Effect of HIV Infection and Tuberculosis on Hospitalizations and Cost of Care for Young Adults in the United States, 1985 to 1990

Methods

Data were obtained from the National Hospital Discharge Survey (NHDS), a nationally representative probability survey of discharges from nonfederal short-stay general and specialty hospitals in 50 states and the District of Columbia. From a total of approximately 6000 U.S. short-stay hospitals, a sample of about 500 hospitals was obtained for each year of the analysis [22]. The NHDS survey design has previously been described in detail [22]. Before 1988, a two-stage survey was used to select hospitals and sample discharges within hospitals. In 1988, the design was changed to a three-stage sample to select geographic locations within regions, hospitals, and discharges within hospitals [22]. Modification of the survey design in 1988 may affect the evaluation of trends [22]. Each hospital discharge record lists from 1 to 7 diagnoses, coded according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) [23].

Tuberculosis hospitalizations were defined as those for which discharge records listed a diagnosis of “tuberculosis” (ICD-9-CM codes 010-018). Records listing “primary tuberculosis complex” (code 010.0), “primary tuberculosis infection” (010.9), or “tuberculosis of the skin” (017.0), which account for approximately 11% of records listing tuberculosis, were excluded because they could have represented tuberculosis infection without disease. Hospitalizations for pulmonary tuberculosis were those for which discharge records listed a diagnosis of pulmonary tuberculosis without extrapulmonary tuberculosis. Extrapulmonary tuberculosis hospitalizations were those for which discharge records listed a diagnosis of extrapulmonary tuberculosis with or without pulmonary tuberculosis.

Human immunodeficiency virus hospitalizations were defined as those for which records listed HIV infection (ICD-9-CM codes 042-044, 279.19 or 795.8) [24]. The number of AIDS-related hospitalizations was estimated from the number of records listing ICD-9-CM codes designated for AIDS (042 or 279.19) and the number listing an HIV infection code concurrently with an illness that was a part of the 1987 CDC AIDS surveillance definition [25]. Hospitalizations related to “drug abuse or drug dependence” were defined as those for which records listed a diagnosis of “drug dependence” (ICD-9-CM code 304) or “nondependent drug abuse” (305.2-305.9) and did not include those in which records listed alcohol abuse or dependence.

Analyses included hospitalizations for which records listed HIV infection or active tuberculosis among adults 15 to 44 years of age (young adults) during 1985-1990, and were based on 3990 sampled records, corresponding to a weighted national estimate of 479 700 hospitalizations or patients. The unit of analysis was the “hospitalization” or the “patient,” not the “person,” because persons could have more than one hospitalization. Hospitalization rates were estimated by using the number of NHDS hospitalizations and U.S. census population data [26]. Geographic regions were those defined by the U.S. Census Bureau.

To estimate the direct costs of inpatient care, data on hospital daily charges (which included facility fees) were obtained from 1990 statewide hospital billing information for HIV infection and tuberculosis from 18 states: Arizona, California, Colorado, Florida, Illinois, Maine, Maryland, Massachusetts, Nevada, New Hampshire, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Vermont, Washington, and Wisconsin (Personal communication, 2 October, 1992. Spitzer M, Codman Research Inc.).

Inpatient HIV care costs were estimated by aggregating regional estimates. Regional rather than statewide estimates were derived because of the observed variation in cost by region and because estimates of the number of NHDS hospitalization days were available for regions, not states. We calculated the HIV care cost for each region by multiplying its number of NHDS HIV hospitalization days by its estimated daily cost of HIV care. This regional daily cost of HIV care was estimated from the statewide billing database by determining an average of the state-specific daily HIV care costs weighted by the number of HIV hospitalization days in the state. Because the daily cost of HIV inpatient care was similar for persons with and without AIDS, these groups were analyzed in aggregate. The same procedure was used for estimating tuberculosis inpatient care costs.

Although cost data were available for only 18 states, these states accounted for 71% of AIDS cases and 72% of tuberculosis cases among adults 15 to 44 years of age within the jurisdiction of NHDS during 1985-1990 [8, 27]. However, because cost data were not available for all states, we did a sensitivity analysis to assess the effect of geographic variations in costs on our national cost estimate.

Two methods were used to estimate inpatient care costs for the 6-year period 1985-1990. In method I, the 1990 hospital costs were applied to all years of data. Regional weighted average 1990 daily costs for HIV infection and tuberculosis were, respectively, $820 and $740 in the North, $1460 and $1190 in the Midwest, $1310 and $1160 in the South, and $1710 and $1210 in the West. In method II, costs were adjusted to account for inflation; costs for 1985-1989 were deflated according to the Consumer Price Index for Hospital and Related Services [28]. The estimates from both method I and method II were further adjusted by 10% to account for the potential undercoding of HIV infection [21, 29]. To estimate the increase in inpatient costs during the 6-year period, method II was used.

