Severity of Inpatient Hospitalizations Among Undocumented Immigrants and Medi-Cal Patients in a Los Angeles, California, Hospital: 2019

Abstract

Objectives. To compare the severity of inpatient hospitalizations between undocumented immigrants and Medi-Cal patients in a large safety-net hospital in Los Angeles, California.

Methods. We conducted a retrospective analysis of all 2019 inpatient stays at a Los Angeles hospital (n = 22 480), including patients of all races/ethnicities. We examined 3 measures by using insurance status to approximate immigration status: illness severity, length of hospital stay, and repeat hospitalizations. We calculated group differences between undocumented and Medi-Cal patients by using inverse probability weighted regression adjustment separately for patients aged 18 to 64 years and those aged 65 years and older.

Results. Younger undocumented patients had less severe illness and shorter lengths of stay than their Medi-Cal counterparts. Older undocumented immigrants also had less severe illness, but had similar lengths of stay and were more likely to have repeated hospitalizations.

Conclusions. While existing work suggests that undocumented immigrants could have more severe health care needs on account of their poorer access to medical care, we did not see clear health disadvantages among hospitalized undocumented immigrants, especially younger patients. There were fewer differences between undocumented and Medi-Cal patients who were older. (Am J Public Health. 2021;111(11):2019–2026. https://doi.org/10.2105/AJPH.2021.306485)

There is growing interest in how immigration status affects the health care access and utilization for 10.5 million undocumented immigrants¹ in the United States. Undocumented immigrants have significantly lower levels of insurance coverage and lower use of the emergency department, and are less likely to have a usual source of care compared with both their documented counterparts and the US-born.^2,3 These disparities are attributable to lower household incomes and ineligibility for Medicaid and Medicare, as well as lack of private insurance from employers. Even when health care is available, undocumented immigrants may be reluctant to access it because of the fear of deportation for themselves or their families.^4,5 Undocumented immigrants also face challenges accessing the limited care that is available to them because of economic barriers and unfamiliarity with the health care system.⁶

Despite worse access to and lower utilization of health care, undocumented immigrants do not seem to have widespread negative physical health. In a review of 45 studies, Hamilton et al.⁷ found that the large majority did not observe undocumented immigrants to have significantly worse health than their documented counterparts. In some studies, undocumented immigrants seemed to have better health outcomes for conditions such as hypertension, asthma, and other self-reported chronic conditions. A nationally representative sample confirmed the diverging trends between health insurance coverage and health status: Ro and Van Hook⁸ found that undocumented immigrants had lower odds of being currently insured but also had lower odds of disability and poor or fair self-rated health compared with those born in the United States.

While undocumented immigrants’ limited health care access and robust physical health patterns may appear contradictory, these findings rely on self-reported health status in surveys and may capture their health care utilization when individuals are relatively healthy. Undocumented immigrants tend to be younger than their documented counterparts,⁸ suggesting that some chronic health conditions may not have yet emerged or may have not reached the point of requiring serious medical intervention. Undocumented immigrants are also more likely to be positively selected on health, given the costs and risks inherent in unauthorized migration.⁹ Undocumented immigrants’ younger age and positive health selection may buoy their general health status, and, as a result, they may not seek regular medical care.

What is unknown, however, is the health status of undocumented immigrants when they reach the point of needing intensive medical care and their health needs when they are in close contact with the health care system. As immigrants stay longer in the United States, their positive health advantage erodes.¹⁰ In the case of undocumented immigrants, initial health advantages can obscure longer-term care needs as their health deteriorates over time. In the general population, those who are uninsured or do not have regular medical care are more likely to enter the health care system in poorer health and have worse outcomes for both overall health and specific diseases.¹¹ Given their overall lack of medical care access, it is possible that undocumented immigrants will display poorer health outcomes than other groups when they are at the point of needing higher-level care. Moreover, consistent underutilization of preventive services may leave them with more advanced stages of disease when they eventually need medical attention.

