INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

After a generation of decline, tuberculosis (TB) case rates, like those of many other infectious diseases, rose through the 1980s and early 1990s (1, 2). Public-health responses have focused on interrupting the chain of transmission by treating active cases that present for care, tracing their contacts, and chemoprophylaxis (3, 4). Yet much of the spread of TB may take place before the first contact with the medical system. Further, if contact-tracing efforts miss symptomatic TB patients, these patients may not suspect their diagnosis and may therefore delay seeking care long enough to spread the disease. Delays of more than 2 mo have been shown to spread the disease to domestic contacts (5). Moreover, delays in diagnosis may also worsen the course of the disease and even increase the case fatality rate (6). Understanding what proportion of symptomatic TB patients delay seeking care and why they do so may prove crucial to controlling epidemic TB.

Little is known about the prevalence of delay in seeking care among TB patients in the United States, or about what the roots of the delay might be. International studies indicate that a greater degree of concern about TB symptoms, age, and male gender predict less delay until presentation to the physician (9, 10). Studies of Latinos in Orange County and Texas found average delays in seeking care of more than 8 mo. In open-ended interviews, subjects attributed their typical symptoms (e.g., chronic cough, weight loss) to mild conditions not requiring the care of a physician (11). More general studies of access to care and delay in care-seeking among patients without TB confirm that self-perceived need for care and physician ratings of the severity of symptomatology are strong predictors of prompt care-seeking (12). Aday and Andersen concluded that more equitable access to care occurs when need for care or severity of illness predicts utilization better than potential access barriers (e.g., ethnicity, income, insurance status, appointment waiting time) (19).

Los Angeles County has the second largest number of active TB cases of any city in the United States (20). In cooperation with the public-health authorities responsible for controlling TB in Los Angeles County, we surveyed a consecutive sample of patients with active TB to answer the following research questions: (1) What proportion of active-TB patients were identified through contact-tracing efforts and what proportion sought care as a result of symptoms? (2) How much clinically significant delay occurs among symptomatic TB patients? (3) Which access barriers best explain delay? In particular, does less perceived need for care or less severe disease best explain delay, or are social factors like unemployment or lack of a regular source of care more important?

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sampling and Data Collection

The sampling and data-collection methods used in the study, which were reviewed and approved by the relevant institutional review boards, have been reported previously (21). We prospectively identified 526 consecutive cases of confirmed active TB, reported between April and September 1993, as they were reported to the Los Angeles County TB control authority by care providers, governmental agencies, and laboratories, as required by law. We excluded children, prisoners, deceased patients, and those who could not speak one of five common languages in our study population: English, Spanish, Mandarin, Tagolog, or Vietnamese.

Each potential respondent received a letter describing the study and providing a phone number to call if they did not wish to participate. Trained bilingual interviewers contacted all eligible patients, using the phone numbers provided by the county TB registry, clinics, directory assistance, or friends or relatives listed in medical records. Once they had contacted a potential respondent, the interviewers identified themselves as calling from the University of California at Los Angeles, and again explained the purpose of the study in the respondents' native language. Subjects were assured that all results would be confidential and not shared with their providers or the TB control authority. We also sought, at soup kitchens, shelters, and a specialized county clinic for the homeless, the 12% of our target population who were homeless (11% of eventual respondents). Interviewers made an average of 16 attempts to contact respondents. The average time from confirmed TB diagnosis to interview was 2 mo.

