Elder Abuse Workshop: Working Lunch - Measurement and Data: Working with Large Data Sets
Panel Presentations:
Jim Robinson — Using Nursing Home MDS and OSCAR Information
The Center for Health Systems Research and Analysis (CHSRA) assists in nursing home monitoring and investigations by:
- Supporting OIG CIA (Office of Inspector General Corporate Integrity Agreements) monitoring by providing data analysis information to support quality monitors such as the Long Term Care Institute.
- Supporting IIG/DOJ/MFCU (Independent Inspector General/Department of Justice/Medicaid Fraud Control Units) investigations by providing OSCAR (Online Survey and Certification and Reporting System) data on deficiencies and QI (quality indicator)/QM (quality measure) outliers and identifying residents whose MDS (minimum data set) profiles suggest potential care problems.
CHSRA also analyzes data for nursing home corporations by identifying facilities and running deficiency and QI/QM data, and by comparing the corporation with a comparison group in the areas of number of health deficiencies; Health Deficiency Index, which is adjusted for state-to-state differences; percent of facilities with severe deficiencies; and Severe Deficiency Index, to name just four. For selected facilities, individual residents can be identified for review using MDS data.
Conclusion: OSCAR and MDS data are useful for exploratory problem identification and ongoing quality monitoring. It should be noted that QI is an indicator only; deficiency citations are subject to an appeals process.
Erica Smith — The National Crime Victimization Survey (NCVS) and the National Incident-Based Reporting System (NIBRS)
The two main sources of national-level data on crimes against older persons are NCVS and NIBRS.
NCVS is the nation's primary source of information on criminal victimization and represents an annual survey of 77,000 households comprising nearly 134,000 persons on the frequency, characterization and consequences of criminal victimization in the United States. The survey enables the Bureau of Justice Statistics (BJS) to estimate the likelihood of victimization by rape, sexual assault, robbery, assault, theft, household burglary and motor vehicle theft for the population as a whole, as well as for certain segments of the population, including the elderly. The NCVS shows that older persons have lower violent crime rates and that although property crime rates are lower for older persons, they are disproportionately affected by property crime.
NIBRS includes detailed information from FBI (Federal Bureau of Investigation) administrative records collected by police departments on many categories of crime, including concurrent offenses, weapons, injury, location, property loss, and characteristics of the victims, offenders and arrestees. NIBRS shows that older persons account for a relatively small percentage of police-reported crime and reinforces the finding that older persons are disproportionately affected by property crime.
Both NCVS and NIBRS have distinct advantages for examining crimes against the elderly. NCVS contains information on both reported and unreported victimizations and offers direct access to victims. However, aggregating data on violence against the elderly over a number of years still does not always yield enough cases for a reliable analysis of the relationship between the victim and the perpetrator. Advantages of NIBRS include the ability to examine offender characteristics with greater validity and reliability, and the fact that the data are collected yearly. Although NIBRS data include enough cases involving older victims to analyze the victim-offender relationship, the numbers may not reflect the unwillingness of family members to report family.
In terms of limitations, NCVS allows for proxy interviews, which may be a problem if the abuse is occurring in the home by another household member. Also, interviews are not conducted on persons in institutional populations. NIBRS only covers crimes reported to the police, and then only from participating jurisdictions, so it is not truly a national survey.
Wendy Verhoek-Offendahl — Evaluation of Data Linkage Methodology to Improve Estimates of Elder Mistreatment
Goal: To apply linkage methodology to improve estimates of elder mistreatment.
Study Design:
- Develop, pilot test and evaluate the feasibility of data linkage methodology statewide that could be employed nationally to improve estimates of incidence and prevalence of elder mistreatment from existing sources.
- Assess and further elucidate risk factors for elder mistreatment that may be used to detect previously unrecognized cases.
- Develop standardized operational definitions that can be applied nationally for each type of elder mistreatment: physical abuse, sexual abuse, psychological abuse, neglect and financial exploitation.
Observations: Linkage of multiple data sources has been used to improve estimates of incidence and prevalence in public health surveillance of disease and more recently to improve estimates of intimate partner violence and child mistreatment. Such linkage methodology may be useful to gain estimates of elder mistreatment, which is an under-studied and underreported problem.
Conclusions: None yet. Progress to date includes:
- Central database development under way.
- Assembly of electronic data for 2006-2007 in progress.
- APS record abstraction for supplemental variables under way.
- Record abstraction at psychiatric hospital in pilot phase.
- Working group assembled to clarify case definition used by each data source.
Discussant, XinQi Dong
Dr. Dong explained that the human rights issue in China informed his interest in this area. After receiving his M.D. he felt that he had a deeper obligation to society and began working on domestic violence and gun control. Then he became committed to elder abuse.
Dr. Dong stated that Dr. Robinson did a wonderful job of describing OSCAR, which is a comprehensive source with great potential for linking. The QI and QM are markers for potential problems, but what needs to be explored is how predictive they are. As for MDS, in coming years MDS will be more clinically relevant and integrate more selected scales in terms of resident and family voices.
Ms. Smith presented on NIBRS — those data give us more information about offender characteristics as related to violent crime. NCVS gives a national representative sample, but we struggle with aggregate and proxy data and whether it is precise. The six-month intervals are good to minimize telescoping.
Dr. Verhoek-Offendahl's very ambitious work uses comprehensive, all-inclusive data sources. It is critical to have standard, operational definitions.
We often struggle with definitions. Right now we are looking at the tip of the iceberg. The question is whether the definition from the tip affects the base and can we capture the entire broad spectrum. Dr. Dong described his own work in Chicago with a population-based study using a biracial population. The goal is to get solid empirical data to look at how well measures predict adverse outcomes or consequences. We may find that mortality, for example, is an indicator of abuse. Data linkage is also an issue because minor differences in methods in linking can cause large variations. We must consider ways to deal with data linkage methods to minimize inconsistencies.
Question/Comment Period
Dr. Verhoek-Offendahl noted the availability of Link-Solv software for probabilistic linkage to find cut points. But because of the small numbers in Rhode Island it was necessary to use client identifiers. As many variables as possible are used for matching. Ms. Smith pointed out that BJS does not use linking, but it does have survey software that is available free that can be adapted to the victimization setting. Dr. Robinson noted that they have linked individuals with hospital records with some success and would welcome another data set with outcome information. Dr. Acierno asked if it is possible to get information on the top 5 percent of problems in nursing homes. Dr. Robinson replied that security considerations limit much of what can or cannot be done. He noted that he could access the Medicare comparison site and make comparisons at the facility level. In terms of linking with CDC or other health databases, there is a potential for linkage, but it is technically much harder than it might appear.
Dr. Catherine Hawes, referring to Dr. Dong's comment that if certain numbers indicated a change in mortality, that might be the gold standard, asked why mortality should be the standard. A person might instead be scared, and we would be under-recognizing the impact of psychological violence. Dr. Dong agreed that there should be multiple domains to measure instrument performance.
In response to a question, Ms. Smith stated that she did not know why it was so costly to add questions to surveys. She noted that they run a supplement in partnership with the National Center for Health Statistics to see how to capture yearly information while bringing the cost down. It might be possible to look at a core group of questions and do a modular run around those.
Dr. Stahl, in summarizing the panel, said that he is intrigued by the possibility of using secondary data through data linkages with major data sources specifically addressing elder abuse. He noted the difficulty in adding on to surveys. Citing as an example the longitudinal Heath and Retirement Survey, Dr. Stahl said that even if he could convince a grantee to write a module to add questions, there would be concern about jeopardizing the survey with questions on a topic as sensitive as elder mistreatment. Overall, he concluded, the methodologies are very impressive.