Changes in the Incidence and Duration

7/28/2019 Changes in the Incidence and Duration

1/9

special article

T he n e w e n g l a n d j o u r n a l o f medicine

n engl j med 360;17 nejm.org april 23, 20091740

Changes in the Incidence and Duration

of Periods without InsuranceDavid M. Cutler, Ph.D., and Alexander M. Gelber, Ph.D.

From the Department of Economics, Har-vard University, and the National Bureauof Economic Research both in Cam-bridge, MA. Address reprint requests to Dr.Cutler at the Department of Economics,

Harvard University, 1875 Cambridge St.,Cambridge, MA 02138, or at [email protected].

N Engl J Med 2009;360:1740-8.Copyright 2009 Massachusetts Medical Society.

A B S T RA C T

Background

Policymakers have recently proposed ways of providing health care coverage for anincreased number of uninsured persons. However, there are few data that show howthe incidence and duration of periods in which persons do not have insurance havechanged over time.

Methods

We used two data sets from the Survey of Income and Program Participation of theU.S. Census Bureau: one that covered the period from 1983 through 1986 (25,946persons), and another that covered the period from 2001 through 2004 (40,282 per-sons). For each set of years, we estimated the probability that a person would beuninsured for some period of time and the probability that a person would subse-quently obtain private or public insurance. We also estimated the probabilities thatpersons in various demographic groups would become uninsured over the courseof a year and would remain uninsured for various amounts of time.

Results

The percentage of the population that lost insurance in a 12-month period increasedfrom 19.8% in 19831986 to 21.8% in 20012004 (P = 0.04). The percentage thatwas uninsured for a period of time increased markedly among persons with thelowest educational level and predominantly represented loss of private coverage.The percentage of new uninsured periods that ended within 24 months increasedfrom 73.8% to 79.7% between the two study periods (P


2/9

Changes in the Incidence and Duration of Periods without Insurance

n engl j med 360;17 nejm.org april 23, 2009 1741

The number of Americans who do nothave health insurance for an entire yearhas increased markedly in recent years,

from 32 million (15% of the nonelderly popula-tion) in 1988 to 45 million (17% of the nonelderlypopulation) in 2007.1 Lack of insurance for anentire year reflects only one dimension of insur-

ance coverage, however. Data from the 1980s sug-gested that about twice as many people were unin-sured at some point in a year than were uninsuredfor the entire year.2

Being uninsured (i.e., having what we term anuninsured period), even temporarily, is harmfulto a persons health.3,4 Uninsured periods of longerduration have a larger adverse effect on healththan periods of shorter duration5-7 and may alsobe concentrated in a group that is initially lesshealthy.3 The status of the uninsured has becomeparticularly important as policymakers consider

incremental expansions of health coverage, suchas enhanced access to coverage through the Con-solidated Omnibus Budget Reconciliation Act(COBRA), which currently ensures coverage foronly 18 months.8

Over the past several decades, economic andsocial factors have contributed to both longer andshorter durations of uninsured periods. The in-crease in temporary and contingent work,9 alongwith increased cost-sharing for insurance,10 maylead to more long-term uninsured periods. On theother hand, eligibility for public insurance hasbeen expanded to include not only single womenand children in low-income families but alsodual-parent families and families at higher in-come levels, a change that should lead to shorteruninsured periods. The net change over time inthe incidence and duration of periods withoutinsurance is therefore uncertain.

Data from the National Survey of AmericasFamilies (NSAF) show that there were relativelysmall changes in the distribution of insurancecoverage between 19981999 and 20012002, with

a slight trend toward shorter uninsured periodsamong children11; however, the period of timecovered by that study (3 years) is limited. Datafrom the Survey of Income and Program Partici-pation (SIPP) of the U.S. Census Bureau in themid-to-late 1980s and early 1990s show that me-dian durations of uninsured periods were in therange of 4 to 8 months, with no discernible trendover time.12-15 However, these analyses have notbeen updated with the use of more recent SIPP

data. Research on the loss of insurance has nottracked its evolution over time, nor has a hazardmodel for losing insurance been estimated.16

We examined changes over time in the inci-dence and duration of periods without insurance,using data from 1983 through 1986 and 2001through 2004. The overall economy was better in

20012004 than in 19831986, but in each case itwas in recovery from a recession. We estimatedhazard models for both gaining and losing insur-ance, controlling for employment changes in or-der to minimize the effect of economic factors,and examined population groups separately ac-cording to age and educational level. Finally, weevaluated transitions to and from public and pri-vate insurance in the two survey periods.

