Causes of death among United States decedents with ALS: An eye toward delaying mortality

Abstract

Objective

To report multiple cause of death (MCOD) occurrence among patients in the United States with amyotrophic lateral sclerosis (ALS).

Methods

Using death certificate data for all ALS deaths from 50 U.S. states and the District of Columbia, 2011–2014, we tabulated MCOD, used association rules mining (ARM) to determine if MCOD occurred together, and calculated standardized mortality odds ratios (SMOR) for select causes, comparing ALS with other U.S. decedents.

Results

Among 24,328 death certificates, there were 25,704 MCOD, excluding ALS. ALS was listed as the sole cause of death in n = 11,263 (46%). The most frequent causes of death co‐occurring with ALS were respiratory failure (n = 6503; 25.3%), cardiovascular disease (n = 6077; 12.6%), pneumonia (n = 1345; 5.2%), and pneumonitis (n = 856; 3.3%). The SMORs among ALS decedents compared with non‐ALS decedents for falls and accidents were 3.4 (95% CI 2.6, 4.3) and 3.0 (95% CI 2.2, 4.2), respectively. From ARM analysis, falls and accidents were both associated with injuries. The most common causes identified were weakly to very strongly associated with being an ALS decedent compared with other U.S. deaths, with SMOR point estimates ranging from 1.3 to 51.1.

Interpretation

This study provides information about the natural history of ALS. With knowledge that some causes of death may be preventable, healthcare providers may be able to optimize patient care and possibly postpone mortality and reduce morbidity. Moreover, this study located gaps in data; medical certifiers completing death certificates for ALS decedents should ensure all MCOD data are recorded.

Introduction

Amyotrophic lateral sclerosis (ALS) is a progressive and fatal neurodegenerative disease involving the cortical, brainstem, and spinal motor neurons. ¹ This neuronal cell loss causes weakness in muscles involved in speaking, swallowing, limb movement, and breathing. Further, up to 50% of individuals with ALS develop changes in cognitive function, indicating that the disease is not just limited to the motor system. ² Tracheostomy‐free median survival is often cited in the range of 2–5 years, with variability based on factors like onset age and whether symptoms begin in cranial or limb muscles. Respiratory failure is the most common cause of death among patients with ALS, due to muscle paralysis, yet other comorbid conditions such as cardiovascular disease may contribute to ALS deaths. ³ In the United States, death certificates contain a field for the underlying cause of death, which is defined as the disease or injury which initiated the series of events leading directly to death. ⁴ The instructions for completing the death certificate include directions to report the full sequence of events from the underlying to the immediate cause of death. ⁵ Thus, death certificates contain fields for all causes of death, also known as multiple causes of death (MCOD). While death certificate data have been used for prior ALS studies, ⁶ few studies have examined MCOD. ⁷ , ⁸ , ⁹ , ¹⁰ However, complete information on cause of death has the potential to guide clinical care of patients with ALS, specifically by highlighting preventable causes of death that could influence future patient management. For the present study, we used a retrospective design to examine a large number of U.S. death certificates from 2011 to 2014. Deceased patients with ALS were compared with a control population consisting of deaths from all other causes. Our objectives for the present analyses were to (1) tabulate MCOD occurrences among U.S. ALS patients, with the aim of determining if certain causes of death predominate or occur together, (2) to estimate standardized, cause‐specific mortality among patients with ALS, and (3) identify preventable causes of death.

Methods

Identification of ALS deaths

The use of de‐identified national mortality data for this study was approved by the Institutional Review Board for the Centers for Disease Control and Prevention (CDC). We previously obtained multiple causes mortality data for all U.S. deaths involving motor neuron disease (MND, coded as G12.2 in the International Classification of Disease, 10th Revision [ICD‐10]) from the National Center for Health Statistics (NCHS), CDC. ¹¹ To be included in the initial listing of MND deaths, G12.2 must have appeared in the underlying or multiple causes of death fields. Although an ICD‐10 code exists for ALS (G12.21), coded causes of death from NCHS are only available with a maximum of four characters. Consequently, we then reviewed the verbatim (uncoded) causes of death, also obtained from NCHS, as they appeared on the death certificate to exclude records for MNDs other than ALS. ¹¹ Thus, due to the way the ALS decedents were identified, for the rest of this paper G12.2 is synonymous with ALS. We define preventable as a death that may be related to ALS symptoms and has an associated intervention but is not a result of ALS natural progression in a nonmechanically ventilated patient.

