Secondary Logo

Predictors of 2010–2011 Michigan Medicaid Beneficiary Adverse E-Code Health Care Encounters

Corser, William D., PhD, RN*; Huebner, Marianne, PhD; Zhu, Qi, MS

doi: 10.1097/PTS.0000000000000206
Original Articles
Free
SDC

To inform Medicaid medication management and public health policymaking, the authors analyzed the major predictive factors influencing program-approved therapeutic use or poisoning E-coded encounters leading to emergency department visits and hospital admission for the totality of Michigan Medicaid beneficiaries during a 12-month 2010–2011 period. The analytic cohort was composed of 26,134 approved E-code encounters submitted for 19,865 discrete Michigan Medicaid beneficiaries.

More than 1% of all beneficiaries experienced at least one adverse medication/agent-related E-code encounter during the period. More such encounters and costlier approved encounters were recorded female subjects, African Americans, dually eligible adults, urban elderly, those with fee-for-service Medicaid coverage, and those residing in urban-density counties.

Especially notably for patient safety policymakers, more than 9% of total E-coded encounters for children and adults were primarily attributed by providers to likely preventable poisoning causes such as exposure to household cleaning agents/gases, cosmetic products, illicit drug/alcohol, or secondary tobacco smoke. Encounter costs for the total sample totaled $37 million but ranged considerably up to more than a quarter million dollars.

In view of the future expanding Medicaid-covered beneficiary cohorts, the authors propose several key patient safety/public health policy implications for researchers and policymakers striving to serve lower-income health care consumer groups.

From the *Statewide Campus System, College of Osteopathic Medicine; and

Department of Statistics and Probability, Michigan State University, East Lansing, Michigan; and

SRO Technical Service, University of Michigan, Ann Arbor, Michigan.

Correspondence: William D. Corser, PhD, RN, Michigan State University, East Lansing, MI (e-mail: corser@msu.edu).

The authors disclose no conflict of interest.

Funding: Department of Health and Human Services/ MI Department of Community Health. “2013 Health Information Technology Resource Center.” 01/01/2013-12/31/2013. (Corser, W., and Given, C, Co-PIs).

The inappropriate management, prescription, and/or dispensation of both prescribed and over-the-counter medications has long been cited in larger scale studies as an integral factor contributing to increasing levels of emergency department (ED) visits and hospital admissions for health-care consumers between formal health-care encounters.1–8 This national injury phenomenon has been shown to be even more severe for vulnerable lower income consumers often lacking resources, support, or health literacy to help them avoid such adverse medication-related encounters.9–12

An external cause of injury, or E-code diagnosis code, is a supplemental code assigned by servicing providers to standard condition code(s) during typical care provision to elaborate on the additional reason(s) for required health care treatments.13 E-code events have been systematically defined as a key injury measure by the World Health Organization as conditions documented as due to a drug, medicinal, or biologic substance.13 An E-coded encounter is further coded as a therapeutic use (i.e., codes E930–E949) encounter if the prescribed drug/substance seems to have been taken or used as prescribed.13,14 When a substance has not been taken or used as recommended, the encounter is typically E-coded as a poisoning (i.e., codes E850–869) encounter.13,14 For decades, experts in numerous countries have agreed that improved predictive analyses of these types of E-code encounters are required to inform the contemporary development of better-targeted medication management strategies, regimens, and policies to enable consumers to avoid such adverse health-care encounters.15–23

Unlike the reported multivariate predictive analyses in this paper, which used complete statewide Medicaid Program claims data, most E-code researchers to date have used inconsistent data sources (e.g., health-care system claims data,18–26 individual consumer medical records19,26,27) and/or have focused exclusively on children and adolescents22,28 and/or been conducted in various other countries such as Germany,24 New Zealand,25 and The Netherlands.26,27 Such methodological and setting variations have made it especially difficult for Medicaid and other types of health-care policymakers in the United States to incorporate empiric results into drug safety program components and practice guidelines and policies for our nation’s most vulnerable health-care consumers.17,20,24

