Abstract
Background: Poor and uninsured people have increased risk of medical and psychiatric illness, but they might be more reluctant to seek care than those with higher incomes. Little information exists about the biopsychosocial problems and concerns of this population in primary care.
Methods: We surveyed 500 consecutive patients (aged 18 to 64 years) in a primary care clinic serving only uninsured, low-income patients. We used self-report questions about why patients were coming to the clinic, a chronic illness questionnaire, the Primary Care Evaluation of Mental Disorders Patient Health Questionnaire, and items from the Childhood Trauma Questionnaire. Providers completed a questionnaire naming problems elicited from patients.
Results: Patients reported their most common chronic medical problems to be headaches, chronic back problems, and arthritis. The most common concerns patients wanted to discuss with providers and that providers elicited from patients were problems with mood. Compared with patients without current major mental illness, patients with a current major mental illness reported significantly (P < .001) more concerns, chronic illnesses, stressors, forms of maltreatment and physical symptoms.
Conclusion: The illness content of this uninsured, low-income population is dominated by emotional distress and physical pain. These needs place a serious burden on providers and can complicate management of chronic medical illnesses. Recommendations for specialized interview training and integrating mental health services are discussed.
People without health insurance often have different health-care–seeking behaviors than people who are insured. The uninsured are less likely to pursue care for acute illnesses, chronic illnesses, preventive care, or for serious or morbid symptoms1–3 Their hesitance to seek health care becomes apparent after short periods without health insurance and is more pronounced after extended uninsured periods.4,5 In midlife, those who are uninsured experience a greater decline in overall health status than those who are insured.6 The most vulnerable sectors of the uninsured population are those persons living in poverty. In 2000, 41% of the population living below federal poverty guidelines, aged 18 and 64 years, lacked health insurance.7
The relation between poverty and poor health is well established.8,9 Compared with middle- and upper-income populations, indigent populations have more medical illness and mental illness, diminished psychological, social, and physical functioning,10–12 and greater mortality rates.13,14 Those living in poverty, even when covered by health insurance, are more reluctant to seek health care than persons with higher incomes.15–17
These features of low-income and uninsured populations suggest that the characteristics of indigent, uninsured primary care populations might be different from general primary care populations. There is growing evidence to support this hypothesis.18–22 We could find only limited documentation, however, of the full content of medical or psychiatric problems, common co-occurring illnesses, or the reasons why patients in low-income uninsured, primary care populations seek care.
The present study builds on data from two earlier studies of an uninsured, low-income, primary care adult population. To better understand provider time use, we studied the frequency of diagnoses and of diagnostic combinations in all patient visits during a 1-year period between 1998 and 1999.23 The most common visit diagnoses recorded by providers were depression (23%), hypertension (12%), sinusitis (7.4%), tobacco abuse (6.7%) and anxiety (6.0%). In one half of the visits, providers recorded two or more problems. Fifty percent of these multiple problem visits included depression, anxiety, or alcohol abuse. In a second study we surveyed 500 consecutive patients (only one visit per patient was included) to assess the prevalence of mental disorders and compared findings with results from a parallel study of 3,000 patients from a representative primary care sample.22 Twice as many low-income, uninsured primary care patients had mental disorders compared with the general primary care sample (34% vs. 15%). Within the low-income sample, those with mental disorders had significantly more medical problems, lower functional status and greater disability than patients without mental disorders. Because our visit data included multiple visits by the same patients and our prevalence study only assessed mental disorders, we did not have a measure of the relative prevalence of all health care problems or of problem combinations.
In the current study, we sought to describe the biopsychosocial content of this population. Because middle- and high-income, insured people have different health-care–seeking behaviors than low-income and uninsured people, we wanted to learn what concerns prompted them to seek care. Because mental disorders are so prominent in this population and because they often complicate overall health care efforts, we studied which problems were most associated with mental disorders. We believed this information would be helpful in designing health care efforts to respond to the unique needs of poor, uninsured working-age patients seeking primary health care.
