Jump to:

Lorillard

Medicaid Expenditures Attributable to Smoking: Database & Software Documentation for Oklahoma

Date: 26 Mar 1998
Length: 95 pages
98720794-98720888
Jump To Images
spider_lor 98720794_0888

Fields

Author
Harrison, G.W.
Area
LEGAL DEPT FILES/BASEMENT GMP
Type
REPT, OTHER REPORT
BIBL, BIBLIOGRAPHY
CHAR, CHART/GRAPH/MAPS
LIST, LIST
Site
G29
Named Person
D, P.I.
Ever
Harrison, G.W.
N, P.S.
Novotny
Rice
Schultz
Tucker, A.
Request
R1-167
Date Loaded
10 Apr 2002
Document File
98720792/98721189/Iron Workers V. Philip Morris Defendants Motion to Exclude the Damages Model and Testimony of Stanley Roberts and John Dements Volume 3 of 3
Named Organization
Adl, A.D.Little
Agency for Health Care Policy + Research
Armed Forces
Employer Group
Hhs, Dept of Health and Human Services
Hispanic Ancestry Group
Iadl
Inter Univ Consortium for Politica
Korean
Medicaid
Medicare
Natl Center for Health Services
Natl Center for Health Statistics
Ok Medicaid
States
Union Group
US
US Government
Vet Admin
Vietnam
Litigation
Feda/Produced
Author (Organization)
Univ of SC
Characteristic
ILLE, ILLEGIBLE
MISS, MISSING PAGES
UCSF Legacy ID
sqs53c00

Document Images

Text Control

Highlight Text:

OCR Text Alignment:

Image Control

Image Rotation:

Image Size:

