Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Apr 1.
Published in final edited form as: Health Commun. 2020 Dec 1;37(4):467–475. doi: 10.1080/10410236.2020.1851862

Examining the Examiners: How Medical Death Investigators Describe Suicidal, Homicidal, and Accidental Death

Adam S Miner a,b,1,*, David M Markowitz c,1, Brian L Peterson d, Benjamin W Weston e
PMCID: PMC8939265  NIHMSID: NIHMS1787724  PMID: 33950764

Abstract

This study describes differences in medicolegal death investigators’ written descriptions for people who died by homicide, suicide, or accident. We evaluated 17 years of death descriptions from a midsized metropolitan midwestern county in the United States to assess how death investigators psychologically respond to different manners of death (N = 10,408 cases). Automated text analyses suggest investigators describe accidental deaths with more immediacy relative to homicides, while they also described suicidal deaths in less emotional terms than homicides as well. These data suggest medicolegal death investigators have different psychological reactions to circumstances and manners of death as indicated by their professional writing. Future research may surface context-specific psychological reactions to vicarious trauma that could inform the design or personalization of workplace coping interventions.

Keywords: medical examiners, medicolegal death investigation, suicide, homicide, accident, cause of death, natural language processing, health communication

Official death investigations have important personal, medical, and legal ramifications. In concert with autopsies and other sources of information, medicolegal death investigations are used to officially certify a decedent’s manner of death (e.g. natural, accident, suicide or homicide) (Hanzlick, 2006; Hanzlick et al., 2002; NHS National Quality Board, 2017). Although conducting death investigations is psychologically stressful (Brondolo et al., 2018; Timmermans, 2005), little is known about the relationship between death investigator psychological functioning and the decedent’s cause of death.

Establishing the circumstances of a person’s death is important for families’ emotional closure, legal proceedings, and reporting public health concerns (National Institute of Justice, 2019; NHS National Quality Board, 2017; Roberts et al., 2017). Medical death investigations typically occur when someone dies in unusual circumstances (Hanzlick, 2006). An official medicolegal death investigation is one aspect of certifying a death. Other important steps include autopsies, which are typically conducted by forensic pathologists. Although reporting methods vary regionally, medical death investigations in Milwaukee County, Wisconsin, USA (2010 census population = 947,735) are conducted under the authority of a county-based Medical Examiner’s office. They are performed by Medicolegal Death Investigators who have a variety of professional backgrounds (e.g., forensic nurses, police officers). The medicolegal death investigation includes uncovering circumstances of death by gathering social and medical information, discovering activities before death, scene investigation, contacting family members, and documenting official findings. Sometimes, investigators will aid in the external physical inspection of the decedent, however the autopsy is performed by a forensic pathologist. One aspect of official documentation is a free-text description of the circumstances of death which is written by an individual death investigator. This official documentation informs medical, legal, insurance-related, and public health decisions.

The psychological functioning of medical death investigation is important to understand because investigators’ perception and subsequent description of death has personal (e.g., internalized negative cognitions) and societal impact (e.g., for families and the creation of official records). Because medicolegal death investigation is uniquely stressful, psychological coping is important to personal and professional duties (Brondolo et al., 2018; Farrow et al., 2009; National Institute of Justice, 2019). Interacting with violent and recent deaths is associated with negative psychological reactions and secondary trauma (aka vicarious trauma; Chung et al., 2000; McCann & Pearlman, 1990). In a sample of medical examiners, increased negative cognitions were linked to symptoms of depression and posttraumatic stress disorder (Brondolo et al., 2018). Additionally, rates of secondary trauma are higher when medical workers regularly interact with interpersonally violent cases (Passmore et al., 2020; Raunick et al., 2015).

Despite the personal and social importance of medical death investigation, we are unaware of prior work that has assessed whether psychological experiences of medical death investigators are expressed through official written reports. Understanding the psychological experiences of death investigators via language may facilitate early detection of maladaptive coping, which may inform the design and deployment of prophylactic coping strategies, such as supervisory social support (Dickinson & Wright, 2008; Molnar et al., 2017). In this paper, we use language patterns from death investigator reports to infer how such professionals internalize different death types. This work is timely and important as medical death investigators are expected to be objective in their reporting, and therefore, the psychological impact of their work is often overlooked or masked by the profession. We take this opportunity to examine how words reflect the social and psychological experience of reporting on death in a large database of medical death investigator writing.

The psychology of language: content and style

Inferring the internal states of communicators from word patterns is supported by prior work that uses language as a lens into psychological processes (Tausczik & Pennebaker, 2010). For example, language patterns can reveal social status (Kacewicz et al., 2014; Markowitz, 2018), personality dimensions (Ireland & Mehl, 2014), and interpersonal attraction (Ireland et al., 2011). Work in this tradition often considers two classes of words to reveal psychological dynamics: content words and function words (Pennebaker, 2011). Content words (e.g., nouns, verbs) describe what people are talking about, whereas function words (e.g., pronouns, articles, prepositions) describe how a communicator is speaking, or their style. Content words can be counted to form themes within a piece of text or identify key topics, and style words are often counted to reveal important attentional aspects of a writer’s psychology at a particular moment in time (Chung & Pennebaker, 2007).

We focus the current evaluation primarily on function words, which are often context-independent and linked to psychological processes such as depression, anxiety, and distress (Markowitz & Hancock, 2017; Stirman & Pennebaker, 2001). Function words may be uniquely effective in surfacing psychological trends in medical documentation, as medicolegal death investigation data is rife with inconsistent medical terminology (Lathrop et al., 2009). Pronouns, for example, indicate a person’s attention. High-status individuals tend to use more “we” words (e.g., we, our, us) than low-status people, as those of high rank are often required to care for more people than the self (Kacewicz et al., 2014).

