246 Digital Healthcare Empowering Europeans R. Cornet et al. (Eds.) 2015 European Federation for Medical Informatics (EFMI). This article is published online with Open Access by IOS Press and distributed under the terms of the Creative Commons Attribution Non-Commercial License. doi:10.3233/978-1-61499-512-8-246 Can ehealth Reduce Medical Expenditures of Chronic Diseases? Masatsugu TSUJI a Sheikh AbuTAHER a, b and Yusuke KINAI a a Graduate School of Applied Informatics, University of Hyogo, Kobe, Japan b Jahangirnagar University, Department of Finance and Banking, Dhaka, Bangladesh Abstract. The objective of this research is to evaluate empirically the effectiveness of ehealth in Nishi-aizu Town, Fukushima Prefecture, based on a mail survey to the residents and their receipt data of National Health Insurance from November 2006 to February 2007. The residents were divided into two groups, users and non-users, and sent questionnaires to ask their characteristics or usage of the system. Their medical expenditures paid by National Health Insurance for five years from 2002 to 2006 are examined. The effects were analyzed by comparison of medical expenditures between users and non-users. The interests are focused on four namely heart diseases, high blood pressure, diabetes, and strokes. A regression analysis is employed to estimate the effect of ehealth to users who have these diseases and then calculate the monetary effect of ehealth on reduction of medical expenditures. The results are expected to be valid for establishment of evidence-based policy such as reimbursement from medical insurance to ehealth. Keywords. Heart diseases; high blood pressure; diabetes; strokes; panel data analysis. Introduction This paper analyzes what kind of diseases ehealth is effective for, and then how much ehealth reduces actual medical expenditures by examining Nishi-aizu Town, Fukushima Prefecture, Japan, as a case study. ehealth studied here is to monitor the health condition of the elderly at home by transmitting users health-related data on blood pressure, ECG, and blood oxygen to a remote medical institution. The system is equipped with a simple device that records an elderly person s condition or a patient s illness in graphs that are then used for diagnosis and consultation. Reports sent by the medical institution also help users to enhance their daily health consciousness and maintain good health. It is obvious that this system is quite simple, however, our previous research 1, 2, 3 demonstrated that ehealth actually reduced medical expenditures of. This paper studies in more detail by focusing on specific chronic diseases such as heart disease, high blood pressure, diabetes, and strokes. The analysis and results obtained here provide the rigorous economic foundation of ehealth for Japan as well as for the world.
1. Methods M. Tsuji et al. / Can ehealth Reduce Medical Expenditures of Chronic Diseases? 247 This paper examines the relationship between medical expenditures of Nishi-aizu s residents and ehealth by focusing on ehealth intervention between two groups; (i) users and (ii) non-users. Their receipts from those stored in the town office are examined. The receipts of National Health Insurance of each month are kept at the town office. This paper uses the following data: (i) name of resident, (ii) birth date, (iii) either regular outpatient treatment or hospitalized patient treatment, (iv) name(s) of major disease(s), (v) date of initial treatment, (vi) number of days spent for treatment, and (vii) score (amount) of medical expenditure. Two types of estimation for each disease are conducted; with full sample and selected samples. The former indicates that the analysis covers all samples, while the latter covers only those who have been treated or examined each disease. The estimation model is panel data analysis with the one-way fixed effect model considering time effect. A user dummy variable is added as an explanatory variable, which enables to estimate whether there is difference between users and non-users in medical expenditures of each disease. Twelve categories of diseases treated are shown Table 1. In particular, our interests lie in heart diseases, high blood pressure, diabetes, and strokes, since these diseases are categorized into which this paper and the ehealth system in Nishi-aizu place main targets. Table 1. Diseases treated within five years User Non-user Total Heart diseases 44 23 67 High blood pressure 100 74 174 Diabetes 15 21 36 Strokes 14 10 24 Respiratory diseases 9 10 19 Cancer 8 3 11 Gastropathy 25 13 38 Lumbago, Arthritis 45 43 88 Ophthalmic diseases 57 46 103 Kidney diseases 3 1 4 Anal diseases 9 7 16 Others 19 7 26 Based on the data obtained, the objectives of the estimation consists of the following two questions: (i) how the experience of treatment of four provided effect to total medical expenditures, in other words, whether there is difference in medical expenditures among users and non-users due to ; and (ii) how medical expenditures related to were different among two groups, user and non-user. In so doing, panel data analysis is employed with the one-way fixed effect model. In addition, the individual effect, or dummy variables might cause serious multicollinearity with each characteristic, and accordingly only the time effect is considered.
