Anders Hammerich Riis

Background: The increasing aging population and limited health care resources have placed new demands on the healthcare sector. Reducing the number of hospitalizations has become a political priority in many countries, and special focus has been directed at potentially preventable hospitalizations. Objectives: We aimed to develop an artificial intelligence (AI) prediction model for potentially preventable hospitalizations in the coming year, and to apply explainable AI to identify predictors of hospitalization and their interaction. Methods: We used the Danish CROSS-TRACKS cohort and included citizens in 2016-2017. We predicted potentially preventable hospitalizations within the following year using the citizens’ sociodemographic characteristics, clinical characteristics, and health care utilization as predictors. Extreme gradient boosting was used to predict potentially preventable hospitalizations with Shapley additive explanations values serving to explain the impact of each predictor. We reported the area under the receiver operating characteristic curve, the area under the precision-recall curve, and 95% confidence intervals (CI) based on five-fold cross-validation. Results: The best performing prediction model showed an area under the receiver operating characteristic curve of 0.789 (CI: 0.782–0.795) and an area under the precision-recall curve of 0.232 (CI: 0.219–0.246). The predictors with the highest impact on the prediction model were age, prescription drugs for obstructive airway diseases, antibiotics, and use of municipality services. We found an interaction between age and use of municipality services, suggesting that citizens aged 75+ years receiving municipality services had a lower risk of potentially preventable hospitalization. Conclusion: AI is suitable for predicting potentially preventable hospitalizations. The municipality-based health services seem to have a preventive effect on potentially preventable hospitalizations.

Purpose: To evaluate the effect of a transitional care intervention (TCI) on readmission among older medical inpatients. Methods: This non-randomised quasi-experimental study was conducted at Horsens Regional Hospital in Denmark from 1 February 2017 to 31 December 2018. Inclusion criteria were patients ≥ 75 years old admitted for at least 48 h. First, patients were screened for eligibility. Then, the allocation to the intervention or control group was performed according to the municipality of residence. Patients living in three municipalities were offered the hospital-based intervention, and patients living in a fourth municipality were allocated to the control group. The intervention components were (1) discharge transportation with a home visit, (2) a post-discharge cross-sectorial video conference and (3) seven-day telephone consultation. The primary outcome was 30-day unplanned readmission. Secondary outcomes were 30- and 90-day mortality and days alive and out of hospital (DAOH). Results: The study included 1205 patients (intervention: n = 615; usual care: n = 590). In the intervention group, the median age was 84.3 years and 53.7% were females. In the control group, the median age was 84.9 years and 57.5% were females. The 30-day readmission rates were 20.8% in the intervention group and 20.2% in the control group. Adjusted relative risk was 1.00 (95% confidence interval: 0.80, 1.26; p = 0.99). No significant difference was found between the groups for the secondary outcomes. Conclusion: The TCI did not impact readmission, mortality or DAOH. Future research should conduct a pilot test, address intervention fidelity and consider real-world challenges. Trial registration: Clinical trial number: NCT04796701. Registration date: 24 February 2021.

Purpose This paper describes the open cohort CROSS-TRACKS, which comprises population-based data from primary care, secondary care and national registries to study patient pathways and transitions across sectors while adjusting for sociodemographic characteristics. Participants A total of 221 283 individuals resided in the four Danish municipalities that constituted the catchment area of Horsens Regional Hospital in 2012–2018. A total of 96% of the population used primary care, 35% received at least one transfer payment and 66% was in contact with a hospital at least once in the period. Additional clinical information is available for hospital contacts (eg, alcohol intake, smoking status, body mass index and blood pressure). A total of 23% (n=8191) of individuals aged ≥65 years had at least one potentially preventable hospital admission, and 73% (n=5941) of these individuals had more than one. Findings to date The cohort is currently used for research projects in epidemiology and artificial intelligence. These projects comprise a prediction model for potentially preventable hospital admissions, a clinical decision support system based on artificial intelligence, prevention of medication errors in the transition between sectors, health behaviour and sociodemographic characteristics of men and women prior to fertility treatment, and a recently published study applying machine learning methods for early detection of sepsis. Future plans The CROSS-TRACKS cohort will be expanded to comprise the entire Central Denmark Region consisting of 1.3 million residents. The cohort can provide new knowledge on how to best organise interventions across healthcare sectors and prevent potentially preventable hospital admissions. Such knowledge would benefit both the individual citizen and society as a whole.

