|Home | About | Journals | Submit | Contact Us | Français|
To examine dental utilization transition dynamics between 2004 and 2006 in the context of changing dental coverage status with data from the Health and Retirement Study (HRS) for persons ages 51 years and older.
We estimate a multivariable model of dental use transitions controlling for dental coverage and retirement transitions and other potentially confounding covariates.
We find that elderly persons losing dental coverage between the 2004 and 2006 survey periods were more likely to stop dental use between periods, and those gaining coverage were more likely to start dental use between periods, than those without coverage in both periods.
Regular dental care is imperative for maintaining good oral health. The likelihood of seeking dental care has been shown to be highly correlated with having health insurance coverage, meaning that persons who experience changes in dental insurance status may have irregular dental care utilization patterns. Indeed, it has been shown that persons who expect a change in their dental insurance status modify their use patterns to “stock up” prior to losing coverage. (1)
For most in the United States, dental insurance coverage is job-based, meaning that those who change jobs or leave employment altogether are most at-risk of changes in dental coverage status and therefore at highest risk of irregular utilization patterns. Working-age adults who lose dental coverage as a consequence of a change in employment status may be able to obtain coverage from a spouse or through another employer. However, older adults around the age of retirement may not have such options, as the vast majority become eligible for Medicare coverage at age 65, which, apart from a small but growing percentage of beneficiaries in Medicare Advantage programs, does not offer dental benefits (2,3,4).
Previous evidence has shown that those who are retired have lower levels of dental utilization and lower rates of coverage than those who are not retired (5–8). The transition from work to retirement is associated with a loss in dental coverage (6). Taken together, these findings imply that the transition from work to retirement may lead to irregular patterns of dental care utilization. This could be problematic if those nearing retirement age could avoid high-cost treatments later in life through regular preventive care (9).
The purpose of this article is to examine dental utilization transition dynamics in the context of changing dental coverage status among a population around the age of retirement. We use data from the Health and Retirement Study (HRS) to assess the characteristics of persons aged 51 years and older based on whether they had, maintained or changed their dental use status between the 2004 and 2006 waves of the HRS. In particular, we assess how changes in dental coverage and changes in retirement status affect the relative likelihood of having irregular dental utilization patterns.
The Health and Retirement Study (HRS), administered by the Institute for Social Research (ISR) at the University of Michigan and sponsored by the National Institute on Aging, is a longitudinal household survey useful for the study of aging, retirement, and health among older populations in the United States.(10) Response rates for the HRS are quite high; in 2004 the overall response rate for persons interviewed in previous waves was 95%, while the overall response rate (including among first-time sample members) was 88%.
The HRS contains a large battery of questions at the individual and household level, including information about demographics; income and assets; physical and mental health; cognition, family structure and social supports; health care utilization and costs; health insurance coverage; labor force status and job history; and retirement planning and expectations. The RAND Corporation has created an analytic file of key HRS variables that are consistent across waves of the HRS; those variables are used in this analysis when possible. HRS identifier variables contained in that file ensure that our analysis across survey waves is based on the same individual in both periods.
This analysis focuses on dental insurance coverage reported in the HRS for the two-year periods prior to the 2004 and 2006 surveys. We did not include earlier waves of the HRS in our analysis because dental coverage was not measured consistently prior to the 2004 HRS. Dental coverage in the 2004 and 2006 HRS was identified in one of two ways: either (1) the respondent reported seeing a dentist for dental care during the two-year period preceding the survey and having expenses at least partially covered by insurance, or (2) the respondent did not see a dentist but reported that they would expect any costs to be covered by insurance if he or she did need to see a dentist. Using the coverage data that were available, we calculated national estimates of the number of those persons 51 years and older covered by dental insurance in both survey periods, and those gaining or losing dental coverage between survey periods by retirement status and other characteristics.
In each two-year period, we use a binary measure of whether the individual had any dental care or not. To construct utilization transitions, we defined “stopping use” to include persons who had at least one dental visit in the two-year period preceding the 2004 HRS interview, but did not have use in the two years preceding 2006. “Starting use” was defined to be persons who did not have a dental visit in the period preceding 2004, but did in the period between 2004 and 2006. The other two potential transition categories included those reporting a visit in both 2004 and 2006, and those reporting not having a visit in either period.
