Original Contribution

The relationship between chronic illness, chronic pain, and socioeconomic factors in the ED^?

Owen Hanley MPH ^a,b, James Miner MD ^a,?, Erik Rockswold MA ^a, Michelle Biros MD, MS ^a

^aDepartment of Emergency Medicine, Hennepin County Medical Center, Minneapolis, MN 55415, USA ^bDepartment of Environmental and Occupational Health, University of Minnesota School of Public Health, Minneapolis, MN 55364, USA

Received 29 April 2009; revised 30 September 2009; accepted 2 October 2009

Abstract

Objectives: The study aimed to determine the prevalence of chronic illness and chronic pain in emergency department (ED) patients across demographic backgrounds.

Methods: This was a cross-sectional study at an urban, level I trauma center with 98 000 annual visits. This was a prospective sample of adult patients presenting to the ED during a randomized distribution of daily 8-hour periods between June 4 and August 26, 2007. Prevalence of chronic illness was compared to subject demographics using logistic regression, and prevalence of chronic pain was compared using ordinal logistic regression.

Results: Six thousand nine hundred sixty-one patients presented during the data collection periods; 3882 were eligible, and 3132 (82%) were enrolled (51.7% male; age, 41.1 +- 15.8 years; range, 18-98 years). Chronic illness was reported in 36.3% of patients and chronic pain in 34.9% of patients. Chronic illness was associated with homelessness (odds ratio [OR], 1.75; 95% confidence interval [CI], 1.17- 2.61), family income less than $25 000 (OR, 2.27; 95% CI, 1.60-3.22), and lack of access to primary care facilities (OR, 2.68; 95% CI, 2.25-3.21). Chronic pain was associated with homelessness (OR, 2.56; 95% CI, 1.79-3.64), family income less than $25 000 (OR, 2.54; 95% CI, 1.91-3.39), and lack of access to primary care facilities (OR, 1.47; 95% CI, 1.26-1.70).

Conclusions: Patient housing situation, family income, and perceived access to primary care medical facilities were associated with higher self-reported rates of chronic illness and chronic pain.

(C) 2011

Introduction

The “demographic transition” is a phenomenon of proportional age restructuring that is occurring worldwide. The term describes the increasing percentage of middle-aged

^? This work was presented at the 2008 SAEM Annual Meeting in Washington, DC, in May 2008.

* Corresponding author.

E-mail address: [email protected] (J. Miner).

(25-49 years) and older (50+) persons and a decline in the percentage of younger age groups. Developed nations were the first to experience this transition as public health improvements coupled with increasing medical knowledge drastically reduced the infant mortality rate and infectious disease mortality [1]. Simply put, people are living longer and more productive lives than in any prior period in history [2]. Chronic health problems have replaced many infectious disease agents as the largest contributors to mortality [3]. In the United States, much of our health care efforts and dollars are channeled toward these chronic conditions.

0735-6757/$ - see front matter (C) 2011 doi:10.1016/j.ajem.2009.10.002

Studies have shown that a small number of repeat patients represent a substantially larger proportion of total patient visits in emergency department (ED) settings [4]. Other studies have focused on socioeconomic status and ethnicity as an important indicator of health, demonstrating the groups that use medical services disproportionately to their population [5,6]. To address chronic health problems, it is important to characterize the populations presenting to the ED.

Only a small number of studies have been done that quantify the relationship between socioeconomic status and health [7-9]. Wealth is one of the best predictors of lifespan and years of productive life [10,11]. Although this has been documented in Developing countries, the issue is no less salient in the United States. Epidemiologic methods have been historically used to identify proximate causes of disease and pain without attempting to identify the root causes of these medical problems [12].

