Introduction
We measure population health by comparing the status of health over time. Infant mortality is one of those measurements that serve as a barometer for this comparison and makes understanding the effect of health systems and programs between comparable periods (Reidpath and Allotey).
Infant Mortality Rate (IMR) is defined as “the number of deaths in children under 1 year of age per 1000 live births in the same year” (Reidpath and Allotey). The IMR reflects the overall health scenario of a country and statistical measures of developed and developing cuontries indicate that a high or very high IMR is experienced in developing countries while a low IMR is prevalent in developed countries. A low IMR indicates that the health infrastruture is effective while a high IMR indicates that the health infrastructure does not address basic population health requriements . IMR is influenced by calamities and natural disasters or civil unrest such as war in the short term.
IMR is also associated with other factors that are indicative of economic health. For example, low IMR countries have high economic development, high standard of general living conditions, high rate of social well being, and high degree of environmental protection . However, IMR is not a wholistic approach and the World Health Report of 2000 makes no reference in measuring it.
Martiner reports that infant mortality rates have fallen worldwide. Countries that traditionally had high IMR have reduced IMR indicating substantial health progress but are still far behind developed countries. This means that high IMR is concentrated in the Sub-Sahara Africa, African and Asian countries.
This paper deals with the statistical evidence pointing to the state of health of Sri Lanka, as indicated by the measurement of the country’s IMR. The data on IMR is gathered and hypothesized to be related to changes in per capita income as well as health care and educational expenditures. It is theorized that the higher per capita income and more funds poured on health care programs and education in Sri Lanka would decrease IMR significantly. Data on the infant mortality rate (IMR), real Gross Domestic Product per capita in rupees (GDPPC), as well as educational and health expenditures per capita in rupees (EDUCPC and HEXPPC) for Sri Lanka, covering the period 1951 to 1981 were acquired.
Methodology
A statistical model shall be developed to determine the effect of several factors to Sri Lanka’s IMR. There will be two parts of the analysis. The first is a trend analysis showing the time series data for both metrics, shown from 1951 to 1981. The second part of the analysis is a regression analysis. The regression analysis will be done in two parts. Firstly, an estimate using the regression model of the form shown below will be used. The “t” subscript corresponds to year t. The coefficients that shall be obtained shall be interpreted for their statistical significance. Also, a prediction will be made for IMR when Sri Lanka’s GDP per capita level is 750 rupees under the assumption that HEXPPC is at its mean value.
IMRt =α + β1GDPPCt + β2HEXPPCt +ut
Where:
- IMR = Infant Mortality Rate per 1000 live births
- GDPPC = Real GDP per capita in rupees
- EDUCPC = Real Educational Expenditures per capita in rupees
- HEXPPC = Real Health Expenditures per capita in rupees
The second part is the re-estimation of the model including the EDUCPC variable and comment on any changes to the results and goodness of fit:
IMRt =α + β1GDPPCt + β2HEXPPCt + β3EDUCPCt +ut
Results
The graph below shows the trend for Sri Lanka’s IMR from 1951 to 1981. The graph shows a downward trend in Sri Lanka’s IMR in the 30-year observation period. Based on the statistical evidences presented by Martiner, it could be inferred that Sri Lanka’s health programs have achieved significant success which has led to a reduction of IMR. The linear trend line indicates that for every year, Sri Lanka manages to reduce IMR by 1.349 points thus bringing IMR from a high of 82 in 1951 to 29.5 in 1981.
Figure 1 Sri Lanka's IMR from 1951 to 1981
The statistical results of the first regression model yields the results shown in the table below. This regression yields the following coefficients:
Regression Model IMRt =α + β1GDPPCt + β2HEXPPCt +ut
Results IMRt =145.0590 – 0.06625 GDPPCt – 3.0739 HEXPPCt
This regression result shows that every one unit increase in the GDPPP of Sri Lanka reduces the IMR by 0.06625 units while an increase of one unit of HEX reduces IMR by 3.0739 units.
For this model, the test statistics are:
- T-test – the t-stat shows that the degree of freedom (df) is 2 for the regression equation. At a df of 2, the critical t-stat based on statistical tables is 2.920. We will say that the variable is significant if it its absolute value is greater than the calculated t-stat value. Based on the results, GDP and HEX have t-statistics whose absolute values are both greater than 2.920. Therefore the two are significant explanatory variables of IMR.
- F-test – the f-stat is used for calculating the P-values. A variable is significant if it has a P-value that is greater than 1. From what the results show, based on the f-statistic from which the P-values were computed, both factors are significant and therefore are good predictors of IMR.
- Goodness of Fit – the R-squared results show that this regression model captures 92% of all changes in the IMR as explained by the two explanatory variables.
Using a GDPPPP of 750 rupees and a HEXPPP of 13.70548 (the average HEXPPP over the 30-year observation period), we predict that the IMR will be 53.24.
The statistical results of the first regression model yields the results shown in the table below. This regression yields the following coefficients:
Regression Model IMRt =α + β1GDPPCt + β2HEXPPCt + β3EDUCPCt +ut
Results IMRt =139.7691 – 0.06453 GDPPCt – 1.91515 HEXPPCt – 0.4268 EDUCPCt
This regression result shows that every one unit increase in the GDPPP of Sri Lanka reduces the IMR by 0.06453 units while an increase of one unit of HEX reduces IMR by 1.915 units. An increase one unit in EDU reduces IMR by 0.4268.
For this model, the test statistics are:
- T-test – the t-stat shows that the degree of freedom (df) is 2 for the regression equation. At a df of 2, the critical t-stat based on statistical tables is 2.920. Based on the results, all variables have t-statistics whose absolute values are both greater than 2.920. Therefore the three variables are significant explanatory variables of IMR.
- F-test – the f-stat is used for calculating the P-values. A variable is significant if it has a P-value that is greater than 1. From what the results show, based on the f-statistic from which the P-values were computed, all factors are significant and therefore are good predictors of IMR.
- Goodness of Fit – the R-squared results show that this regression model captures 93.6% of all changes in the IMR as explained by the three explanatory variables.
Conclusions
The statistical results show that increasing the spending on healthcare programs and educational programs and increasing the per capita household income improves the quality of life in Sri Lanka as indicated by a decrease in the IMR.
The two regression models improve the interpretation of IMR changes, with the additional explanatory variable increasing the goodness-of-fit from 92% to 93% and make the effect of the explanatory variables more precise.
Annex 4 Data Set
Bibliography
Balbina. "Why Infant Mortality Rate Is an Important Health Indicator?" 1 May 2011. Balbina20's Blog. http://chintu7875.typepad.com/blog/2011/05/why-infant-mortality-rate-is-an-important-health-indicator.html. 4 November 2013.
Berkeley, University of. "Regression Analysis." n.d. University of Berkeley. http://elsa.berkeley.edu/sst/regression.html. 3 November 2013.
Cottrell, A. "An Introduction to Regression Analysis: Basic Concepts." 2011. WFU. http://users.wfu.edu/cottrell/ecn215/regress.pdf. 3 November 2013.
Martiner, Ramon. "Infant Mortality rate in the World: Magnitude, inequities and trends." 18 February 2013. Health Intelligence. http://healthintelligence.drupalgardens.com/content/infant-mortality-rate-world-magnitude-inequities-and-trends. 3 November 2013.
Reidpath, D.D. and P. Allotey. "Infant Mortality as an Indicator of Population Health." 7 May 2002. J. Epidemiol Community Health. http://jech.bmj.com/content/57/5/344.long. 3 November 2013.