SESUDAAN, a software program designed to analyze complex multistage surveys and available through Statistical Analysis Systems (Cary, North Carolina), was used to compute the variances for the estimates of totals and proportions of hospitalizations with specified characteristics and to assess P values for the comparisons of means and proportions [30, 31]. The geometric mean duration of stay was used to determine the duration of hospitalization because of the non-normal distribution of the data. When the relative standard error of the estimate was 30% or greater, the estimate was excluded from the tables.

Results

Trends and Demographics

During 1985-1990, an estimated 418 200 HIV hospitalizations occurred among adults 15 to 44 years of age (young adults) in the United States. During the same period, an estimated 77 700 tuberculosis hospitalizations occurred among young adults, 16 200 (21%) of which were hospitalizations for patients with both tuberculosis and HIV infection. From 1985 to 1990, the rate of HIV hospitalization increased sixfold, from 18 to 102 per 100 000 young adults (Figure 1). From 1988 to 1990, the rate of tuberculosis hospitalization increased twofold, from 8 to 16 per 100 000 young adults (Figure 2). The rate of hospitalization for young adults without HIV infection or tuberculosis declined from 1.2 to 0.99 per 100 000 during 1985-1990.

Of young adult patients with HIV infection, 95% were 25 to 44 years of age, 58% were white, 84% were male, and 40% were from the Northeast (Table 1). Of patients with tuberculosis, 82% were 25 to 44 years of age, 29% were white, 70% were male, and 38% were from the Northeast (Table 1). Of patients with both HIV infection and tuberculosis, 95% were 25 to 44 years of age, 47% were black, 83% were male, and 54% were from the Northeast. Patients with HIV infection or tuberculosis were more likely to be 35 years of age or older, black, male, and from the Northeast than were those without HIV infection or tuberculosis (Table 1).

Comorbidity

During 1985-1990, the overall prevalence of tuberculosis among young adult HIV-infected patients was 4% (Table 2); tuberculosis prevalence increased from 2.4% in 1985-1988 to 5.1% in 1989-1990 (P = 0.003). Of patients with AIDS, 4% were hospitalized with tuberculosis (including 1.5% with extrapulmonary tuberculosis). Seventy-three percent of HIV-infected patients and 73% of HIV-infected patients with tuberculosis were hospitalized with AIDS (Table 2). Drug dependence was a concurrent diagnosis for 15% of patients with HIV infection and for 27% of those with both HIV infection and tuberculosis (Table 2).

Of young adult patients with tuberculosis, 23% had extrapulmonary tuberculosis, 21% had HIV infection, and 15% had AIDS (Table 2). Among patients with tuberculosis, HIV prevalence increased from 10.6% in 1985-1988 to 28% in 1989 to 39% in 1990 (P < 0.001). The prevalence of HIV infection was higher among patients with extrapulmonary tuberculosis (31%) than among those with pulmonary tuberculosis (18%) (P = 0.01). Of those with tuberculosis, 15% had a diagnosis of drug dependence or abuse, 10% had a diagnosis of alcohol dependence or abuse and 5% were diagnosed with diabetes. The number of sampled patients with tuberculosis and risk factors for tuberculosis, such as neoplasms, was too small to provide national estimates.

Duration of Hospital Stay

An increase in the duration of hospital stay was significantly associated with both HIV infection and tuberculosis. Among patients without tuberculosis, the duration of stay, as measured by the geometric mean duration of stay, rose in association with the increasing severity of HIV-AIDS status (trend test, P < 0.001) (Figure 3). Among patients with tuberculosis, the geometric mean duration of stay was 1.7-fold longer for HIV-infected patients without AIDS than for patients without HIV infection (P < 0.001).

The geometric mean duration of stay was approximately twofold greater for patients with tuberculosis than for those without tuberculosis in each subgroup defined by HIV-AIDS status: AIDS, HIV infection without AIDS, and no HIV infection (P < 0.001) (Figure 3). Among patients with AIDS, the geometric mean duration of stay was significantly longer for those with extrapulmonary tuberculosis and pulmonary tuberculosis than for those without tuberculosis (Table 3); similar results were observed for patients without HIV infection. Among HIV-infected patients without AIDS, the geometric mean duration of stay was 2.3-fold longer for those with pulmonary tuberculosis than for those without tuberculosis (P < 0.001) (Table 3).

Funding and Cost of Care

During 1985-1990, the proportion of HIV hospitalizations among young adults that was funded by the government increased from 33% to 54% (trend test; P = 0.03). During the same period, the proportion of tuberculosis hospitalizations funded by the government was 45% (range, 30% to 52%); there were no statistically significant trends.

