Health insurance coverage is important for enabling access to medical care, but obtaining quality care depends also on maintaining stable patient-provider relationships [1, 2]. Among the most important of these relationships are those between patients and usual source of care providers (USCP), who help monitor preventive care and manage chronic illness. Gaps in insurance coverage can disrupt these relationships, which take time to establish, and therefore may affect access even after coverage is regained. While previous studies have demonstrated that individuals with insurance disruptions report worse access to care compared to those insured continuously [3–6], additional studies are needed to determine whether this reflects barriers to care experienced while uninsured, or whether prior coverage disruptions have lasting effects even after coverage is regained. In this study, we investigated the association between having a USCP and past disruptions in insurance coverage among insured adults using a longitudinal, nationally representative sample.
METHODS
We used data from the Medical Expenditure Panel Survey (MEPS), a nationally representative household survey of the U.S. non-institutionalized population. At five in-person interviews, the MEPS collects information on health insurance coverage, utilization, expenditures, and experiences with care covering a 2-year period. A new panel is drawn every year and followed for 24 months. We pooled the most recent panels (2012–2017) and restricted our sample to adults ages 18–64 years with health insurance coverage at the fourth interview, which resulted in a sample of 33,030 adults.
The main outcome variable records whether individuals had a USCP at the fourth interview, conducted late in the second year of each panel. Health insurance status (whether someone has coverage) was reported at each interview and recorded for each of the 24 months. Using these monthly indicators, we grouped people into five mutually exclusive categories based on the number of months they were uninsured prior to round 4: 0 (continuously covered), 1–3 months, 4–7 months, 8–11 months, or ≥ 12 months.
We estimated the association between the number of months without insurance before the fourth interview and having a USCP at the fourth interview using a multivariable logistic regression model controlling for survey year, age, sex, self-rated health, household income relative to the federal poverty line, and attitudes about risk and health insurance. All estimates were weighted to be nationally representative and variance estimates were adjusted for the complex sample design of MEPS using Taylor series linearization [7].
RESULTS
Compared to people who were continuously insured, people with previous insurance coverage disruptions, even short ones, were less likely to have a USCP (Table 1). For example, only 68.8% of individuals who were uninsured for 1 to 3 months had a USCP, compared to 79.7% of those who were continuously insured. Among individuals who were uninsured for 12 or more months, only 60.4% had a USCP.
Table 1.
Sample Size and Characteristics of Insured Adults Ages 18–64 by Number of Months Uninsured Prior to the Interview (95% Confidence Intervals in Parentheses)
| Duration of health insurance coverage disruption prior to round 4 interview | |||||
|---|---|---|---|---|---|
| Continuously insured | 1–3 months | 4–7 months | 8–11 months | ≥ 12 months | |
| Sample size | 27,368 | 1916 | 1658 | 996 | 1092 |
| Percent with a USCP | 79.7% | 68.8%*** | 65.6%*** | 60.6%*** | 60.4%*** |
| (78.8%, 80.6%) | (65.6%, 72.0%) | (62.5%, 68.6%) | (55.8%, 65.4%) | (55.8%, 65.0%) | |
| Percent of sample from each MEPS panel | |||||
| Panel 17 | 18.8% | 15.5%** | 17.2% | 20.6% | 18.9% |
| (17.9%, 19.7%) | (12.9%, 18.1%) | (14.6%, 19.7%) | (16.8%, 24.5%) | (14.8%, 23.0%) | |
| Panel 18 | 19.2% | 18.6% | 19.8% | 21.8% | 29.7%** |
