SciELO - Scientific Electronic Library Online

 
vol.23Agreement between nutritional status and perception of body image in indigenous Khisêdjê of the Xingu Indigenous ParkSpatio-temporal analysis of leptospirosis in Eastern Amazon, State of Pará, Brazil author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand

Journal

Article

Indicators

Related links

Share


Revista Brasileira de Epidemiologia

Print version ISSN 1415-790XOn-line version ISSN 1980-5497

Rev. bras. epidemiol. vol.23  Rio de Janeiro  2020  Epub June 01, 2020

http://dx.doi.org/10.1590/1980-549720200047 

ORIGINAL ARTICLE

Decrease in blood donation rates in Japan: a time series analysis

Redução na taxa de doadores de sangue no Japão: uma análise de série temporal

IShikoku Medical College - Utazu, Kagawa, Japan.

IIDepartment of Hygiene, Faculty of Medicine, Kagawa University - Miki, Kagawa, Japan.


ABSTRACT:

Objective:

To clarify that one of the causes for the decrease in blood donation (BD) rates was the introduction of the 400 ml BD program in 1986.

Method:

BP rates were monitored over 48 years (1965-2012) and were divided into pre- and post-intervention periods prior to analysis. An interrupted time series analysis was performed using annual data on BD rates, and the impact of the 400 ml BD program was investigated.

Results:

In a raw series, autoregressive integrated moving average analysis revealed a significant change in slope between the pre- and post-intervention periods in which the intervention factor was the 400 ml BD program. The parameters were as follows: intercept (initial value) = 0.315, confidence interval (CI) = (0.029, 0.601); slope (pre-intervention) = 0.316, CI = (0.293, 0.340); slope difference = -0.435, CI = (-0.462, -0.408); slope (post-intervention) = -0.119, CI = (-0.135, -0.103); all, p = 0.000; goodness-of-fit, R2 = 0.963. After adjusting for stationarity and autocorrelation, the parameters were as follows: intercept (initial value) = -0.699, CI = (-0.838, -0.560); slope (pre-intervention) = 0.136, CI = (0.085, 0.187); slope difference = -0.165, CI = (-0.247, -0.083); slope (post-intervention) = -0.029, CI = (-0.070, 0.012); all, p = 0.000 (except for slope (post-intervention), p = 0.170); goodness-of-fit, R2 = 0.930.

Conclusion:

One of the causes for decrease in BD rates may be due to the introduction of the 400 ml BD program in Japan.

Keywords: Blood donors; Time series studies; Japan

RESUMO:

Objetivo:

Esclarecer que uma das causas para a diminuição das taxas de doação de sangue (BP) foi a introdução do programa de doação de sangue BP de 400 mL em 1986.

Método:

As taxas de BP foram monitoradas ao longo de 48 anos (1965-2012) e divididas em períodos pré e pós-intervenção antes da análise. Uma análise de séries temporais interrompidas foi realizada usando dados anuais sobre as taxas de BP, e investigamos o impacto do programa de BP de 400 mL.

Resultados:

Em uma série bruta, a análise integrada autorregressiva de médias móveis revelou uma mudança significativa na inclinação entre os períodos pré e pós-intervenção em que o fator de intervenção foi o programa de 400 mL da BP. Os parâmetros foram os seguintes: intercepto (valor inicial) = 0,315, intervalo de confiança (IC) = (0,029, 0,601); inclinação (pré-intervenção) = 0,316, IC = (0,293, 0,340); diferença de inclinação = -0,435, IC = (- 0,462, -0,408); inclinação (pós-intervenção) = -0,119, IC = (-0,135, -0,103); todos, p = 0,000; poder explicativo do modelo, R2 = 0.963. Após o ajuste para estacionariedade e autocorrelação, os parâmetros foram os seguintes: intercepto (valor inicial) = -0,699, CI = (-0,838, -0,560); inclinação (pré-intervenção) = 0,136, IC = (0,085, 0,187); diferença de inclinação = -0,165, IC = (-0,247, -0,083); inclinação (pós-intervenção) = -0,029, IC = (-0,070, 0,012); tudo, p = 0,000 (com exceção da inclinação (pós-intervenção), p = 0,170); poder explicativo do modelo, R2 = 0.930.

