A new socioeconomic status measure for vaccine research in children using individual housing data: a population-based case-control study

Background We recently developed HOUSES, an individual housing-based socioeconomic status (SES) measurement for health disparities research. We assessed whether HOUSES was associated with risk of pertussis and pertussis vaccine up-to-date status in children. Methods The study utilized a previous population-based case-control study cohort assembled during the 2004–2005 pertussis outbreak. We collected data on pertussis vaccine status (up-to-date status) at the time of the index date. Using a z-score for housing value, actual square footage, and numbers of bedrooms and bathrooms, HOUSES was formulated in continuous variable and categorized into quartiles. Vaccine up-to-date status was compared among subjects with different SES as measured by HOUSES using a chi-square test and logistic regression models. Results Of the 391 eligible pediatric subjects (median age of 13.1 years with male sex of 55 %), 363 (93 %) were successfully geocoded to formulate HOUSES index. HOUSES was not associated with the risk of pertussis (p = 0.82). Pertussis vaccine up-to-date statuses were 79, 86, 83, and 94 % for children in the first (the lowest SES), second, third, and fourth quartiles of HOUSES, respectively (p = 0.03). HOUSES as a continuous variable was associated with pertussis vaccine up-to-date status (adjusted OR: 1.15 per increment of one unit of HOUSES, 95 % CI: 1.04–1.27, p = 0.008). Conclusion While HOUSES is not associated with the risk of pertussis, it predicts vaccine up-to-date status among children with different SES. HOUSES may be a useful tool for vaccine delivery research among children.


Background
Pertussis is re-emerging as an infectious disease causing significant morbidity, through the social burden of lost workdays and school-days, and mortality, especially among infants [1]. Pertussis outbreaks are on the rise with more than 48,000 reported cases per the CDC in the United States in 2012, as compared with 27,550 in 2010. These statistics are staggering when bearing in mind the historic low of 1010 cases in 1976. It is estimated that 50 % of pertussis cases affect children under the age of 16, a population in whom the effects of the infection can be serious [1]. In 2010, there were 6257 pertussis cases reported in California, including nine infants under 2 months of age who died having not received the pertussis vaccine [2].
Known risk factors for pertussis include household exposure to susceptible individuals, close proximity to a cough under 5 ft from a carrier, being female, and being an infant [3][4][5]. Recently, asthma has been shown to increase the risk of pertussis. Additionally, the efficacy of the acellular pertussis vaccine has been called into question and possibly implicated as an important factor in recent outbreaks [6,7], and notably in the 2012 outbreak in Southeast Minnesota [8], wherefrom our study cohort derives.
Socioeconomic Status (SES) is an important factor to be considered in understanding the epidemiology of all infectious diseases, including pertussis, because environment and resources impact multiple domains related to the risk of contracting and spreading an infectious disease. Indeed, the risk of acute and chronic respiratory diseases has been shown to increase among populations of lower SES, especially during childhood [9][10][11].
While SES is an important factor to be considered for the epidemiology of infectious diseases, it is frequently unavailable in commonly used datasets. As a result, area (census)-level SES measures are increasingly used as a proxy for individual SES [12][13][14][15][16][17][18]. For example, in our previous report showing the relationship between asthma and the increased risk of pertussis, only about 30 % of subjects had parental education levels available in medical records [19]. To address this concern, we developed an SES index derived from publicly available housing features including: square footage, assessed housing value, and the number of bedrooms and bathrooms in the residence where a patient lives [20]. This index, the HOUsing-based measure of SocioEconomic Status (termed as HOUSES), has proven its utility as a measure for SES in predicting various health outcomes including: low birth weight, obesity, household smoking exposure, invasive pneumococcal diseases, asthma control status, and general health status in children [20][21][22][23], as well as the risk and mortality of rheumatoid arthritis [24] and post-myocardial infarction mortality in adults [25].
Due to the large number of subjects with missing SES data in our previous study, we did not adequately address the role of SES in the risk of pertussis and upto-date pertussis vaccination status. Thus, we sought to assess the effect of HOUSES on the risk of pertussis and up-to-date pertussis vaccination status. We hypothesized that HOUSES might impact the risk of pertussis through differential pertussis vaccine up-to-date status (up-take rate) among children with different SES measured by HOUSES.

Methods
The study was approved by both the Institutional Review Boards at Mayo Clinic and Olmsted Medical Center.

Study population and setting
Rochester, Minnesota, is centrally located in Olmsted County. During the study period, characteristics of the City of Rochester and Olmsted County populations were similar to those of the U.S. Caucasian population, with the exception of a higher proportion of the working population employed in the health care industry [26]. Health care in Olmsted County, Minnesota, is virtually self-contained within the community. Uniquely, when patients first register care with any health care provider in the community, they are asked to authorize the use of their medical records for research. Those who grant such authorization (and 95 % do), are assigned a unique identifier with the Rochester Epidemiology Project (REP), a program which has been continuously funded by the NIH since 1966. All clinical diagnoses and information from every episode of care are contained within detailed patient-based medical records. This unique longitudinal population-based resource has been the source of over 2000 publications on the epidemiology of disease.

Study design
The study was designed as a population-based casecontrol study. We utilized a previous population-based study cohort, which included 164 pertussis cases and 328 age-and gender-matched controls among Olmsted County, Minnesota, residents during the 2004-2005 pertussis outbreak. In this study, we only evaluated the pediatric cohort, defined as individuals under 18 years of age at the time of the index date.

Study subjects: cases and controls
Details about the study cohort have been previously described [19]. Briefly, a case was defined as isolation of Bordetella pertussis by PCR from upper respiratory tract or serologic testing among Olmsted County residents between January 1, 2004, and December 31, 2005. An age-and gender-matched control group was randomly selected from a list of Olmsted County residents with negative PCR results within a month of the index date of a pertussis case. Two controls were matched for each case with regard to gender and birthday (within 1 month for patients younger than 2 years of age, within 3 months for patients between ages 2 and 5 years, within 6 months for patients between ages 6 and 12 years, and within 1 year for patients older than 12 years). Additional criteria for controls required that participants be residents of Olmsted County for least 1 year prior to the study period; the same exclusion criteria for the cases were also applied.

Dependent variable
We collected pertussis vaccination status at the time of the index date of pertussis for both cases and controls from the medical record (DPT, DTaP, and Tdap). We defined pertussis vaccine as up to date if subjects completed all age-appropriate pertussis vaccinations as recommended by the Advisory Committee on Immunization Practices (ACIP). In general, up-to-date vaccine status meant one through five immunizations appropriate for age in the DTap (or DTP before 1997) series for children younger than 6 years of age. Since routine use of Tdap for adolescents aged ≥11 years old was recommended by ACIP in March, 2006, we did not consider Tdap for up to date vaccine status in the present study [27].

Socioeconomic status measurement (independent variable)
We utilized two individual-level SES variables including maternal and paternal education level and an individualhousing based SES measure (HOUSES). For both cases and controls, maternal and paternal education levels and home addresses were abstracted from the entire medical record. HOUSES is a composite index derived from housing features of real property data and home address information as charted in the patient's medical record at the time of their pertussis event. Development and initial testing of the index were completed in both Olmsted County, Minnesota, and Jackson County, Missouri. Results from that study have been reported in previous publications [20][21][22][23]. Briefly, in calculating HOUSES, each subject's address at the index date of their pertussis event was geocoded. Geocoding allowed us to match study subject address to real property data. Once geocoded, we applied principal components factor analysis on the basis of real property data features of housing and neighborhood SES items. Factor analysis results were pared down to four real property feature variables, including market housing value, square footage of the housing unit, the number of bedrooms, and number of bathrooms. We then formulated a standardized HOUSES index score by transforming the four variables to z-scores (i.e., standardized the index to allow for comparisons across different study settings) and summing the z-scores to the HOUSES index. Essentially, a higher HOUSES index score represents a higher SES.

