Damian Sendler: Patients in psychiatry are classified based only on their own self-reported symptoms, rather than on objective diagnostic testing. In some cases, this technique has proved helpful in selecting therapy options; in the vast majority of cases, the treatment is ineffective or even non-existent. We must improve. In order to achieve this goal, researchers must establish clear and robust ties between clinical symptoms and the underlying neurobiological dysfunction (what physicians refer to as the etiology), so that objective tests and targeted treatments can eventually replace symptom-based differential diagnosis.
Damian Jacob Sendler: As a result, researchers are unable to identify these connections themselves. There is no biological specificity in the diagnostic and statistical manual of mental disorders (DSM), the most widely used psychiatric lexicon. Co-morbidity, or the risk of being diagnosed with two or more illnesses, is a significant factor in determining whether a person will be diagnosed with one or more of these conditions. The symptoms that define these disorders are exceedingly diverse, which means that two people with the same diagnosis might have quite distinct physical appearances. To acquire a diagnosis, patients are often only need to fulfill a small portion of a broad range of possible symptoms. In certain circumstances, such as ADHD and schizophrenia, two people might be diagnosed with the same condition with no overlapping symptoms.
Damian Sendler
Dr. Sendler: If taxonomic difficulties are a problem, it is not unexpected that the findings are not specific when researchers employ these categories as independent variables in their study For example, it is difficult to tell whether the observed differences between cases and controls are due to an illness of interest, or one of the numerous comorbid disorders that are likely to be found in the case group. Possibly because of this, intriguing etiologic leads might turn out to be unclear after additional investigation. The same is true for intriguing genetic connections and brain markers that were originally tied to a single condition but are now visible in several disorders.
Our current categorization system for mental disorders is in need of data-driven revision, as are the research methods that go along with it. Massive datasets are almost certainly required to disentangle comorbid and heterogeneous symptoms, find the most biologically valid features in the clinical phenomenology, distinguish between environmental and genetic influences, as well as handle high-dimensional problems like predicting treatment response to drugs with wildly divergent receptor affinity profiles.
Scientists are increasingly resorting to “big data” in order to overcome these restrictions. Researchers have undertaken comprehensive diagnostic interviews with hundreds of patients in the hunt for natural diagnostic boundaries that might explain patterns of comorbidity and family risk across illnesses (). It has been proposed that “mega-analyses” of imaging data may be performed by pooling existing information from several research locations in an attempt to better control causes of known heterogeny (eg. comorbidity, age, gender) (). IMAGEN, a huge multi-center study that tracked the progress of 2,000 14-year-olds in the hopes of uncovering risk factors for mental illness, is even more ambitious, since it involves collecting diagnostic, genetic, cognitive, and neurological data on the same people (). These methods are leading to some exciting discoveries, but they are not suitable for everyone due to their logistical complexity and high cost.
Psychology has embraced online data collecting in the face of comparable limitations in power and approach. Mobile phone apps have been created to track thousands of people’s performance in games that mimic popular cognitive activities. There have been recent studies to examine age-related spatial working memory impairment () and how reward receipt correlates with transient changes in happiness () (). Single-session tests may be done online, but “citizen scientist” initiatives including prominent media websites have also been used to conduct long-term research. Online data collecting is now simpler than ever, thanks to Amazon’s Mechanical Turk program, which started in 2005 and has since become a popular crowdsourcing tool.
Research in mental health might benefit greatly from this kind of data collecting, since the general population that can be accessed through Turk has a wide spectrum of symptoms and severity levels (). This hypothesis was recently explored in a research that sought to identify one specific aspect of mental health symptoms by looking at how it connects to an underlying neural process (). Compulsive behaviors such as addiction and OCD have been related to deficits in goal-directed control, which we have demonstrated may lead individuals to stay trapped in their routines. Compulsions in mental illness were formerly thought to be caused by abnormalities in the brain’s reward system, but new case-control studies show that these impairments are present in a wide range of diseases, including those that don’t exhibit obsessive symptoms ().
