Medical insurance and physician-induced demand in China: the case of hemorrhoid treatments

See the following website for details: http://​www.​chinadaily.​com.​cn/​a/​201905/​21/​WS5ce3f51ca31048​42260bcef1.​html.

Average costs of new and traditional hemorrhoid treatments in our sample are respectively USD2582 and USD1750. According to the IMF, China’s GDP per capita in 2018 was USD9608. Thus, hemorrhoid treatments are affordable for most Chinese. See the following website for details of China’s GDP per capita: https://​cn.​knoema.​com/​pjeqzh/​gdp-per-capita-by-country-statistics-from-imf-1980-2024?​country=​China. Moreover, in accordance with a recent disclosure by China’s government, the annual per-capital income in China’s poverty-stricken population (about 122 millions) is CNY10740 (about USD 1500) in 2020 (See the website for details: https://​baijiahao.​baidu.​com/​s?​id=​1696268826517225​237&​wfr=​spider&​for=​pc). That said, even for the poverty-stricken population in China, the treatment fee of hemorrhoids, which is not more than USD2582, would be less than 2-years income of them. In this sense, the treatment fee of hemorrhoids is affordable for the mass. Taken the merits of affordability and non-fatality in treating hemorrhoids into account, the involved patients have enough time to communicate with the physicians, and modify their irrational expectations. In this sense, using the cases of hemorrhoids as sample can facilitate us to exclude noisy factors such as lack of communication, high medical costs, and patients' irrational expectations.

For details, see the official documents http://​hrss.​gd.​gov.​cn/​gkmlpt/​content/​1/​1301/​post_​1301815.​html#1268 and newspaper report http://​epaper.​oeeee.​com/​epaper/​A/​html/​2016-04/​14/​content_​27413.​htm.

The conservative treatment for grade I has been agreed internationally; thus, there are no controversial options for comparison and it is excluded from the RCTs.

ICER represents the additional cost of one unit of outcome gained by one strategy compared with another. This widely used methodology in cost-effectiveness analysis is used to examine the costs and health outcomes of alternative strategies. QALY provides an economic evaluation of health outcomes, estimated by multiplying the length of life in each health state by its self-reported health-related quality of life (HRQoL) weight. The weights are on a scale ranging from 1 (full health) to 0 (a health state equivalent to being dead), with negative values indicating a health state considered to be worse than being dead.

For details, see the official documents http://​www.​gz.​gov.​cn/​gzybj/​gkmlpt/​content/​5/​5495/​post_​5495074.​html#14461.

The social medical insurance covers more than 95% Chinese in China up to the end of 2020. (see the website for details: http://​www.​gov.​cn/​xinwen/​2021-03/​08/​content_​5591551.​htm). However, it is worthy to note that in a long history, the social medical insurance in China supports only the local medical cares. Since 2013, some city governments start to do some attempts on introducing the social medical insurance for cross-city patients.

For details, see the official documents on the following website: http://​hrss.​gd.​gov.​cn/​gkmlpt/​content/​1/​1301/​post_​1301815.​html#1268 and newspaper report http://​epaper.​oeeee.​com/​epaper/​A/​html/​2016-04/​14/​content_​27413.​htm.

The treatment fees covered by social medical insurance in China are always directly paid by the government. That said, it is unnecessary for patients to pay the fees in advance.

For example, we could only confirm a case’s therapeutic effect in year level instead of in day/month level.

If we attribute the cases including both traditional and new schemes to the group involved with new schemes, the regression results would remain unchanged.

In this case, one might argue that the observations treated meanwhile by two options should be attributed to the new option group, since they are involved to be induced to adopt new treatment. For robustness, in the unreported tables, we also attempt to redo regressions by taking these observations into the new option group, and the related results are still consistent.

The main reason of using only in-patients data lies in that the two medical options (the traditional one and the new one) for hemorrhoid are mainly for the in-patients requiring surgical treatments. That said, the out-patients always do not require surgical treatments and medical options, which deters us in looking into the SID behaviors. Therefore, we exclude the outpatients cases.

In this paper, the number of visits include the visits related to only hemorrhoids. That said, we exclude visits related to other diseases from our sample. In our sample, the average visit of local patients in the previous year is 0.3303, which is far higher than non-local patients’ 0.1847. It is worthy to note that someone might attribute the cause of high SID in cross-city patients to the less communication with the physicians of them, which could be denoted by the low number of visits, no matter what reason induces the less communication. Therefore, we have to involve the number of visits in our regressions for ruling out the concern.

In accordance with the public disclosure of National Bureau of statistics of the People's Republic of China, the share of people with high education is 6.5% in China, while in Guangdong province the share is 5.8%, which is close to the share in our sample (see the following website for details: http://​www.​stats.​gov.​cn/​tjsj/​ndsj/​2018/​indexch.​htm). The majority of civil servants in China experience high education. In the recent 5 years, 95.28% of Chinese civil servants requires the education degrees of undergraduate and above (see the following website for details: https://​www.​offcn.​com/​gjgwy/​2020/​0922/​68405.​html).

Here we treat local patients with or without insurance and civil servants equally in control group. We could observe from the results of Table 5 in the context that only the coefficients of Non_localinsu are significantly positive in contrast to those of Local_insu and Free_medical, enlightening that the policy could only dramatically drives the SID on cross-city patients with insurance. We also attempt to use only the local patients with insurance as the control group in some unreported tables, the results still are consistent.

ssThe policy is indeed enacted in Oct, 2015, that said, the policy might have some effects in 2015. For robustness, we hereby attempt to redefine the “Policy” variable, denoted as Policy1. The “Policy1” variable is defined as zero for years 2013 and 2014, 0.25 for year 2015; and one for years 2016, 2017 and 2018. We repeat the regressions in Table 11 of “Appendix”, the results of which are consistent with those based on the Eq. (2).