Skip to content
COSMOHS logo

Method

Data and methods

Which data are used, how they are linked, and how they are analysed.

Study population

The study includes everyone born between 1930 and 2006 who was registered as a resident of Sweden at some point from 1982 onwards — approximately 9 million people — identified from the Swedish Total Population Register. The end year is updated at each new register linkage. Everyone diagnosed with HIV in Sweden is identified from InfCareHIV. The date of HIV diagnosis is defined as the date of the first positive HIV test in Sweden, or the first registered positive HIV-RNA result, whichever came first.

InfCareHIV

InfCareHIV is the Swedish national HIV quality register. Data from the register may be used for research where ethical approval has been granted. It was established in 2003 and has had national coverage since 2008; earlier data, including data on people who have since died, have been added retrospectively.

Every clinical centre (21) caring for people with HIV in Sweden reports to InfCareHIV, and more than 99 % of everyone diagnosed and living in Sweden is included. Around 8,500 people currently living with HIV are registered, and around 13,000 in total including those who have died or emigrated. Coverage has been validated repeatedly against HIV diagnoses reported to the Public Health Agency of Sweden.

The register holds demographic data, virological data, CD4 and CD8 cell counts, antiretroviral treatment, and health-related quality of life measured through a health questionnaire. Socioeconomic data such as income, education and civil status are not held in InfCareHIV — they come from the registers below, which is part of why the linkage matters.

Linked registers

Statistics Sweden

Visit ↗
  • Total Population Register (TPR) — births, deaths, immigration, emigration, internal migration, marriage and divorce, country of birth, sex and citizenship.
  • Longitudinal integrated database for health insurance and labour market studies (LISA) — education, employment, income and social benefits.

The National Board of Health and Welfare

Visit ↗
  • The National Patient Register — inpatient and specialised outpatient care.
  • The Swedish Cancer Register — all cancer diagnoses.
  • The Prescribed Drug Register — dispensed prescriptions.
  • The Cause of Death Register — underlying and contributing causes of death.

National quality registers

Linkage and pseudonymisation

The study population is identified from the Total Population Register at Statistics Sweden. Linkage using personal identity numbers is carried out by the National Board of Health and Welfare in contact with the respective registers. Only pseudonymised files (anonymised for the recipient), identified by index numbers, are returned to the researchers. The key linking index numbers to personal identity numbers is held by Statistics Sweden, which enables the cohort to be updated every two to three years.

Analysis

Analysis begins with a data integration and preprocessing phase — cleaning the data, handling missing values, outliers and inconsistencies. Statistical methods include descriptive statistics and tests of difference in distribution, survival analysis for time-to-event data, and logistic and Cox regression, with adjustment for potential confounders such as underlying comorbidity and socioeconomic position. Questionnaire data (PROM and PREM) are analysed using regression to examine the association between stigma, treatment results and health. Machine learning methods are used for prediction modelling.

Because COSMOHS analyses a whole population rather than a sample, statistical power is less central than in a conventional study. As an illustration of scale, the study is able to detect a hazard ratio of 1.5 for cardiovascular disease between people starting treatment with CD4 counts below 500 and at or above 500 cells/µL, with power above 80 %.

Strengths

  • Access to individual-level data on the entire Swedish adult population, with more than 40 years of follow-up.
  • A heterogeneous HIV cohort — a large proportion of migrants and of women, diverse routes of HIV acquisition, and a range of socioeconomic backgrounds — allowing analysis of subpopulations that are often invisible in HIV research.
  • The ability to control for sociodemographic differences, comorbidity and HIV-specific cofactors, and so to assess the impact of HIV infection itself more precisely.
  • Patient involvement through a diverse reference group.

Limitations

  • Data on smoking are missing for most of the HIV-negative population and for part of the population living with HIV. Prescriptions for chronic obstructive pulmonary disease and other smoking-related conditions will be used as a proxy for smoking history in sensitivity analyses.
  • Data on lifestyle factors are limited to what the linked quality registers happen to record.