Methods
Setting
This study was conducted in April and May of 1999 at the Marillac Clinic serving Grand Junction, Colo, and the surrounding Mesa County. At the time of this study, Marillac Clinic served only those with incomes of less than 150% of federal poverty guidelines and who had no form of health insurance (no Medicaid, no Medicare). In 1999 Mesa County had a population of 113,000; 14.5% lived below the poverty level, 16% lacked health insurance, 90% were white, and 8% were Hispanic.24 The Human Subjects Review committee of St. Mary’s Hospital, Grand Junction, approved our study. St. Mary’s Hospital, the regional medical center for Western Colorado, is adjacent to and closely affiliated with the Marillac Clinic.
Selection and Description of Patients
All consecutive patients (patients with repeated visits were excluded) aged 18 years and older who spoke English or Spanish and with clinic medical appointments were invited to participate. Participants were given a $5 coupon to a grocery store. Five hundred eighty-nine patients were invited, 68 refused, and 21 were missed, for an enrollment of 500 patients (85%). This was 19% of the patients seen in 1999. The mean age (SD) of those who refused was 40 (11.2) years compared with 38 (12.1) years for participants, (t(588) = 1.88), and 59% of those who refused were women compared with 68% of participants (χ2 = 2.42; df = 1). More information about the enrollment procedures is available elsewhere.22
Data Collection
Patients
Before seeing their provider Marillac patients completed a questionnaire. During consent, patients could approve or not allow providers to be given information from the survey. Providers were not allowed to see patient responses until after the visit, however. Patients were asked to “please list all the health concerns you hoped to discuss with your medical provider today.”
The questionnaire included the Primary Care Evaluation of Mental Disorders Patient Health Questionnaire (PRIME-MD PHQ).25 The PHQ includes a question asking, “During the last 4 weeks, how much have you been bothered by any of the following problems?” to assess the presence of 13 physical symptoms. Response options were (1) not bothered, (2) bothered a little, (3) bothered a lot. One question on the PHQ assessed physical and sexual abuse, and we divided it into two questions, one assessing physical abuse since age 18 years and the other assessing sexual abuse since age 18 years. The PHQ includes a section assessing whether patients were not bothered, bothered a little, or bothered a lot by up to 10 current stressors in the last 4 weeks. The PHQ assesses the presence of four threshold disorders: major depression, anxiety, panic disorder, or bulimia without physician involvement in assessment. The term threshold refers to meeting criteria for DSM IV disorders. More details about the use of these questions and population percentages with each diagnosis are described in our earlier study22 and in the PHQ validation study.25
To assess risk for childhood maltreatment, we used seven items taken from the Childhood Trauma Questionnaire, a valid and reliable measure of early abuse and neglect.26 Although we were not able to include the entire instrument, these seven questions had the highest item-total correlations with the five subscales (emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect) based on a secondary analysis of a large primary care population.27 Respondents are presented with a series of statements about childhood experiences that are endorsed on a 5-point Likert-type scale, with response options ranging from never true to very often true. One question, used elsewhere,28 assessed the presence of 16 current chronic illnesses. Other questions assessed demographic variables.
When patients consented, all diagnostic information was given to their primary care provider after the visit for use in the next visit. All patients endorsing any level of suicidal ideation were called by their primary care provider or the first author (LBM).
Providers
Before the visit, providers were given a separate sheet of paper with the patient’s chart and asked to “please list all problems and concerns you have elicited.” Providers filled out these sheets during and after the visit and placed them in a collection box in the provider work area. Numbers were placed on the pages that matched numbers on the patient questionnaire so they could be combined for analysis.
Problem Coding
The clinic database was queried to find the 30 most frequent diagnoses. This list was compared against the top diagnoses in the National Ambulatory Medical Care Survey (NAMCAS)29 with missing items added. The resulting list was integrated into the 17 global categories from the International Classification of Diseases, 9th Edition (ICD-9)30 to create global categories for those not covered by the specific Marillac and NAMCAS items, including categories for signs, symptoms, and ill-defined conditions. We added three items: medication check or refill, socioeconomic concerns (eg, “no money for medicine” or “can’t pay rent”), and other tests. The final list of 58 items was used to categorize patient responses about concerns they wanted to discuss on the day of their visit and provider elicitation of patients’ concerns. The first author or a nurse categorized responses on all the surveys. The clinic medical director clarified coding questions. The list of the 58 categories is available from the first author.
