Neuroimaging in the Acute Psychiatric Setting: Associations With Neuropsychiatric Risk Factors

The debate regarding the appropriateness and clinical utility of neuroimaging in psychiatric populations stems from limitations of imaging to meaningfully guide diagnosis and treatment.

Structural and functional neuroimaging continue to play an important role in psychiatric research as the field works toward neurobiological understandings of etiology, diagnosis, prognosis, and targets for intervention and management (1–4).

A predominant focus in the past three decades has been on the first-episode psychosis population. In a systematic review of the clinical utility of neuroimaging in first-episode psychosis, Forbes et al. (5) found that structural abnormalities were not uncommon but rarely deemed the cause of symptoms or required further intervention (median=3.5%); the investigators concluded that unless neurological or cognitive abnormalities are present, routine imaging should not be ordered. While guidelines differ internationally, ultimately, most conservative recommendations conclude that imaging is warranted if clinical presentation, history, signs, and symptoms raise concerns about intracranial pathology (6–8).

In psychiatric populations outside of first-episode psychosis, patients are more likely to have neuroradiological abnormalities. Beyer et al. (9) studied the use of computerized tomography (CT) and MRI in a large naturalistic cohort comprising all inpatients and outpatients of a psychiatric clinic who underwent brain scanning over a 10-year period (N=2,922) and found that 32.8% had clinically relevant abnormalities on imaging associated with older age and psychiatric diagnosis. Mueller et al. (10) studied brain imaging in consecutively hospitalized psychiatric patients over 4 years (N=435), with a focus on stratifying by demographic data and neurological signs and symptoms; the investigators found that 14.3% and 16.3% of brain scans were abnormal or equivocal, respectively, and that abnormal scan results were more likely among those with focal neurologic signs or advanced age but that these categories alone were not sufficiently predictive of imaging outcomes. In a smaller South African study (N=53), Juby and Paruk (11) found 62.0% of MRI scans to be abnormal among patients in a psychiatric facility; of these, 54.0% required referral to other specialties, a large proportion of whom had comorbid HIV.

Another debate surrounds the presence and relevance of incidental findings on neuroimaging in research and routine care. In a systematic review and meta-analysis of incidental brain MRI findings among people without neurological symptoms who underwent imaging for screening or research purposes, Morris et al. (12) found a prevalence rate of 2.7%, which increased with age. In a more recent systematic review and meta-analysis, Gibson et al. (13) found a prevalence rate of 1.4% of potentially serious incidental findings on brain MRI and 1.7% when indeterminate findings were included. In the Rotterdam Scan Study, Bos et al. (14) found that in a cohort of participants middle-aged and older (mean±SD age=64.9±10.9 years), 9.5% of participants had incidental findings, most commonly meningiomas and cerebral aneurisms, and 3.2% of participants required referral to other medical specialties. In their follow-up study, the majority of those referred for further investigation or management had conditions that did not require treatment.

Incidental findings are infrequently studied in a clinical context; however, in the psychiatric population, the prevalence is potentially higher. Lubman et al. (15) found a higher prevalence of incidental findings among patients with chronic schizophrenia (50.0%; requiring referral, 20.0%) compared with patients with first-episode psychosis (22%; requiring referral, 8.5%) and control subjects (23.7%; requiring referral, 5.0%). Importantly, the differences could not be accounted for solely due to older age.

Nevertheless, while imaging guidelines in psychiatry largely focus on first-episode psychosis, there is a lack of similar guidelines for patients with nonpsychotic or established illnesses. A set of “red flags” has been proposed when considering neuroimaging use for patients with known psychiatric illnesses encompassing neurological signs or symptoms, preexisting neurological conditions or brain pathology, significant change in clinical presentation, family history of neurological disorders, history of head injury or seizures, or acute-onset delirium-like features, as well as for patients prior to electroconvulsive therapy (16).