Pronouns are particularly useful to indicate psychological distress and trauma. For example, after the terrorist attack of September 11th, bloggers wrote with a more psychologically distant style than before the attacks (Cohn et al., 2004). To manage traumatic events, people often experience a reduction in self-references (e.g., I, me, my); distancing the self from trauma is one way to cope with the experience. In work exploring the relationship between language and self-harm, linguistic markers of psychological distress have been observed prior to an eventual suicide. Suicidal poets used a more immediate writing style relative to poets who had accidental deaths (Stirman & Pennebaker, 2001), and the lyrics of suicidal musicians were more emotional than non-suicidal musicians (Markowitz & Hancock, 2017).

Together, the prior evidence suggests that psychological experiences can be inferred through word patterns, and that style words offer an important lens into people experiencing trauma. A different dynamic — one that is paramount to our interests — explores how those who are reporting on trauma and death, but not experiencing it themselves, use words as a reflection of their psychological experience and their occupation as a medical death investigator. Deaths by homicide, suicide and accidents may be uniquely stressful and may engender different psychological responses (Cerel et al., 2016; MacNeil, 2008).

Language dimensions: Prior research

We draw on prior evidence to evaluate how medical death investigators write about death as a reflection of the victim’s death type (e.g., homicide, suicide, accident). Four language dimensions were evaluated based on prior work: first-person singular pronouns (self-references), psychological distancing, analytic thinking, and emotional writing.

First, we evaluate medical death investigators’ attention and focus through first-person singular pronouns. Communicators who use high rates of first-person singular (e.g. I, me, my) tend to focus on the self instead of others or their social world (Pennebaker, 2011).

Second, we assess an index of psychological distancing (Pennebaker & King, 1999). Recall, self-references and more immediate writing styles reveal an introspective disposition and more psychological attention on a personal experience, whereas a distant and less self-reflective style pushes an event away from the individual psychologically.

Third, we evaluate the medical death investigators’ thinking style as indicated by a function word index of analytic thinking. Prior work by Pennebaker and colleagues (2014) observed that students who wrote college admissions essays with more articles (e.g., a, the) and prepositions (e.g., above, below) relative to storytelling words (e.g., pronouns, adverbs) had higher grades at the end of college (Pennebaker et al., 2014). This analytic thinking index reveals if a person’s thinking style is formal, concrete, and categorical (high rates of analytic thinking) relative to simple, abstract, and reflecting a story (low rates of analytic thinking). Although prior work suggests that suicidal musicians tend to write lyrics with a narrative and less analytic style compared to non-suicidal musicians (Markowitz & Hancock, 2017), it is unknown how medical death investigators write about decedents. Therefore, we explore the rate of analytic thinking to appraise how medical death investigators construe death — either formally and factually, or as a story.

Fourth, we evaluate rates of emotion in medical death investigator writing by assessing words with positive or negative affect. Emotionality is especially relevant in this population, as death investigators experience emotionally demanding work, for example when investigating the murder of a child (Roach et al., 2017). Although death investigation is psychologically impactful on the medical professional, the relationship between secondary trauma (i.e., indirect exposure to trauma) and psychological writing style reflecting one’s experience as a medical investigation professional has not been well described.

Taken together, the purpose of this study is to assess the association between cause of death and language styles used to describe those deaths by professionals. We believe this is an important investigation because it is currently unclear how people who write about traumas and death for a living, such as medical examiners, might face psychological costs as a result of their profession. We seek to understand how medical examiners internalize their work psychologically through language patterns. We draw on prior evidence to assess four language dimensions (self-references, psychological distancing, analytic thinking, emotion) across medical death investigator disclosures about different death types. We do not make formal predictions about how these dimensions will transpire across the death types, since this research enterprise is the first of its kind with this unique population.

RQ: How do medical death investigators of homicide, suicide, and accident write as a reflection of their psychological experience?

Method

This study is a retrospective analysis of written medical death investigation reports. The data were generated during the course of normal government death investigations. Text synopses of investigated deaths from a single county in the United States and a host of victim metadata (e.g., race, age, sex) was obtained through request and approval from the Milwaukee County Medical Examiner’s office. This study was deemed exempt from Institutional Review Board review at Stanford University.

Database information

We obtained the electronic records database from the Milwaukee County Medical Examiner (Milwaukee County Medical Examiner, 2020). Individual death related data are publicly available upon reasonable request from the Office of the Medical Examiner.

This study assessed county records in Milwaukee County, Wisconsin from year 2000–2017 (no data were available in 2001). A total of 12,025 cases were retrieved and descriptions with less than 15 words were removed to prevent small word counts from skewing the results (final N = 10,408 cases in the dataset) (Markowitz, 2019b). This criterion resulted in a focal dataset covering 283,444 words. Demographic information (e.g., age, race) and cause of death (i.e., suicide, homicide, accident) were included in the data, though some reports did not contain victim-level demographics.

Automated text analysis

The text synopsis from each medical death investigation was quantified with Linguistic Inquiry and Word Count (LIWC: Pennebaker et al., 2015). This tool transforms text into numbers by identifying words in a piece of text that correspond with LIWC’s internal dictionary of social (e.g., words related to people such as brother, sister, woman), psychological (e.g., words related to emotion such as hate, happy, or disgust), and health-related categories (e.g., words related to the body, such as skin or blood). LIWC quantifies words as a percent of the total word count. For example, the phrase “The victim was shot” contains four words and several dimensions that are incremented by the LIWC dictionary, including but not limited to: articles (the; 25% of the total word count), adverbs (was; 25% of the total word count), and negative emotion terms (victim; 25% of the total word count).

We used the standard LIWC2015 dictionary in all analyses unless otherwise noted. On average, medical death investigators wrote 27.23 words per death description (SD = 10.51).

Preprocessing

Two changes were made to the text to ensure that the effects reflected psychological processes and not medical death investigation conventions. First, the abbreviation for medical examiners (ME) would be captured by LIWC as a self-reference. “ME” was replaced with “M.E.” to prevent this convention from affecting the results. Second, medical death investigators often describe where an event occurred (e.g., a crash occurred on I-95). This convention would also be incorrectly quantified by LIWC as a self-reference. Any phrase containing “I-” was replaced with the word “Interstate.”

Language dimensions

First-person singular pronouns

Rate of self-references (e.g., I, me, my) indicates a writer’s focus and attention (Pennebaker, 2011). Text from our database that contains a high rate of self-references includes “Deceased was shot in the chest when he answered his door. I spoke with B of I Technician, [name], on [date] at 0950 hours; she informed me deceased has been positively identified by his right index print.”

Psychological distancing

The rate of psychological distancing was calculated by adding the standardized scores of articles (e.g., the, a) and words greater than six letters, but subtracting standardized scores of self-references (e.g., I, me), present tense verbs (e.g., try, wait, do), and words that represent “discrepancies from reality” (e.g., would, could, should) (Cohn et al., 2004, p. 689). High scores on this standardized composite suggest that the text is more psychologically distant (less immediate) than text that scores low on this index (less psychologically distant, more immediate). An example synopsis of an accident death with high rates of psychological distancing includes: “Deceased was involved in an accident; he was driving a motorcycle and he struck a pickup truck.”

Analytic thinking

To evaluate the thinking style of medical death investigators while they wrote the synopses, several function word categories were combined in an index (Pennebaker et al., 2014). We therefore evaluate how style words relate to each cause of death and this dimension is calculated on a scale of 0 (low analytic thinking) – 100 (high analytic thinking).

The language features comprising the analytic thinking index include high rates of articles (e.g., a, the) and prepositions (e.g., above, below), but low rates of pronouns (e.g., he, she, they), adverbs (e.g., almost, indeed, just), auxiliary verbs (e.g., been, done, must), negations (e.g., no, not, nor), and conjunctions (e.g., and, but, while). Articles and prepositions make concrete references to verbal categories, while pronouns and other storytelling words are often observed in narratives. An example of a low analytic thinking writing style from our analysis is illustrative: “This 31-year-old male was drinking heavily and arguing with his girlfriend when he grabbed his military gear and weapons, stated he was going to kill himself, and then inflicted a witnessed gunshot wound to his head.”

Emotion

Finally, we evaluated the general rate of emotion, plus individual valence measures of positive and negative affect, in each text description. A suicide synopsis with high rates of emotion includes, “This 62-year-old male who was depressed, frustrated, and in pain, left a suicide note and was found with an empty bottle of Morphine and Gabapentin nearby.”

Analytic approach

We used separate general linear models to evaluate how language patterns differed across death causes. Since we observed that demographic variables such as victim race (categorical), gender (categorical), and age (continuous) were not evenly distributed across death causes (Table 1), we controlled for these variables as fixed effects in all models. We also controlled for victim death year (categorical) as a fixed effect to account for any writing style changes that might occur over time. Treating death year as a continuous variable produced substantively equivalent results.

Table 1.

Database descriptive information.

Victim information Suicide (n = 1,394) Homicide (n = 1,374) Accident (n = 7,640)
M(SE) M(SE) M(SE) F p η2p
Age (years) 44.43 (0.62) 31.53 (0.62) 58.97 (0.26) 952.77 < .001 0.155
Gender n n n χ2(4) p φ
Male 1,053 1,149 4,385 449.58 < .001 0.208
Female 341 224 3,249
Unknown 0 1 6
Race n n n χ2(14) p φ
Black or African American 187 1,005 1,597 1,979.84 < .001 0.436
Asian 27 17 62
Eastern Indian 5 11 9
Hispanic 75 114 405
Mixed race 6 12 30
Native American 6 6 39
White 1,087 204 5,475
Unknown 1 5 23
Open in a new tab

The data were analyzed in R (version 4.0.0). To evaluate the robustness of our effects, we provided bootstrapped p-values and 95% Confidence Intervals, based on 1,000 resamples, which are reported in a data table of our results. Crucially, all effects were maintained with no substantive changes in significance levels.

Results

A correlation matrix of our key variables is located in Table 2. A summary of our findings is located in Table 3.

Table 2.

Correlation matrix of language variables.

Self-references Psychological distancing Analytic thinking Affect Positive affect
Self-references --
Psychological distancing −.464** --
Analytic thinking −.015 .380** --
Affect .003 −.114** −.028** --
Positive affect .010 −.101** .030** .441** --
Negative affect −.002 −.071** −.046** .871** −.055**
Open in a new tab

Note.

*

p < .05,

**

p < .01.

N = 10,408.

Table 3.

General linear models connecting language patterns to death type.

Group 1 Group 2 M diff SE t p p boot 95% CIboot R 2
Self-references Accident Homicide 0.01 0.01 0.01 .989 .988 [−0.016, 0.014] 0.01
Suicide Homicide 0.01 0.01 0.78 .434 .484 [−0.013, 0.025]
Suicide Accident 0.01 0.01 1.05 .295 .362 [−0.007, 0.021]
Psychological distancing Accident Homicide −0.28 0.07 −3.90 < .001 .001 [−0.412, −0.129] 0.02
Suicide Homicide −0.07 0.09 −0.75 .454 .447 [−0.224, 0.115]
Suicide Accident 0.21 0.06 3.30 .001 .002 [0.075, 0.346]
Analytic thinking Accident Homicide −1.36 0.47 −2.90 .004 .005 [−2.217, −0.463] 0.02
Suicide Homicide 0.13 0.57 0.23 .821 .785 [−0.915, 1.154]
Suicide Accident 1.49 0.42 3.52 < .001 .001 [0.666, 2.29]
Affect Accident Homicide −0.02 0.09 −0.18 .854 .890 [−0.213, 0.177] 0.01
Suicide Homicide −0.41 0.11 −3.71 < .001 .002 [−0.64, −0.202]
Suicide Accident −0.40 0.08 −4.84 < .001 .001 [−0.546, −0.236]
Positive affect Accident Homicide 0.04 0.04 1.01 .314 .209 [−0.026, 0.111] 0.06
Suicide Homicide −0.12 0.05 −2.26 .024 .002 [−0.198, −0.047]
Suicide Accident −0.16 0.04 −4.19 < .001 .001 [−0.217, −0.108]
Negative affect Accident Homicide −0.07 0.08 −0.81 .421 .453 [−0.247, 0.096] 0.01
Suicide Homicide −0.31 0.10 −3.05 .002 .004 [−0.508, −0.12]
Suicide Accident −0.24 0.07 −3.26 .001 .001 [−0.385, −0.107]
Open in a new tab

Note. Models control for year of death (categorial), victim race (categorial), victim gender (categorial), and victim age (continuous) as fixed effects. Mdiff. = Mean difference between Group 1 and Group 2. pboot = recomputed p-values based on 1,000 resamples. 95% CIboot = percentile-based 95% Confidence Intervals for the mean difference.