For example, 3 earlier descriptive studies have indicated that elders (covered by different governmental/private insurance programs) are more relatively more vulnerable to experiencing E-coded encounters,14,22,26 whereas 2 other studies have indicated that children generally experience more severe E-coded encounters than adults.21,28 Despite these contradictory findings, several common factors have been more often associated with increased per-consumer E-code encounter rates. Factors such as possessing impaired cognition,6,26 increased comorbidity,14,26 rural residence,21 being a younger female,21,28 and taking more prescribed medications.6,26 Notably, the types of prescribed medications most often shown in one earlier study to contribute to disproportionately higher rates of medication-related ED visits have included psychotropic, neurological, and analgesic medications.21

It is also important for readers to consider that several analytic groups using administrative health-care system claims data have found significant variability in prevalent provider E-coding practices across providers or agency settings, complicating the meaningful comparison of results from different settings.15,17,18,21,24 These coding inconsistencies may be in part due to the fact that additional E-code entries over and above diagnostic codes have rarely been used to evaluate prevalent care processes and have not generally affected providers’ service reimbursement.15,18,21 Also, unlike the results reported in this paper, American researchers working with Medicaid beneficiaries have been generally unable to distinguish total submitted claims data from claims that were both submitted and approved.16,17,22,23,28

Still, several research groups have concluded that the use of E-codes to examine and monitor rates of medication-specific injuries may represent the best available population health measure.17,25 To at least partially compensate for likely regional and setting-specific coding variations as shown in earlier works, the authors of this paper also confined their analyses to more adverse approved E-code claims leading to costlier ED visits and hospital admissions submitted for a statewide sample of Medicaid payments and used the most complete statewide government-provided source of E-code claims data submitted for the totality of Medicaid-enrolled billing providers.

Back to Top | Article Outline

Purpose and Research Questions

The purposes of these analyses was to determine the primary factors leading to the number and types of therapeutic use or poisoning approved E-coded encounters and encounter days associated with ED visits and/or hospital admissions for the totality of Michigan Medicaid beneficiaries over a 12-month period.

For these analyses, the following 2 research questions were addressed:

  1. “What types of Michigan Medicaid beneficiaries are more likely to experience one or more E-coded ED visits and hospital admission encounters?”
  2. “What primary factors drive per beneficiary numbers of E-coded ED visits and hospital admissions and number of E-coded encounter days attributable to either therapeutic use or poisoning causes?”

Overall, the authors hypothesized that they would find more ED and hospital E-coded encounters experienced by Medicaid beneficiaries with more of the following: a) prescribed medications; b) chronic health conditions; and c) physical, cognitive, and/or mental impairments, relationships that have been demonstrated in earlier studies using varied data sources.1,4,7,10–12

Back to Top | Article Outline

METHODS

The analytic data set was extracted from 12 complete months (July 1, 2010, through June 30, 2011) of the Michigan Medicaid Data Warehouse (MMDW)29 claims data concerning a total of 26,143 approved E-code claims submitted for an analytic cohort of 19,865 discrete Medicaid beneficiaries. The predictors of overall E-coded encounter claim rates and results obtained from beneficiary subgroups, such as a) Medicaid-Medicare dually eligible, b) children, and c) those experiencing 3 or more E-code encounters during the analytic period, were analyzed.

In essence, the only type of MMDW claims data that the authors found to be significantly incomplete within the analytic sample was encounter cost data from approximately 80% of dual-eligible beneficiaries. More than 98% of E-code claims leading to an ED visit or hospital admission were submitted from acute care hospital or office-based settings, with the remaining claims coming from nursing home or other settings. Each beneficiary claim during the period could be easily classified as either an ED visit or hospital admission-related claim since they were consistently coded by billing providers with specific service date, service location, and rendering provider codes.