The extant literature has provided only limited or dated information on this topic. One study found uninsured undocumented immigrants hospitalized in Florida to have higher case mix indexes (a measure of disease severity) but shorter hospital stays than immigrants with legal status.¹² Among patients in select cites in California and Texas, undocumented immigrants had the same levels of hospitalizations as documented immigrants, with the exception of childbirth.¹³ Treatment of end-stage renal disease among undocumented immigrants has been widely studied as an example of a life-threatening condition that requires intensive medical treatment but is highly mediated by access to health care. Undocumented immigrants with end-stage renal disease are more likely to receive emergency-only hemodialysis than standard hemodialysis,¹⁴ which is associated with increased mortality, health care utilization, and costs.^15,16

In this study, we explored the severity of hospitalizations among undocumented immigrants of all races/ethnicities who have required inpatient stays at the largest safety net hospital in Los Angeles County, California. Los Angeles County has the largest population of undocumented immigrants in the country; there are nearly 880 000 undocumented immigrants, representing a wide number of countries of origin.¹⁷ We used insurance status to approximate immigration status, leveraging the county’s unique health plans for undocumented immigrants to accurately estimate the likely undocumented population. Addressing this knowledge gap will reveal potential unmet needs of undocumented immigrants in the hospital setting during critical periods of illness as well as possible disparities in inpatient health care.

METHODS

The study was a retrospective analysis of all inpatient stays at Los Angeles County and University of Southern California (LAC+USC) Medical Center from January 1, 2019, to December 31, 2019. We merged 2 data sources: (1) the LAC+USC internal electronic medical records system using Cerner PowerInsight and (2) Vizient Health System Data, a hospital billing and administrative claims database. All data were de-identified to conform to Health Insurance Portability and Accountability Act requirements.

We examined the data at both the encounter and patient level. For encounter data, there were 29 765 total inpatient hospitalizations at LAC+USC in 2019. We limited the sample to hospitalizations with patients aged 18 years and older and who were full-scope Medi-Cal patients (California’s Medicaid program) or undocumented (coding detailed in the “Variables” section). Our final analytic sample consisted of 22 480 inpatient encounters. In our analyses, we separated the sample by age (18‒64 years and ≥ 65 years) because of differences in health status at older ages and public insurance coverage (18‒64 years = 18 244; ≥ 65 years = 4236). For the older adults, we limited the sample to those who only had Medi-Cal or a combination of Medi-Cal and Medicare to better isolate a low-income comparison group. For outcomes that used patient-level data (e.g., unique number of patients who were admitted to LAC+USC in 2019), there were 15 876 patients (18‒64 years = 12 910; ≥ 65 years = 2966).

Variables

Outcomes

We examined 3 outcomes indicating the severity of the hospitalization: illness severity, length of hospital stay, and repeat hospitalization.

We measured illness severity by the relative risk of mortality, which we calculated with a proprietary algorithm that predicts the risk for mortality based on patient demographics, clinical characteristics, procedures, and comorbidities for each Medicare Severity‒Diagnosis Related Group. For example, the risk prediction for a liver transplant (Medicare Severity‒Diagnosis Related Group 5 or 6) was a function of hemodialysis, cachexia, complication of transplanted organ or tissue, ventilator on admission day, and type 2 diabetes. The risk for mortality was averaged over all encounters in the hospital for each diagnosis group in 2019. Encounters that were 75% of the mean were coded “lower risk” relative to the mean. Encounters that were within 75% and 125% of the mean were coded “similar risk,” and encounters more than 125% of the mean were considered “higher risk.” We dichotomized the outcome to hospitalizations lower or similar to the mean versus those with higher risk of mortality.

Length of hospital stay was the total number of days of the inpatient admission, with longer encounters indicating sicker patients. This was an encounter-level variable, meaning that each hospitalization had its own length of stay. To account for outliers, we conducted additional analyses with encounters longer than 21 days removed.

Repeat hospitalization was a patient-level variable and was dichotomized to patients who had 1 inpatient admission in 2019 versus those who had more than 1 admission. We assumed that those with more than 1 admission were sicker and needed more comprehensive care. Though it is possible that repeated hospitalizations might be attributable to unrelated or isolated events (e.g., injuries), it is more likely that repeated hospitalizations result from unresolved or ongoing illness.