In this study, we chose to use self-reported chronic disease as an indicator of ongoing Health care needs. Furthermore, we used chronic pain as an indicator of the need for ongoing health care. A large number of ED patients have underlying chronic pain syndromes, with rates as high as 40% noted in one review [13]. A key step in the prevention of chronic pain may be the adequate treatment of acute pain, and lack of access to medical care may be detectable by measuring the occurrence of chronic pain. The objective of this study is to determine patient characteristics associated with higher prevalence of chronic illness and chronic pain [5,8,14,15] to quantify some of the social mechanisms associated with the distribution of chronic illness and chronic pain differentially throughout the population [16-21]. This information can be used to better understand the numbers and characteristics of patients presenting to urban EDs with chronic illness and pain and to raise questions about the relationship between socioeco- nomic conditions and health status as seen from the ED.

Methods

Data collection procedures for this study were approved by the institutional review board at Hennepin County Medical Center (HCMC). The subsequent analysis of de- identified data was approved by the institutional review board at the University of Minnesota. This study used a cross-sectional design ascertaining demographic character- istics, socioeconomic factors, and a prospective patient survey of self-reported chronic illness or chronic pain.

Study population

The HCMC is a public, urban level I trauma center in downtown Minneapolis, Minn. Data collection was con- ducted in the HCMC ED and consisted of a survey

administered to all eligible patients by trained research associates during one randomly assigned 8-hour shift each day (7:00 AM to 3:00 PM, 3:00 PM to 11:00 PM, or 11:00 PM

to 7:00 AM) between June 4 and August 26, 2007. All adults 18 years and older were approached to take the survey. The patient was excluded if he or she was in critical condition (could not respond to survey), a prisoner, intoxicated or otherwise under the influence of impairing drugs, refused, or had already completed the survey on a prior visit. Patients were also excluded if they could not speak English. Written consent was obtained before the interview.

Data collection

Data collection was conducted by trained research associates who were composed of 45 volunteer medical, public health, and undergraduate students who were part of the Research Associate Program at HCMC. These inter- viewers were trained in ascertaining study eligibility, consent processes, survey administration as well as answering and clarifying patient questions concerning survey questions. Data were initially entered into an Excel spreadsheet (Microsoft Corp, Redmond, Wash) and then transferred to STATA 10.0 (STATA Corp, College Station, Tex) for analysis.

Statistical analysis

The study population was first characterized with simple descriptive statistics. Percentage distributions were calculat- ed for each survey question to clearly illustrate the proportion of patients in each question category. To quantify and describe the relationship between socioeconomic status and the prevalence of chronic conditions, 4 different indicators of socioeconomic status were applied: family income, access to primary care, health insurance, and housing status. For purposes of analysis, reported family income was divided into 3 categories: less than $24 999, $25 000 to $74 999, and

$75 000 and higher. These were based on rough low-, middle-, and high-income brackets established by the Minnesota Census. Access to primary care was determined by patient response (yes or no). Housing status was categorized by patients as homeless, halfway house, living with friends, renting, property owner, and nursing home resident. Health insurance was determined by patient response (yes or no, and type). The survey question pertaining to chronic illness was answered by either yes or no. Prevalence odds ratios were estimated using logistic regression to describe the relationship between socioeco- nomic indicators and self-reported chronic illness. The independent categories of exposures in this study are family income, access to primary care, health insurance, self- reported health, and housing status. Odds ratios were calculated for each of these exposures separately. The reference level for family income was the highest income

bracket (N$75 000). The reference level for housing status was “property owner,” the reference level for access to primary care was “yes” (care available), the reference level for health insurance was “yes,” and the reference level of self-reported health was “poor.”

Table 1 Description of excluded patients by reason, frequency, and proportion

Exclusion criteria*	Frequency (n)	Proportion (%)
Under 18	932	30.3
Non-English speaking	445	14.5
Intoxicated/under the	184	6.0
influence (drugs)
Critical condition	633	20.6
Prisoner/in custody	66	2.1
Sexually assaulted	29	0.9
Vulnerable adult	104	3.4
Already in study	158	5.1
Unable to contact a	528	17.1
Total	3079
*A small number of patients were excluded without recording a reason (b1%). a We could not interview some patients because they were being seen by providers or were transferred or discharged from the department before data collection could occur.