After obtaining informed consent, interviewers used a structured, symptom-based interview to elicit ethnicity, country of birth, reason for seeking care, legal status, English proficiency, number of potentially exposed contacts, TB-related symptoms, and length of delay in seeking care from each respondent. In this population of active-TB patients, we defined TB symptoms as one or more of the following occurring in the previous 2 yr and lasting 2 wk or more: cough, hemoptysis, fatigue, fever, weight loss, swollen glands, diarrhea, anorexia, night sweats, pleuritic pain, or lymphadenopathy. Delay times were measured by a series of questions ascertaining how long before obtaining care the respondent had noted symptoms of TB. The number of potentially exposed contacts was determined by the number of persons living in the respondent's usual sleeping place or who worked in the same room for more than 1 h each day. Perceived access barriers and need were assessed through a battery of questions focused on the reasons for delaying seeking care and the need for obtaining care. Potential reasons for care-seeking delay were elucidated both in an open-ended manner and by specific probes for language barriers, waiting time for an appointment or in the care provider's office, cost, transportation difficulties, absence of a regular provider, child-care or work conflicts, fear of treatment or diagnosis, fear of immigration authorities, and low perceived need for care. For example, subjects were asked, "What was your main reason for not seeking care sooner?", and their unprompted responses were recorded and categorized. As an example of specific probes, self-reported need for care was assessed by asking respondents whether any of the following reasons explained any delay in seeking care: "Did you think the symptom was not serious enough?" "Did you think that you could treat the condition yourself?" and "Did you think that the doctor couldn't do anything about it?" All identifying information linking interview or registry data to individual subjects was destroyed at the conclusion of data collection, and was replaced with an anonymous unique identifier.

A registry maintained by the county TB control authority provided additional demographic data. The registry also provided clinical information, such as radiography and microbiology results, that might serve as markers for severity or need for care. Previous studies of the same registry population showed reasonably good agreement between measures of human immunodeficiency virus (HIV) status, acid-fast bacilli (AFB) smear status, and culture results abstracted from the registry, medical records, and laboratories (22).

Data Analysis

Previous research has shown that exposed domestic contacts of active-TB patients frequently show signs of infection after 2 mo of exposure. Thus, we defined clinically significant delay as a period of more than 60 d from the onset of any TB symptom (as defined earlier) to the first medical visit for the symptom. We classified symptomatic respondents into those with and without clinically significant delay.

Predictors of delay were divided into measures of perceived access barriers and measures of need for care. Measures of access barriers could be patient characteristics (e.g., employment status, insurance status) or reasons for delay (e.g., child-care conflict, transportation problems). Measures of need were either self-reported (as described earlier) or derived from the clinical information in the county TB registry (e.g., AFB smear status, presence of a cavitary lesion on radiography). Serious symptoms were defined as cough, coughing up of blood, fever, and night sweats. The registry also provided information about whether the respondent was homeless at the time of diagnosis. These data were based on public-health nurses' fairly strict definition of homelessness: patients had to be living in a shelter or in public places at times of diagnosis. Chi-square, t or Wilcoxon's two-sample tests were used to examine the bivariate associations between predictors and delay. Analyzing delay with a continuous outcome measure and nonparametric comparisons did not change the direction or statistical significance of the bivariate results. Age, gender, and bivariate predictors showing a statistical association with delay at the p < 0.10 level were subjected to separate, all-possible-subsets regression and stepwise backward regression for selection of variables (23). These variables were then reestimated with the full data set, using a logistic-regression model.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Of the 526 eligible cases, 161 could not be contacted and 54 refused to participate, for an overall response rate of 60% (n = 313). Nonrespondents did not differ significantly from respondents by age, gender, ethnicity, employment, homelessness, anatomic site of disease, or culture results. The 313 participating patients formed the study population for our first research questions evaluating the relative prevalence of discovery through case tracking and seeking care for symptoms of TB. Most patients (71%) reported that they had sought care for symptoms (Table 1). Another 15% were diagnosed in the course of seeking care for a medical condition other than TB. Only 1% reported that public-health authorities had identified them as contacts of infected patients, although 10% had participated in TB screening programs.