Immigration status

We compared undocumented immigrants to full-scope Medi-Cal patients by using insurance status as a proxy to determine immigration status. We coded a patient as having undocumented status if the primary insurance for the encounter was restricted-scope Medi-Cal, which provides health services to low-income Los Angeles County residents who meet the income threshold for Medi-Cal but do not meet immigration status requirements as either US nationals, citizens, or lawful permanent residents. These services include access to county facilities that provide preventive, emergency, diagnostic, specialty, inpatient, and pharmacy services, as well as a local health care program (My Health LA) that offers primary and preventive health care services through community clinic partners.¹⁸

Given the income requirement for restricted-scope Medi-Cal eligibility, we chose full-scope Medi-Cal (hereafter referred to as Medi-Cal) patients as a comparison group of low-income patients who are either US-born or foreign-born with authorized status. Because of the citizenship requirements for federal health insurance, nearly all of the Medi-Cal patients are either US-born or documented. The other option for a comparison group is individuals on other forms of insurance (e.g., private insurance), but we do not know either the poverty level or the immigration status of these patients.

Covariates

We included age as a continuous variable, race/ethnicity (Hispanic [reference], non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, non-Hispanic other), language (English [reference], Spanish, other), gender (male [reference], female), and homeless status (housed [reference], nonhoused). We included language preference to account for reduced English proficiency, which can have an impact on treatment utilization and quality.¹⁹ As the primary safety net hospital in the county, LAC+USC has a substantial homeless patient population. We included homeless status to account for the substantial barriers to regular medical care that unhoused individuals face, as well as their complex health conditions.²⁰ For encounter-level outcomes, we included an indicator of whether the encounter was the only admission for the patient in 2019 (reference) or if it was a repeat hospitalization. For instance, if an encounter was 1 of 2 or more hospitalizations from a patient in 2019, all encounters from that patient were coded “1.” If an encounter was the only hospitalization from a patient in 2019, it was coded “0.”

Analysis

Because of the inherent compositional differences between undocumented and Medi-Cal patients, we estimated group differences in our outcomes of interest by using inverse probability weighted regression adjustment (IPWRA). This approach weights observations based on the inverse of their conditional probability of “treatment” exposure (i.e., being undocumented). The “treatment” model is estimated first, and the predicted probability of being “treated” (i.e., undocumented) is used to compute the inverse-probability weights. These weights are then applied to the outcome models (i.e., illness severity, length of stay, repeat admission), creating a pseudo-matched sample, such that undocumented and Medi-Cal patients are balanced in their covariates. In conventional multivariate regression analysis, covariates are included in the model to account for compositional differences but each observation is weighted equally. In IPWRA models, observations that have a higher likelihood of being “treated” (i.e., undocumented) are up-weighted while those that have lower likelihood are down-weighted. IPWRA is also known as the doubly robust method because it models both the outcome and propensity for “treatment” within the same framework, and only one needs to be correctly specified to produce unbiased results.²¹

We ran each weighted outcome model on the undocumented group first and then the comparison group. We calculated average treatment effects (ATEs), a common postestimation approach for treatment effects models such as the IPWRA,²² for each of our outcomes. The ATEs take a counterfactual approach and assume that each subject has a pair of potential outcomes: the outcome if they were undocumented or the outcome if they were on Medi-Cal. We calculated the predicted means for the outcome for each observation assuming they were undocumented (using their own covariate values) and then again for each observation assuming they were on Medi-Cal. For each observation, the difference between the undocumented predicted mean and Medi-Cal predicted mean is called the effect of “treatment.” We averaged this difference across the entire population to calculate the ATE.

For the “treatment” model (i.e., undocumented vs Medi-Cal), we included age, race, language, gender, and homeless status as predictors of undocumented status. For the outcome models (i.e., illness severity, length of stay, repeat admission), we included the same covariates as the “treatment” model as well as whether the encounter was a repeat hospitalization for the illness severity and length-of-stay outcomes. For illness severity and repeat admissions, we conducted logistic regressions, whereas for length of stay, we conducted a linear regression model. We calculated robust standard errors to correct for potential for heteroskedasticity. We did this separately for patients aged 18 to 64 years versus those aged 65 years or older. We conducted our analyses by using Stata version 16 (StataCorp LP, College Station, TX). We provide the coefficients for the IPWRA models in the Tables C, D, and E (available as supplements to the online version of this article at http://www.ajph.org).

