Cervical cancer is largely preventable through Pap testing with prompt and proper follow-up for abnormal results. Not all women, however, receive Pap tests at appropriate intervals, and Appalachia Ohio is an area where women are particularly at risk for Pap test non-adherence (8
). In this randomized trial, we tested the efficacy of an LHA intervention designed to increase the uptake of Pap test screening among women in need of the test who were recruited from 14 separate health clinics in Ohio Appalachia. The LHA intervention was hypothesized to promote Pap test screening by improving knowledge, reducing barriers, and increasing positive beliefs about screening. The impact of the intervention on knowledge, barriers, and beliefs is the topic of another manuscript; however, evidence from previous studies of LHA interventions documents this process (40
Overall, the results demonstrate an increase in screening among women who received the LHA intervention, although results from MRR and self-report differed in effect size and level of statistical significance of the intervention effect. Participants in the intervention group were more than twice as likely to report having had a Pap test within the study period (p=0.008). However, according to MRR, although a higher proportion of participants in the intervention group received a Pap test, the difference between the two arms was not significant (p=0.135). The effect size for our LHA intervention was similar to that reported by several previous studies using medical record review (17
), however, other studies have shown an even larger intervention effect (19
). Perhaps the lack of statistical significance for the intervention effect based on MRR in our analysis stemmed also from our definition of usual care. We defined usual care as receiving an NCI brochure and a letter from the participant’s physician. In reality, however, usual care may include neither, and we may have ‘watered down’ our intervention effect by defining usual care in such a manner.
Self-report bias in screening has been extensively studied, and likely explains why the overall proportion of women who reported having had a Pap test during the study period was notably higher than that confirmed by MRR (56
). However, self-report and MRR discordance between the intervention and control groups did not appear to be differential among those who had information for both report types. That is, a slightly higher proportion of women in the control
group who did not have a Pap test according to MRR reported having had a Pap test. For self-report bias to explain the difference in magnitude and significance of the intervention effect between self-report and MRR data in this study, the opposite would have had to be true.
Therefore, we sought alternate explanations for the discordance of results in MRR and self-report. It is likely that women who did not complete the follow-up survey were less likely to have had a Pap test within the study period. This fact may be particularly important among the women in the intervention group who may have refused the follow-up survey because they did not want to indirectly disappoint the LHA who had spent time trying to assist them during the past year. Thus, social desirability bias may have been introduced into the self-report population, and the self-reported intervention effect may have been artificially inflated.
Another possibility stems from the MRR of Pap test completion in the clinics in which this study was conducted. While in most settings MRR is the gold standard, Ohio Appalachia represents a unique setting. Many clinics in our study had inadequate resources to maintain complete and up-to-date medical records. Most of the clinics lacked computerized records, and many were understaffed in administrative positions. Still another site was damaged by local flooding and reported losing a portion of their medical records in the natural disaster. Thus, there is the possibility that for some of the women who reported having had a Pap test, but for whom there was no medical record of the test, a Pap test was performed but the record was lost or destroyed. In this scenario, the intervention effect in the MRR analysis could have been underestimated. The latter possibility should be considered in future studies that plan to use MRR as the gold standard in highly rural, understaffed, and underfunded clinics where lack of computerized medical record technology is prevalent.
Secondary outcomes showed that based on MRR data, having a Pap test in the previous three years was the strongest predictor of obtaining a Pap test over the course of the intervention. This underscores the importance of promoting regular screening by health care providers to establish patterns of healthy behavior. In addition, based on self-report, women in this study with a history of risky-sexual behavior (i.e., having had sex before age 18 or with a partner with an STI) were more likely to adhere to risk-appropriate screening guidelines, suggesting that providers may stress the importance of regular screening to high-risk women or that women who engage in risky sexual behavior may be more acutely aware of their cervical cancer risk. The finding that significant predictors of Pap adherence for MRR were dissimilar than those of self-report further highlights the need to examine outcomes of studies conducted in rural populations using a variety of methods.
This study is not without limitations. Primarily, medical records in the Appalachian Ohio clinics included in our study may have been incomplete. Computerized medical records were used in only two (14.3%) of the participating clinics, and other clinics reported having had problems finding or maintaining medical records for all of the women enrolled in our study. If records of Pap tests for participants were differentially lost, the effect of the intervention could have been dampened. Additionally, we did not have follow-up survey data or MRR data for all women which could have introduced bias, although we obtained this information for the great majority of participants. However, internal validity of this study was likely protected by randomization and statistical control for measured baseline demographics. Furthermore, results are not generalizable to populations outside of Ohio Appalachia and, specifically, the clinics who participated in our study.
In spite of limitations, this study is one of the few LHA interventions conducted in Appalachia, an area which suffers from health disparities. Moreover, we recruited women from 14 clinics across four regions of Ohio Appalachia, representing the great variation in this area. Additionally, we captured information on participant characteristics to assess factors associated with both the intervention and outcome. Lastly, this is one of few studies that compared MRR to self-reported Pap test receipt among an underserved population served by rural health clinics. Because Pap tests must be obtained from a health care provider, the use of medical records to verify receipt of Pap tests has traditionally been the gold standard for research. Data about Pap testing from Appalachian clinics have never been evaluated for use in research studies nor compared to self-reported Pap test adherence.
Ultimately, results from our study suggest increased uptake in Pap testing among women who were randomized to receive an LHA intervention. Thus, the LHA intervention appears to be transferable to the Ohio Appalachian population in need of Pap testing; however, future studies should explore methods to strengthen Pap test uptake in the Ohio Appalachian region. Indeed, results of MRR, which is typically considered the gold standard, were not significant. Future studies should continue to assess the effect and cost-effectiveness of LHAs on uptake of cervical cancer screening in vulnerable populations, evaluate the effect of varying definitions of usual care, and examine the reliability of MRR in clinic populations with poor administrative oversight, funding, and access to computerized medical charting.