Damian Jacob Markiewicz Sendler: Web-based data collection was used to collect data from nearly 2,000 subjects in order to isolate the contribution of different aspects of psychopathology to goal-directed control. We suspected that this apparent lack of specificity might be due to diagnostic categories rather than the brain-behavior links. An online behavioral task that measured goal-directed control was used in our work, and its neurological roots are likewise well-studied (). Self-report data supplied by individuals was utilized to analyze the symptoms and severity of nine distinct characteristics of psychopathology supported by the DSM-IV-TR, rather than comparing one illness to another (depression, addiction, social anxiety, etc.). For example, if we assume that eating disorder symptoms are separate from OCD symptoms, then the association between goal-directed control and mental symptoms is not particular to one condition over the other when we analyse the data in the typical method. Co-occurrence of illnesses and overlap of symptoms between them, as detailed above, led researchers to this conclusion.
Damian Jacob Sendler
For this reason, we were able to evaluate how individual symptoms (rather than illnesses) naturally co-occur and assess if the adoption of the dimensional method would be more accurate than the current standard of care. When we looked at how respondents’ responses to more than 200 items were intercorrelated using factor analysis, we discovered evidence of three cross-diagnostic mental dimensions: (Figure 1). To our knowledge, these symptom dimensions were more explicit in their link to our independent evaluation of goal-directed control compared to methods that view eating disorders as different from OCD, for instance. Compulsive behavior was a greater predictor of goal-directed control deficits than any of the measures that quantify severity of DSM categories, including OCD, according to fundamental science expectations. An important first step in figuring out the causes of mental health issues and the most effective ways to treat them is to refine psychiatric categorization via the use of adequately powered online samples.
General population samples collected via Turk can answer many questions in psychiatry, such as how symptoms might relate to neural or cognitive constructs that can be assessed through behavioral testing, but other questions involving treatments and diagnosed patients are not so readily answered by such a sample population. However, alternative online testing methods give a wide range of possibilities.
One of the most significant and readily accessible goals for large, data-driven psychiatry is identifying which patients will benefit most from certain treatments. Even in laboratory research, computational methods that demand big datasets have demonstrated promising results. A recent research found that electroencephalography (EEG) indicators outperformed clinically-determined treatment strategies for depression (). Co-morbidity, age of beginning, and other baseline self-report variables (e.g., self-report measures of depression severity) have previously been found to predict chronicity of depression (). (). Despite the importance of lab-based therapy research, they are expensive, time-consuming, and difficult to scale up to obtain the necessary sample volumes for predictive analysis. In addition to being an attractive supplement, online data collecting allows us to target people who may never proceed beyond primary care. Advertising on internet forums and respected patient information sites, for example, might target those who look up details about antidepressants (a usual habit before beginning a new treatment).
Hybrid techniques that combine web-based testing with clinical collaboration may be even more beneficial. If primary care facilities, which are where the majority of antidepressants are administered, advertised online research, an unprecedented number of treatment-naive patients may be given access to novel treatments.
Damien Sendler: Online testing may allow for much tighter temporal tracking of symptoms than is possible in conventional laboratory investigations, in addition to enhancing patient accessibility and sample sizes. To better understand the temporal course of therapy response/non-response, smartphone applications may be used to track daily swings in symptoms. There are currently CBT providers working in this field, not just monitoring symptom changes, but also giving creative and cost-effective CBT options that have shown preliminary effectiveness for drug use and depression. ” (). It is possible to study cognitive and even neurological indicators of psychopathology in relation to changes in state vs changes in trait using this kind of time course monitoring. When symptomatology is low or high, applications that ask users to report on their mood on a monthly basis might be used to remember people for lab testing, which could provide further information about cause and effect. Patients needn’t travel to have cognitive testing done, of course.
Identifying biomarkers that may predict who is at risk of acquiring a condition is perhaps even more crucial than tracking symptom changes in diagnosed individuals. Thus, early intervention could enhance treatment outcomes in the long run, as some think. With this kind of study, you need the most extensive databases. Despite the fact that only a tiny percentage of healthy young people will acquire a mental health issue, we may anticipate numerous independent variables (genetic and environmental) to contribute to individual risk within that subset. Recruiting and maintaining big samples is made easier thanks to the Internet, but there is a caveat: the web does not provide access to all of the data one could need. In the absence of brain scans or blood draws, we may use cognitive tasks for which the neurological correlates have already been established in prior research (). Large-scale online testing may be used to develop reliable cognitive targets, which can then be transferred to smaller in-person samples where brain-imaging methods may provide further predictive value, in the absence of cognitive tasks that map onto well-defined neural circuits or genes.
We may perform naturalistic tests using prominent social networking and search websites thanks to the Internet’s influence on psychiatry’s research environment. If you’re interested in doing real-world experimentation based on the structure of a website, clicks, etc. and related drug usage (which may be deduced from Facebook “likes”), you could do so (). Lottery sales have been found to be impacted by unexpected favorable occurrences in the real world, as shown by other natural experiments (). Researchers now have a new way to analyze real-world clinically relevant behavior changes and identify risk factors for relapse as the popularity of online gambling rises. When Facebook released a research a few years ago that modified more than 650,000 individuals’ “News Feeds” to exclude good or bad messages from other users, it raised major concerns about informed consent in a commercial context, which is why this technique isn’t without its own ethical issues (). The problem may be alleviated by explicitly defining consent parameters, which will open up an important new research resource.
While the use of completely computerized testing enhances methodological consistency and consequently repeatability of study results, it also introduces a large degree of untracked variability to testing circumstances. To put it simply, it eliminates the inherent variability in verbal task instructions, eliminates the possibility that patient groups are routinely coached or given supplemental or special instructions, and eliminates other forms of unintentional experimenter influence that don’t typically appear in research papers Data from online research projects may thus be considered generalizable.
There’s little doubt that self-reporting for online psychiatric research will become the primary method of diagnosis in the future. This, on the other hand, is the most contentious aspect of the project. There is a strong belief that skilled raters are needed for properly determining the presence or absence of DSM illnesses, or differential diagnosis. A compelling case may be made that clinician-rated measures are less accurate than self-reported data in several critical areas. Another kind of noise is added by clinicians—inter-rater discrepancies in the interpretation of patients’ answers (e.g., test, re-test). Diagnostic interviews will be phased out in favor of online data collecting in psychiatry, which is more scalable and ambitious. These findings might help us develop our self-report instruments, which are increasingly being used as a primary research tool and as a way to better understand the underlying psychological and biological processes.
Online testing has the potential to boost generalizability in addition to uniformity. These samples are more typical of the general population of the United States than those collected on university campuses (). For those who can’t leave their homes or go to a university campus, Internet-based research eliminates a significant barrier to involvement. As a result, research may include patients with the most severe disabilities, who would otherwise be left out of trials conducted in person. In a similar vein, it gives researchers access to hard-to-find populations, such as Mechanical Turk users, who, although unhappy and nervous like the rest of us, exhibit higher levels of social anxiety.
Research in psychiatry is undergoing a major shift because to the Internet. Classification must be reworked in order to relate mental states to their neurobiological genesis using “big data.” Treatment development research will experience a resurgence this way. If you’re looking for enormous datasets, the Internet isn’t your only option, but it offers a number of advantages over conventional lab methods, including the fact that it’s less expensive. We will be able to perform hybrid studies that combine field experiments with controlled manipulation, as well as investigate symptoms in the wild and in more depth than we have ever been able to in the past thanks to this technology. For psychiatry, this is an exciting time: an opportunity (and a challenge) to generate large and daring new ideas, the fruits of which may make a meaningful impact in the practice of the profession.