Predicting treatment-seeking for visual hallucinations among Parkinson's disease patients




While much research has been conducted towards understanding the basis of visual hallucinations in Parkinson's disease, little has focused on characterizing the content and patients' emotional experience. These factors are likely very influential on a patient's decision to seek treatment, a critical aspect of any symptom from the clinical perspective.


A retrospective chart analysis was performed on Parkinson's disease patients seen in a community-based Parkinson's Disease and Movement Disorder Clinic between 2005 and 2010.


The study consisted of 334 patients with Parkinson's disease, among whom 10.5% had visual hallucinations. Hoehn and Yahr disease stage (P = 0.001), concurrent presence of dementia (P = 0.001),and sex (P = 0.031) were significant onset predictors. The most significant determinant of treatment-seeking was emotional reaction, namely whether hallucinations were bothersome (P = 0.008). However, the specific type of content during hallucinations was sometimes more influential and contradicted emotional response.


Although treatment-seeking can be predicted by how individuals feel about hallucinations, a patient's decision may not be logically consistent. We suggest that clinicians offer treatment based on patients' recollections and opinions.

VISUAL HALLUCINATIONS (VH) are reportedly observed in approximately one-third of Parkinson's disease (PD) patients,[1] although some variation has been suggested by different sources.[2, 3] The presence of this non-motor symptom is now well recognized and no longer considered a side-effect of anti-parkinsonian medication.[4]

The basis of VH in PD is still not completely understood. It has been suggested that they may occur due to deficiencies within the lower levels of visual processing, such as with attention,[5, 6] object and space perception,[5] and image recognition speed.[6] Cognitive impairments have been commonly reported as risk factors for VH,[1, 3, 7] indicating the involvement of higher order processing as well. Further, VH in PD have been related to other complications, such as sleep disturbance.[8]

In rare cases, hallucinations of other modalities have manifested in PD.[9] Experiencing hallucinations in more than one sensory medium does not indicate presence or risk of greater neuropsychological impairment,[10] however recent findings suggest that individuals with VH are more likely to develop hallucinations of other modalities with time progression.[11]

Relatively little research has been conducted on the content of VH experienced by PD patients. The available information suggests an individual will most likely endorse stereotyped content of hallucinations.[12] Between individuals, the observed entities do vary, ranging from people and animals to buildings and scenery.[2] These experiences tend to last for minutes and insight is usually retained.[13]

Previous research also claims that instances of VH in PD are devoid of emotional content.[12] However, complex hallucinations like those encountered by PD patients have been reported to invoke emotional responses, either during hallucinations[13] and/or afterwards.[2]

This study observes various factors contributing to a patient's desire to treat VH in a clinical sample. It further focuses on hallucination content and personal emotional reflections on VH. To the best of our knowledge, these factors have not been analyzed concurrently.

Since these factors lead to a horrid experience, they would be the greatest motivation for treatment. The data was also used to analyze correlations between the prevalence of visual hallucinations and patient sex, age, disease progression, and presence of dementia.


Patient source

All 334 individuals in the study were patients diagnosed with idiopathic PD and regularly followed in a community-based PD and movement disorders center from 2005 to 2011. The PD diagnoses were based on the UK Parkinson's Disease Society Brain Bank clinical diagnostic criteria.[14] Atypical parkinsonism was excluded for consistency. Altogether, 11 met the exclusion criteria and we removed them from analyses, except where stated otherwise.

The method of this study conforms to the provisions of the Declaration of Helsinki in 1995 (as revised in Tokyo 2004) and has been approved by the ethics committee of the Centenary Hospital part of the Rouge Valley Health System. All participants gave informed consent and anonymity has been preserved.