Analysis
The entire patient sample was analyzed as a single unit to describe the relative prevalence of all problems and concerns. The sample was also analyzed to compare two subgroups—those meeting criteria for threshold diagnoses and those not meeting criteria for threshold diagnoses. This subgroup analysis was performed because in our earlier study22 we found that the threshold group had considerably lower functional status, greater disability, and higher counts of medical problems than patients who did not have a threshold disorder. One hundred sixty-eight patients (34% of the 500 patients sampled) met threshold criteria for one or more diagnoses.22
To test differences in demographics between patients with and without threshold diagnoses, chi-square analyses with corrections for continuity and t tests were used for categorical and continuous variables, respectively. Odds ratios (ORs) and 95% confidence intervals (CIs) were constructed to examine the odds of endorsing a particular health problem, concern, stressor, or maltreatment items in the group of patients with a threshold diagnosis in comparison with those without a threshold diagnosis. t Tests were used to determine whether the number of chronic health problems, concerns, elicited concerns, stressors, and endorsed maltreatment items differed between the threshold groups. Lastly, a logistic regression analysis was performed to determine which item counts best discriminated between patients with and without a threshold diagnosis. Backward and forwards stepwise techniques were used to arrive at the set of independently significant predictors (P < .05). The odds ratios for the analyses represent the increase in odds associated with a 1-point increase in the count.
Results
Table 1 displays the population demographics. We compared the demographic variables (age, education, martial status, ethnicity, household size, household income, and personal income) in the patients who met criteria for threshold diagnoses and patients in the rest of the population. We found no significant differences except in personal income. Forty-three percent of patients reporting incomes less than $5,000 a year had one or more threshold disorders compared with 29% of patients with incomes greater than $5,000 per year (χ2 = 9.09; df = 1; P < .003).
Demographics of Study Population (n = 500).
Table 2 shows patient reports of current chronic illnesses. Pain-related illnesses are the most frequent problems. Seven of the 10 most common illnesses are reported by a higher percentage of patients with one or more threshold diagnoses. The population percentages for other health problems not in the table are physical disability, such as loss of an arm or leg or eyesight or hearing or a birth defect, 6.5%; chronic liver disease, 6.1%; angina, heart failure, or coronary artery disease, 4.0%; neurological condition, such as epilepsy, convulsions, fainting spells, or Parkinson’s disease, 2.7%; stroke or major paralysis, 2.3%; Cancer diagnosed within the last 3 years but not skin cancer, 2.1%.
Percentage of Patients Responding to the Request to Indicate Whether They Now Have Any of the Following Health Problems.
Table 3 shows the frequencies of the concerns patients hoped to discuss on the day of the visit and the concerns and problems providers elicited from their patients. Not reported in the table are assorted signs, symptoms, and ill-defined conditions that patients hoped to discuss and that providers elicited. Thirty-eight percent of patients and 33% of providers reported one or more symptoms, making this category the largest of responses. There was no significant difference between the percentage of patients with threshold diagnoses (42%) or patients without threshold diagnoses (36%) indicating hopes to discuss one or more symptoms (OR = 1.3, 95% CI, 0.9–1.9). Providers elicited one or more symptoms from a higher percentage of patients in the threshold group than in the non-threshold group, however, 42% vs 29%, (OR = 1.7, 95% CI, 1.2–2.5; P < .01).
What Patients Want to Discuss and What Providers Elicit from Their Patients.
Table 4 shows the prevalence of common stressors bothering patients a lot. Financial concerns are most common, and all stressors are reported significantly more often in patients with major mental disorders. Table 5 shows patient reports about physical and sexual violence in adulthood and emotional, social, physical, and sexual maltreatment during years when patients were growing up. All forms of maltreatment during adulthood and childhood are more common in patients with mental disorders.
Patients Indicating Whether They Have Been Bothered by Any of the Following Problems (Stressors) in the Last 4 Weeks.
Patient Reports of Maltreatment During Adulthood or Childhood.
Table 6 examines the relation between numbers of items endorsed and the presence of a mental disorder. Patients with mental disorders endorsed significantly more items for all variables in Tables 1 through 4 and on the symptom checklist in the PRIME-MD PHQ25(not shown). Logistic regression analysis of the six total scores showed that item count increases in any one of four categories (patient concerns, stressors, maltreatment items, and physical symptoms) were significant independent predictors of having a threshold diagnosis.
