Intibucá is one of the 18 states (“departments”) into which the nation of Honduras is divided. Concepción is one of 16 municipalities in the department, with a population of approximately 10,458 people
]. The nearest hospital is located in La Esperanza, the capital of Intibucá, approximately 35 kilometers away or three hours by bus. The mountainous terrain and limited transportation hinder access to care for Hondurans in this region.
Community health workers and data analysis staff hired by Shoulder to Shoulder conducted the census from November of 2009 to April of 2010 using a standardized form. Shoulder to Shoulder staff traveled from home to home collecting vital registration information such as births, deaths, and geographic location, as well as supplemental health information, including immunization information for each child in the home recorded on vaccination cards. These vaccination cards are assigned to individual children, and the parent is responsible for bringing the card to the health center at the time of immunization. Township divisions are slightly different in our two data sources; certain communities are included in one township in the Shoulder to Shoulder database, but another township according to the Honduran government. For the purposes of this analysis, we used Shoulder to Shoulder township boundaries.
Shoulder to Shoulder database
The data used for this study are not openly available; we received permission to use vaccine data for individuals in rural western Honduras from Shoulder to Shoulder, the custodian of the relevant health records. Data analysis staff systematically entered census information into Shoulder to Shoulder’s database, which organizes health and demographic information for quick access and updates. The Shoulder to Shoulder database system is written in Microsoft Access 2007. The system interface is via 30 forms that are supported by approximately 150,000 lines of Visual Basic Code, which provide navigation, error checking and selected context sensitive help. The database is structured around eight main tables and a series of support tables. Shoulder to Shoulder employs six to eight full-time Honduran staff members dedicated to the database. They update the database daily with new patient information, and Shoulder to Shoulder uses these data to perform analyses that direct community interventions.
Immunization data from the Shoulder to Shoulder database
Immunization data were obtained from immunization cards during the census period and extracted from the Shoulder to Shoulder database. A database query was performed to produce an immunization table restricted to children less than 5 years of age (as of May 1, 2010) who live in the nine townships of Concepción: Calucica, Colomarigua, Concepción, El Guachipilincito, El Guajiniquil, El Rodeo, Jiquinlaca, San Nicolás, and Santiago. From this table, 14 children were found to be missing immunization records. These children were noted for subsequent data collection endeavors, but excluded from this analysis.
In order to assess the number of children who had received the accurate doses of each vaccine, a reference table was constructed to specify the appropriate dosage at each age. Because all data pertained to children less than 5 years of age, the analysis was restricted to the five vaccines administered to this age group: BCG, pentavalent (diphtheria, tetanus, whooping cough, hepatitis B, and Hib), OPV (oral polio vaccine), MMR (measles, mumps, and rubella), and DPT (diphtheria, pertussis, and tetanus booster). A second reference table was created to organize the immunization data for each child by location (department, township, and village), carnet number (a number assigned by Shoulder to Shoulder to each person registered in the database), gender, name, date of birth, age in years, and received doses of BCG, pentavalent, OPV, MMR, and DPT.
A third table cross referenced data from the reference tables to produce raw results. By creating a formula that pinpointed the number of doses the child received, and then subtracting the number of doses he or she should have received according to the first reference table, a fourth table of missing doses was constructed, organized, and sorted by township.
A final table used township population size and dosage information from the second table to calculate immunization rates and coverage. We did this by restricting the count to a single township and excluding children for whom we had no immunization data. For each percentage, the denominator was the total number of children per township.
Immunization data from health center records
Through interviews with Shoulder to Shoulder community health workers, data analysis staff, and nurses, we discovered that the immunization cards were not always accurate or up-to-date, as they were contingent upon mothers bringing the cards in to be updated with every vaccine dose. We therefore utilized a second source of data: immunization records kept by the health centers in their “Listados de Niños para Vigilancia Integral” (LINVIs), or “Lists of Children for Integrated Monitoring.” We again received permission to use these vaccine data for individuals in rural western Honduras from Shoulder to Shoulder, the administrator of these health records. LINVIs are handwritten tables of immunization information kept within the three local health centers servicing the townships. The townships are divided into villages. Each LINVI contains all children less than 5 years of age who live within a specific village, organized by mother’s name and birth date. A nurse manages and updates a separate LINVI for each village serviced by the health center by recording vaccine dates. By counting the number of times a child received a vaccine by a certain date, we could estimate the received and missing doses for each child.
We collected data in November 2010 by copying all LINVI entries dating back to 2005 into a spreadsheet containing columns for: name of the village and township, name of child, name of mother, date of birth, and number of doses received of BCG, pentavalent, OPV, MMR, and DPT vaccines. Since LINVI entries did not include the carnet numbers of each child, names or dates of birth were used to identify and match children with those recorded in the database. We then applied the same calculations to the raw data collected at the three health centers. Our data entry staff used matched records to add additional and missing information to our database, including names of newborns, names of mothers, changes in location, and deaths. Children who were not found in the database were noted and copied into a new spreadsheet to be assigned a carnet number and entered into the database.
Immunization data from Shoulder to Shoulder database and health center records
In a final assessment, we combined data from the database with data from the LINVIs, in order to create a more accurate and comprehensive table of immunization data. For children who overlapped and were present in both data sets, the most updated information was used. If a child received one oral polio vaccine according to a LINVI but three according to the database, we recorded three doses. We removed duplicate records in the cases of children who were present in two or more LINVIs, since this typically indicated children had moved to a different community.
The joint effects of database, township, and vaccine on percent immunization were evaluated by logistic regression. The main effects of each factor and all two-way interactions were modeled. F-statistics were generated for the following effects: (1) database differences overall, in each township, and for each vaccine; (2) township differences overall, from each database, and for each vaccine; and (3) vaccine differences overall, in each township, and between databases. We obtained publicly available immunization percentages from the WHO
] and compared these to mean immunization percentages across databases and townships for each vaccine using confidence interval inclusion. The analyses were implemented using SAS PROC GLIMMIX. An alpha level of 0.05 denoted statistical significance.