Health and Care Data: Approaches to data linkage for evidence-informed policy.

An indispensable prerequisite for answering research questions in health services research is the availability and accessibility of comprehensive, high-quality data. It can be assumed that health services research in the coming years will be increasingly based on data linkage, i.e., the linking, or connecting, of several data sources based on suitable common key variables. A range of approaches to data collection, storage, linkage and availability exists across countries, particularly for secondary research purposes (i.e., the use of data initially collected for other purposes), such as health systems research. The main goal of this review is to develop an overview of, and gain insights into, current approaches to linking data sources in the context of health services research, with the view to inform policy, based on existing practices in high-income countries in Europe and beyond. In doing so, another objective is to provide lessons for countries looking for possible or alternative approaches to data linkage. Thirteen country case studies of data linkage approaches were selected and analysed. Rather than being comprehensive, this review aimed to identify varied and potentially useful case studies to showcase different approaches to data linkage worldwide. A conceptual framework was developed to guide the selection and description of case studies. Information was first identified and collected from publicly available sources and a profile was then created for each country and each case study; these profiles were forwarded to appropriate country experts for validation and completion. The report presents an overview of the included countries and their case studies (Chapter 2), with key data per country and case study in the appendices. This is followed by a closer look at the possibilities of using routine data (Chapter 3); the different approaches to linkage (Chapter 4); the different access routes for researchers (Chapter 5); the use of data for research from electronic patient or health records (Chapter 6); foundational considerations related to data safety, privacy and governance (Chapter 7); recent developments in cross-border data sharing and the European Health Data Space (Chapter 8); and considerations of changes and responses catalysed by the COVID-19 pandemic as related to the generation and secondary use of data (Chapter 9). The review ends with overall conclusions on the necessary characteristics of data to inform research relevant for policy and highlights some insights to inspire possible future solutions - less or more disruptive - for countries looking to expand their use of data (Chapter 10). It emphasises that investing in data linkage for secondary use will not only contribute to the strengthening of national health systems, but also promote international cooperation and contribute to the international visibility of scientific excellence.

Health and care data: approaches to data linkage for evidence-informed policy

An indispensable prerequisite for answering research questions in health services research is the availability and accessibility of comprehensive, high quality data. It can be assumed that health services research in the comingyears will be increasingly based on data linkage, i.e., the linking, or connecting, of several data sources based on suitable common key variables. A range of approaches to data collection, storage, linkage and availability exists across countries, particularly for secondary research purposes (i.e., the use of data initially collected for other purposes), such as health systems research. The main goal of this review is to develop an overview of, and gain insights into, current approaches to linking data sources in the context of health services research, with the view to inform policy, based on existing practices in high-income countries in Europe and beyond. In doing so, another objective is to provide lessons for countries looking for possible or alternative approaches to data linkage. Thirteen country case studies of data linkage approaches were selected and analyzed. Rather than being comprehensive, this review aimed to identify varied and potentially useful case studies to showcase different approaches to data linkage worldwide. A conceptual framework was developed to guide the selection and description of case studies. Information was first identified and collected from publicly available sources and a profile was then created for each country and each case study; these profiles were forwarded to appropriate country experts for validation and completion.

Trajectories in hypnotic use and approaching death: a register linked case-control study.

PURPOSE: Whether the association between hypnotic and increased mortality risk is created by causation or confounding, has been long debated. We further examined the possibility of confounding by indication with a comprehensive approach. METHODS: The National FINRISK Study cohorts of 1997, 2002, and 2007 (25,436 participants aged 25-74) were followed up until July 2012. There were 1822 deaths, and at least one gender, baseline age and cohort matched 'control' was found for 1728 'cases' yielding a final analytical sample of 3955 individuals. An index age, equivalent to the age at death of their respective cases' was set for each control. Hypnotic drug purchases were followed from the Finnish nationwide register during a 36-month run-up period before the date of death/index date. The prevalence and incidence of hypnotic purchases were compared between cases and matched controls. Moreover, latent developmental trajectories of purchases were modelled and their relations with specific and all-cause death risks were analysed. RESULTS: An increasing difference between cases and controls was observed as regards the use of hypnotic drugs. During the last 30 months before the date of death/index date, the rate ratio of incident purchases between cases and controls was 2.37 (95% CL, 1.79-3.12) among older and 3.61 (95% CL, 2.37-5.89) among younger individuals. The developmental trajectories of hypnotic drug purchases were differently and by interpretation plausibly associated with specific mortality risks. CONCLUSIONS: In most cases the association between hypnotics and mortality risk is created by symptomatic treatment when death is approaching.