Chronic pain was categorized into 4 levels of patient self- reported experience ranging from “always/everyday” to “never/rarely.” Ordinal logistic regression was used to establish coefficients relating chronic pain to family income, access to primary care, and housing status. Euler’s number was then used to take the natural logarithmic base of the coefficients, which were then reported as odds ratios [odds ratio = e^ (coef.)]. Ordinal logistic regression is used when the outcome variable is in rank-ordered categories [22]. Reference categories were again selected to compare the

odds of having chronic pain at different levels of exposure.

The reference categories for this regression scheme were identical to those used in the logistic regression analysis. It is important to note that any exposure level could be chosen as

Table 2 Characterization of patient population by patient self-report

No response

180

5.7

No response indicates that a patient chose not to respond to question.

n = 3132
Mean age, 41.1 +- 15.8				Age range, 18-98
Characteristic	Total	%		Characteristic	Total	%
Sex			Family income
Male	1536	49.0	$0-$24 999		1399	44.7
Female	1430	45.7	$25 000-$74 999		786	25.1
No response	166	5.3	N $75 000		179	5.7
			Do not know		587	18.7
Ethnicity			No response		181	5.8
Asian	48	1.5
Black/African American	1194	38.1	Access to primary care
Native American	236	7.5	Yes		1724	55.0
White	1190	38.0	No		1233	39.4
Hispanic	179	5.7	No response		175	5.6
Other	112	3.6
No response	173	5.5	No. of people living with
			Alone		724	23.1
Educational level			1-2		1143	36.5
b8th grade	406	13.0	3-4		616	19.7
High school	1632	52.1	5-6		240	7.7
Associate degree	447	14.3	N7		233	7.1
Bachelor’s degree	346	11.0
Advanced degree	114	3.6	No response		186	5.9
No response	187	6.0
			Insurance type
Housing status			None		705	22.5
Homeless	152	4.9	Private		772	24.6
Halfway house	149	4.8	Medicare/Medicaid		1201	38.3
Living with friends	478	15.3	Other		276	8.8
Renting	1618	51.7	No response		178	5.7
Property owner	510	16.3
Nursing home	45	1.4

the reference level. The effects of 5 potentially confounding variables, age, sex race, education level, and number of people reported in the subjects household, were controlled for in the models.

Results

All patients (N = 6961) admitted to the HCMC ED during the data collection shifts were screened for eligibility. Table 1 shows the number and proportion of patients excluded from the study for each of the above-mentioned exclusion. Three thousand eight hundred eighty-two (55%) patients met all study Eligibility requirements, 750 of whom declined to participate in the study. Three thousand one hundred thirty-two consented to participate and completed the survey, yielding a response rate of 81%.

Patient sex, ethnicity, family income, access to primary care, housing status, number of people living in a residence, Type of insurance, and educational level are described in Table 2. Patient-reported overall health, chronic illness, and chronic pain are described in Table 3.

Logistic regression analyses were used to more clearly elucidate how socioeconomic status corresponds to chronic illness. Table 4 provides the results of the logistic regression used to obtain odds ratios for family income, housing status, access to primary care, self-reported health, and health insurance as they relate to chronic illness. The odds ratios given in Table 4 estimate the odds of having chronic illness at a nonreference level compared to the reference level. Among the potential exposure variables, the odds of having a chronic illness was higher in groups with lower family income, less stable housing status, less access to primary care, and poor self-reported health.

n = 3132 Total (n) %

Overall health Poor 371 11.8 Fair 848 27.1 Good 1349 43.1 Excellent 378 12.1 No response 186 5.9 Chronic illness Yes 1137 36.3 No 1828 58.4 No response 167 5.3 Chronic pain Never 996 31.8 Rarely 877 28.0 Most days 566 18.1 Always 526 16.8 No response 167 5.3

No response indicates that a patient chose not to respond to question.