Self-references

There was no significant relationship between rates of self-references and death types (ps > .295).

Psychological distancing

Medical death investigators describe accident deaths with less psychological distancing (more immediacy) than homicide deaths, after controlling for victim-level variables (p < .001). Further, medical death investigators describe suicides with more psychological distancing than accidents (p = .001). Suicides and homicides were described with similar rates of psychological distancing (p = .454).

Analytic thinking

Homicide deaths were described with higher rates of analytic thinking than accident deaths, after controlling for victim-level variables (p = .004). Suicide deaths were described with greater rates of analytic thinking than accidents (p < .001). Suicides and homicides were described with similar rates of analytic thinking (p = .821).

Affect

Suicide deaths were described with lower rates of emotion than homicides and accidents, after controlling for victim-level variables (ps < .001). Accidents and homicides tend to be described with similar rates of emotion (p = .854).

Positive affect

Consistent with the overall emotion effect, medical death investigators describe suicides less positively than homicides and accidents, after controlling for victim-level variables (ps < .024). Accidents and homicides tend to be described with similar rates of positive affect (p = .314).

Negative affect

Medical death investigators describe suicide deaths with reduced levels of negative affect than homicide and accident deaths, after controlling for victim-level variables (ps < .002). Accidents and homicides tend to be described with similar rates of negative affect (p = .421).

Taken together, medical death investigators — whose professional goal is to objectively report on death — psychologically manage and respond to death causes in different ways, as reflected by language patterns. We observed that medical examiners who reported on suicide and accidents showed the largest degree of difference in verbal style compared to other combinations of death types. That is, examiners reporting on suicide contained a less emotional, more analytic, and more psychologically distant communication style than accidents. Traumas are often difficult for people to manage for themselves psychologically and prior work suggests this distress can appear in word patterns (Cohn et al., 2004). Here, we observed that psychological distress can impact how medical examiners write, even if they are indirectly related to an event (e.g., they were not involved with a death). Therefore, medical examiners use psychological strategies to appraise suicides differently than accidents and such effects are revealed through language taken from an ecologically valid sample. This suggests that words might provide an important lens to understand the psychological effects of occupational distress on medical examiners and others in high-stakes or high-risk professions.

Content patterns

Using the Meaning Extraction Helper (Boyd, 2018), we also assessed the top 25 content words within each death type to evaluate if medical examiners wrote about different topics as a reflection of their death determination (Table 4)1. Indeed, examiners who wrote about suicide discussed methods of this death type (e.g., hang, self-inflict) more than those in the other death types. Gun shots were prevalent in suicides and homicides, while accidents described falls, fractures, and people with a history of mishaps. Together, these data — in conjunction with the other language results — support the idea that medical examiners write with a different style and content patterns when describing various death types.

Table 4.

Top 25 Content Words Across Death Types

Suicide (n = 1,394) Homicide (n = 1,374) Accident (n = 7,640)
Word Raw n Cases with word % Word Raw n Cases with word % Word Raw n Cases with word %
year-old 1,044 1,040 74.61 male 815 808 58.81 year-old 5,315 5,297 69.33
male 834 834 59.83 year-old 797 794 57.79 male 3,238 3,232 42.30
head 431 429 30.77 shot 589 573 41.70 female 2,476 2,470 32.33
deceased 461 382 27.40 wound 339 334 24.31 deceased 2,554 2,260 29.58
hang 314 313 22.45 die 304 300 21.83 fall 2,342 2,183 28.57
female 278 278 19.94 deceased 344 297 21.62 die 2,055 2,038 26.68
wound 266 262 18.79 year 275 274 19.94 fracture 1,995 1,863 24.38
[Examiner] 230 230 16.50 gunshot 271 270 19.65 history 1,777 1,693 22.16
gunshot 226 223 16.00 [Examiner] 233 233 16.96 year 1,391 1,375 18.00
convey 226 218 15.64 head 229 228 16.59 unresponsive 1,323 1,315 17.21
self-inflict 218 218 15.64 multiple 221 220 16.01 [Examiner] 1,215 1,215 15.90
suicide 229 216 15.49 sustain 170 170 12.37 hip 1,081 1,038 13.59
autopsy 215 213 15.28 female 170 168 12.23 sustain 981 975 12.76
year 208 206 14.78 convey 181 167 12.15 hospital 880 848 11.10
history 203 198 14.20 [Hospital] 149 148 10.77 convey 756 720 9.42
apparent 189 188 13.49 hospital 146 142 10.33 hospice 738 715 9.36
gsw* 185 184 13.20 gsw* 142 139 10.12 drug 756 706 9.24
note 163 158 11.33 autopsy 137 137 9.97 abuse 716 701 9.18
depress 153 152 10.90 chest 132 132 9.61 autopsy 685 674 8.82
shot 129 119 8.54 [Examiner] 124 124 9.02 care 683 650 8.51
gun 122 117 8.39 time 121 118 8.59 decedent 680 640 8.38
dead 120 117 8.39 decedent 107 105 7.64 st 645 626 8.19
die 115 113 8.11 street 109 99 7.21 bed 614 596 7.80
facility 112 112 8.03 stab 106 98 7.13 death 615 567 7.42
basement 109 109 7.82 victim 106 97 7.06 vehicle 607 497 6.51
Open in a new tab

Note. Percentages refer to the percent of cases within each death type that contain a particular word.

