Objectives
This is a system dynamics modelling study based on the settings of the EHCV scheme in Hong Kong. The study aims to evaluate the long-term effectiveness of the enhanced strategies of the EHCV scheme in reducing the reliance on of public primary care services. The objective of the modelling study is to examine changes in the ratio of visits between public and private doctors in primary care (as a metric of reliance on the public sector) given different strategic scenarios.
Design of scenarios
A qualitative phase was initiated to develop a conceptual model i.e. stock-and-flow as a preliminary sketch of the behaviour system which illustrates the pathway of individuals using the voucher and factors affecting the loop in order to design scenarios for testing. The factors were identified from the previous studies, the interim report of the EHCV scheme, a recent survey from the Hong Kong Medical Association, and some unpublished cross-sectional studies [21-25]. An expert panel was formed to comment on the importance of the factors in the system of using vouchers and to identify what adjustment or intervention should be introduced. Five independent private primary care professionals who have worked for at least 10 years in the private sector were invited for interviews through the networks of the investigators. They were also enrolled in the EHCV scheme and experienced the use of the claiming system. Their comments were noted in Chong et al. [26] and scenarios for increasing voucher amounts, lowering age eligibility criteria, and designating vouchers for chronic disease conditions were considered in simulation testing.
System Dynamics model
A System Dynamics (SD) model for testing different strategic scenarios was developed for computational simulations. In public health studies, SD modelling, involving the development of computational simulation, can address the dynamic complexity of different types of care in a health system and systematically evaluate policies for influencing changes. In our study, the SD model was stratified into 5 levels:
Generation of visits to healthcare services (Figure 1): A population model including new births and deaths was built. In the population model, total numbers of visits from different age groups (i.e. <60, 60–64, 65–69, and ≥70 years old) were generated depended on their average numbers of visits and population sizes. The numbers of visits to public and private healthcare services were generated assuming baseline proportions of visits to public and private which is independent to the uses of vouchers.
Generation of visits of using EHCV (Figure 2): Based on the settings of the voucher (e.g. voucher amounts and the purposes of uses), the expected numbers of visits for different services (i.e. non-preventive services, chronic conditions follow-up, dentistry, vaccination, and others) were generated. Inflation in service prices from supplier induced demand was considered according to the trends reported in the historical claims information. The expected total number of voucher visits was the sum of expected numbers of visits for different services.
Generation of visits of using vouchers for chronic diseases (Figure 3): A designating voucher for chronic condition follow-up was suggested for a scenario testing. In the SD model, the total number of voucher visits for chronic conditions was determined by the eligible population size with chronic diseases in different age groups and their utilization patterns. The utilization patterns of the voucher varied by the voucher amount based on the assumption collected from the previous cross-sectional surveys.
Generation of actual number of visits using vouchers (Figure 4): Based on a proportion of the eligible population willing to join the scheme, the actual number of visits was determined with the corresponding total expenditure for the model calibration.
Changes in utilization of primary healthcare services (Figure 5): Based on utilization from the voucher uses, the net changes of visits to public and private providers were calculated based on the difference in the number of visits shifted from public to private and the baseline number of visits in each sector. The ratio of visits between public and private sectors over time was determined.
A set of corresponding deterministic and differential equations of SD model was developed and computerized to depict the flow of individuals seeking healthcare services in public and private sectors. The unit of time in the simulation settings was set as weeks. Time-based inputs were repeated for every week over a 15-years trend of aging (2017-2032). Euler’s method was used to solve the differential equations. As utilization rate is the most widely used outcome measurement for different voucher programs [27], the primary endpoint in the simulations is the ratio of visits to public doctors and private doctors by years as a metric of reliance.
Scenario settings
The settings of tested scenarios were assumed plausibly as below:
Increasing voucher amounts: We tested the annual voucher amounts of increasing to $3,000, $4,000, and $5,000 at year 2021, 2025, and 2029 respectively, keeping the rate of increment similar pace to the historical trends of the current EHCV scheme.
Lowering the age eligibility: We tested including the elderly aged 60-64 years into the scheme starting from year 2021.
Designating vouchers for chronic conditions follow-up: On top of the original EHCV, we tested an additional voucher for visits of chronic disease conditions. An extra amount of $2,000 was given to eligible population every year starting from 2021 for treating their chronic diseases.
Data collection for model inputs
The data and statistics for inputs of the SD model were collected from different sources:
Demographics statistics from Hong Kong Census and Statistics (C&S) Department [1, 28]
Public healthcare services utilization statistics from Hospital Authority Statistical Reports and Thematic Household Survey Report [3, 29, 30]
Baseline EHCV statistics from literature, the Department of Health's (DH) interim report, the published survey from Hong Kong Medical Association [19-23], and the cross-sectional studies (unpublished). The cross-sectional studies include i. a repeated cross-sectional survey of elderly person aged 70 or above assessing their changes in attitudes towards, and usage of, vouchers among elderly persons in the community, and ii. a public opinion survey of the general public examining the potential use of voucher in primary care system e.g. enhancement of voucher for preventive care and chronic disease management from the general public perspectives.
Statistics of claimed amount from DH’s administrative data and other relevant published statistics [31, 32].
Baseline calibration
Baseline scenario followed the current EHCV scheme: $250 annually for elderly aged 70 years or above in 2009-2011; $500 in 2012, $1,000 in 2013, and $2,000 in 2014-2016 annually; and $2,000 annually for elderly aged 65 years or above after 2016. For the calibration of SD model, the model simulated total expenditure from the voucher visits was validated against the DH published total expenditure of voucher claims from 2009 to 2016 [31, 32]. The mean absolute percentage error (MAPE) of the calibrated model should be kept below 40%. R-square was also obtained to assess the fitness of model calibration. Based on calibrated model, the numbers of visits that uses the vouchers for different services were simulated until year 2032. The variable specifications for model development was listed in Additional file 1.