248 M. Tsuji et al. / Can ehealth Reduce Medical Expenditures of Chronic Diseases? 2. Results Samples of two groups are selected according to the following way: (i) User group, 412 users were selected from the list of registered users in the town according to the year they registered. Then we sent questionnaires to them and 311 replies were received. After checking the replies, 199 replies remain as valid. The rate of valid reply is 38.05%; and (ii) Non-user group, 450 residents who do not use ehealth were selected from the list of National Health Insurance out of total 3,528. Questionnaires were sent to 450 and we received 239 replies. Again by checking the replies, we had 209 significant replies. The rate of significant reply is 46.44%. 2.1 Heart diseases The estimation of heart diseases in Table 2 shows that for full sample case, Sex (p<5%), Age (p<1%), Income (p<5%), and (p<1%) are found to be significant. The question of aims to examine whether respondents reorganize they have. If they do, they have more incentive to use ehealth than those who do not. Although the user dummy variable is not significant for full sample case, but for selected sample it is strongly negative at the 1% significance level. This result can be interpreted that users medical expenditures of heart diseases are lower than those of non-users by approximately JPY39,080.9 (US$390.81) per year. The significance levels of some user characteristics become lower in the selected sample than in the full sample. This seems to be reasonable; since the number of observations is smaller in the latter than in the former and being heart disease might weaken the effect of aging or other factors to medical expenditures. Table 2. Result of estimation I (Heart diseases) Variables Coef. S. E. t val. p val. Coef. S. E. t val. p val. Sex 572.58 286.49 2.00 0.046 * 1136.15 1348.47 0.84 0.400 Age 45.34 17.36 2.61 0.009 *** 75.06 81.26 0.92 0.356 Education 90.45 185.25 0.49 0.625 194.34 1060.51 0.18 0.855 Working -265.28 304.95-0.87 0.384 149.64 1324.46 0.11 0.910 No. of family 80.37 76.63 1.05 0.294 988.26 353.48 2.80 0.006 *** Income -5.94 2.47-2.40 0.016 ** -24.19 11.64-2.08 0.039 ** chronic 835.36 285.06 2.93 0.003 *** 292.06 1364.25 0.21 0.831 diseases User dummy -60.14 289.93-0.21 0.836-3908.09 1309.69-2.98 0.003 *** Constant -2300.88 1414.66-1.63 0.104 1399.02 7195.71 0.19 0.846 Adjusted R2 0.0164 0.0454 Number of Observation 1545 315 2.2 High blood pressure Table 3 summarizes the result of estimation of high blood pressure. Sex (p<1%), Age (p<1%), Number of family (p<1%), Income (p<1%) and (p<1%) are significant. The user dummy variable has negative coefficient at the less than 10% significance level in the full sample, and it amounts to JPY8,660.7 (US$86.60). Moreover, the difference is larger in the selected sample estimation, which
M. Tsuji et al. / Can ehealth Reduce Medical Expenditures of Chronic Diseases? 249 amounts to JPY21,859.3 (US$218.60). In both cases, users medical expenditures of high blood pressure are found to be lower than those of non-users. As for the estimated coefficients of user dummy, the same argument is applicable, that is, coefficient of the user dummy variable is larger in selected sample than in full sample. Table 3. Result of estimation II (High blood pressure) Variables Coef. S. E. t val. p val. Coef. S. E. t val. p val. Sex 2408.42 447.69 5.38 0.000 *** 2229.11 681.87 3.27 0.001 *** Age 147.93 27.13 5.45 0.000 *** 163.67 39.73 4.12 0.000 *** Education -72.52 289.48-0.25 0.802-559.20 421.98-1.33 0.185 Working -46.35 476.54-0.10 0.923 903.69 704.09 1.28 0.200 No. of family 421.12 119.74 3.52 0.000 *** 577.38 176.52 3.27 0.001 *** Income -10.39 3.86-2.69 0.007 *** -13.03 6.76-1.93 0.054 * 3696.00 445.45 8.30 0.000 *** 3943.53 668.43 5.90 0.000 *** User dummy -866.07 453.07-1.91 0.056 * -2185.93 670.07-3.26 0.001 *** Constant -8428.52 2210.64-3.81 0.000 *** -6688.60 3381.95-1.98 0.048 ** Adjusted R2 0.0919 0.0764 Number of Observation 1545 