Patient Characteristics Panel Report

(printer-friendly version - PDF, 297 kb)

Contents on this Page:

1. Introduction
   1. Panel Purpose and Process
   2. Discussion of Data Sources
   3. Conceptual Framework and Guiding Principles
2. Subpanel Discussions and Variable Templates
   1. Clinical Characteristics Subpanel
   2. Comorbidities Subpanel
   3. Sociodemographic Subpanel
3. Citations
4. History of the Panel
   1. Members and Affiliations
   2. HSR/RTI Contact Information

Views expressed are those of the panel participants and do not represent official positions of the Federal Government

I. Introduction

A. Panel Purpose and Process

The Patient Characteristics panel was charged with identifying specific characteristics of HIV/AIDS patients that result in a greater need for services. To determine these characteristics, the panel members identified and conducted three sequential tasks:

1. To ensure that the panel considered a comprehensive set of variables prior to assessing their validity, feasibility, and interdependence, the panel identified an initial set of variables thought to be important determinants of severity of need (SON).

2. The panel members split into three groups to better focus their efforts on measuring three types of patient-specific factors that may significantly increase the depth, scope, or necessary utilization of care and services for people with HIV:

• A clinical subpanel
• A comorbidities subpanel
• A demographic subpanel.

3. Each subpanel was responsible for completing a template for each variable in their area of consideration. The template defined the variable; identified the rationale for its inclusion; identified the potential sources of data for measuring the variable; assessed the validity, reliability, and potential bias of each variable; and suggested ways to address any underlying bias. The goal was to evaluate the value of each variable and develop a final set of recommendations for inclusion in the SON index. The full Patient Characteristics panel then reconvened to discuss the recommendations of the subpanel and identify the variables that should be included in the panel’s final recommendation to the larger SON expert panel.

Within each subpanel, the variables were evaluated based on their importance in determining resource needs for CARE Act services and in determining the current quality, cost, and availability of data used to measure the variables. The variables would then be forwarded to either the Area Characteristics panel, so they could consider how the variables might be better measured or applied at the area level, or to the Associated Costs panel, so they could cost out the impact of the characteristics on SON.
As identified in Table 1, the panel evaluated 20 variables, 5 of which were forwarded for possible inclusion in a SON index. Variables were excluded based on either a lack of sufficient data, or a lack of a sufficient rationale for inclusion. Variables that simply lacked sufficient data should be reconsidered in the future for possible inclusion if data becomes available. One variable the panel considered (insurance) was not forwarded because it was being considered by another panel.

Table 1. Variables considered and forwarded for possible inclusion in an HIV/AIDS SON index, by the Patient Characteristics Subpanel

In an effort to prioritize their work before breaking into smaller subpanels, panelists first met as a full group to develop a list of variables to evaluate subjectively each variable’s contribution to SON. Before voting to rate the variables, the group eliminated 12 of 20 variables. Panelists were asked to score each variable from 1 to 5 based on how well each measured the theoretical concept of SON, with 1 indicating a variable of the highest importance and 5 indicating a variable of the lowest importance. These scores were then compiled and the averages ranked. The results are presented in Table 2.

Table 2. Area characteristics variables (considered important by the panel with sufficient data currently available) forwarded to the full panel and panelists’ priority score

^ back to top

B. Discussion of Data Sources

Once the variables were defined, the panel then identified the potential sources of data. Panelists were able to share knowledge of the existing sources, discuss challenges inherent in using the sources, and identify critical gaps within the sources. As part of this discussion, the panel asked for clarification on the use of past studies and asked if they could recommend new data to collect, either through surveys or by adding data elements to existing surveys. Health Systems Research (HSR) and RTI International (RTI) clarified that the panel should focus on existing data sources, with an emphasis on those that are of high quality, lower in cost, and available and updated on a regular basis. Although past studies were used in part to determine the relevance of data elements, the Panel agreed that future data would need to come from sources that are regularly updated. The panel could make recommendations to the Health Resources and Services Administration (HRSA) to collect new data, but the inclusion of a variable could not be dependent upon proposed collection of new data.

