Search

Abstract Background The potential for artificial intelligence (AI) to transform healthcare cannot be ignored, and the development of AI technologies has increased significantly over the past decade. Furthermore, healthcare systems are under tremendous pressure, and efficient allocation of scarce healthcare resources is vital to ensure value for money. Health economic evaluations (HEEs) can be used to obtain information about cost-effectiveness. The literature acknowledges that the conduct of such evaluations differs between medical technologies (MedTechs) and pharmaceuticals, and poor quality evaluations can provide misleading results. This systematic review seeks to map the evidence on the general methodological quality of HEEs for AI technologies to identify potential areas which can be subject to quality improvements. We used the 35-item checklist by Drummond and Jefferson and four additional checklist domains proposed by Terricone et al. to assess the methodological quality of full HEEs of interventions that include AI. Results We identified 29 studies for analysis. The included studies had higher completion scores for items related to study design than for items related to data collection and analysis and interpretation of results. However, none of the studies addressed MedTech-specific items. Conclusions There was a concerningly low number of full HEEs relative to the number of AI publications, however the trend is that the number of studies per year is increasing. Mapping the evidence of the methodological quality of HEEs of AI shows a need to improve the quality in particular the use of proxy measures as outcome, reporting, and interpretation of the ICER.

Abstract Background Osteoarthritis (OA) affects 20% of the adult Danish population, and the financial burden to society amounts to DKK 4.6 billion annually. Research suggests that up to 75% of surgical patients could have postponed an operation and managed with physical training. ERVIN.2 is an artificial intelligence (AI)-based clinical support system that addresses this problem by enhancing patient involvement in decisions concerning surgical knee and hip replacement. However, the clinical outcomes and cost-effectiveness of using such a system are scantily documented. Objective The primary objective is to investigate whether the usual care is non-inferior to ERVIN.2 supported care. The second objective is to determine if ERVIN.2 enhances clinical decision support and whether ERVIN.2 supported care is cost-effective. Methods This study used a single-centre, non-inferiority, randomised controlled in a two-arm parallel-group design. The study will be reported in compliance with CONSORT guidelines. The control group receives the usual care. As an add-on, the intervention group have access to baseline scores and predicted Oxford hip/knee scores and HRQoL for both the surgical and the non-surgical trajectory. A cost-utility analysis will be conducted alongside the trial using a hospital perspective, a 1-year time horizon and effects estimated using EQ-5D-3L. Results will be presented as cost per QALY gain. Discussion This study will bring knowledge about whether ERVIN.2 enhances clinical decision support, clinical effects, and cost-effectiveness of the AI system. The study design will not allow for the blinding of surgeons. Trial registration ClinicalTrials.gov NCT04332055 . Registered on 2 April 2020.

Abstract Introduction Rapid identification and treatment of stroke is crucial for the outcome of the patient. We aimed to determine the performance of glial fibrillary acidic protein (GFAP) independently and in combination with the Prehospital Stroke Score (PreSS) for identification and differentiation of acute stroke within 4.5 h after symptom onset. Patients and methods Clinical data and serum samples were collected from the Treat-Norwegian Acute Stroke Prehospital Project (Treat-NASPP). Patients with suspected stroke and symptoms lasting ≤ 4.5 h had blood samples collected and were evaluated with the National Institutes of Health Stroke Scale prospectively. In this sub study, NIHSS was retrospectively translated into PreSS and GFAP was measured using the sensitive single molecule array (SIMOA). Results A total of 299 patients with suspected stroke were recruited from Treat-NASPP and included in this study (44% acute ischemic stroke (AIS), 10% intracranial hemorrhage (ICrH), 7% transient ischemic attack (TIA), and 38% stroke mimics). ICrH was identified with a cross-fold validated area under the receiver-operating characteristic curve (AUC) of 0.73 (95% CI 0.62–0.84). A decision tree with PreSS and GFAP combined, first identified patients with a low probability of stroke. Subsequently, GFAP detected patients with ICrH with a 25.0% sensitivity (95% CI 11.5–43.4) and 100.0% specificity (95% CI 98.6–100.0). Lastly, patients with large-vessel occlusion (LVO) were detected with a 55.6% sensitivity (95% CI 35.3–74.5) and 82.4% specificity (95% CI 77.3–86.7). Conclusion In unselected patients with suspected stroke, GFAP alone identified ICrH. Combined in a decision tree, GFAP and PreSS identified subgroups with high proportions of stroke mimics, ICrH, LVO, and AIS (non-LVO strokes).

Abstract Background Depression and anxiety are prevalent mental disorders among the working population with potentially high personal and financial cost. The overall aim of this study was to test the applicability of an outreach collaborative model for early identification and treatment of clinical and sub-clinical mental disorders among Danish employees. This applicability was examined by I) investigating the fractions of identified and treated clinical and subclinical cases, II) describing the distribution and characteristics of cases identified and III) investigating the effect of allocated treatment. Methods A longitudinal study design with four assessments (T0-T3) over 16 months was applied. Self-reporting questionnaires probing for psychopathology were distributed to all employees in six consecutively enrolled companies at the four time points. Employees meeting the screening criteria at T1 were assessed diagnostically. Subjects diagnosed with a clinical mental disorder were allocated to outpatient psychiatric treatment, and subjects with subclinical conditions were allocated to preventive cognitive behavioural therapy. Follow-up was conducted 6 and 12 months after initiation of treatment. We used chi-squared test and F-test to compare the different groups on baseline characteristics and mixed effects linear regression to analyse the treatment effects. Results Forty (6.8%) of the 586 responders at T1 were diagnosed with a clinical mental disorder and referred to outpatient psychiatric treatment. Thirty-three (5.6%) were affected by a subclinical condition and referred to preventive treatment. Nearly two-thirds (63%) of the employees diagnosed with a clinical condition had never received treatment before. Symptom severity decreased significantly for both treated groups until follow-up. When compared to a composed control group, subclinical cases displayed a more rapid initial significant symptomatic decrease on the global symptom scale (coefs = − 0.914, 95% CI [− 1.754, − 0,075]) and anxiety sub-scale (coefs = − 1.043, 95% CI [− 2.021, − 0.066]). This did not apply to the clinical cases as no significant difference in change were identified. Conclusions The outreach collaborative model demonstrated an applicability to identify both clinical and subclinical cases, among these a high number of employees with an unmet need for treatment. We found evidence of a positive initial effect on symptomatology from the allocated preventive treatment among the subclinical cases, but not for clinical cases. Trial registration Retrospectively registered at December 18, 2018 at clinicaltrials.gov, identifier: NCT03786328.