Statistics from Altmetric.com
Cohort study, with 4–10 years of follow up, in which the Israeli Draft Board Registry was linked with the National Psychiatric Hospitalization Registry.
Boys who were 16–17 years of age and had pre-induction assessments (1985–91) to determine their eligibility for military service. Exclusion criteria were detectable signs or symptoms of mental illness or mental retardation, admission to a psychiatric hospital before or up to 1 year after the assessment, and incomplete test data. 509 boys were admitted and 9215 boys were not admitted for schizophrenia to a psychiatric hospital during follow up.
Assessment of risk factors
Data from the draft board assessment were used. General intelligence (ie, arithmetic, verbal abstraction and categorisation, non-verbal abstract reasoning and problem solving, and ability to understand and do verbal instructions) was assessed by a cognitive test battery. Behaviour (ie, social functioning, individual autonomy, organisational ability, and physical activity) was assessed in a structured interview.
Main outcome measure
Psychiatric hospital admission for schizophrenia diagnosed with the ICD-9.
When cases were matched to control participants according to the high school attended at the time of assessment, people in the schizophrenia group had lower scores for behavioural and intellectual functioning than those in the control group. The most pronounced differences existed for social functioning, for which the cut off point of the 2 lowest quintiles accurately predicted 43% of people in the schizophrenia group and 93% of those in the control group. In a conditional logistic regression model, the strongest predictors of hospital admission for schizophrenia 4–10 years later were poor social functioning (p<0.001), poor organisational ability (p<0.001), and low intellectual functioning (p=0.006) (table⇓). The results were not influenced by the time lag between testing and first hospital admission.
Poor social functioning, poor organisational ability, and low intellectual functioning in Israeli adolescent boys predicted psychiatric hospital admission for schizophrenia 4–10 years later.
This remarkable study by Davidson et al adds to the life course view of schizophrenia. Can we now predict who will develop it?
Prediction is an imperfect art even when genetic risk is known or elegant analyses of unique data such as these are presented. The matching within individual school classes meant that the main sample was drawn from classes in which 1 boy would inevitably develop schizophrenia. This was excellent for isolating the behavioural and intellectual precursors of schizophrenia, but shortened the odds for prediction. In practice, such classes cannot be defined, a priori, and teachers know that boys at the bottom of the majority of classes will not get ill. The lowest score (1) in social functioning here, however, represented behaviour so poor or bizarre that such boys were really akin to a recognisable but rare morbid or clinic population that needed help. In this group, schizophrenia was extremely likely. About 10% of all schizophrenia could be predicted from them, which is a worthwhile target. The authors are considering how their model can be improved and extended.
Moving from the computer room to the population is a giant leap. Large scale, population based clinical programmes to detect pre-psychotic individuals are, we feel, not yet justified. Small changes in a predictive model can amount to massive shifts into and out of the false positive and negative categories. People in the false positive category are particularly vulnerable to the costs rather than the benefits of interventions, however benign these may seem. On the other hand, interventions may be helpful for other adverse outcomes in high risk groups. Studies such as this begin to add evidence to the growing debate over detection and intervention at the pre-psychotic (rather than the early or peri-psychotic), high risk stage. Other factors concern the costs, effects, and alternative targets of any interventions. All empirical signposts that shed light rather than heat are welcome. This study adds science to the developing art of prediction and ushers in the question not can we, but should we?
We are grateful to the authors for helpful and enthusiastic discussion of these comments.
Source of funding: not stated.
For correspondence: Dr M Davidson, Chaim Sheba Medical Center, Tel-Hashomer 52621, Israel. Fax +972 3 534 5920.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.