Comorbidity Profiles of Posttraumatic Stress Disorder Across the Medical Phenome

Emily M. Hicks, Maria Niarchou, Slavina Goleva, Dia Kabir, Julia Ciarcia, PTSD & Trauma EHR Working Group, Jordan W. Smoller, Lea K. Davis, Caroline M. Nievergelt, Karestan C. Koenen, Laura M. Huckins, Karmel W. Choi
doi: https://doi.org/10.1101/2023.08.25.23294572

Abstract

Background Prior epidemiological research has linked PTSD with specific physical health problems, but the comprehensive landscape of medical conditions associated with PTSD remains uncharacterized. Electronic health records (EHR) provide an opportunity to overcome prior clinical knowledge gaps and uncover associations with biological relevance that potentially vary by sex.

Methods PTSD was defined among biobank participants (total N=123,365) in a major healthcare system using two ICD code-based definitions: broad (1+ PTSD or acute stress codes versus 0; NCase=14,899) and narrow (2+ PTSD codes versus 0; NCase=3,026). Using a phenome-wide association (PheWAS) design, we tested associations between each PTSD definition and all prevalent disease umbrella categories, i.e., phecodes. We also conducted sex-stratified PheWAS analyses including a sex-by-diagnosis interaction term in each logistic regression.

Results A substantial number of phecodes were significantly associated with PTSDNarrow (61%) and PTSDBroad (83%). While top associations were shared between the two definitions, PTSDBroad captured 334 additional phecodes not significantly associated with PTSDNarrow and exhibited a wider range of significantly associated phecodes across various categories, including respiratory, genitourinary, and circulatory conditions. Sex differences were observed, in that PTSDBroad was more strongly associated with osteoporosis, respiratory failure, hemorrhage, and pulmonary heart disease among male patients, and with urinary tract infection, acute pharyngitis, respiratory infections, and overweight among female patients.

Conclusions This study provides valuable insights into a diverse range of comorbidities associated with PTSD, including both known and novel associations, while highlighting the influence of sex differences and the impact of defining PTSD using EHR.

Introduction

Post-traumatic stress disorder (PTSD) occurs following highly stressful and/or traumatic events. Not only does PTSD impair occupational performance, social relationships, and overall quality of life, it frequently presents with comorbidities1, defined broadly here as co-occurring health conditions. Common PTSD comorbidities include psychiatric conditions such as depression and anxiety2,3, and physical health problems including cardiovascular disease4 and gastrointestinal conditions5. However, the comprehensive landscape of PTSD comorbidities, particularly in healthcare contexts, has remained incompletely characterized6. This limits clinical insights into complex needs of patients with PTSD (i.e., what other medical concerns to consider in formulating a treatment plan) as well as empirical insights into medical conditions that may serve as risk factors for PTSD and/or share underlying biological mechanisms7.

Electronic health records (EHR) provide a unique opportunity to uncover such associations with clinical and potential biological relevance. EHR offer large samples and longitudinal data including all diagnoses that a patient receives while active in that healthcare system, recorded using standardized diagnostic billing codes that can be examined in tandem across multiple systems. Such codes can be used in different ways to identify post-traumatic psychopathology8. While PTSD is designated by specific billing codes, providers may choose to assign a broader code to reflect diagnostic uncertainty, higher likelihood of insurance coverage, stigma reduction, or other factors, meaning that some PTSD cases may be missed when simply focusing on a narrow set of codes. In addition to identifying specific codes, we can also consider how frequently they are assigned to an individual. Accordingly, a broader definition might include a single diagnosis of any stress-related disorder versus a narrow definition requiring multiple diagnoses of PTSD specifically. This has implications for epidemiologic9 and genomic studies that leverage these definitions to identify PTSD cases and controls from EHR (for example, GWAS10,11), where a broader definition can inclusively capture more cases, enhancing discovery power, but may also introduce non-specific associations12. How we identify individuals with post-traumatic psychopathology using EHR, i.e., with more inclusive versus narrow definitions, may impact our ability to resolve relevant comorbidities, and the translatability of resulting genetic findings.

In addition to a need to resolve wide-ranging PTSD comorbidities, important sex differences may also exist in such comorbidities. The burden of PTSD varies by sex3,13, with rates of lifetime PTSD about twice as high in females compared to males14. Understanding why PTSD prevalence differs by sex is complex; rather than simply capturing intrinsic biological sex differences in risk, differences may reflect gendered environments including trauma exposure (e.g., interpersonal violence)15–19 or available supports. Male, female, and intersex patients may also have different experiences within healthcare systems, including where and how care is received, access to care, and their diagnostic journey for PTSD. As such, the medical conditions that co-occur with PTSD may also differ20.

In this multi-site study, we sought to investigate the relationships between different PTSD phenotypes and the medical phenome using data from 123,365 patients enrolled in biobanks across three major EHR-based healthcare systems. Our objectives were to comprehensively identify medical comorbidities associated with PTSD, and to examine whether these associations varied based on broad versus narrow definitions, and by sex. Finally, we explored how each PTSD definition relates to polygenic signals to understand implications for downstream genomic and clinical research.

Methods

Samples

Our study includes 123,365 biobank research participants with available genetic and longitudinal EHR data21 from three major health systems—Vanderbilt University Medical Center (VUMC, n=73,488), Mount Sinai (MSSM, n=31,704), and Mass General Brigham (MGB, n=29,885). Sample characteristics (Table 1) and phenotype category codes (phecodes) frequencies (Supplementary Table 1) and study information (Supplementary Material 1) for each site are provided.

View this table:
Table 1. Sample characteristics by site.