Item Counts and Relation with Mental Illness.
Discussion
Patients in this low-income, uninsured primary care population report their most common chronic health problems to be headaches, chronic back problems and arthritis. A problem with mood was the most common, single health problem patients wanted to discuss with their providers (29%) and that providers elicited (9.8%) from their patients. Consistent with studies of symptom complaints in primary care,31,32 patients with current major mental illness reported significantly higher counts of current chronic illnesses, physical symptoms, maltreatment experiences, and current stressors and they had significantly more concerns to discuss and significantly more concerns elicited. Increasing item counts in any one of four variables (patient concerns, stressors, maltreatment items, and physical symptoms) was an independent predictor of mental illness. An examination of these patients’ current stressors and maltreatment history provides insight into this population’s psychosocial morbidity highlighting financial strain, social and family stress, violence, and sexual abuse in childhood and adulthood.
The most prevalent patient health concerns in this population are mental disorders and various forms of chronic pain, and these two illnesses are likely to occur in patients at the same time. One recent large nationally representative study in the United States used the same question as we did to assess prevalence of chronic medical illnesses. They found pain complaints to be the largest illness category across all socioeconomic strata, with population percentages inversely proportional to family income.33 The high prevalence of pain-related problems (headaches, back pain, gynecological pain) in patients with mental disorders (and vice versa) is consistent with evidence showing that patients with persistent pain are more likely to have concurrent depression or anxiety.34,35 Patients with depression and anxiety have also been found to have a higher prevalence of medically unexplained symptoms, particularly pain.32
The well-established trend for most patients to use general medical providers as their de facto36,37 mental health providers might be more pronounced for a low-income population with mental disorders. Recent evidence suggests they might be more likely to lose insurance coverage or have benefits decreased.38,39 Addressing these problems might be an essential component for improving the health and well-being of this population. Mental illness is associated with more disability40,41 than most biomedical problems and can impede efforts to manage common medical illnesses, such as hypertension and diabetes.42–44
The needs of this population present a formidable challenge to health care providers who might have limited time, training, and interest to address psychosocial problems.45,46 Current pressures to see more patients, fill out more forms, yet provide high-quality, evidence-based care make addressing these needs difficult. Workable solutions should accommodate the provider’s as well as the patient’s needs if changes are to be lasting and effective. The recent Institute of Medicine recommendations, Crossing the Quality Chasm: A New Health Care System for the 21st Century,47 provides some recommendations that can help both patients and providers. Two of its 10 core recommendations are for providers to share decision making with patients and for clinicians to cooperate with one another. Below we discuss these recommendations in light of this study’s findings, keeping in mind that any changes should enhance rather than complicate the lives of indigent-care providers.
Using communication skills48 that emphasize shared decision making throughout the interview improves health outcomes and conserves resources,49,50 helps manage time,51–53 and can be especially important in serving a population that feels little control in day-to-day life. An element long recognized as essential to effective communication is discovering the reason(s) for the patient visit.54 Patients with long problem lists and mental disorders, however, are often seen as difficult55 or frustrating56 by physicians who do not want to lose control of time and who have limited training to treat mental health problems. Training providers to be proactive in eliciting a full list of patient concerns up front, in combination with time-management strategies (eg, prioritization, negotiation), can help providers feel comfortable in collaborative agenda setting, knowing visit lengths need not be longer.57
Making an effort at the beginning of the interview to elicit all patient concerns has several advantages. Our results are consistent with other studies of patients who have multiple medically unexplained symptoms,58,59 suggesting that many patients will acknowledge emotional pain if given the opportunity. Counting patient complaints during a full elicitation of patient concerns may serve as a simple screen for mental illness.60 The US Preventive Services Task Force now recommends screening for depression in primary care.61 This practice can be especially valuable in underserved settings. A full understanding of patient concerns up front might decrease the probability of “Oh, by the way” comments in the closing moments of the interview.51 When patients prioritize the management of current chronic illness lower than psychosocial issues, physicians gain valuable insight into patient readiness to address problems that might be of greater importance to providers than to patients. Patients who feel that all their concerns and expectations62,63 are addressed and that their preferences64,65 influence decision making are more satisfied and have better health outcomes.50,62 Addressing indigent patient expectations might help providers establish a foundation for a relationship from which subsequent efforts can be devoted to preventive and chronic health care.66