Odds ratio		95% confidence interval
Family income
b $25 000	2.27	1.60-3.22
$25 000-$74 999	1.36	0.94-1.96
N$75 000	1.0 (reference)
Housing status
Homeless	1.75	1.17-2.61
Halfway house	1.48	0.97-2.25
Living with friends	1.17	0.88-1.56
Renting	1.11	0.89-1.39
Property owner	1.0 (reference)
Nursing home	1.83	0.87-3.85
Access to primary care
Yes	1.0 (reference)
No	2.68	2.25-3.21
Self-reported health
Poor	1.0 (reference)
Fair	0.42	0.32
Good	0.13	0.10
Excellent	0.05	0.03
Health insurance
Yes	1.0 (reference)
No	1.03	0.94-1.34
Analysis adjusted for age, race, education level, number of people in home, and sex.

Table 5 displays the results of the ordinal logistic regression analysis used to estimate odds ratios correlating chronic pain with family income level, housing status, access to primary care, self-reported health, and health insurance. Family income, housing status, and access to primary care all show an increased rate of chronic pain at their lowest levels (family income b$25 000, housing status = homeless, access to primary care = no). These odds ratios mean that as prevalence of reported chronic pain increases, family income level decreases. Self-reported health and health insurance did not correspond with chronic pain.

Table 3 Patient-reported health assessment

Table 4 Logistic regression odds ratios for prevalence of chronic illness as it relates to family income, housing status, and access to primary care

Discussion

This study finds an association between socioeconomic factors and prevalence of chronic health conditions. We have used 3 proxies for socioeconomic status (family income, housing status, and access to primary care) and related them to prevalence of self-reported chronic illness and chronic pain among patients presenting to the ED [23,24]. The odds ratios derived in Table 5 illustrates the association each of these variables has on the chronic health outcomes. Access to primary care is the most signifi- cant factor associated with the perceived prevalence of

Odds ratio		95% confidence interval
Family income
b $25 000	2.54	1.91-3.39
$25 000-$74 999	1.66	1.23-2.24
N$75,000	1.0 (reference)
Housing status
Homeless	2.56	1.79-3.64
Halfway house	1.31	.90-1.91
Living with friends	1.30	1.0-1.68
Renting	1.33	1.10-1.62
Property owner	1.0 (ref)
Nursing home	2.23	1.14-4.36
Access to primary care
Yes	1.0 (ref)
No	1.47	1.26-1.70
Self-reported health
Poor	1.0 (ref)
Fair	1.22	0.94
Good	1.02	0.81
Excellent	1.13	0.91
Health insurance
Yes	0.93
No	1.0 (ref)	0.80-1.01
Analysis adjusted for age, race, education level, number of people in home, and sex.

chronic illness, followed by family income and housing status. This suggests that those with no access to a primary care are more likely to see themselves as having chronic illnesses. In the case of chronic pain, the odds ratio for housing status was the largest, followed by family income and then access to primary care. This suggests that the environment in which a person resides and income within families are more strongly correlated with perceived chronic pain and that factors associated with the perception of chronic illness are different from those associated with chronic pain. It is possible that these patient characteristics are Surrogate markers for barriers to treatment rather than the occurrence of disease, but that cannot be determined from our study (eg, is no access to primary care associated with chronic illness, or does chronic illness decrease a patient’s access to primary care).

Table 5 Ordinal logistic regression coefficients for chronic pain as it relates to family income, housing status, and access to primary care

The goals of this study were threefold. The first was to better characterize the HCMC patient population. More information on the demographic makeup of ED patients will enable the provision of better treatment and guide future research efforts. The second goal of this study was to elucidate more clearly the relationship between socioeco- nomic status and chronic conditions. The information obtained in this study could be used to address the health needs of populations with low socioeconomic status. Improving and maintaining health care access may be

a means of decreasing chronic illness among patients. Addressing a patient’s housing status and state of poverty may be an important part of preventing or addressing a patient’s chronic pain.

Our final goal was to extend some of the issues raised by social epidemiologists to a common health care setting [25]. Instead of looking at individual risk factors for diseases, we aimed to look at the association of general socio- economic issues with chronic illness and chronic pain. Further research concerning the effects of unstable living conditions, access to primary care, and low income on chronic illness and chronic pain will be necessary to identify specific risk factors. Interventions at this higher level in the causal chain is challenging yet crucial to effect changes in the health of this population.