Conclusão:

Uma das causas para a diminuição das taxas de BP pode ser devido à introdução do programa de doação de sangue BP de 400 ml no Japão.

Palavras-chave: Doadores de sangue; Estudos de séries temporais; Japão

INTRODUCTION

In Japan, blood donations, which are only conducted by the Japanese Red Cross Society, began in 1965 (one donation volume: 200 ml) and are used for treating diseases and injuries1. Therefore, it is important to ensure a stable supply of blood. Considering the physical fitness and nutritional status of the Japanese in 1965, the blood donation volume was determined to be 200 ml once, which was about half that of developed countries in Europe and the United States. The volume was corrected to 400 ml in 1986, due to the improvement of Japanese physical strength and nutritional status.

Many studies have shown that blood donation rates are an index to measure the level of social capital2,3,4. Starting in the late 1980s, blood donation rates have continued to decline in Japan5, which prompted the view of declining altruism6. However, only a few studies have shown that factors other than altruism are related to blood donation rates7.

This study focused on the cause for decrease in blood donation rates. Thus, the blood donation rate was investigated using an interrupted time series (ITS) analysis8,9,10,11. The possible cause hypothesized for decrease in blood donation rates was the introduction of the 400 ml blood donation program in 1986.

METHODS

STUDY DESIGN

Blood donation rates (the number of blood donors/total population × 100 :%) were monitored over 48 years (1965-2012) using data from the Welfare Work White Paper in Japan12. Prior to analysis, pre- and post- intervention periods were created, in which the intervention factor was the 400 mL blood donation program introduced in 1986. ITS analysis was performed using annual data on blood donation rates, and the trends and impact of introducing the 400 mL blood donation program were investigated.

DATA ANALYSIS OF CLINICAL QUALITY MEASUREMENT

ITS is an analysis used to observe long-term phenomena and evaluates changes due to certain interventions13,14,15. The following linear regression model is used to estimate the level and trend in the dependent variable before intervention, as well as changes in the level and trend following intervention (Equation 1):

Yt = β0 + β1×time before intervention t + β2×intervention t +β3×time after intervention t + et (1)

In which:

Yt =

the outcome,

t =

time in years at time t from the start of the observation period to the last time point in the series.

All “t” is the number of years elapsed since 1965, which was set to 0. Furthermore, intervention is a measure of time t specified as a dummy variable in which the value is 0 (when occurring before intervention) and 1 (when occurring after intervention) and was implemented in the series. In this model:

β0 =

the baseline level of outcome at the beginning of the series;

β1 =

the slope prior to intervention;

β2 =

the change in the level immediately after intervention (pre-intervention = 0, post-intervention = 1);

β3 =

the change in slope from pre- to post-intervention;

β1 + β2 =

the post-intervention slope;

et =

the random error term

STATISTICAL ANALYSIS

Firstly, the ITS data (raw data: between 1956 and 2012) was analyzed using auto regressive moving average model (ARIMA (p, d, q))16. Secondly, if there was non-stationarity, autocorrelation17, or seasonality, the data were re-adjusted and re-analyzed. At the time of analysis, stationarity in the raw series was evaluated using Dickey-Fuller test with no divergence. Durbin-Watson statistic was then used to check for autocorrelation. Next, data were collected (between 1965 and 2012 over 48 years), a sufficiently large number of data to verify the seasonal variation. Finally, Pearson’s correlation coefficient was used to analyze the correlation between the donation rate and the number of blood donors (400 mL), and the total donation volume was clarified. Modeling and statistical tests were carried out using SPSS 25 (USA) and XLSTAT 2018.5 (USA).