We could not define delay in seeking care for symptoms in 65 patients without significant TB symptoms, and the analysis of delay is therefore based on the 248 symptomatic cases. The characteristics of these 248 patients are summarized in the first column of Table 2. More than three quarters were under 65 yr of age, and men made up two thirds (64%) of the study population. Immigrants made up 72% of the study population of TB cases in Los Angeles County. Many of the immigrants (35% of immigrants, 20% overall, not shown) had an undocumented legal status, although 20% declined to describe the legality of their immigration status. More than one in four of the overall sample described their ability to communicate in English as poor. The vast majority (89%) of the sample were nonwhite (11% white, 12% African-American, 48% Latino, 26% Asian, 3% other, not shown). The socioeconomic variables describe a disadvantaged population. The median annual household income was between $5,000 and $10,000, and 46% of the sample reported a household income of under $5,000/yr. Similarly, 49% were unemployed and 48% lacked any public or private health insurance. Slightly more than one in 10 (11%) were homeless.

In this population of active-TB patients, 88% had an abnormal and suggestive chest radiograph and 75% had positive cultures, and virtually all (97%, not shown) had either an abnormal chest radiograph or culture. Abnormalities indicating a high degree of infectiousness were common. Half of the study population had positive smears for AFB, and a third had cavitary lesions on chest radiography. Only about 14% were coinfected with HIV. Over half (57%) were told they might have TB when they first sought care.

Of the 248 patients with TB symptoms, 50 (20%) met our definition of a medical-care delay likely to result in infection of contacts, consisting of a period of 60 d from symptom onset to first medical visit. Almost a third (n = 73, 30%) delayed more than 30 d. The distribution of the delay was highly skewed; the average delay was 74 d (SD = 216), during which subjects exposed an average of eight people either at home or at work. Many (37%) thought in retrospect that they had delayed too long in seeking care (not shown).

The right-hand columns of Table 2 summarize the relationship between delaying care for more than 60 d and patient characteristics, perceived access barriers, and need factors in the 248 symptomatic patients. There was no association between delay and most demographic variables, such as age, gender, and ethnicity. Unemployed patients delayed seeking care 25% of the time, as compared with 14% of the time for those with jobs.

Several perceived access barriers were statistically associated with delay in seeking care. For example, patients who lacked a regular doctor delayed 23% of the time, as compared with only 13% for those who had a regular doctor. Similarly, patients who were unsure of where to go for care delayed 33% of the time, as compared with 16% for those who knew where to go. Patients who cited cost and waiting time in the office and for an appointment were also more likely to delay for more than 2 mo in seeking care. Fear of immigration authorities, though low in prevalence, was strongly associated with delay. Almost half (47%) of those who feared immigration authorities delayed in seeking care, as compared 18% of patients without such fear. Transportation problems did not affect delay.

One of the three measures of perceived need for care was also statistically associated with delay in seeking care. Patients who felt that they could treat their condition themselves delayed 31% of the time, as compared with only 14% of patients who didn't feel this. However, patients who described their health as excellent or good, or who felt that their symptoms were not serious, were statistically no less likely to seek care in less than 60 d than were sicker patients. Patients with serious symptoms and positive cultures, AFB smears, and X-ray results also did not seek care sooner than those without such findings.

Both the best of the all-possible-subjects regression model and the stepwise backwards logistic-regression model revealed that three of these variables statistically significantly predicted delay when controlling for age and gender (Table 2). The final logistic-regression model (not shown) had the following statistically significant variables: uncertainty about where to go for care (OR = 2.4; 95% CI: 1.2 to 4.9), unemployment (OR = 2.3; 95% CI: 1.2 to 4.7), and belief in the efficacy of self-treatment (OR = 2.1; 95% CI: 1.1 to 4.4). Linear regression of the different variables on continuous delay (rather than on a 60-d threshold) did not change the direction or significance of these multivariate results.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We found that the vast majority of active-TB patients in Los Angeles County had their disease diagnosed after they had sought care for symptoms, rather than as a result of screening or contact-tracking efforts. Of the symptomatic patients, one in five had delayed seeking care for a period long enough to result in the spread of infection. During that medical-care delay, patients exposed an average of eight other people to TB.