*

gsw = gunshot wound

Discussion

To our knowledge, this is the first quantitative assessment of the psychological relationship between medical death investigators’ official reports and decedent cause of death. Differences in writing style were observed between descriptions of death by homicide, suicide and accident. Medical death investigators generally write about suicide and homicide in an analytic manner (e.g., using more articles and prepositions relative to pronouns and other storytelling words) versus those who die by accident. Accidents are also described with more emotion than suicide. Therefore, language may serve as a window into the psychology of medical death investigators at the time they are reporting on death causes.

It is important to highlight the text analysis approach in this study relative to other approaches that typically assess well-being in medical settings. We used function words and emotion terms to reveal patterns in medical death investigators’ thinking (Pennebaker et al., 2014). Content word patterns also revealed that what medical examiners wrote about reflected the cause of death for each case, further supporting the idea that verbal patterns betray psychological aspects of their profession. This text analysis method is unique, especially for health communication and medical fields, which often rely on self-reported surveys and interviewing techniques to infer how people are thinking and feeling under distress in their job (Kroska et al., 2017; Wolfe et al., 2018). We argue that counting words from actual work products to understand social and psychological processes for medical personnel provides an important opportunity to evaluate, in the moment, how people might be appraising or internalizing distress.

We make two important theoretical contributions in the current paper. First, this work bridges a gap between health communication scholarship on self-focused versus other-focused trauma narratives. Prior research has established a link between personal writing style and personal trauma (Markowitz & Hancock, 2017; Stirman & Pennebaker, 2001). Also, prior work has established a link between health professionals’ vicarious trauma and exposure to workplace trauma (Molnar et al., 2017). However, prior work has not established a relationship between personal writing style and regular interactions with other people’s personal trauma (i.e., secondary trauma). We therefore find these results to be important for informing how medical and health professionals appraise trauma in the scope of their work, using words to unobtrusively infer the psychological states of communicators.

Second, this research addresses a gap in workplace-based burnout and depression research. Although prior work has established a link between cognitive and emotional distress and death investigation (Farrow et al., 2009; Roach et al., 2017), the methods of assessment are survey-based, and thus separated from both the workplace environment and internal processes of the respondent. Our approach, assessing the actual work product of death investigators, provides a contextual assessment of latent psychological processes which may provide a more honest or genuine reflection of psychological processes. Thus, our findings suggest a previously undescribed connection between personal psychological processes and workplace-based secondary trauma, using quantitative, ecologically valid methods. These findings answer a call from trauma researchers and governmental agencies to improve methods for identifying mechanisms of action in work settings prone to vicarious trauma (Molnar et al., 2017; National Institute of Justice, 2019).

Relatedly, health communication research has often explored the impact of expressive writing paradigms on health outcomes. Expressive writing tasks have people write about trauma or a painful experience over several days, with evidence suggesting the writing helps to improve psychological and health functioning (Pennebaker, 1997). The words within these essays also help to reveal the extent and type of trauma people experienced (Pennebaker, 2011). People who write about painful first-hand experiences tend to feel better over time and their verbal output reveals their thoughts and feelings as a result of the trauma. This pattern occurs with average citizens when writing about themselves, but we extend the expressive writing model to medical professionals and show that writing about an indirect traumatic experience, or a secondary trauma, can serve expressive writing functions and reveal psychological coping and distress management processes as well.

Limitations

This study has several limitations worth resolving in future research. First, data were collected from a single county, and significant variation in trauma reporting exists both between US counties and between countries (MacNeil, 2008; Mann et al., 2006; Weinberg et al., 2013). The United States, for example, has over 2,000 medical examiner and coroner offices, managing approximately 500,000 cases a year (Weinberg et al., 2013). Although there are some standardized approaches to death investigation across the US, prior research has highlighted the importance of understanding local practice variation (Arias et al., 2016; Slavova et al., 2015). Our findings may not generalize to other counties or countries, but this initial examination provides a springboard for other counties or research teams to adopt similar approaches and evaluate how their personnel respond to trauma.

Second, although our approach to text analysis is established and validated from social science disciplines (Tausczik & Pennebaker, 2010), the dictionary-based method (i.e., using collections of predefined words) neglects contextual embeddings. For example, the phrase “I am not happy” would be miscalculated by LIWC, incrementing categories such as negations (e.g., not) and positive emotion terms (e.g., happy). Additionally, context appropriate terms such as ‘victim’ may have different emotional valence in the medicolegal context than other written contexts. Since the present analyses relied on theory and prior empirical evidence to ground the selection of language dimensions, plus the number of cases is quite large, we believe context-related issues are minimal but noteworthy for transparency.

Third, no information was gathered about individual medical death investigators, which suggests there may be unaccounted for individual effects based on medical investigator identity. For example, years of experience may account for an unknown amount of writing style variation. These data were unavailable to the authors, but future work might assess individual level variation. Finally, our effect sizes are relatively small but consistent with other evaluations of natural language (Kramer et al., 2014; Markowitz, 2019a). Experimental research should complement this observational study to evaluate the directionality of the patterns and effects.

There are potential alternative explanations to these findings that merit further research as well. Deaths by homicide, suicide, and accident have intrinsic differences which may be related to differences in death investigator descriptions. For example, suicide is self-directed violence, whereas homicides necessarily involve multiple actors and different levels of agency. Although we controlled for death year and victim race, age, and gender in our analyses, more nuanced analysis may uncover group memberships of both the investigator and the decedent which influence written descriptions of death. Larger and more diverse datasets from different counties across the US are warranted in future work, especially in light of existing disparities in medical access and treatment based on individual and social factors such as socioeconomic status, race, gender, or age (Manuel, 2018).

Future directions

These findings are aligned with explicit goals from national quality improvement initiatives. A report from the National Health Services National Quality Board argues that systematic approaches are needed to derive and share learnings from death investigations, with the express goal of improving medical death investigation training (NHS National Quality Board, 2017). The psychological experiences of death investigators are important to the investigators themselves, the families they interact with, and the counties in which they serve. There are two proximal areas of future research that appear important and feasible: 1) early detection of changes in death investigator psychological coping, and 2) further assessment of geographic variability in descriptions of death.