Evaluation of VH

Each patient was questioned in detail about their VH by the treating neurologist in every clinic visit. Symptoms were endorsed by accompanying caregivers. This was considered to verify the presence and severity of hallucinations and to guide treatment options. Each patient, even demented ones, understood their VH as ‘hallucination’, felt it as real but considered it as hallucination because of the information from the caregiver. By suggestion based on previous research,[15, 16] all patients were rated on the VH component of the Parkinson's Psychosis Rating Scale (PPRS)[17] for measuring severity. The PPRS is aimed at rating the quality, severity, and frequency of six components, one of which is the VH component. The symptoms are scored on a four-point scale, from 1 (absent) to 4 (severe).[17]

For those with VH, the age and Hoehn and Yahr PD (H&Y) stage at onset were recorded. The contents of all hallucinations were categorized as ‘Familiar people,’ ‘Unfamiliar people,’ ‘Other animate objects,’ or ‘Other inanimate objects’, and if they were bothersome/threatening or not. Also, the number of different types of contents was recorded as a measure of diversity of VH experienced.

Hallucinations of other modalities

All patients were independently screened for auditory, olfactory, and somatosensory hallucinations. This information was also recorded in the same manner as the data for VH.

Patients with hallucinations excluded from analysis

Patients were screened for history of drug abuse as a confounding factor. Those who experienced hallucinations potentially as a result of factors other than PD, in particular, due to the introduction or dose escalation of medicine, were excluded. Furthermore, individuals with prior history of potentially hallucinating disorders, like urinary tract infection or psychosis, were excluded.

Studied parameters and statistical analysis

Several regression analyses were performed to identify correlations between observed variables. Logistic regression analyses included all subjects barring those excluded to observe the predictive value of sex, age, PD stage, and dementia on VH presence. All available data within only the group of patients with VH was specifically analyzed to establish any correlation among them. Dementia was diagnosed using the DSM-IV criteria.[18]

In each analysis, the full model was reduced as required by step-wise removal of highly insignificant factors in the model and by non-parametric tests. This was to ensure that only factors pertinent to the outcome variable were observed and not skewed by factors without substantial predictive power. For all significant variables in each analysis the level of significance and odds ratio (OR), along with 95% confidence intervals (CI), are provided.


Of the 334 PD patients, 45 (13.5%) reported experiencing VH. Eleven of them were excluded (eight classified as drug-induced VH and two as having history of potentially hallucinating disorders). The remaining 34 of 323 patients (10.5%) were identified to have hallucinations that could not be attributed to obvious external factors. These patients were on commonly used anti-parkinsonian drugs, including levodopa, pramipexole, rasagiline, and amantadine. They could have been under isolation or a combination of drugs. We did note that throughout the assessment, the dosage of each medication for every patient was constant without any fluctuation; however, the combination and dosage varied depending upon the condition of the PD. There were no patients on anticholinergic medications. No patient had history of drug abuse. Out of the 34 patients, 29 were rated two out of four on the PPRS while the other five were rated three out of four. Much of the categorized data obtained in the study are presented in Table 1.

Table 1. Summary of results for groups with no VH, with VH due to Parkinson's disease, and with treatment-induced VH
 No VH (n = 289)VH due to PD (n = 34)Drug-induced VH (n = 8)
  1. aBased on onset age and stage for those with VH, current for those without.
  2. H&Y, Hoehn and Yahr; PD, Parkinson's disease; VH, visual hallucinations.
Sex: female41.9%55.9%37.5%
Mean agea73.9 ± 10.676.0 ± 8.975.0 ± 2.7
H&Y stagea   
Stage 253.6%14.7%12.5%
Stage 336.0%61.8%75.0%
Stage 410.4%23.5%12.5%
Dementia present15.2%41.2%12.5%
VH bothersome58.8%87.5%
VH threatening14.7%12.5%
VH treatment wanted58.8%100%
VH content: Familiar people8.8%0.0%
VH content: Unfamiliar people76.5%62.5%
VH content: Other animate objects44.1%25%
VH content: Other inanimate objects14.7%37.5%

Factors relating to VH presence

The logistic regression analysis with sex, age, PD stage, and dementia as predictive factors on VH presence consisted of the 323 included patients. The age and stage at onset were used. For patients without hallucinations, the current age and stage was applied, as these values represented the highest age and stage reached by patients without having hallucinations. Only PD stages 2, 3, and 4 were analyzed. Patients were grouped into stage 1 with stage 2 and those in stage 5 with stage 4. This was to simplify the analysis and avoid imbalance in the data due to very few individuals being in stages 1 (n = 3; of whom one had VH) and 5 (n = 2; none with VH).