Purpose: To examine the occurrence and severity of coronary artery disease (CAD) in patients with rheumatoid arthritis (RA) compared to non-RA patients in a population referred for coronary computed tomography angiography (CTA) due to chest pain. Patients and Methods: In this cross-sectional study, 46,210 patients from a national CTA database were included. Patients with RA were stratified on serology, treatment with conventional synthetic or biological disease-modifying antirheumatic drugs (DMARDs), and the need for relapse or flare treatment with intraarticular or -muscular glucocorticoid injections (GCIs). Primary outcomes were coronary artery calcium score (CACS) > 0 and CACS ≥ 400, and secondary outcome was obstructive CAD. Associations between RA and outcomes were examined using logistic regression and results were adjusted for age, sex, cardiovascular risk factors and comorbidities. Results: A total of 395 (0.9%) RA patients were identified. In overall RA, crude odds ratio (OR) for having CACS > 0 was 1.48 (1.21– 1.82) and 1.52 (1.15– 2.01) for CACS ≥ 400, whereas adjusted ORs were 1.08 (0.86– 1.36) and 1.21 (0.89– 1.65), respectively. Seropositive RA patients had adjusted OR of 1.16 (0.89– 1.50) for CACS > 0 and 1.37 (0.98– 1.90) for CACS ≥ 400. Patients who had received ≥ 1 GCI in the period of 3 years prior to the CTA had an adjusted OR of 1.37 (0.94– 2.00) for having CACS > 0 and 1.46 (0.92– 2.31) for CACS ≥ 400. Conclusion: This is the first large-scale, CTA-based study examining the occurrence and severity of CAD in RA patients with symptoms suggestive of cardiovascular disease. A higher prevalence of coronary artery calcification was found in RA patients. After adjusting for age, sex, cardiovascular risk factors and comorbidities, the tendency was less pronounced. We found a trend for increased coronary calcification in RA patients being seropositive or needing treatment with GCI for a relapse or flare.

Purpose: This paper describes the open cohort CROSS-TRACKS, which was set up in Denmark in 2015. The cohort comprises population-based data from primary care, secondary care, and national registries. In addition, the paper outlines the new research opportunities provided by the cohort; the possibility to study patient pathways and transitions across sectors while adjusting for sociodemographic characteristics. The new data source is especially interesting for studies of potentially preventable admissions and readmissions. Participants: A total of 221,283 individuals resided in the four Danish municipalities that constituted the catchment area of Horsens Regional Hospital in 2012–2018. A total of 96% of the population used primary care, 35% received at least one transfer payment, and 66% was in contact with a hospital at least once in the period. Additional clinical information is available for hospital contacts (eg alcohol intake, smoking status, body mass index, and blood pressure). A total of 27% (n=9,986) of individuals aged ≥65 years had at least one potentially preventable hospital admission, and 69% (n=6,881) of these individuals had more than one. Findings to date: The cohort is currently used for research projects in epidemiology and artificial intelligence. These projects comprise a prediction model for potentially preventable hospital admissions, a clinical decision support system based on artificial intelligence, prevention of medication errors in the transition between sectors, health behaviour and sociodemographic characteristics of men and women prior to fertility treatment, and a recently published study applying machine learning methods for early detection of sepsis. Future plans: The CROSS-TRACKS cohort will be expanded to comprise the entire Central Denmark Region consisting of 1.3 million residents. The cohort can provide new knowledge on how to best organise interventions across healthcare sectors and prevent potentially preventable hospital admissions. Such knowledge would benefit both the individual citizen and society as a whole. Strengths and limitations of this study: The CROSS-TRACKS cohort is a population-based open cohort containing routinely collected data from both primary and secondary care combined with sociodemographic register data. The cohort is readily available for research projects: the data sources have already been linked, patient pathways have been identified, and data have been remodelled and prepared for cutting-edge use of artificial intelligence and to enable easier and simpler querying across source systems. The cohort is currently limited to the catchment area of Horsens Regional Hospital, but it covers both urban and rural areas and it has a total population of 221,283 individuals by 2018. The cohort will be expanded to include additional catchment areas in the Central Denmark Region, comprising a total population of 1.3 million individuals. The cohort offers new opportunities for health research by tracking the patients’ pathway across healthcare sectors to provide new knowledge on the best way to organise future interventions.