975 2.3. Diabetes The result of diabetes is shown in Table 4, which summarizes that Age (p<10%), Education (p<5%), Working (p<5%), Number of family (p<5%), and (p<5%) are significant. The coefficients of user dummy variable are negatively significant for both full and selected sample estimation, and the difference between two groups amounts to JPY8,784.5 (US$87.85) and JPY37,639.9 (US$376.40), respectively. The main reason for Nishi-aizu Town to introduce ehealth was to manage diabetes. In this account, ehealth in Nishi-aizu is considered to be successful. Table 4. Result of estimation III (Diabetes) Coef. S. E. t val. p val. Coef. S. E. t val. p val. Sex -97.08 300.06-0.32 0.746-408.77 1776.25-0.23 0.818 Age 31.18 18.18 1.71 0.087 * 61.11 108.29 0.56 0.573 Education -413.88 194.02-2.13 0.033 ** -1215.02 1266.35-0.96 0.338 Working 722.28 319.40 2.26 0.024 ** 2171.53 1987.25 1.09 0.276 No. of family -193.26 80.26-2.41 0.016 ** -1158.43 483.21-2.40 0.017 ** Income -2.74 2.59-1.06 0.289-9.64 16.19-0.60 0.552 686.11 298.56 2.30 0.022 ** -649.29 1842.70-0.35 0.725 User dummy -878.45 303.66-2.89 0.004 *** -3763.99 1802.45-2.09 0.038 ** Constant 101.37 1481.66 0.07 0.945 9386.29 8446.43 1.11 0.268 Adjusted R2 0.0167 0.0428 Number of Observation 1545 245
250 M. Tsuji et al. / Can ehealth Reduce Medical Expenditures of Chronic Diseases? 2.4. Strokes The result of estimation of strokes is shown in Table 5. Neither the coefficient of user dummy variable in full sample nor in selected sample is significant. Accordingly, ehealth in Nishi-aizu town provides a positive effect especially to the except strokes and these effects are higher in particular for the people affected by those diseases. Table 5. Result of estimation IV (Strokes) Coef. S. E. t val. p val. Coef. S. E. t val. p val. Sex -276.53 176.10-1.57 0.117-1201.41 1316.01-0.91 0.362 Age 27.27 10.67 2.56 0.011 ** -84.73 78.68-1.08 0.283 Education 56.36 113.87 0.49 0.621 328.28 979.60 0.34 0.738 Working 22.34 187.45 0.12 0.905 1849.92 1284.99 1.44 0.152 No. of family 84.10 47.10 1.79 0.074 * -56.47 365.12-0.15 0.877 Income -1.04 1.52-0.69 0.493 38.97 26.86 1.45 0.148 356.87 175.22 2.04 0.042 ** 1264.20 1368.64 0.92 0.357 User dummy -22.80 178.22-0.13 0.898-1046.93 1451.75-0.72 0.472 Constant -1308.04 869.58-1.50 0.133 10748.15 5942.92 1.81 0.072 Adjusted R2 0.0067 0.0141 Number of Observation 1545 195 3. Conclusion Using rigorous regression analysis, the results obtained thus far are interesting, because town s original goals were to manage namely high blood pressure and diabetes. The analysis demonstrates by rigorous methods that the medical expenditures of town related to these diseases are reduced. It is important to note that solely introducing ehealth does not contribute to its success, but that ehealth has been supported by eagerness of all related staff of the town office. In Nishi-aizu town, there are six public nurses in charge of ehealth, and they always grasp health conditions of all users. It is clear from authors previous studies 1, 2, 3 that ehealth is useful for consultation and maintaining the good health of the elderly and patients suffering from who are in stable condition. However, it is not effective at curing disease. It therefore has the psychological effect of providing a sense of relief to its users by the knowledge of being monitored by a medical institution 24-hours-a-day. This paper enables to provide concrete amounts for the reduction of users medical expenditure of some. References [1] Akematsu, Y. Tsuji, M. Empirical Analysis of the Reduction of Medical Expenditures. Telemedicine and Telecare 2009, 15(3), 109-111. [2] Akematsu, Y. Tsuji, M. Measuring the Effect of Telecare on Medical Expenditures without Bias Using the Propensity Score Matching Method. Telemedicine and e-health 2012, 18 (10), 743-747. [3] Akematsu, Y. Tsuji, M. Relation between telecare implementation and number of treatment days in a Japanese town. Telemedicine and Telecare 2013, 19 (1), 36-39.