Additionally, there are inherent difficulties that come with investigating and diagnosing neurological and neurocognitive disorders, particularly for people who are acutely mentally unwell. Diagnostic overshadowing, or the tendency to attribute somatic symptoms to psychiatric illness, contributes to the health disparity between patients with mental illness and the general population (17). The insidious nature and overlap of symptoms only compound the challenges (18). Studies have repeatedly found delays from symptom onset to diagnosis of neurodegenerative disorders, on average 1–3 years among older patients and up to 3.4 years among people with younger-onset dementia (19–22).

In this context, we examined a general adult psychiatric population in a naturalistic hospital setting to review MRI ordering practices, identify links with neuropsychiatric risk factors, and assess the outcomes of structural imaging in terms of urgency. Secondary aims of the present study were to review the rates of imaging findings that influence psychiatric management and the role of CT scans preceding MRI, as well as to consider the generalizability of findings to broader adult psychiatric populations.

Methods

Participants

A retrospective file review was undertaken for inpatients of the acute adult mental health unit at Royal Melbourne Hospital, identifying 100 consecutive patients who had a brain MRI as part of their clinical care from October 1, 2018, which was anticipated to span approximately the previous 3–4 years. The imaging reports and clinical files for the associated 100 admissions were reviewed. This study was authorized by the Human Research and Ethics Committee at Melbourne Health. There were no specific exclusion criteria.

Study Variables

Information was ascertained from scanned medical records and imaging reports, including demographic characteristics, indication for admission, psychiatric diagnosis at the time of discharge, age at onset, and duration of psychiatric disorder, calculated from age at onset to age at time of MRI.

Specific psychiatric phenomenology commonly associated with brain pathology was recorded, including catatonia and visual hallucinations. History of neurological or intracranial conditions was determined, and neurological examination findings were included after excluding those related to medication side effects. The presence of cognitive impairment was noted; objective cognitive impairment was defined in accordance with Neuropsychiatry Unit Cognitive Assessment Tool (NUCOG) criteria (score <80), and subjective cognitive impairment was based on documented concerns of the treatment team, patient, or family (23).

Indications for Neuroimaging

Indications for imaging were established from the radiology referral form or clinical file. Radiology results were reviewed to identify patients who had a CT scan preceding MRI during the same admission. All MRI scans were inpatient scans performed on-site at the Royal Melbourne Hospital and reported by radiologists or neuroradiologists as part of routine clinical care.

Indications for MRI were categorized as follows:

• First-episode psychosis: first presentation with a diagnosis of psychosis.

• Atypical psychiatric presentation: clinical presentation inconsistent with the typical illness pattern for the psychiatric diagnosis or inconsistent with relapse patterns of the patient’s mental illness.

• Known neurological condition: presence of established neurological diagnosis or intracranial abnormality known prior to the current admission.

• Physical, neurological, or cognitive abnormalities: presence of neurological, cognitive, pathological, or physical abnormalities that may affect cerebral structures, including specific events such as head trauma or indeterminate findings from brain CT.

• Screening: routine imaging undertaken for work-up of psychiatric symptoms without the referrer conveying particular suspicion of brain pathology.

Neuroimaging Outcomes

Urgency of findings.

The presence of any abnormality independent of acute, subacute, or chronic time course was determined from MRI reports. Abnormal findings were classified according to the urgency of follow-up referral required, a method used previously in several studies investigating incidental findings on brain MRI (15, 24, 25) and described below.

• No referral necessary: findings are common in asymptomatic individuals (e.g., cavum septum pellucidum).

• Routine referral: findings do not require immediate or urgent medical review but should be noted by the patient’s treating practitioner and monitored or reviewed in a subacute manner (e.g., old infarct or possible vascular disease).

• Urgent referral: findings require prompt but nonemergent medical review or further investigation within days to weeks (e.g., arteriovenous malformation).

• Immediate referral: requires immediate medical review (e.g., acute subdural hematoma).

Psychiatric relevance.