The Shikoku Medical College Ethic Screening Committee determined that medical ethical approval was not required since all the data used in this study was already officially released.

RESULTS

The collected data, including blood donation rates (%), are presented in Table 1. The variables are as follows: year, blood donation rate (%), time period (the order from the beginning to the end of this investigation period), phase (pre-investigation (0) and post-investigation (1)), and interact (pre-intervention (0) and post-intervention (same as time period)).

Table 1. Data collection. 

Year OC TP PH IA Year OC TP PH IA Year OC TP PH IA
1965 0.4 1 0 0 1981 5.9 17 0 0 1997 4.8 33 1 33
1966 1.0 2 0 0 1982 6.1 18 0 0 1998 4.9 34 1 34
1967 1.6 3 0 0 1983 6.5 19 0 0 1999 4.9 35 1 35
1968 2.0 4 0 0 1984 7.0 20 0 0 2000 4.7 36 1 36
1969 2.2 5 0 0 1985 7.2 21 0 0 2001 4.6 37 1 37
1970 2.3 6 0 0 1986 7.1 22 1 22 2002 4.6 38 1 38
1971 2.5 7 0 0 1987 6.8 23 1 23 2003 4.4 39 1 39
1972 2.7 8 0 0 1988 6.5 24 1 24 2004 4.3 40 1 40
1973 3.1 9 0 0 1989 6.4 25 1 25 2005 4.2 41 1 41
1974 3.5 10 0 0 1990 6.3 26 1 26 2006 3.9 42 1 42
1975 3.4 11 0 0 1991 6.6 27 1 27 2007 3.9 43 1 43
1976 3.7 12 0 0 1992 6.2 28 1 28 2008 4.0 44 1 44
1977 4.1 13 0 0 1993 5.8 29 1 29 2009 4.2 45 1 45
1978 4.4 14 0 0 1994 5.3 30 1 30 2010 4.2 46 1 46
1979 4.8 15 0 0 1995 5.1 31 1 31 2011 4.1 47 1 47
1980 5.3 16 0 0 1996 4.8 32 1 32 2012 4.2 48 1 48

OC: outcomes: blood donation rates (%); TP: time periods; PH: phase: pre-intervention; 0: post-intervention; equation 1: to investigate the effect of introducing the 400 mL blood donation program; IA: Interact: pre-intervention; 0: Post-intervention; time period.

First, the ITS data (raw data: between 1965 and 2012) were analyzed using ARIMA (p, d, q). Based on these results, the most compliant model was ARIMA (0, 0, 0). During this study, a consistent increase in blood donation rates was observed before intervention, and a consistent decrease rate after (Figure 1). In a raw series, ARIMA revealed a significant change in slope between the pre- and post-intervention periods (Table 2). The parameters were as follows: intercept coefficient (initial value) = 0.315, confidence interval (CI) = (0.02, 0.601); slope (pre-intervention) = 0.316, CI = (0.293, 0.340); slope difference = -0.435, CI = (-0.462, -0.408); slope (post-intervention) = -0.119, CI = (-0.135, -0.103); all, p = 0.000; ­goodness-of-fit, R2 = 0.963.

Figure 1. Time series data with regression lines for the pre- and post-intervention periods. 

Table 2. Interrupted time series analysis; ARIMA (0, 0, 0). 

Coefficient; 95%CI p
initial value or slope (SE)
Intercept 0.315 (0.146) 0.029 ~ 0.601 < 0.001
Slope pre-intervention 0.316 (0.012) 0.293 ~ 0.340 < 0.001
Slope differences -0.435 (0.014) -0.462 ~ -0.408 < 0.001
Slope post-intervention -0.119 (0.008) -0.135 ~ -0.103 < 0.001

Goodness-of fit: R2 = 0.963; 95%CI: 95% interval of confidence.

Second, we found that the raw series was stationarity using the Dickey-Fuller test without divergence. Since the time series plot of blood donation rate did not diverge, it was considered that the donation rate was either “having unit root” or “satisfying stationarity”. Here, as it was found by the Dickey-Fuller test that the blood donation rate had “no unit roots”, it could be judged that the donation rate satisfied the stationarity legally. It was also confirmed that the raw series was autocorrelated using Durbin-Watson statistics. Since seasonality was not established, the raw series was converted to a logarithmic series and a single moving average was obtained. The adjusted model was ARIMA (1, 0, 0) with the following parameters: intercept (initial value) = - 0.699, CI = (-0.838, -0.560); slope ­(pre-intervention) = 0.136, CI = (0.085, 0.187); slope difference = -0.165, CI = (-0.247, -0.083); slope (post-intervention) = -0.029, CI = (-0.070, 0.012); all, p = 0.000 (except for slope (post-intervention), p = 0.170); goodness-of-fit, R2 = 0.930 (Table 3). Therefore, an increase and decrease in blood donation rates was observed before and after intervention, respectively.

Table 3. Interrupted time series analysis (adjusted; ARIMA (1, 0, 0)). 

Coefficient; 95%CI p
initial value or slope (SE)
Intercept - 0.699 (0.071) - 0.838 ~ - 0.560 < 0.001
Slope pre-intervention 0.136 (0.026) 0.085 ~ 0.187 < 0.001
Slope differences - 0.165 (0.042) - 0.247 ~ - 0.083 < 0.001
Slope post-intervention - 0.029 (0.021) - 0.070 ~ 0.012 0.170

Goodness-of fit: R2 = 0.963; 95%CI: 95% interval of confidence.

Finally, Pearson’s correlation coefficient was used to analyze the correlation between the donation rate and number of blood donors (400 mL) and was found to be -0.9480 (p < 0.001). The total donation volume between 1986 and 2012 was nearly constant every year (1,845,000 (minimum) - 2,167,000 liters (maximum)) (Table 4).

Table 4. Changes in blood donation rate, number of blood donors, blood donation volume. 

Year OC NBD TDV Year OC NBD TDV
1986 7.1 617 1,845 2000 4.7 2,726 2,076
1987 6.8 1,049 1,861 2001 4.6 2,727 2,088
1988 6.5 1,251 1,859 2002 4.6 2,752 2,133
1989 6.4 1,447 1,892 2003 4.4 2,766 2,078
1990 6.3 1,600 1,938 2004 4.3 2,686 2,018
1991 6.6 1,813 2,160 2005 4.2 2,760 1,960
1992 6.2 1,903 2,167 2006 3.9 2,761 1,842
1993 5.8 2,062 2,117 2007 3.9 2,932 1,887
1994 5.3 2,356 2,022 2008 4.0 3,030 1,970
1995 5.1 2,642 1,940 2009 4.2 3,160 2,070
1996 4.8 2,662 1,917 2010 4.2 3,270 2,070
1997 4.8 2,665 1,935 2011 4.1 3,300 2,020
1998 4.9 2,710 2,095 2012 4.2 3,320 2,040
1999 4.9 2,768 2,129

OC: Outcomes: blood donation rates (%); ML: milliliter; NBD: number of blood donors (400 mL) (thousands unit); TDV:Total donation volume (1,000 liters unit).

DISCUSSION

ITS analysis is a powerful quasi-experimental design for assessing the longitudinal impact of an intervention. In this study, the intervention was the 400 mL blood donation program introduced in 1986. Using this analysis, our results suggest that one of the causes for decrease in blood donation rates may be due to the introduction of the 400 mL blood donation program in Japan.

Based on our analysis, the donation rate consistently increased between 1965 and 1985 prior to the 400 mL blood donation program. The slope of the fitted line was 0.316 (adjusted, 0.136). However, the donation rate consistently decreased between 1986 and 2012 after the 400 ml blood donation program was implemented. The slope of the fitted line was -0.119 (adjusted, -0.029). This suggests that the blood donation rate may have decreased due to the 400 ml blood donation program. In this study, the slope after the intervention presented significant difference in the ARIMA (0, 0, 0) model (Table 2) and no significant difference in the ARIMA (1, 0, 0) model (Table 3). In addition, ARIMA (0, 0, 0) model (R2 = 0.963) was more fit than ARIMA (1, 0, 0) model (R2 = 0.930). This was probably because ARIMA (1, 0, 0) model had excessively incorporated the rise in blood donation rates since 2009 on into the model, causing goodness-of-fit to decrease. There is also a strong inverse correlation between the decrease in the blood donation rate and increase in the number of 400 ml blood donations (Pearson’s correlation coefficient = -0.9480, p < 0.001). This result suggests that the implementation of the 400 ml blood donation program may have decreased the blood donation rate. Furthermore, despite the decrease in blood donation rate, the total donation volume between 1986 and 2012 was nearly constant every year. Therefore, there was no risk of shortage in blood supply.

When using ITS, the upper limit effect may affect the possible outcomes18; however, we did not observe any upper limit effects in our study. If the upper limit effect was present, the blood donation rate would be 7.2% (1985) from 1986, and the slope of the regression line would be zero. However, the slope of the regression line was negative (< 0).

Some studies have shown that the decline in blood donations was due to the decline in altruism2,3,4,6. However, in this study, we found that one of the causes for the decline in blood donation rates might be due to the 400 mL blood donation program.

Our study has several limitations. Firstly, we did not investigate other intervention factors19 (e.g., the promotion of high-unit formulations, efficiency of inventory adjustments). In future studies, confounding factors should be controlled. Secondly, since blood donation rates vary among different regions and age groups, future studies should stratify the analyses by these categories20,21.

CONCLUSION

One of the causes for the decrease in blood donation rates may be due to the introduction of the 400 mL blood donation program in Japan.

REFERENCES

1. The Japanese Red Cross Society. “Watching to blood” [Internet]. [cited on Aug 25, 2018]. Available at: Available at: http://www.jrc.or.jp/donation/first/flow/Links ]

2. Guiso L, Sapienza P, Zingales L. The role of social capital in financial development. Am Econ Rev 2004; 94(3): 526-56. http://doi.org/10.3386/w7563 [ Links ]

3. Buonanno P, Montolio D, Vanin P. Does social capital reduce crime? J Law Economics 2009; 52(1): 145-70. https://doi.org/10.1086/595698 [ Links ]

4. Gonçalez TT, Di Lorenzo Oliveira C, Carneiro-Proietti AB, Moreno EC, Miranda C, Larsen N, et al. Motivation and social capital among prospective blood donors in three large blood centers in Brazil. Transfusion 2013; 53(6): 1291-301. https://doi.org/10.1111/j.1537-2995.2012.03887.x [ Links ]

5. Japan. Ministry of Health, Labor and Welfare. Current status of blood business. 1966-2013. Japan: Ministry of Health, Labor and Welfare; 2015. [ Links ]

6. Haruya S. Social capital in Japan reconsidered. Kansai University Institutional Repository 2010; 150: 1-31. [ Links ]

7. Lacetera N, Macis M, Slonim R. Economic reward to motivate blood donations. Science 2013; 340(6135): 927-8. http://doi.org/10.1126/science.1232280 [ Links ]

8. Fretheim A, Tomic O. Statistical process control and interrupted time series: a golden opportunity for impact evaluation in quality improvement. BMJ Qual Saf 2015; 24(12): 748-52. https://doi.org/10.1136/bmjqs-2014-003756 [ Links ]

9. Kontopantelis E, Doran T, Springate DA, Buchan I, Reeves D. Regression based quasi-experimental approach when randomization is not an option: interrupted time series analysis. BMJ 2015; 350: h2750. https://doi.org/10.1136/bmj.h2750 [ Links ]

10. Jandoc R, Burden AM, Mamdani M, Lévesque LE, Cadarette SM. Interrupted time series analysis in drug utilization research is increasing: systematic review and recommendations. J Clin Epdimiol 2015; 68(8): 950-6. https://doi.org/10.1016/j.jclinepi.2014.12.018 [ Links ]

11. Bernal JL, Cummins S, Gasparrini A. Interrupted time series regression for the evaluation of public health interventions: a tutorial. Int J Epidemiol 2017; 46(1): 348-55. https://doi.org/10.1093/ije/dyw098 [ Links ]

12. Japan. Ministry of Health, Labor and Welfare. [Internet] [cited on Aug 25, 2018]. Available at: Available at: https://www.mhlw.go.jp/stf/seisakunitsuite/bunya/kenkou_iryou/iyakuhin/kenketsugo/genjyou.htmlLinks ]

13. Kassakian SZ, Yackel TR, Deloughery T, Dorr DA. Clinical Decision Support Reduces overuse of red blood cell transfusions: interrupted time series analysis. Am J Med 2016; 129(6): 636.e13-20. https://doi.org/10.1016/j.amjmed.2016.01.024 [ Links ]

14. Komen J, Forslund T, Hjemdahl P, Andersen M, Wettermark B. Effects of policy interventions on the introduction of novel oral anticoagulants in Stockholm: an interrupted time series analysis. Br J Clin Pharmacol 2017; 83(3): 642-52. https://doi.org/10.1111/bcp.13150 [ Links ]

15. Belemsaga DY, Goujon A, Tougri H, Coulibaly A, Degomme O, Duysburgh E, et al. Integration of maternal postpartum services in maternal and child health services in Kaya health district (Burkina Faso): an intervention time trend analysis. BMC Health Serv Res 2018; 18: 298. https://doi.org/10.1186/s12913-018-3098-6 [ Links ]

16. Maaskant JM, Tio MA, Van Hest R, Vermeulen H, Geukers VGM. Medication audit and feedback by a clinical pharmacist decrease medication errors at the PICU: An interrupted time series analysis. Health Sci Rep 2018; 1(3): e23. https://dx.doi.org/10.1002%2Fhsr2.23 [ Links ]

17. Morgan OW, Griffiths C, Majeed A. Interrupted time-series analysis of regulations to reduce paracetamol (Acetaminophen) poisoning. PLoS Med 2007; 4(4): e105. https://doi.org/10.1371/journal.pmed.0040105 [ Links ]

18. Devkaran S, O’Farrell PN. The impact of hospital accreditation on quality measures: an interrupted time series analysis. BMC Health Serv Res 2015; 15: 137. https://doi.org/10.1186/s12913-015-0784-5 [ Links ]

19. Penfold RB, Zhang F. Use of interrupted time series analysis in evaluating health care quality improvements. Acad Pediatr 2013; 13(6 Suppl.): S38-44. https://doi.org/10.1016/j.acap.2013.08.002 [ Links ]

20. Wagner AK, Soumerai SB, Zhang MS, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002; 27(4): 299-309. https://doi.org/10.1046/j.1365-2710.2002.00430.x [ Links ]

21. Lopez Bernal JA, Gasparrini A, Artundo CM, McKee M. The effect of the late 200s financial crisis on suicides in Spain: an interrupted time-series analysis. Eur J Public Health 2013; 23(5): 732-6. https://doi.org/10.1093/eurpub/ckt083 [ Links ]

Financial support: none

Recebido: 31 de Outubro de 2018; Revisado: 20 de Fevereiro de 2019; Aceito: 20 de Março de 2019

Corresponding author: Yutaka Owari. 62-1, Hamagobann-cho, Utazu, Kagawa 769-0205, Japan. E-mail: owari@med.kagawa-u.ac.jp

Conflict of interests: nothing to declare

Authors’ contributions: Mr. Owari, hand in hand with his supervisors, developed this idea from conceptualization to final proposal. Dr. Miyatake and Suzuki supervised proposal development, oversaw study design and data analysis.

Creative Commons License This is an open-access article distributed under the terms of the Creative Commons Attribution License