HIV disease burden, cost, and length of stay in Portuguese hospitals from 2000 to 2010: a cross-sectional study

Background The number of HIV-related hospitalizations has decreased worldwide in recent years owing to the availability of highly active antiretroviral therapy. However, the change in HIV-related hospitalizations in Portugal has not been studied. Using comprehensive hospital discharge data from mainland Portuguese hospitals, we examined trends in HIV-related inpatient admissions, length of stay (LOS), Elixhauser comorbidity measures, in-hospital mortality, and mean cost from 2000 to 2010. Methods The hospital administrative data from inpatient admissions and discharges at 75 public acute care hospitals in the Portuguese National Health Service from 2000 to 2010 were included. HIV-related admissions were identified using the International Classification of Diseases, 9th Revision, Clinical Modification diagnosis codes 042.x–044.x. The effect of Elixhauser comorbidity measures on extending the LOS was assessed by comparing admissions in HIV patients with and without comorbidities using the Mann–Whitney U test. Multivariate logistic regression was performed to estimate the odds of having a decreased discharge. Results A total of 57,027 hospital admissions were analyzed; 73% of patients were male, and the mean age was 39 years. The median LOS was 11 days, and the in-hospital mortality was 14%. The mean cost per hospitalization was 5,148.7€. A total of 83% of admissions were through the emergency room. During the period, inpatient HIV admissions decreased by 22%, LOS decreased by 9%, and in-hospital mortality dropped by 12%. Elixhauser comorbidities increased the median LOS in nearly all admissions. Conclusions Despite small regional variations, a strong, consistent decrease was observed in the hospital admission rate, mean cost, length of stay, and mortality rate for HIV-related admissions in Portugal during 2000–2010. Electronic supplementary material The online version of this article (doi:10.1186/s12913-015-0801-8) contains supplementary material, which is available to authorized users.


Background
Since the HIV epidemic emerged in Portugal in 1983, 39,347 cases have been reported, and as of 2010, 16,370 patients have developed advanced HIV infection [1]. Although the incidence of advanced HIV infection is declining worldwide, the population of persons living with HIV (PLWHIV) continues to rise [2]. Studies show that the number of HIV-related hospitalizations has decreased in recent years primarily because of the availability of highly active antiretroviral therapy (HAART) [3][4][5][6]. This decrease has been accompanied by an increase in the population of PLWHIV worldwide, which was estimated at 34 million people (31.6-35.2 million) through 2010, a 17% increase since 2001 [7]. HIV-infected individuals are surviving and experiencing longer life expectancies, as well as age-associated comorbidities [8]. Accordingly, the number of hospitalizations has emerged as an important outcome measure and contributor to healthcare expenditures in this population.
To determine the relationship between non-AIDS comorbidities and the progression of HIV disease, in the present study we employed the Elixhauser Comorbidity Index (ECI), which is a comorbidity classification system used to describe and compare patient populations. The ECI can adjust for confounding effects caused by the potential chronic disease burden when a condition is associated with a particular outcome [9]. In a previous study, we found that the ECI was associated with substantial increases in the frequency and duration of hospitalization, and mortality [10]. The ECI evaluates 30 comorbidities including HIV; however, we applied the ECI without including HIV as a comorbidity.
The impact on the Portuguese population of the reduction in HIV/AIDS-related morbidity and hospitalization in the HAART era has not been thoroughly characterized, and only one known study, which primarily examines mortality, has been published [11]. An understanding of the factors affecting HIV-related hospitalizations would help optimize resource allocation among regions and could potentially reduce the hospitalization rate for PLWHIV. The Portuguese health system comprises three coexisting systems: the National Health Service (NHS), special social health insurance programs for certain professions (health subsystems), and elective private health insurance. The NHS establishes the right of all citizens to health protection; all citizens have a guaranteed universal right to health care through the NHS. The patient pays a specific fixed cost per use for certain health care services delivered by NHS facilities, but HIV drugs are provided free of charge to all HIV patients [12,13].
Using comprehensive hospital discharge data from Portuguese hospitals, we examined the trends in HIVrelated inpatient admissions, length of stay (LOS), mean costs, ECI, and HIV in-hospital mortality from 2000 to 2010.

