitive decline, role transitions, or deaths of loved ones.^{37, 38} However, it is unclear how often PTSD symptoms occur or if they impact medical care in the final months and days of life, as death nears.

United States (US) PTSD prevalence rates are 3.5% current and 6.8% lifetime, but as low as 2.5% in the older population.^39–42 Studies have quoted increased PTSD rates in various life-threatening medical conditions: 2.3–18% in cancer,^43–45 30–54% in human immunodeficiency virus (HIV),^{46, 47} and in 3.7–59% of patients admitted to the intensive care unit (ICU).^48–50 These rates vary, as different tools and time periods were used to diagnose PTSD. The gold standard for PTSD diagnosis is a structured clinical interview, and symptoms must persist for at least one month.¹⁵ PTSD rates are lower when using these criteria.^{45, 48}

PTSD is often studied within Veterans Affairs (VA), due to combat trauma exposure. VA is also the largest national healthcare system to routinely screen all patients for PTSD.⁵¹ Outside of VA, PTSD screening varies and is often under-diagnosed.^52–54 Therefore, to test our hypothesis that PTSD symptoms worsen at the end of life, manifested by increased healthcare utilization or increased medication use, we sought to study PTSD in VA. We conducted an exploratory, secondary analysis of data describing end-of-life care received by veterans, dying in VA medical centers from the Best Practices for End-of-Life Care for Our Nation’s Veterans (BEACON) trial.⁵⁵

METHODS

RESULTS

DISCUSSION

Our findings provide new empirical evidence supporting the hypothesis that individuals with PTSD have increased symptom burden in the final months and days of life. In comparing veterans with and without PTSD dying in a VA medical facility, those with PTSD were more likely to have previously been admitted to a VA facility in the past 12 months and were more likely to have received antipsychotics in the last 7 days of life. PTSD, by definition, requires exposure to a traumatic event involving near death or threatened death.¹⁵ Subsequent trauma can cause a recurrence or worsening of PTSD symptoms.^{61, 62} Therefore, we hypothesized that as death nears, PTSD symptoms would worsen. Obtaining evidence that PTSD impacts the end of life will facilitate future work exploring this impact and developing interventions to improve the end-of-life experience.

Because prospective data collection in hospice and palliative care populations can be challenging,^{63, 64} we used medications and healthcare utilization as proxies for PTSD symptom burden. Patients with psychiatric illness typically have increased healthcare utilization.^3–6 Those in mid-life with PTSD may have the highest utilization.⁶⁵ Because this utilization is associated with symptom burden,^65–67 we hypothesized that continued or worsening PTSD symptoms would thus contribute to increased end-of-life healthcare utilization.

Moving beyond symptoms, because PTSD is frequently comorbid with other psychiatric illnesses,^{17, 18, 68} we needed to determine if PTSD was a separately detectable issue at the end-of-life. Importantly, we found that only PTSD and schizophrenia were associated with administration of antipsychotics in the last 7 days of life. The schizophrenia finding was unsurprising, as the data did not consider medications prior to the last 7 days of life. While patients with schizophrenia would continue to receive antipsychotics in the final days of life, antipsychotics are not recommended for PTSD treatment⁶⁹ nor are they routinely prescribed for PTSD in the VA setting.^{70, 71}

For example, antipsychotic prescription rates in PTSD in VA in 2009 were 5.1% for second-generation antipsychotics in veterans age 65 and older⁷¹ and 13.9% for any antipsychotic overall.⁷⁰ Our data was collected between 2005 and 2011, and the rates of receiving any antipsychotic in the last 7 days of life with and without PTSD were 26.3% and 15.6%, respectively. Antipsychotics have been associated with PTSD in patients with dementia.⁷¹ However, we controlled for dementia, and it was not associated with antipsychotic administration.

Antipsychotics are commonly prescribed in palliative care and hospice settings^{72, 73} and are recommended for short-term treatment of delirium.^{74, 75} However, black box warnings recommend against the long-term use of antipsychotics in elderly patients or those with dementia.^{76, 77} Antipsychotics are less commonly prescribed upon hospice admission than opioids or benzodiazepines. ^{73, 78} However, prescriptions increase as death nears,^{73, 78} initiated an average of 2–5 days prior to death.⁷³ For additional context, in veterans receiving palliative care, but not actively dying, there was no difference in antipsychotic prescription rates between patients with or without PTSD.⁷⁹ We cannot be certain, but these studies suggest that the antipsychotics provided to veterans with PTSD in our study were newly ordered in response to symptoms in the imminently dying.

To potentially explain our observed association with PTSD and antipsychotics, PTSD has been associated with anesthesia emergence delirium^80–82 and ICU delirium.^{83, 84} However, to date, PTSD has not been identified as a risk factor for terminal delirium,^{85, 86} which is treated first-line with antipsychotics.⁷⁴ A detailed discussion about the psycho-neuro-pathophysiology of PTSD and how it may predispose to delirium is beyond the context of this paper but has been described elsewhere.^87–91 Briefly, in a semi-conscious state, an unknown stimulus, such as monitor beeping, is often interpreted as something dangerous, something to fear, especially in patients with PTSD. Generally, patients with PTSD are also less able to extinguish a fear response. Therefore, it is more likely that the continued fear response signal will cause physical agitation, such as pulling at tubes or lines. While some literature suggests that ICU exposure causes PTSD and not the opposite,^{48, 50} controlling for pre-existing PTSD or psychiatric illness removes this association.^{84, 92, 93}

Due to the high psychiatric comorbidity in PTSD and in our study, it was unsurprising that PTSD was not associated with the administration of opioids or with benzodiazepines in the last 7 days of life. By controlling for comorbid psychiatric diagnoses, we were attempting to isolate the effects of PTSD on end-of-life medication administration. However, in patients with multiple psychiatric diagnoses, attributing one specific symptom to one specific diagnosis can be challenging.^{94, 95} Still, we feel that the correlations observed between opioids and depression and benzodiazepines with depression and anxiety suggest that in the final days of life, psychiatric illness may contribute to symptom burden.

We also considered that symptom burden may lead to increased healthcare utilization. We found that VA hospitalizations in the last 12 months of life were increased in comparing patients with and without PTSD; ER visits were not statistically different. Generally, patients with psychiatric illness, including those with PTSD, have increased healthcare utilization,^65–67 if they have adequate healthcare access and do not feel stigmatized.^{96, 97} As VA provides expertise in veteran issues, especially in PTSD,⁵¹ veterans may feel less stigma and more comfortable seeking VA care.^{98, 99} In contrast, in Washington state, patients with psychiatric illness had decreased inpatient and ICU utilization, with increased ER use in the last 30 days of life.¹⁰⁰ However, all psychiatric illnesses in this study were grouped together, which may not be ideal, as end-of-life care patterns can be more nuanced. For example, pre-existing depression was associated with earlier hospice use and less acute care in patients with lung cancer.¹⁰¹ A similar study found that patients with pre-existing depression or post-cancer depression both were more likely to use hospice than those without depression.¹⁰² However, patients with lung cancer and increased depression or anxiety symptoms were more likely to receive chemotherapy in the last 14 days of life.¹⁰³ Thus, further study of psychiatric illness and its impact on end-of-life healthcare utilization is needed.

It was unexpected that advanced directive documentation was more common in veterans with PTSD, because we had hypothesized that avoidant coping in PTSD¹⁵ would decrease advance care planning (ACP). However, it is possible that repeated health system contact via higher utilization led to this finding.^{104, 105} In contrast, there was a trend towards veterans with PTSD being less likely to have a DNR order. Thus, it is difficult to discern the care choices made, or if PTSD impacted these choices. It is also possible that the BEACON intervention itself, emphasizing ACP and decreasing unwanted aggressive end-of life care, impacted our ACP findings.

We acknowledge that defining PTSD based on the patient problem list is a limitation. Our PTSD rate was 8.7%, with a median age of 65 for those patients, compared to lifetime PTSD prevalence of 3.7% for US veterans age 65 or older.⁶⁸ In VA, using ICD-9 codes to detect PTSD has a 72–84% positive predictive value.¹⁰⁶ Also, using ICD-9 codes for PTSD has been demonstrated to adequately separate patients with the most PTSD symptoms (true positives) from those with the least symptoms (true negatives), ¹⁰⁷ symptoms being a dimension of interest in our study. However, without further information, PTSD rates using the problem list would be less than lifetime rates. Thus, our PTSD rate is likely an underestimate. However, an underestimate would bias the findings toward the null.

Further, our study focused on predominantly male veterans, dying within VA medical centers. Our findings only generalize to similar settings, as patients dying at home may have different end-of-life symptoms and utilization patterns. Non-VA healthcare use was unaccounted for and could have influenced our utilization findings. Additionally, descriptions of patient symptoms, rationales for medication administration, and PTSD severity were lacking. Other possible confounders, such as chronic pain, traumatic brain injury, or socioeconomic disadvantage, could not be accounted for. The strengths of this study include its large sample size and detailed information about end-of-life care processes metrics. Also, with 31.6% African-Americans, this sample was more racially representative than other VA-based studies, important because of the higher PTSD rates noted in minorities.^{18, 68}

In conclusion, our findings provide new evidence suggesting increased symptom burden for individuals with PTSD at the end-of-life and that PTSD may be a separate risk factor for terminal delirium. Future studies should examine the impact of PTSD and other psychiatric illnesses on the end-of-life experience in all patient populations, not only veterans. The impact of symptom severity, prior treatment for psychiatric illness, physical and psychiatric comorbidities, and type of trauma experienced should also be considered.

Electronic supplementary material

Funding Information

This research was funded by a grant from the Veterans Administration Health Services Research and Development (HSR&D) Program, “Impact of An Intervention to Improve Care at Life’s End in VA Medical Centers” (IIR 03-126, PI: KL Burgio, Co-PI: FA Bailey). Dr. Bickel was funded by the National Institute on Aging of the National Institutes of Health under award number 5T32AG044296. The views expressed are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs, the United States Government, or the National Institutes of Health.

Compliance with Ethical Standards