Other pertinent variables
Other information abstracted from the medical records included sociodemographic variables: age, gender, ethnicity, parental education levels, birth weight, gestational age, comorbid conditions, and smoking status at the time of index date (either active or passive exposure to tobacco smoke).

Data analysis
HOUSES was analyzed as both a continuous variable and as categorical variables (quartiles). We compared the frequency of subjects with different SES measured by HOUSES in quartiles between cases and controls to determine relationship between HOUSES and the risk of pertussis. Similarly, we determined the relationship between HOUSES and pertussis vaccine up to date status. Adjusting for age and sex, data were fitted to logistic regression models to determine the association of HOUSES with the risk of pertussis and pertussis vaccine up-to-date status by calculating odds ratios and their corresponding 95 % confidence intervals (CI). Similar analysis was performed for association of maternal and paternal education level. The weighted kappa index for the agreement between HOUSES and maternal and paternal education levels in quartiles was also calculated.

Study subjects
The characteristics of the subjects included in the original case-control study are summarized in Table 1

Socioeconomic status and the risk of pertussis
The proportions of pertussis cases in the first, second, third, and four quartiles of HOUSES were 28, 33, 31 and 31 %, respectively (p = 0.91). Of the 130 pediatric pertussis cases, 102 (78.5 %) had pertussis vaccine up-to-date status whereas 216 of 261 control subjects (82.8 %) had vaccine up-to-date status (p = 0.158). Maternal (p = 0.40) and paternal (p = 0.19) education levels were not associated with the risk of pertussis.

Socioeconomic status and pertussis vaccine up-to-date status
Out of 55 subjects who missed a pertussis vaccine, 31 missed at least one DTP and 28 at least one DTaP, including 4 who missed both. The pertussis vaccine upto-date statuses were 63 (79 %), 74 (86 %), 70 (83 %), and 91 (94 %) for children in the first (the lowest SES), second, third, and fourth quartiles (the highest SES) of HOUSES, respectively (