Multiple cause of death analyses

To tabulate MCOD occurrences for the present analysis, we first created a long‐format data set of MCOD codes (i.e., a data set with a record for every MCOD from each decedent). Thus, the unit of analysis here is MCOD, not the decedent. Since by study design, all decedents included an MCOD for ALS, we next deleted MCOD for G12.2. All MCOD codes were then recoded into groupings developed by CDC's National Center for Health Statistics to allow the identification of causes of death with public health and medical importance. ¹² In addition to those 113 cause groupings, we created ALS‐relevant groupings for respiratory failure, pneumonia, aphagia and dysphagia, foreign object in the respiratory tract or esophagus, dementia, injury, poisoning and other external causes, and falls. We selected these additional causes due either to their association with ALS mortality in other studies or our a priori expectations of MCOD codes among ALS decedents. We reported the frequency of a disease grouping if it occurred >50 times (which we define as common causes of death) and placed other disease groupings and the cause groupings “All other diseases” and “Symptoms not elsewhere classified” in an “other causes” category (defined as uncommon causes of death). We also reported the most common MCOD codes occurring within each of these disease groupings.

Association rules mining analysis

To detect disease groupings that occur together in patients with ALS, we used association rules mining (ARM), a machine‐learning technique used to find causal relationships between objects. ¹³ In ARM, a computer algorithm is used to create rules describing which objects tend to occur together. ARM was originally developed to analyze consumer shopping patterns but can be applied to other domains, including studies in medicine. ¹⁴ Rules are stated as “if item X, then item Y” and an example from grocery retail is “if a shopping cart contains cornflakes, then it also contains milk.” Association rules are evaluated using three metrics ¹³ : (i) Support is the number of transactions that include items in the X and Y parts of the rule as a percentage of the total number of transactions (i.e., a measure of how frequently the collection of items occur together as a percentage of all transactions); (ii) Confidence is the fraction of the time the rule is true; and (iii) Lift is the ratio of observed support to expected support; if the lift is near 1, then it is likely occurring by chance, but the larger lift the more likely the rule is “true.” We screened rules by setting their minimum support to 0.005, minimum confidence to 0.5, and minimum lift to 1.1 and report the Fisher's exact test result for the statistical significance of the lift for each rule. The “arules” package ¹⁵ in the R statistical programming environment, Version 4.0 ¹⁶ was used for the ARM analysis.

Standardized mortality odds ratio analyses

For selected MCOD groupings, we then calculated standardized mortality odds ratios (SMORs), a measure of the association between a cause of interest and membership in the group under study (here, ALS decedents). ¹⁷ The first step in computing an SMOR is to calculate the ratio of deaths from a cause of interest (e.g., respiratory failure) serving as “cases” to deaths from a second cause (which can be thought of as a “control” cause) in the study population. This ratio is then compared with the equivalent ratio from a standard population. For this analysis, we considered deaths from select comorbidities as cases and deaths from all cancers (with exceptions) as controls, among ALS decedents compared with all U.S. deaths, 2011–2014. We chose all cancers (ICD‐10 “C” codes) as “control” causes, with the exception of prostate (C61), melanoma (C43), and tongue (C01 and C02), as those three may be associated with ALS. ¹⁸ Thus, here the SMOR is the comparison of the ratio of cause of interest to cancer deaths in the study population to the same ratio in the control population.

To calculate SMORs, we used logistic regression with case status (coded as either the cause of interest or cancer) as the outcome variable and exposure (ALS vs. all U.S. decedents) and age at death (categorized as <60, 60–69, 70–79, and ≥80 years) as predictors. U.S. MCOD data were obtained from the Centers for Disease Control and Prevention's Wide‐ranging Online Data for Epidemiologic Research (CDC WONDER) ¹⁹ for 2011–2014. We used SAS, Version 9.4 software ²⁰ for all analyses, except for the ARM analysis.

Results

ALS deaths

During 2011–2014, 24,328 ALS decedents were identified by death certificates (Table 1). Using the same data, we previously reported statistically stable annual ALS mortality rates that were also consistent with MND mortality rates reported for a similar period. ¹¹ , ²¹ Male (n = 13,611; 55.9%), White (n = 22,245; 91.4%), and non‐Hispanic persons (n = 23,103; 95.0%) predominated. Few decedents (n = 1761; 7.2%) were younger than age 50, and the majority were categorized as age 60–69 (n = 7635; 31.4%) or 70–79 years (n = 6931; 28.5%). These patterns of patient factors are consistent with those observed in U.S. ALS prevalence and incidence data. ²² , ²³ Among all decedents, there were 25,704 MCODs recorded, excluding ALS. Of note, ALS was the sole cause of death for 11,263 (46%) decedents.

Table 1.

Characteristics of ALS‐associated deaths in the United States during 2011–2014.

	n	%
All deaths	24,328	100.0
Sex
Male	13,611	55.9
Female	10,717	44.1
Race
White	22,245	91.4
Black	1540	6.3
Other	543	2.2
Hispanic origin
Hispanic	1225	5.0
Non‐Hispanic	23,103	95.0
Age group, years
≤49	1761	7.2
50–59	4461	18.3
60–69	7635	31.4
70–79	6931	28.5
≥80	3540	14.6

Frequencies of MCOD

Table 2 shows the frequency of disease groupings occurring more than 50 times (n = 23). Respiratory failure was the most frequent MCOD (n = 6503); it with major cardiovascular disease excluding cardiac arrest (n = 4084), cardiac arrest (n = 1993), pneumonia (n = 1345), and pneumonitis (n = 846) comprising 58% of all causes (excluding G12.2). Figure 1 shows the degree of overlap of these MCODs in the same decedent. Respiratory failure and major cardiovascular disease mostly occurred in isolation, but also were the most frequent intersection of MCOD. Pneumonia and respiratory failure were the second (n = 449) and cardiac arrest and other major cardiovascular diseases were the third (n = 334) most frequent intersections. There were 1002 deaths were due to pneumonia and/or pneumonitis in the absence of respiratory failure and cardiovascular disease (Fig. 1). There were 5519 (21.5%) MCOD categorized as “other causes,” and the remaining causes in Table 2 were relatively rare and ranged from 0.2 to 3.3% of the total.

Table 2.

Occurrence of selected causes of death.

Cause	N	%
Respiratory failure a	6503	25.3
Major cardiovascular disease, excluding cardiac arrest	4084	15.9
Cardiac arrest	1993	7.8
Pneumonia	1345	5.2
Pneumonitis	846	3.3
Chronic lower respiratory diseases	842	3.3
Diabetes mellitus	747	2.9
Aphagia and dysphagia	541	2.1
Malignant neoplasms	515	2.0
Foreign object in the respiratory tract or esophagus b	488	1.9
Septicemia	477	1.9
Dementia c	403	1.6
Other diseases of the respiratory system	254	1.0
Nephritis, nephrotic syndrome, and nephrosis	209	0.8
Nutritional deficiencies	203	0.8
Injury d	125	0.5
Other disorders of the circulatory system	121	0.5
Anemias	114	0.4
Poisoning and other external causes e	106	0.4
Parkinson's disease	75	0.3
Falls	74	0.3
Other infectious diseases	66	0.3
Accidents	54	0.2
Other causes	5519	21.5

Causes occurring fewer than 50 times were aggregated as “other causes.”

^a J96 (respiratory failure not elsewhere classified) and R09.2 (respiratory arrest).

^b T17.2–T17.9 (foreign object in the respiratory tract), T18.1 (foreign object in esophagus), W79 and W80 (inhalation and ingestion of food/other objects causing obstruction of respiratory tract).

^c F01.0 (vascular dementia of acute onset), F01.1 (multi‐infarct dementia), F01.2 (subcortical vascular dementia); F01.3 (mixed cortical and subcortical vascular dementia), F01.8 (other vascular dementia); F01.9 (vascular dementia, unspecified); F03 (unspecified dementia), G30 (Alzheimer's disease), G31 (other degenerative diseases of the nervous system, not elsewhere classified).

^d All ICD‐10S codes.