Because dental insurance is oftentimes tied to one’s employer, understanding how coverage relates to retirement is important. As a generally healthier elderly population with a longer life span than previous generations increases, gradual transitions to retirement have become more common.(11) For this reason, we split retirement status into two categories: fully or partially retired. Survey respondents are designated as fully or partly retired based on employment, labor force, and self-reported retirement status variables in the RAND HRS. (10) Persons designated as fully retired in our study were screened to make certain that they were not self-employed or working for pay. Persons not identified as fully retired who report being partly retired or who report retirement and also report either working or looking for work are defined as partly retired. Persons not classified as fully or partly retired are designated as either in or out of the labor force. Those classified as in the labor force (1) report working for pay or (2) have a labor force status of working full-time, part-time, or unemployed. Those identified as not in the labor force report being disabled, not in the labor force, or having never been in the labor force.
We estimate a multivariable model of dental use transitions controlling for dental coverage and retirement transitions and other potentially confounding covariates. Using Andersen’s conceptual framework to guide the selection of these other covariates, we include self-reported 2006 HRS data for predisposing factors of age, gender, race/ethnicity, education, and household size; an enabling factor for income; and need factors of health and dentate status. (12) Given the dichotomous dependent variables for dental use transitions, we use logistic regressions to measure the association of coverage status between survey periods on dental use transitions while controlling for potential demographic and other confounders.
Previous research (5) confirms the correlation between variables such as income, education, retirement and dental coverage.(13–15) To omit them could potentially bias our parameter estimates of the impact of labor force transitions on dental coverage.
Observations with any missing data were omitted from the regression analysis. The HRS core sample design is a multistage area probability sample of households, so all estimates and statistics reported were computed taking into account this design with the use of the software packages SUDAAN and STATA. (16–17) The 2006 respondent weights in the HRS were used for all the estimates. Our study was reviewed by the University of Maryland (UMB) Institutional Review Board (IRB), and it was determined that the protocol does not require IRB review.
Our sample consists of 16,345 individuals interviewed in both the 2004 and 2006 HRS representing 74,047,165 members of the community-based population who were age 51 and above at the time of the 2004 interview. Excluded from this sample were 646 individuals who were not respondents in both 2004 and 2006 and 1,478 beneficiaries who either had zero weights and/or did not have dental visit data in both HRS waves. More than half of the participants were female (57.6%, N=9,410). Nearly fourteen percent (N=2,260) of the participants were non-Hispanic Black and 9 percent (N=1,471) were Hispanic. About 12 percent (N=1,892) of the participants were age 75 or older, 36.4 percent (N=5,952) were between the ages of 65 and 74, and 35.9 percent (N=5,864) were between the ages of 51 and 64.
Dental use transitions are reported by population characteristics (Table 1) and by retirement transitions and dental coverage transitions (Table 2). Labor force status and transitions are reported by population characteristics and by dental coverage and transitions in Table 3. Tables 4 and and55 show the adjusted and unadjusted odds ratio estimates of the probability of stopping and starting dental care use between the 2004 and 2006 survey periods. Unadjusted odds ratios were estimated from logistic equations without controls for other variables and provide a straightforward comparison to the adjusted logistic estimates incorporating controls. We focus on the adjusted estimates and point out that, unless otherwise noted, results for the unadjusted estimates did not differ from the adjusted estimates. Differences that do appear are typically caused by correlations between covariates present in the full regression models but omitted from the unadjusted models. Unless otherwise stated, all reported results are significant at least at the .05 level.
As shown in Tables 1 and and2,2, use patterns tended to be fairly consistent at the individual level. That is, the majority of persons who used dental care in 2004 also used care in 2006 (58.14 percent), though about one in four did not use care in either period (24.69 percent). However, there was also a fair amount of transition in utilization; approximately 13.5 percent of older adults with a dental visit in the two years prior to the 2004 survey did not have a visit in the two years prior to the 2006 survey (i.e. stopped use between the two survey waves). Approximately 25 percent of older adults without a dental visit in the two years prior to the 2004 survey did have a visit in the two years prior to the 2006 survey (i.e. started use between the two survey waves).
Table 3 shows that more than half of the older adults (55 percent) were out of the labor force in 2006. Most of them were out of the labor force in both 2004 and 2006 (47 percent) while another 8 percent of the elderly had left the labor force between periods. Table 3 also shows that a disproportionately high percentage of those who lost dental coverage between periods had exited the labor force (13.0 percent) between periods compared to those who were covered (7.7 percent) or not covered (7.3 percent) in both periods, or who had gained coverage between periods (7.1 percent).