Our data suggest that several barriers to care impede symptomatic TB patients from obtaining treatment. Patients who delayed thought that they could treat their symptoms themselves, or were disconnected from the medical-care system either by unemployment or by lack of a regular source of care. Fear of immigration authorities, cost, and office or appointment waiting time predicted delay at the bivariate level, though these measures may be operating through other variables and require further investigation.

More worrisome was the lack of a strong relationship between need for care (at least as evidenced by serious symptoms and positive AFB smear, culture, and radiography results) and the promptness with which TB patients sought care. This is particularly concerning because these measures of severity also predict infectivity (24). In most previous studies of delay in seeking care for TB, evaluated need and severity of illness strongly predicted care utilization. If access to care had been distributed evenly throughout our sample, we would have expected that TB patients with more severe illness would have been more likely to promptly seek medical care. Instead, perceived access barriers appear to explain more of the delay than does need for care of health status, suggesting that subgroups of the TB population are facing inequitable barriers to care.

The present study was among the first to examine delay in seeking care for symptomatic TB, and has several potential limitations. It was a cross-sectional study, thus limiting our ability to determine cause-and-effect relationships. Our outcome measure of delay in seeking care was necessarily self- reported, and it may therefore suffer from recall bias or misclassification of non-TB symptoms as those of TB. Also, although we went to unusual effort by attempting to contact patients an average of 16 times and updating phone numbers from multiple clinical sites and telephone directories, we were unable to contact 40% of our target population. However, respondents and nonrespondents did not differ on a range of characteristics measured by the county TB registry. Nonrespondents most probably had higher rates of address changing and no telephone access, attributes that may be associated with lower incomes and poorer access to care. If nonrespondents did have greater perceived barriers in access to care than did respondents, our study may have underestimated the prevalence and effect of those perceived barriers. Finally, we chose to analyze the outcome variable dichotomously, on the ground that exposures to TB occurring less than 2 months after its development are less likely to result in spread of the disease, thus limiting our ability to analyze variation for shorter and longer intervals. However, analysis of continuous delay did not change the direction or significance of the multivariate results.

Because our sample was drawn from the Los Angeles County TB registry, to which all laboratories, physicians, and hospitals must report active-TB patients by law, it is likely to be quite representative of TB patients in the County. Unpublished internal TB-control studies, using pharmacy data from the largest hospital in Los Angeles, indicate a 97% reporting rate for the study time period. However, the TB-patient population of Los Angeles does differ from the national TB population in many respects, particularly in the number of immigrant cases, limiting the generalizability of our study results (25).

Conclusions

Control of TB has relied primarily on the early and aggressive identification and treatment cases of active TB (3, 26). Our results have implications for public-health policies aimed at this goal. First, policies to control the spread of TB should encourage increased access to care. This is because perceived barriers to access, such as cost and waiting time, are also associated with delay in obtaining care. In particular, our results support policies that reduce copayments, increase availability of medical care, and decrease waiting time in the office and for an appointment at likely sites of first TB care. Policies that increase the fear of immigration authorities among patients seeking care for TB, like the recent California ballot proposition, may increase delay in seeking care and exacerbate epidemic TB, though further studies would shed light on the mechanism of this bivariate association (27). Contact tracking and screening programs, although important, can only have a limited effect in identifying active-TB patients, since most patients seek care as a result of symptoms.

Our data also support public-education efforts to encourage persons with TB who are unaware of the grave significance of their symptoms, or who have no regular source of care, to seek carea difficult challenge in this population. Targeted campaigns aimed at high-risk groups in HIV clinics, or in churches or schools serving the immigrant community, may be a first step. However, public-education efforts cannot represent a complete solution to the TB epidemic. In an era in which budgetary constraints increasingly force public-health departments to reduce personal health services to populations at high risk for TB, these results sound a cautionary note.