Research Spotlights: September 2022

Risk of “long COVID” and connection with psychological distress prior to SARS-CoV-2 infection

Symptoms of COVID-19 that extend beyond four weeks from onset of SARS-CoV-2 infection are often referred to as “long COVID” and have been identified as an emerging and important health concern. Long COVID is characterized by a host of symptoms including fatigue; shortness of breath or difficulty breathing; persistent cough; muscle, joint, or chest pain; smell or taste problems; confusion, disorientation, and brain fog; memory issues; depression, anxiety, and changes in mood; headache; and heart palpitation. Relatively little is known about the psychological factors associated with increased risk of long COVID symptoms. Recently published research supported by the NICHD, NHLBI, NIEHS, NCI, Veteran’s Association, and others examined whether pre-infection psychological distress, early in the pandemic, including depression, anxiety, loneliness, perceived stress, and worry were associated with an increased likelihood of developing post-COVID-19 conditions among individuals who were infected with SARS-CoV-2 after the baseline measures of psychological distress.

From April 2020 to May 2020, participants from three large, ongoing longitudinal studies (Nurses’ Health Study II, Nurses’ Health Study 3, and the Growing Up Today Study) were invited to complete a baseline questionnaire, which contained validated measures for various types of psychological distress. Researchers included participants who did not report a history of a positive SARS-CoV-2 test at baseline, returned at least one follow-up questionnaire, and reported a positive result on a SARS-CoV-2 test, for a total of 3,193 participants (6% of total sample population) for analysis.

The results found that in individuals with a positive SARS-CoV-2 test, those that had an increased risk ratio for depression, anxiety, worry about COVID-19, perceived stress, and/or loneliness were associated with post-COVID-19 conditions, when adjusted for sociodemographic factors, health behaviors, and comorbidities. Individuals with 2 or more types of distress prior to COVID-19 infection were at almost a 50% increased risk for post-COVID conditions. All types of psychological distress were associated with an increased risk of daily life impairment among individuals with post-COVID-19 conditions. The most commonly reported post-covid symptoms were fatigue (56%), smell or taste problems (45%), shortness of breath (26%), confusion/disorientation/brain fog (25%), and memory issues (22%).

Study limitations include a lack of sample diversity, self-reported SARS-CoV-2 infection status, and lack of randomness for missing data. The sample was predominantly White and female and comprised of primarily health care workers. Additionally, data were more likely to be missing among younger, unpartnered, health care workers, with higher distress at baseline. However, results were similar in analyses using multiple imputation.

In summary, psychological distress early in the pandemic was associated with increased risk for post-COVID-19 conditions and impairment in daily activities. These results indicate that addressing long COVID and its impact on the population is an important public health concern. This research highlights the need for a better understanding of how psychological factors contribute to long COVID symptoms and the need for interventions to help address the negative and sustained impact of COVID on mental and physical wellbeing.

Citation:
Wang S, Quan L, Chavarro JE, Slopen N, Kubzansky LD, Koenen KC, Kang JH, Weisskopf MG, Branch-Elliman W, Roberts AL. Associations of Depression, Anxiety, Worry, Perceived Stress, and Loneliness Prior to Infection with Risk of Post-COVID-19 Conditions. JAMA Psychiatry. 2022 Sep 7. doi: 10.1001/jamapsychiatry.2022.2640. Epub ahead of print. PMID: 36069885.

New frontier in music-based interventions: Measuring music exposure dosage to decrease delirium in older, mechanically ventilated patients in intensive care units

Mechanical ventilation for acute respiratory failure is used in intensive care units (ICUs) for one million adult patients annually. While in the ICU, up to 80% of those patients develop delirium, a syndrome of acute brain dysfunction resulting in a severe state of confusion, during their hospital stay. Older ventilated adults with delirium are at increased risk of longer hospital stays, in-hospital complications, elevated costs of care, and risk of death. Long-term post-ICU complications associated with delirium include cognitive impairment and dementia. Despite the high prevalence of delirium in ICUs, there are no effective pharmacological agents available for treatment, indicating the need for non-pharmacological therapies to prevent and/or manage delirium. In a recent study supported by the NIA, researchers investigate the use of music to reduce delirium and coma duration and severity in mechanically ventilated ICU patients and improve their post-ICU brain health. Prior research using a music-listening intervention in this patient population demonstrated a reduction in anxiety, pain, sympathetic nervous system activity, and the attenuation of inflammatory mediators, however it has not yet been established that music can reduce delirium in ICU patients who are mechanically ventilated. Additionally, this study will measure the dosage of music exposure while evaluating the potential of music to decrease or potentially prevent delirium.

The researchers are conducting a two-arm, parallel-group, clinical trial to assess the effectiveness of a music intervention compared to a silence-track control as treatment for older patients in the ICU with delirium. Adult patients (n = 160, age = 50 years and older), within 72 hours of ICU admission, will be randomized to one of two groups, a music intervention group or control group. Those in the music arm of the study will receive a dose of slow tempo (60 to 80 beats per minute) instrumental music through noise-canceling headphones for one hour twice daily for seven days. The control arm will receive a placebo consisting of a silent audio track administered in the same frequency and duration as the intervention group. Delirium, pain, and anxiety were evaluated daily. The study’s primary outcome of improving delirium as measured by the number of days patients were alive, not comatose, and free of delirium during the seven days of either exposure to music or silence.

The current study builds upon the researchers’ previous work, where they conducted a three-arm pilot acceptability and feasibility trial among mechanically ventilated, critically ill patients testing two music listening interventions (patient-preferred or slow-tempo relaxing). That study showed high feasibility and acceptability of 80% in each music arm, but also demonstrated a reduction in delirium duration in patients assigned to the slow-tempo relaxing music group. A secondary outcome of the current study is to investigate the downstream effects of relaxing, slow-tempo music on brain health outcomes. To better understand the mechanism(s) and components of the music intervention, a specially designed computer application will be used for both groups that tracks the length and frequency of listening sessions. For those in the music intervention arm, the computer application gathers data about the music selection from an extensive playlist. Three months after hospital discharge, the effects of exposure to music on cognition, as measured by memory, attention, information processing, speed, and executive cognitive function of recipients of music versus silent track were assessed as well as patient mood and anxiety.

In summary, this study is one of the first that will measure dosage of music exposure while evaluating the potential of music to decrease or potentially prevent delirium. Prior studies have been limited by methodological challenges, including small sample size, lack of blinding, and exclusion of the critically ill. These limitations are addressed in the current study protocol including blinding of outcome assessors, twice daily delirium assessments, assessments of pain, anxiety, and cognition post hospital discharge, as well as inclusion of critically ill ICU patients. Prior research has indicated that music listening activates areas of the brain involved with memory, cognitive function, and emotion. Interventions that increase brain activity in the areas related to memory, while reducing brain dysfunction, music-listening interventions may help retain cognitive function in older patients, especially those that experience critical illness or injury. The long-term goal is that the results from this trial may lead to the development of music algorithms and implementation of music-listening intervention protocols in a busy ICU. This study is currently ongoing and recruiting patients, thus the results and the dataset from this trial were not accessible at publication.

Citation:
Seyffert S, Moiz S, Coghlan M, Balozian P, Nasser J, Rached EA, Jamil Y, Naqvi K, Rawlings L, Perkins AJ, Gao S, Hunter JD 3rd, Khan S, Heiderscheit A, Chlan LL, Khan B. Decreasing delirium through music listening (DDM) in critically ill, mechanically ventilated older adults in the intensive care unit: a two-arm, parallel-group, randomized clinical trial. Trials. 2022 Jul 19;23(1):576. doi: 10.1186/s13063-022-06448-w. PMID: 35854358; PMCID: PMC9295531.

Study finds that machine learning models of brain function are not one-size-fits-all

Machine learning has been employed by researchers in various field of science to help untangle and understand complex relationships and processes. In neuroscience, it has been used to understand how the brain gives rise to complex human behaviors, including working memory, impulsivity, and various mental/psychological disorders. Using these tools, researchers can make models of these relationships with the goal of making predictions about an individual’s behavior and health. Previous linear modelling work has relied on two assumptions: 1) a single brain network is associated with a given phenotype, with activity patterns within that network varying across individuals; and 2) larger and more heterogeneous samples will more be more accurate and reliable to capture this single model. However, although there are several published models that have demonstrated good generalizability, they do not account for brain–phenotype relationships in all individuals. In a recent study supported by the NIGMS, NCATS, NIMH, NIH Blueprint for Neurosciences, and others sought to answer the question of for which individuals do the models fail, and if systematic bias plays a role.

The fact that there is model failure, with some individuals fitting a model better than others suggests that one brain–phenotype relationship does not fit all, and that systematic bias may determine who is fit and who is not. This failure may lead to imprecise, misleading, and potentially harmful model interpretations. However, whether brain–phenotype models are affected by bias in phenotype measurement and, if so, how this bias affects model failure remains unanswered. To answer this question, researchers first trained models that could use patterns of brain activity to predict how well an individual would score on a variety of cognitive tests. When tested, the models correctly predicted how well most individuals would score, however for some, they were incorrect, wrongly predicting that they would score poorly when they actually scored well, and vice versa. The research team then looked at which individuals the models failed to categorize correctly. They found that there was consistency in which individuals were being misclassified across cognitive tasks and analyses, such that those misclassified in one dataset had something in common with those misclassified in another dataset. Next, they investigated if these similar misclassifications could be explained by differences in those individuals' brains, however there were no consistent brain differences. The researcher, however, found misclassifications were related to sociodemographic factors such as age, education, and clinical factors such as symptom severity. For example, models employed in the study associated more education with higher scores on the cognitive tests. In turn, this meant that any individuals with less education who scored well did not fit the model's profile and were thus often predicted to be low scorers in error. It is important to note, that the models did not have access to sociodemographic information. In fact, these sociodemographic variables are embedded in the cognitive test score, through biases in how cognitive tests are designed, administered, scored, and interpreted, seeping into the obtained results. These results indicate that researchers need to carefully consider what is actually being measured by a given test and, therefore, what a particular model is really predicting.

In summary, the researchers determined that the models failed for individuals that did not fit a stereotypical profile. Additionally, the results indicated that the machine leaning models were not only reflecting cognitive ability but were instead reflecting more complex profiles that incorporated cognitive ability and various sociodemographic and clinical factors. The authors of this study provide recommendations for how to mitigate this problem including: 1) employing strategies that minimize bias and maximize the validity of the measurements being used at the study design phase and 2) after data collection researchers should use statistical approaches that correct for any stereotypical profiles that remain, whenever possible. Taking these measures should lead to models that better reflect the cognitive construct under study, however it was noted that fully eliminating bias would be unlikely and should be acknowledged when interpreting model output. Additionally, for some measures, it may be necessary to utilize more than one model. In conclusion, disentangling the relationship(s) between the construct of interest and a range of sample-dependent demographic and clinical variables presents an important and broad opportunity for future research.

Citation:
Greene AS, Shen X, Noble S, Horien C, Hahn CA, Arora J, Tokoglu F, Spann MN, Carrión CI, Barron DS, Sanacora G, Srihari VH, Woods SW, Scheinost D, Constable RT. Brain-phenotype models fail for individuals who defy sample stereotypes. Nature. 2022 Sep;609(7925):109-118. doi: 10.1038/s41586-022-05118-w. Epub 2022 Aug 24. PMID: 36002572; PMCID: PMC9433326.