傳統在全民健保資料庫的分析研究上，大都從單一特定疾病統計，或共病性疾病間的網絡著手。 近年來，使用人工智慧分析大數據，且應用在醫療領域上已較過往更為可信。 當人工智能學習了中醫十年 (2004-2013年) 健保資料庫門診開立處方的數據，而能預測中醫處方所屬年份的分群現象，可供當作此研究的起始點。
本論文研究，接續了2017年《確認與中醫處方有關的環境和社會經濟變數》初步研究結果中，利用階層式分群法得到社會總體經濟指標-國內生產毛額、國民所得毛額，與AI預測處方年份分群有相似的結果。 進而引發蒐集更多具代表性社會經濟變量，來確認影響此分群現象最為關鍵的因子，並新增了變量關係網絡分析法協助確認。 總共收集- 國民所得、家庭收支、物價統計、衛生統計，教育及環保統計六大類，合計有2500項變量。 結果顯示，其中符合統計檢定有19項，三分之一與教育方面有關，包含教育服務業生產總值 (毛額)、受雇人員報酬，並政府經費運用國人教育等各方面。 進一步把這19項變量計算相關係數並繪製網絡圖。 得到教養設備及用具、發電機、院所病床數、醫療保健服務業、行動電話，傳播業及教育服務業這七項變量，在關係網絡圖中佔有重要的位置及功能。
越來越多的研究實證，開發中國家的教育與人口健康議題，一直存有正向的關係。 但通常有時間延遲及衡量上的困難。 近日更有回顧性文章整理出，成人或全人教育比單靠經濟因素，更有直接且明顯影響全民健康的效果。 因此，從本研究中，亦可佐證政府愈重視全民的教育，愈能提升人民的健康狀態。

The conventional way to employ National Health Insurance Research Database (NHIRD) focuses on single specific disease statistics or comorbidity study by network analysis. In recent years, the use of artificial intelligent (AI) to analyze big data and develop applications in medical field has become more credible. When AI trained with ten years (2004-2013) traditional Chinese medicine (TCM) prescription data extracted from NHIRD, it acquired capability to group TCM prescriptions into clustered years. Therefore we took this as the initiative of our current studies.
In this research, continuous to our prior study of “Identification of environmental and socioeconomic variables that are associated with TCM prescriptions” in 2017, we generated hierarchical clustering-derived data patterns for overall social economic indicators--gross domestic product and gross national income, which are similar to the AI-predicted data patterns for TCM prescriptions grouped by years. As a result, it led us to collect more representative socioeconomic variables to identify the mostly influential factors correlated with the clustering patterns of TCM prescriptions. A new method of network analysis to dissect variable relationships was introduced to facilitate the validation. A total of 2,500 variables in 6 categories, including national income, household income and expenditure, retail price, healthcare, education and environmental protection statistics, were collected. The results showed that there are 19 variables meet the statistical significance, nearly one-third of which are related to education statistics, including education service gross production value, employee income, government funding for national education. These 19 variables were further computed for correlation co-efficiencies and graphed for network mapping. Subsequently, seven indicators as education equipment and device, power generator, ward bed counts, healthcare service, mobile phone, mass communication and education service were weighted to have more important positions and functions in correlation network.
A growing body of research has shown that education and population health issues always exist in a positive correlating manner; however, there are often time delays in such measuring. Moreover, recent retrospective articles have compiled that continuous education or holistic education has more direct and obvious effects on population health than economic factors alone. Therefore, our results are further supportive to the belief that the more the government attaches importance to the development of national education, the more it can improve the health of the people.

1. Preston, S.H., The changing relation between mortality and level of economic development. Population studies, 1975. 29(2): p. 231-248.
2. Bhargava, A., et al., Modeling the effects of health on economic growth. Journal of health economics, 2001. 20(3): p. 423-440.
3. Bloom, D.E. and D. Canning, Population health and economic growth. Health and growth, 2009: p. 53.
4. Acemoglu, D. and S. Johnson, Disease and development: the effect of life expectancy on economic growth. Journal of political Economy, 2007. 115(6): p. 925-985.
5. Dever, G.A., An epidemiological model for health policy analysis. Social indicators research, 1976. 2(4): p. 453-466.
6. Herzer, D. and P. Nunnenkamp, Income inequality and health: Evidence from developed and developing countries. Economics: The Open-Access, Open-Assessment E-Journal, 2015. 9(2015-4): p. 1-56.
7. Dalgaard, C.-J. and H. Strulik, Optimal aging and death: understanding the Preston curve. Journal of the European Economic Association, 2014. 12(3): p. 672-701.
8. Deaton, A., The great escape: health, wealth, and the origins of inequality. 2013: Princeton University Press. Figure 3 Longer lives, richer lives.
9. 「全民健保20週年叢書系列-金色挑戰 (全民健保納保及財務平衡施政紀實)」2015: p. 46 (衛生福利部中央健康保健署網站-圖書分類)
10. 「全民健康保險簡介中英文版」 (圖六、健保開辦後死亡率降低), 2016: p. 103.
11. 周毓媜, 「確認與中醫處方有關的環境和社會經濟變數」2017, 碩士論文(Identification of environmental and socioeconomic variables that are associated with TCM prescriptions).
12. 中華經濟研究院網站, 「臺灣重要經濟變動指標-經濟展望中心編製 PDF」 2018 (表一至表十一，http://www.cier.edu.tw/public/Data/86128282771.pdf).
13. Ivković, A.F., Limitations of the GDP as a measure of progress and well-being. Ekonomski vjesnik/Econviews-Review of Contemporary Business, Entrepreneurship and Economic Issues, 2016. 29(1): p. 257-272.
14. 中華民國統計資訊網：「PC-AXIS 總體統計資料庫」、「臺灣物價統計資料庫」，以及「縣市重要統計指標-人口概況」 (http://statdb.dgbas.gov.tw/).
15. 行政院主計總處網站,「家庭收支調查-102年調查報告 (電子書)」 附錄二、調查方法 (https://win.dgbas.gov.tw/fies/a11.asp?year=102).
16. 主計總處統計專區網站,「物價指數-簡介，訂定物價指數連動調整條款之一般原則 PDF」 (https://www.stat.gov.tw/ct.asp?xItem=42469&ctNode=486&mp=4).
17. Galili, T., dendextend: an R package for visualizing, adjusting and comparing trees of hierarchical clustering. Bioinformatics, 2015. 31(22): p. 3718-3720.
18. Johnson, K.P., M. Kennedy, and K.G. McCracken, Reinterpreting the origins of flamingo lice: cospeciation or host-switching? Biology letters, 2006. 2(2): p. 275-278.
19. Scornavacca, C., F. Zickmann, and D.H. Huson, Tanglegrams for rooted phylogenetic trees and networks. Bioinformatics, 2011. 27(13): p. i248-i256.
20. Venkatachalam, B., et al., Untangling tanglegrams: Comparing trees by their drawings. IEEE/ACM Transactions on Computational Biology and Bioinformatics (TCBB), 2010. 7(4): p. 588-597.
21. Sokal, R.R. and F.J. Rohlf, The comparison of dendrograms by objective methods. Taxon, 1962: p. 33-40.
22. Saraçli, S., N. Doğan, and İ. Doğan, Comparison of hierarchical cluster analysis methods by cophenetic correlation. Journal of Inequalities and Applications, 2013. 2013(1): p. 203.
23. Baker, F.B., Stability of two hierarchical grouping techniques Case I: Sensitivity to data errors. Journal of the American Statistical Association, 1974. 69(346): p. 440-445.
24. Newman, M.E., Finding community structure in networks using the eigenvectors of matrices. Physical review E, 2006. 74(3): p. 036104.
25. 李丞華, et al., 「全民健保中醫門診利用率及其影響因素」 台灣公共衛生雜誌, 2004. 23(2): p. 100-107.
26. Lutz, W. and E. Kebede, Education and Health: Redrawing the Preston Curve. Population and Development Review, 2018. 44(2): p. 343-361.