RESULTS

Table 1 provides descriptive information for our sample at the encounter level. The same descriptive statistics at the patient level are provided in Table A (available as a supplement to the online version of this article at http://www.ajph.org). Nearly one third of encounters were for undocumented patients (31.6%). The mean age for younger patients (18‒64 years) was 45.8 years for undocumented patients and 41.6 years for Medi-Cal patients. Among older patients (≥ 65 years), the mean age was 73.1 years for undocumented and 73.8 years for Medi-Cal. Undocumented patients in both age groups were predominantly Hispanic (18‒64 years: 91.8% vs ≥ 65 years: 86.5%). Medi-Cal patients in both age groups were also predominantly Hispanic, but comprised a lower proportion (18‒64 years: 62.5% vs ≥ 65 years: 57.2%).

TABLE 1—

Descriptive Table of Inpatient Encounters in Los Angeles County and University of Southern California Medical Center: 2019

	Age 18–64 y			Age ≥ 65 y
Undocumented (n = 6 074), Mean or No. (%)	Medi-Cal (n = 12 170), Mean or No. (%)	Total (n = 18 244), Mean or No. (%)	Undocumented (n = 1 226), Mean or No. (%)	Medi-Cal (n = 3 010), Mean or No. (%)	Total (n = 4 236), Mean or No. (%)
Mean age, y	45.9	41.4	42.9	73.1	73.8	73.6
Race/ethnicity
Hispanic	5 576 (91.80)	7  609 (62.52)	13 185 (72.27)	1 061 (86.54)	1 723 (57.24)	2 784 (65.72)
Non-Hispanic White	6 (0.10)	508 (4.17)	514 (2.82)	11 (0.90)	88 (2.92)	99 (2.34)
Non-Hispanic Black	75 (1.23)	1 893 (15.55)	1 968 (10.79)	19 (1.55)	443 (14.72)	462 (10.91)
Non-Hispanic Asian	241 (3.97)	461 (3.79)	702 (3.85)	85 (6.93)	370 (12.29)	455 (10.74)
Non-Hispanic other	176 (2.90)	1 699 (13.96)	1 875 (10.28)	50 (4.08)	386 (12.82)	436 (10.29)
Gender
Female	3 008 (49.52)	4 964 (40.79)	7 972 (43.70)	704 (57.42)	1 386 (46.05)	2 090 (49.34)
Male	3 066 (50.48)	7 206 (59.21)	10 272 (56.30)	522 (42.58)	1 624 (53.95)	2 146 (50.66)
Language
English	859 (14.14)	8 229 (67.62)	9 088 (49.81)	93 (7.59)	1 106 (36.74)	1 199 (28.31)
Spanish	5 007 (82.43)	3 630 (29.83)	8 637 (47.34)	1 040 (84.83)	1 534 (50.96)	2 574 (60.76)
Other	208 (3.42)	311 (2.56)	519 (2.84)	93 (7.59)	370 (12.29)	463 (10.93)
Homeless	407 (6.70)	2 289 (18.81)	2 696 (14.78)	59 (4.81)	401 (13.32)	460 (10.86)
Illness severity (risk for mortality)
Below or similar	5 550 (91.37)	10 804 (88.78)	16 354 (89.64)	1 032 (84.18)	2 314 (76.88)	3 346 (78.99)
Above	524 (8.63)	1 366 (11.22)	1 890 (10.36)	194 (15.82)	696 (23.12)	890 (21.01)
Mean length of stay, days	4.5	6.1	5.6	5.2	6.8	6.3
2019 repeat encounter
First encounter	4 286 (70.56)	8 624 (70.86)	12 910 (70.76)	800 (65.25)	2 166 (72.00)	2 966 (70.00)
Second or higher	1 788 (29.44)	3 546 (29.13)	5 334 (29.24)	426 (34.75)	844 (28.00)	1 270 (30.00)

Undocumented patients primarily spoke Spanish (18‒64 years: 82.5%; ≥ 65 years: 84.8%), whereas Medi-Cal patients had differences in language preference according to age group. Notably, homelessness was nearly 3-fold greater among Medi-Cal patients compared with undocumented patients across age groups (18‒64 years: 18.8% vs 6.6%; ≥ 65 years: 13.3% vs 4.8%).

Compared with Medi-Cal patients, undocumented patients had lower illness severity and average length of stay. Among younger patients, 8.6% of undocumented patients had higher-than-average relative mortality risks compared with 11.2% of Medi-Cal patients. Among older patients, 15.8% of undocumented patients had higher-than-average mortality risk compared with 23.1% of Medi-Cal patients. The average length of stay in the hospital was 4.6 days for undocumented and 6.1 days for Medi-Cal among younger patients and 5.2 days and 6.8 days, respectively, among older patients. Among younger patients, inpatient encounters over the study period were comparable between undocumented and Medi-Cal patients, in which approximately 70.7% were first encounters and 29.1% were repeat hospitalizations. However, among older patients, undocumented patients had more repeat hospitalizations compared with Medi-Cal patients (34.8% vs 24.8%).

Group Differences in Inpatient Measures

The results of the IPWRA are presented in Table 2. Among patients aged 18 to 64 years, encounters with undocumented patients had lower illness severity and shorter lengths of stay than encounters with Medi-Cal patients. The probability of having an encounter with a higher-than-average risk for mortality was 8.4% for undocumented patients and 11.6% for Medi-Cal patients. Encounters for undocumented patients therefore had 3.2% lower probability of a higher-than-average mortality risk compared with encounters for Med-Cal patients. The average length of stay for undocumented patients’ encounters was 1.1 days shorter than for Medi-Cal patients (5 days vs 6.1 days). Among younger patients, the probability of having a repeat hospitalization in 2019 did not significantly differ between undocumented and Medi-Cal patients. Both groups had a probability of repeat hospitalization that was around 22%.

TABLE 2—

Group Differences Between Undocumented and Medi-Cal Patients in Inpatient Measures by Age Group, Los Angeles County and University of Southern California Inpatient Admissions: 2019

	Illness Severity (Above-Average Risk for Mortality)		Length of Stay, Days		Repeat Patient
Predicted Probability (95% CI)	P	Predicted (95% CI)	P	Predicted Probability (95% CI)	P
Aged 18–64 y		< .001		< .05		.97
Undocumented	8.4 (7.3, 9.5)		4.9 (4.2, 5.6)		22.4 (20.0, 24.9)
Medi-Cal	11.6 (11.0, 12.2)		6.1 (5.9, 6.3)		22.4 (21.4, 23.4)
Difference	3.2 (1.9, 4.4)		1.2 (0.4, 1.9)		−0.1 (−2.7, 2.6)
Aged ≥ 65 y		< .001		.33		< .05
Undocumented	15.7 (13.4, 18.1)		6.0 (5.0, 7.0)		26.5 (22.4, 30.6)
Medi-Cal	22.2 (20.7, 23.7)		6.6 (6.2, 7)		22.0 (20.2, 23.8)
Difference	6.5 (3.7, 9.3)		0.5 (−0.6, 1.6)		−4.5 (−9.0, −0.1)

Note. CI = confidence interval. Predicted probabilities calculated from inverse probability weighted regression adjustment models. Covariates for “treatment” model (i.e., undocumented vs Medi-Cal) were age, race, language, gender, and homeless status. Covariates for outcome models (i.e., illness severity, length of stay, repeat admission) were age, race, language, gender, homeless status, and length of stay.

Among patients aged 65 years and older, encounters for undocumented patients had similarly lower illness severity. Undocumented patients’ encounters therefore had 6.5% lower probability for a higher-than-average risk encounter (15.7% vs 22.2%). There was no significant difference in predicted length of stay. However, the probability of having a repeated hospitalization was higher for undocumented patients compared with Medi-Cal patients. Undocumented patients had a 26.5% probability of having a repeat hospitalization in 2019 while Medi-Cal patients had a 22.0% probability, with a statistically significant difference in probabilities of 4.5%.

Sensitivity Checks

We conducted several sensitivity checks to confirm the robustness of our findings. We also limited the sample to those who identified as Hispanic/Latino, which was 72% of the sample aged 18 to 64 years and older and 62% of sample aged 65 years and older. For encounter-level outcomes (relative risk of mortality and length of stay), we kept only 1 encounter per patient. The results were qualitatively similar for all of these checks (Table B, available as a supplement to the online version of this article at http://www.ajph.org).

For the length-of-stay outcome, we removed outliers whose inpatient stays were 21 days or longer. For encounters with patients aged 18 to 64 years, the difference in predicted length of stay dropped to 0.5 days but remained significantly different from zero. For patients aged 65 years and older, the difference dropped to 0.1 and remained nonsignificant.

DISCUSSION

We compared the severity of inpatient stays between undocumented and full-scope Medi-Cal patients at the largest safety-net hospital in Los Angeles County. Our results provide insight into the health status of undocumented immigrants at the point of needing inpatient medical care, by using measures of illness severity, length of hospital stay, and repeat hospitalizations. In our study, younger undocumented patients who were hospitalized had less severe illness and spent less time in the hospital compared with Medi-Cal patients. While existing work suggests that undocumented immigrants could have higher risk for poorer inpatient outcomes on account of their limited access to medical care,²³ the younger undocumented patients in this study did not present with more severe health problems upon hospital admission. Our findings add to those of others that have found positive health trends among undocumented immigrants compared with documented immigrants or US-born counterparts.^7,8

These trends were consistent among older undocumented immigrants as well, but older undocumented patients were more likely to have repeat hospitalizations than their Medi-Cal counterparts. Lower illness severity among older undocumented patients, however, suggests that their higher likelihood of repeat hospitalizations may represent a higher willingness to access care rather than comparatively poorer health. While qualitative work has suggested that barriers to receiving regular care can exacerbate older undocumented immigrants’ existing chronic conditions,²⁴ we did not find this to be the case from our data.

There are several caveats to our interpretations. First, the Medi-Cal patients at LAC+USC may be a unique, low-income sample. The high proportion of homelessness, for example, indicates that this population has complex health care needs. The longer length of stay among Medi-Cal patients may also be attributable to discharge planning issues, such as a long wait for other health facilities or a lack of a discharge destination. Thus, the relatively positive outcomes among undocumented patients may not be attributable to any inherent health advantages per se, but rather that the comparison group suffers from serious health conditions. Conversely, our sample of undocumented immigrants included those who had access to county and community facilities via local health plans that provide preventive care. Los Angeles County has actively addressed the health care needs of the undocumented population, which suggests that this undocumented patient population may be receiving regular care that contributes to their better health at the point of hospitalization.

Limitations

We do not have direct information on the immigration status of patients. However, we believe restricted-scope Medi-Cal is a valid approximation of undocumented status. Patients who are not insured at the time of admission but are Medi-Cal eligible are coded under a separate payment source that provides qualified individuals immediate access to temporary Medi-Cal while applying for permanent Medi-Cal or other health coverage (e.g., hospital presumptive eligibility). There may also be immigrants who use restricted-scope Medi-Cal who are not undocumented, such as those on student and work visas, and certain permanent legal residents who have not met the 5-year residency requirement for public insurance. Alternatively, there may be undocumented immigrants who do not qualify for restricted Medi-Cal because of their high incomes. We believe these comprise a very small number in our sample, however.

We also acknowledge that we lack important variables, such as nativity, country of origin, or ethnic subgroups (e.g., Mexican, Salvadoran, Chinese) and did not control for them in our analysis. We did include language use in our IPWRA models, however, which up-weights individuals in the full-scope Medi-Cal comparison who may be foreign-born and prefer Spanish. We also found similar results when we limited the analyses to Hispanics/Latinos, confirming that these trends are consistent for Hispanic/Latino undocumented immigrants. In addition, we did not have information on baseline health status. While this was indirectly incorporated in our illness severity measures, this is a limitation for length of stay and repeat encounters. Finally, our illness severity outcome might be biased by undiagnosed disease. Yet comprehensive evaluation upon admission often leads to diagnosis of various chronic conditions, which are then captured in the illness severity score. Nevertheless, we acknowledge the limitations of unknown chronicity of medical conditions and adequacy of treatment in the outpatient setting. Future research, including qualitative interviews, would enrich research with respect to nativity status, length of time in the United States, and perceptions around health and health care utilization among undocumented immigrants.

Conclusions and Future Directions

We found that undocumented immigrants in Los Angeles County do not have poorer health outcomes at the point of hospitalization. Contrary to some popular narratives, undocumented immigrants do not appear to overburden the health care system. We note, however, that the county provides public or subsidized access to health care services for its undocumented residents. While we do not have information on primary care utilization or outpatient care, future research could examine whether local policies contribute to positive hospitalization characteristics among undocumented immigrants or whether our findings hold in other locales with weaker safety-net programs.