| (18.4%, 20.1%) | (15.7%, 21.5%) | (17.2%, 22.4%) | (18.4%, 25.1%) | (25.5%, 33.9%) | |
| Panel 19 | 19.8% | 21.7%** | 24.1% | 23.2% | 21.7% |
| (19.0%, 20.6%) | (18.8%, 24.7%) | (20.9%, 27.3%) | (19.9%, 26.6%) | (18.0%, 25.4%) | |
| Panel 20 | 21.1% | 23.1% | 19.1% | 17.4%* | 15.1%** |
| (20.2%, 22.0%) | (19.4%, 26.7%) | (16.3%, 21.9%) | (14.0%, 20.8%) | (12.5%, 17.6%) | |
| Panel 21 | 21.1% | 21.1% | 19.8% | 17.0%* | 14.7%** |
| (20.4%, 21.9%) | (18.3%, 23.9%) | (16.5%, 23.2%) | (13.3%, 20.7%) | (11.1%, 18.3%) | |
| Mean age | 41.6 | 38.5*** | 36.7*** | 37.3*** | 38.4*** |
| (41.3, 41.9) | (37.7, 39.4) | (35.9, 37.6) | (36.1, 38.4) | (37.2, 39.6) | |
| Percent female | 51.9% | 52.6% | 52.3% | 53.2% | 51.9% |
| (51.3%, 52.5%) | (49.9%, 55.3%) | (49.2%, 55.4%) | (49.0%, 57.5%) | (48.3%, 55.5%) | |
| Household income relative to poverty | |||||
| < 100% | 6.9% | 15.6%*** | 18.1%*** | 19.2%*** | 18.6%*** |
| (6.3%, 7.5%) | (13.3%, 17.9%) | (15.2%, 21.1%) | (15.9%, 22.4%) | (15.2%, 22.0%) | |
| 100–125% | 2.1% | 5.6%*** | 5.9%*** | 4.8%** | 5.0%** |
| (1.9%, 2.3%) | (4.2%, 6.9%) | (4.1%, 7.7%) | (3.0%, 6.5%) | (3.4%, 6.6%) | |
| 125–200% | 7.7% | 15.9%*** | 16.7%*** | 15.9%*** | 19.0%*** |
| (7.3%, 8.2%) | (13.6%, 18.1%) | (14.4%, 18.9%) | (12.9%, 19.0%) | (15.7%, 22.4%) | |
| 200–400% | 28.5% | 31.5% | 29.3% | 40.1%*** | 38.1%*** |
| (27.5%, 29.5%) | (27.9%, 35.1%) | (26.3%, 32.3%) | (35.2%, 45.0%) | (33.1%, 43.2%) | |
| > 400% | 54.7% | 31.5%*** | 30.0%*** | 20.0%*** | 19.2%*** |
| (53.5%, 56.0%) | (27.4%, 35.6%) | (26.0%, 34.0%) | (15.8%, 24.3%) | (15.5%, 23.0%) | |
| Self-rated health | |||||
| Excellent | 33.4% | 30.3%* | 29.0%** | 29.5% | 30.1% |
| (32.5%, 34.4%) | (27.2%, 33.4%) | (26.0%, 32.0%) | (25.5%, 33.6%) | (26.0%, 34.3%) | |
| Very good | 35.6% | 30.6%** | 30.9%** | 31.3%* | 32.7% |
| (34.7%, 36.5%) | (27.8%, 33.4%) | (27.7%, 34.0%) | (27.2%, 35.4%) | (28.6%, 36.8%) | |
| Good | 22.5% | 26.0%* | 25.5%* | 28.5%** | 23.9% |
| (21.8%, 23.2%) | (23.3%, 28.7%) | (22.8%, 28.2%) | (24.7%, 32.2%) | (21.1%, 26.8%) | |
| Fair | 6.8% | 9.7%*** | 11.9%*** | 8.0% | 11.1%*** |
| (6.4%, 7.2%) | (8.1%, 11.3%) | (9.6%, 14.2%) | (5.8%, 10.2%) | (8.7%, 13.5%) | |
| Poor | 1.6% | 3.4%*** | 2.7%* | 2.7% | 2.2% |
| (1.5%, 1.8%) | (2.4%, 4.4%) | (1.9%, 3.6%) | (1.4%, 4.0%) | (1.2%, 3.1%) | |
| Percent “agree” or “strongly agree” with: | |||||
| I don’t need health insurance | 12.5% | 19.0%*** | 18.1%*** | 16.8%** | 19.8%*** |
| (12.0%, 13.1%) | (15.6%, 22.5%) | (15.5%, 20.7%) | (13.9%, 19.8%) | (16.4%, 23.3%) | |
| Health insurance is not worth the cost | 22.6% | 29.4%*** | 26.9%* | 26.0% | 29.5%** |
| (21.7%, 23.4%) | (26.5%, 32.3%) | (23.6%, 30.2%) | (22.3%, 29.7%) | (25.3%, 33.6%) | |
| I can overcome illness without medical help | 16.3% | 19.8%** | 22.3%*** | 20.4%* | 20.6%* |
| (15.6%, 17.0%) | (17.4%, 22.1%) | (19.5%, 25.0%) | (16.5%, 24.4%) | (16.5%, 24.8%) | |
| I am more likely to take risks than others | 22.0% | 27.8%** | 25.2%* | 23.7% | 25.0% |
| (21.3%, 22.7%) | (24.0%, 31.7%) | (22.0%, 28.3%) | (19.4%, 27.9%) | (21.3%, 28.7%) | |
MEPS panels 17–21 covering 2012–2017
2Percent of those who “strongly agree” or “agree” with the statements
Statistical significance for differences relative to continuously insured (t-tests) are indicated by *p < 0.05, **p < 0.01, ***p < 0.001
Prior insurance disruptions were still strongly associated with a reduced likelihood of having a USCP even after controlling for demographic, socioeconomic, health, and attitudinal variables in a multivariable logistic regression analysis (Table 2). The magnitude of association was greatest for adults with the longest disruptions in coverage. For example, compared to people who were continuously insured, the odds of having a USCP was 38% lower among people who were uninsured for 1 to 3 months (odds ratio (OR): 0.62; 95% CI: 0.51–0.75) and 56% lower for those uninsured for ≥ 12 months (OR: 0.44; 95% CI: 0.35–0.55).
Table 2.
Association Between Duration of Prior Insurance Coverage Disruption and Having a Usual Source of Care Provider (USCP) Among Currently Insured Adults Aged 18–64 Years
| Prior health insurance coverage disruption | Odds ratios (95% confidence intervals) |
|
|---|---|---|
| Unadjusted | Adjusted2 | |
| Continuously insured | Ref | Ref |
| 1–3 months uninsured | 0.563(0.482, 0.658) | 0.619 (0.507, 0.756) |
| 4–7 months uninsured | 0.485(0.421, 0.559) | 0.580 (0.492, 0.685) |
| 8–11 months uninsured | 0.392(0.318, 0.483) | 0.472 (0.371, 0.601) |
| ≥ 12 months uninsured | 0.389(0.321, 0.472) | 0.431 (0.346, 0.536) |
Data source: MEPS panels 17–21 covering 2012–2017
2Odds ratios from multivariable logistic regression model are adjusted for sex, age, household income relative to the federal poverty line, self-rated health, and attitudes about healthcare and risk
DISCUSSION
Prior research has established that being uninsured is strongly associated with poor access to healthcare, sub-optimal utilization, and, ultimately, poor health outcomes. Extending this literature, our study suggests that lapses in insurance coverage, even short ones, may continue to negatively affect access to care even after coverage is regained. One possible explanation for this is that disruptions in coverage may force some patients to change provider networks. For these patients, this may mean finding new providers, which takes time and effort. Regardless of the causal mechanism, if disruptions in health insurance coverage have lingering effects on access to care as our analysis suggests, policy makers working to improve the U.S. healthcare system may need to go beyond strategies that expand coverage opportunities for the uninsured and develop strategies that also help insured individuals at risk of losing coverage maintain it.
Supplementary Information
Below is the link to the electronic supplementary material.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Arnold LD, McGilvray MM, Cooper JK, James AS. Inadequate cancer screening: Lack of provider continuity is a greater obstacle than medical mistrust. J Health Care Poor Underserved. 2017;28(1):362. [DOI] [PMC free article] [PubMed]
- 2.Sabety AH, Jena AB, Barnett ML. Changes in health care use and outcomes after turnover in primary care. JAMA internal medicine. 2021;181(2):186-194. [DOI] [PMC free article] [PubMed]
- 3.Cowburn S, Carlson M, Lapidus J, Heintzman J, Bailey S, DeVoe J. Insurance continuity and human papillomavirus vaccine uptake in Oregon and California federally qualified health centers. Am J Public Health. 2014;104(9):e71-e79. [DOI] [PMC free article] [PubMed]
- 4.Kogan MD, Alexander GR, Teitelbaum MA, Jack BW, Kotelchuck M, Pappas G. The Effect of Gaps in Health Insurance on Continuity of a Regular Source of Care Among Preschool-aged Children in the United States. JAMA. 1995;274(18):1429-1435. [PubMed]
- 5.Rogers MA, Lee JM, Tipirneni R, Banerjee T, Kim C. Interruptions In Private Health Insurance And Outcomes In Adults With Type 1 Diabetes: A Longitudinal Study. Health Aff (Millwood). 2018;37(7):1024-1032. [DOI] [PMC free article] [PubMed]
- 6.Yabroff KR, Zhao J, Halpern MT, Fedewa SA, Han X, Nogueira LM, Zheng Z, Jemal A. Health Insurance Disruptions and Care Access and Affordability. U.S. Am J Prev Med. 2021 61(1):3-12. [DOI] [PubMed]
- 7.Stata Corporation. Stata Survey Data Reference Manual, Release 16. 2020.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.