Association Between Unstable Work and Occupational Wellbeing Among Artists in Finland: Results of a Psychosocial Survey.

OBJECTIVE: Although artistic work is in transition, the occupational wellbeing of artists has been less studied than wellbeing among other workers. This study aimed to explore the relationship between work characteristics and occupational (psychosocial) wellbeing of artists. METHODS: A national questionnaire was sent to all artists (theatre artists, writers, and visual artists) reached by four major labor unions in Finland. Type of employment (permanent full-time work vs other), working field (own field of art vs other), regularity of working hours (regular vs irregular), and control of workload were assessed. The wellbeing outcomes were work engagement, recovery from work, and experience of stress and low mood. RESULTS: Full-time permanent employment, regular working hours, and working in one's own field of art were positively associated with work engagement. Furthermore, regular working hours were positively associated with recovery and negatively associated with subjective report of low mood. Ability to control workload was positively associated with recovery and negatively associated with stress and low mood. Higher age was associated with lower stress and better recovery. CONCLUSIONS: Artists with regular working hours, secure employment, ability to control workload, working in one's own field of art, and higher age reported better wellbeing in this study. The late stages of career appear to guarantee more stability and wellbeing than the more insecure beginning of a career.

Use of antidepressants among 0-26 year olds in Finland during 1997-2007

Background and Objectives: The use of antidepressants has increased in many Western countries among children, adolescents and young adults. The objective of this study was to analyse changes in incidence and prevalence of antidepressant use among patients aged 0 to 26 years from year 1997 to 2007. Methods: This is an observational population-based drug utilization study using a nationwide prescription register including data on reimbursed purchases of antidepressants for outpatient use. Results: The incidence and prevalence of antidepressant use increased about three foldduring the study period. Among children less than 14 years of age both incidence and prevalence were low during the study period. In the age group 14-26 years, the incidence increased from 6.4 to 16.1 (per 1000 person-years) among men and from 9.7 to 28.3 among women, while the prevalence went up from 10.4 to 30.4 among men and from 15.5 to 56.8 among women. By the age of 26 years, 11.6% of men and 17.8% of women had purchased antidepressants at least once. The first antidepressant prescriptions were mainly prescribed by a specialist for children aged 7 to 15 years and by a non-specialist for the older age groups Conclusions: Use of antidepressants has increased remarkably among adolescents and young adults during recent years. Further studies are warranted to analyse whether the increase in use is due to an increase in depression incidence, a lower threshold of prescribing medications or a lack of psychosocial treatment resources (AU)

Long-term excess mortality in 623 patients with brain arteriovenous malformations.

OBJECTIVE: Long-term follow-up studies in patients with brain arteriovenous malformations (AVM) have been scarce and without proper statistical estimates of mortality. We performed a retrospective survival study in 623 consecutive patients with AVMs admitted to the Department of Neurosurgery in Helsinki University Hospital between 1951 and 2005. METHODS: Patients were followed from admission until death or the end of 2005. Patient survival was estimated using the relative survival ratio, which provides a measure of the excess mortality experienced by the patients compared with the general Finnish population matched by age, sex, and calendar time. RESULTS: Median follow-up was 11.9 years, and total follow-up was 10,165 person-years. Treatment was conservative in 155 patients. Total AVM occlusion was attained in 356 patients, and partial occlusion was obtained in 94 patients. Overall, 206 deaths were observed. Of these, 100 were related to AVMs. Diagnosis of AVM was associated with significant long-term excess mortality, with cumulative relative survival ratios of 0.85 (95% confidence interval, 0.81-0.88) and 0.69 (95% confidence interval, 0.62-0.75) at 10 and 30 years after admission, respectively. Men had higher excess mortality than women. The excess in mortality was highest in conservatively treated patients, intermediate in patients with partially occluded AVMs, and lowest in those with totally occluded AVMs. The subgroup with the best outcome consisted of those with totally occluded unruptured AVMs, which did not demonstrate excess mortality after the first year. CONCLUSION: AVMs are associated with long-term excess mortality that may be reduced by active, even partial, treatment. Male patients have a higher excess mortality rate than female patients.