PTSD exposures

PTSD case-control phenotypes were defined using two ICD code-based definitions. Individuals who met broad PTSD criteria (PTSDBroad) had > 1 PTSD (309.81, F43.1, F43.10, F43.11, F43.12) or stress-related disorder code (310.2, F07.81, 309, 309.1, 309.2, 309.21, 309.22, 309.23, 309.24, 309.28, 309.29, 309.3, 309.4, 309.82, 309.83, 309.89, 309.9, F43, F43.0, F43.2, F43.20, F43.21, F43.22, F43.23, F43.24, F43.25, F43.29, F43.8, F43.9), while controls had no such code; individuals who met narrow PTSD criteria (PTSDNarrow) had > 2 PTSD codes only, while controls had no codes for PTSD, other stress-related disorders, or post-concussive injury.

Main PheWAS analyses

We used a phenome-wide association (PheWAS) design to test logistic regression associations between PTSDNarrow or PTSDBroad with all phecodes as previously defined22. Cases for each outcome phecode had > 2 ICD codes belonging to that phecode category, while controls had 0 codes in that category (individuals with only one code were excluded). We excluded outcome phecodes derived from the same ICD codes used to define PTSDNarrow (phecode 300.9 (“Posttraumatic Stress Disorder”)) and PTSDBroad (phecodes 300.8 (“Acute Reaction to Stress”), 304 (“Adjustment Reaction”), and 300.9 (“Posttraumatic Stress Disorder”)) to eliminate circularity of exposure and outcome. We adjusted the models for age, genetically inferred ancestry (European [EUR], African [AA], Americas [AMR], East Asian [EAS], where available at each site), sex, and top 5 PCs to adjust for residual differences. As a sensitivity analysis, each site also tested PheWAS models in which genetic PCs were not included as a covariate. We applied a Bonferroni threshold to establish significance, correcting for 1324 tests (pBONF<3.7×10−5; PTSDNarrow) and 1367 tests (pBONF<3.8×10−5; PTSDBroad). We assessed enrichment of phecode categories of associations reaching Bonferroni significance, and the top 25% of associations, using a binomial test.

For all phecodes appearing in > 2 sites in the analysis (1649 phecodes were available for >2 sites; 765 were available for all three sites), we meta-analyzed across sites using fixed-effects in the PheWAS R package23. To assess heterogeneity, we calculated I2 statistics in METAL24 to identify associations that differed in effect sizes between health systems. Phecodes with I2<0.1 were considered significantly heterogeneous. While we observed significant heterogeneity among effect size estimates across sites for both PTSDNarrow and PTSDBroad PheWAS, directions of effect and top associations across sites remained largely uniform (Supplementary Table 2).

Sex-stratified and sex-interaction PheWAS Analyses

We conducted sex-stratified PheWAS analyses (n=13,228 female, n=11,614 male), adjusting for age, ancestry, and top 5 PCs. Patient sex was categorized as follows: MGB: sex assigned at birth, VUMC: sex assigned at birth, Mount Sinai: karyotypic sex; a small number of intersex individuals and individuals who declined to state their sex were excluded from this analysis. For all Bonferroni-significant phecodes from the full or sex-stratified analyses, we conducted an analysis including a sex-by-diagnosis interaction term in each logistic regression. Interaction effect sizes were meta-analyzed using an inverse-variance weighted model using a “STDERR” scheme with METAL software24.

Polygenic score analyses

We assessed genetic correlates of PTSD definitions in individuals of European (EUR) and African ancestries (AA) based on PTSD polygenic scores (PGS). We calculated PGS inferring SNPs weights from European PGC-PTSD GWAS summary statistics (discovery n=174,659)25 as implemented in PRS-CSx26 and summing allele counts weights. For each available ancestry group within each health system, we normalized the PGS and tested their association with PTSDNarrow or PTSDBroad using logistic regression, including sex, age, and within-ancestry genotype-derived principal components (PCs). We estimated phenotypic variance explained by PTSD scores using Nagelkerke’s R2,27. We meta-analyzed PGS associations with fixed effects using the metafor package28.

Results

Sample

To examine phenome-wide comorbidities of PTSD, we conducted analyses included 123,365 patients (3,026 and 14,899 meeting the PTSDNarrow versus PTSDBroad definitions, respectively) across three major healthcare systems: VUMC, MGB and MSSM (Table 1). Prevalence of PTSDNarrow and PTSDBroad at MGB (4.32% and 22.14%, respectively) were higher than the other two sites (VUMC: 1.9% and 9.37%, MSSM: 1.69% and 6.98%).

Narrowly and broadly defined PTSD definitions have wide-ranging correlates across the medical phenome

807 of 1324 (61%) phecodes were significantly associated with PTSDNarrow and 1136 or 1367 (83%) with PTSDBroad (pBonf<3.8×10−5, 3.7×10−5, respectively; full summary statistics: Supplementary Table 2). Top PTSDNarrow associations included psychiatric and pain-related phecodes, whereas top PTSDBroad associations included psychiatric and pain-related phecodes as well as respiratory, genitourinary, and circulatory phecodes (Figure 1). Phecode associations with PTSD were primarily positive: we observed an increased risk of each of these phecodes among individuals with PTSD. Only a very small minority of phecodes was negatively associated with PTSD, including regional enteritis (Crohn’s) (p=0.019, OR=0.77; p=0.0061, OR=0.86 for PTSDNarrow and PTSDBroad, respectively) and prostate cancer (p=0.014, OR=0.61; p=0.022, OR=0.87) (Supplementary Table 2). Notably, these negative directions of effect were consistent across all sites.

Figure 1.
Figure 1.

