Our principal finding is that adherence to WCC schedules decreases after 6 months of age in this low-income urban population. This drop in adherence in a largely publicly insured population (> 90% at each time point) has not previously been described, and reveals a missed opportunity in preventive health services for these high-risk children. Why this drop should occur specifically after 6 months is not clear from our data. Other studies have found that parents and providers have different expectations of well child visits, and that parents' expectations are poorly met . We hypothesize that care for younger infants is perceived as a higher priority, either because of greater parental uncertainty or greater perceived vulnerability of the infant. Other researchers have shown a trend toward greater utilization of preventive services under 1 year of age, as compared to after 1 year [9, 12], but did not address this trend specifically. Of concern, the great majority of well child visits in the age range we studied are associated with immunizations. In a population already at higher risk of under-immunization and unrecognized developmental delays, missed visits likely contribute to these problems. The rise in adherence in the 18-< 24 month age group may be due to the front-loaded structure of the EPSDT schedule: with only one recommended visit in that time period, it is easier to adhere to the schedule in that time period than 6-< 12 or 12-< 18 months, when 2 visits are expected in each.
In addition to the subjects' changing adherence trajectory, we found that adherence was predicted most strongly by maternal prenatal care adherence, number of other children, maternal marital status, and low income. Prenatal care adherence and birth order as predictors of adherence are consistent with previous findings [26, 28]. Birth order also influences early immunization status in this same cohort . We hypothesize that single mothers and mothers of more than one child may be less able to attend preventive visits due to competing needs of other children. An alternative but not exclusive explanation is that mothers of more than one child feel more confident in the care of their younger infants. In either case, mothers must feel there is value in a health service if they are to prioritize it among the many other needs of their children.
Sensitivity analyses testing the robustness of our findings against alternative specifications of the outcome and missing data maintained significant associations between parity, prenatal care adherence, and income, and adherence, depending on the specification. In contrast, while marital status maintained the direction of its effect, it was no longer significantly associated with the outcome. Further study is needed to clarify this relationship.
The finding that mothers in the lowest income bracket are more likely to adhere to WCC may be associated with the fact that these families are more likely to be eligible for other kinds of services. Indeed, mothers in this category were themselves more likely to have health insurance as the income threshold for insurance coverage in adults is considerably lower than for young children.
There are some limitations to this study. First, our study population was comprised primarily of African Americans (92%) from the Philadelphia metropolitan area. Hence, our results have limited generalizability to other races/ethnicities or to suburban or rural settings. While non-English speakers were excluded, they represent a small proportion of the Medicaid-eligible population in the hospital from which patients were recruited. The number of individuals lost to follow-up may have also introduced additional bias in the sample, however, comparisons between retained subjects and those lost do not reveal significant differences between them. Second, while we attempted to include only those subjects and time periods in which children received care principally within the EMR system, it is possible that some subjects may have attended well child visits outside the system to which we had access. Consequently, the rates of adherence we report may be an underestimate. However, our results are consistent with published adherence rates for similar populations [3–5]. Because the majority of Medicaid enrollees in the Philadelphia region are enrolled in managed care organizations which require an assigned primary care provider, patients should see that provider for all their primary care needs within a given period, which mitigates this underestimation. We restricted outcome data to those children in whom we could be reasonably confident that they had been consistently assigned to the practice from which we gathered the EMR data. Third, because this is a secondary analysis of a dataset not specifically designed to assess well child care adherence, the survey did not include maternal report of well child visits, which would obviate the need for abstracting data from the medical record and excluding subjects for whom records were not available, yielding a larger sample size.
Finally, as with any survey, there is potential for biased recall or social desirability bias. Among the predictors, the most likely to be sensitive to this would be income. Indeed, this was the variable with the greatest number of missing observations; however, our sensitivity analyses showed that our findings were robust to these missing data. Despite these limitations, this study furthers our understanding of predictors of WCC adherence in young children.