Comparison of cancer registry and clinical data as predictors for breast cancer survival.

BACKGROUND: In spite of the increasing amount of clinically relevant information for survival from breast cancer, the amount of data recorded in a population-based cancer registry is limited and the registry-based survival predictions are routinely made without clinical details. OBJECTIVE: To find out how important is the role of screening and clinical tumor characteristics in addition to cancer registry information in describing the breast cancer survival. METHODS: A representative clinical database on 483 breast cancer patients diagnosed during 1996-1997 in Tampere University Hospital Area was linked with Finnish Cancer Registry data and a survival model including the available registry variables was compared to models including screen-detection information and clinical tumor characteristics also. RESULTS AND CONCLUSION: Estimates of registry stage and age act as surrogates for clinical variables and mammography-detection. The surrogacy was found to be almost complete indicating that clinical variables are not necessarily needed when making breast cancer mortality predictions based on a population-based cancer registry. In cases with dissimilar staging cancer registry stage gave a better picture of the breast cancer survival than the clinical stage.

Predicting impacts of mass-screening policy changes on breast cancer mortality.

The aim of this study is to present a methodology for taking into account the mass-screening invitation data in breast cancer mortality predictions, particularly in assessing impacts of screening policy changes on the short-term predictions. The methodology is applied to a database that includes observed year- and age-specific screening invitation schemes in Finnish municipalities from the time period 1987-2001. The target year for predictions is 2012. To predict mortality, breast cancer incidence and patients' survival from breast cancer are modelled with the screening data included. The knowledge of breast cancer survival together with the other cause survival is then used to calculate the number of breast cancer deaths caused by observed (1987-2001) and predicted (2002-2012) incident cases in Finland. Survival from breast cancer was estimated with a parametric mixture model where the patient population is assumed to be a combination of cured and uncured patients. This approach provides a way of modelling the hazard of fatal cases and the proportion of cured cases simultaneously. In other cause survival, the patients' hazard was allowed to differ from that of the general population. Breast cancer mortality predictions are presented according to three alternative future scenarios of screening policy. The results show no major differences between predictions yielded by alternative scenarios: Any policy change would have at the most a 3.0 per cent impact on breast cancer mortality in the near future.

No long-term excess mortality in 280 patients with ruptured distal anterior cerebral artery aneurysms.

OBJECTIVE: The aim of this study was to assess the long-term excess mortality after the rupture of distal anterior cerebral artery (DACA) aneurysms compared with that of a matched general Finnish population in an unselected, population-based series. METHODS: We identified 280 consecutive patients (119 men, 161 women) treated for ruptured DACA aneurysms (clipped, 262; coiled, 10; no intervention, 8) at two neurosurgical centers serving solely the southern and eastern parts of Finland from 1976 to 2003. All patients were followed from subarachnoid hemorrhage until death or the end of 2004. No patients were lost to follow-up. Long-term excess mortality was estimated using the annual relative survival ratio compared with the general Finnish population matched by age, sex, and calendar time. RESULTS: The median follow-up period was 9.6 years (range, 0.1-29 yr). The 3-year cumulative relative survival ratio was 0.84 (95% confidence interval, 0.78-0.88), implying 16% excess mortality in the patient group during the first 3 years after subarachnoid hemorrhage. The annual relative survival ratio attained 1.0 at the fourth year of follow-up, indicating no excess mortality thereafter. There were four episodes of recurrent subarachnoid hemorrhage and only one from a treated DACA aneurysm, with a 10-year cumulative risk of 1.4% (95% confidence interval, 0.0-3.0). Cardiovascular disease and cancer were the leading causes of death after 10 years of follow-up. CONCLUSION: After surviving 3 years after the rupture of a DACA aneurysm, the patients' long-term survival became similar to that of the matched general population. Rebleeding of treated DACA aneurysm was rare.