A) PheWAS meta-analysis results for the narrow PTSD definition. B) PheWAS meta-analysis results for the broad PTSD definition. Phecodes are organized across the y-axis by their category, with the strength of association shown in the x-axis using –log10p. Red dashed line indicates the Bonferroni significance threshold. Bonferroni (**) and nominally significant (*) phecode category enrichments for the top 25% of phecode associations for PTSDNarrow (mental disorders: p<0.0001, symptoms: p=0.0001, neurological: p=0.0019, respiratory=0.0081) and PTSDBroad (mental disorders: p<0.0001, symptoms: p<0.0001, respiratory: p=0.0007, neurological: p=0.0269, musculoskeletal: p=0.0324, endocrine/metabolic: p=0.0312)

Narrowly-defined PTSD shows strong associations with neuropsychiatric traits

PTSDNarrow was associated with over 85% of phecodes in the mental disorders category (Supplementary Table 3). Among the top 25% of phecodes associated with PTSDNarrow, ‘mental disorders’ and ‘symptoms’ categories were significantly enriched after Bonferroni correction (Supplementary Table 3). The strongest associations with PTSDNarrow fell in the mental disorders category, including personality disorders (Antisocial/borderline personality disorder, Personality disorders); mood disorders (Mood disorders, Depression); suicide-related phecodes (Suicidal ideation, Suicidal ideation or attempt); substance use disorders (Substance addiction and disorders, Alcohol-related disorders); and anxiety disorders (Agorophobia, social phobia, and panic disorder, Anxiety disorder, Generalized anxiety disorder). Beyond the mental disorders category, PTSDNarrow was associated with over 85% of phecodes in the infectious diseases category (Supplementary Table 3). Phecodes with the largest odds ratios in association with PTSDNarrow were Anxiety Disorders, Antisocial/borderline personality disorder, personality disorder, mood disorder, suicidal ideation, suicide or self-inflicted injury, poisoning by psychotropic agents, and depression.

Additionally, PTSDNarrow showed strong associations with phecodes in the neurological and symptoms categories pertaining to pain, including Chronic pain, Back pain, Pain, and Nonspecific chest pain. Several genitourinary phecodes also emerged in association with PTSDNarrow, for example, Other symptoms/disorders of the urinary system, Urinary tract infection, and Dysuria.

Broadly-defined PTSD shows strong associations with neuropsychiatric traits and respiratory, digestive, genitourinary, and cardiovascular conditions

PTSDBroad was significantly associated with all phecodes in the mental disorders and infectious disease categories and over 90% of phecodes in the respiratory, injuries & poisonings, symptoms and digestive categories (Supplementary Table 3). Among the top 25% of significant phecodes, mental disorders, symptoms, and respiratory PheCode categories were significantly enriched (Supplementary Table 3). The top phecode associations with PTSDBroad included mood disorders (depression, suicidal ideation, anxiety, bipolar disorder), substance use disorders (Substance addiction and disorders, alcohol related disorders alcoholism, tobacco use disorder), and other psychiatric disorders (schizophrenia, ADHD). The phecodes with the largest odds ratios in association with PTSDBroad were antisocial/borderline personality (OR=18.8 (18.61-19.00)), personality disorder (OR=13.21 (13.07-13.34)), suicidal ideation (OR=9.4 (9.30-9.51)), and anxiety disorders (OR=9.37 (9.33-9.41)).

In addition to mental disorders, top associations with PTSDBroad were related to a wide range of medical phecode categories, including sleep (insomnia, sleep disorders, malaise and fatigue), pain (back pain, chronic pain, abdominal pain, pain, pain in joint), respiratory system (other symptoms of respiratory system, cough, acute upper respiratory infections of multiple or unspecified sites, shortness of breath, pneumonia), digestive system (nausea and vomiting, GERD, diseases of esophagus, esophagitis), genitourinary system (urinary tract infection, other symptoms disorders of the urinary system, dysuria), cardiovascular system (nonspecific chest pain, cardiac dysrhythmias, hypertension, hypotension, arrhythmia, palpitations), and asthma conditions.

Clinical phenome associations were largely shared but showed some differences across PTSD definitions

Significant phecode associations were largely shared between PTSDNarrow and PTSDBroad definitions, especially the top-ranking associations (Figure 2). However, many more phecodes were significantly associated with PTSDBroad (1136/1367 tested) compared to PTSDNarow (807/1324), and 334 Bonferroni-significant phecodes were associated with PTSDBroad but not PTSDNarrow. The strongest such associations included phecodes related to skin (neoplasms of skin, dyschromia, actinic keratosis, other non-epithelial cancer of skin, dermatitis), lung transplant and chronic pulmonary heart disease; however, these phecodes were not among the strongest hits for PTSDBroad. Only four phecodes were significantly associated following Bonferroni correction with PTSDNarrow but not PTSDBroad. These included relatively uncommon genitourinary phecodes such as urethral hypermobility/ISD, pelvic peritoneal adhesions, as well as amblyopia, and other disorders of middle ear and mastoid.

Figure 2.
Figure 2.

Top PheWAS hits are shared between narrowly and broadly defined PTSD. Phecodes ranked by p value are plotted for PTSDBroad (x-axis) and PTSDNarrow(y-axis). Phecodes with Bonferroni significant association with PTSDNarrow (blue), PTSDBroad (red), both (purple) or neither (gray) are indicated. Number of shared and distinct phecode associations for PTSDNarrow and PTSDBroad are shown in the inset.

Sex moderates associations between broadly-defined PTSD and several other phenotypes

We conducted sex-stratified PTSDNarrow and PTSDBroad PheWAS, and meta-analyzed across sites (Supplementary figures 1-2; Supplementary Table 4), and performed a PTSD-by-sex interaction test restricted to only those phecodes with a Bonferroni-significant main effect within either sex (Figure 3). There were 16 phecodes that were significantly different between males and females coded with PTSDBroad after meta-analysis. Phenotypes that were more common among male patients with PTSDBroad included osteoporosis, respiratory failure, subarachnoid and intracranial hemorrhage, and pulmonary heart disease. Phenotypes that were more common among female patients with PTSDBroad included urinary tract infection and other urinary symptoms, acute pharyngitis, acute respiratory infections, and overweight, though directions of effect were not consistent between sites for some phecodes. For PTSDNarrow, other symptoms/disorder of the urinary system was the only phecode with a significant sex interaction, driven by higher prevalence in males at MGB despite opposing associations observed at the other two sites.

