Page 76 - GHES-2-4

Global Health Economics and
Sustainability
Assessing Vietnam’s pandemic lockdown

for these interventions was −1.14 (95% CI: −3.5 – 1.22). 4. Conclusion and policy implications
However, in HCM City, an exception occurred where
the local government needed to develop its own, This study examines the effectiveness of lockdown
more stringent policy (Directive No.11/CT-UBND) to directives at the province level in Vietnam. For this
control the infection rate effectively. This intervention, purpose, we applied the ITS model, along with graphical
referred to as HCM-CT11-1 in the forest plot, had a analysis, to examine the changes in the trendline of newly
significant impact, with a trendline value of −59.14 (95% confirmed COVID-19 cases following the implementation
of interventions in 49 provinces and centrally run cities
CI: −86.7 – −31.58) for the post-intervention period.
from April 27, 2021, to October 1, 2021. The empirical
3.3. Limitations of the technical analysis analysis reaffirms the evidence from previous research
that government lockdown policies are effective in curbing
While ITS analysis with regression-based models is widely the spread of the virus. However, the effectiveness of these
used for evaluating public health interventions, it does policies is influenced by various factors.
not account for some key characteristics of time series
data, such as autocorrelation, overdispersion, and seasonal Geographically, the model results reveal that provinces
trends. in the Northern Midlands and Central Highlands were
more successful in implementing lockdown interventions,
Autocorrelation in time series data can lead to an largely due to their low population density. In addition,
underestimation of true variability, which in turn affects provinces in the Red River Delta, North Central, and
the validity of statistical inferences. We conducted the Central Coastline achieved better outcomes in controlling
Durbin-Watson test (Durbin & Watson, 1950) across 49 the COVID-19 pandemic than the Southern provinces.
provinces and found that 18 models exhibited significant
autocorrelation. This finding indicates that residual Regarding policy stringency, Directive No.16/CT-TTg
autocorrelation is a notable issue in our analysis. To address in targeted areas is only effective when the number
this limitation, future studies could explore methods such as of infections is relatively low. For example, Northern
autoregressive integrated moving average (ARIMA) models provinces applied this directive effectively, maintaining
(Schaffer et al., 2021), which can account for autocorrelation low infection rates and preventing the pandemic from
and seasonal characteristics. Overdispersion can affect the spreading as it did in the Southern provinces.
accuracy of statistical inferences by inflating the variance. However, when the number of new confirmed cases
Although our current analysis does not directly address exceeds a certain threshold, Directive No.16/CT-TTg for
overdispersion, future research should consider using specific locations becomes less effective. Therefore, local
models that can handle overdispersed data to enhance the governments must act decisively by applying province-/
robustness of the findings. city-wide Directive No.16/CT-TTg. Ha Noi, Da Nang,
Seasonal trends can significantly impact the outcomes Khanh Hoa, and Phu Yen successfully implemented
of health programs and policies. While techniques such Directive No.16/CT-TTg province- or city-wide after
as seasonal decomposition or the use of Fourier terms recognizing that targeted lockdowns were insufficient to
in regression models are often employed to address this control the spread of the virus.
issue, the unique circumstances of the lockdown period Delayed implementation of Directive No.16/CT-TTg
likely overshadowed any typical seasonal patterns in our at a province- or city-wide level can have harmful effects
study. Given the limited timeframe and the atypical citizen when it is no longer effective. This was evident in the three
behavior observed during the lockdown, we believe that Southeast provinces of Binh Duong, Dong Nai, and HCM
seasonality is not a major factor in this analysis. However, City. By the end of the study period, Binh Duong and Dong
future studies with longer observation periods should Nai had failed to reverse the trendline. In HCM City, the
adjust for seasonal trends to better isolate the true effects local government had to apply Directive No.11/CT-UBND
of the interventions. with the highest level of stringency to successfully reverse
In summary, while our analysis provides valuable the trend.
insights into the effects of public health interventions Furthermore, it is essential to note that timely testing and
during the pandemic, certain limitations exist. Addressing case detection are crucial for the effective implementation
these issues in future research by incorporating models like of lockdown policies. This study did not include testing
ARIMA, which can handle autocorrelation and seasonality, capacity in its measures due to limited data availability.
and using techniques to adjust for overdispersion and However, based on the model results, the Mekong River
confounding factors, will enhance the robustness and Delta provinces failed to detect COVID-19 cases early,
validity of the findings. allowing the virus to spread silently throughout the region.

Volume 2 Issue 4 (2024) 12 https://doi.org/10.36922/ghes.3423