The micro-simulation model is a computer program designed to use individual-level data. It is a special type of simulation technology. Through a simulation of each individual’s relevant behaviour (such as medical behaviour), it implements relevant policies on individuals, estimates and predicts the future development trend of a group under certain conditions, judges the impact of policy adjustments on individual distributions, and infers and synthesises the macro effect of policy implementation.6–9
The realisation of the micro-simulation model depends on the quality of the data files. The development of database technology directly affects the accuracy, efficiency, and practicability of the model. The idea of modelling is: sampling the individuals to be studied to obtain a micro database, that is, to establish the environment for the model simulation. Then the simulation model is constructed according to the behaviour of individuals in various systems of society, that is, the main behaviour patterns of individuals in the model are constructed. Computer technology is used to simulate the changes in individual characteristics in response to changes in the relevant policy parameters and characteristics, that is, the simulation results are obtained through the operation of the model. Based on the statistics, inferences, analyses, and the synthesis of characteristic indicators, the impact of policy adjustments on individuals on a micro level is obtained, the effect of policy implementation at all levels is analysed, and the simulation results are summarised and analysed.
The application process is as follows. (1) Based on the change law of the relevant characteristics of insured individuals, the number and characteristics of the SIMIS insured population from 2020 to 2025 are estimated. (2) Based on the national guiding policy on serious illness insurance, this study designs the payment policy of SIMIS in Shanghai according to large expenses, and constructs a micro-simulation model. (3) The micro database is used to determine the medical treatment distribution, medical consumption distribution, basic medical insurance payment proportion, and serious illness insurance payment proportion of patients with large expenses. A random method is used to simulate the medical consumption and medical insurance payment process of each insured person, and to summarise and analyse the effect after the implementation of a serious illness insurance policy.
2.1 Data
The individual data (n = 381,363) of medical insurance payments of 2% of the insured population of Shanghai from 2011 to 2016 were randomly selected from the population with basic medical insurance, covering UEBMI and URRBMI. The treatment categories were outpatients and inpatients. The database provides the following information: (1) identification number, (2) age, (3) gender, (4) insurance type: URRBMI or UEBMI, (5) diagnosis of inpatients, (6) total medical expenses, (7) URRBMI or UEBMI fund payment expenses, (8) self-payment expenses under the URRBMI or UEBMI, and (9) total self-payment expenses. The data include information on all personal characteristics and medical insurance payments.
2.2 Scheme design and model construction
2.2.1 Payment scheme design
This study is based on the guidelines of SIMIS in China that determine serious illness in patients based on large expenses. The setting of the threshold payment of SIMIS in most parts of China is mainly based on disposable income per capita per year before the implementation of the policy. This study assumes that the SIMIS policy in Shanghai was implemented in 2016. In 2015, the per capita disposable income of rural family residents in Shanghai was 25 520 yuan. Based on this, the threshold payment for SIMIS was 25 000 yuan. In view of the high level of economic development in Shanghai, there is no ceiling limit for the total SIMIS payment and the payment proportion is set at 60% according to the national policy guidelines.1
SIMIS is a secondary payment based on basic medical insurance. The payment mode is illustrated in Figure 1, where total medical expenses can be divided into two parts: inside and outside medical insurance.
2.2.2 Inside medical insurance
URRMBI payment. According to the policy, total expenses for an inpatient are split into two tiers: total self-payment and URRBMI payments. The URRBMI payment includes the threshold to trigger the URRBMI fund, self-payment under the URRBMI, and URRBMI fund payments.
SIMIS payment. The expenses described by SIMIS are called serious illness expenses. The SIMIS payment comes from self-payment under the URRBMI threshold payment. If the annual serious illness expenses of inpatients exceed 25 000 yuan, the SIMIS fund is paid proportionally without a ceiling limit; if not, the reimbursement scheme falls under URRBMI payments.
2.2.3 Outside medical insurance
Total self-payment. Expenses outside the medical insurance system are called total self-payments, which the medical insurance cannot reimburse.
2.3 Construct micro-simulation model
The micro-simulation model (Figure 2) is constructed according to the designed SIMIS payment system, which is divided into four modules: micro data, medical service utilisation, policy implementation, and effect analysis module.10–13
The micro-data module is mainly a micro database obtained by a random sampling of 2% of the population with basic medical insurance in Shanghai. To ensure the accuracy of the analysis, this study focuses on the medical treatment of patients with serious illness expenses of more than 10 000 yuan. The micro database provides the following information: 1) the distribution parameters of the admission rate by age and sex, 2) the distribution parameters of each patient visit type, and 3) the annual medical consumption growth rate parameters and medical insurance payment proportion parameters for individual patients. The micro database was updated to 2025.
The effect analysis module mainly analyses the effect of policy implementation by summarising the individual medical consumption and SIMIS results for each target year.
2.3.1 Simulation process
The threshold to trigger the designed SIMIS fund is an annual serious illness expense exceeding 25 000 yuan that will be paid again. Therefore, the payment categories can be divided into three: only outpatient medical services (OOMS), only inpatient medical services (OIMS), and both outpatient and inpatient medical services (BOIMS).