Imaging outcomes were further assessed for psychiatric relevance—that is, whether abnormalities were directly related to the clinical psychiatric symptoms, contributed to an understanding of the psychiatric presentation, or were in line with changes that could indicate an early neurodegenerative process, leading to alterations in the psychiatric management plan. This was determined from documentation in the patient’s clinical file, including subspecialty opinions, further investigations, or follow-up arranged as a result of MRI findings.

Data Analysis

Univariate analyses were used to identify associations between demographic or clinical variables and imaging results. Odds ratios were used to investigate the association between binary variables. Mean differences were calculated for continuous variables. Statistical inference was performed by using confidence intervals at the 95% level. All statistical analyses were performed by using SPSS, version 26.0 (SPSS, Chicago).

Results

Demographic and Clinical Data

There were 100 consecutive individual patients included in the analysis (male, N=61; female, N=39). The age of participants ranged from 18 to 70 years (mean±SD=46.61±13.27). The length of hospital stay ranged from 1 to 111 days (mean=25.08±19.85). Demographic and clinical data are summarized in Table 1.

During the study period, a total of 2,735 unique patients (total patient admissions, N=3,551) were admitted to the psychiatric ward, based on hospital admissions data. Hence MRI was performed for 3.7% of patients during their admission.

Psychiatric Diagnosis and Phenomenology

A primary psychotic disorder was the reported discharge diagnosis for 59 patients (schizophrenia, N=34; first-episode psychosis, N=13; schizoaffective disorder, N=10; drug-induced psychosis, N=1; delusional disorder, N=1). Affective disorders accounted for 25 cases (bipolar affective disorder, N=12; major depressive disorder, N=8; depression with psychotic features, N=5). A diagnosis of stress-related disorder was reported for seven patients (acute stress reaction, N=5; adjustment disorder, N=1; posttraumatic stress disorder, N=1). Four patients were diagnosed with neurocognitive impairment (acquired brain injury, N=1; frontotemporal dementia, N=1; vascular dementia, N=1; Huntington’s disease, N=1). “Other” diagnoses (steroid-induced mania, cholesteatoma, delirium, peri-ictal psychosis, and autoimmune-related psychosis) were given to five patients. Age at onset ranged from 15 to 63 years (mean=36.82±14.04), and duration of psychiatric illness ranged from 0 to 44 years (mean=9.79±11.89). There were 10 patients with visual hallucinations, nine patients with catatonia, and one patient with both.

Neurological History and Examination

Neurological examination was abnormal for 45 patients; of these cases, four were directly attributed to extrapyramidal medication side effects. There were 26 patients with a history of neurological or intracranial conditions; of these, 17 had an abnormal neurological examination documented. There were 24 patients with an abnormal neurological examination not explained by medication or a preexisting condition. Details of neurological examination findings and preexisting conditions are presented in Appendix S1 in the online supplement.

Cognitive Impairment

Subjective cognitive impairment was documented for 66 patients. The NUCOG was completed for 49 patients, with scores ranging between 43.5 and 99, out of 100 total (mean=78.80±14.13). Objective cognitive impairment was documented for 31 patients, and their mean age was 54.1±10.57 years. The mean age of patients without cognitive impairment or who did not undergo testing was 43.4±13.07 years.

MRI

Indications for neuroimaging.

Multiple indications for MRI (versus a single indication) were common, with 57 patients having two or more indications (two indications, N=37; three indications, N=18; four indications, N=2). An indication for the vast majority of patients (N=91) was a physical, neurological, or cognitive abnormality; among these patients, 40 had no other indication, 36 had an atypical psychiatric presentation, 19 had first-episode psychosis, and 17 had a known condition. The next largest indication category was atypical psychiatric presentation (N=41); among patients with an atypical psychiatric presentation, one had no other indication, 19 had first-episode psychosis, and eight had a known condition. Of patients with indications for a known neurological or intracranial condition (N=20), all had another indication for scanning, including five who had first-episode psychosis. One patient alone fell into the “screening” category for first-episode psychosis. The overlap in indication categories is summarized in Appendix S2 in the online supplement.

Outcomes and psychiatric relevance.