Figure 3
Figure 3

Phecodes with significant sex interaction with PTSDNarrow (A) and PTSDBroad (B) in at least one site or in meta-analysis across sites. Interaction term (PTSD * Sex) coefficients and standard errors are plotted for each site independently (MGB=blue, VUMC=green, MSSM=purple) and for a meta-analyzed interaction term coefficient (red). Bonferonni significant interaction terms are indicated with a filled circle. Positive interaction term indicates a higher prevalence in males and a negative interaction term indicates a higher prevalence in females.

Explanatory power of polygenic scores varies according to PTSD definition

We tested PGS associations with both PTSDNarrow and PTSDBroad (Table 2). PTSD PGS significantly associated with both PTSDNarrow (meta-analytic B=0.17, SE=0.02, p=5.6E-14, R2 range= 0.08-0.29%; total cases: 11,075) and PTSDBroad (meta-analytic B=0.07, SE=0.01, p=2.7E-11, R2 range=0.08-0.43%; total cases: 2,116) in EUR subsamples, although the proportion of variance explained (R2) was low. PGS were not significantly associated among AA individuals for either PTSDNarrow (B=0.06, SE=0.09, p= 0.30, R2 range=0.0002-0.04%) or PTSDBroad(B=0.01, SE=0.09, p=0.53, R2 range=0.006-0.32%), likely due to poor representation of African ancestry alleles in existing PTSD GWAS.

View this table:
Table 2. PTSD polygenic score (PGS) Meta-analysis association statistics.

Discussion

This study examined phenome-wide relationships between EHR-based PTSD and over 1300 medical phecodes among over 123,000 participants across 3 health systems, including 14,899 and 3,026 individuals meeting broad and narrow definitions for PTSD, respectively. We identified known and novel PTSD comorbidities that replicated across multiple health systems and contextualized the relative strength of PTSD-related associations across the medical phenome. Additionally, we identified several phenotypes for which males and females exhibit differing relationships with PTSD. These results characterize a wide range of PTSD-related diagnostic comorbidities in a health system setting and provide a useful catalog of phenotype-phenotype relationships for interpreting future biobank-based genetic association studies of PTSD.

Contextualizing known and novel PTSD comorbidities

Our results build upon known psychiatric29–31 and non-psychiatric PTSD comorbidities. Psychiatric disorders were consistently identified as the strongest comorbidities for both broad and narrow PTSD definitions; among non-psychiatric disorders, we observed a strong association of both PTSD definitions with pain phenotypes, which encompass physical sensory components, as well as cognitive and emotional components32–36. Given inflammatory/immune links to PTSD37, we were surprised to find that autoimmune disorders such as rheumatoid arthritis and multiple sclerosis were not significantly associated with PTSD. This suggests these comorbidities may be less common on a health system scale, or that these comorbidities present with specific PTSD subtypes (for example, those with a specific immune component or etiopathology) that are not well represented or captured in our biobanks. Unexpectedly, we did not find significant associations between PTSD and decreased libido38. However, this phecode is relatively scarce in our biobanks (N=162-214), decreasing our ability to detect associations. We identified novel associations of genitourinary phenotypes with PTSD, with a particularly strong association with PTSDBroad, with significance levels comparable to respiratory and cardiovascular comorbidities. It is possible that these associations stem from shared risk factors, i.e., that sexual assault or related traumatic experiences lead to both genitourinary and psychiatric outcomes39, or are artefactual associations resulting from cross-disorder correlations, where conditions that appear significant are not directly associated with the exposure but rather linked to other diseases. It is also possible these associations reflect “healthcare pleiotropy” – i.e., the association stems from an artifact of the clinical settings where traumatic exposures are more commonly/routinely discussed (i.e., OB/GYN clinics).

Sex differences in PTSD comorbidities

Epidemiological evidence indicates that females exposed to traumatic events are at higher risk for developing PTSD compared to trauma-exposed males. We identified several phenotypes for which associations with PTSDBroad differed between males and females. For females, obesity had a stronger association with PTSD than males. This relationship has been previously reported for some studies of PTSD40,41 and other psychiatric phenotypes42. We also found stronger association of urinary symptoms with PTSD in females. This may be due to genito-urinary symptoms that occur at higher rates following exposure to abuse38; for example, interstitial cystitis/bladder pain syndrome43, vulvodynia44, and chronic pelvic pain45,46, coupled with a greater prevalence of sexual trauma among females17,47 or perhaps due to the clinical setting where trauma and genitourinary symptoms are assessed. Respiratory failure, cranial hemorrhage, and pulmonary heart disease were more prevalent among male PTSDBroad cases. Together these results may reflect differential patterns of diagnosis of these disorders or PTSD, differences in the care received between sexes (for example, in medical fields with sex specificity like gynecology) or sex differences in PTSD pathology and the downstream system-wide consequences of having PTSD.

Scope of PTSD phenotyping and implications for genetic studies

This study explored the role of phenotypic specificity in resolving comorbid phenotypes. A narrow definition requiring multiple PTSD codes likely captures individuals with more severe and specific symptomatology. In contrast, the broad definition, consisting of a single code inclusive of stress-related reactions, may include individuals with a broader range of symptom severity. Psychiatric genetic studies increasingly tend to include broader definitions which facilitate greater sample sizes, but at the potential cost of specificity12. Both PTSD definitions showed strikingly high levels of comorbidities. However, we found top PTSDNarrow associations were concentrated in mental disorders and pain-related symptoms, whereas PTSDBroad associations were spread across more phecode categories such as respiratory, genitourinary and circulatory conditions. Generally, the top phecodes were consistent across both definitions, with the broader definition associated with 300 additional phenotypes.