Informing future studies of burnout or mental health

An explicit need to better understand and improve a resilient forensic workforce is a key finding from a recent US Department of Justice report to Congress (National Institute of Justice, 2019). Using written descriptions of death by investigators may provide a novel, ecologically valid assessment of psychological processing which may inform the cognitive and emotional processes medical death investigators experience. In future work, these insights may inform tailored content development and strategies for forensic workplace wellness programs, which aim to mitigate burnout risk and employee turnover (National Institute of Justice, 2019; Panagioti et al., 2017; Raunick et al., 2015). This may be especially important for suicide-related deaths because families and healthcare professionals report stigmatizing personal or social perceptions that those effected failed in some way to prevent the death (Cvinar, 2005; Turecki & Brent, 2016). Workplace-based vicarious trauma interventions are already underway, such as the US Department of Justice based Office for Victims of Crime Vicarious Trauma Toolkit (Department of Justice, 2019). Another area of inquiry may assess whether people speak differently as they gain experience on the job, or whether death investigators speak differently about death than people not directly involved in death investigation.

Geographic variability and population predictions

Future research should assess death investigator language across additional counties that vary on community makeup (e.g., socioeconomic status, race, gender; Manuel, 2018). Insights would allow for greater precision when designing or implementing programs to aid in death worker job-related coping, and may surface geographically-bound cultural differences that affect death reporting and county-level resource allocation (e.g., funding for substance use programs and law enforcement; Fletcher et al., 2018; Fontanella et al., 2018, 2019; Hurstak et al., 2018). Additionally, understanding relevant cultural variability is an important aspect of developing predictive models that may rely on human-generated data. In Denmark for example, suicide risk is predicted at a population level using machine learning techniques (Gradus et al., 2019; McCoy et al., 2016). Deeper understanding of predictive models and their use in health-related decision making has been called for (Shah et al., 2019). Medical death investigations may be a uniquely helpful epidemiologic source of information because written descriptions of death or dying in hospital records are often inadequate, unclear, and vague (Lathrop et al., 2009; Wentlandt et al., 2018).

Conclusion

Understanding how and why people die is important for families, healthcare systems, governments, and law enforcement (National Institute of Justice, 2019; NHS National Quality Board, 2017; Timmermans, 2005). Despite the importance of death investigation, little is known about the psychological experiences of death investigators themselves, whose profession dictates they write about psychologically taxing content such as death causes and manners. We found that medical death examiners describe manners of death differently in key psychological ways. Extending this work may improve our understanding of psychological stress inherent to death investigation and inform early interventions for workplace stress.

Acknowledgments

We thank Dr. Kristin Sainani for feedback on project conceptualization and analytic plan. We thank Taylor & Francis editing services.

Funding details

This work was supported by a National Institutes of Health, National Center for Advancing Translational Science, Clinical and Translational Science Award (KL2TR001083 and UL1TR001085), and the Stanford HAI Seed Grant Program. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Footnotes

Declaration of interest statement

Brian L. Peterson has a competing interest as an employee of the Milwaukee Country Medical Examiner Office.

1

Note, the Meaning Extraction Helper lemmatizes words while counting them. Lemmatization is the process of transforming words to their “basic form.” For example, the words drive, driving, and drove are converted to drive by the Meaning Extraction Helper to identify higher order word categories (Boyd, 2018).

Data availability statement

Due to restrictions on use, data for this study is not being made publicly available. Data is available upon reasonable request from the Milwaukee Country Medical Examiner Office.

References

  1. Arias E, Heron M, National Center for Health Statistics, Hakes J, & US Census Bureau. (2016). The validity of race and Hispanic-origin reporting on death certificates in the united states: An update. Vital and Health Statistics. Series 2, Data Evaluation and Methods Research, (172), 1–21. [PubMed] [Google Scholar]
  2. Boyd RL (2018). Meaning Extraction Helper. Accessed September 17, 2020. https://www.ryanboyd.io/software/meh/
  3. Brondolo E, Eftekharzadeh P, Clifton C, Schwartz JE, & Delahanty D (2018). Work-related trauma, alienation, and posttraumatic and depressive symptoms in medical examiner employees. Psychological Trauma, 10(6), 689–697. 10.1037/tra0000323 [DOI] [PubMed] [Google Scholar]
  4. Cerel J, Maple M, van de Venne J, Moore M, Flaherty C, & Brown M (2016). Exposure to suicide in the community: Prevalence and correlates in one U.S. State. Public Health Reports, 131(1), 100–107. 10.1177/003335491613100116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Chung C, & Pennebaker JW (2007). The psychological functions of function words. In Fielder K (Ed.), Social Communication (p. 12). Psychology Press. [Google Scholar]
  6. Chung C, Chung C, & Easthope YM (2000). Traumatic stress and death anxiety among community residents exposed to an aircraft crash. Death Studies, 24(8), 689–704. 10.1080/074811800750036578 [DOI] [PubMed] [Google Scholar]
  7. Cohn MA, Mehl MR, & Pennebaker JW (2004). Linguistic markers of psychological change surrounding September 11, 2001. Psychological Science, 15(10), 687–693. 10.1111/j.0956-7976.2004.00741.x [DOI] [PubMed] [Google Scholar]
  8. Cvinar JG (2005). Do suicide survivors suffer social stigma: A review of the literature. Perspectives in Psychiatric Care, 41(1), 14–21. 10.1111/j.0031-5990.2005.00004.x [DOI] [PubMed] [Google Scholar]
  9. Department of Justice. (2019, December). Vicarious trauma toolkit. Accessed September 17, 2020. https://vtt.ovc.ojp.gov/about-the-toolkit
  10. Dickinson T, & Wright KM (2008). Stress and burnout in forensic mental health nursing: A literature review. The British Journal of Nursing, 17(2), 82–87. 10.12968/bjon.2008.17.2.28133 [DOI] [PubMed] [Google Scholar]
  11. Farrow R, Arensman E, Corcoran P, Williamson E, & Perry IJ (2009). Irish coroners’ attitudes towards suicide and its prevention. Irish Journal of Medical Science, 178(1), 61–67. 10.1007/s11845-008-0261-9 [DOI] [PubMed] [Google Scholar]
  12. Fletcher A, Coster J, & Goodacre S (2018). Impact of the new medical examiner role on patient safety. BMJ, 363, k5166. 10.1136/bmj.k5166 [DOI] [PubMed] [Google Scholar]
  13. Fontanella CA, Saman DM, Campo JV, Hiance-Steelesmith DL, Bridge JA, Sweeney HA, & Root ED (2018). Mapping suicide mortality in Ohio: A spatial epidemiological analysis of suicide clusters and area level correlates. Preventive Medicine, 106, 177–184. 10.1016/j.ypmed.2017.10.033 [DOI] [PubMed] [Google Scholar]
  14. Fontanella CA, Warner LA, Steelesmith DL, Bridge JA, Brock GN, & Campo JV (2019). A national comparison of suicide among Medicaid and non-Medicaid youth. American Journal of Preventive Medicine, 56(3), 447–451. 10.1016/j.amepre.2018.10.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Gradus JL, Rosellini AJ, Horváth-Puhó E, Street AE, Galatzer-Levy I, Jiang T, Lash TL, & Sørensen HT (2019). Prediction of sex-specific suicide risk using machine learning and single-payer health care registry data from Denmark. JAMA Psychiatry, 77(1), 25–34. 10.1001/jamapsychiatry.2019.2905 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hanzlick R (2006). Death investigation: Systems and procedures. CRC Press. [Google Scholar]
  17. Hanzlick R, Hunsaker JC, & Davis GJ (2002). A guide for manner of death classification. National Association of Medical Examiners. [Google Scholar]
  18. Hurstak E, Rowe C, Turner C, Behar E, Cabugao R, Lemos NP, Burke C, & Coffin P (2018). Using medical examiner case narratives to improve opioid overdose surveillance. The International Journal on Drug Policy, 54, 35–42. 10.1016/j.drugpo.2017.12.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ireland ME, & Mehl MR (2014). Natural language use as a marker of personality. The Oxford Handbook of Language and Social Psychology, 549, 201–218. 10.1093/oxfordhb/9780199838639.013.034 [DOI] [Google Scholar]
  20. Ireland ME, Slatcher RB, Eastwick PW, Scissors LE, Finkel EJ, & Pennebaker JW (2011). Language style matching predicts relationship initiation and stability. Psychological Science, 22(1), 39–44. 10.1177/0956797610392928 [DOI] [PubMed] [Google Scholar]
  21. Kacewicz E, Pennebaker JW, Davis M, Jeon M, & Graesser AC (2014). Pronoun use reflects standings in social hierarchies. Journal of Language and Social Psychology, 33(2), 125–143. [DOI] [Google Scholar]
  22. Kramer ADI, Guillory JE, & Hancock JT (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences of the United States of America, 111(24), 8788–8790. 10.1073/pnas.1320040111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kroska EB, Calarge C, O’Hara MW, Deumic E, & Dindo L (2017). Burnout and depression in medical students: Relations with avoidance and disengagement. Journal of Contextual Behavioral Science, 6(4), 404–408. 10.1016/j.jcbs.2017.08.003 [DOI] [Google Scholar]
  24. Lathrop SL, Davis WL, & Nolte KB (2009). Medical terminology coding systems and medicolegal death investigation data: Searching for a standardized method of electronic coding at a statewide medical examiner’s office. Journal of Forensic Sciences, 54(1), 207–211. 10.1111/j.1556-4029.2008.00903.x [DOI] [PubMed] [Google Scholar]
  25. MacNeil MS (2008). An epidemiologic study of aboriginal adolescent risk in Canada: The meaning of suicide. Journal of Child and Adolescent Psychiatric Nursing, 21(1), 3–12. 10.1111/j.1744-6171.2008.00117.x [DOI] [PubMed] [Google Scholar]
  26. Mann NC, Guice K, Cassidy L, Wright D, & Koury J (2006). Are statewide trauma registries comparable? Reaching for a national trauma dataset. Academic Emergency Medicine, 13(9), 946–953. 10.1197/j.aem.2006.04.019 [DOI] [PubMed] [Google Scholar]
  27. Manuel JI (2018). Racial/ethnic and gender disparities in health care use and access. Health Services Research, 53(3), 1407–1429. 10.1111/1475-6773.12705 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Markowitz DM (2018). Academy awards speeches reflect social status, cinematic roles, and winning expectations. Journal of Language and Social Psychology, 37(3), 376–387. [DOI] [Google Scholar]
  29. Markowitz DM (2019a). What words are worth: National Science Foundation grant abstracts indicate award funding. Journal of Language and Social Psychology, 38(3), 264–282. [DOI] [Google Scholar]
