Fulfilling the Promise of Artificial Intelligence in the Health Sector: Let's Get Real.

OBJECTIVES: This study aimed to showcase the potential and key concerns and risks of artificial intelligence (AI) in the health sector, illustrating its application with current examples, and to provide policy guidance for the development, assessment, and adoption of AI technologies to advance policy objectives. METHODS: Nonsystematic scan and analysis of peer-reviewed and gray literature on AI in the health sector, focusing on key insights for policy and governance. RESULTS: The application of AI in the health sector is currently in the early stages. Most applications have not been scaled beyond the research setting. The use in real-world clinical settings is especially nascent, with more evidence in public health, biomedical research, and "back office" administration. Deploying AI in the health sector carries risks and hazards that must be managed proactively by policy makers. For AI to produce positive health and policy outcomes, 5 key areas for policy are proposed, including health data governance, operationalizing AI principles, flexible regulation, skills among health workers and patients, and strategic public investment. CONCLUSIONS: AI is not a panacea, but a tool to address specific problems. Its successful development and adoption require data governance that ensures high-quality data are available and secure; relevant actors can access technical infrastructure and resources; regulatory frameworks promote trustworthy AI products; and health workers and patients have the information and skills to use AI products and services safely, effectively, and efficiently. All of this requires considerable investment and international collaboration.

How do Patient-reported Outcome Scores in International Hip and Knee Arthroplasty Registries Compare?

BACKGROUND: Patient-reported outcome measures (PROMs) are the only systematic approach through which the patient's perspective can be considered by surgeons (in determining a procedure's efficacy or appropriateness) or healthcare systems (in the context of value-based healthcare). PROMs in registries enable international comparison of patient-centered outcomes after total joint arthroplasty, but the extent to which those scores may vary between different registry populations has not been clearly defined. QUESTIONS/PURPOSES: (1) To what degree do mean change in general and joint-specific PROM scores vary across arthroplasty registries, and to what degree is the proportion of missing PROM scores in an individual registry associated with differences in the mean reported change scores? (2) Do PROM scores vary with patient BMI across registries? (3) Are comorbidity levels comparable across registries, and are they associated with differences in PROM scores? METHODS: Thirteen national, regional, or institutional registries from nine countries reported aggregate PROM scores for patients who had completed PROMs preoperatively and 6 and/or 12 months postoperatively. The requested aggregate PROM scores were the EuroQol-5 Dimension Questionnaire (EQ-5D) index values, on which score 1 reflects "full health" and 0 reflects "as bad as death." Joint-specific PROMs were the Oxford Knee Score (OKS) and the Oxford Hip Score (OHS), with total scores ranging from 0 to 48 (worst-best), and the Hip Disability and Osteoarthritis Outcome Score-Physical Function shortform (HOOS-PS) and the Knee Injury and Osteoarthritis Outcome Score-Physical Function shortform (KOOS-PS) values, scored 0 to 100 (worst-best). Eligible patients underwent primary unilateral THA or TKA for osteoarthritis between 2016 and 2019. Registries were asked to exclude patients with subsequent revisions within their PROM collection period. Raw aggregated PROM scores and scores adjusted for age, gender, and baseline values were inspected descriptively. Across all registries and PROMs, the reported percentage of missing PROM data varied from 9% (119 of 1354) to 97% (5305 of 5445). We therefore graphically explored whether PROM scores were associated with the level of data completeness. For each PROM cohort, chi-square tests were performed for BMI distributions across registries and 12 predefined PROM strata (men versus women; age 20 to 64 years, 65 to 74 years, and older than 75 years; and high or low preoperative PROM scores). Comorbidity distributions were evaluated descriptively by comparing proportions with American Society of Anesthesiologists (ASA) physical status classification of 3 or higher across registries for each PROM cohort. RESULTS: The mean improvement in EQ-5D index values (10 registries) ranged from 0.16 to 0.33 for hip registries and 0.12 to 0.25 for knee registries. The mean improvement in the OHS (seven registries) ranged from 18 to 24, and for the HOOS-PS (three registries) it ranged from 29 to 35. The mean improvement in the OKS (six registries) ranged from 15 to 20, and for the KOOS-PS (four registries) it ranged from 19 to 23. For all PROMs, variation was smaller when adjusting the scores for differences in age, gender, and baseline values. After we compared the registries, there did not seem to be any association between the level of missing PROM data and the mean change in PROM scores. The proportions of patients with BMI 30 kg/m 2 or higher ranged from 16% to 43% (11 hip registries) and from 35% to 62% (10 knee registries). Distributions of patients across six BMI categories differed across hip and knee registries. Further, for all PROMs, distributions also differed across 12 predefined PROM strata. For the EQ-5D, patients in the younger age groups (20 to 64 years and 65 to 74 years) had higher proportions of BMI measurements greater than 30 kg/m 2 than older patients, and patients with the lowest baseline scores had higher proportions of BMI measurements more than 30 kg/m 2 compared with patients with higher baseline scores. These associations were similar for the OHS and OKS cohorts. The proportions of patients with ASA Class at least 3 ranged across registries from 6% to 35% (eight hip registries) and from 9% to 42% (nine knee registries). CONCLUSION: Improvements in PROM scores varied among international registries, which may be partially explained by differences in age, gender, and preoperative scores. Higher BMI tended to be associated with lower preoperative PROM scores across registries. Large variation in BMI and comorbidity distributions across registries suggest that future international studies should consider the effect of adjusting for these factors. Although we were not able to evaluate its effect specifically, missing PROM data is a recurring challenge for registries. Demonstrating generalizability of results and evaluating the degree of response bias is crucial in using registry-based PROMs data to evaluate differences in outcome. Comparability between registries in terms of specific PROMs collection, postoperative timepoints, and demographic factors to enable confounder adjustment is necessary to use comparison between registries to inform and improve arthroplasty care internationally. LEVEL OF EVIDENCE: Level III, therapeutic study.

