Secondary data analysis was conducted on data from the Pakistan Demographic and Health Survey (PDHS) 2006-07. The PDHS is a publicly available dataset produced by ORC Macro for the Measure DHS (Demographic and Health Surveys) Project. It is funded by the US Agency for International Development (USAID). The PDHS contains a wide range of health-related information and is considered to be free of systematic bias .
The PDHS is the largest household survey conducted in Pakistan. It is comprised of a nationally representative sample of over 95,000 households collected using a stratified two-stage cluster sampling procedure. However, for security and political reasons, some areas, such as Federally Administered Tribal Areas (FATA), Federally Administered National Areas (FANA) and Azad Jammu and Kashmir (AJK), were not part of the survey. Amongst several modules on fertility, nutrition, reproductive health and malaria in the PDHS, one is on child immunization . The detailed methodology of the survey design, data collection and management has been described elsewhere .
Our data was limited to mothers with a last-born child (youngest child) between the ages of 12 to 23 months, resulting in a sample size of 2,435. The reason for the selection of this age group was that, until 2006, the course of basic vaccinations (i.e., 12 doses for seven vaccine-preventable diseases) for children was completed by the age of 9 months. Some previous studies have also used the same age-group for studying the utilization of immunization [10, 11, 14]. We computed the dependent variable, “immunization status” by using twelve doses of 5 vaccines, i.e. polio (4 doses), BCG (1 dose), DPT (3 doses), HBV (3 doses) and measles (1 dose). The data on Haemophilus influenza type b (Hib) vaccination were not available in the PDHS 2006-07, as it was first introduced in 2008 in the national EPI. Similarly, no information was available on the Measles 2 vaccination, as it was added to the routine immunization schedule in Pakistan in 2012.
We selected twelve variables, which showed “Received: BCG, polio (0,1,2,3), DPT (1,2,3), HBV (1,2,3) and measles”. These variables had five response categories: No, vaccination date on card, reported by mother, vaccination marked on card and DK (don’t know). We recoded each variable in a similar way. No and DK responses were recoded as “0” and considered as “not received the vaccine”, while the other responses “vaccination date on card, reported by mother, vaccination marked on card” were recoded as “1” and considered as “received the vaccine”. Later, we added all twelve vaccine variables and labeled them “Immunization status”. The immunization status was recoded as “0” if the child had received all twelve doses of the above-mentioned vaccinations and categorized as “complete immunization”, and “1” if the child had missed one or more vaccinations, and categorized as “incomplete immunization”.
We defined “complete immunization” in accordance with the WHO, which considers a child completely immunized if “he or she has received a BCG vaccination against tuberculosis; three doses of DPT vaccine to prevent diphtheria, pertussis, and tetanus (DPT); at least three doses of polio vaccine; and one dose of measles vaccine.” Previously published studies have defined “complete immunization” in a similar way [22, 23].
Based on a literature review [9–15] and available data within the PDHS 2006-07, 14 independent variables were identified. These were: mother’s age, father’s age, mother’s education, father’s education, father’s occupation, wealth index (measured on the basis of household assets and ownership of a number of consumer items and divided into quintiles from one [poorest] to five [richest]), sex of child, birth order of child, place of residence, region, access to information, seeking formal advice/treatment, use of antenatal care and place of delivery. The independent variables, such as sex of child, place of residence (urban/rural) and region (Sindh, Baluchistan, North Western Frontiers Province vs. Punjab [because Punjab is the most developed region having good social indicators]) were used as such. Some independent variables were also recoded, such as mother’s and father’s age (15-24, 25-34 and ≥ 35), mother’s and father’s education (no education, up to primary, up to secondary and higher), father’s occupation (not working, manual worker, clerical/sales/service and management/professional),wealth index (poor, middle and rich) and birth order of the child (1-3, 4-6 and ≥ 7).
The remaining independent variables were computed and recoded, and limited to categories between two and four because of the small number of cases in some categories. The variable: “seeking formal advice/treatment” was computed using three variables from the PDHS, a) Seek advice/treatment from doctor, b) Seek advice/treatment from Midwife/LHV/nurse and c) Seek advice/treatment from Lady Health Worker (LHW). After computation, the variable was recoded into “0” if the mother has not taken advice/treatment from any of these sources and categorized as “no”, and “1” if the mother has taken advice from one or more of these sources and categorized as “yes”. Similarly, the variable “access to information” was computed and recoded using three variables: a) Access to radio, b) Access to television and c) Access to computer. The data was recoded into “0” if the mother has no access to any of these sources and categorized as “no”, and “1” if the mother had access to one or more of the above-mentioned sources and categorized as “yes”.
All the data were weighted and analyzed using SPSS version 17 to account for selection probability, non-response, and sampling differences between regions to produce national estimates of the population. Descriptive statistics for both groups (incomplete immunization and complete immunization) were presented as frequency distributions and percentages. Simple binary logistic regression analysis was carried out to examine the relationship between “incomplete immunization” and the independent variables. In the multiple logistic model, three variables (mother’s age, mother’s education, wealth quintile) were adjusted and fixed. We entered all the independent variables that were significant at the 0.05 level one by one into the model. We also assessed the multicollinearity between the variables and highly correlated variables were eliminated from the logistical model. Multicollinearity was assessed between mother’s and father’s age through Pearson correlation and it was significant at the 0.01 level, so the father’s age variable was eliminated from the model.
This study is based on secondary analysis of publicly available data; hence no ethical approval was required from our institutions. Permission to use the PDHS-2006-07 data was obtained from Measure DHS.