The counties of beneficiaries’ home residences were divided into urban and rural-density categories using zip code data and the federal Metropolitan Statistical Area (MSA) definitions from the U.S. Census Bureau.30 An MSA has been defined as a geographical region with a relatively high population density at its core and close economic ties throughout the county by the Federal Office of Management and Budget and used by the U.S. Census Bureau and other federal agencies for statistical purposes.30

A conservative analytic approach was used to identify beneficiaries who were either dually eligible or may have been covered by multiple forms of insurance in addition to Medicaid by using other insurance claims codes. Such beneficiaries with multiple forms of insurance during the period were excluded from the total analytic sample. Standard ICD-9 code,31 National Drug Code (NDC)32 and National Drug Data File HIC333 therapeutic classification code categories were used to collapse less frequent diagnoses and medications into pertinent medication or bodily system categories. Because of the overt overdispersion of E-code encounters (i.e., conditional variance greatly exceeded the conditional mean), a series of negative binomial regression models34 were used to analyze the total E-code claim count per beneficiary as outcomes measure and the number of associated per encounter day(s). The E-coded claims data in the working data set were analyzed using both PASW 18.035 and SAS 9.236 software programs.

An E-coded ED visit was identified using the standard Medicaid Provider Manual claims definition37 of an ED visit for an emergency medical condition as a sudden onset of a physical or mental condition, which causes acute symptoms, including severe pain, where the absence of immediate medical attention could reasonably be expected to place the person’s health in serious jeopardy. An E-coded hospital admission was similarly identified using the standard Medicaid Provider Manual37 claims definition of a (stay in) a hospital for bed occupancy with the expectation that he (beneficiary) will stay at least overnight, even when it later develops that he can be discharged or is transferred to another hospital and does not use the bed overnight.

Back to Top | Article Outline

RESULTS

Medicaid Beneficiary Cohort Characteristics

The baseline characteristics of the total beneficiary sample (n = 19,865) and key sample subgroups are depicted in Table 1. This statewide cohort of those beneficiaries (representing approximately 1% of total beneficiaries at the time) with at least one E-coded encounter experienced a mean number of 1.46 (SD, 1.30; range, 1–23) such encounters during the 12 months. It is important to note that children aged between 1 and 17 years were overrepresented in experiencing more E-code encounters (20% of total encounters vs. 3% in children younger than 1 year), as were female subjects (60% vs. 40%) and beneficiaries living in an urban MSA county (81% of total encounters). African American and dually eligible beneficiaries experienced disproportionately more E-code encounters compared with the total sample cohort (Table 1).

TABLE 1

TABLE 1

Back to Top | Article Outline

Types of E-Code Claims

The types and numbers of approved E-code claims for the 12-month period are depicted in Table 2. Therapeutic use E-code claims were more likely billed for individuals experiencing multiple encounters during the period than poisoning encounters (72% vs. 28%). Beneficiaries with certain ICD-931 diagnoses such as mental/nervous disorders and circulatory/respiratory systems conditions accounted for higher proportions of E-coded encounters. Similar to the results of at least 2 earlier studies,15,26 beneficiaries taking certain types of medications such as pain/anticholinergics/antispasmodics and nervous system/psychotropic agents accounted for a disproportionately higher total number of E-coded encounters.

TABLE 2

TABLE 2

Also notably for patient safety policymakers, more than 9.0% (N = 2364) of total E-code encounters involved likely avoidable exposures to the following: a) secondary tobacco smoke, b) household agents/gases, or c) illicit drugs or alcohol (see lower rows of Table 2). This finding generally reflects the results from earlier studies.15,19–21,28 The types of reported household agents and substances included cosmetics, suntan lotions, household cleaners, caffeine-containing products, paint removers, and varied approved/illicit types of home heating fuels. As could perhaps be expected, children recovered sooner from their respective E-code encounters if hospitalized than either dually eligible or adult beneficiaries experiencing 3 or more E-code encounters. E-coded ED visits and hospital admissions were of course quite collinear to each other, with 41.6% of E-coded hospitalizations preceded by an ED visit.

Back to Top | Article Outline

E-Code Claim Approved Coverage Amounts

The process of obtaining and analyzing cost data for approved claims was complicated by the fact that approximately 9785 (37.4% of total E-code claims) had no approved state and/or federal amount data specified in the MMDW29 records. On subsequent examination, it appeared that the great majority (80% or 6353) of missing E-code encounters claims came from dually eligible beneficiaries for which Medicare coverage data were not currently available and/or claims in which the perceived likelihood of obtaining additional Medicaid reimbursement may have been rather low. Using available cost data, however, federal/and/or state approved claim amounts for the total sample and several key sample subgroups were described in Table 3. In a complete case analysis, approved claims totaled more than $37 million, averaging $3702 per individual encounter (SD $10,207) and running up to more than $253,000.

TABLE 3

TABLE 3

In summary, therapeutic use E-code encounter were consistently more costly than poisoning claims. Fee-for-service Medicaid beneficiary E-code encounter costs were also somewhat higher than that for managed care-covered beneficiaries. As might be expected, hospital only E-code encounters were also more costly than either the ED only or both ED and hospital encounters because direct admission hospitalized beneficiaries likely presented themselves in more emergent states that required more immediate and complex acute care attention.

Back to Top | Article Outline

Predictive Influences of Greater Per-Beneficiary Claim Counts and Encounter Days

As shown in Table 4, beneficiaries who were older, African American, urban-dwelling, and/or living in 1 of 5 Detroit area counties significantly experienced both more E-code encounters and total encounter-related hospital days. It was also found that therapeutic use E-coded hospital encounters generally consumed more encounter days than poisoning E-code encounters. In addition, disproportionately higher per beneficiary E-code encounter counts and encounter days were coded to certain medication categories (e.g., psychotropic, nervous system, oncology meds) and diagnostic categories (respiratory, circulatory, digestive systems) (not specifically reported in Table 4).

TABLE 4

TABLE 4

Back to Top | Article Outline

DISCUSSION

In summary, our results tangibly demonstrate that a) older, b) African Americans, c) those with ongoing mental health needs, and/or d) adult beneficiaries living in Detroit and urban counties were more susceptible to experiencing multiple E-code encounters, also accruing greater encounter-associated health-care costs. At the same time, younger beneficiaries tended to recover sooner when a therapeutic medication or poisoning-related hospitalization was deemed necessary. E-code cohort rates and patterns during the 12-month analytic period were significantly influenced by certain sociodemographic, clinical, and care delivery factors in somewhat similar patterns to those found in earlier works.14,21,26,38

Although the costs of E-coded encounters measured from these Medicaid program claims files was sizable, it should be considered why our encounter rates seemed to be somewhat lower (using various admission/ED measures and settings) than those reported from earlier works.24,38 For example, one 2010 American group38 estimated that more than 4.7% of all national ED visits were due to medication-related problems, with 2 other European research groups measuring rates of more than 1.41%24 and 5.6%, respectively,26 of total hospitalizations were specifically medication-related admissions, rates far higher than these we found in this statewide Medicaid-covered cohort.

As indicated earlier, this general disparity in E-coded encounter rates could certainly be attributable to nationally demonstrated variations in provider coding practices and/or lower relative Medicaid reimbursement rates relative to private insurers. Regardless, the actual fiscal and care delivery burden imposed on this and other states’ Medicaid programs and health-care systems from more severe E-coded health care encounters may, in fact, be far more severe than even these statewide Michigan program claims data suggest.

Back to Top | Article Outline

Limitations

Several limitations to these analyses should be acknowledged. First, these E-code claims were derived from Medicaid beneficiaries in one especially economically deprived Midwest state. Second, both E-code and administrative Medicaid program claims data will continue to possess some inherent ongoing limitations related to missing data for certain beneficiary groups, provider coding variability, lags in claims data availability, and so on.17,21–23,25,36 Because we were using administrative claims files, it was of course realistically difficult to precisely discern whether medication-related E-code encounters were primarily due to drug-to-drug or drug-to-condition interactions, medication prescribing errors, or beneficiary nonadherence/mismanagement.

