The risk of a chronic clinical condition following a previous hospitalisation by a psychiatric disorder: a linkage nationwide study in Brazil

Multimorbidity is the occurrence of more than one chronic condition at a time. This can include conditions like cardiovascular disease, a mental health condition of long duration, such as a mood disorder or dementia or an infectious disease of long duration, such as Tuberculosis.

Multimorbidity increases with age and also presents certain distinct patterns related to sex. On the other hand, mental disorders are commonly found together and usually have consequences on other chronic physical condition that the patient may have. Mental health disorders, in particular depression, represent an important risk factor for premature mortality. Multimorbidity can also increase the treatment costs, as subjects with depression will cost 50% more to health systems than those with one chronic condition alone. Cardiovascular diseases are also found to co-occur with depression, this might happen because these patients show difficulties in following their GP 's recommendations. It is of note that women seem to be at higher risk of morbidity from cardiovascular disease, while they are also at a higher risk of depression suggesting gender-specific mechanisms. Analysis of a recent health survey in adult Brazilians observed that the prevalence rate of multimorbidity are higher among women compared to men, among older people and those with lower educational level.

We at the Center for Data Integration and Knowledge for Health (CIDACS) have access to very large datasets, which would commonly be known as Big Data. This data comprises information from Brazilian subjects that were listed or benefited by one of the several social programmes or were seen in any of the numerous public hospital of the country's universal health system.

This project aims at increasing knowledge over the relationship of mental disorders and other chronic conditions to ameliorate the lives of those affected. More specifically we want to (A) estimate the the risk of hospitalisations or death by diabetes mellitus, cardiovascular diseases or stroke following a hospitalisation due to depressive disorders, alcohol and substance use-related disorders, and schizophrenia; (B) To estimate the risk of the occurrence or death by tuberculosis following a hospitalisation due to depressive disorders, alcohol and substance use-related disorders, and schizophrenia; (C) To investigate how these chronic conditions goes together in clusters and how these patterns evolve over time and ageing.

This project will analyse electronic data routinely collected within the Brazilian Public Health System (SUS). It involves enriching this type of data by linking them to other Brazilian governmental databases destined to national social protection programmes, which have important socio-economic variables of the individuals. A very large dataset with over 114 million people (The 100 Million Brazilian Cohort), which more than half of the Brazilian population, will be consolidated to the study of multimorbidity and its socio-economic determinants nationwide. Therefore, establishing foundations on the magnitude of the problem, in particular on the most impoverished population and for further studies on this subject in Brazil and other similar middle-income countries.

We aim to study the risk of hospitalisation or death by diabetes mellitus, cardiovascular disease, stroke, or tuberculosis associated with previous hospitalisation by the following psychiatric disorders: depression, alcohol and substance use-related, and schizophrenia. We also intend to identify disease clusters and their related patterns, and how these patterns interact over time to influence the formation of such clusters. "Big data" is seen as having great potential to answer numerous questions and CIDACS collection of Brazilian data is unique in low-middle income countries. Our access to this collection of data allows for unprecedented study of morbidity and mortality in a nationwide scope.

Multimorbidity is a concerning and growing problem. It affects subjects from their younger ages to elders, although elders are known to present more commonly with more than one chronic condition. It can affect genders distinctly, as women seem to be at higher risk of multimorbidity from cardiovascular disease and depression. This project will use very large data sets comprised of information of subjects who accessed one of the numerous social programmes in Brazil or were seen in any of the numerous public hospital of the country's universal health system. The uniqueness of the data at hand, with its nationwide nature, ensures large impact. Products of this project will be not only of relevance locally in Brazil, but also internationally, as patterns of multimorbidity will be generalisable to other upper middle income countries and to other countries that also have a Universal Health System like the UK.

More importantly, patients can potentially and immediately benefit from this work, as their doctors, exposed to our findings, will have information for better, more precise, directives on actions regarding the presence of mental disorders in subjects who also have chronic conditions. Communication arising from this project will also allow the public to be more aware about their medical conditions and search medical assistance more often. Medical and non-medical staff members are also potential beneficiaries, as findings from this project can help improve treatment and early detection of crucial interactions between chronic conditions.

The use of nationwide information on hospitalisation will also permit better policy making. The government will be able to address regional differences highlighted in our work and provide more precise action on health, where these are needed. This is of particular importance in any large and heterogeneous country like Brazil. We will analyse our data at the municipality level, so that mayors of these municipalities will have access to information tailored for their particular area.

The United Kingdom will also benefit from this research, as we anticipate that some of our results will be generalisable. Brazil is a very heterogeneous country and results from the richer areas can be generalisable to other countries. Furthermore, the computational (machine and deep learning) models to be deployed in this project will be reported and released in public domain, so they can be immediately accessed and used for similar research in other settings and scenarios.