A survival analysis of hospitalization among patients with acquired immunodeficiency syndrome

Deep Multi-Dictionary Learning for Survival Prediction With Multi-Zoom Histopathological Whole Slide Images

Survival prediction based on histopathological whole slide images (WSIs) is of great significance for risk-benefit assessment and clinical decision. However, complex microenvironments and heterogeneous tissue structures in WSIs bring challenges to learning informative prognosis-related representations. Additionally, previous studies mainly focus on modeling using mono-scale WSIs, which commonly ignore useful subtle differences existed in multi-zoom WSIs. To this end, we propose a deep multi-dictionary learning framework for cancer survival prediction with multi-zoom histopathological WSIs. The framework can recognize and learn discriminative clusters (i.e., microenvironments) based on multi-scale deep representations for survival analysis. Specifically, we learn multi-scale features based on multi-zoom tiles from WSIs via stacked deep autoencoders network followed by grouping different microenvironments by cluster algorithm. Based on multi-scale deep features of clusters, a multi-dictionary learning method with a post-pruning strategy is devised to learn discriminative representations from selected prognosis-related clusters in a task-driven manner. Finally, a survival model (i.e., EN-Cox) is constructed to estimate the risk index of an individual patient. The proposed model is evaluated on three datasets derived from The Cancer Genome Atlas (TCGA), and the experimental results demonstrate that it outperforms several state-of-the-art survival analysis approaches.

Model Comparison for Breast Cancer Prognosis Based on Clinical Data

We compared the performance of several prediction techniques for breast cancer prognosis, based on AU-ROC performance (Area Under ROC) for different prognosis periods. The analyzed dataset contained 1,981 patients and from an initial 25 variables, the 11 most common clinical predictors were retained. We compared eight models from a wide spectrum of predictive models, namely; Generalized Linear Model (GLM), GLM-Net, Partial Least Square (PLS), Support Vector Machines (SVM), Random Forests (RF), Neural Networks, k-Nearest Neighbors (k-NN) and Boosted Trees. In order to compare these models, paired t-test was applied on the model performance differences obtained from data resampling. Random Forests, Boosted Trees, Partial Least Square and GLMNet have superior overall performance, however they are only slightly higher than the other models. The comparative analysis also allowed us to define a relative variable importance as the average of variable importance from the different models. Two sets of variables are identified from this analysis. The first includes number of positive lymph nodes, tumor size, cancer grade and estrogen receptor, all has an important influence on model predictability. The second set incudes variables related to histological parameters and treatment types. The short term vs long term contribution of the clinical variables are also analyzed from the comparative models. From the various cancer treatment plans, the combination of Chemo/Radio therapy leads to the largest impact on cancer prognosis.

Length of mechanical restraint following haloperidol injections versus oral atypical antipsychotics for the initial treatment of acute schizophrenia: a propensity-matched analysis from the Japanese diagnosis procedure combination database

Differences in effectiveness between haloperidol injection and oral atypical antipsychotics in the acute-phase treatment of schizophrenia are not well examined. We retrospectively investigated whether these treatment options affected the length of mechanical restraint. We used the Japanese Diagnosis Procedure Combination Database to identify schizophrenia patients who were involuntarily hospitalized and receiving mechanical restraint between July and December, 2006-2009. Data included patient demographics, use of antipsychotics, and number of days on which patients underwent mechanical restraint. Propensity score matching was performed to compare the number of days of mechanical restraint between the haloperidol injection group and the oral atypical antipsychotics group. We used survival analysis to examine whether the initial difference in treatment affected the number of days of mechanical restraint. Cox regression was performed to compare the concurrent effects of various factors. Among 1731 eligible patients, 574 were treated with haloperidol injections and 420 with atypical antipsychotics. Matching produced 274 patients in each group. Cox regression analysis showed that the initial therapeutic agents did not significantly affect the number of days of mechanical restraint. The results indicate that atypical antipsychotics were as effective as haloperidol injections in the acute-phase treatment of schizophrenia.

Survival patterns among HIV+ individuals based on health care utilization

The purpose of this study was to determine if HIV+ persons who first obtained health care in New Orleans through public hospital inpatient services had a higher risk of death or disease progression than patients who first entered care through public outpatient services. The sites included the largest HIV outpatient clinic in the Gulf South, two early intervention sites and a public hospital. A medical record review on patients who attended these sites from July 1995 through December 1999 and were enrolled in the Adult Spectrum of Disease (ASD) Study was conducted (n = 3402). The multivariate analysis examined the associations between inpatient services and the main effects. Kaplan-Meier analysis and Cox proportional hazards regression were performed. Risk of death or disease progression was analyzed for three different endpoints: time from study entry to death, time from HIV to AIDS, and time from AIDS to death. The multivariate analysis showed that patients first entering care through inpatient services were significantly more likely to be African American, have AIDS, and use drugs. The risk of death or disease progression was significantly higher for all three endpoints. Results from this study indicate that HIV+ individuals receiving initial care through public hospital inpatient services may require more effective early intervention.

How good are US studies of HIV costs of care?

BACKGROUND

Valid, timely estimates of the costs of HIV care are needed by health planners and policy makers.

OBJECTIVE

To perform a methodologic critique of published estimates of resource utilization and costs of HIV care.

DATA SOURCES

MEDLINE database for 1990-1998.

DATA SELECTION

Included articles focused on adults with a spectrum of HIV disease in which the authors developed their own resource use and cost data. Thirty one articles met these criteria.

DATA EXTRACTION

Studies were compared based on: (1) utilization and cost estimates, in 1995 dollars; (2) study period; (3) research design; (4) sampling frame; (5) sample size and patient characteristics; (6) data sources and scope of services; and (7) methods used in the analysis.

DATA SYNTHESIS

The most recent estimates pertain to the first half of 1995, before the use of protease inhibitor therapy. We found wide variations in the estimates and identified three major sources for this: (1) patient samples that were restricted to subgroups of the national HIV-infected population; (2) utilization data that were limited in scope (e.g., inpatient care only); and (3) invalid methods for estimating annual or lifetime costs, particularly in dealing with decedents.

CONCLUSIONS

To accurately estimate resource use and costs for HIV care nationwide, a nationally representative probability sample of HIV-infected patients is required. Even in research that is not intended to provide national estimates, the scope of utilization data should be broadened and greater attention to methodologic issues in the analysis of annual and lifetime costs is needed.

Methods for analyzing health care utilization and costs

Important questions about health care are often addressed by studying health care utilization. Utilization data have several characteristics that make them a challenge to analyze. In this paper we discuss sources of information, the statistical properties of utilization data, common analytic methods including the two-part model, and some newly available statistical methods including the generalized linear model. We also address issues of study design and new methods for dealing with censored data. Examples are presented.

On the use of survival analysis techniques to estimate medical care costs

Measurement of treatment costs is important in the evaluation of medical interventions. Accurate cost estimation is problematic, when cost records are incomplete. Methods from the survival analysis literature have been proposed for estimating costs using available data. In this article, we clarify assumptions necessary for validity of these techniques. We demonstrate how assumptions needed for valid survival analysis may be violated when these methods are applied to cost estimation. Our observations are confirmed through simulations and empirical data analysis. We conclude that survival analysis approaches are not generally appropriate for the analysis of medical costs and review several valid alternatives.

Enteral glutamine supplementation for very-low-birth-weight infants decreases hospital costs

BACKGROUND

There is growing evidence that glutamine may be a conditionally essential amino acid for critically ill patients, including preterm infants cared for in neonatal intensive care units (NICUs). In a randomized study of 68 very-low-birth-weight (VLBW) infants, we found evidence of lower morbidity in a group fed glutamine-supplemented preterm infant formula from postnatal day 3 to day 30 than in a group fed a standard formula. We report here the effects of the glutamine supplementation on hospital costs in these infants.

METHODS

The costs were analyzed by log-rank tests and Kaplan-Meier plots.

RESULTS

The median costs for hospitalization, radiology, pharmacy, laboratory, and the NICU, and the median number of utilization units were reduced with glutamine supplementation.

CONCLUSIONS

This study provides the first evidence for decreased hospital costs in VLBW neonates who receive enteral glutamine supplementation.

Predictors of health care utilization in the chronically ill: a review of the literature

The objective of this paper is to identify predictors of health care utilization in the chronically ill. This paper reviews 53 studies on hospitalizations and physician visits, published between 1966 and 1997 and identified by MEDLINE and ClinPSYCH databases. Studies with both univariate and multivariate analyses were included. On the basis of the Andersen-Newman model of health care utilization, the effects of predisposing, enabling and need variables are examined. Most studies reviewed indicate that predisposing factors such as age, sex, and marital status are not predictors of hospital utilization in the chronically ill. The enabling factors income, insurance and social support have not been shown to affect health care utilization, but characteristics of the hospitals could have an effect. Need factors such as disease severity, symptom severity and complications adversely affected health care utilization in the chronically ill, while disease duration and comorbidity do not have such an effect. Quality of life and perceived health might affect hospital utilization and physician use. Finally, depression and psychological distress proved to be among the strongest predictors of hospitalizations and physician visits. In conclusion, both disease severity and psychological well-being are most important in health care utilization. Intervention programs to support depressed or psychologically distressed patients should be considered. These could both help the patient and reduce health care utilization costs.

Hospital care for persons with AIDS in the European Union

This study estimates the current and future hospital resources for AIDS patients in the European Union (EU), using multinational scenario analysis (EU Concerted Action BMH1-CT-941723). In collaboration with another EU-project ('Managing the Costs of HIV Infection'), six national European studies on the utilization of hospital care for AIDS have been selected to provide the data for our analysis. The selection criteria involve recentness, quality, comparability, accessibility and representativeness. Baseline hospital resource utilization is estimated for hospital inpatient days and outpatient contracts, using a standardized approach controlling for two severity stages of AIDS (chronic stage and late stage). The epidemiological part of the study is based on standard models for backcalculating HIV incidence and projecting AIDS incidence, prevalence and mortality. In the next step, baseline resource utilization is linked to epidemiological information in a mixed prevalence and mortality-based approach. Several scenarios render different future epidemiological developments and hospital resource needs. For the year 1999, hospital bed needs of 10,000-12,700 in the EU are indicated, representing an increase of 20-60% compared to the estimated current (1995) level. The projected range for 1999 corresponds to a maximum of 0.65% of all hospital beds available in the EU. The growth in the number of outpatient hospital contacts is projected to possibly exceed that of inpatient days up to 1.82 million in 1999. Our methodology illustrates that estimation of current and future hospital care for AIDS has to be controlled for severity stages, to prevent biases. Further application of the multinational approach is demonstrated through a 'what-if' analysis of the potential impact of combination triple therapy for HIV/AIDS. Estimation of the economic impact of other diseases could as well benefit from the severity-stages approach.

The influence of health-related quality of life and social characteristics on hospital use by patients with AIDS in the Boston Health Study

OBJECTIVES

The authors examine whether health-related quality of life (HRQL) and social factors were independent predictors of future hospital use for persons with acquired immunodeficiency syndrome (AIDS).

METHODS

A panel of 305 patients with AIDS treated at three provider settings in the Boston, Massachusetts area were enrolled during 1990 and 1991. Data were collected at baseline study enrollment and again 4 months later. Patient interviews, hospital bills, and medical charts were used to measure hospital use (admissions and days during the 4 months after enrollment), sociodemographic characteristics (age, gender, race, education, insurance, homelessness, alcohol use, and AIDS risk factors), disease burden (patient severity and a three-level opportunistic diseases and complications score), HRQL (patient-reported symptoms, activities of daily living, neuropsychological status, and global health assessment), system of care, and use of prophylactic drugs. Logistic regression was used to estimate the odds of admission. Total days of hospital care by patients with at least one admission were analyzed using multiple linear regression. Clinical models of hospital use were developed first from the variables measuring disease burden and system of care. Models estimating the associations between hospital use and all other predictor variables measured at baseline then were estimated using stepwise techniques, controlling for variables in the core model.

RESULTS

Patients were more likely than their reference groups to be hospitalized if they had serious opportunistic diseases (adjusted odds ratio [OR] = 2.7), had poorer neuropsychological status (OR = 1.9), were non-white (OR = 2.0), or were homeless (OR = 3.3) (all P < or = 0.05). Activities of daily living were associated moderately (OR = 1.3; P = 0.07). Only system of care and neuropsychological status predicted total hospital days.

CONCLUSIONS

The results indicate that future hospital use by persons with AIDS may be influenced by social and other health-related factors in addition to the more clinically related characteristics that are recorded in a medical chart. It therefore may be appropriate to assess these factors when considering options for intervention or when comparing patterns of use among patient groups or settings.

The severity classification system for acquired immunodeficiency syndrome hospitalizations. Association with survival after discharge and inpatient resource use

The Severity Classification for AIDS Hospitalizations (SCAH) was applied to a longitudinal person-based data set of Maryland adult residents diagnosed with acquired immunodeficiency syndrome (AIDS) between 1983 and 1989 to predict long-term survival. In contrast to other AIDS severity measures, SCAH can be applied to administrative data bases for analyses of large populations. Although SCAH was created to predict inpatient mortality using cross-sectional hospital discharge data, the models used in this study show SCAH stage at first AIDS hospitalization to predict long-term survival in persons with AIDS, even after adjusting for sociodemographic and treatment variables. Additional models in the study show SCAH stage at first hospitalization has a strong association with inpatient length of stay and associated charges, making it useful for health care resource planning.

Lifetime hospitalization profiles for symptomatic, HIV-infected persons

We explored the relationship between the incidence of hospitalization and disease progression in a group of 140 symptomatic, HIV infected patients by linking hospitalizations to the time of diagnosis, the time of death, or both. The relationship could best be described by positively skewed U-patterns or (weak) J-patterns with a high use of resources immediately following diagnosis and preceding death. The lifetime hospitalization profiles differed according to the type of insurance, age, the initial diagnosis in the CDC-IV stage and the length of survival. The results not only confirm general hypotheses posed by other research groups, but also demonstrate the existence of variations among subgroups of patients. The results can be used to improve economic assessments of the impact of AIDS in The Netherlands and the European Union. The method used has the advantage of being based on a bottom-up approach to resource utilization, involving the use pf prospective data for the patients' full lifespans, and can easily be applied to other areas of health services research.

Inpatient health care utilization for patients with HIV and AIDS in the Edinburgh City Hospital

A retrospective analysis of HIV-positive patients admitted to the City Hospital, Edinburgh by 31st December 1992, 7 years after the inception of the HIV clinic, revealed that 373 patients, 72% of them injection drug users (IDUs), had required 2069 admissions (5.5/admitted patient, 3.3/clinic patient or 0.5/clinic patient/year) and had utilized 21,934 bed days (59 bed days/admitted patient, 35 days/clinic patient or 5 bed days/clinic patient/year). The average length of stay (ALOS) was significantly longer for AIDS than HIV (non-AIDS) admissions (14.0 vs 9.5, P < 0.0001) as it was for admissions with a CD200 diagnosis (a CD4 count below 200 cells per mm3 on two consecutive occasions) compared to those without (12.1 vs 10.0 days, P = 0.004). There was no gender effect on ALOS but there was a significant effect of risk activity; homo/bisexuals had a significantly longer ALOS than drug users (P < 0.0001). Homo/bisexual patients with AIDS or a CD200 diagnosis had longer ALOS than drug users (15.7 vs 13 days and 15.8 vs 10.8). By 1992 each member of the clinic was on average utilizing one admission per year and 11.6 bed days per year. The number of admissions in that year for patients without a CD200 or AIDS diagnosis was however low (0.5 and 0.75 admissions/patient/year) compared to patients with an AIDS or CD200 diagnosis (2.6 and 1.6 admissions/patient/year). The annual number of occupied bed days/living patient was greatest for those with AIDS (60 vs 5 days) or with a CD200 diagnosis (23.5 vs 4.1 days).(ABSTRACT TRUNCATED AT 250 WORDS)

Trends in hospital resource utilization by HIV-infected persons, January 1987-June 1990

Trends in the utilization of various hospital resources by HIV-infected persons between January 1987 and June 1990 have been studied to support health care planning. Data on 126 asymptomatic and symptomatic HIV-infected persons have been recorded at a patient level and analyzed at half-yearly intervals. At a hospital level, increasing utilization trends were observed. At the patient level, a decreasing utilization intensity was measured for admissions, inpatient days, inpatient diagnostic examinations, and outpatient consultations. Increasing utilization intensity was measured for care during admissions. A constant utilization intensity was observed for outpatient diagnostic examinations and inpatient medication days. Use of interventions tended to increase at the end of the study period. Discriminating between trends in the utilization of different hospital resources can improve the management of hospital health care demands of HIV-infected people.

Hospital bed usage by people with HIV disease: experience in a provincial setting

In order to describe hospital bed usage by people with HIV disease in a provincial setting, a retrospective analysis of admissions to hospital wards in Oxford was undertaken for people admitted to hospital with all HIV-related illnesses or complications of HIV-related treatment. A total of 83 people were identified as having been admitted to hospital between January 1986 and the end of August 1990. Average length of hospital stay, the number of admissions per observed person-year and the in-patient days per observed person-year decreased. Of the 2,446 days spent in hospital, 913 were by people with an AIDS diagnosis; 1,533 days were spent by people who did not fulfil the World Health Organisation/Centers for Disease Control (WHO/CDC) classification for AIDS but who were admitted because of their HIV disease. AIDS is an end-point of infection with HIV. Pre-AIDS morbidity, a spectrum of illness of increasing severity from minor illness up to the point of WHO/CDC level AIDS, is a major determinant of hospital care and has previously been underestimated. In order to calculate the best estimates of hospital care needed, HIV disease should be regarded as a single entity and the artificial barrier dividing HIV illnesses from AIDS should be discarded.

Patterns of hospital care for patients with HIV infection and AIDS

OBJECTIVE

The aims were to describe the use of inpatient resources by patients with HIV infection and patients with AIDS; to examine trends in service use over time; and to provide data for planners concerned with service provision for HIV infection.

SETTING

An inner London health district reporting 9% of AIDS cases nationally.

DESIGN

Data on survival times and inpatient and day care use of resources were derived from existing patient records or collected prospectively between 1983 (when the first case of AIDS was diagnosed in the district) and 31 March 1990.

SUBJECTS

A total of 488 HIV positive patients of whom 396 had been diagnosed as having AIDS were studied.

MEASUREMENTS AND MAIN RESULTS

Inpatient days consumed per annum; trends in the number of bed days per person year with AIDS; the lifetime inpatient use per AIDS patient; and the influence of survival on service use estimates were determined. Altogether 16.4% of a total 17,785 hospital inpatient days were attributable to HIV positive patients who did not fulfil the criteria for AIDS. For patients with AIDS, there was an initial increase in the intensity of inpatient use in 1987 when a dedicated HIV ward was opened. After 1988, however, inpatient use fell to 30.8 bed days per person year with AIDS. Patients diagnosed after April 1987 had noticeably longer survival times than those diagnosed earlier (a median of 17-18 months compared with a median of 10-11 months). From 214 lifetime service use records, it was estimated that patients with short survival (less than six months) would consume 36 days of inpatient care, while those expected to survive longer would consume approximately twice that number of days, irrespective of how much longer they survived.

CONCLUSIONS

The data indicate less intensive use of inpatient care by AIDS patients over time, and hence the apparent ability to manage an increasing AIDS patient workload without a comparable increase in occupied bed days. Increases in the size of that workload and changes in the survival profile of patients, together with a relatively constant rate of service demand by longer survivors, however, will continue to place increasing strains on finite inpatient resources. Further research is needed to establish the extent to which the greater use of outpatient and community services can offset this.

Modelling length bias in a longitudinally-linked record system of HIV cases

We compare numbers of hospital admissions for intravenous drug using (IVDU) HIV patients and other HIV patients in acute-care facilities in New York State. Data consist of routinely collected hospital-discharge reports from New York's Statewide Planning and Research Cooperative System, linked into longitudinal case histories. Because recognition of an IVDU depends on an opioid diagnosis on any record in the case history, the observed distribution of the number of admissions per case for recognized IVDU's is biased towards greater numbers of admissions. We develop and apply a model to overcome this biasing. Our findings reveal that the mean numbers of admissions for the two groups differ significantly, but less so than without recognition of the length biasing.