These association patterns have significant implications for our understanding of broad vs. narrow definitions of psychiatric disorders and may shape future analytical and inclusion/exclusion criteria, as well as how we should interpret associations specific to PTSDBroad. The most significant of these were skin conditions (dyschromia, actinic keratosis etc.), which may reflect racial differences in PTSD prevalence, trauma exposure, or diagnostic rates15. To assess relevance for future studies, we consider each definition. If PTSDBroad captures both full PTSD and sub-threshold PTSD cases (or, if an accurate diagnosis is obtained using only one ICD code, PTSDBroad may capture primarily full PTSD cases), then we might conceptualize these associations as true PTSD associations, discoverable only thanks to this increase in power. However, if the broad category involves decreased diagnostic accuracy, PTSDBroad may simply add ‘noise’ to the analysis, rather than identifying sub-threshold or otherwise ‘missed’ cases, and these additional associations may not contribute meaningfully to our understanding of PTSD and its comorbidities. At the genetic level, we observed that a PTSD PGS (trained on a GWAS based on mainly studies of clinically ascertained PTSD25) significantly associated with both PTSDBroad and PTSDNarrow status, though more strongly PTSDNarrow than PTSDBroad. In the future, as sample sizes for PTSD GWAS increase, focusing on more strictly defined cases may become more feasible and may lead to increased predictive power of PGS.

Several factors suggest that the former, rather than latter, conceptualization is correct. First, addition of truly random cases (i.e., introduction of ‘noise’) should reduce PheWAS discovery power48 rather than identify additional significant associations. Second, both PTSDBroad and PTSDNarrow have significant psychiatric associations; it does not appear that one group is obviously more ‘accurately’ ascertained that the other. Third, we note a primary difference between the two groups reflects frequency of diagnosis; a single ICD code for PTSD does not necessarily imply an incorrect diagnosis, sub-clinical presentation, or less severe symptoms (this could be examined through in-depth chart reviews in the future); rather, these patients may leave the medical system or be referred elsewhere for treatment. In particular, we note that MSSM is associated with a VA, where veterans experiencing combat-related PTSD may pursue PTSD care instead of the initial diagnostic site.

Strengths and limitations

Our study has several limitations to note when assessing and interpreting our results. First, while studying the full clinical phenome using EHR is a strength of our study, our biobank-based patient populations may not be fully representative of the general population. To some extent, we mitigate this concern by comparing effect sizes and meta-analyzing48 across three healthcare settings that represent distinct geographical areas (Boston, New York, Nashville), with differing demographics and healthcare needs (e.g., the proportion of veterans seeking care at these clinics differs significantly), and differing proportions of individuals with PTSD. In particular, one site (MGB) has a much higher rate of PTSD diagnosis (PTSDNarrow: 4.3% vs 1.9%, 1.7% at VUMC, MSSM, respectively). Consequently, we observe some heterogeneity in effect size of associations; however, effect directions are largely consistent across all three sites, which is highly unlikely to be due to chance (655 concordant across three sites vs. 169 discordant: binomial p=4.84×10−236; for codes present in two sites only: 340 concordant vs. 126 discordant, binomial p=9.06×10−24).

Given the majority of individuals are seeking care in a hospital setting for some symptom or condition, EHR-based analyses likely underestimate the proportion of healthy individuals and will not identify sub-threshold PTSD or those with well-managed PTSD receiving care in a different system. Our analysis also does not adjust for utilization rates. Relatedly, individuals in our study may have uneven access to healthcare or to certain specialists within a healthcare system: psychiatric or mental health services in particular may be unevenly accessed due to differences in insurance coverage or stigma or immigration status. Even when healthcare is accessed, patients may receive differing levels of care and stereotyping bias or institutionalized racism in healthcare may lead to under-or over-diagnosis48–50. These issues may be particularly exacerbated in psychiatry, where certain presentations of PTSD may be more readily diagnosed, potentially biasing our comorbidity results towards that subtype of PTSD.

Another limitation is the lack of temporal resolution. Understanding whether these associated phecodes occur prior to, or following, PTSD onset would ideally allow us to delineate associations into antecedents versus potential consequences of PTSD diagnosis. However, this is difficult in practice: patients may report longstanding existing diagnoses to a clinician upon enrollment at a new clinic, or receive correct diagnoses after long diagnostic journeys. In the latter case, a patient may experience initial symptoms several years before an accurate diagnosis of PTSD, such that other diagnoses received may either be comorbid with PTSD or represent inaccurate diagnoses given to explain some of the PTSD symptoms (for example, MDD, or adjustment reaction disorder). Difficulty in pinpointing disease onset precludes analysis of temporal relationships. Informed by this comprehensive map of associations, studies should be repeated in longitudinal cohorts with detailed and repeated psychiatric questionnaires that provide nuanced temporal resolution and can help shed light on the temporal nature of these phenotype-phenotype relationships in PTSD. Further exploration of the genetic factors contributing to PTSD and its comorbidities is also warranted.