The proportion of hospitalizations funded by the government was significantly higher for those listing HIV infection or tuberculosis than for those listing other diagnoses (Table 4). This proportion was twofold higher for hospitalizations listing either HIV infection or tuberculosis and threefold higher for hospitalizations listing both diagnoses than for those listing neither (Table 4). The proportion of hospitalizations in state or local government-owned facilities was twofold higher for tuberculosis hospitalizations than for others; there was no difference in this proportion between HIV hospitalizations and others (Table 4).

From 1985 to 1990, HIV hospitalization days for young adults increased 5.7-fold, and average daily hospital fees increased 1.5-fold, resulting in an estimated 7.7-fold increase in the estimated national cost of inpatient care. This estimated cost rose from $0.26 to 0.29 billion in 1985 to $0.67 to 0.74 billion in 1987 to $2.02 to 2.22 billion in 1990. During this 6-year period, HIV hospitalizations among young adults totaled 5 793 000 days, resulting in an estimated national cost of $5.74 to 7.38 billion (Table 5).

From 1985 to 1990, tuberculosis hospitalization days for young adults increased 2.1-fold and estimated national inpatient care costs increased 3.2-fold, from $0.09 billion to $0.31 billion. During this 6-year period, tuberculosis hospitalization days among young adults totaled 1 107 900, at an estimated national cost of $0.89 to $1.07 billion (Table 5).

The estimated national cost of inpatient care for HIV infection or tuberculosis or both totaled $6.41 to $8.10 billion during 1985-1991, of which 5% was for hospitalization of HIV-infected patients with tuberculosis. Sensitivity analyses showed that varying our assumptions about regional costs did not alter the national estimates of inpatient care costs by more than 7%. The HIV and tuberculosis inpatient care costs estimated for young adults accounted for 79% and 35%, respectively, of the estimated $7.3 to $9.4 billion HIV care costs and the estimated $2.5 to 3.1 billion tuberculosis care costs for persons of all ages in the United States during 1985-1990.

Discussion

Ours is the first study to use both cost data and a nationally representative sample of hospitals to estimate hospitalizations and their cost for HIV infection and tuberculosis. Our data show an increase in HIV and tuberculosis hospitalization rates, an increase in the prevalence of comorbidity of HIV infection and tuberculosis among inpatients, and escalating costs of HIV and tuberculosis inpatient care for persons 15 to 44 years of age in the United States.

The estimated direct cost of HIV inpatient care increased 7.7-fold from 1985 to 1990. A major factor contributing to this increase was the sixfold increase in HIV hospitalization rates during this period. Similarly, tuberculosis hospitalization rates increased twofold from 1988 to 1990. The change in the NHDS design in 1988 may explain the discontinuity in the trends of tuberculosis hospitalization rates between 1985-1987 and 1988-1990 [22].

These data provide a minimum estimate of HIV and tuberculosis hospitalizations. A diagnosis of HIV infection listed on the computerized records has been shown to be highly predictive for HIV infection: Ninety-five percent of hospitalized persons with a listed diagnosis of HIV infection were confirmed to have HIV infection [29]. However, not all persons with HIV- or tuberculosis-related disease may have had HIV infection or tuberculosis listed on discharge records. Although an adjustment was made in our analyses to account for the potential undercoding of HIV infection [21, 29], no such adjustment was made for tuberculosis. In addition, records that may have represented inactive tuberculosis, accounting for 11% of the records listing tuberculosis, were excluded.

The increase in the rate of HIV hospitalization observed during the study period (1985-1990), has continued during 1991 and 1992 (NHDS 1991-92. Personal communication, January 1994, Kozak, J, NCHS, CDC). Recent evidence, however, indicates that from 1985 to 1992, the rate of increase in HIV hospitalization decreased; furthermore, during 1991-1992, the annual number of hospitalization days declined (NHDS, 1991-92. Personal communication, January 1994. Kozak J, National Center for Health Statistics, CDC). Similarly, from 1991 to 1992, the frequency of hospitalization declined among a nonrandom sample of persons with AIDS [32].

Despite an apparent recent slowing in the rate of increase of HIV hospitalizations and a likely increase in outpatient care [32], inpatient care continues to account for a large portion of HIV health care costs. In our study, HIV inpatient care costs totaled $5.7 to $7.4 billion during 1985-1990. In 1992, inpatient care accounted for an estimated 65% of costs among persons with AIDS and 46% of costs among HIV-infected persons with T-cell counts lower than 200/µL [32].