Methods

Data Set

We used data from the SIPP, a longitudinal, strat-ified, random sample of families in the UnitedStates.17 All participants provided informed con-sent. The SIPP full panel that began in the fall of1983 (the 19831986 SIPP) enrolled persons fromOctober 1983 through January 1984 and followedthem for 32 months, whereas the SIPP full panelthat began in 2001 (the 20012004 SIPP) enrolledpersons from October 2000 through January 2001and followed them for 36 months. We addressedthis difference in panel lengths by estimating ahazard model for the duration of uninsured peri-ods. To make length-of-time windows equal forthe two survey periods, when we analyzed data onwhether the person was uninsured at all, we useddata from only the first 32 months of the 20012004 SIPP panel. The use of panels of data cover-ing at least 32 months allowed an analysis of lon-ger periods than those that could be analyzedwith the use of the Current Population Survey orthe NSAF. Survey participants were asked about anumber of socioeconomic variables, including age,

race or ethnic group, sex, geographic location,education, work income, number of hours workedper week, capital income, net wealth, and partici-pation in government programs, such as Medi-care, Medicaid, and the State Childrens HealthInsurance Program (SCHIP). Every 4 months (theperiod of time that was termed a wave), respon-dents were asked to report whether each memberof the family had had health insurance in each ofthe previous 4 months and, if so, what type of

The New England Journal of Medicine

Downloaded from nejm.org on February 28, 2013. For personal use only. No other uses without permission.

Copyright 2009 Massachusetts Medical Society. All rights reserved.


3/9

Th e n e w e n g l a n d j o u r n a l o f medicine


health insurance. In the third wave of the survey(12 months from the beginning of the survey pe-riod), persons 15 years of age or older reportedtheir health status on a f ive-point scale in whichthe possible responses were excellent, very good,good, fair, and poor.

We used the educational level of the familys

highest earner as the main socioeconomic vari-able of interest. We did not use income to catego-rize families, since a persons cash income maydecline with increased health benefits,18,19 andeducational level is a better proxy for long-termearnings.20,21

Our sample was restricted to persons whowere 61 years of age or younger at the time ofthe first survey for each of the 19831986 and20012004 panels. The age restriction ensuredthat no one would become old enough during thesurvey period to be eligible for Medicare. We also

excluded persons who were removed from thesample when the 19831986 panel was reducedowing to budget cutbacks and persons who werein the military or who were receiving veteranspayments, as well as their relatives, since they arecovered by TRICARE (formerly known as the Civil-ian Health and Medical Program of the UniformedServices [CHAMPUS]) or the Civilian Health andMedical Program of the Department of VeteransAffairs (CHAMPVA). These exclusions resulted ina final sample of 25,946 people in the 19831986panel and 40,282 people in the 20012004 panel.The sample of persons with an uninsured periodcomprised 8706 persons in the 19831986 paneland 14,443 persons in the 20012004 panel whoreported being uninsured for some period of time(and who met the other criteria listed above).

Statistical Analysis

We used a Cox proportional-hazards model to es-timate the probability of becoming uninsured in19831986 and 20012004, using as covariates apersons age at the time of the first survey (


4/9



being uninsured are shown for each survey inTable 2. The coefficient estimates were in the ex-pected direction in all cases. Whites and personswith a higher educational level generally had low-er hazard rates for loss of coverage than personsof other races or ethnic groups and those with alower educational level. When they obtained cov-erage after an uninsured period, younger persons,persons with a lower educational level, and non-whites were more likely to obtain public coverage.

Probability of Losing Coverage

The predicted probability that an insured personwould lose coverage at some time in the subse-quent 12-month period, with the demographicmix of the population held constant, is shown inTable 3. The percentage of persons who lost cov-erage in a 12-month period increased from 19.8%in 19831986 to 21.8% in 20012004 (P = 0.04).The increase was particularly large among per-

sons with a lower educational level an increaseof 8.3 percentage points among persons withouta high-school or GED diploma (P


5/9



component of older white men who had a historyof employment in manufacturing industries andsecure health insurance coverage. By the early2000s, the people in that group had largely re-tired and, because of our age criterion for eligi-

bility, were no longer included in the sample;the group therefore comprised younger persons.Nevertheless, the greatly increased incidence ofloss of insurance among persons with a lowereducational level is not a consequence of demo-

Table 2. Hazard Ratios for Losing or Gaining Private or Public Insurance, 19831986 and 20012004.*

VariableHazard Ratio for Losing

Insurance (95% CI)Hazard Ratio for Gaining

Insurance (95% CI)

19831986(N = 24,858)

20012004(N = 38,704)

19831986(N = 8,706)

20012004(N = 14,443)

Private coverage

Age

1861 yr 1.00 1.00 1.00 1.00


6/9



graphic change. When only persons 20 to 40 yearsof age were included in the analysis, the percent-age of persons with less than 12 years of educa-tion who became uninsured over a 12-month in-terval increased from 24.2% in the 1980s to 35.1%in the 2000s (P


7/9



health resulted primarily from the fact that theywere more likely than those in better health toobtain public insurance. Tables 2 and 3 in theSupplementary Appendix show hazard ratios withrespect to health status.

