Data Sources/Available Data Elements/Diagnosis-Procedure Codes Studied
Data Sets/Sources and Available Data Elements
Data sources for investigating dental care provided in EDs are numerous. The data sources to a large extent are chosen depending on the specific research question being asked. Some researchers are assessing the issue at a national level. Others, either due to a local focus or possibly in some cases just using the data that are most convenient to them, elect to use administrative data from a single hospital or group of hospitals, or in some cases from patient interviews. Stakeholders at a county level will use data for a county. State oral health programs (SOHPs) will usually obtain and use state level data to elucidate a problem or implement interventions or have others influence policymakers to address the problem at the state level. The source and characteristics of the data used directly impact the research questions that can be addressed and the inference of results to specific populations. Some commonly used datasets are summarized in the following sections.
The National Emergency Department Sample (NEDS)
The Nationwide Emergency Department Sample (NEDS) includes data sampled from a family of state inpatient (SID) and state emergency department (SEDD) databases including software developed by the Healthcare Cost and Utilization Project (HCUP).(95) NEDS is a stratified sample of about 20% of U.S. hospital EDs and contains data from 950 hospitals in 30 states. NEDS data can be used to generate national and regional estimates of ED use. Further information on NEDS can be found in Appendix 1. Investigators using this dataset include Allareddy et al.(21, 22), Chi et al.(89), Laurence et al.(38, 39), Nakao et al.(40), Nalliah et al.(23), Walker et al.(37), Wall and Vujicic(24).
Medical Expenditure Panel Survey (MEPS)
In 1996 the Medical Expenditure Panel Survey began collecting data on use, frequency of use, and costs of health services used by American families and individuals, and how these services are paid for.(96) The surveys include data from families, individuals, doctors, hospitals, pharmacies, and employers across the United States. MEPS surveys include a household component (including additional information from health care providers) and an insurance component. Further information on MEPS can be found in Appendix 2. Investigators using this dataset include Dismuke et al.(92), Fields et al.(26), Chevarly et al.(91), Newacheck et al.(27), Romaire et al.(41, 42)
National Hospital Ambulatory Medical Care Survey (NHAMCS)
The National Hospital Ambulatory Medical Care Survey (NHAMCS) includes data on utilization and care provided in hospital emergency and outpatient departments and in ambulatory surgery centers. The hospital component of the survey includes data from a national probability sample of visits to emergency and outpatient departments and to ambulatory surgery facilities in non-institutional hospitals in all 50 States and the District of Columbia. The freestanding ambulatory surgery component includes data from a national probability sample of visits to ambulatory surgery centers in all 50 States and the District of Columbia. Data include demographic characteristics of patients, expected source(s) of payment, patients' complaints, diagnoses, diagnostic/screening services, procedures, medication therapy, disposition, types of providers seen, causes of injury (emergency department and ambulatory surgery center only), and certain characteristics of the facility, such as geographic region and metropolitan status. See Appendix 3 for additional information on NHAMCS. Investigators using this dataset include Lee et al.(28), Lewis et al.(30, 31), Okunseri et al.(32-36), Wall(29),
Other National Databases
Other national datasets containing relevant health and dental care utilization data, including use of EDs for care, have also been used to explore various research questions. Flores and Tomany-Korman analyzed 2003-2004 National Survey of Children’s Health to examine racial/ethnic disparities in health and dental care.(43)
State Level Data/Sources
Some investigators, especially those connected to SOHPs, are more interested in state level data. These data may vary by availability and the content of datasets. To the extent that data are available and consistent across states, and similar methodology is employed, comparisons among states can be made. A common source of state level data is the state’s emergency room discharge database (SEDD).
State Emergency Department Databases (SEDD)
The State Emergency Department Databases (SEDD) are part of the family of databases including software developed by the Healthcare Cost and Utilization Project (HCUP) The SEDD includes data on emergency visits at hospital emergency departments that do not result in hospitalization. Data on patients admitted to a hospital after an ED visit are included in the State Inpatient Databases (SID). The SEDD files include all ED patients regardless of payer, and include clinical and non-clinical data. Thirty-two states currently participate in SEDD. Further information on SEDD can be found in Appendix 4.
Other State Emergency Department Data Sources
States that don’t participate in SEDD may still maintain and make available their own ED databases. Different researchers have used SEDD data or state ED databases to study ED dental care at the state level. Singhal et al. used California 2006-2011 SEDD data and Medicaid enrollment and reimbursement data from the California Department of Health Care Services in a study of the effects of changes in state Medicaid benefits.(55) Sun et al. used a combination of 2010 claims data from 45 of Oregon’s 60 hospitals (including all payer groups) and the Oregon Payer All Claims file (which includes procedure, prescription, repeat ED visits, and costs data not available in hospital supplied data), and also interviewed purposive samples of ED dental visitors and community stakeholders in six counties.(44) Anderson et al. studied ED use for non-traumatic dental care in New Hampshire using the New Hampshire Hospital Discharge dataset, focusing on ED visits not resulting in hospital admission.(87) Likewise, DeLia et al., in their study of ED dental care, used the New Jersey Discharge Data Collection System, which contains billing records for inpatient and ED care for all of the state’s hospitals.(93) Hom et al. used the North Carolina Emergency Room Discharge Database to investigate the relation of dental related ED visits to insurance mix of patient populations across North Carolina EDs.(45) Shortridge and Moore used 2005 SEDD data from three states to assess similarities and differences in ED dental care among the states.(57)
Available state level datasets other than hospital ED discharge data also are used. Lee et al. used a combination of North Carolina birth records, Medicaid data, WIC files, and the area resource file in comparing North Carolina Medicaid children participating in WIC to those not participating in WIC for dental care access, dental procedures, and ED visits specifically for caries.(53) Wallace et al. analyzed a combination of data from before and after an elimination of Medicaid dental benefits, including Oregon Health Plan eligibility data, fee-for-service claims data, and encounter data from managed care organizations, as well as patient survey data in their research on changes in dental care utilization and accessing medical care settings for dental care among Oregon Medicaid patients.(56) Martin et al. used South Carolina Medicaid data in their study of dental care utilization for Medicaid children younger than four years.(46) Okunseri et al. used Wisconsin state Medicaid data combined with county Wisconsin DHPSA data and US Department of Agriculture Urban Influence Codes (a measure of county rurality) in their study of Medicaid dental treatment provided in EDs and physicians’ offices.(50) Pajewski and Okunseri did a similar analysis of Wisconsin Medicaid data focusing on adult Medicaid patients.(51) Medicaid data availability presents an opportunity for investigating ED access for dental care in the primary care sector as medical and dental data for Medicaid subjects can be linked. For example, follow-up dental care subsequent to ED visits for dental problems can be explored. When using state Medicaid data, issues related to Medicaid data analysis must be addressed, including changes in eligibility affecting numerator and denominator determination in calculation of rates, and the use of procedure codes instead of diagnostic codes in dental insurance data.
DeVoe et al. collected their own Oregon state level dataset of a demographic subpopulation, sending mail surveys to a sample of randomly selected families in the Oregon food stamp program.(54) Similarly, Kempe et al. surveyed a statewide random sample of Colorado Child Health Plan Plus (CHP+) enrollees to assess health access changes, including changes accessing dental care and EDs for health care, before to one year after enrollment.(52)
Local Level Data/Sources
Many studies have used administrative data or collected patient data in single hospitals or EDs.(7, 10-12, 67-76, 80, 81) Other local studies have used administrative data from a hospital system of two or more hospitals, or a group of hospital systems. Davis et al. analyzed hospital administrative data from five major hospital systems in the Minneapolis-St. Paul metropolitan area.(77) Fox et al. conducted chart reviews in a two-hospital system to study the effects of an instituted prescribing guideline on opioid prescribing for patients with oral pain complaints.(78) Other studies utilize smaller geographic areas within states. Roghmann and Goldman used emergency logs of Rochester, New York hospitals along with Medicaid data and ED patient interviews to explore the effects of a new neighborhood health center on ED dental emergency visits.(84) Feinglass et al. studied enrollees in Access DuPage, a healthcare program for low-income, uninsured residents of DuPage County, IL.(85) Hong et al. used the Kansas City, Missouri data subset from state 2001-2006 hospital discharge data in one study,(82) and combined 2001-2010 Kansas City hospital discharge data with the 2007-2011 American Community Survey 5-Year Estimates data to explore community level predictors in another study.(83) Lave et al. used a random sample survey of new enrollees in Western Pennsylvania health insurance programs for low-income uninsured residents studying changes in health care access during the first year following enrollment.(86)