In Table 4 the odds of stopping use between the 2004 and 2006 survey periods were lower for the 65 to 69 age group compared to the oldest age group (75 years and over). Hispanics and Black and Other non-Hispanics were more likely to drop use than White non-Hispanics. Females were found to be less likely to stop use than males. The odds of stopping dental use were higher for persons in the lowest three income groups compared to those with the highest incomes, and the odds of stopping use were higher for those elderly persons with a high school degree or less education compared to college graduates. Similarly, the likelihood of stopping use was higher for persons missing their permanent teeth, persons who were widowed or divorced, in households of three or more persons (unlike the unadjusted estimate), and for those in self-reported good or fair/poor health compared to persons with teeth, persons who were married, those in single-person households, and persons in excellent/very good health. Other household size and age effects found in the unadjusted estimates became statistically insignificant after controlling for other explanatory variables in the logistic model.
In comparison to the elderly who remained in the labor force between 2004 and 2006, the odds of stopping use were lower for those fully or partially retired over the two periods (unlike the unadjusted odds for both groups). Other retirement effects found in the unadjusted estimates became statistically insignificant after controlling for other explanatory variables in the logistic model.
The effect of coverage transitions was strong and persisted even after controlling for other confounders. In comparison to the elderly without dental coverage between 2004 and 2006, the odds of stopping use were higher for those losing coverage between periods, and lower for those with coverage in both periods.
In Table 5 the odds of starting dental use between the 2004 and 2006 survey periods were lower for individuals in families below the poverty line compared to high income individuals and for persons with a high school degree or less education compared to college graduates. Persons without teeth were also less likely to start use than persons with teeth as were individuals in fair/poor health compared to those reporting excellent/very good health. Age, marital status, household size, retirement, and other income and health status effects found in the unadjusted estimates were no longer significant in the multi-variable model.
Individuals who were covered in both periods or who gained coverage between the two periods were more likely to start use than persons without coverage in both periods.
In both the starting and stopping use models we tested for pair-wise interaction terms between age, income, and dentate status with retirement status using STATA stepwise logistic regression. In no case did we find any of the interaction terms, tested as a group, to be statistically significant at or below the 0.05 level.
The focus of our study is on transitions in dental use among an elderly population over a four year period between 2002 and 2006. Of the 74 million persons 51 years and over in the community population over this period, we found that three-fourths had at least one visit over this four year period. In most cases, persons who used dental services in one two-year period also did so in the next, and those who did not use services in the first period did not use them in the second period. However, nearly one in five older Americans either started or stopped dental use during our period of study, with about half stopping use and the other half starting use. Despite this the likelihood that a person without dental use in 2004 would start dental use in 2006 was nearly twice the likelihood of a person with dental use in 2004 stopping dental use by 2006.
Our findings show that after adjusting for coverage transitions and other explanatory variables, retirement transitions do not have an independent effect on use transitions, either stopping or starting use, in most cases. The exception is for the fully or partially retired in both periods who were actually less likely to stop use than those continuously in the labor force. Given that retirees may have more time to seek dental care than those who are in the labor force, this result is not entirely unexpected.
A key finding from our study is the increased likelihood of stopping use among those losing coverage and the increased likelihood of starting use among those persons gaining dental care coverage relative to those who remain without coverage between survey periods. This finding is not surprising given our cross-sectional findings from the same data in our previous research on the correlation between dental coverage and use. (7) We also find that elderly persons with dental coverage in both periods are less likely to stop use and more likely to start use, relative to the same reference group. Correlations between losing and gaining coverage and respective exits from and entrances into the labor force from Table 3 and from our previous research suggest reasons for these coverage transitions. (6)
Interestingly though, we also find an unexpected positive, though statistically insignificant coefficient estimate, for persons gaining dental coverage relative to those never covered in the stopping use equation. For most, gaining coverage occurs through a job-based arrangement, as Medicare, apart from those enrolled in Medicare Advantage programs, does not offer dental coverage, and Medicaid dental coverage is not uniformly available across states (2,3,4). In fact the State of California recently cut dental benefits under its Medicaid program for budgetary reasons. We looked at those gaining coverage in 2006 with dental use in 2004 and found that of those not stopping dental use in 2006, 57 percent were working for pay in 2006 while of those stopping dental use in 2006, only 48 percent were working for pay during that period. Those who gained coverage were less likely to stop dental use (19%) if they were working and were more likely to stop dental use (25%) if they were not working, so it does not appear that work interferes with arranging dental care. We do not fully understand why there is a tendency to stop dental use among those gaining coverage, so we will have to explore this finding in a future study. It may be that these individuals only seek dental care when needed. On the other hand, retirees may reenter the work force because of financial difficulties. Gaining coverage as a secondary benefit associated with returning to work may not be enough to outweigh the difficulty that caused the retiree to return to work.