Page 1: sqs53c00
1 1 Executive Summary 1 The statistical calculation of the fraction of medical expenditure attributable to smoking requires the use of large databases and customized computer programs. This report 1 documents the data that were used and the software that was developed. The final data and 1 computer programs used in our damage assessment are also available in electronic form. 1 1 I ! I I I 1 1 ~10 OD ~ ~ ~ I PnGE 3 oF 96 MARCH 26, 1998 I I
Page 2: sqs53c00
I I I I I I I I i- 1 I I I I I I Medicaid Expenditures Attributable to Smoking: Database Sr Software Documentation for Oklahoma by Glenn W. Hanisont March 26, 1998 t Harrison is Dewey H. Johnson Professor of Economics, Department of Economics. College of Business Administration; Universitv of South Carolina. This research was undertaken in support of litigation efforts of the Attomev-General of the State of Oklahoma. Address correspondence to Harrison at 14 Lyme Bay, Columbia, SC 29208; phone (803) 777-4943; private FAX (803) 749-8924; E-mail; GLErrNWH@wOAIANBT.ATT.NET. PAGE I oF 96 IvinRCH 26, 1998 I
Page 3: sqs53c00
Medicaid Expenditures Attributable to Smoking: Database & Software Documentation for Oklahoma by Glenn W. Harrison I i I I I I I I I I I I I I This report documents the computational steps required to calculate Medicaid expenditures attributable to smoking in Oklahoma. It complements a report by Harrison [1998] describing the procedures in "words" and drawing conclusions with respect to damages. The first stage of our analysis was to access the prirnary database on Medicaid medical expenditures, the National Medical Erpenditures Survey (NMES) for 1987. The second stage was to construct a version of this survey that linked the smoking history of natural mothers and fathers to the expenditure and other characteristics of children and teenagers. The third stage was to access the database used for Nursing Homes, the National Health and Nutrition Fxamination Survey (NHANES). The fourth stage was to undertake the statistical calculations of the fraction of Medicaid expenditures attributable to smoking. The fifth-stage was to assemble a database of medical expenditures under Medicaid along with demographics of the Oklahoma Medicaid population. This database contains individual claimant information on Medicaid expenditures and basic demographics. Each of these stages is documented in sections 1 through 5, respectively. The electronic computer files are described in section 6. Appendix A lists the names of the variables in the original NMES database. Appendices B and C list the source code of all computer programs developed. A "ZIP disk" accompanying this report provides electronic versions of all programs and data. PAGE 4 oF 96 MAxcH 26, 1998
Page 4: sqs53c00
I. I I I differs from the SAMh3EC definitions and mappings is the HCFA 64 category for Dental Services. We allocate that to OTHER PROFESSIONAL SERVICES, whereas the SAMMEC classification puts it into OTHER SERVICES, which is effectively an a priori catch-all for ` expenditures that will have a zero SAF. Table 2 shows the mapping from the original NMES database categories into the expenditure categories of the NMES-based statistical model, as well as the mapping to the HCFA 64 categories. ~ 110 ~ o) v ND ~ CD C:o ~ PAGE 20 OF 96 NhARCH 26, I99$ -P~-
Page 5: sqs53c00
I I I I I I I I NMES variable of tlie same name. Our re-coding sets all "inapplicable" and "not ascertain" responses to be "no." Hence the variable should now be interpreted as our knowing that the subject smoked, with the alternative being a subject that almost certainly did not smoke. The second constructed variable is NYRSM, which is`tftee number of years that the subject smoked. It is 0 for someone that had never smoked, and is otherwise given by the difference between AGESTOP and AGESMOKE. The third constructed variable is CIGSO, which denotes the typical number of cigarettes per day that the subject smoked in the past. This variable is obtained directly from CIGSSMOK. - - The fourth constructed variable is SMNOW, and denotes if the subject smokes now or not It is derived from NMES variable SMOKENOW, with the same re-coding used for SMOXED. The fifth constructed variable is CIGSI, and denotes the number of cigarettes that the subject smoked per day at the time of the survey. This is zero for a subject who was not a current smoker, of course. This variable is obtained directly from CIGSADAY. C. Passive Smoking Informadon_ The NMES database identifies the "original dwelling unit" in which each respondent lived at the time of the survey. Thus it is possible to identify the aggregate amount of current smoking by all residents of that dwelling unit The program GETPASS.BAS extracts the information on dwelling units from the NMES database, and then aggregates the smoking information for each member of that dwelling unit. These data are then added to the record of each individual resident in that dwelling unit The variables PCIGSO and PCIGS 1 denote the average number of cigarettes smoked daily in the past and currently by all residents of the dwelling unic To be conservative, the analysis focuses only on "current passive smoking" in the dwelling unit (PCIGS 1), and ignores effects from past passive smoking (PCIGSO). We make no assumption about the extent to which the smokers living in any dwelling unit actually smoke in the home or not. 1~0 If they do not, then there should be no statistical link between their smoking and health co . V N PAGE 14 OF 96 • MnitcH 26, 1998 ~ Co 0 cc
Page 6: sqs53c00
expenditures on other members of the household. The value of PCIGS I used for each individual is the value for their dwelling unit minus any own-smoking by that individual. This simple transformation makes it easier to determine the additive effects of passive smoking. The program GETPASS.BAS reads in the data file NMESLRAW generated by GETHSETLBAS from NMES_RAW, and produces the new data file NMESP.RAW. In all subsequent steps the file NMESP.RAW is "re-named" to NMES.RAW (usually in a separate sub-directory from the program GETNMES.BAS). I I :' I I I t I I I D. Reading in the NMES Database The LI1vmEP program NMES.LIM reads in the NMES database NMES.RAW and undertakes some simple transformations of the variables in preparation for the statistical analyses. It creates another ASCII file called NMESLITE.RAW, which contains the data used in the main statistical analysis described later. The program NMES.LIM can be run using 8mb of RAM, avoiding the use of a temporary hard disk storage file (use the LIMDEP option /800000 when invoking LIMDEP at the DOS command line). For users with limited RAM,, there are instructions for how to set up the temporary storage file in the initial READ command (just remove the question mark in column 1 beside this row). The file created by NMES.LIM is designed for use in determining the SAFs for Medicaid. A comparable analysis can be undertaken for health expenditures paid by private insurance. The file NME2.LIM generates a file called NMESLIT2.RAW which ran be used instead of the file NMESLITE.RAW. These steps are also described below. E. Random Draws from the NMES Database The bootstrap method for generating confidence intervals for the SAF values requires that we randomly draw samples with replacement front the NMES database. This is accomplished with the LIMDEP programs BSA.LIM and BSC.uB4. The first program accesses the NMESLITE.RAW file discussed above, draws 10 thousand random observations with PAGE 15 oF 96 MARCH 26, 1998 I
Page 7: sqs53c00
I ' I I I 1 , I I I I Mississippi. Tennessee, Arkansas, Louisiana, Oklahoma, Texas, Delaware, the District of Columbia. Florida, Georgia, Maryland, North Carolina. South Carolina. Virginia and West Virginia. We also use the variables NEAST and WEST to pick out two of the other°large census regions. INCITY and INSUB denote if the respondent lives in the inner city or in the suburbs, respectively. Each is derived from NMES variable POPDNSTY if it takes on values of I("core metro") or 2("other metro"). WAGE is a measure of the respondent's wage and salary income, and derives from the NMES variable AWAGE: The values shown here are annual wage and salary income in dollars, which we convert later to thousands of dollars to ease numerical scaling problems during estimation of the statistical models. OWNHOME is a binary indicator if the respondent owns their own home, and is derived from values of RENTROWX. JOB is a binary indicator if the respondent was employed during the final round of the NMES survey, as indicated by the value 1 for the NMES variable EMPLOYX4. General Health and Physical Gha>•acteristics PREG is a binary indicator to denote if the respondent was pregnant at any time during 1987, as given by NMES variable PREG19B7. A measure of obesity is constructed using the Body Mass Index of the respondent, denoted BMI in our database. The BMI is given by the weight of the respondent in grams divided by the square of the height of the respondent in meters. The NMES variables TALL and WEIGHT provide the basic data needed to compute the BMI, after conversion to metric counterparts. Most of the controversy over the use of the BMI, or any such similar scalar measure of weight status, concerns the magical threshold at which someone is deemed significantly overweight or significantly underweight. To remain agnostic about that threshold we just use the BMI directly as well as a squared term to allow for some non-linear effects from large BMI deviations from the mean. PAGE 12 OF 96 • MaxCH 26, 1998 I
Page 8: sqs53c00
1 I I I i I I I I I I I I I I I TOTALSP9. Medicaid'payments are always indicated by the last three characters SP4 (e.g.. TOTALSP4). Medicare payments are indicated by the last three characters SP3. Since some of the variables presented on this file. are subsets of other variables (e.g., DENTORTH is a subset of DENT pertaining to orthodontic services), merely summing all expenditure variables on this file for all persons will result in overestimating personal and family total health care expenditures. a D The categories extracted are as follows: DRVISEXP, DRVISSP4: visits to physicians, either doctors of Medicine or Osteopathy. NDRVSEXP, NDRVSSP4: visits with other medical providers, such as optometrists, podiatrists, chiropractors, physical therapists, speech therapists, audiologists, occupational therapists, nurses, nurse practitioners, paramedics, health aides. physician assistants, psychologists, and psychiatric social workers. O O p DRTELEXP, DRTELSP4: telephone calls to ambulatory care physicians. NDRTLEXP, NDRTLSP4: telephone calls to other medical providers. OPDDREXP, OPDDRSP4: visits to a physician, including hospital stays where the person, was admitted and discharged in the same day. 0 OPDNDEXP, OPDNDSP4: visits to other medical providers, including nurses, physician assistants, physical therapists, podiatrists, chiropractors, psychologists, and social workers (other than those included in NDREVSEXP and NDRVSSP4 above). 0 EROMSEXP, EROMSSP4: emergency room visits. 0 HSXMDEXP, HSXMDSP4: expenses for physicians who billed separately for any inpatient services provided during the hospital stay. See the note below this list for an explanation of which variables were extracted. \10 W O HSXFCEXP, HSXFCSP4: basic hospital facility expenses, including all expenses for v N direct hospital care, including room and board, diagnostic and laboratory O work, x-rays, and similar charges, as well as any physician services included in cJJ O the hospital charge. See the note below this list for an explanation of which Cr! PAGE 9 OF 96 MARCH 26, 1998 I
Page 9: sqs53c00
U I I I I I I I I I I I I I I I t. The primary NMES data source for our purposes is file DA6247. which contains the main information on expenditures. Additional data is extracted from files DA9674 (information: on smoking habits and other personal characteristics) and DA9695 (information on-income and education levels). If data are available for a given respondent from one file but not another, then we still write out a record for that respondent, noting the missing values explicitly. Thus there is no attempt to impute values or make any judgements at this stage, just data extraction. The end result is an ASCII file called NMES.RAW that contains one row of data for each respondent Each column refers to a variable, such as age or gender. This file can then be processed by the statistical gackage used in the analysis, LIhmEP version 7.0. All LIMDEP commands are contained in ASCII files with the suffix LIM, and can be executed at the DOS command line. The LiMIDEP program files are documented in detail later. B. Variables Extracted from the NMES Database The NMES database provides information on health expenditures, personal characteristics of the respondent, and health characteristics of the respondent. Each is now described, with edited extracts from the codebook files for each variable. Medica! Expenditures TOTALEXP is a constructed variable that sums health care expenditures at the - person level. Specifically, TOTALEXP includes expenses for ambulatory physician (DRVISEXP) and non-physician (NDRVSEXP) services in a clinic or office setting; including telephone calls with a charge (DRTELEXP and NDRTLEXP), ambulatory hospital outpatient physician (OPDDREXP) and non-physician (OPDNDEXP) visits, emergency room visits (EROMSEXP), inpatient hospital and physician services (HOSMDEXP and HOSFCEXP), home health care services (DRHOMEXP and NDMHMEXP), prescribed medicines (PMEDSEXP), dental services (DENTEXP), and medical equipment purchases I'D and rentals (MEXPIEXP, MEXP2EXP, and MEXP3EXP). Co Nine sources of payment for TOTALEXP are indicated by TOTALSPI through ~ C) PAGE 8 OF 96 MAttCH 26, 1998 O t1o I
Page 10: sqs53c00
I I I t I I t Smokixg Histoiy " The NMES contains considerable information on the smoking history of the respondent Given their obvious importance for our purposei, we first define the variables as :. they are in the NMES and then discuss our transformations of them separately. The exatt wording of the smoking questions is presented in Harrison [1998]. The first NMES variable is SMOKED, and indicates if the respondent ever reported that they had smoked 100 or more cigarettes in their life. Although a coarse indicator, it is meant to pick out folks that have any noticeable smoking history. It is coded I for "yes" and 2 for "no," but there is also a=I code for "inapplicable" and a -9 code for "not ascertain." The latter code is used primarily for respondents that refused to answer the question. The second NMES variables is AGESMOKE, and gives the age in years that the respondent started smoking. The -1 and -9 codes are as defined above for SMOKED. There are naturally a large number of "inapplicable" responses here since any subject with a "no" or "inapplicable" for the SMOKED question was not asked this question. The third NMES variable is CIGSSMOK, and denotes the number of cigarettes that the respondent smoked per day in the past. The same qualification applies here about the relatively large number of "inapplicable" responses due to the absence of a positive response to the SMOKED question. The fourth NMES variable is SMOKENOW, and indicates if the respondent smokes now. This is coded similarly to SMOKED. The fifth NMES variable is CIGSADAY, and tells us how many cigarettes a day the respondent smoked at the time of the survey. • The sixth NMES variables is AGESTOP, which tells us the age at which the respondent stopped smoking. As expected, there were a large number of "inapplicable"responses to this question, since it was not asked of folks who had never smoked or who smoked at the time of the survev. We use these NMES variables to construct several variables for our regression purposes. The first constructed variable is SMOKED, which is just a simple re-coding of the PAGE 13 OF 96 MARCH 26, 1998 t

Text Control

Highlight Text:

OCR Text Alignment:

Image Control

Image Rotation:

Image Size: