Research
Research Studies in Health
Alternative and Complementary Medicine Data Source
Submitted April 22, 2002
Introduction and Data Source
Little research using claims–based utilization data is available to assess the cost and cost–effectiveness of the complementary and alternative medical therapies (CAM). This research project uses data submitted by one insurer (the "source") to examine the utility of standard eligibility and claims extracts for CAM analyses, and to explore issues likely to arise should these analyses be expanded.
In late August of 2001, Jim Conner submitted to the Society of Actuaries, and to the researchers, draft specifications he had discussed with the systems manager at the data source. After review and comment on the choice of data elements, field definitions, format, media and level of detail required, Mr. Conner submitted a revised set of specifications to the systems manager in early September. The systems manager then "created a custom procedure to extract enrollment, claims data and miscellaneous code information in a fixed record format." These data were to cover a period including calendar years 1998, 1999 and 2000, the period for analysis in this study. The data were provided on CD–ROM, accompanied by a cover letter and list of "standard remark codes" required for translating a comment field on the claims records.
The data were submitted as a Microsoft Access database file containing three tables which summarized claims, "encounters" (i.e., membership or exposure information) and provider types (i.e., descriptions of twenty–one codes designating the type of provider associated with a claim). For claims and exposure data, an attempt was made to adhere to the Society's specification for the Medical Large Claims Experience Study, since the data source is also a participant in that study.
The key deliverable product of this research is a set of spreadsheets summarizing claims and exposure information for CAM use, compared with conventional medical use. The Society of Actuaries provided spreadsheet templates designating the desired information to be derived and presented by the researchers. Several telephone conferences and exchanges of memoranda made some modifications and expansions to the original templates.
The original templates and scope of work envisioned four data extracts, each to be repeated for four different subgroups of members. The analysis of CAM is limited to chiropractic and acupuncture services because these provider types are readily determined from the available data.
The reader is referred to the METHODS section of this report for a description of the information contained in Extract 1 through Extract 4. Each set of extracts is prepared for the following four subgroups of members: all members combined, adult males, adult females, and children. Two additional analyses were requested during the course of the study: Extract 3, limited to all members who were continuously enrolled throughout the study period; and Extract 3, analyzed separately for each diagnosis category considered in the study. After reviewing the draft report, Society staff requested inclusion in the final report of one additional table, which was originally presented during the course of work to guide the analysis: Office Visit Comparison–CPT Code vs. Service Date Count.
Methods
Membership Data
The membership ("encounter") table contains 27,142 records, many of which are exact duplicates (i.e., each value of each field is identical to those of at least one other record). The number of distinct records is 21,388 (i.e., at least one field contains values which are not identical). Duplication aside, each member appears to be represented by at least two records. Investigation reveals that a member is represented by multiple records mainly because multiple values of "coverage termination date" are present. Typically, a member is represented by two records identical in all fields except that one record has a blank value for coverage termination date, while the other record has a value of "12/31/9999". Each of these values is taken to indicate no specific termination date and that coverage remains in effect.
The combination of two fields, insured ID and relationship to insured, appears adequate to identify an individual. The membership table contains 13,120 distinct combinations of these two fields, which is taken to be the number of members initially represented. Members retained for the study number 11,581. Reasons for eliminating members from the study are summarized in the following table.
Distinct Combinations of Insured ID and Member Relationship: |
13,120 |
| Coverageis effective after conclusion of study |
–1,432 |
| Coverage termination date precedes coverage effective date |
–45 |
| Inconsistent member–months of coverage |
–32 |
| Multiple birth dates |
–25 |
| Undefined value for member relationship |
–5 |
Members retained in study |
11,581 |
A member's months of coverage for each year are calculated from coverage effective date and coverage termination date. If records for a member contained multiple termination or effective dates which result in differing values for months of coverage, that member is dropped from the study.
Claims Data
The claims table contains 345,197 records. These records represent 9,036 distinct combinations of Insured ID and Member Relationship, all but 49 of which match a combination present in the membership file. Of these records, 13,611 have an "incurred from" date falling outside the three year study period. The included claim records are distributed by year, as follows:
- 1998 – 101,147
- 1999 – 119,387
- 2000 – 111,052
Each record contains a field for provider type, which is used to determine whether services were provided by a chiropractor (21,058 records) or an acupuncturist (984 records).
A claim record is retained for the study if its "incurred from" date falls within the study period, and if it matches a member for whom coverage was in effect during the "incurred from" year. After eliminating claims not satisfying these two criteria, claim records retained for the study are distributed by year, as follows:
- 1998 – 101,147
- 1999 – 119,387
- 2000 – 111,052
The record count by provider type is 17,364 chiropractor and 883 acupuncturist.
The procedure code is used to determine whether a record represents an office visit. A record with a procedure code from 99201 to 99215 or from 99450 to 99499, inclusive, is counted as an office visit. The researchers and staff of the Society of Actuaries, after discussion with a representative of the source, are aware that this approach significantly under–counts office visits.
The Extracts
The data extracts are numbered 1 through 4. Staff of the Society of Actuaries and their advisors provided spreadsheet templates defining the desired analysis. Extracts 1 through 4 are described below for the analysis of all members. Three other groups (adult males, adult females, and children) are separately analyzed, and their Extracts 1 through 4 are separately presented. The analyses are parallel, and the extracts for these additional groups are not separately described below except for one distinction (i. e., gender analysis not possible for children). Two additional versions of Extract 3 (one limited to members continuously enrolled throughout the study period and one analyzed by diagnosis category) are described.
Throughout these extracts the number of members or users is computed with reference to member–months (e.g., number of members covered for a given year is the total member–months of coverage for that year divided by twelve). Similarly, computations of gender percentages and average ages are weighted by member–months of coverage.
All costs are totals of paid charges.
Each extract presents a separate analysis for each study year.
Extract 1
Extract 1 presents, for each year, an analysis of total CAM usage. Separately presented is CAM usage by type of therapy (chiropractor or acupuncturist). The number of members using CAM (or a CAM therapy), along with the number of office visits, are included, as are the range and average number of visits per user. The number of CAM users is compared with total membership, and the proportion using CAM is noted. CAM costs are separated into visit and non–visit costs, and the average cost per visit is computed. Total medical costs are presented, and the portion of those costs expended on CAM (or a CAM therapy) is calculated. The reader is again advised to interpret the number of office visits cautiously, as they appear to be significantly under–counted.
Extract 2
Extract 2 presents analysis by combination of CAM therapy and diagnosis category. The diagnosis categories considered for analysis are those developed for the Society's prior "Group Medical Insurance Large Claims Database Collection and Analysis" study. Selected for analysis in this extract are the five most frequently occurring diagnostic categories for chiropractor and the four most frequently occurring categories for acupuncturist.
Each CAM therapy was analyzed for the following three diagnosis categories:
- Skeleton & Muscle System Conditions
- Injury & Poisoning
- Symptoms & Ill–Defined Conditions
-
- In addition, chiropractor is analyzed for:
- Congenital & Perinatal Conditions
- Nervous System Conditions
-
- Acupuncturist is also analyzed for:
- Mental Disorders, Drug, Alcohol Conditions
Each of the nine analyses in Extract 1 is for a subgroup of members having at least one claim falling within the diagnosis category under consideration and utilizing the CAM therapy under analysis. The CAM user information is similar to that presented in Extract 1: number of users, number of visits, average number of visits per user, visit costs, non–visit costs and average costs per visit. These visits and costs are sums for claims within both the diagnosis category and CAM therapy under analysis. Non–CAM visits and costs are derived from claims, for the subgroup of members described above, which fall within the diagnosis category being considered but which do not utilize a CAM therapy (those claims which may use the alternative CAM therapy are not included). The Non–CAM presentation is similar to the CAM information and includes visit costs, non–visit costs, number of visits, costs per visit, and visits per user.
Extract 3
Extract 3 presents an analysis for both CAM therapies and for all diagnoses, combined. Total medical costs, CAM costs and non–CAM costs are subtotaled for members who use only CAM services, for those who use both CAM and conventional services and for those who use no CAM services. The number of member–months of coverage is presented for these three groups, and adds a category for members with no CAM use but with some other medical use. Corresponding total medical costs per member per month are computed. This extract further presents average ages and proportion of females (weighted by member–months of coverage) for each of the four subgroups. Average ages are included for all members and for members who are not children.
Gender analysis could not be performed for the subgroup of children. Although the claims table contains a field indicating gender of the claimant, the membership file contains a field only for gender of the primary insured (employee). The gender of an employee's spouse was imputed to be opposite the gender of the employee. The gender of child members without a record in the claim file could not be determined.
Extract 4
Extract 4 compares CAM costs with non–CAM costs. For each category, totals are presented for charges, allowed charges, ineligible charges and paid charges. In addition, allowed and ineligible charges are presented as a percentage of total charges, and paid charges are expressed as a percentage of allowed charges.
Extract 3–All Continuously Enrolled Members
This separate version of Extract 3 is identical to that described above except for the group of members included in the analysis. With the exception of this extract, all extracts include all relevant members with some coverage during the study period. The present extract is limited to those members with a full twelve months of coverage during each study year.
Extract 3–All Members–By Diagnosis Category
This extract is a version of Extract 3, with separate analysis for each of the six diagnosis categories included in Extract 2 and listed above. The prior description of Extract 3 applies here. For each analysis, a CAM user is a member with at least one claim utilizing a CAM therapy for the diagnosis category under consideration. A non–CAM user utilizes only non–CAM services for the diagnosis category being analyzed (even if he or she uses CAM therapies for other diagnoses).
Office Visit Comparison–CPT Code vs. Service Date Count
As noted in the discussion above, determination of office visits by CPT code significantly undercounts visits to chiropractors and acupuncturists. During the course of this work, the researchers compared the CPT code count with a count of service dates (when the "incurred from" date is the same as the "incurred to" date) in an attempt to estimate the extent of the undercount. The table "Office Visit Comparison–CPT code vs. Service Date Count" presents the results of this analysis. The great discrepancy between the CPT code count and the service date count, ranging by year from a factor of 8.3 to a factor of 11.4, for combined CAM procedures seems to confirm a significant undercount by CPT code. In contrast, the ratio of the counts for members using some medical services but no CAM services does not exceed 2.3 in any year.
Recommendations
This work presents a preliminary view of complementary and alternative practices in the medical care system. It is intended as a first look, which might help guide additional analysis of the importance of CAM.
Further work in this area could include data from additional insurers. The definition of CAM could be broadened. The current project was limited to consideration of services provided by chiropractors and acupuncturists because these services were readily identifiable within the limited scope of the project. Future work could involve consideration of appropriate services to categorize as CAM, and of methods required to identify those services from available data.
This research illustrates that office visits can be difficult to determine by examination of procedure codes, perhaps because chiropractors and acupuncturists use such codes differently from more conventional medical practitioners. Future work could attempt to develop methods to determine more reliably whether a claim represents an office visit.