top of page



Analysis of Duration – I amalgamated the datasets on the different delusions that I have collected and after cleaning the data, I investigated on an exploratory basis, various finding for duration of delusions, as well as for those cases with delusions present for less than 3 months and less than one month -

Table showing the duration of delusions for the different binary variables in years. Information shows the median and maximum values for the total dataset, for cases with duration less than 3 months and for cases with duration less than one month.


Obviously, as we consider smaller timeframes the median durations decrease. If we consider the whole dataset there appears to be a greater variation in median values. There is a negative correlation between age and duration for the whole dataset (corr=-0.002) and for delusions that last less than 3 months (corr =-0.037)


As I coded the variables, when information on such variables were not reported, I coded the variable as ‘0’, to indicate the absence of the variable. This left missing values of gender, 50 in total, which accounted for 1.9% of the total. As females accounted for 58.75%, I placed all the missing gender values into the male level.


Because the log of the duration is a continuous variable I used t-tests to compare the differences across the levels of the other variables (13 binary), so the corrected p-value used is 0.004. There was a trend for longer delusional duration in those with a psychiatric history (1.7 v 1.2, p=0.0044, df=818). Duration of delusions were longer in those with other delusions (1.8 v 1.2, p=0.0017, t=818) and in those who were not on psychiatric treatment (1.7 v 1.0, p=0.0001, t=818). There is a small negative correlation between age and duration (-0.015)


When we look at the duration of delusions for each of the different delusion type -

Picture 1.png

Duration of misidentification delusions seems to be shorter than the others, whilst the others appear to have roughly the same duration. On running simple linear regressions with the log of the duration as the dependent variable (bear in mind, I have not yet imputed missing values), duration rose by an average of 0.7 years in those with a medical history (b=0.7, p=0.006), by 1.4 years in those with a psychiatric history (b=1.4, p=0.004), by 1.5 years in those with siblings (b=1.5, p=0.042), by 1.5 years in those with other delusions and by an average of 0.6 years in those who responded to treatment.


Now as the effects of evolution act, in part, via age and gender, I created a model with the log of the duration as the independent variable, with dependent variables being age, gender, delusion type, psychiatric history, presence of other delusions and response to treatment. This regression model did not produce any interactions that were significant. This result was the same when I added the variables gender and age to the analysis. This shows that the average duration does not vary with the levels of the categorical variables or with age.


That left missing observations for age and duration, and these were imputed using STATA multiple imputation commands. Now to ensure I could use linear regression, I generated the log of duration data, which was normally distributed. After this I ran simple linear regressions using the mi estimate command. The only variables that were significantly related to the log of the duration were (results have been exponentiated)

  • Psychiatric history                coef = 1.29, SE = 0.13, 95% CI 1.01 – 1.67

  • Other delusions                   coef = 1.72, SE = 0.12, 95% CI 1.35 – 2.20

  • Response to treatment         coef = 0.68, SE = 0.13, 95% CI 0.53 – 0.87

  • Folie delusions                     coef = 1.48, SE = 0.16, 95% CI 1.07 – 2.03

  • Somatic delusions                coef = 1.88, SE = 0.18, 95% CI 1.32 – 2.68

  • Erotomania                         coef = 2.17, SE = 0.25, 95% CI 1.32 – 3.57

  • Jealousy                              coef = 2.42, SE = 0.27, 95% CI 1.35 – 4.33

  • Religious delusions               coef = 2.60, SE = 0.35, 95% CI 1.26 – 5.38

For the whole dataset, delusions tend to be longer if there is previous psychiatric history and if patients have other delusions. All the other delusional types seem to last longer than misidentification delusions.


If we look at the relationship between duration and age, we get the following graph –




This seems to show a quadratic relationship, with the red line indicating the quadratic fitted values. When we do the same for the different delusional types, we get -





Which shows a similar pattern for some delusional types.


Looking at the whole dataset where information on duration was available, we can see that the values of the distribution follow an exponential pattern. Hence the log of this was used to ensure a greater normality to the data. Having delusions that last longer in those with a psychiatric history or other delusions make sense – there is more psychopathology going on. As is it being longer on those without treatment. One finding that takes some explaining is that delusions were longer if cases had siblings. Another interesting finding is that misidentification delusions were shorter. This may be because they are so far from normal experience, that they attract attention far earlier and hence get treated. The lack of interactions seems interesting considering the research literature, but the dataset is probably too small and doesn’t reflect the information needed. After correcting for missing values there seems to longer delusions in those with delusions of erotomania, jealousy, religious and somatic. Which is in keeping with the other findings.


When I looked at the data where delusions lasted less than 3 months, on one t-test the only variable related to duration was the presence of a medical history. When it came to correlations and interactions, none were significantly related to the log of the duration. Age and gender were not related to delusional duration.



The corresponding results for duration delusions of less than 3 months are (again values are exponentiated)


  • Medical history                    coef = 0.64, SE = 0.21, 95% CI 0.42 – 0.98

  • Other delusions                   coef = 0.62, SE = 0.23, 95% CI 0.40 – 0.97

  • Folie delusions                     coef = 2.47, SE = 0.35, 95% CI 1.24 – 4.95

  • Somatic delusions                coef = 1.87, SE = 0.25, 95% CI 1.14 – 3.07

  • Erotomania                         coef = 2.51, SE = 2.24, 95% CI 1.11 – 5.67


Discussion – From an evolution point of view there are a few interesting points. Duration seems to rise slightly with age, but it seems that duration of delusions is highest in midlife but lower at the other ends of the life span. This pattern seems to be replicated across the different delusional types. From an evolutionary point of view having long lasting somatic delusions may lead to such people being the recipient of help. Longer durations of erotomania may help you get a partner, whilst being jealous for longer may help you keep such a partner. Longer lasting religious delusions, well this depends on the function that religion has on people. Having other delusions than the main one is linked to longer delusional duration – this may mean that multiple evolutionary pressures are operating and hence there may be a greater benefit.


Moving away from an evolutionary view, it makes sense that having a psychiatric history is linked to longer durations, as many are not insightful into their condition and hence do not seek treatment, or it is not recognized. Those that respond to treatment have shorter durations and this makes sense. The shorter duration of delusions in the young may simply reflect the fact that patients have not been alive long enough. Though in the elderly have shorter duration of delusions, may reflect different pathological processes, which start later in life.


When we consider results for durations of less than three months, i.e., a picture that may resemble acute and transient psychosis we see a lot of the same findings. Though this time having other delusions is associated with reduced duration of delusions. Maybe this results in a more dramatic picture that results in an increased likelihood of treatment. Now why would having physical disorders result in reduced duration of delusions? Are such people more likely to receive mental health treatment? Would they be more likely to see a professional and hence problems are picked up before they become an issue? This makes more sense than saying physical disorders are protective of delusions, as the literature is full of reports of delusions being linked with physical disorders.


Limitations here include the relatively small numbers of people that have delusions of less than three months (n=105) and even less if we consider one month (n=64). Some have said that a six-month cut off may define brief psychotic disorder, and this may be worth a look. The case reports often do not mention issues with stress, the presence of life events which limits analysis.

Picture 2.png
Picture 3.png
bottom of page