In general our regression results were unchanged when we subset our sample on the basis of either dentate status, age, or income. The only exception we found was that the estimated coefficient for gaining dental coverage in the stopping dental use regression became statistically significant only for the subset of those 65 years and older.
The model results could be biased from omitting unobserved relevant variables measuring access to care or supply constraints on dental utilization. The lengthy two-year recall period in the HRS could affect the accuracy of selfreporting dental use and dental use transitions. Without data on clinical oral health status, type of dental coverage, and number and type of dental procedures in the HRS, it is difficult to fully understand why individuals stop or start dental use over a four-year period. Further insight into dental use transitions could be found by incorporating transitions in explanatory variables such as changes in dentate status or location, or by adding more waves of HRS data to the longitudinal model.
Further analysis is needed to identify the reasons for dental coverage changes and to assess the accuracy of measuring dental coverage in the HRS. We show an association between losing coverage and exiting from the labor force between periods, but there could be other reasons for coverage losses from the retirement of a spouse or reductions in hours worked, income or wealth that could be explored in the future. There may be confusion among Medicare beneficiaries regarding whether or not they have dental benefits under their Medicare coverage. We plan to analyze potential measurement error in dental coverage and its effect on our results when an improved HRS measure of dental coverage becomes available in the future.
While the transition to retirement does not have an independent effect on dental use transitions, retirement has previously been shown to be strongly associated with a loss in dental coverage. Therefore, the changes in dental use identified in our study may highlight the vulnerabilities that those reaching retirement age face. Forty-five percent of the 74 million individuals in our HRS sample were in the labor force in 2006, assuming that the partly retired in our sample retain some attachment to the labor force. This estimate is comparable to the actual 48 percent labor force participation rate (LFPR) for the civilian population 50 years and over in 2008, and in line with the projected 50 percent rate for the same population in 2015. For those 65 and older the actual LFPR of 16 percent in 2008 is projected to increase to 20 percent by 2015.(18,19,20) This increase may mitigate to some degree the tendency our study shows for elderly persons to stop their dental visits as they exit the labor force and lose their dental coverage at retirement.
Currently, Medicare does not offer a dental benefit, but covers the vast majority of retirees.  While many current retirees have supplemental retiree health insurance coverage (and possibly dental coverage through such a plan), the offers of such coverage have declined precipitously in recent years, meaning that future retirees will be more dependent on Medicare. Recent estimates show that nearly one quarter of Medicare beneficiaries are enrolled in Medicare Advantage plans with about 40 percent of these plans offering preventive dental care and close to 20 percent offering comprehensive dental care (2, 3). While the full implementation status of health reform is unclear, at this point it does not seem likely that a dental benefit will be added to Medicare in the near future and may even be eliminated if Medicare Advantage plans are phased out. State budgetary pressures make the addition of a dental benefit to Medicaid highly unlikely, and increase the possibility of such coverage being cut from state plans where it does currently exist.
Thus, in the future, it is possible that many more elderly adults reaching retirement age will experience a loss in dental coverage. This loss is associated with stopping use. While we were only able to look at a short time horizon with the HRS data and therefore do not know the longer-term use patterns of those who lose coverage around retirement age, even short-term lapses in preventive coverage can result in more invasive and costly procedures in the future. For retirees on fixed incomes, the high cost of dental procedures could have important financial consequences, and the delay of care could lead to worse overall health status and affect more than only dental costs.
This investigation (Dental Coverage Transitions, Utilization and Retirement) was supported by the National Institute on Aging (NIA) (R01AG026090) and the National Institute of Dental & Craniofacial Research (NIDCR) (R01DE021678) of the National Institutes of Health.
The HRS (Health and Retirement Study) is sponsored by the National Institute of Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan.
Richard J. Manski, Professor and Director, Division of Health Services Research, University of Maryland Dental School.
John Moeller, Research Professor, Division of Health Services Research, University of Maryland Dental School.
Patricia A. St. Clair, RAND Corporation.
Jody Schimmel, Mathematica Policy Research, Inc.
Haiyan Chen, Assistant Professor, Division of Health Services Research, University of Maryland Dental School.
John V. Pepper, Department of Economics, University of Virginia.