Conclusions and future directions

This multi-site study investigated relationships between PTSD phenotypes and medical comorbidities using data from over 123,000 patients across three major EHR-based healthcare systems. Our findings provide insights into the diagnostic complexity of PTSD and its associations with the clinical phenome. By considering both broad and narrow PTSD definitions, we comprehensively identified known and novel diagnostic comorbidities, particularly in the domains of mental disorders, pain, respiratory, genitourinary, and cardiovascular conditions. These associations were observed across multiple health systems, highlighting their generalizability. Importantly, we observed sex differences in the associations between PTSD and certain medical phenotypes, underscoring the need to consider sex-specific factors. Furthermore, our genetic analyses revealed significant associations between polygenic scores derived from GWAS and both broad and narrow definitions of PTSD. Future research should include longitudinal analyses to examine temporal relationships between PTSD and medical comorbidities, providing insights into the directionality of these associations. Investigating sex-specific factors, such as hormonal influences, trauma response differences, or care-seeking patterns, will enhance our understanding of sex disparities in PTSD and associated medical conditions, while mechanistic studies could explore underlying biological pathways and shared genetic factors contributing to the observed associations. Overall, this study deepens our understanding of the diagnostic complexity of PTSD within EHRs and provides insights for research towards integrated treatment approaches that address both PTSD diagnosis and its comorbidities.

Acknowledgements

This collaborative work was supported in part by the PsycheMERGE grant MH118233 (to JWS and LKD) and administrative supplement to PGC PTSD R01 (R01MH106595). KWC was supported in part by funding from the National Institute of Mental Health (K08MH127413) and a NARSAD Brain and Behavior Foundation Young Investigator Award. EMH, JC and LMH acknowledge funding from NIMH (R01MH124839, R01MH118278) and NIEHS (R01ES033630). LKD was supported by NIMH (R01MH118223); NHGRI and the Office of Research on Women’s Health (RM1HG009034).

MSSM: This work was supported in part through the computational resources and staff expertise provided by Scientific Computing at the Icahn School of Medicine at Mount Sinai under award number S10OD018522 from the Office of Research Infrastructure of the National Institutes of Health. This work was supported through the resources and staff expertise provided by the Charles Bronfman Institute for Personalized Medicine and the BioMe Biobank Program at the Icahn School of Medicine at Mount Sinai.

BioVU: The dataset(s) used for the analyses described were obtained from Vanderbilt University Medical Center’s BioVU which is supported by numerous sources: institutional funding, private agencies, and federal grants. These include the NIH funded Shared Instrumentation Grant S10RR025141; and CTSA grants UL1TR002243, UL1TR000445, and UL1RR024975. Genomic data are also supported by investigator-led projects that include U01HG004798, R01NS032830, RC2GM092618, P50GM115305, U01HG006378, U19HL065962, R01HD074711; and additional funding sources listed at https://victr.vumc.org/biovu-funding/. This publication was supported by Grant No. R01 HG011405, partially funded by the Office of Research on Women’s Health, Office of the Director, NIH and the National Human Genome Research Institute. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Office of Research on Women’s Health or the National Human Genome Research Institute.

Disclosures: JWS is a member of the Scientific Advisory Board of Sensorium Therapeutics (with equity), and has received grant support from Biogen, Inc. He is PI of a collaborative study of the genetics of depression and bipolar disorder sponsored by 23andMe for which 23andMe provides analysis time as in-kind support but no payments.

Supplementary Tables

  1. Rank of top phecodes by site

  2. Full summary statistics for PTSD PheWAS

  3. PTSD PheWAS hits by PheCode category and Cross-category enrichment of PheCode categories for PTSD PheWAS hits

  4. Full summary statistics for sex-stratified PTSD PheWAS

Supplementary Figures

  1. Male and female stratified PheWAS plot for PTSD Narrow

  2. Male and female stratified PheWAS plot for PTSD Broad.

Data Availability

All data produced in the present work are contained in the manuscript and supplemental materials.