Seventy-nine patients had MRI scans with one or more abnormalities; of these patients, no referral was necessary for 22 patients, routine referral was necessary for 43 patients, and urgent referral was necessary for 14 patients. No patients required immediate referral. Types of abnormalities are summarized in Appendix S3 in the online supplement.

There were 32 scans with findings relevant to the patient’s psychiatric presentation; of these cases, 12 patients required urgent follow-up, with findings including arachnoid cyst, cholesteatoma, intracranial hypertension, possible frontotemporal dementia, possible Lewy body dementia, possible toxic disorder (chronic or acute disulfiram toxicity and carbon monoxide poisoning), progressive global atrophy, severe chronic obstructive hydrocephalus, and vascular lesions.

Twenty patients required routine follow-up, with findings including generalized atrophy, alcohol-related dementia, possible demyelinating disease, prominent areas of calcification or ischemia excessive for their age, regional atrophy or hypoperfusion, traumatic changes, and stable known disorders (Huntington’s disease, multiple sclerosis, tuberous sclerosis, and ocular abnormalities).

Forty-seven scans had abnormalities not deemed to be relevant to the patient’s psychiatric presentation; of these cases, urgent referral was required in two: one for incidental arteriovenous malformation and another for severe chronic obstructive hydrocephalus overdue for neurosurgical review.

Demographic Variables Associated With MRI Findings

The mean age of patients with a normal MRI was 39.19±14.43 years, and the mean age for those with any abnormality was 48.58±12.30 years. The average length of hospital stay was 8.3 days (for admissions ≤35 days). Four percent of patients had long-stay admissions (>35 days) (26). For patients with MRI scans of psychiatric relevance, the mean age was 51.66±12.58 years; the mean age at onset of their mental illness was 41.41±13.75 years; the mean duration of illness since diagnosis was 10.25±12.17 years; the mean NUCOG score was 76.27±13.73; and the mean length of hospital stay was 27.84±23.29 days.

There was an association between older age and both abnormal MRI (mean difference=9.39 years; 95% confidence interval [CI]=−15.61, −3.17) and psychiatrically relevant MRI findings (mean difference=7.42 years; 95% CI=−12.90, −1.95), as well as longer duration of psychiatric illness (mean difference=5.04 years; 95% CI=−9.45, −0.63). Psychiatrically relevant findings were also associated with older age at onset of psychiatric disorder (mean difference=6.74 years; 95% CI=−12.60, −0.89). Total NUCOG score and length of admission were not associated with MRI results. The associations between demographic characteristics and neuroimaging results are summarized in Table 2.

Neuropsychiatric Risk Factors Associated With MRI Findings

Cognitive impairment was more common among patients with an abnormal MRI (N=29, 36.71%) than those with no MRI abnormality (N=2, 9.52%; odds ratio [OR]=5.51; 95% CI=1.20, 25.37). Cognitive impairment was also associated with a psychiatrically relevant abnormality on MRI (N=15, 46.88%) compared with no psychiatrically relevant abnormality on MRI (N=16, 23.53%; OR=2.87; 95% CI=1.18, 7.00). There was a weak association (OR=2.36; 95% CI=1.00, 5.57) between an abnormal neurological examination and psychiatrically relevant MRI findings (N=19, 59.38%) compared with the relative lack of association among those with an abnormal examination and no psychiatrically relevant MRI findings (N=26, 38.24%).

There was no association between MRI results and catatonia or visual hallucinations.

Forty-one patients had a brain CT scan prior to an MRI scan, and there was no association with MRI results. The associations between clinical factors and neuroimaging results are summarized in Table 3.

Discussion

The main finding of this study was the high proportion of psychiatrically relevant findings on brain MRI despite a low rate of scanning, with associations between abnormal scans and cognitive or neurological impairment, older age, and longer duration of psychiatric disorder. This review of indications for imaging, urgency of follow-up, and relevance to psychiatric management highlights the importance of subacute findings in psychiatric populations.