Earlier we reported that 90% of the Marillac patients preferred their mental health providers and medical providers communicate with one another about their care.22 Sharing the care of complicated patients can ease the burden on providers and improve the care for patients. Models that integrate mental health practitioners into primary care67–69 have been shown to be effective in reducing symptoms,28,70 and cost-effective71–73 in reducing patient disability and risk of unemployment.72,74,75 The recent Surgeon General’s report on mental health advocated for integration of mental health and primary care services as essential in improving health care outcomes in the United States.76 Similar collaborative and multifaceted approaches can be useful for patients with persistent pain.77 Systems serving a low-income, uninsured population can also benefit from hiring a case manager with strong community connections to assist with self-sufficiency concerns.78,79 Poverty plus financial strain (40% of this sample, 66% of patients with psychiatric illness) contribute to the incidence and maintenance of unemployment and psychiatric illness.12,80 Transforming health care systems to sustain efforts to address chronic illnesses, however, requires adjustment in multiple domains.23,81–83
Limitations
The Marillac Clinic population is mostly white and located in a semi-rural setting and might not be representative of often more diverse urban populations. Although our results suggest that the content of a low-income, uninsured primary care population is different from what has been reported in general primary care,29,84 more research is needed. Earlier studies of the content of primary care use provider billing data that might not include all secondary diagnoses, such as psychosocial problems and chronic illnesses.85 Patients might have underreported or overreported concerns on survey questions or might write down more concerns than they express verbally. Providers might not have allowed patients to complete their agendas52,86 or they might have screened patient responses in some other way. We asked providers to list problems they elicited, which might not be the same as the problems they addressed or recorded in medical records. Some of the Marillac providers had received training in patient-centered interviewing. Studies in settings where providers are not trained in patient-centered skills might show a greater discrepancy between patient concerns and concerns the provider elicited.
Further studies using the same methods in both general primary care and underprivileged settings are needed to make accurate comparisons of the prevalence of common illnesses and patient concerns. Studies using direct observation will shed more light on the dynamics of the provider-patient relationship in indigent care settings.
Conclusion
Patients who lack health insurance and live in poverty often seek health care because they suffer emotional and physical pain. Without care, these patients are more likely to experience declining health, prolonged disability, and unemployment. Health care systems serving a low-income uninsured population might be more effective when integrating multifaceted, interdisciplinary approaches. Physicians who support patient autonomy87,88 through shared agenda setting and treatment planning can provide to their indigent patients a health ingredient often absent in their worlds—a sense of control in life.89,90
Acknowledgments
The authors wish to thank the Marillac Clinic medical assistants, administrative staff, community volunteers, and clinicians who contributed to this study in many essential ways. Because no additional financial support was used to fund this study, the role of the entire clinic community was critical for successful completion of the study. We thank the Marillac patients for taking the time to describe their health care needs. We also thank Mark Doescher, MD, Barry Saver, MD., and Sharon Dobie, MD for their help with the conceptual phase of this work. Jurgen Unutzer, MD, helped us find the chronic illness questionnaire and Robert Spitzer, MD, helped us with the PRIME-MD PHQ.
The Marillac Clinic and the Mesa County Consortium on Health expresses their appreciation for support from the Robert Wood Johnson 2000 Local Initiative Funding Partners Program to match funding from local contributors lead by the Colorado Trust. This funding provides 4 years of funding for training and clinical service to better meet the needs of this population.
Notes
Between August, 1998 and July, 1999 Mr. Mauksch was on leave from the University of Washington Department of Family Medicine as a Consultant in Collaborative Care to the Marillac Clinic. Funding for his position came from the Laura Jane Musser Fund, the Victim/Witness Assistance and Law Enforcement Fund–21st Judicial District of Colorado, the Sisters of Charity, and St. Mary’s Hospital. Portions of this research were presented at the 2002 meeting of the Collaborative Family HealthCare Association, Clearwater, Florida.
- Received for publication September 12, 2002.
- Revision received September 12, 2002.