The analysis indicated that sex, PD stage, and presence of dementia were predictive of presence of VH (Table 2). The OR showed greater likelihood among women, those in stage 3 and 4 compared to stage 2, and those with dementia.

Table 2. Comparison of factors with significant predictive power on the presence of VH and the presence of dementia
Wald χPOR95%CIWald χPOR95%CI
  1. aOverall significance of full models for both visual hallucinations and dementia was P < 0.001.
  2. bCalculated using values of age and Hoehn and Yahr stage at visual hallucination onset.
  3. CI, confidence interval; OR, odds ratio; PD, Parkinson's disease; VH, visual hallucinations.
PD stageb14.6280.0010.9030.637
2 vs 3b13.552<0.0017.5902.580–22.3300.3880.5330.7970.391–1.626
2 vs 4b10.7300.0019.0562.423–33.8430.1240.7251.1840.462–3.036
3 vs 4b0.1250.7231.1930.449–3.1750.7720.3790.6730.278–1.627

Due to dementia being a significant factor according to existing literature and our own observations, we analyzed it independently in the overall sample. This was to compare the correlation of hallucinations with sex, age, and PD stage to the correlation of dementia with these same factors, as dementia may have confounded their respective significance values (Table 2). This indicated that only sex may have overlapped between hallucinations and dementia.

Content of VH

Most subjects reported hallucinations of unfamiliar people within and outside the home, followed by animate objects other than humans, such as animals and insects (Table 1). Very few recollected hallucinations of familiar people or of other objects that would fit in no category, such as cobwebs and smoke. Most individuals (n = 23) endorsed only one type of hallucination content. A few patients endorsed some diversity, recounting two (n = 7), three (n = 1), four (n = 2), or five (n = 1) different types.

Covariates that could significantly predict the experience of content of ‘Unfamiliar people’ and ‘Other animate objects’ are presented in Table 3. The two content groups were inversely related to one another. Content of unfamiliar people was also related to hallucinations being bothersome while content with other animate objects was related to treatment-seeking. The large CI indicated that some of these correlations, although significant, may be unreliably quantified.

Table 3. Covariates of significant relevance with visual hallucination content
Hallucination contentCovariatesWald χPOdds ratio95%CI
  1. Overall significance for reduced model for the first 3 covariates was P = 0.001. For the last 2 factors, it was P = 0.008.
  2. CI, confidence interval; VH, visual hallucination.
Unfamiliar peopleVH bothersome4.8170.02824.9081.411–439.829
VH content: Other animate objects5.0960.0240.0310.001–0.632
Other animate objectsVH treatment wanted4.5920.0326.8811.179–40.162
VH content: Unfamiliar people3.7120.0540.1340.017–1.305

None of the observed covariates were significantly relevant to hallucinations in the other two categories. Due to few individuals experiencing either type of content, any results obtained from such analysis would be questionable, whether they showed significance or not.

Response to VH

Table 4 shows the results of analysis of significant predictive covariates of bothersome VH and treatment-seeking. While the two were strongly related to one another, they each had other unshared relevant factors. Each related well to only one of the two most populated content groups. Treatment-seeking was additionally related to the number of different types of content encountered by patients, however it was only nearly significant and only apparent in an analysis completely isolated from other significant and non-significant factors. Patients were offered quetiapine for treatment in all cases as it posed lower risk of worsening motor symptoms. Although significance or near-significance was detected for these correlations, the wide CI suggest lack of reliability in the given exact OR.

Table 4. Covariates of significant relevance with VH being bothersome and covariates leading to treatment-seeking behavior
 CovariatesWald χPOdds ratio95%CI
  1. aOverall significance for reduced model, including these 2 covariates was P < 0.001.
  2. bOverall significance for reduced model, including these 2 covariates was P = 0.001.
  3. cOverall significance for reduced model, including this covariate only was P = 0.003.
  4. CI, confidence interval; VH, visual hallucination.
BothersomeVH treatment wanteda7.7610.00527.1352.661–276.722
VH content: Unfamiliar peoplea5.7320.01723.9721.779–323.017
Treatment-seekingVH bothersomeb7.0900.00811.7631.917–72.196
VH content: Other animate objectsb3.9640.0466.7621.030–44.389
VH diversityc3.6890.0558.3360.958–72.563

A subset of five individuals who found their hallucinations bothersome also expressed being threatened or frightened. This was related to the diversity of hallucinations encountered (P = 0.036), individuals being 2.445 (95%CI: 1.060–5.640) times as likely to feel threatened for each additional type of content encountered. We acknowledge, however, that these findings may not be entirely reliable due to the small sample. These were the only individuals (n = 5) who were judged to have lost insight and given a rating of three on the PPRS. Remaining patients with hallucinations were rated two out of four.

Auditory hallucinations

Only three individuals endorsed auditory hallucinations. None had any other mode of hallucinations besides. A single patient, hearing voices that would comment on her thoughts and activities, expressed being bothered by the occurrences. The others respectively noted hearing voices of a deceased family member and hearing household noises, such as the doorbell or telephone ringing, but neither mentioned any distress.

Multimodal hallucinations

Another three individuals in the patient sample reported having multimodal hallucinations. All had VH accompanied by somatic hallucinations. One individual reported feeling insects on his skin, another felt that there was someone else in bed, and the third experienced her arms becoming longer.


VH prevalence

Previous studies have reported varying values of prevalence of VH in PD, spanning from about 6% to 60%.[3] Our prevalence rate falls within the lower end of this range, likely due to lower H&Y stages of the sample. Those with advanced PD are normally placed in long-term care institutions and less often visit outpatient settings.

Influences on VH presence

Our analysis indicated that sex, PD stage, and concurrent dementia were predictive of VH. Women, individuals in stages 3 and 4, and those with cognitive impairments were significantly more likely to have VH.

Although questionable, the findings do encourage studying sex differences among PD patients with hallucinations, of which there is little research to our knowledge. One study found the prevalence to be equal in both sexes, but the sample was limited to patients in nursing homes.[19] Considering that sex differences in PD have been noted in several non-motor symptoms of the disease, such as drooling,[20] depression, cognition (also observed in our sample), verbal fluency, and visuospatial cognition, among others,[21] it may be worthwhile to explore this among those who hallucinate.

Previous studies have reported age to be of significance to hallucinations in PD,[2, 22] however, it is suspected that the more relevant factor is the duration of illness.[2, 4] This may in fact contribute to the relevance of H&Y stage, identified to be the most significant predictor of VH in our sample. Although this correlation has been reported[2] and the symptom now serves as a milestone for advanced disease,[23] it remains unclear why hallucination occurrence would relate to the motor symptoms on which the scale is heavily based.[24] Perhaps it is an effect of the widespread neural deterioration in advanced disease. Nevertheless, it appears that at stage 3 or somewhere between stages 2 and 3 lays a threshold after which the likelihood of hallucination onset increases dramatically and remains consistent with further disease progression.

Content of VH

As expected from previous research,[12] most individuals had stereotyped hallucinations and encountered only one type of content. In agreement with a thorough analysis of VH content by Barnes and Davis,[2] the most common entity in hallucinations was unfamiliar people, followed by animate objects, such as animals and insects. Also, as we expected, a fewer number of individuals saw familiar people and other inanimate objects not fitting any category.

The inverse correlation between the two most common types of VH content may support the tendency of having stereotyped hallucinations. However, it is not apparent what gave rise to these two groups with two distinct types of content.

It may as well be reasoned that content with unfamiliar people and other animate objects leads, respectively, to hallucinations being bothersome and to desiring treatment for hallucinations. It is understandable that most individuals, although not all, would respond negatively to such unwanted experiences.

Factors affecting severity and perception of VH

Finding hallucinations bothersome was strongly related to wanting treatment to ease or be rid of the symptom. However, the correlation was not perfect, in that some patients openly expressed negativity towards the experience and yet did not request or refused treatment.

A portion of the disparity between bothersome hallucinations and treatment-seeking appears to be due to the content of hallucinations. While seeing unfamiliar people was often reported as bothersome, there were a substantial number of individuals who experienced this and yet did not take treatment, resulting in poor correlation between this content category and treatment-seeking. Conversely, hallucinations of other animate objects often resulted in requesting treatment, despite a noticeable number of these individuals expressing no botheration.

A greater diversity in the content experienced, which was itself related to hallucinations perceived to be threatening on top of bothersome, also appeared to be somewhat related to patients wanting treatment. While most individuals reported only one type of content, it seemed that those who tended to have a variety of content in their hallucinations were more likely to feel threatened and lose insight. This could indicate that more negative emotional responses and loss of insight occur with varied content. However, the small sample size is again a limitation.

The variability in patients finding their hallucinations bothersome and seeking treatment may be due to different emotional perception thresholds, ways of expressing their sentiments, and opinions on taking medications and their consequent effects. Evidently, patients' choices of wanting treatment were not always logically consistent. Analysis did indicate that the type of content encountered had substantial impact on the discrepancy between hallucinations being bothersome and treatment-seeking; however it failed to explain why the content would relate in this manner. It may be that some element of the content experienced gives rise to counterintuitive emotional response and decision-making. Alternatively, the content may be influenced by emotional predisposition. Overall, treatment-seeking appears dependent on whether or not hallucinations are bothersome, although for some individuals the content will dictate this decision more directly than emotional perception of the content.

Equally interestingly, many of the analyzed variables had no significance. Sex, age, PD stage, and dementia presence within the VH group had no bearing on the type of content, whether they could predict VH to be bothersome or if treatment was desired. It seems then that the significant factors for hallucination presence define only a threshold. The content and perception of content seems to largely remain the same after onset, although this needs to be confirmed longitudinally in all individuals, and these are the factors that predict treatment-seeking. The decision to initiate treatment should not be at all based on external measures, but rather solely on the patients' perception and opinion.

Hallucinations of other modalities

There were very few individuals with hallucinations of other modalities, insufficiently so to merit analysis. However, we interestingly observed that none of the individuals with auditory hallucination had VH. This is not very common, as most cases of auditory hallucinations present in addition to VH.[11] A few individuals had somatic hallucinations alongside VH. Previous research has suggested that duration of disease results in hallucinations of more modalities,[11] but this could not be confirmed.


Several characteristics commonly noted for PD, such as PD stage, and perhaps sex to an extent, serve as predictors for presence of hallucination based on regression analyses. However, these factors only indicate a threshold and make no contribution towards the type of content, severity, or patient reaction to the hallucinations. Treatment-seeking is largely based on whether patients find the hallucinations bothersome; however, the content may sometimes be the direct determinant of this and contradict patients' reports of the hallucinations being bothersome or not. Although the decision to provide treatment is of great interest to clinicians, a patient's choice regarding this is not always predictable or logically consistent, suggesting that such judgment should be based on the patient's own preference in addition to their recollection and emotional response.


We thank Muhammad Husain Cader, Haris Vaid, Mujtaba Morshed and Raza Akhter for their extensive contribution to data collection and formatting for this study. The authors have no conflict of interest to declare.