Discussion
Our study results showed that conventional SES measures such as parental education levels are not commonly available in medical records, making it difficult to apply them to clinical studies. While HOUSES index is not associated with the risk of pertussis, it predicts pertussis vaccine upto-date status suggesting HOUSES index might be a useful tool for vaccine delivery research. Previous studies have shown epidemiologic risk factors for pertussis such as gender [28], vaccination status [28], day care [29][30][31][32][33], age [5,34], minority ethnic groups, and exposure to tobacco smoke [35]. While vaccination can prevent pertussis infection [36,37], there is little literature demonstrating reliable factors that predict who gets the vaccine, or describing populations that carry greater percentages of individuals under-vaccinated for the pertussis vaccine. While lower SES is often related to limited access to health care, there have been few studies directly linking SES to pertussis vaccine uptake, and the studies that do exist use SES proxies such as occupation, smoking status, and parental income [38]-variables which we have found to be temporally dynamic and fluctuant, inaccurately recorded in medical records, and frequently absent. For example, in this study dataset, subject ethnicity was 18.7 % unknown and parental education level was 66-75 % unknown. The unavailability of conventional SES measures for clinical research has been a significant barrier to health disparities research, in particular for research on pediatric vaccination.
HOUSES shows promise to overcome the unavailability of conventional SES measures in clinical and administrative datasets and, therein, may enhance a large-scale health disparities research based on administrative or clinical datasets concerning vaccine delivery as shown in our study results. Again, of our 391 pediatric subjects, the addresses for 363 (93 %) subjects were successfully geocoded and formulated for HOUSES index while maternal and paternal education levels were only available for 34 and 25 % of subjects, respectively. As such, maternal and paternal education levels in our study were not associated with pertussis vaccine up to date status. This lack of association might be due to a type 2 error stemming from missing data points in a significantly reduced sample size. In addition, the pattern of missing was not at random. To underscore the point, a large prospective study by Henderson, et al., did show an association between maternal education and pertussis vaccination [39]. Importantly, while maternal education levels did not show any patterns in missing data in relation to SES measured by HOUSES, paternal education levels among children from lower SES background measured by HOUSES were more likely to have missing data on paternal education levels than those from higher SES (data not shown). Therefore, conventional SES measures such as parental education level are frequently unavailable for clinical research and available SES measures might potentially bias the results. In this respect, HOUSES offers a propitious new model for SES measures with superior availability.
In addition, the ease of formulating HOUSES further supports its utility in disparities research concerning vaccine delivery. The weighted kappa index for the agreement between HOUSES and maternal and paternal education levels in quartiles was 0.30 (95 % CI: 0.22-0.37) and 0.38 (95 % CI: 0.25-0.51) suggesting moderate concordance. This result suggests that there is little redundancy in capturing one's SES measured by HOUSES index compared to other SES measures and HOUSES can be a supplementary SES measure even if other SES measures are available.
Regardless of the merit of HOUSES as compared to other SES metrics, HOUSES was associated with pertussis vaccine up-to-date status with an adjusted OR of 1.15  per increment of one unit of HOUSES. The association between HOUSES index and pertussis vaccine up-todate status was almost a dose-response relationship as shown in Table 3. The difference in pertussis vaccine up-to-date status between the first and fourth quartiles of HOUSES index was 15 %. When we adjusted the results for potential confounding factors such as comorbidity, asthma status, and even the risk of pertussis (a proxy for unknown risk factors for pertussis), the results were still significant. However, although HOUSES was associated with vaccine up-to-date status, it was not associated with the risk of pertussis. One possible explanation for this finding has to do with the waning efficacy of the pertussis vaccine over time as recently reported. For example, a previous study reported waning humoral immunity in adolescents [7]. Given that the median age of our pediatric subjects was 13.1, it is possible that a large proportion of our adolescent population may have had waned antibody titers after their initial primary series of pertussis vaccine [6]. This could be why current pertussis epidemiology primarily affects adolescents as shown in our community and elsewhere [7].
One important advantage of the HOUSES index over other conventional SES measures is its built-in ability to facilitate a geospatial analysis to identify "hot spots" scaled by vaccination rate (vaccine topography). The HOUSES index is linked to a parcel identification number, the unique identification code for a parcel containing one (in the case of single family dwellings) or more residential units. Further studies might also examine the relationship of HOUSES to vaccine up-to-date status. The flu vaccine, for example, with an associated geospatial analysis of "hot spots" in a community where flu transmission is more likely, could facilitate real-time intervention by a public health team by contacting or reminding those without vaccinations, or possibly setting up extra vaccination sites in the geographical location of greatest need as demonstrated by HOUSES.
The strength of our study addresses an apparent need in health research: the paucity of readily available SES measures in commonly used clinical datasets. Our novel application of housing data to assess pertussis vaccination up-to-date status eliminates recall and social desirability bias of prior indices for measuring SES. This study is based on a previous population-based casecontrol study cohort and has epidemiological advantages as our study population is in a relatively self-contained geographic location and all medical events are linked to each resident and health care provider through the work of the REP program. This increases the likelihood that our dataset includes almost all the cases of pertussis that occurred during the outbreak of 2004-2005.
However, as a retrospective study, there are inherent limitations in validity, reliability, and availability of pertinent data as shown in parental education levels. Furthermore, Olmsted County, Minnesota, is a relatively homogenous population with a small minority population; this too limits the external validity of our findings. Our study only included parental education levels as a reference SES measure as other measures were not available. Furthermore, while HOUSES can explain some aspects of the material deprivation of families, it cannot describe or estimate the deleterious effects of immaterial deprivation (psychosocial, etc.). HOUSES is best understood as a tool to look at the association between

Conclusion
In conclusion, our results demonstrate a significant association between socioeconomic status, as measured by HOUSES, and pertussis vaccination up-to-date status. Further studies are necessary to apply HOUSES to the uptake of other vaccines such as the influenza vaccine and to develop strategies for improving vaccine uptake rate through targeting (or hot spotting) underserved populations.
Abbreviations HOUSES: HOUsing-based measure of SocioEconomic Status; OR: Odds ratio; SES: Socioeconomic status