2.3.2 Insured population of SIMIS estimation
The target population for SIMIS is the population that does not participate in UEBMI. The estimation process was conducted as follows. (1) Based on the total registered residence population published in Shanghai Statistical Yearbook in 2010–2019,14 the size of the population with registered residency in Shanghai in 2020–2025 is estimated by fitting an exponential curve. Based on the changing trend in the registered resident population in different age categories, the number of registered residents in Shanghai in 2020–2025 years is estimated. (2) Based on the changes in the composition of the insured population by type and age group,15 it is possible to estimate the target insured population with serious illness insurance before 2025. (3) When the participation rate of basic medical insurance in Shanghai is 97%, the actual participation in URRBMI can be estimated. Table 1 shows the estimation results for the insured SIMIS population.
Table 1
Estimation results of insured population of SIMIS
Year
|
Insured persons
|
Uninsured persons
|
2016
|
3,370,900
|
435,000
|
2017
|
3,438,500
|
436,900
|
2018
|
3,420,400
|
439,100
|
2019
|
3,486,800
|
441,300
|
2020
|
3,569,900
|
443,100
|
2021
|
3,539,700
|
444,900
|
2022
|
3,516,400
|
446,600
|
2023
|
3,593,800
|
448,300
|
2024
|
3,675,000
|
450,000
|
2025
|
3,648,600
|
451,700
|
2.3.3 Population estimation of seriously ill patients
We summarise the total medical expenses, URRBMI fund payment expenses, self-payment under the URRBMI expenses, and total self-payment expenses of the insured in each year according to the ID code of the insured. If the total amount of serious illness expenses exceeds 10 000 yuan, they are screened and the admission rate is calculated. We find the individual consumption data in outpatients and inpatients according to the patient’s ID code, and summarise the number of patients admitted to OOMS, OIMS, and BOIMS from 2011 to 2016 by age and gender, and determine the distribution of patient admission types. During the simulation, the patients in the current year are determined in combination with the distribution of patient admission rates and admission types.
Estimation of admission rate. The actual data show that the admission rate of patients generally increases by a certain amount each year. Therefore, when constructing the admission rate of patients in all the forecast years, a small increase is assumed; for example, rate2017 = rate2016 + (the sum of the added value of admission rate in each year from 2011 to 2015) / 5. The estimated results of the admission rates are given in Table 2.
Table 2
Parameter estimation of patient admission rate from 2011 to 2025 (%)
Gender
|
Age
|
2011
|
2012
|
2013
|
2014
|
2015
|
2016
|
2017
|
…
|
2025
|
1
|
1
|
0.1691
|
0.1782
|
0.2128
|
0.1937
|
0.6576
|
0.6785
|
0.6993
|
…
|
0.8455
|
1
|
2
|
5.5039
|
6.7469
|
7.1574
|
9.7208
|
7.3669
|
7.4407
|
7.5145
|
…
|
8.0313
|
2
|
1
|
0.0604
|
0.1097
|
0.1401
|
0.1916
|
0.4497
|
0.4653
|
0.4809
|
…
|
0.5901
|
2
|
2
|
8.4681
|
8.0847
|
16.3417
|
11.8208
|
7.6279
|
7.6641
|
7.7002
|
…
|
7.9532
|
Note
1 = male, 2 = female; age 1 < 60 years, age 2 ≥ 60 years.
Estimation of admission-type distribution. As there are different types of admissions, patients need to bear different payment percentages of medical service expenses. According to the actual data from 2011 to 2016, the proportion of BOIMS increased slightly, the proportion of OIMS decreased slightly, and the proportion of OOMS increased slightly. In 2017, for example (Table 3), the annual visit types of male patients under the age of 60 were OOMS admission, OIMS admission, and BOIMS admission, with the possibility of 0.113 991, 0.122 510, and 0.763 499, respectively.
Table 3
Distribution of visit types in 2016 and 2017
Gender
|
Age
|
A2016
|
B2016
|
C2016
|
A2017
|
B2017
|
C2017
|
1
|
1
|
0.115 207
|
0.124 424
|
0.760 369
|
0.113 991
|
0.122 510
|
0.763 499
|
1
|
2
|
0.063 366
|
0.009 901
|
0.926 733
|
0.063 580
|
0.009 231
|
0.927 189
|
2
|
1
|
0.133 333
|
0.180 000
|
0.686 667
|
0.128 784
|
0.178 859
|
0.692 357
|
2
|
2
|
0.036 585
|
0.006 098
|
0.957 317
|
0.036 808
|
0.005 730
|
0.957 462
|
Note
1 = male, 2 = female; age 1 < 60 years, Age 2 ≥ 60 years; A is OOMS admission, B is OIMS admission, and C is BOIMS admission.
OOMS = only outpatient medical services, OIMS = only inpatient medical services, BOIMS = both outpatient and inpatient medical services.
2.4 Simulation of medical expenses of BOIMS with serious diseases
2.4.1 Total medical expense forecast
The model is based on the level of annual medical expenses (4 groups), gender (2 groups, male = 1; female = 2), age (2 groups, 0–59 years = 1; 60 + years = 2). The influencing factors are classified into 16 groups. The annual average medical expenses for each group are calculated. Since the data from 2013 to 2016 are relatively complete, the average value of the growth rate of medical expenses in 2013–2014, 2014–2015, and 2015–2016 is taken as the estimated growth rate of medical expenses in 2017. Using the growth rate of medical expenses and the distribution of patients’ medical expenses in 2016, we calculate the distribution of patients’ medical expenses in 2017. Using the same method, the distribution of patient medical expenses from 2018 to 2025 is obtained. Table 4 lists some of the estimated results.
Table 4
Growth rate and average cost estimation of BOIMS medical expenses in 2017
Gender
|
Age
|
Expense
|
Corresponding frequency
|
Average cost in 2016
|
Growth rate
|
Average cost in 2017
|
1
|
1
|
1
|
0.00—
|
33194.15
|
0.011 19
|
33565.75
|
1
|
1
|
2
|
0.25—
|
45858.05
|
0.018 05
|
46685.56
|
1
|
1
|
3
|
0.50—
|
62357.56
|
-0.012 79
|
61560.03
|
1
|
1
|
4
|
0.75—
|
138356.19
|
-0.010 01
|
136970.70
|
1
|
2
|
1
|
0.00—
|
38228.55
|
0.017 82
|
38909.90
|
…
|
…
|
…
|
…
|
…
|
…
|
…
|
2
|
2
|
4
|
0.75—
|
165315.70
|
-0.013 38
|
163103.52
|
Note
The corresponding frequency represents the relative frequency of the four medical cost indicators, from low to high.
BOIMS = both outpatient and inpatient medical services
When these parameters are used to predict patients’ medical expenses, every seriously ill patient is assigned two uniform random numbers, ran01 and ran02. For example, if a 65-year-old male patient enters the hospital in 2017 and the uniform random number is 0.00 ≤ ran01 < 0.25, the total annual medical costs in the hospital is estimated to be cost17 = (ran02 + 0.5) × 38 909.90 yuan. The random number ran02 is the dispersion of increasing the estimated cost of the same unit, and the number 0.5 ensures that the average cost of this unit is 38 909.90 yuan.
2.4.2 Simulation of URRBMI fund payment and serious illness expenses
Before the payment of SIMIS, total personal medical expenses are divided into three parts: the URRBMI fund payment expenses, the serious illness expenses, and total self-payment expenses. For the simulation of the URRBMI fund payment, according to the BOIMS consumption data from 2011 to 2016, the proportion of URRBMI fund payments in the total medical expenses is analysed. The number of classification factors is consistent with the analysis of total expenses. Based on the actual data, the payment proportion in 2014–2016 is relatively stable, so the average proportion in 2014–2016 is taken as the proportion parameter of URRBMI fund payment expenses in 2017–2025. Some of the estimates are presented in Table 5. The proportion of serious illness expense estimations is consistent with that of the URRBMI fund payment expenses.
Table 5
Proportion of URRBMI fund payment in total medical expenses from 2014 to 2016
Gender
|
Age
|
Expense
|
Corresponding frequency
|
2014
|
2015
|
2016
|
Means
|
1
|
1
|
1
|
0.00—
|
0.5270
|
0.5785
|
0.5660
|
0.5572
|
1
|
1
|
2
|
0.25—
|
0.5377
|
0.5160
|
0.5822
|
0.5453
|
1
|
1
|
3
|
0.50—
|
0.5438
|
0.5454
|
0.5663
|
0.5518
|
1
|
1
|
4
|
0.75—
|
0.5781
|
0.4848
|
0.4924
|
0.5184
|
1
|
2
|
1
|
0.00—
|
0.6142
|
0.6335
|
0.6435
|
0.6304
|
…
|
…
|
…
|
…
|
…
|
…
|
…
|
…
|
2
|
2
|
4
|
0.75—
|
0.5750
|
0.6334
|
0.5848
|
0.5978
|
Note
Simulation of medical expenses of OOMS and OIMS with serious diseases, where OOMS = only outpatient medical services, and OIMS = only inpatient medical services
URRBMI = urban and rural residents’ basic medical insurance
For the simulation of two types of medical treatment of seriously ill patients, OOMS and OIMS, the process is consistent with the simulation of BOIMS. (1) The number of patients is determined based on the distribution of visit types. (2) The growth trend of total medical expenses in each group from 2011 to 2016 is analysed to obtain the distribution of the growth rate of total medical expenses. (3) Using the data from 2011 to 2016, the proportional distribution of URRBMI fund payment expenses and serious illness expenses in the total medical expenses are determined to simulate all kinds of medical expenses. 3) The simulation focuses on the medical behaviour of patients whose annual serious illness expenses are more than 10 000 yuan.