  30. Markowitz DM (2019b). Putting your best pet forward: Language patterns of persuasion in online pet advertisements. Journal of Applied Social Psychology, 50(3), 222. 10.1111/jasp.12647 [DOI] [Google Scholar]
  31. Markowitz DM, & Hancock JT (2017). The 27 Club: Music lyrics reflect psychological distress. Communication Reports, 30(1), 1–13. 10.1080/08934215.2016.1210663 [DOI] [Google Scholar]
  32. McCann IL, & Pearlman LA (1990). Vicarious traumatization: A framework for understanding the psychological effects of working with victims. Journal of Traumatic Stress, 3(1), 131–149. 10.1002/jts.2490030110 [DOI] [Google Scholar]
  33. McCoy TH Jr, Castro VM, Roberson AM, Snapper LA, & Perlis RH (2016). Improving prediction of suicide and accidental death after discharge from general hospitals with natural language processing. JAMA Psychiatry, 73(10), 1064–1071. 10.1001/jamapsychiatry.2016.2172 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Milwaukee County Medical Examiner. (2020). Accessed September 17, 2020. https://county.milwaukee.gov/EN/Medical-Examiner
  35. Molnar BE, Sprang G, Killian KD, Gottfried R, Emery V, & Bride BE (2017). Advancing science and practice for vicarious traumatization/secondary traumatic stress: A research agenda. Traumatology, 23(2), 129–142. 10.1037/trm0000122 [DOI] [Google Scholar]
  36. National Institute of Justice. (2019, December 19). Report to Congress: Needs Assessment of Forensic Laboratories and Medical Examiner/Coroner Offices | National Institute of Justice. National Institute of Justice. Accessed September 17, 2020. https://nij.ojp.gov/library/publications/report-congress-needs-assessment-forensic-laboratories-and-medical [Google Scholar]
  37. NHS National Quality Board. (2017, March 15). national guidance on learning from deaths. National Guidance on Learning from Deaths. Accessed September 17, 2020. https://www.england.nhs.uk/publication/national-guidance-on-learning-from-deaths/ [Google Scholar]
  38. Panagioti M, Panagopoulou E, Bower P, Lewith G, Kontopantelis E, Chew-Graham C, Dawson S, van Marwijk H, Geraghty K, & Esmail A (2017). Controlled interventions to reduce burnout in physicians: A systematic review and meta-analysis. JAMA Internal Medicine, 177(2), 195–205. 10.1001/jamainternmed.2016.7674 [DOI] [PubMed] [Google Scholar]
  39. Passmore S, Hemming E, McIntosh HC, & Hellman CM (2020). The relationship between hope, meaning in work, secondary traumatic stress, and burnout among child abuse pediatric clinicians. The Permanente Journal, 24, 19.087. 10.7812/TPP/19.087 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Pennebaker JW (1997). Opening Up: The Healing Power of Expressing Emotions. Guilford Press; New York. [Google Scholar]
  41. Pennebaker JW (2011). The secret life of pronouns: What our words say about us (Vol. 211). Bloomsbury Publishing. [Google Scholar]
  42. Pennebaker JW, Chung CK (2011). Expressive writing: Connections to physical and mental health. The Oxford handbook of health psychology. 916:417–437. [Google Scholar]
  43. Pennebaker JW, & King LA (1999). Linguistic styles: Language use as an individual difference. Journal of Personality and Social Psychology, 77(6), 1296–1312. 10.1037/0022-3514.77.6.1296 [DOI] [PubMed] [Google Scholar]
  44. Pennebaker JW, Boyd RL, Jordan K, & Blackburn K (2015). The development and psychometric properties of LIWC2015. [Google Scholar]
  45. Pennebaker JW, Chung CK, Frazee J, Lavergne GM, & Beaver DI (2014). When small words foretell academic success: The case of college admissions essays. PLOS ONE, 9(12), e115844. 10.1371/journal.pone.0115844 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Raunick CB, Lindell DF, Morris DL, & Backman T (2015). Vicarious trauma among sexual assault nurse examiners. Journal of Forensic Nursing, 11(3), 123–128; quiz E1. 10.1097/jfn.0000000000000085 [DOI] [PubMed] [Google Scholar]
  47. Roach J, Cartwright A, & Sharratt K (2017). Dealing with the unthinkable: A study of the cognitive and emotional stress of adult and child homicide investigations on police investigators. Journal of Police and Criminal Psychology, 32(3), 251–262. 10.1007/s11896-016-9218-5 [DOI] [Google Scholar]
  48. Roberts AP, Morrow G, Walkley M, Flavell L, Phillips T, Sykes E, Kirkpatrick G, Monkhouse D, Laws D, & Gray C (2017). From research to practice: results of 7300 mortality retrospective case record reviews in four acute hospitals in the North-East of England. BMJ Open Quality, 6(2), e000123. 10.1136/bmjoq-2017-000123 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Shah NH, Milstein A, & Bagley PhD SC (2019). Making machine learning models clinically useful. JAMA. 10.1001/jama.2019.10306 [DOI] [PubMed] [Google Scholar]
  50. Slavova S, O’Brien DB, Creppage K, Dao D, Fondario A, Haile E, Hume B, Largo TW, Nguyen C, Sabel JC, Wright D, & Council of State and Territorial Epidemiologists Overdose Subcommittee. (2015). Drug overdose deaths: Let’s get specific. Public Health Reports, 130(4), 339–342. 10.1177/003335491513000411 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Stirman SW, & Pennebaker JW (2001). Word use in the poetry of suicidal and nonsuicidal poets. Psychosomatic Medicine, 63(4), 517–522. 10.1097/00006842-200107000-00001 [DOI] [PubMed] [Google Scholar]
  52. Tausczik YR, & Pennebaker JW (2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24–54. 10.1177/0261927X09351676 [DOI] [Google Scholar]
  53. Timmermans S (2005). Death brokering: Constructing culturally appropriate deaths. Sociology of Health & Illness, 27(7), 993–1013. 10.1111/j.1467-9566.2005.00467.x [DOI] [PubMed] [Google Scholar]
  54. Turecki G, & Brent DA (2016). Suicide and suicidal behaviour. The Lancet, 387(10024), 1227–1239. 10.1016/s0140-6736(15)00234-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Weinberg M, Weedn VW, Weinberg S, & Fowler D (2013). Characteristics of medical examiner/coroner offices accredited by the National Association of Medical Examiners. Journal of Forensic Sciences, 58(5), 1193–1199. 10.1111/1556-4029.12165 [DOI] [PubMed] [Google Scholar]
  56. Wentlandt K, Toupin P, Novosedlik N, Le LW, Zimmermann C, & Kaya E (2018). Language used by health care professionals to describe dying at an acute care hospital. Journal of Pain and Symptom Management, 56(3), 337–343. 10.1016/j.jpainsymman.2018.05.013 [DOI] [PubMed] [Google Scholar]
  57. Wolfe AW, Blithe SJ, & Mohr B (2018). Dirty workers’ management of hidden emotions: Selling intimacy and seeking social support through the shroud of secrecy. The Journal of Communication, 68(1), 194–217. 10.1093/joc/jqx002 [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Due to restrictions on use, data for this study is not being made publicly available. Data is available upon reasonable request from the Milwaukee Country Medical Examiner Office.

RESOURCES