Health care variation: time to act.

Geographic variation in health care use has been demonstrated in many countries over many years. Such variation can be warranted - in response to patient need or preference for care - or unwarranted. Unwarranted variation raises concerns about equity and appropriateness of care. Recent analyses of health care provision in the Australian atlas of healthcare variation show that when routinely available Australian data are mapped by residence of patient, there are wide variations in rates of use of diagnostic tests, dispensing of prescriptions for a range of indications, surgical procedures and hospital admission rates. Despite the wealth of studies demonstrating variation in care internationally, there is relatively little research that explores the best ways of responding to unwarranted variation. Recommendations for action in the Australian Atlas focus on some approaches that could be used in Australia.

Variation in coronary angiography rates in Australia: correlations with socio-demographic, health service and disease burden indices.

BACKGROUND: Variation in the provision of coronary angiography is associated with health care inefficiency and inequity. We explored geographic, socio-economic, health service and disease indicators associated with variation in angiography rates across Australia. METHODS: Australian census and National Health Survey data were used to determine socio-economic, health workforce and service indicators. Hospital separations and coronary deaths during 2011 were identified in the National Hospital Morbidity and Mortality databases. All 61 Medicare Locals responsible for primary care were included, and age- and sex-standardised rates of acute coronary syndrome (ACS) incidence, coronary angiography, revascularisation and mortality were tested for correlations, and adjusted by Bayesian regression. RESULTS: There were 3.7-fold and 2.3-fold differences between individual Medicare Locals in the lowest and highest ACS and coronary artery disease mortality rates respectively, whereas angiography rates varied 5.3-fold. ACS and death rates within Medicare Locals were correlated (partial correlation coefficient [CC], 0.52; P < 0.001). There was modest correlation between ACS and angiography rates (CC, 0.31; P = 0.018). The proportion of patients undergoing angiography who proceeded to revascularisation was inversely correlated with the total angiogram rate (CC, -0.71; P < 0.001). Socio-economic disadvantage and remoteness were correlated with disease burden, ACS incidence and mortality, but not with angiography rate. In the adjusted analysis, the strongest association with local angiography rates was with admissions to private hospitals (71 additional angiograms [95% CI, 47-93] for every 1000 admissions). CONCLUSION: Variation in rates of coronary angiography, not related to clinical need, occurs across Australia. A greater focus on clinical care standards and better distribution of health services will be required if these variations are to be attenuated.

Pay-for-performance and patient safety in acute care: A systematic review.

Pay-for-performance (p4p) has been tried in all healthcare settings to address ongoing deficiencies in the quality and outcomes of care. The evidence for the effect of these policies has been inconclusive, especially in acute care. This systematic review focused on patient safety p4p in the hospital setting. Using the PRISMA guidelines, we searched five biomedical databases for quantitative studies using at least one outcome metric from database inception to March 2023, supplemented by reference tracking and internet searches. We identified 6,122 potential titles of which 53 were included: 39 original investigations, eight literature reviews and six grey literature reports. Only five system-wide p4p policies have been implemented, and the quality of evidence was low overall. Just over half of the studies (52 %) included failed to observe improvement in outcomes, with positive findings heavily skewed towards poor quality evaluations. The exception was the Fragility Hip Fracture Best Practice Tariff (BPT) in England, where sustained improvement was observed across various evaluations. All policies had a miniscule impact on total hospital revenue. Our findings underscore the importance of simple and transparent design, involvement of the clinical community, explicit links to other quality improvement initiatives, and gradual implementation of p4p initatives. We also propose a research agenda to lift the quality of evidence in this field.

Data Rich, Information Poor: Can We Use Electronic Health Records to Create a Learning Healthcare System for Pharmaceuticals?

Judicious use of real-world data (RWD) is expected to make all steps in the development and use of pharmaceuticals more effective and efficient, including research and development, regulatory decision making, health technology assessment, pricing, and reimbursement decisions and treatment. A "learning healthcare system" based on electronic health records and other routinely collected data will be required to harness the full potential of RWD to complement evidence based on randomized controlled trials. We describe and illustrate with examples the growing demand for a learning healthcare system; we contrast the exigencies of an efficient pharmaceutical ecosystem in the future with current deficiencies highlighted in recently published Organisation for Economic Co-operation and Development (OECD) reports; and we reflect on the steps necessary to enable the transition from healthcare data to actionable information. A coordinated effort from all stakeholders and international cooperation will be required to increase the speed of implementation of the learning healthcare system, to everybody's benefit.