A method for generating synthetic longitudinal health data.

Getting access to administrative health data for research purposes is a difficult and time-consuming process due to increasingly demanding privacy regulations. An alternative method for sharing administrative health data would be to share synthetic datasets where the records do not correspond to real individuals, but the patterns and relationships seen in the data are reproduced. This paper assesses the feasibility of generating synthetic administrative health data using a recurrent deep learning model. Our data comes from 120,000 individuals from Alberta Health's administrative health database. We assess how similar our synthetic data is to the real data using utility assessments that assess the structure and general patterns in the data as well as by recreating a specific analysis in the real data commonly applied to this type of administrative health data. We also assess the privacy risks associated with the use of this synthetic dataset. Generic utility assessments that used Hellinger distance to quantify the difference in distributions between real and synthetic datasets for event types (0.027), attributes (mean 0.0417), Markov transition matrices (order 1 mean absolute difference: 0.0896, sd: 0.159; order 2: mean Hellinger distance 0.2195, sd: 0.2724), the Hellinger distance between the joint distributions was 0.352, and the similarity of random cohorts generated from real and synthetic data had a mean Hellinger distance of 0.3 and mean Euclidean distance of 0.064, indicating small differences between the distributions in the real data and the synthetic data. By applying a realistic analysis to both real and synthetic datasets, Cox regression hazard ratios achieved a mean confidence interval overlap of 68% for adjusted hazard ratios among 5 key outcomes of interest, indicating synthetic data produces similar analytic results to real data. The privacy assessment concluded that the attribution disclosure risk associated with this synthetic dataset was substantially less than the typical 0.09 acceptable risk threshold. Based on these metrics our results show that our synthetic data is suitably similar to the real data and could be shared for research purposes thereby alleviating concerns associated with the sharing of real data in some circumstances.

The Healthcare Cost Burden in Adults with High Risk for Cardiovascular Disease.

OBJECTIVE: We calculated the short- and long-term care resource use and costs in adults with high-risk conditions for cardiovascular disease (HRCVD) as defined by the Canadian Cardiovascular Society dyslipidemia guidelines. METHODS: We linked Alberta health databases to identify patients aged ≥ 18 years with HRCVD between fiscal year (FY) 2012 and FY2016. The first HRCVD event was the index event. Patients were categorized into (1) primary prevention patients and (2) secondary prevention patients at the index event and were followed until death, they moved out of the province, or they were censored at March 2018. We calculated the resource use and costs for each of the 5 years after the index event. RESULTS: The study included 459,739 HRCVD patients (13,947 [3%] were secondary prevention patients). The secondary prevention patients were older (median age 61 years vs. 55 years; p < 0.001), and there were fewer females in this group (30.4% vs. 51.3%; p < 0.001). The total healthcare costs in the first year decreased over time (FY2012: 1.16 billion Canadian dollars (CA$); FY2016: CA$1.05 billion; p < 0.001). An HRCVD patient incurred CA$12,068, CA$5626, and CA$4655 during the first, second, and fifth year, respectively (p for trend < 0.001). During the first year, healthcare costs per secondary prevention patient (CA$36,641) were triple that for a primary prevention patient (CA$11,299; p < 0.001), primarily due to higher hospitalization costs in secondary prevention patients (CA$26,896 vs. CA$6051; p < 0.001). CONCLUSIONS: The healthcare costs for HRCVD patients were substantial but decreased over time. The costs were highest in the year following the index event and decreased thereafter. Secondary prevention patients incurred higher costs than the primary prevention patients.

Outcome and Cost of Optimal Control of Dyslipidemia in Adults With High Risk for Cardiovascular Disease.

BACKGROUND: We assessed the impact of optimal dyslipidemia control on mortality and costs in adults at high risk for cardiovascular disease (HRCVD). METHODS: We linked Alberta health databases to identify patients aged ≥ 18 years with HRCVD between April 2012 and March 2017. The first HRCVD event was considered the index event. Patients were categorized into (1) optimal control and (2) suboptimal control of dyslipidemia based on biomarkers and lipid-lowering therapy during the year post-index event. We measured the association between optimal dyslipidemia control and mortality and health care costs using difference-in-difference and propensity score-matching methods. RESULTS: The study included 459,739 patients with HRCVD (43,776 [9.5%] optimal patients). The optimal patients were older (median age = 62 vs 55 years; P < 0.001), included fewer female patients (37.7% vs 52%; P < 0.001), and featured a higher proportion of secondary prevention patients (15.7% vs 1.7%; P < 0.001). Compared with suboptimal patients, the optimal patients had lower adjusted mortality (0.7% vs 1.9% at 1-year and 2.9% vs 5.1% at 3-year post-index event; both P < 0.001), and higher adjusted health care costs (CA$3758 and CA$6844 at 1-year and 3-year post-index event, respectively; both P < 0.001). Among the secondary prevention group, the optimal patients had lower adjusted mortality (2.4% and 5% absolute reduction at 1-year and 3-year post-index event, respectively; both P < 0.001) at no additional costs. The results were robust across 5 definitions of optimal dyslipidemia control. CONCLUSIONS: Patients with optimal dyslipidemia control have lower mortality and incur modestly higher costs. However, secondary prevention patients experience lower mortality at no additional costs.