^e All ICD‐10T codes, excluding those codes, used to define “Foreign object in respiratory tract or esophagus.”

Figure 1.

Upset plot shows overlap of select major causes of death in the same decedent. The horizontal bars show the number of each cause of death. The dots and lines represent subsets of causes. The histogram represents the number of causes in each subset. Major CV disease, major cardiovascular disease excluding cardiac arrest.

The most frequent MCOD codes occurring within each major cause grouping are shown in Table S1. Some groupings were dominated by a single MCOD (e.g., “protein‐calorie malnutrition” comprised 92% of the category “nutritional deficiencies”). Others, such as “other infectious diseases,” contained many singular MCOD codes that occurred with relatively low frequency. These results allow inspection of cause of death in greater detail than by major category. MCOD categories associated with preventable interventions include foreign objects in respiratory tract or esophagus, nutritional deficiencies, injury, poisoning and other external causes, falls, and accidents. Of interest, among 488 deaths due to foreign objects in the respiratory tract or esophagus, 154 deaths did not include respiratory failure. Similarly, of the 125 deaths due to injury, 96 were without respiratory failure. These results indicate that preventable deaths frequently occur independent of respiratory failure.

Association rules mining

Table 3 shows the association rules meeting our criteria (n = 10) and are sorted by descending lift. Again, lift compares the frequency of a rule with how often it would be expected to occur by chance. The most meaningful rules involved falls or accidents. If a patient had a fall or accident, then they also tended to have an injury. Lift for both rules was very high (>36) and statistically significant (p < 0.001). Lift for other rules in Table 3 was much smaller (i.e., between 1 and 2) but was also statistically significant. Among these other rules, “if pneumonia, then respiratory failure” and “if diabetes mellitus then major cardiovascular disease excluding cardiac arrest” had the greatest support (0.089 and 0.074, respectively) and count (n = 562 and n = 464, respectively). Nonetheless, these results suggest that the most frequent causes of ALS deaths (respiratory failure, cardiac arrest, and other major cardiovascular diseases) occur together only rarely with preventable causes of ALS deaths, based on our inclusion criteria for the rules shown in Table 3.

Table 3.

Association rules sorted by descending lift.

Rules (left hand side → right hand side)	Support	Confidence	Lift	Count	p value for lift
[Falls] → [Injury]	0.009	0.757	44.899	56	<0.0001
[Accidents] → [Injury]	0.005	0.623	36.941	33	<0.0001
[Chronic lower respiratory diseases, Diabetes mellitus] → [Major CV disease]	0.007	0.742	1.925	46	<0.0001
[Diabetes mellitus] → [Major CV disease]	0.074	0.704	1.827	464	<0.0001
[All other diseases, Diabetes mellitus, Respiratory failure] → [Major CV disease]	0.005	0.647	1.679	33	<0.0001
[Diabetes mellitus, Respiratory failure] → [Major CV disease]	0.017	0.588	1.526	110	<0.0001
[Other disorders of the circulatory system] → [Major CV disease]	0.010	0.570	1.479	65	<0.0001
[Cardiac arrest, Chronic lower respiratory diseases] → [Major CV disease]	0.006	0.500	1.297	38	0.0270
[Pneumonia] → [Respiratory failure]	0.089	0.578	1.211	562	<0.0001
[Pneumonitis, Symptoms not elsewhere classified] → [Respiratory failure]	0.006	0.527	1.105	39	0.0227

Major CV disease, major cardiovascular disease excluding cardiac arrest.

Comparisons with national deaths

Table 4 shows the odds of death for the causes shown in Table 2, except for malignant neoplasms and other causes. The odds of death involving aphagia and dysphagia, foreign object in the respiratory tract or esophagus, respiratory failure, pneumonitis, chronic lower respiratory disease, pneumonia, major cardiovascular disease (excluding cardiac arrest), nutritional deficiencies, or cardiac arrest were 9.8 to 51.1 times greater for ALS than of all U.S. deaths. These associations are very strong and statistically significant (p < 0.001). Ten causes were strongly associated with ALS, with SMORs ranging from 2.9 to 7.8 and were statistically significant (p < 0.001). Of these, accidents and falls are preventable. The remaining four causes were either weakly associated (SMOR 1.3–1.4; p < 0.05) or not associated with ALS.

Table 4.

Standardized mortality odds ratios (SMORs) for select major categories of cause of death.

Cause of death	SMOR (95% CI)	Model p value
Aphagia or dysphagia	51.1 (44.6, 58.7)	<0.0001
Foreign object in respiratory tract or esophagus	44.7 (39.1, 51.1)	<0.0001
Respiratory failure	37.2 (33.7, 41.1)	<0.0001
Pneumonitis	30.5 (27.0, 34.4)	<0.0001
Chronic lower respiratory disease	30.1 (26.7, 33.9)	<0.0001
Pneumonia	16.7 (15.0, 18.7)	<0.0001
Major cardiovascular disease, excluding cardiac arrest	15.2 (13.8, 16.9)	<0.0001
Nutritional deficiencies	13.9 (11.8, 16.5)	<0.0001
Cardiac arrest	9.8 (8.8, 10.9)	<0.0001
Other disorders of the circulatory system	4.7 (3.8, 5.8)	<0.0001
Diabetes mellitus	4.6 (4.1, 5.2)	<0.0001
Other infectious diseases	4.3 (3.3, 5.6)	<0.0001
Septicemia	4.2 (3.7, 4.8)	<0.0001
Dementia	4.1 (3.4, 4.8)	<0.0001
Falls	3.4 (2.6, 4.3)	<0.0001
Anemia	3.3 (2.7, 4.1)	<0.0001
Accidents	3.0 (2.2, 4.2)	<0.0001
Parkinson's disease	2.9 (2.3, 3.8)	<0.0001
Nephritis	1.4 (1.2, 1.7)	<0.0001
Injury	1.3 (1.1, 1.6)	0.0083
Other lower respiratory disease	1.3 (1.1, 1.5)	0.001
Poisoning or other external cause	0.9 (0.7, 1.2)	0.5042

Discussion

In this review of death certificates for all ALS deaths in the U.S., 2011–2014, we confirmed the most frequent causes of death identified in other studies and, due to the large number of certificates reviewed, also identified some additional nuance regarding ALS deaths. This study is unique in that it analyzed multiple cause of death data. Consistent with other studies, ³ , ⁷ , ⁸ , ²⁴ respiratory failure was the most frequent cause of death (excluding ALS), followed by cardiovascular disease and cardiac arrest. Similarly, deaths from heart failure ⁹ and cardiac arrest ²¹ , ²⁵ have been identified as common causes in other studies of ALS decedents. Pneumonia, the fourth most frequent MCOD, may be caused by weakness of respiratory muscles that leads to infection ⁹ or aspiration of gastric contents. ²⁴

Other common causes found here have not been identified in other studies but could be expected in patients with ALS. To our knowledge, pneumonitis has not been previously identified as a common cause of death among ALS decedents, but similar to pneumonia may be related to aspiration of gastric contents. ²⁴ Foreign object trapping in the respiratory tract or esophagus may be due to weakness in and lack of control of swallowing muscles.

Most causes in Table 4 were strongly associated with ALS when compared with all U.S. deaths. This is not unexpected, as the health and functional challenges of patients with ALS differ considerably from those of the general population in many aspects. However, it is important to note that ALS is largely a disease that affects older populations with similar susceptibility to chronic illnesses as the general population. Cardiovascular disease, diabetes, and cancer were major causes of death in the present analysis. Upon diagnosis, management of ALS can quickly become the focus of clinical care, with a possible lapse in the care of other chronic conditions. Physicians should be aware of other underlying conditions that can lead to a negative outcome.

This study adds knowledge regarding the circumstances leading to death in patients with ALS and suggests preventable causes of death. By highlighting such causes, greater emphasis during clinic visits could be made to decrease the likelihood of death from these causes. That cardiovascular deaths ranked number two behind respiratory failure supports the continued role of an internist or cardiologist for risk factor modification, but additional research is needed in this area to understand the impact of this care in persons with ALS.

This analysis suggests that in most individuals with ALS, deaths are linked to ALS disease progression and related symptoms. Several causes in Table 4 may be at least partly preventable, including aphagia or dysphagia, foreign object in the respiratory tract or esophagus, pneumonitis and pneumonia, nutritional deficiencies, falls, and accidents. These causes represent 14% of all MCOD in this study, occurring at a rate of about 900 deaths per year. Further, each may have preventable interventions. For example, interventions including gastrostomy tubes and laryngotracheal separation may reduce the chance of pneumonitis, complications of aphagia and dysphagia, foreign objects in the respiratory tract or esophagus, and nutritional deficiencies. These interventions may be associated with increased survival. ²⁶ , ²⁷ However, provider recommendations and patient preference will impact the decision to pursue invasive therapies.

Notably, falls may be a preventable cause of death; the age‐adjusted odds of death involving a fall were 3.4 times greater among ALS decedents (Table 4). Further, injuries tended to occur with falls and accidents (Table 3), indicating that objects and situations in patients' environments that could result in falls or accidents be mitigated. In fact, the prevention of falls in this patient population is an important component of the American Academy of Neurology Quality Care Guidelines. ²⁸ However, there may be barriers to the implementation of strategies to reduce the risk of falls. These include a patient's preference for the use of assistive devices, the home configuration to support the use of assistive devices, caregiver availability to support safe transfers and mobility, and financial and insurance barriers for the early use of more costly durable medical equipment. ²⁹

Suicide was excluded from Tables 2 and 4 due to its rarity (n = 30). Results from other studies suggest that suicide ³ , ³⁰ or psychological predisposition for suicide ³¹ may be more common among patients with ALS than in the general population. It is curious that suicide appears to be uncommon in U.S. death certificates, but this may be related to the misclassification of this cause. ³² In addition, some patients elect to withdraw from fluid, nutrition, or ventilatory support as a means of hastening death. ³³ In such circumstances, it is possible that death certificates would not include a code for suicide, thereby resulting in an undercount of this cause. Additional research on suicide among U.S. patients with ALS using a prospective design should be considered.

A large limitation of this study is the quality of MCOD entered on death certificates. Although ALS is the underlying cause initiating a series of events leading to the immediate cause of death, we found that 46% of death certificates had ALS as the sole cause, suggesting that the immediate cause of death frequently goes unreported in this population. While it is likely these deaths are the result of respiratory failure or cardiovascular disease, it remains that in a significant proportion of ALS decedents, we do not know other causes. The reason for incomplete cause of death fields is unknown, but the missing information does hamper public health efforts at identifying preventative causes of death in this patient population. We speculate that the medical certifier completing a death certificate may frequently be unaware of other causes or neglect to report them which, again, may result in an undercount of the true number of major or minor causes. To allow improved surveillance of ALS mortality, medical certifiers are encouraged to report as completely as possible all relevant factors in ALS deaths.

While the accuracy of cause of death on U.S. death certificates is thought to be good for cancer, ³⁴ it may be less reliable for other causes, particularly among older persons with multiple diseases. ³⁵ The sensitivity of death certificates in the U.S. for ALS has been estimated as 0.85–0.87. ¹¹ , ³⁶ Several prospective studies of patients with ALS have examined the cause of death determined from autopsy ⁹ , ²⁴ , ³⁷ or examination of medical records. ³ , ⁷ Those studies have a considerable advantage over the present one in that they were not subject to limitations of death certificate data. However, those studies had many findings corresponding to ours, and an advantage of our study is its large sample size.

In conclusion, using a national sample of death certificates, we found that causes of death in ALS decedents were consistent with causes reported in other studies. Due to our study's large sample size, we were able to detect additional causes not found in other studies and calculate their odds ratios using national mortality data as a comparison. Using association rules mining, a machine‐learning method, we found patterns in the data in which certain MCOD categories occurred together. Medical examiners completing death certificates for ALS decedents should ensure all causes of death are entered. The results from our study inform us about the natural history of ALS. With knowledge that some causes of death may be preventable, clinicians may be able to optimize patient care with increased awareness and training.

Conflicts of Interest

The authors report no competing interests.

Disclaimer

The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the Agency for Toxic Substances and Disease Registry.

Author Contributions

All authors contributed to the research project conception, design, and execution. TL, BD, and FB performed the statistical analysis. All authors contributedto preparation, review, and critique of the manuscript.

Supporting information

Supplemental Table 1