References

  1. 1.↵
    Sareen, J. et al. Physical and mental comorbidity, disability, and suicidal behavior associated with posttraumatic stress disorder in a large community sample. Psychosom. Med. 69, 242–248 (2007).
    OpenUrlAbstract/FREE Full Text
  2. 2.↵
    Flory, J. D. & Yehuda, R. Comorbidity between post-traumatic stress disorder and major depressive disorder: alternative explanations and treatment considerations. Dialogues Clin. Neurosci. 17, 141–150 (2015).
    OpenUrlPubMed
  3. 3.↵
    Qassem, T., Aly-ElGabry, D., Alzarouni, A., Abdel-Aziz, K. & Arnone, D. Psychiatric Co-Morbidities in Post-Traumatic Stress Disorder: Detailed Findings from the Adult Psychiatric Morbidity Survey in the English Population. Psychiatr. Q. 92, 321–330 (2021).
    OpenUrl
  4. 4.↵
    Pietrzak, R. H., Goldstein, R. B., Southwick, S. M. & Grant, B. F. Physical Health Conditions Associated with Posttraumatic Stress Disorder in U.S. Older Adults: Results from Wave 2 of the National Epidemiologic Survey on Alcohol and Related Conditions. J. Am. Geriatr. Soc. 60, 296–303 (2012).
    OpenUrlCrossRefPubMed
  5. 5.↵
    Pacella, M. L., Hruska, B. & Delahanty, D. L. The physical health consequences of PTSD and PTSD symptoms: a meta-analytic review. J. Anxiety Disord. 27, 33–46 (2013).
    OpenUrlCrossRefPubMedWeb of Science
  6. 6.↵
    Qureshi, S. U., Pyne, J. M., Magruder, K. M., Schulz, P. E. & Kunik, M. E. The link between post-traumatic stress disorder and physical comorbidities: a systematic review. Psychiatr. Q. 80, 87–97 (2009).
    OpenUrlCrossRefPubMedWeb of Science
  7. 7.↵
    Johnston, K. J. A. & Huckins, L. M. Chronic Pain and Psychiatric Conditions. Complex Psychiatry 24–43 (2023) doi:10.1159/000527041.
    OpenUrlCrossRef
  8. 8.↵
    Harrington, K. M. et al. Validation of an Electronic Medical Record-Based Algorithm for Identifying Posttraumatic Stress Disorder in U.S. Veterans. J. Trauma. Stress 32, 226–237 (2019).
    OpenUrlCrossRef
  9. 9.↵
    Livingston, N. A. et al. Identifying Posttraumatic Stress Disorder and Disparity Among Transgender Veterans Using Nationwide Veterans Health Administration Electronic Health Record Data. LGBT Health 9, 94–102 (2022).
    OpenUrl
  10. 10.↵
    Smoller, J. W. The use of electronic health records for psychiatric phenotyping and genomics. Am. J. Med. Genet. B Neuropsychiatr. Genet. 177, 601–612 (2018).
    OpenUrlPubMed
  11. 11.↵
    Stein, M. B. et al. Genome-wide association analyses of post-traumatic stress disorder and its symptom subdomains in the Million Veteran Program. Nat. Genet. 53, 174–184 (2021).
    OpenUrlCrossRef
  12. 12.↵
    Cai, N. et al. Minimal phenotyping yields genome-wide association signals of low specificity for major depression. Nat. Genet. 52, 437–447 (2020).
    OpenUrlCrossRef
  13. 13.↵
    Duncan, L. E. et al. Largest GWAS of PTSD (N=20 070) yields genetic overlap with schizophrenia and sex differences in heritability. Mol. Psychiatry 23, 666–673 (2018).
    OpenUrlCrossRefPubMed
  14. 14.↵
    Olff, M. Sex and gender differences in post-traumatic stress disorder: an update. Eur. J. Psychotraumatology 8, 1351204 (2017).
    OpenUrl
  15. 15.↵
    Marchese, S. & Huckins, L. M. Trauma Matters: Integrating Genetic and Environmental Components of PTSD. Adv. Genet. 2200017 (2022) doi:10.1002/ggn2.202200017.
    OpenUrlCrossRef
  16. 16.
    Kim, G. S. & Uddin, M. Sex-specific and shared expression profiles of vulnerability and resilience to trauma in brain and blood. Biol. Sex Differ. 11, 13 (2020).
    OpenUrl
  17. 17.↵
    Elliott, D. M., Mok, D. S. & Briere, J. Adult sexual assault: Prevalence, symptomatology, and sex differences in the general population. J. Trauma. Stress 17, 203–211 (2004).
    OpenUrlCrossRefPubMedWeb of Science
  18. 18.
    Kucharska, J. Sex differences in the appraisal of traumatic events and psychopathology. Psychol. Trauma Theory Res. Pract. Policy 9, 575–582 (2017).
    OpenUrl
  19. 19.↵
    Mundy, S. S., Foss, S. L. W., Poulsen, S., Hjorthøj, C. & Carlsson, J. Sex differences in trauma exposure and symptomatology in trauma-affected refugees. Psychiatry Res. 293, 113445 (2020).
    OpenUrl
  20. 20.↵
    Muniz Carvalho, C. et al. Disentangling sex differences in the shared genetic architecture of posttraumatic stress disorder, traumatic experiences, and social support with body size and composition. Neurobiol. Stress 15, 100400 (2021).
    OpenUrl
  21. 21.↵
    Kohane, I. S. et al. What Every Reader Should Know About Studies Using Electronic Health Record Data but May Be Afraid to Ask. J. Med. Internet Res. 23, e22219 (2021).
    OpenUrlPubMed
  22. 22.↵
    Wu, P. et al. Mapping ICD-10 and ICD-10-CM Codes to Phecodes: Workflow Development and Initial Evaluation. JMIR Med. Inform. 7, e14325 (2019).
    OpenUrlCrossRef
  23. 23.↵
    Carroll, R. J., Bastarache, L. & Denny, J. C. R PheWAS: data analysis and plotting tools for phenome-wide association studies in the R environment. Bioinformatics 30, 2375–2376 (2014).
    OpenUrlCrossRefPubMedWeb of Science
  24. 24.↵
    Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191 (2010).
    OpenUrlCrossRefPubMedWeb of Science
  25. 25.↵
    Nievergelt, C. M. et al. International meta-analysis of PTSD genome-wide association studies identifies sex-and ancestry-specific genetic risk loci. Nat. Commun. 10, 4558 (2019).
    OpenUrlCrossRefPubMed
  26. 26.↵
    Ruan, Y. et al. Improving polygenic prediction in ancestrally diverse populations. Nat. Genet. 54, 573–580 (2022).
    OpenUrl
  27. 27.↵