The burden of HIV and tuberculosis inpatient care costs appears to be principally and increasingly borne by the government, which was the principal payer for 62% of HIV and tuberculosis hospitalizations. These data are consistent with data from previous studies that have shown an increase in the percentage of AIDS hospitalizations financed by Medicaid in California and New York, and of HIV hospitalizations financed by the government nationally during 1983-1988 [17, 33].

Although cost data for our study were not available from all states, the hospital billing database included states that accounted for 71% of AIDS cases and 72% of tuberculosis cases in adults 15 to 44 years of age within the jurisdictions included in the NHDS [8, 27]. Sensitivity analyses showed that varying our assumptions about regional and yearly costs had a minimal effect on the estimated national inpatient care costs. Thus, the national cost estimates appear to be insensitive to the relatively small regional and yearly differences in hospital charges per day compared with the large and increasing number of days of hospitalization.

Estimates of national annual inpatient costs have not been included in several recent studies of the cost of HIV care [19, 21, 32]; therefore, we compared our findings with available estimates from earlier studies [18, 20]. Our estimated 1990 HIV inpatient care costs of $2.02 to $2.22 billion were lower than previously estimated 1991 HIV costs of $8.5 billion, of which 80% to 90% was for inpatient care [20]. Probably the principal reason for this difference is that in the previous study, the projected number of hospitalization days was greater than the actual number. In fact, the number of hospitalization days used for the previous study's “medium” and “high” estimates was 3 and 6 times higher, respectively, than those determined by NHDS (NHDS 1991 tabulation. Unpublished data. NCHS, CDC). Our estimate is closer to the previous study's “low” estimate of $3.5 million [20]. The remaining difference between the previous estimate and ours may be related to the difference in study year (1991 vs. 1990), age of population (all ages vs. 15 to 44 years), and type of charges (all vs. inpatient only); each of these factors accounted for an estimated 1.1 to 1.2-fold difference between the estimates.

Our estimated cost for 1987 inpatient HIV care was $0.67 to 0.74 billion, higher than the $0.49 billion previously estimated for 1987 inpatient AIDS care [18]. This difference was expected because our analysis was based on HIV care and charge data, whereas the previous analysis was based on AIDS care and costs estimated by hospitals. Also, our study was based on a representative sample of U.S. hospitals, whereas the previous study was based on selected hospitals.

Most studies have estimated HIV inpatient care costs from charges [20, 21, 32], and some have used costs reported by hospitals [18, 19]. It is possible that neither charges nor hospitals' reported costs reflect the actual opportunity costs of providing services because of cost shifting [34]. Cost shifting may be practiced by hospitals to maximize their revenues from third-party payers and to provide services to uninsured patients who are unable to pay [35]. Despite the limitations in estimating costs, it is clear from the increasing number of HIV and tuberculosis hospitalization days in our study that inpatient care costs for both illnesses have increased.

In our study, we found an increase in the prevalence of comorbidity of HIV infection and tuberculosis among patients from 1985 to 1990. Among HIV-infected patients, the prevalence of tuberculosis increased twofold; among patients with tuberculosis, the prevalence of HIV infection increased 3.7-fold. Since 1990, explosive outbreaks of multidrug-resistant tuberculosis among persons hospitalized with HIV infection have resulted in significant morbidity and mortality [2, 11-16]. As the prevalence of comorbidity of tuberculosis and HIV infection increases among inpatients, the potential for the nosocomial transmission of tuberculosis can be expected to increase. Clearly, there is an urgent need for strict attention to guidelines for hospital infection control of tuberculosis [36].

Infection with HIV was more prevalent among patients with extrapulmonary tuberculosis than among those with pulmonary tuberculosis. This finding is consistent with previous reports, and most likely reflects the increased risk for disseminated tuberculosis disease in persons with HIV infection [6].

Drug dependence or abuse was a concurrent diagnosis in more than one fourth of the patients hospitalized with both HIV infection and tuberculosis. The simultaneous presence of HIV infection, tuberculosis, and drug abuse or dependence complicates clinical management [37] and warrants coordinated prevention and treatment of these problems.

The findings of this study underscore the need to increase prevention of infection and disease from HIV infection and tuberculosis to reduce morbidity, hospitalizations, and the cost of care. Adherence to guidelines for the prevention, early identification, and treatment of HIV infection and tuberculosis is critical [38-43]. The early identification of HIV infection may reduce morbidity and hospitalizations for HIV-related illnesses, including tuberculosis. Many cases of tuberculosis in patients with HIV infection are believed to result from the reactivation of latent tuberculosis infection [44]; therefore, coinfected persons should be identified early and given preventive therapy in order to prevent future cases of tuberculosis. Moreover, the early identification of tuberculosis may facilitate the early identification of HIV infection, the initiation of antiviral therapy, and counselling to prevent the transmission of HIV infection.