STUDY QUESTION: Does male alcohol consumption affect fecundability? SUMMARY ANSWER: In data pooled across Danish and North American preconception cohort studies, we found little evidence of an association between male alcohol consumption and reduced fecundability. WHAT IS KNOWN ALREADY: Experimental and clinical studies have shown that alcohol affects male reproductive physiology, mainly by altering male reproductive hormones and spermatogenesis. However, few epidemiologic studies have examined the association between alcohol consumption and male fertility. STUDY DESIGN, SIZE, DURATION: Data were collected from two ongoing prospective preconception cohort studies: the Danish ‘SnartForaeldre’ (SF) study (662 couples) and the North American ‘Pregnancy Study Online’ (PRESTO) (2017 couples). Participants included in the current analysis were enrolled from August 2011 through June 2019 (SF) and from June 2013 through June 2019 (PRESTO). PARTICIPANTS/MATERIALS, SETTING, METHODS: Eligible men were aged ≥18 years in SF and ≥21 years in PRESTO, in a stable relationship with a female partner and not using contraception or receiving fertility treatment. In both cohorts, alcohol consumption/serving size was self-reported as number of beers (330 mL/12 oz.), glasses of white or red wine (120 mL/4 oz. each), dessert wine (50 mL/2 oz.) and spirits (20 mL/1.5 oz.). Overall alcohol consumption was categorized as none, 1–5, 6–13 and ≥14 standard servings per week. Total menstrual cycles at risk were calculated using data from female partners’ follow-up questionnaires, which were completed every 8 weeks until self-reported pregnancy or 12 menstrual cycles, whichever came first. Analyses were restricted to couples that had been trying to conceive for ≤6 cycles at study entry. Proportional probability regression models were used to compute fecundability ratios (FRs) and 95% confidence interval (CIs). We adjusted for male and female age, female partner’s alcohol consumption, intercourse frequency, previous history of fathering a child, race/ethnicity, education, BMI, smoking and consumption of sugar-sweetened beverages and caffeine. MAIN RESULTS AND THE ROLE OF CHANCE: The cumulative proportion of couples who conceived during 12 cycles of follow-up were 1727 (64.5%). The median (interquartile range) of total male alcohol consumption was 4.5 (2.0–7.8) and 4.1 (1.0–8.6) standard servings per week in the SF and PRESTO cohorts, respectively. In pooled analyses, adjusted FRs for male alcohol consumption of 1–5, 6–13 and ≥14 standard servings per week compared with no alcohol consumption were 1.02 (95% CI: 0.90–1.17), 1.10 (95% CI: 0.96–1.27) and 0.98 (95% CI: 0.81–1.18), respectively. For SF, adjusted FRs of 1–5, 6–13 and ≥14 standard servings per week compared with no alcohol consumption were 0.97 (95% CI: 0.73–1.28), 0.81 (95% CI: 0.60–1.10) and 0.82 (95% CI: 0.51–1.30), respectively. For PRESTO, adjusted FRs of 1–5, 6–13 and ≥14 standard servings per week compared with no alcohol consumption were 1.02 (95% CI: 0.88–1.18), 1.20 (95% CI: 1.03–1.40) and 1.03 (95% CI: 0.84–1.26), respectively. LIMITATIONS, REASONS FOR CAUTION: Male alcohol consumption was ascertained at baseline only, and we did not distinguish between regular and binge drinking. In addition, we had insufficient numbers to study the effects of specific types of alcoholic beverages. As always, residual confounding by unmeasured factors, such as dietary factors and mental health, cannot be ruled out. Comorbidities thought to play a role in the reproductive setting (i.e. cancer, metabolic syndrome) were not considered in this study; however, the prevalence of cancer and diabetes was low in this age group. Findings for the highest categories of alcohol consumption (6–13 and ≥14 servings/week) were not consistent across the two cohorts. WIDER IMPLICATIONS OF THE FINDINGS: Despite little evidence of an association between male alcohol consumption and reduced fecundability in the pooled analysis, data from the Danish cohort might indicate a weak association between reduced fecundability and consumption of six or more servings per week. STUDY FUNDING/COMPETING INTEREST(S): This study was supported by the National Institutes of Health (R01-HD060680, R01-HD086742, R21-HD050264, R21-HD072326, R03-HD090315), the Novo Nordisk Foundation, Oticon Fonden, Politimester J.P.N. Colind og hustru Asmine Colinds mindelegat and Erna og Peter Houtveds studielegat. PRESTO receives in-kind donations from FertilityFriend.com, Kindara.com, Swiss Precision Diagnostics and Sandstone Diagnostics for the collection of data pertaining to fertility. Dr Wise serves as a consultant on uterine leiomyomata for AbbVie.com. All other authors declare no conflict of interest.

Background: Among men and women diagnosed with colorectal cancer (CRC), 20– 50% will develop a cancer recurrence. Cancer recurrences are not routinely captured by most population-based registries; however, linkage across Danish registries allows for the development of predictive models to detect recurrence. Successful application of such models in population-based settings requires validation against a gold standard to ensure the accuracy of recurrence identification. Objective: We apply a recently developed validation study design for prospectively collected validation data to validate predicted CRC recurrences against gold standard diagnoses from medical records in an actively followed cohort of CRC patients in Denmark. Methods: We use a Bayesian monitoring framework, traditionally used in clinical trials, to iteratively update classification parameters (positive and negative predictive values, and sensitivity and specificity) in an adaptive validation substudy design. This design allows determination of the sample size necessary to estimate the corresponding parameters and to identify when validation efforts can cease based on predefined criteria for parameter values and levels of precision. Results: Among 355 men and women diagnosed with CRC in Denmark and actively followed semi-annually, there were 63 recurrences diagnosed by active follow-up and 70 recurrences identified by a predictive algorithm. The adaptive validation design met stopping criteria for the classification parameters after 120 patients had their recurrence information validated. This stopping point yielded parameter estimates for the classification parameters similar to those obtained when the entire cohort was validated, with 66% less patients needed for the validation study. Conclusion: In this proof of concept application of the adaptive validation study design for outcome misclassification, we demonstrated the ability of the method to accurately determine when sufficient validation data have been collected. This method serves as a novel validation substudy design for prospectively collected data with simultaneous implementation of a validation study.