    Nagelkerke, N. J. D. A note on a general definition of the coefficient of determination. Biometrika 78, 691–692 (1991).
    OpenUrlCrossRefWeb of Science
  28. 28.↵
    Viechtbauer, W. Conducting Meta-Analyses in R with the metafor Package. J. Stat. Softw. 36, (2010).
  29. 29.↵
    Kang, B., Xu, H. & McConnell, E. S. Neurocognitive and psychiatric comorbidities of posttraumatic stress disorder among older veterans: A systematic review. Int. J. Geriatr. Psychiatry 34, 522–538 (2019).
    OpenUrl
  30. 30.
    Facer-Irwin, E., Karatzias, T., Bird, A., Blackwood, N. & MacManus, D. PTSD and complex PTSD in sentenced male prisoners in the UK: prevalence, trauma antecedents, and psychiatric comorbidities. Psychol. Med. 52, 2794–2804 (2022).
    OpenUrl
  31. 31.↵
    Pietrzak, R. H., Goldstein, R. B., Southwick, S. M. & Grant, B. F. Psychiatric Comorbidity of Full and Partial Posttraumatic Stress Disorder Among Older Adults in the United States: Results From Wave 2 of the National Epidemiologic Survey on Alcohol and Related Conditions. Am. J. Geriatr. Psychiatry 20, 380–390 (2012).
    OpenUrlCrossRefPubMedWeb of Science
  32. 32.↵
    Asmundson, G. J., Coons, M. J., Taylor, S. & Katz, J. PTSD and the Experience of Pain: Research and Clinical Implications of Shared Vulnerability and Mutual Maintenance Models. Can. J. Psychiatry 47, 930–937 (2002).
    OpenUrlCrossRefPubMedWeb of Science
  33. 33.
    Sharp, T. J. & Harvey, A. G. Chronic pain and posttraumatic stress disorder: mutual maintenance? Clin. Psychol. Rev. 21, 857–877 (2001).
    OpenUrlCrossRefPubMedWeb of Science
  34. 34.
    Brennstuhl, M.-J., Tarquinio, C. & Montel, S. Chronic Pain and PTSD: Evolving Views on Their Comorbidity: Chronic Pain and PTSD: Evolving Views on Their Comorbidity. Perspect. Psychiatr. Care 51, 295–304 (2015).
    OpenUrlPubMed
  35. 35.
    Giordano, N. A. et al. Differential Pain Presentations Observed across Post-Traumatic Stress Disorder Symptom Trajectories after Combat Injury. Pain Med. 22, 2638–2647 (2021).
    OpenUrl
  36. 36.↵
    Gasperi, M., Afari, N., Goldberg, J., Suri, P. & Panizzon, M. S. Pain and Trauma: The Role of Criterion A Trauma and Stressful Life Events in the Pain and PTSD Relationship. J. Pain 22, 1506–1517 (2021).
    OpenUrl
  37. 37.↵
    Katrinli, S., Oliveira, N. C. S., Felger, J. C., Michopoulos, V. & Smith, A. K. The role of the immune system in posttraumatic stress disorder. Transl. Psychiatry 12, 313 (2022).
    OpenUrl
  38. 38.↵
    Yehuda, R., Lehrner, A. m. y. & Rosenbaum, T. Y. PTSD and Sexual Dysfunction in Men and Women. J. Sex. Med. 12, 1107–1119 (2015).
    OpenUrl
  39. 39.↵
    Vézina-Gagnon, P. et al. Childhood sexual abuse, girls’ genitourinary diseases, and psychiatric comorbidity: A matched-cohort study. Health Psychol. 40, 104–112 (2021).
    OpenUrl
  40. 40.↵
    Perkonigg, A., Owashi, T., Stein, M. B., Kirschbaum, C. & Wittchen, H.-U. Posttraumatic Stress Disorder and Obesity. Am. J. Prev. Med. 36, 1–8 (2009).
    OpenUrlCrossRefPubMed
  41. 41.↵
    Tronieri, J. S., Wurst, C. M., Pearl, R. L. & Allison, K. C. Sex Differences in Obesity and Mental Health. Curr. Psychiatry Rep. 19, 29 (2017).
    OpenUrl
  42. 42.↵
    Lopresti, A. L. & Drummond, P. D. Obesity and psychiatric disorders: Commonalities in dysregulated biological pathways and their implications for treatment. Prog. Neuropsychopharmacol. Biol. Psychiatry 45, 92–99 (2013).
    OpenUrlCrossRefPubMed
  43. 43.↵
    McKernan, L. C. et al. Posttraumatic stress disorder in interstitial cystitis/bladder pain syndrome: Relationship to patient phenotype and clinical practice implications. Neurourol. Urodyn. 38, 353–362 (2019).
    OpenUrl
  44. 44.↵
    Iglesias-Rios, L., Harlow, S. D. & Reed, B. D. Depression and Posttraumatic Stress Disorder Among Women with Vulvodynia: Evidence from the Population-Based Woman to Woman Health Study. J. Womens Health 24, 557–562 (2015).
    OpenUrl
  45. 45.↵
    Lamvu, G., Carrillo, J., Ouyang, C. & Rapkin, A. Chronic Pelvic Pain in Women: A Review. JAMA 325, 2381 (2021).
    OpenUrlCrossRefPubMed
  46. 46.↵
    Meltzer-Brody, S. et al. Trauma and Posttraumatic Stress Disorder in Women With Chronic Pelvic Pain. Obstet. Gynecol. 109, 902–908 (2007).
    OpenUrlCrossRefPubMedWeb of Science
  47. 47.↵
    Tolin, D. F. & Foa, E. B. Sex differences in trauma and posttraumatic stress disorder: A quantitative review of 25 years of research. Psychol. Bull. 132, 959–992 (2006).
    OpenUrlCrossRefPubMedWeb of Science
  48. 48.↵
    Dueñas, H. R., Seah, C., Johnson, J. S. & Huckins, L. M. Implicit bias of encoded variables: frameworks for addressing structured bias in EHR–GWAS data. Hum. Mol. Genet. 29, R33–R41 (2020).
    OpenUrl
  49. 49.
    Huckins, L. M. Thoughtful Phenotype Definitions Empower Participants and Power Studies. Complex Psychiatry 8, 57–62 (2022).
    OpenUrl
  50. 50.↵
    Huckins, L. M. et al. What next for eating disorder genetics? Replacing myths with facts to sharpen our understanding. Mol. Psychiatry 27, 3929–3938 (2022).
    OpenUrl
Back to top
Posted August 28, 2023.
Download PDF

Supplementary Material

Data/Code
Email
Comorbidity Profiles of Posttraumatic Stress Disorder Across the Medical Phenome
Emily M. Hicks, Maria Niarchou, Slavina Goleva, Dia Kabir, Julia Ciarcia, PTSD & Trauma EHR Working Group, Jordan W. Smoller, Lea K. Davis, Caroline M. Nievergelt, Karestan C. Koenen, Laura M. Huckins, Karmel W. Choi
medRxiv 2023.08.25.23294572; doi: https://doi.org/10.1101/2023.08.25.23294572
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area