Methods
This cross-sectional study examined trends in HIVrelated inpatient admissions in Portugal. Hospital administrative data were collected from public acute care hospitals in the Portuguese NHS database maintained by the Central Health System Administration (ACSS). Anonymity was maintained for all hospitals and patients. The dataset included 10,586,118 discharges across 75 hospitals in mainland Portugal from January 2000 to December 2010. The dataset also included the principal diagnosis, defined as the principal condition causing the patient's admission to the hospital, and up to 19 additional diagnoses.
To capture all of the potential HIV hospital admissions, eligible cases met one of the following criteria: (1) principal diagnosis of HIV/AIDS as defined by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) with diagnosis codes of 042.x-044.x; (2) principal diagnosis of advanced HIV infection (i.e., pneumonia or opportunistic infection as defined by the Major Diagnostic Category 25, "Human Immunodeficiency Virus Infection" [14][15][16] and a secondary diagnosis of HIV/AIDS; or (3) principal diagnosis unrelated to HIV/AIDS and a secondary diagnosis of HIV/AIDS. Comorbidities were defined as diagnoses other than HIV/AIDS as defined above.
The country-wide distribution of hospitals was determined using the Nomenclature of Territorial Units for Statistics II (NUTS II), which is a geocode standard developed by the European Union for referencing the administrative divisions of countries for statistical purposes. The NUTS II level identifies the region by a two-digit or letter-digit code [17] and is only applied to the mainland regions of Portugal, namely Norte, Centro, Lisbon, Alentejo, and Algarve ( Figure 1).
From 2000 to 2010, five administrative regulations were issued by the Ministry of Health to assign and update the costs for inpatient admissions according to diagnosis-related groups (DRGs). Two DRGs were used Figure 1 Portugal Nomenclature of Territorial Units for Statistics II (NUTS II) regional map. The Norte, Centro, Lisbon, Alentejo, and Algarve regions were included in the present study.  [18]. The mid-year population of PLWHIV was calculated per year by summing the new and old cases, and dividing the result by two.

Analysis
Variations in the HIV-related hospitalization rates according to clinical characteristics were assessed by univariate analysis and a linear trends test. We first examined the number of HIV-related admissions per year according to the NUTS II region, median LOS, inhospital mortality, mean cost, admission type, age, and sex. The units of analysis comprised each combination of sex (two categories), age (four categories), NUTS II region (five categories), year (11 categories), LOS, mortality, and mean cost. The results are represented as the frequency, median, and interquartile range, or as the mean and 95% confidence interval (CI), as appropriate.
The impact of the ECI on the LOS was determined by comparing admissions in patients with and without comorbidities using the Mann-Whitney U test at a 0.05 level of significance. Multivariate logistic regression was performed to estimate the odds of having a decreased discharge status according to all HIV-patient and independent variables, including age (first age category served as the standard: 0 to 17 years, 18 to 45 years, 46 to 65 years, and 66 years and older), gender (males compared with females), LOS, and Elixhauser comorbidities (each comorbidity admission was compared with another admission without that comorbidity), with a p-value < 0.05 indicating statistical significance. The variables (comorbidities) included in multivariate logistic regression analysis were selected based on the results of a univariate analysis. Despite the large number of variables (comorbidities), their use is relevant given their influence on hospital outcomes, particularly on mortality, LOS, and the use of hospital resources. In addition, the study population is quite large and includes many deceased patients; therefore, many cases exist for each predictor and independent variable. The statistical analyses were performed using SPSS version 21 (SPSS IBM, New York, USA).

Ethical standards
This study employed secondary anonymous data and was performed in accordance with the ethical standards established by the 1964 Declaration of Helsinki and its later amendments. The data were used for scientific research purposes under the Medicine Faculty of University of Porto and Central Health System Administration agreement. During this time, 8,203 deaths were reported, and the mortality rate decreased from 14.9% to 13.1% (Table 2).

Results
When examined by the NUTS II region (Table 3), admissions were most frequent in Lisbon and Norte, which was expected as these are the most populous areas in Portugal. From 2000 to 2010, the number of inpatient admissions decreased by 35% in Lisbon, 22% in Algarve, and 4% in Norte, but increased by 31% in Centro and approximately 20% in Alentejo. The in-hospital mortality rate in Norte decreased by 40% over the period, from 19.1% in 2000 to 11% in 2010, while in Centro, the mortality decreased by only 24%, dropping from 10% in 2000 to 7.7% in 2010. In Lisbon, Alentejo, and Algarve, the mortality rates increased by 3%, 34%, and 52%, from 14.5%, 17%, and 8. The median LOS was 11 days (interquartile range 5-22 days). Between 2000 and 2010, the median LOS across mainland Portugal decreased by 25% (from 12 to 9 days) in the Norte region, 18% (from 11 to 9 days) in the Centro and Alentejo regions, and 8% (from 12 to 11 days) in the Lisbon region, while the median LOS in the Algarve region increased by 8% (from 11 to 13 days). The median LOS was significantly higher in men (12 days; interquartile range 5-23 days) than in women (10 days; interquartile range 2-23 days). This difference was statistically significant (p < 0.001) using the Mann-Whitney U test (data not shown).
Hospitalization occurred primarily in two ways: through the emergency room or through an ambulatory consultation. When the data were examined according to the site of admission, we found that 83% of hospitalizations originated from the emergency room.  Table S1).
The mean cost per HIV admission decreased during the study period and substantially decreased in the Norte, Centro, and Lisbon regions. The mean cost in the Norte region was 5,499€ (95% CI 5,555€-5,444€) over the entire study period, with 2003 and 2004 as the most expensive years (7,137€ and 7,181€; 95% CI 7,006€-7,269€ and 7,058€-7,305€, respectively). In Centro, the overall mean cost was 4,739€ (95% CI 4,822€-4,657€), with 2003, 2004, and 2005 as the most expensive years (6,393€, 6,593€, and 6,474€; 95% CI 6,183€-6,603€, 6,372€-6,814€ and 6,243€-6,704€, respectively). In  Table 4). The mean daily cost was 253€ and increased slightly over the decade from 248€ to 260€ (4% increase), while the estimated cost per year decreased considerably from 19 to 15 million euros over the decade (12% decrease). Overall, an estimated 208 million euros were spent on patients with HIV and  related diseases. DRGs 710 (21%) and 714 (19%) were the most expensive (Additional file 1: Table S2 and  Table S3). Based on the ECI, drug abuse was identified in 30% of patients, liver disease in 24%, and fluid and electrolyte disorders in 9%. Nearly every comorbidity significantly increased the LOS. The comorbidities associated with the greatest increase in the median LOS were paralysis (12 days), valvular heart disease (8 days), and hemorrhage (8 days; Table 5).When the Elixhauser comorbidities were examined by gender, because of the predominantly male population, most comorbidities occurred in men (73%). When the number of comorbidity cases was examined by sex per thousand episodes (men 41.834 and women 15.210), twice as many women as men were diagnosed with chronic pulmonary disease, uncomplicated hypertension, and depression.
We also examined the number of etiologies causing the hospitalizations; the cause of hospitalization was categorized as related or not directly related to HIV. HIV- related hospitalizations decreased consistently over the study period from 82% to 75%, while hospitalizations unrelated to HIV increased from 18% to 25%. Deaths directly caused by HIV decreased from 16% to 14%, while deaths from causes unrelated to HIV increased from 9% to 12% over the decade (Additional file 1: Table S4).
In the multivariate regression analysis (adjusted and unadjusted) of HIV admissions between 2000 and 2010, we examined the variables influencing the risk of death. The risk of death increased primarily with age. Men had a 1.5 odds ratio for death compared with women. The LOS did not affect the odds ratio for death. Comorbidities, such as metastatic cancer, weight loss, fluid and electrolyte disorders, lymphoma, pulmonary circulation disorders, and cardiac arrhythmias, increased the death odds ratio more than two times compared with admissions without those comorbidities (Table 6).

Discussion
The decline in HIV-related hospital admissions during the final 4 years of the study period (2007)(2008)(2009)(2010) was in opposition to the trend over the first 7 years (2000)(2001)(2002)(2003)(2004)(2005)(2006) in which the number of episodes increased. The HIV incidence appears to have decreased by nearly 50% in one decade. These data indicate success in controlling HIV in Portugal. This success is good considering that the worldwide HIV incidence is also decreasing (nearly 20% over the study period) [19]. Because HIV is a chronic disease, affected patients experience numerous acute exacerbations. In the present study, 80% of patients were hospitalized because of an acute illness, which highlights the challenge of clinical management and control of HIV. These acute episodes are more common in men because they form the majority of HIV patients in Portugal, as in the rest of Europe [20]. Worldwide, half of HIV patients are men, though in certain regions, such as Africa and the Caribbean, women predominate [7]. Over the study decade, the median age increased by 10 years, confirming that many PLWHIV are surviving longer [2,21,22].
The overall inpatient LOS decreased during the decade, though the rate increased during the most recent years. This trend may reflect improved access to outpatient care and better adherence to HAART [23]. Fleishman and Hellinger [3] studied trends in HIVrelated inpatient admissions in seven states in the USA (1996)(1997)(1998)(1999)(2000) and found a comparatively shorter LOS (median 6 days) that was half the LOS observed in this study. Characteristics unique to the Portuguese NHS may explain this difference. One reason could be that to discharge a patient in Portugal, several guarantees should be met, such as adequate family support, and continued care at home or at a center providing social assistance; these requirements likely increase the LOS. When the LOS for HIV-related admissions was compared with the LOS for all other admissions, HIVrelated admissions showed a greater median LOS than all other admissions. While the LOS for HIV-related admissions decreased compared with all other admissions, this trend was not statistically significant.
Trends in mortality varied across regions, with Norte and Centro differing significantly from Lisbon, Alentejo, and Algarve. Overall, the in-hospital mortality rate decreased by 12%. In the Norte and Centro regions, the decrease between 2000 and 2010 was approximately 40% and 24%, respectively. In Lisbon and Algarve, the mortality rate increased steadily during the period; this result contrasts with the overall decreased mortality over the same period. Many mechanisms may underlie this discrepancy; the treatment methods and patient follow-up protocols are similar across the country, while the absolute frequency of admissions varies by region. The Alentejo and Algarve regions had fewer episodes; therefore, one death may have greater weight than a death in the northern regions where there are more HIV inpatient admissions. Thus, the discrepancy may be due to significant differences in the population itself, rather than in the quality of care. Our results show a substantial decrease in the mortality rate. This decrease might be due to HAART or improved monitoring of HIV patients through outpatient hospital services and health centers in the area of residence. These data confirm previously published data showing that the HIV-related mortality in Europe has decreased since 1996 [24,25].
The mean estimated cost increased from 2000 to 2001, stabilized with little fluctuation from 2001 to 2005, and decreased steadily from 2006 to 2010. There were no regional differences, and the admission complexity did not appear to vary over the period. Even considering that the prices are set by the Ministry of Health, the mean cost showed a clear decreasing trend, which may be due to the decreased number of cases or improved clinical management over the decade. However, the mean daily cost per admission increased over the same period. These results contradict the lower daily costs reported by Perelman et al. [26].
Regression analyses for the risk of HIV inpatient mortality were performed for age, sex, LOS, and ECI. The risk of death increased over four-fold with age, but the difference was not significant in patients older than 18 years, as indicated by the overlap in the 95% CIs. Men had a greater risk of death than women did, and the LOS did not influence the risk of mortality. Most comorbidities increased the risk of death. Metastatic cancer increased the risk three-fold. Several comorbidities did not influence the risk of death, including complicated or uncomplicated hypertension, chronic pulmonary disease, obesity, nutrient-deficiency anemia, psychosis, and depression.
This study used comprehensive discharge data compiled in mainland Portugal. Thus, the findings are more generalizable than results based on data from a single hospital. However, this study has limitations, due primarily to the nature of the data [27]. First, we cannot identify the number of unique patients, the number of times that one patient was hospitalized at a single facility, or the number of years in care. Important clinical information, mainly complementary exams (CD4 cell count, viral load) and the therapy used, was not available. To track the long-term outcomes and quality of care, an electronic database or electronic health record system specially designed to capture these important components of HIV-related care is necessary.

Conclusions
We found a strong, consistent decrease in the hospital admission rate, cost, LOS, and mortality rate during