The panelists identified various challenges associated with using the data sources considered, and did not include key elements or cofactors. In general, many data sources:

• Did not identify people with HIV
• Did not appropriately capture CD4 counts
• Did not involve a large enough sample
• Were not available at the county level
• Were nor complete and accurate for other reasons
• Did not include those “not in care”
• Did not provide unduplicated counts
• Were not updated quickly and consistently.

Ultimately, the panel felt that the following four Centers for Disease Control and Prevention (CDC) data sources, despite their limitations, might be the most appropriate:

• HIV/AIDS Reporting System (HARS) National Surveillance Database. HARS contains the records of all persons confidentially reported by name that have HIV/AIDS
  • Advantages: Captures race/ethnicity, sex, disease status, insurance data, in care/not in care, some clinical data, zip code, and county; states are coming online constantly.
  • Limitations: Degree of completeness varies; CDC does not accept HIV data from 13 states that are using coded identifiers (must be name-based data); data are based on information at time of initial diagnosis.
• Morbidity Monitoring Project (MMP). MMP is a surveillance system designed to collect information from HIV/AIDS patients who received care from randomly selected HIV care providers.
  • Advantages: Locally representative (19 states, 1 territory, 5 cities, LA county); supplements core HIV/AIDS surveillance data with linked medical record abstractions and patient interviews; provides data to estimate quality of care, clinical outcomes, risk behaviors, health care utilization, and unmet needs among HIV-infected persons receiving medical care.
  • Limitations: Not nationally representative; county-level data unavailable except for LA county. Data may not be available for several years.
• National Health Interview Survey (NHIS). NHIS is the principal source of information on the health of the civilian noninstitutionalized population of the United States.
  • Advantages: Displays health characteristics by many demographic and socioeconomic characteristics.
  • Limitations: Does not identify people with HIV; only publishes national-level data.
• National Health and Nutrition Examination Survey (NHANES): NHANES is a unique data source in that it combines interview survey data with physical examination data. It is the primary source of national-level serologically-based estimates.
  • Advantages: Allows the estimation of serologic prevalence among self identified risk groups (for example those who have ever injected drugs, or those who have ever had a same-sex attraction); includes demographic, socioeconomic, nutrition, and health-related data, as well as biomedical measurement data, prevalence, and risk factors; identifies people with HIV.
  • Limitations: Only provides national estimates; estimates among some subgroups are too unstable to use due to sample size limitations.

^ back to top

C. Conceptual Framework and Guiding Principles

The panel developed a conceptual framework and series of guiding principles that guided the subsequent discussions of the panel. The panel started by describing the patient-specific factors they believed would significantly increase the depth, scope, or necessary utilization of care and services for people with HIV. When creating this list, the panel used the following seven guiding principles:

• Defining target level of care: Recommendations would differ depending on whether the group considered the impact of patient characteristics on optimal care versus usual care. For instance, active injection drug use may result in lower outpatient costs (and predictably poor outcomes) because the usual level of care by these patients is low, but may result in much higher cost to provide optimal care for the same patient if outreach, harm reduction, counseling, on-site methadone or buprenorphine, and multiple other services are provided. When identifying the factors that affect utilization, the panel used a criterion of impact on usual care, assuming those data might be more available.
• Focusing on direct medical services and those that support access to care and treatment: Although the panel would have preferred to use optimal care, the established guidelines and the limited ability to estimate additional costs required for optimal care, restricted their consideration to the usual medical services. Panelists chose to focus on the likely impact on categories of service that are funded by the CARE Act within the scope of ambulatory care services, including antiretroviral (ARVs) therapies, case management, mental health care, substance use/harm reduction services, adherence services, and oral health care. The final SON estimates of care utilization will need to take into account the entire spectrum of critical services, including those provided outside the medical clinic (e.g., referrals for mental health treatment, drug abuse treatment, adherence support), to be reasonable criteria for determination of funding. The panel recognized that limited data may exist on how to estimate the impact on resource needs for several of these measures.
• Type of increased need: Increases in SON may be manifested as increased need for care visits, screening, monitoring, or treatment and may take the form of higher costs for medications, laboratory tests, or human resources.
• Duration of need: We have only included those conditions that increase SON over time or those with potential for chronic increased SON. This assumption considers that acute conditions will likely be less cost-intensive over time and be more evenly distributed across all HIV populations (e.g., HIV-related acute pneumonias or medication-related hepatitis or acute sexually transmitted infections [STIs]). The one exception would be hospitalization for new opportunistic infections (OIs); however, because inpatient services are not paid for by the CARE Act, this exception is not relevant to the model. This assumption is critical because funding allocations based on severity will likely be based on retrospective data that will affect allocations over several years.
• Exclusion of unrelated comorbidities: Only comorbidities caused by or complicated by HIV and its treatment have been included in the list. This list is incomplete; additional categories may determine whether specific conditions result in increased need because of increased monitoring (visits and/or labs), treatment (services, medication, or screening tests to confirm diagnosis), or supportive services important for successful treatment (mental health, adherence services, drug treatment, case management), as well as any correlation between these services.
• Impact of late entry to care and poor adherence to care: The panel considered factors resulting in late entry into care and poor treatment adherence because both are associated with higher morbidity, more frequent complications, and more expensive monitoring and medications. Panelists asserted that these factors would balance the savings from visits, labs, and medications foregone during the period when the patient was out of care.
• Medication complexity and interactions: Panelists initially considered any factor that increased a patient’s number of prescriptions as a factor that also increases cost of care and complexity, because additional work is required to (1) coordinate prescribing with other clinicians, (2) increase monitoring frequency and types of laboratory tests, (3) manage more side effects with advice, visits, and symptom-controlling medications, (4) and monitor and manage drug-drug interactions.

The group then used these seven guiding principles to identify and characterize the patient-specific factors they believed would significantly increase the depth, scope, or necessary utilization of care and services for people with HIV. The patient-specific determinants of resource needs are summarized in Table 3.

^ back to top

II. Subpanel Discussions and Variable Templates

The following section outlines the rationale for recommending each of the selected variables and explains why other variables were excluded. The section is divided into two subsections: the first subsection discusses variables forwarded by the panel as potential elements for a SON index, and the second subsection discusses variables that were not forwarded. A short description is provided for each variable followed by a standardized data evaluation template that guided the investigation and discussion.

^ back to top

A. Clinical Characteristics Subpanel

The Clinical Characteristics subpanel was charged with identifying clinical variables measured at the patient level that would be predictive of need. The subpanel forwarded the HIV/AIDS disease progression variable, which all members agreed was most important. They felt ARV drug resistance was likely predictive of higher need, but was unable to identify a data source to measure it. The subpanel recommended that the comorbidities subpanel evaluate the risk behavior and exposure categories. The full panel agreed with the recommendations of the subpanel, which are outlined in Table 4.

Table 4. Clinical Characteristics Subpanel variables for consideration

1. Variables forwarded for consideration

HIV/AIDS disease progression: The subpanel forwarded HIV disease progression based on the assessment that patients with more advanced destruction of their immune system would require greater resources. Panelists identified the CDC HIV/AIDS reporting system (HARS) as a data source that currently provided information on patient CD4 counts as well as HIV/AIDS designation. Although the variable was forwarded, panelists were concerned that CD4 counts as reflected in HARS might not reflect current severity of need as they were only a measure of the patient’s low point (nadir) of immune destruction. Many patients in fact rebound quite dramatically from this nadir following the use or ARVs. The panel agreed that to measure disease progression requires repeated observations. In addition, HARS provides only a snapshot and does not reflect the ongoing costs over time. Finally, some states have limited data due to less-sophisticated reported systems. However, despite these limitations, the subgroup felt that nadir measures of CD4 counts provide better information on disease progression than no information at all.

It is important to note that CDC is currently transitioning states to reporting of HIV/AIDS cases through eHARS. A complete transition is expected within the next 2 years. Implementation of eHARS will allow states to report repeated CD4 observations for individual patients into the database. This will be especially beneficial for states with lab-based reporting for they will always have the most current CD4 counts for infected patients in their state. States currently have different levels of CD4 reporting which include those who report all counts, <500, <200, <50. However, all states with lab based reporting of CD4 counts, report CD4 counts of at least <50 which will catch persons most in need of ARV.

As a related issue to using CD4 counts, the panel noted that antiretroviral (ARV) therapy is helping HIV-positive persons live longer, meaning patients now need more care for chronic illness rather than for acute terminal illness. CARE Act grantees are more likely to be disabled (and less likely to have access to other kinds of resources) and are more likely to experience increased utilization in terms of visits, complications, and resources than in the past. Some panelists recommended use of the Morbidity Monitoring Project (MMP), along with HARS, to look at HIV/AIDS disease progression as a proxy for resource utilization.

In contrast other panelists were concerned that data from the MMP would not be useful for allocating CARE Act resources because it is currently fielded in only a subset of areas (i.e., 19 states, 1 territory, 5 cities, and 1 county), and because date from it may not be available for several years.

Further, the panel was concerned that rewarding jurisdictions with a greater number of cases with advanced HIV disease might penalize jurisdictions that successfully identify infections earlier or prevent people from progressing to advanced disease states. The panel noted that this may not be a major concern because the CARE Act does not pay for inpatient services, so a large portion of the shift of costs from early-stage patients to late-stage patients would not be included in the allocation. However, they recommended that HRSA remain cognizant of the issue. Potentially, the panel thought that jurisdictions that make an effort to diagnose people and identify HIV cases early could receive a supplemental allocation.

Nonetheless, the panelists agreed that the CARE Act should ultimately pay extra for patients in need, even if they are in need because of their State’s poor health policies. Therefore, the panel voted to forward disease stage variable to the Associated Cost panel to cost out the potential impact on SON.

HIV Disease Progression Template

2. Variables not forwarded for consideration

The Clinical subpanel did not believe that risk behaviors and exposure categories would have an ongoing contribution to the SON, except as they manifest as comorbidities. Therefore, they recommended that the Comorbidities subpanel consider both factors among their other variables.

Drug resistance: Panel members concluded that the increased risk of resistance from previous ARV history would require more intensive clinical care including more expensive drugs, more frequent outpatient visits and laboratory tests, a higher level of expert care, and additional medications to treat side effects of ARVs. They also agreed that there are probably significant differences among communities in terms of the frequency of resistance and frequency of patients with extensive ARV histories. Potential data sources for looking at these issues include cohort studies and cost-effectiveness models. However, the panel had the following unresolved questions regarding ARV drug resistance:

• The panel doubted that the per-patient impact on costs of developing drug resistance could be quantified given present data sources.
• The panel did not think a consensus existed on what would constitute appropriate care and drug therapy for patients with drug resistance.

The panel was uncertain that the costs associated with care of patients who have exhausted available treatment options could be estimated with present data sources. The subpanel discussed whether existing studies could be used to at least estimate the proportions of patients with primary drug resistance. They also recommended that future studies evaluate the costs of secondary resistance, which they thought was even more difficult to assess than primary resistance.

The panel felt drug resistance was important to consider, but due to data limitations did not forward it for inclusion.

Drug Resistance Template

^ back to top

B. Comorbidities Subpanel

The Comorbidities subpanel was charged with identifying specific comorbidity variables measured at the patient level that would potentially increase need. They agreed that as a complex, multisystem illness, HIV is influenced by other factors, such as general health and behaviors. Research indicates that the number of people with co-occurring conditions, such as coinfection with hepatitis C, is rising. Mental illness and substance abuse in particular, create unique challenges and can impact entry into care, adherence to treatment, level of risk behaviors, and utilization of services. Access to primary care, mental health, and substance abuse treatment, in addition to food, transportation, and housing, can in turn encourage entry and retention in care (Messeri et al., 2002; Lo et al., 2002: Wells et al., 2001).

The panel felt that certain comorbidities can significantly impact resource needs for CARE Act services. Unfortunately, it is difficult to measure the impact of comorbidities on SON, because challenges exist in the source, validity, reporting period, and definitions used to document them. Because of data limitations, estimates are largely reported among the general population instead of among the HIV-positive population. Keeping these limitations in mind, the panel made the recommendations outlined in Table 5.

Table 5. Comorbidities Subpanel Variables for Consideration

Of the comorbidities considered, the group only recommended including the intravenous drug use (IDU) exposure category. Several variables, including age-related comorbidities (including diabetes and cardiovascular disease), hepatitis C (which was seen largely as a subset of IDU exposure), mental illness (initially added by the panel for consideration), and substance abuse other than IDU, were not forwarded because the panel felt that data sources did not currently exist to measure them adequately. The panel was particularly conflicted about excluding mental illness and substance abuse because of their heavy impact on cost of care. There was also significant discussion surrounding the inclusion of IDU as a variable without also adjusting for need for other substance abuse services. Ultimately, the panel agreed to submit the IDU and other substance abuse and mental illness variables to the SON working group for further discussion. The panel strongly suggested that these variables should be considered in the future when better data become available.

Other variables, including tuberculosis (TB), gonorrhea, and syphilis, were not forwarded because the occurrence of illness is rare or because treatment is relatively cheap and available. There was some debate about whether to forward gonorrhea and syphilis because epidemiologically they are measures of risky behavior and possibly surrogate measures of HIV infection. However, the panel was not able to identify any potential data set that appropriately measures prevalence of gonorrhea and syphilis among HIV-positive persons.

It should be noted that the panel spent time discussing the potential use of Centers for Medicare and Medicaid (CMS) data before deciding to rely solely on CDC data for the included variables. Panelists acknowledged that it is theoretically possible to create a valid and reliable measure of comorbidities in AIDS patients who are eligible for Medicaid via the disability mechanism, but the measure is not feasible for this project. Challenges to using CMS data include the following:

• The data might accomplish what the group ideally wants (i.e., a consistent measurement of comorbidities within a defined and consistent population), but the process would be technically complex, and therefore time consuming and costly. Assuming the data could be acquired for free, the panel would recommend that HRSA dedicate future resources to think through the issues of identifying comorbidities using claims data, and develop such estimates using the MAX data. These estimates could be updated annually or biannually if resources allowed.
• A few issues of validity would persist regardless of the quality of the analysis:
• The MAX files may not include patients enrolled in Medicaid managed care. This could create systematic bias in the estimates between states.
• The MAX files obviously do not include any information on patients who do not enroll.
• It might be impossible to identify AIDS patients in the MAX files if patients received services only for comorbidities and not directly for HIV/AIDS during the year, or if they received treatment for comorbidities (e.g., sexually transmitted diseases) anonymously or at a facility that did not submit a claim to Medicaid.

1. Variables forwarded for consideration

IDU exposure category: Intravenous drug use is related to resource needs in several ways. IDU exposure increases the need for substance abuse services, the likelihood of extremely high rates of hepatitis C, the tendency to enter care at a late stage of disease progression, and the overall cost of primary care. The IDU exposure category is measured in the CDC HARS data, although the precise exposure category is missing for many patients. The panels felt strongly that the IDU-exposure category as measured by HARS was a strong indicator of severity of need.

However, the panel struggled with including a measure of IDU use without also including a measure of other substance abuse. The panel was concerned that adjusting for IDU use but not for other drug use (such as methamphetamines or crack cocaine) would weight the index in favor of the Northeastern US, because that is the region with the greatest number of IDUs.

In an attempt to address this issue the contractor worked with the CMS representative to evaluate the number of patients in Medicaid who had an HIV or AIDS diagnosis and who had a claim for substance abuse or mental health services. This data was not available for inclusion in this report.

The contractor also worked with Substance Abuse and Mental Health Services Administration (SAHMSA) to determine sources for measuring substance abuse and mental health services. Currently, the National Survey on Drug Use and Health (NSDUH) does not capture HIV infection. It is possible to estimate need for substance abuse services for several fairly general groups, but it is not possible to generate an estimate aggregated at a lower than national level. Fortunately, starting in 2007, NSDUH added a question asking respondents if a medical professional had ever told them they had HIV/AIDS. Assuming reasonably valid reporting on this question is available they will be able to cross substance use and mental health treatment need data with the HIV/AIDS variable.
In light of these issues and findings, the panel recommended including an estimate of IDU risk based on exposure category from the HARS data, with the caveat that the NSDUH data will be available soon and will potentially provide a reliable, valid source of data to estimate drug abuse of all substances among people with HIV. They felt this data should be evaluated as soon as it is ready and incorporated into the SON index if possible. They also recommended that data from CMS Medicaid files be evaluated once available.

Injection Drug Use Template