Discussion

Our analysis of uninsured periods suggests sev-eral conclusions. First, the incidence of uninsuredperiods is rising over time, especially among per-

sons with a lower educational level. The majorityof uninsured periods involve the loss of privateinsurance, but an increased chance of losing pub-lic coverage over time accounts for the increasein loss of insurance from the earlier survey periodto the later period. Second, when people becomeuninsured, they are uninsured for shorter periods.The percentage of persons who were uninsuredfor 2 years or longer, for example, decreased from26.2% to 20.3% between the two study periods.Third, uninsured periods are shorter because more

Table 4. Duration of Uninsured Periods, 19831986 and 20012004.*

Variable Duration of Uninsured Period

0 Months 14 Months 512 Months 1324 Months >24 Months

percent19831986

Total sample 66.1 40.8 18.4 14.6 26.2

Age


8/9



people are obtaining public insurance. From themid-1980s to the early 2000s, the percentage ofuninsured periods that ended in coverage withprivate insurance decreased by 6.0 percentagepoints, whereas the percentage that ended incoverage with public insurance increased by 11.7percentage points. Fourth, in the 20012004 sur-vey, persons who were in fair or poor health weresubstantially more likely than persons in good,very good, or excellent health both to lose and togain insurance. This finding is a result of theincrease in public insurance, which persons whoare less healthy are more apt to have.

The increase in the percentage of the popu-lation that was uninsured in any month was

smaller in our analysis than in other analyses ofSIPP data,25 and it is smaller than the increasein the percentage of the population without cov-erage for an entire year that was shown in theCurrent Population Survey.26,27 This differenceis largely a result of the sample weights. The lon-gitudinal and cross-sectional weights in the 20012004 survey give somewhat different estimatesof the percentage of persons who were uninsuredin any month (see the Supplementary Appendix).Since the longitudinal weights are most appro-

priate for the duration analysis, we presented re-sults using these weights. The main results of theanalysis were substantively unchanged when thecross-sectional weights were used.

The increased incidence of uninsured periodsamong persons with the lowest educational levelas compared with those who were more educat-ed parallels and reinforces the economy-wide in-crease in income inequality during this time.28It may also reflect the fact that employers arecharging employees increasing amounts for healthinsurance and that there is a trend toward in-creases in temporary and seasonal employment.9We have performed the analysis using incomeranges in place of educational levels, with similar

results.Many people become uninsured because theylose their job; however, changes in job durationdo not explain our results. When we estimatedour hazard models with adjustment for the em-ployment status of a person and other membersof his or her family, we found a decrease in thelength of uninsured periods between 19831986and 20012004 that was similar to the decrease inthe model that did not control for employmentstatus (data not shown). Similarly, we found that

15

Changein

Probability(percentagepoints)

5

10

0

5

10

150 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2423 2625 27 28 29 30

Public

Private

Months

l

Figure 1. Changes between 19831986 and 20012004 in the Probability of Obtaining Private or Public Insuranceafter Losing Insurance.

On the basis of the hazard model, the probability of obtaining public or private insurance in a given month was esti-mated for each demographic group, with the estimate weighted by the average population share of the demographic

group over both survey periods, and the estimates for the various demographic groups were then added together toobtain the estimate for the overall sample.



Copyright 2009 Massachusetts Medical Society. All rights reserved.


9/9



the increase between 19831986 and 20012004in the probability of losing insurance was similarto the increase in the model that did not controlfor employment status.

The decrease in the duration of uninsuredperiods is not surprising, given the enormous in-crease in Medicaid coverage over time, particu-

larly for children and parents with a lower educa-tional level.29 It is possible that increased Medicaidcoverage could also be partly responsible for therelatively large decrease in private coverage, if in-creased eligibility for public coverage led to anincreased number of people who no longer hadprivate coverage. Our results do not address thisissue.30 We did find that many people, especiallychildren and persons who were less healthy, cy-cled into and out of Medicaid.

Our study has some important limitations.First, the SIPP is subject to seam bias that

is, the tendency for people to report the sameinsurance status in blocks of 4 months. However,

this bias should be the same in each survey.Second, different weights were used in the twosurveys, an issue noted above. Third, our analy-sis focused only on the incidence and durationof periods of no insurance; we did not examinethe health outcomes of persons who were unin-sured. Finally, we looked at data for uninsured

persons and did not analyze data for those whowere underinsured.