Indications

The ambiguity of guideline recommendations for neuroimaging in the psychiatric patient population is well established and has recently been summarized by LeBaron et al. (27). Without consistent guidelines, psychiatrists are left to determine the circumstances in which neuroimaging is more likely to reveal abnormalities pertaining to clinical care, while balancing the drawbacks of anxiety-provoking investigations, consequences of incidental findings, and economic considerations in strained public health systems.

In our study of 100 scans spanning 3.5 years and 3,551 total patient admissions, only one patient had MRI screening for first-episode psychosis requested without the presence of other neuropsychiatric risk factors. In an age of “no unnecessary tests,” this is an important finding because it marks progress in general psychiatric practice in light of research repeatedly demonstrating the limitations of routine imaging in the first-episode psychosis population.

More than one-half of patients (57.0%) had multiple indications, which reveals that patients were not scanned indiscriminately. An indication for the vast majority of patients (91.0%) was a physical, neurological, or cognitive abnormality, raising concern for organic pathology. Of note, for 20 patients, one of the indications for scanning included the presence of a known neurological or intracranial abnormality; however, all of these patients had at least one additional indication.

We also identified that nearly one-half (41.0%) of patients had a CT scan prior to an MRI scan, which was not associated with subsequent abnormality on MRI, nor was an abnormal CT scan more likely to yield an abnormal MRI scan. From this, we can imply that CT rarely assists the treatment team in identifying meaningful findings. Clinicians may keep this in mind when considering judicious allocation of health care resources.

Abnormalities on Neuroimaging and Urgency of Follow-Up

In psychiatric populations, abnormalities on brain MRI have ranged between 14.0% and 50.0%, with the majority of cases not requiring further investigation (9, 10, 15, 25). Such a wide range can be explained by a number of factors, including whether studies report all abnormalities or only those that are clinically significant or require further referral, the resolution of the imaging technique, and patient demographic characteristics. Nevertheless, the rate of abnormal scans in our study was higher than that reported by other investigators, with 79.0% of patients categorized as having “any abnormality” on MRI. The proportion of patients who underwent inpatient MRI during the study period was small (3.7%), when similar studies have reported a scanning rate of 5.3% (10). When combined with the rates of pathology found, this indicates the highly selective use of MRI.

Categorizing MRI abnormalities by urgency of follow-up referral is useful in determining the severity of findings, compared with abnormalities in the general population, and allows some comparison with prevalence of incidental findings. In a hospitalized group, we expect a higher proportion of patients than healthy control subjects to require referral due to age-related changes and comorbidities. In the present study, of the 79 abnormal scans, no referral was necessary in 22 (27.8%) cases, routine referral was needed in 43 (54.4%), and urgent referral was needed in 14 (17.7%). As expected, no patients required immediate referral because in our center, patients are almost always assessed in the emergency department prior to transfer to the psychiatric ward. More than one-half of those with abnormalities required routine referral, which in some settings may be dismissed as unremarkable; however, in psychiatric populations, chronic or subacute findings can be significant given the high burden of cardiovascular comorbidities, neurocognitive decline, and associated morbidity and mortality. Additionally, we identified an association between MRI abnormalities and longer duration of mental illness. Together, these results highlight the burden of physical comorbidities that contribute not only to a significantly reduced life span but also to additional financial, social, and personal burden for those with severe mental illness when compared with the general population (28, 29). As a specialty, we do our patients an injustice by minimizing these implications. Improving scientific knowledge of the accumulation of brain changes in chronic psychiatric conditions may also present an opportunity to further the pursuit of staging of mental illnesses and, more specifically, tailor treatment.

Psychiatric Relevance

Given the presence of comorbidities and known conditions in naturalistic populations, it was important for us to distinguish MRI abnormalities thought to be relevant to the psychiatric presentation from those that may be incidental or simply associated with known pathology. For example, areas of white matter hyperintensity in a healthy asymptomatic person may be interpreted differently in someone with known multiple sclerosis, but this still does not necessarily mean that the findings are related or relevant to the psychiatric presentation. In our study, 32.0% of patients had psychiatrically relevant findings, which was not always apparent from the MRI report. This highlights the importance of tailoring MRI requests and including enough detailed clinical information for the reporting radiologist to comment specifically on the presence and absence of findings that may be psychiatrically relevant but that may otherwise appear incidental, nonurgent, or of little consequence (30).

