Alcoholism treatment episodes validly defined using mental health care utilization records

Melanie M. Wall, Kurt D. Stromberg, Sandra Pothoff, Robert L. Kane

Research output: Contribution to journalArticlepeer-review

7 Scopus citations


Objective We propose a method for defining and empirically validating episodes of alcoholism treatment from health care utilization records. Study design and setting The study includes utilization records from 86,207 patients enrolled in a large managed behavioral care company who had at least one alcoholism encounter between 1991 and 1998. Treatment episodes are defined as a minimum number of alcoholism treatment encounters with the behavioral care company prior to a "clear zone" of no encounters. Statistical procedures to select a subset of episode definitions from a number of candidate definitions and methods for assessing the convergent and criterion validity of the definitions are presented. Results The percentage of patients having at least one episode of alcoholism treatment varies from 43% to 77%, with the results being more sensitive to the minimum number of encounters required than the length of the clear zone. Criterion validity does not reveal any clear "winning" definitions; positive predictive ability increases most rapidly when going from 2 to 3 encounters required. Conclusion The most robust definitions of an alcoholism treatment episode entail 3 to 4 encounters with a clear zone of 3 to 4 months.

Original languageEnglish (US)
Pages (from-to)373-380
Number of pages8
JournalJournal of Clinical Epidemiology
Issue number4
StatePublished - Apr 2004

Bibliographical note

Funding Information:
This work was supported by a grant from the National Institute of Alcohol Abuse and Addiction (no. 1 R01 AA11781).


  • Alcoholism
  • Episode of care
  • Statistical applications
  • Validation techniques


Dive into the research topics of 'Alcoholism treatment episodes validly defined using mental health care utilization records'. Together they form a unique fingerprint.

Cite this