Table 2.

Health indices by the processes of development

Name of index	Theory, model or framework	Data selection and processing	Formation of index	Testing of index
Miller’s Q index [8,21]	Not reported.	Health data by classes of disease from the Publication Health Services (PHS) publication was used. Mortality rates of both reference and target population were adjusted for age and sex.	The Q index is formed using a mathematical formula.	Diseases ranked by Q were compared (correlation) with the individual data of deaths, inpatients and outpatients that was originally used to compute the Q values.
Gross national health product (GNHP) [8,30]	Not reported.	Data was obtained from two the National Center for Health Statistics (NCHS) publications that provide life-table and disability data for 1971. Population size by age is available in NCHS computer printouts used in computing mortality rates by age. The age group were collapsed to achieve comparability because the two NCHS publication displayed mortality and disability data by different age groups.	Gross national health product was calculated using a mathematical formula.	Not reported.
General index of health [8,32]	Not reported.	Not reported.	Each component is assigned a score between 0 to 10 points. The region with the lowest incidence is assigned 10 points, the highest incidence 0 points and the other regions are ordered by deciles. Scores of each component are then added to have the general index of health with a range between 0 to 30 points.	Not reported.
Life expectancy free of avoidable mortality (LEFAM) [8,33]	Not reported.	Not reported.	Using life table calculation with additional calculation: Subtracting the number of avoidable deaths from the total deaths according to age.	LEFAM is compared with life expectancies by age group. LEFAM and life expectancies is correlated with several variables: mortality rate, infant mortality rate, gross domestic product, hospital beds, health human resources, ambulatory consultations.
Index of child mortality (ICM) [36]	Improved upon the methodological approach to compute an index of health developed by Chandra Sekhar et al. (1991)	Data was assessed for its completeness according to states and years.	Composition of index using factor analysis. The score for ICM is the sumproduct of factor scores and percentage of variation explained by each factor.	Comparison between ICM and under five mortality rate (U5MR).
Healthy life years (HeaLY) [25,37]	Uses the pathogenesis and natural history of disease as the conceptual framework for assessing morbidity and mortality for interpreting the effects of various interventions	Not reported.	Using mathematical formula.	Compare with DALY.
Index of multiple deprivation [8]	Not reported.	Not reported.	'Shrinkage' procedure applied to all data, factor analysis to generate weights to combine indicators, index is ranked then domain standardised and transformed to an exponential distribution; individual domains are weighted (health is 15%) and combined to produced ward index score	Not reported.
Child health index [38]	Not reported.	Indicators were chosen on the basis of historic and routine use to define health outcomes in children and their inclusion as objectives for Healthy People 2010.	Normalisation: Indicators were calculated as rates or percentages. Standard scores were calculated for each state for each health indicator by subtracting the mean of the measures for all states from the observed measure for each state. The resulting measure was then divided by the standard deviation (SD) and multiplied by -1. Weighting: All measures were given the same weight in calculating the overall standard score. Aggregation: Summation for all standard score of each indicator.	Decomposition: Illustrates the differences between the Deep South and other combined regions for each of the health indicators. Link to others: Multivariate analysis was done with variables that was decided to be potential confounders not health outcomes for children.
Composite index of anthropometric failure (CIAF) [40]	Adaptation from Svedberg's framework of anthropometric failure which identified six group of children.	Children with grossly improbable z-scores of anthropometric failure were flagged and excluded.	The CIAF excludes those children not in anthropometric failure and counts all children who have wasting, stunting, or are underweight.	Analysis of variance was used to examine the relationship between undernutrition and standard of living; age-adjusted logistic regressions were used to examine the relationship between undernutrition and morbidity. Children who were not in anthropometric failure (ie, group A) are set as the reference group in each analysis.
Global nutritional index (GNI) [42]	Discuss overall requirements for a well-nourished person, constructed from estimates of nutritional deficits, excess, and food security.	Data were chosen on the basis of comprehensiveness (measuring all aspects of the area), completeness (availability for all countries), and comparability (the appropriateness of comparisons of the measure among countries).	Normalisation: Three indicators were normalised into 0 to 1 scale. Weighting: Because of the lack of an obvious or evidence-based way to weight the three parameters of nutrition, it was decided to weight them equally, as in the HDI. Aggregation: Average value of the three indicators were subtracted from 1 to invert the scale. Because in each indicator a higher value indicated a worse outcome. This invertion made the final score between 0 and 1, with higher scores indicating better nutrition status.	Correlation made between HDI and GNIg.
Inequity-in-health index (IHI) [43]	Developing the IHI using indicators proposed for monitoring progress of the MDG. The index is bi-dimensional composite: estimating inequity in health quantitatively and representing it graphically.	Data selection: (1) Variables were selected if they were individually registered in more than 40% of total countries more than 90 countries). (2) Social determinants of health were not included because we assume the point of view of health outcomes to measure inequity in health. Data exploration: (1) Disparity of countries data were explored using median and Attributable Fraction (AF). (2) Variables were excluded if high uniqueness was verified in at least two factor analysis methods (ie, iterated main factor, maximum likelihood factor method, or main component factor method). Initially, 14 variables was selected, in the end only 6 variables remain.	Normalisation: Attributable Fractions allow relative differences between countries to be estimated. Composition: The scores from the two factors for each variable were obtained by main component analysis and plotted as x and y-axis on a Cartesian plane. Each variable’s area was then calculated as a product of both axes. The sum of all variance was represented as being a circle (360 degree). The percentage of each variable’s area was calculated with respect to the sum of all variable areas. In terms of angles, each vector’s size was the fraction attributed to a specific country in a specific outcome. Each variable therefore had two components represented in the circle: the bi-dimensional score of its variance (angle) and the size of its disparity in health, compared to the best country (attributed fraction vector).	Reliability: Reliability analysis between the three different method getting the area score. Validity: Validity of the index was done by finding linkage with other indicators: Human development index, health gap indicator, human poverty index, life expectancy, and the probability of dying before 40 y of age. Discriminant validity: Discriminate countries by income, region, corruption and level of indebtedness
Wisconsin county health rankings [44]	Adaptation from Kindig and Stoddart population health framework.	Not reported.	Not reported.	Not reported.
MortalityABC index [45]	Draw upon the literature of population health and public health to develop a multidimensional measures covering 3 components of mortality (absolute mortality level, mortality inequality, mortality clustering)	Data was selected according to availability and timeliness.	Comparability: Absolute mortality level (A) and mortality inequality (B) were grouped in tertial classifications (high, medium, low). Mortality clustering (B) were classify into 2 groups: spatial autocorrelation present within the country (significant and not significant). Composition: All the ABC score were paired in three to get the distribution of countries in three-part mortality indicator	Not reported.
EIU outcomes index [46]	Not reported.	DALYs and HALEs was chosen due to expediency. Adult mortality rates and life expectancy at age 60 were added as extra measures since both DALYs and HALE weight young people and children more heavily than older ones.	Not reported.	Compares the outcomes score with spending in health.