Associations With Neuropsychiatric Risk Factors

With regard to neuropsychiatric risk factors, MRI abnormalities were associated with cognitive impairment, older age, and longer duration of psychiatric illness. Findings of psychiatric relevance were associated with cognitive impairment, older age, and older age at onset of psychiatric illness, as well as weakly associated with neurological abnormalities.

We considered that there would likely be an underreporting of cognitive impairment in our sample because several cognitive assessments were unable to be completed due to the patient’s mental state or because scores of particular NUCOG subsections fell below two standard deviations from normal, although the total score was above 80. Furthermore, subjective cognitive impairment was documented for 66.0% of patients. A confounding factor is the accuracy of a cognitive assessment while the patient is acutely mentally unwell. This was often taken into account during admission, with treating teams noting when deficits in attention and concentration were felt to be related to mental illness. Accordingly, for several patients, the treatment team recommended repeat cognitive assessment when the patients’ mental state was more stable. For those whom cognition was felt to be disproportionately impaired, follow-up was arranged. For some patients, a “watch and wait” approach was recommended, with repeat cognitive testing within a 12-month time frame.

Similarly, this study’s ability to identify a stronger association with neurological and imaging results may have been affected because 24.0% of patients did not have any neurological examination documented.

In an acute adult mental health ward, there is typically a wide age range of patients, which was represented in our sample. It was not unexpected that older patients were more likely to have abnormalities on scans due to an accumulation of age-related changes and cardiovascular risk factors, which is consistent with previous studies (9, 10). Additionally, we were able to identify that older patients had a greater number of psychiatrically relevant findings, highlighting not only that age-related incidental findings were accumulating but also that the findings were felt to be relevant to the psychiatric presentation. Of note, patients with a psychiatrically relevant abnormality had an older average age than those without such an abnormality, yet the mean age of these patients was 41.41±13.75 years. This serves as a reminder to pay particular attention to the physical, cognitive, and neurological characteristics of patients who first present around middle age.

There were no associations with visual hallucinations or catatonic symptoms in this study; however, given the low frequency of these symptoms, we anticipate that larger sample sizes may be needed to detect any associations if present.

Strengths and Limitations

The strengths of this study include the naturalistic design reflecting real-world clinical presentations and use of hospital-grade MRI scanners, with imaging reports that can be expected in a tertiary hospital environment. Limitations of this approach include risks of gaps or errors in the documentation relied upon for data collection, use of a single center that could reflect local trends in clinical practice, and a reasonably small cohort. In addition, the presence of neuropsychiatrists and neuroradiologists with associated specialized expertise in identifying radiological changes was noted. While this may have resulted in increased numbers of findings deemed relevant in our cohort, neuroradiology should be considered an important aspect of continuing professional development for general psychiatrists and psychiatrists-in-training, including what constitutes appropriate clinical information to provide to reporting radiologists. In this way, a higher level of clinical care may be provided to improve outcomes for what is generally a disadvantaged population less likely to independently seek investigation and follow-up.

Conclusions

This study supports the utility of brain MRI in psychiatric management if there is evidence of cognitive or neurological impairment, as well as for patients with an older age at onset or longer duration of psychiatric disorder. The task of differentiating between urgency of findings on MRI and subacute but psychiatrically relevant findings is important when applied to naturalistic patient populations. The high proportion of psychiatrically relevant findings in our sample despite a low rate of scanning, along with a review of indications for imaging, indicated that modern psychiatric treating teams can judiciously apply clinical filters to tailor investigations while limiting excessive imaging. Future research in larger cohorts across multiple centers would be beneficial in expanding on our findings and may contribute to shaping more consistent guidelines for neuroimaging in psychiatry.