To what extent does fecundability vary across seasons? After accounting for seasonal patterns in pregnancy planning, we observed higher fecundability in the fall and lower fecundability in the spring, particularly at lower latitudes. In human populations, there are strong seasonal patterns of births that vary across geographic regions and time periods. However, previous studies of seasonality and fecundity are limited because they examine season of birth rather than season of conception and therefore neglect to account for seasonal variation in initiating attempts to conceive or pregnancy loss or differences in gestational length. We conducted a preconception cohort study of 14 331 women residing in North America (June 2013–May 2018: n = 5827) and Denmark (June 2007–May 2018: n = 8504). Participants were attempting to conceive without fertility treatment and had been attempting pregnancy for ≤6 menstrual cycles at enrolment. We collected information on season of each pregnancy attempt using last menstrual period dates over the study period. Pregnancy was reported on female bi-monthly follow-up questionnaires. We fit log-binomial models with trigonometric regression to examine periodic variation in fecundability. We accounted for seasonal variation in initiation of pregnancy attempts by including indicator variables for menstrual cycle of attempt in the regression models. Initiation of pregnancy attempts peaked in September, with stronger seasonality in North America than in Denmark (48 vs. 16% higher probability initiating attempts in September compared with March). After accounting for seasonal variation in initiation of pregnancy attempts, we observed modest seasonal variation in fecundability, with a peak in the late fall and early winter in both cohorts, but stronger peak/low ratios in North America (1.16; 95% confidence interval [CI]: 1.05, 1.28) than in Denmark (1.08; 95% CI: 1.00, 1.16). When we stratified the North American data by latitude, we observed the strongest seasonal variation in the southern USA (peak/low ratio of 1.45 [95% CI: 1.14, 1.84]), with peak fecundability in late November. We estimated menstrual cycle dates between follow-up questionnaires, which may have introduced exposure misclassification, particularly when women skipped follow-up questionnaires. We were unable to measure seasonally varying factors that may have influenced fecundability, including ambient temperature, vitamin D levels or infectious disease. An understanding of how fecundability varies across seasons could help identify factors that can impair reproductive function. Neglecting to account for seasonal variation in initiation of pregnancy attempts could bias estimates of seasonal patterns in fecundability. This is the first preconception cohort study to examine seasonal variation in fecundability after accounting for seasonality in initiation of pregnancy attempts. Fecundability was highest in the fall and lowest in the spring, with stronger effects in southern latitudes of North America, suggesting that seasonal exposures may affect fecundity. This research was funded by the Eunice K. Shriver National Institute of Child Health and Human Development (R21-050264, R01-HD060680, R21-HD072326 and R01-HD086742) and the Danish Medical Research Council (271-07-0338). The authors declare no conflicts of interest.

Objective Rheumatoid arthritis (RA) is a known risk factor for developing coronary artery disease (CAD). The influence of RA on the prognosis after initial CAD diagnosis and treatment is however largely unknown. We examined the risk of major cardiovascular events among RA and non-RA patients with chest pain referred to cardiac CT. Methods This was a follow-up study, using data from the Western Denmark Heart Registry, containing data on CT angiography examinations (Cardiac CT). Information on RA diagnosis and covariates were identified through nationwide administrative registers. The primary outcome was a combined outcome including, myocardial infarction, ischaemic or unspecified stroke, coronary artery bypass grafting, percutaneous coronary intervention, and all-cause mortality. Median time until events or censoring was 3.5 years (min/max: 0.0: 9.2). Cox proportional hazard models were used to examine the association between RA/non-RA patients and outcomes. Results Among 42 257 patients, referred between 2008 and 2016, we identified 358 (0.8%) with RA. An increased risk was seen in RA compared with non-RA (adjusted HR 1.35, 95% CI 0.93 to 1.96). Among patients who had received flare treatment more than once prior to cardiac CT the adjusted HR 1.80 (95% CI 1.08 to 3.00), and among patients with seropositive RA the adjusted HR 1.42 (95% CI 0.93 to 2.16). Conclusion In patients referred to cardiac CT due to chest pain, we found a trend of an association between RA and the combined primary outcome, supporting that RA per se, but in particular seropositive and active RA, may increase the risk of CAD even after initial CAD diagnosis and treatment.