Significant variation in performance does not reflect the effectiveness of the cervical cancer screening programme in Finland.

AIM: To characterise the variation in performance indicators of the Finnish cervical screening programme by screening laboratory and to assess whether the performance affects cervical cancer incidence. METHODS: Cervical cancer screening data from 1999 to 2003 from six well-established laboratories were used to analyse rates for follow-up recommendations, referrals and histologically confirmed dysplastic lesions. Laboratory-specific cervical cancer incidences for 1954-2003 were assessed using the cancer registry files. RESULTS: Differences in follow-up recommendations were up to 3.1-fold and 2.2-fold in referrals; differences in the rates for CIN1, CIN2 and CIN3+ were up to 4.5-, 4.7-, and 1.5-fold, respectively. Pre-screening incidence rates for cervical cancer varied 1.5-fold by laboratory, with no major differences in the incidence trends since the onset of screening. CONCLUSION: The performance of a cervical screening programme differs by screening laboratory but does not materially affect the overall programme effectiveness. This leads to variation in cost-effectiveness and probably in avoidable adverse effects. In cervical cancer screening studies, the outcome should be selected as closely as possible to the true measure of programme effectiveness, prevented invasive cervical cancers and subsequent deaths.

Influence of alternative mammographic screening scenarios on breast cancer incidence predictions (Finland).

BACKGROUND: A population-based early detection programme for breast cancer has been in progress in Finland since 1987. Recently, detailed information about actual screening invitation schemes in 1987-2001 has become available in electronic form, which enables more specific modeling of breast cancer incidence. OBJECTIVES: To present a methodology for taking into account historical municipality-specific schemes of mass screening when constructing predictions for breast cancer incidence. To provide predictions for numbers of new cancer cases and incidence rates according to alternative future screening policies. METHODS: Observed municipality-specific screening invitation schemes in Finland during 1987-2001 were linked together with breast cancer data. The incidence rate during the observation period was analyzed using Poisson regression, and this was done separately for localized and non-localized cancers. For modeling, the screening programme was divided into seven different components. Alternative screening scenarios for future mass-screening practices in Finland were created and an appropriate model for incidence prediction was defined. RESULTS AND CONCLUSION: Expanding the screening programme would increase the incidence of localized breast cancers; the biggest increase would be obtained by expanding from women aged 50-59 to 50-69. The impacts of changes in the screening practices on predictions for non-localized cancers would be minor.

Effects of different phases of an invitational screening program on breast cancer incidence.

The aim of this study was to quantify the effects of separate phases of an invitational screening program on breast cancer incidence at different ages. Our database included detailed municipality-specific information about invitations for mass-screening for breast cancer in 267 Finnish municipalities from time period 1987-2001. The age range was 50-74. For this study, the program was divided into 7 separate phases, and those that had not been invited served as a baseline. The incidence rate was modeled using Poisson regression. To study the differences between age groups within each program phase, an interaction term between age group and phase of screening was included in the model. The modeling was done separately for localized breast cancers, nonlocalized breast cancers and all stages combined. For localized cancers, the risk compared to the noninvited increased during the first years of the 2-year screening rounds in all age groups, and declined below the baseline during the second years. This effect was larger during the first round compared to the subsequent ones. In nonlocalized cancers, a clear incidence peak was detected only during the first year of the first round. A decreasing post screening effect was detected in nonlocalized cancers in women aged 60-69, but not in localized cancers. Cumulating over ages 50-74, extra incidence caused by regular screening in localized cancers was 5.2% if the program was continued up to age 59, and 28.0% if it continued up to age 69. Corresponding figures for nonlocalized cancers were decreases of 19.8% and 20.9%, respectively.