In conclusion, uninsured periods are moreprevalent, but shorter, now than they were twodecades ago. A decline in private coverage, espe-cially for persons with a lower educational level,has been offset by an increase in public coverage.Serious problems could lie ahead if employer-based coverage continues to decline while theavailability of public coverage remains the sameor is reduced.

Supported by grants from the National Institute on Aging

(PO1AG005842, to Dr. Cutler, and T32AG000186, to Dr. Gelber).No potential conflict of interest relevant to this article was

reported.

References

DeNavas-Walt C, Proctor BD, Smith1.JC. Income, poverty, and health insurancecoverage in the United States: 2007. Wash-ington, DC: Government Printing Office,2008.

Swartz K. Dynamics of people with-2.out health insurance: dont let the num-bers fool you. JAMA 1994;271:64-6.

Institute of Medicine, Committee on3.the Consequences of Uninsurance. Care

without coverage: too little, too late. Wash-ington, DC: National Academy Press, 2002.

Hadley J. Sicker and poorer the4.consequences of being uninsured: a re-view of the research on the relationshipbetween health insurance, medical careuse, health, work, and income. Med CareRes Rev 2003;60:Suppl:3S-75S.

Ayanian JZ, Weissman JS, Schneider5.EC, Ginsburg JA, Zaslavsky AM. Unmethealth needs of uninsured adults in theUnited States. JAMA 2000;284:2061-9.

Baker DW, Sudano JJ, Albert JM, Bo-6.rawski EA, Dor A. Lack of health insur-ance and decline in overall health in latemiddle age. N Engl J Med 2001;345:1106-12.

Idem.7. Loss of health insurance and therisk for a decline in self-reported healthand physical funct ioning. Med Care 2002;40:1126-31.

Bodenheimer T. Insuring the unin-8.sured: will the 2004 election provide ananswer? Ann Intern Med 2004;141:556-61.

Farber HS, Levy H. Recent trends in9.employer-sponsored health insurance cov-erage: are bad jobs getting worse? J HealthEcon 2000;19:93-119.

Chernew M, Cutler DM, Keenan PS.10.Increasing health insurance costs and thedecline in insurance coverage. Health ServRes 2005;40:1021-39.

Zuckerman S, Haley J. Variation and11.trends in the duration of uninsurance.Washington, DC: The Urban Institute,2004.

Bureau of the Census. Health insur-12.ance coverage, 1992 to 1993: who loses

coverage and for how long? Washington,DC: Government Printing Office, 1996.

Idem13. . Health insurance coverage, 1991to 1993. Washington, DC: GovernmentPrinting Office, 1995.



Idem.16. Health insurance coverage, 1993to 1995: who loses coverage and for howlong? Washington, DC: Government Print-ing Office, 1998.

Survey of income and program par-17.ticipation (SIPP). Washington, DC: Cen-sus Bureau. (Accessed March 30, 2009, athttp://www.sipp.census.gov/sipp/.)

Summers LH. Some simple econom-18.ics of mandated benefits. Am Econ Rev1989;79:177-83.

Gruber J. The incidence of mandated19.maternity benef its. Am Econ Rev 1994;84:622-41.

Solon G. Intergenerational mobility in20.the labor market. In: Ashenfelter O, CardD, eds. Handbook of labor economics. Am-

sterdam: Elsevier Science BV, 1999:1761-800.

Haider S, Solon G. Life-cycle variation21.in the association between current andlifetime earnings. Am Econ Rev 2006;96:1308-20.

Lunn M, McNeil D. Applying Cox re-22.gression to competing risks. Biometrics1995;51:524-32.

Kalbfleisch JD, Prentice RL. The sta-23.

tistical analysis of failure time data. 2nded. Hoboken, NJ: John Wiley, 2002.

Stata, ed. 9.2. College Station, TX: Stata24.Corporation, 1996.

Bhandari S. People with health insur-25.ance: a comparison of estimates from twosurveys. Washington, DC: Census Bureau,2004.

General Accounting Office. Private26.health insurance: continued erosion ofcoverage linked to cost pressures. Wash-ington, DC: Government Printing Office,1997.

DeNavas-Walt C, Proctor BD, Lee CH.27.Income, poverty, and health insurance cov-erage in the United States: 2005. Wash-ington, DC: Government Printing Office,2006.

Autor DH, Katz LF, Kearney MS. The28.polarizat ion of the U.S. labor market. AmEcon Rev 2006;96:189-94.

National Center for Health Statistics.29.Health, United States, 2006. Washington,DC: Government Printing Office, 2006.

Cutler DM, Gruber J. Does public in-30.surance crowd out private insurance? Q JEcon 1996;11:391-430.Copyright 2009 Massachusetts Medical Society.



Changes in the Incidence and Duration

Documents

Transcript of Changes in the Incidence and Duration