Still, these analyses comprise one of the first statewide projects concerning the multiple predictive factors leading to E-coded ED visits and hospitalizations for a largely complete cohort of Medicaid beneficiaries using the best available measure of medication-related injury. Our results also vividly suggest that considerable ED visit and hospitalization costs are being generated from sometimes preventable medication and household agent-related exposures for our nation’s lower-income citizens. Because more than 9% of total E-code claims came from more likely avoidable causes such as acute alcohol/illicit drug intoxication, exposure to household agents and gases, and secondary tobacco smoke, these findings suggest that additional targeted drug safety and public health policies/educational materials are required to better serve more vulnerable lower-income consumer subgroups.

Our result that beneficiaries living in either the Detroit area or urban-density counties experienced both more E-coded encounters and hospital encounter days suggests that some rural consumers may have been more prone to ignoring or delaying treatment for symptoms because of impaired service access. However, this possibility is not supported by the fact that since urban-dwelling beneficiaries were still more often actually admitted into hospitals than rural beneficiaries in this sample, suggesting that any possible urban–rural differences in time from symptom/signs to access were still probably relatively small compared with encounter-related clinical factors.

Back to Top | Article Outline

Future Work in the Area

The authors would like to propose several patient safety, medication management, and public health policy implications for future E-code health outcomes work. More purposefully constructed study samples could certainly help policymakers more confidently focus medication management interventions for certain types of Medicaid beneficiaries taking higher risk prescribed medications, possessing riskier conditions, and so on requiring increased monitoring and/or consumer education. Similar to 2 earlier studies, our results show that the majority of medication-related E-code encounters were experienced by more vulnerable medically complex consumers psychotropic, neurologic, or analgesic medications.15,26 Unfortunately, Medicaid beneficiaries and other lower-income consumers who are older and/or more cognitively impaired require more consistent monitoring than can be realistically provided in many primary care agency settings. More controlled analyses teasing out the influences of E-coded ED visits versus hospital admissions may also be warranted because some research has suggested that the more prevalent factors contributing to these 2 adverse health-care encounters may be fundamentally different.19,22,38

Despite the incumbent limitations associated with using any types of state government claims data,17,23,25 the inclusion of comparison group claims data from Medicaid beneficiaries who have not experienced an E-code encounter would likely enable future patient safety researchers to conduct more thorough subgroup examinations than these reported in this paper. Matched pair-type analyses could use multivariate logistical and longitudinal modeling procedures to more rigorously examine racial/other health disparities among the most vulnerable consumer subgroups. Analysts could also focus on key primary care delivery system/care coordination factors contributing to greater subsequent ED visit and hospital admission rates to identify the elements that might be most amenable to improving drug and household agent safety levels.

More controlled multilevel and geospatial projects examining E-code variations might also enable health-care planners and policymakers to gauge how beneficiary subgroup outcomes may be variably influenced by regional, urban–rural, provider, and beneficiary-level factors. As has been suggested by these analyses and previous works, E-code encounter rates may also serve as an especially sensitive measure of health-care coordination for our nation’s more complex (i.e., old and poor and/or disabled and poor) dually eligible citizens covered by both Medicaid and Medicare.

Back to Top | Article Outline

CONCLUSIONS

Office and community-based health injury prevention policies designed to serve higher-risk individuals from experiencing avoidable medication and agent-related ED visits and hospitalizations will be less feasibly created or updated under current primary care resource and practice constraints. Ideally, further E-code analyses will provide a key opportunity to improve the evidence-based development and targeting of contemporary initiatives to improve the medication and agent-related safety outcomes and costs of our nation’s most vulnerable adults and children.

Back to Top | Article Outline

ACKNOWLEDGMENT

The authors thank Charles Mundt at Michigan State University Institute for Health Policy.

Back to Top | Article Outline

REFERENCES

1. Bao Y, Shao H, Bishop TF, et al, and others. Inappropriate medication in a national sample of US elderly patients receiving home health care. J Gen Intern Med. 2012;27:304–310.
2. Abramson EL, Bates DW, Jenter C, et al. Ambulatory prescribing errors among community-based providers in two states. J Am Med Inform Assoc. 2012;19:644–648.
3. National Research Council. Preventing Medication Errors: Quality Chasm Series. Washington, DC: The National Academies Press; 2007:480.
4. Walsh KE, Roblin DW, Weingart SN, et al. Medication errors: a multisite study of children with cancer. Pediatrics. 2013;131:e1405–e1414.
5. Budnitz DS, Pollack DA, Weidenbach KN, et al. National surveillance of emergency department visits for outpatient adverse drug events. JAMA. 2006a;296:1858–1866.
6. Lau DT, Kasper JD, Potter DEB, et al. Hospitalization and death associated with potentially inapporiate medication prescriptions among elderly nursing home residents. Arch Intern Med. 2005;165:68–74.
7. Robbins CM, Stillwell T, Johnson D, et al. Integrating patient safety and clinical pharmacy services into the care of a high-risk, ambulatory population: a collaborative approach. J Patient Saf. 2013;9:110–117.
8. Macdonald MT, Lang A, Storch J, et al. Examining markers of safety in homecare using the international classification for patient safety. BMC Health Serv Res. 2013;13:191.
9. van Sluisveld N, Zegers M, Natsch S, et al. Medication reconciliation at hospital admission and discharge: insufficient knowledge, unclear task reallocation and lack of collaboration as major barriers to medication safety. BMC Health Serv Res. 2012;12:170.
10. Olson MD, Tong GL, Steiner BD, et al. Medication documentation in a primary care network serving North Carolina Medicaid patients: results of a cross-sectional chart review. BMC Fam Prac. 2012;13:83.
11. Miranda J, Ong MK, Jones L, et al. Community-partnered evaluation of depression services for clients of community-based agencies in under-resourced communities in Los Angeles. J Gen Intern Med. 2013;28:1279–1287.
12. Ratanawongsa N, Handley MA, Quan J, et al. Quasi-experimental trial of diabetes Self-Management Automated and Real-Time Telephonic Support (SMARTSteps) in a Medicaid managed care plan: study protocol. BMC Health Serv Res. 2012;12:22.
13. Ingenix Incorporated. Assigning E Codes for External Causes. 2012. Available at: http://health-information.advanceweb.com/Article/Assigning-E-Codes-for-External-Causes.aspx. Accessed April 28, 2015.
14. Substance Abuse and Mental Health Services Administration, Center for Behavioral Health Statistics and Quality. The DAWN Report: Emergency department visits involving adverse reactions to medications among older adults. 2011. Rockville, MD. Available at: http://www.samhsa.gov/data/2k10/TDR013AdverseReactionsOlderAdults/AdverseReactionsOlderAdults.htm. Accessed April 28, 2015.
15. Annest JL, Fingerhut LA, Gallagher SS, et al. Strategies to improve external cause-of-injury coding in state-based hospital discharge and emergency department data systems: Recommendations of the CDC workgroup for improvement of external cause-of-injury coding. Morbid Mortal Weekly Recom Reports. 200 57;1–15. Available at: http://www.cdc.gov/mmwr/preview/mmwrhtml/rr5701a1.htm. Accessed April 28, 2015.
16. Guyer B, Berenholz G, Gallagher SS. Injury surveillance using hospital discharge abstracts by external cause of injury (E code). J Trauma. 1990;30:470–473.
17. LeMier M, Cummings P, West TA. Accuracy of external cause of injury codes reported in Washington State hospital discharge records. Injury Prevent. 2001;7:334–338.
18. Marcum ZA, Pugh MJV, Amaun ME, et al. Prevalence of potentially preventable unplanned hospitalizations caused by therapeutic failures and adverse drug withdrawal events among older veterans. J Gerontol Med Sci. 2012;67:867–874.
19. Wadman MC, Muelleman RL, Coto JA, et al. The pyramid of injury: Using E codes to accurately describe the burden of injury. Ann Emerg Med. 2003;2:468–478.
20. Schwartz RJ, Nightingale BS, Boisoneau D, et al. Accuracy of E-codes assigned to emergency department records. Acad Emerg Med. 1995;2:615–620.
21. Xiang Y, Zhao W, Xiang H, et al. ED visits for drug-related poisoning in the United States, 2007. Am J Emerg Med. 2012;30:293–301.
22. Budnitz DS, Shehab N, Kegler SR, Richards CL. Medication use leading to emergency department visits for adverse drug events in older adults. Ann Intern Med. 2007;147:755–765.
23. Crystal S, Akincigil A, Bilder S, et al. Studying prescription drug use and outcomes with Medicaid claims data: Strengths, limitations, and strategies. Med Care. 2007;45(Suppl 2):S58–S65.
24. Stausber J, Hasford J. Drug-related admissions and hospital-acquired adverse drug events in Germany: A longitudinal analysis from 2003 to 2007 of ICD-10 coded routine date. BMC-Health Serv Res. 2011;11:134.
25. Langley J, Stephenson S, Thorpe C, et al. Accuracy of injury coding under ICD-9 for New Zealand public health records. Injury Prevent. 2006;12:58–61.
26. Leendertse AJ, Egberts ACG, Stoker LJ, et al. Frequency of and risk factors for preventable medication-related hospital admissions in the Netherlands. Arch Intern Med. 2008;168:1890–1896.
27. Leendertse AJ, van den Bent PMLA, Poolam JB, et al. Preventable hospital admissions related to medication (HARM): Cost analysis of the HARM study. Value Health. 2011;14:34–40.
28. Cohen AL, Budnitz DS, Weidenbach KN, et al. National surveillance of emergency department visits for outpatient adverse drug events in children and adolescents. J Pediatr. 2008;152:416–421.
29. Michigan Department of Community Health. MDCH Medicaid Data Warehouse. Description. 2012. Available at: http://ihcs.msu.edu/research/medicaid_match_data_warehouse.php. Accessed April 28, 2015.
30. Nussle J. Update of Statistical Area Definitions and Guidance on Their Uses. Office of Management and Budget; 2008:1–2 (federal OMB report). Available at: www.whitehouse.gov/sites/default/files/omb/assets/.../b10-02.pdf. Accessed April 28, 2015.
31. World Health Organization. International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). Geneva, Switzerland: World Health Organization; 1979.
32. U.S. Food and Drug Administration. National Drug Code; 2012. Available at: http://www.fda.gov/Drugs/InformationOnDrugs/ucm142438.htm. Accessed April 28, 2015.
33. FirstDataBank, Inc. National Drug Data File-HIC3. 2010. Available at: http://www.Firstdatabank.com. Accessed April 28, 2015.
34. Jones AM. Health Econometrics. In: Culyer AJ, Newhouse JP, eds. Handbook of Health Economics. Amsterdam, Netherlands: Elsevier North-Holland; 2000.
35. SPSS Inc. PASW STATISTICS 17.0 Command Syntax Reference. Chicago: SPSS Inc.; 2011.
36. SAS Institute Inc. SAS® 9.2 Analytic Software. Cary, NC: SAS Institute Inc.; 2009.
37. Michigan Department of Community Health. Medicaid Provider Manual. 2011. Available at: http://www.michigan.gov/mdch/0,1607,7-132--87572--,00.html. Accessed April 28, 2015.
38. Hohl CM, Zed PJ, Brubacher JR, et al. Do emergency physicians attribute drug-related emergency department visits to medication-related problems? Ann Emerg Med. 2010;55:493–502. e4.
Keywords:

E-code encounters; adverse medication encounters; Medicaid beneficiaries

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved