Incidence of Suicide and Association with Palliative Care among Patients with Advanced Lung Cancer

To the Editor:

Suicide is a leading cause of death in the United States, and rates are increasing. Among veterans, the rate is significantly higher compared with adjusted rates in the U.S. population (1). Increased suicide rates also exist among patients with cancer, and patients with lung cancer are at particularly high risk (2). Among patients with advanced cancer, the desire for hastened death was associated with high symptom burden, and low physical functioning and social support (3). Depression is also a risk factor for suicide among patients with cancer (4), which may contribute to the reduced survival observed among these patients (5). Specialist-delivered palliative care is associated with improvements in factors known to reduce suicide risk (6). This study will characterize suicide rates among patients with advanced lung cancer in the Veterans Health Administration and determine the association with palliative care.

Methods

We identified 23,538 patients with advanced lung cancer (stages IIIB and IV) diagnosed January 2007–December 2013, using the Veterans Affairs (VA) Central Cancer Registry. Registry data meet established standards of the Commission on Cancer. Study methods were reported previously (7). Cause of death (COD) was obtained from the joint Department of Veterans Affairs/Department of Defense Suicide Data Repository (SDR), which combined their efforts in merging suicide and COD data to establish a repository. Acquisition of data for the SDR from the National Death Index was supported with collaboration from the National Center for Health Statistics. Additional repository sources include the Social Security Administration Death Register, Military Mortality Database, and VA User Master File. Repository data were available through December 2013. Patients were considered to have committed suicide if cause of death was listed as ICD9-CM: E950.x or ICD10-CM: X60-X84. Patients with unknown COD were excluded (<1%). Palliative care was defined as a specialist-delivered palliative care encounter received after a lung cancer diagnosis determined on the basis of VA inpatient, outpatient, and fee-based encounter data (8).

Patients were dichotomized by COD, and descriptive statistics are presented. Suicide rates, by year, were calculated per 100,000 person-years (number of suicides divided by person-years of survival). VA general population suicide rates (1) standardized by age, sex, and year of death, to match our cohort distribution, were compared with rates in our cohort. Logistic regression was used to determine the associations between suicide and palliative care, and other relevant patient variables (sociodemographics and tumor-related characteristics). Unadjusted odds ratios, given the small number of events, are reported. Variable categories were collapsed to improve model fit. Patients still alive at censoring were excluded from regression analyses due to unknown COD. All modeling used robust standard errors, and statistical testing was two-sided with a threshold of P < 0.05 (STATA, version 14; StataCorp).

Results

Among 20,900 patients, 88% had lung cancer as COD and 30 deaths were attributable to suicide (Table 1). Among the suicide deaths, 93% (28 of 30) resulted from firearms. The overall rate of suicide in the cohort was 210.2 (95% confidence interval [CI], 147.0–300.7) per 100,000 person-years, which represents a 579% (95% CI, 405–828%) increase in the rate of suicide compared with adjusted rates among veterans who use VA health care (Figure 1). Median time from cancer diagnosis to suicide was 91 days (interquartile range, 53–166). Only 6 of 30 (20%) of the patients with suicide as COD received palliative care compared with 11,854 of 20,870 (57%) among patients with another COD. Palliative care was associated with decreased odds of suicide (unadjusted odds ratio [OR], 0.19; 95% CI, 0.08–0.47; P < 0.001) (Table 2). Non-Hispanic white race (OR, 6.4; 95% CI, 1.53–26.9; P = 0.01) compared with other races and lower Charlson Comorbidity Index (OR, 0.85; 95% CI, 0.76–0.97; P = 0.01) were both associated with increased odds of suicide.

Table 1.

Patient characteristics

Characteristic	Suicide (n = 30)	Other COD (n = 20,870)
Age, mean (SD), yr	67 (8.6)	68 (9.6)
Race
NH-white	28 (93)	14,318 (69)
NH-black	—	3,158 (15)
Hispanic	2 (7)	3,344 (16)
Unknown	—	50 (<1)
Sex, male	30 (100)	20,525 (98)
Marital status
Married	11 (37)	9,002 (43)
Not married	19 (63)	11,825 (57)
Unknown	—	43 (<1)
Income, U.S. dollars
≤44,000	10 (33)	7,063 (34)
>44,000–52,999	11 (37)	6,469 (31)
≥53,000	9 (30)	7,256 (35)
Unknown	—	82 (<1)
Tobacco use
Current	16 (53)	11,714 (56)
Former	11 (37)	7,670 (37)
Never	—	585 (3)
Unknown	3 (10)	901 (4)
Home residence
Urban	19 (63)	14,618 (70)
Rural	11 (37)	5,960 (29)
Unknown	—	292 (1)
FCI score, mean (SD)	2.9 (2.3)	3.2 (2.4)
CCI score, mean (SD)	2.8 (2.8)	4.6 (3.8)
AJCC-TNM stage
IIIB	5 (17)	2,134 (10)
IV	25 (83)	18,736 (90)
Histology
NSCLC	22 (73)	14,267 (68)
SCLC	5 (17)	3,873 (19)
Other	3 (10)	2,730 (13)
Cancer treatment received
Radiation	8 (27)	7,428 (36)
Chemotherapy	14 (47)	9,467 (45)
Cause of death
Lung cancer	—	18,412 (88)
Infection/hematologic malignancy/other	—	1,060 (5)
Nonlung solid tumor	—	865 (4)
Heart disease	—	235 (1)
COPD	—	298 (1)
Suicide	30 (100)	—
Palliative care receipt	6 (20)	11,854 (57)
Interval from cancer diagnosis,* median (IQR), d	81 (16–371)	37 (10–133)
Survival from cancer diagnosis, median (IQR), d	91 (53–166)	117 (47–270)

Definition of abbreviations: AJCC-TNM = American Joint Committee on Cancer-Tumor Node Metastasis; CCI = Charlson Comorbidity Index; COD = cause of death; COPD = chronic obstructive pulmonary disease; FCI = Functional Comorbidity Index; IQR = interquartile range; NH = non-Hispanic; NSCLC = non–small cell lung cancer; SCLC = small-cell lung cancer; SD = standard deviation.

Data represent n (%) unless otherwise noted.

^*Denominator represents patients who received palliative care.

Figure 1.

Incidence of suicide by year of death. The suicide rate among patients with advanced-stage lung cancer (blue column portions) compared with the age, sex, and year of death adjusted rate from the overall Veterans Affairs (VA) population (red column portions). From 2007 to 2013, the suicide rate was 210.2 (95% confidence interval, 147.0–300.7) per 100,000 person-years, which is a 579% difference compared with the expected rate in the VA population over the same time period, which was 36.3/100,000 person-years. Error bars represent 95% confidence intervals.

Table 2.

Unadjusted associations with suicide as cause of death

Variable	Univariate Odds Ratio (95% CI)	P Value
Palliative care	0.19 (0.08–0.47)	<0.001
Patient characteristics*
Age	0.99 (0.96–1.03)	0.70
Race
Other	Reference	—
NH-white	6.4 (1.53–26.9)	0.01
Marital status
Not married	Reference	—
Married	0.76 (0.36–1.60)	0.48
FCI score	0.94 (0.81–1.10)	0.46
CCI score	0.85 (0.76–0.97)	0.01
Tumor characteristics
AJCC-TNM stage
IIIB	Reference	—
IV	0.57 (0.22–1.49)	0.25
Year of cancer diagnosis	1.01 (0.83–1.22)	0.94

Definition of abbreviations: AJCC-TNM = American Joint Committee on Cancer-Tumor Node Metastasis; CCI = Charlson Comorbidity Index; CI = confidence interval; FCI = Functional Comorbidity Index; NH = non-Hispanic.

^*Characteristics that were different among groups (suicide vs. other cause of death) or are known to be associated with suicide rates are included. Sex, income, tobacco use, home residence rurality, and cancer histology were not associated with suicide.

Discussion

This is the first study to report suicide rates among patients with lung cancer receiving VA health care. Rates were considerably higher than age-, sex-, and year of death–adjusted rates among the veteran population who use VA health care. Palliative care was associated with decreased risk of death from suicide. Patients with lung cancer, particularly advanced stage, suffer from significant physical and psychological symptom burden attributable to their disease. Palliative care offers an approach focused on reducing symptom burden and improving quality of life (QOL), which may influence suicide risk.

Using data from the Surveillance, Epidemiology, and End Results (SEER) program on patients with cancer through 2002, a suicide rate nearly twice that in the general population was observed (2). Rates were highest among those with lung and stomach cancer. We found suicide rates among veterans with advanced lung cancer considerably higher than those in the veteran population. Among all veterans, high suicide rates are partially attributed to increased rates of mental health and substance abuse disorders (9). Our study identifies an important subgroup of patients at considerable risk of suicide in need of enhanced detection, focused prevention, and close follow-up, particularly around the time of cancer diagnosis. Although long-term survival among patients with lung cancer is low, suicide is a potentially preventable COD.

Palliative care establishes patient values and preferences while prioritizing QOL by optimizing psychological, social, and spiritual support. A meta-analysis examining palliative care demonstrated improvements in both QOL and symptom burden, but not improved survival (6). However, QOL is an independent prognostic factor for survival in patients with lung cancer (10). Palliative care integration has been suggested by oncologic and thoracic societies; however, uptake has been slow and it continues to be underutilized in patients with cancer. Our results reinforce the importance of timely palliative care among patients with advanced lung cancer.

Study limitations include misclassification bias for COD; however, the SDR is a comprehensive database specifically designed to capture COD. Depression before cancer diagnosis was included in the Functional Comorbidity Index; however, confounding by other comorbid psychiatric conditions was not studied. Suicides, not suicide attempts, were studied. Palliative care delivered by nonspecialists (e.g., primary care clinicians) was not measured. Alcohol and illicit drug use are associated with suicide, but reliable estimates were unavailable. Higher health care use could confound the relationship between palliative care and suicide; however, lung cancer treatments received were similar between groups. The potential for immortal time bias exists as patients had to live long enough to receive a palliative care encounter.

Veterans with advanced lung cancer are at substantially increased risk for suicide. Palliative care is associated with a decreased risk of suicide, and elements of this patient-centered approach may inform enhanced suicide prevention and treatment efforts, which are urgently needed. This study adds to the literature in support of timely palliative care among patients with lung cancer.