Limitations

The results of this study are only applicable to the ED population at HCMC over the summer of 2007 and may not be apply to a wider population. The large sample size and relatively diverse nature of the patient population do, however, imply the results may be similar in other ED populations. Emergency department patients may have a higher incidence of chronic health issues than the general population, and the socioeconomic distribution among patients in an urban county ED may be different from the other populations, further limiting the external validity of our findings.

Both exposure and outcome variables were based on a patient self-report mechanism. There was no method for verifying whether or not patients answered questions truthfully or accurately. It has been noted that patients are more likely to overreport their income and housing status than to underreport it, leading to perhaps even larger actual correlations between socioeconomic factors and chronic health issues than we found among our patients [26]. There was a substantial (20.6%) number of patients who took the survey but either did not know or chose not to reveal their family’s income. There are a number of scenarios that may apply to this group including a group trend toward nonreport because they were wealthy and did not want to disclose their income, were poor and were self-conscious about their income level, or were suspicious of interviewer motivation in a need-based payment setting. We also have no way of knowing the directionality bias of chronic condition self- reports. These 2 factors could bias odds ratios either toward or away from the null.

Another important consideration is the exclusion of patients who do not speak English (11.5% of study population). It is well known that many non-English speakers are recent immigrants to the United States. This group is often represented in Lower socioeconomic status categories [27]. By not including this population in the

survey and subsequent analysis, a large population present- ing to the HCMC ED are not represented. The inclusion of this demographic may change odds ratios.

Although the analysis attempted to control for obvious confounding factors related to chronic conditions such as age, sex, and race, many other less obvious confounders were left out of the model. For example, educational level is closely correlated with both health status and socioeconomic status [28,29]. In attempting causal models, a path between socioeconomic status and health can be made by educational level or knowledge of health systems information (health literacy). Furthermore, if educational level affects socioeco- nomic status directly, it can be directly linked to health. These are important considerations that were not accounted for in the regression models presented in this study.

Another potential confounder pertaining to housing status and family income is the number of people (dependents) living together. A family making $50 000 is considered middle income in this study; however, the number of people supported by that money plays an enormous role in the poverty status of the family. For this reason, census information uses income per number of household depen- dants as their poverty indicator. In piloting this study, we attempted to generate this information by asking patients to report their income in dollars. This failed to generate adequate response. Patients were much more comfortable checking income ranges. We tailored this aspect of the survey to obtain a greater response; therefore, family income brackets described in this article do not reflect poverty status precisely. A better method of reporting incomes in ED surveys would be beneficial to similar research endeavors.

Conclusion

Chronic illness and chronic pain were reported frequently among ED patients. Homelessness, family income less than

$25 000, and perceived lack of access to primary care medical facilities were associated with higher self-reported rates of chronic illness and chronic pain. The associations were different for chronic illness than for chronic pain. It is possible that chronic illness and chronic pain can be addressed at the level of a patient’s socioeconomic status by developing interventions for homeless, lack of financial resources, and lack of Primary care access.

Acknowledgments

Special recognition should be extended to the 45 or so volunteers of the Research Associate Program in the Hennepin County Medical Center Emergency Department who worked tirelessly and diligently to collect the data for this project.

References

1. International Monetary Fund. World economic outlook: the global demographic transition. World Economic and Financial Surveys; 2004. Accessed online at www.imf.org/external/pubs/ft/weo/2004/02, 11/9/2008.
2. UNP9400. 1994. World Population. Prospects 1994. Department of International Economic and Social Affairs, United Nations. (Tape) New York: UN.
3. Mathers C, Bernard C, Iburg K, Inoue M. Global Burden of Disease in 2002: data sources, methods and results. Global Programme on Evidence for Health Policy Discussion Paper Number 54. World Health Organization; 2003.
4. Billings J, Zeitel L, Lukomnik J, Carey TS, Blank AE, Newman L. The impact of socioeconomic status on hospital use in New York City. Health Aff 1993;12(1):162-73.
5. Brunswick A, Aidala A, Dobkin J, Howard J, Titus S, Banaszak-Hall J. HIV-1 of epidemiology 1993;104:107-23.
6. Perneger TV, Whelton PK, Klag MJ. Race and end-stage renal disease. Socioeconomic status and access to health care as mediating factors. Arch Intern Med 1995;155(11).
7. Adler NE, Boyce T, Chesney MA, Cohen S, Folkman S, Kahn RL, Syme SL. Socioeconomic status and health: the challenge of the gradient. Am Psychol 1994;49:15-24.
8. Biros MH, Hoffman PL. The prevalence and perceived health consequences of hunger in emergency department patient populations. Acad Emerg Med 2005;12(4):310-7.
9. Buck CW. Prenatal and perinatal causes of early death and defect. Preventive and community medicine. 2nd ed. Boston (Ma): Little Brown and Company; 1981. p. 149-66.
10. Feinstein JS. The relationship between socioeconomic status and health: a review of the literature. Milbank Q 1993;71(2):279-322.
11. Pappas G, Queen S, Hadden W, Fisher G. The increasing disparity in mortality between socioeconomic groups in the United States. N Engl J Med 1993;329:103-9.
12. Kaufman JS, Oakes JM. Methods in social epidemiology. San Francisco (Calif): Jossey-Bass; 2006.
13. Verhaak PF, et al. Prevalence of chronic benign pain disorder among adults: a review of the literature. Pain 1998;77(3):231-9.
14. Cunningham PJ. What accounts for differences in the use of hospital emergency departments across U.S. communities. Health Aff 2006;25: 324-36.
15. Rahimi AR, Spertus JA. Financial barriers to healthcare and out- comes after acute myocardial infarction. J Am Med Assoc 2007;297: 1063-72.
16. Cassel J. The contribution of Social environment to host resistance. American journal educational groups. Int J Epidemiol 1976;32: 355.
17. Dohrenwend BP, Levav I, Shrout P, Schwartz S, Naveh G, Link B, Skodal A, Stueve A. Socioeconomic status and psychiatric disorders: the causation selection issue. Science 1992;255:946-51.
18. Dutton DB. Social class, health and illness. Applications of social science to clinical medicine and health policy. New Brunswick (NJ): Rutgers University Press; 1986. p. 31-62.
19. Link B, Phelan J. Social conditions as fundamental causes of disease. J Health Soc Behav 1995:80-94 Extra Issue.
20. Rosen G. The evolution of social medicine. The handbook of medical sociology. 3rd ed. Englewood Cliffs (NJ): Prentice Hall; 1979.

p. 23-50.

1. Susser M, Watson W, Hopper K. Sociology in medicine. New York, NY: Oxford University Press; 1985.
2. Juul S. An introduction to Stata for health researchers. College Station (Tex): Stata Press Publication; 2006.
3. Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. Income distribution, socioeconomic status, and self rated health in the United States: multilevel analysis. BMJ 1998;317:917-21.
4. McDonough P, Duncan GJ, Williams D, House J. Income dynamics and adult mortality in the United States, 1972-1989. Am J Public Health 1997;87:1476-83.
5. Mechanic D, Aiken LH. Social science, medicine, and health policy. Applications of social science to clinical medicine and health policy. New Brunswick (NJ): Rutgers University Press; 1986.

p. 1-9.

1. Johnson ME, Fisher DG, Reynolds G. Reliability of Drug users‘ self- report of economic variables. Addict Res Theory 1999;7(3):227-38.
2. Schnepf SV. How different are immigrants? A cross-country and cross-survey analysis of educational achievement; 2004. IZA Discus- sion Paper No. 1398.
3. Blane D. Commentary: explanations of the difference in mortality risk between different seroprevalence and risk behaviors in an urban African American community cohort. Am J Public Health 2003;83: 1390-4.
4. Kaufman JS. Whad’ya know? Another view on cultural literacy. Epidemiology 2002;13:500-3.