Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Adjusting Performance Measures to Ensure Equitable Plan Comparisons

Adjusting Performance Measures to Ensure Equitable Plan Comparisons Adjusting Performance Measures to Ensure Equitable Plan Comparisons Alan M. Zaslavsky Ph.D., Lawrence B. Zaborski, M.S., M.A., Lin Ding, Ph.D., James A. Shaul, M.H.A., Matthew J. Ciof fi, B.S., and Paul D. Clear y, Ph.D. When comparing health plans on scores ance arrangements and health care plans from the Medicare Managed Care (Weinberger, 1999). In 1997, HCFA fund­ Consumer Assessment of Health Plans ed the CAHPS consortium to develop a (MMC-CAHPS ) survey, the results should version of CAHPS suitable for assessment be adjusted for patient characteristics, not of the experiences of Medicare beneficia­ under the control of health plans, that ries in managed care. HCFA now uses that might affect survey results. We developed survey to assess Medicare managed care an adjustment model that uses self-repor ted plans annually. measures of health status, age, education, Several methodological problems com­ and whether someone helped the respondent plicate the measurement and reporting of with the questionnaire. The associations of health care data, par ticularly when repor ts health and education with survey responses draw comparisons among health plans, as differed by HCFA administrative region. is the case in the MMC-CAHPS project. Consequently, we recommend that the case- Among the challenges is the need to adjust mix model include regional interactions. appropriately for patient characteristics Analyses of the impact of adjustment show such as patients’ health and sociodemo­ that the adjustments were usually small but graphic characteristics, which are not not negligible. under the control of plans and which may af fect CAHPS scores. INTRODUCTION There are at least two reasons why it might be desirable to adjust plan CAHPS scores. In 1995, the Agency for Healthcare First, there are certain processes that one Research and Quality (then called the would expect to vary according to the char­ Agency for Health Care Policy and acteristics of patients. For example, one Research) initiated a cooperative agree­ CAHPS question is “…how much of a prob­ ment with RAND, Harvard, and Research lem did you have finding or understanding Triangle Institute to conduct the CAHPS the information…from your health plan?” study. The goals of the CAHPS project Although it is desirable to communicate included developing a standardized sur vey clearly with all patients, it probably is harder that could be used to assess the experience to do so with patients who have less educa­ of consumers in different types of insur- tion than with other patients. Second, certain personal patient characteristics might influ­ The authors are with the Har vard Medical School. Research for ence the response to questions, even if the this article was supported by the Health Care Financing process of care is the same for all patients. Administration (HCFA) under Contract Number 500-95-0057- TO#9 with the Barents Group of KPMG Consulting, Inc. in af fil­ To develop a case-mix adjustment model iation with Har vard Medical School, the MEDSTAT Group, and Westat. The views expressed in this article are those of the for the MMC-CAHPS data, we first authors and do not necessarily reflect the views of Barents reviewed published studies. Next, we ana­ Group of KPMG Consulting, Inc., Har vard Medical School, the MEDSDAT Group, Wesat, or HCFA. lyzed five data sets from sur veys of health HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 109 maintenance organization (HMO) popula­ Joshi, 1997; Lee and Kasper, 1998). tions in dif ferent par ts of the United States Perceived improvement in health also has a and identified potentially important case- strong positive association with health care mix variables (Cioffi et al., 1998). Finally, ratings (Kane, Maciejewski, and Finch, we analyzed MMC-CAHPS data to evalu­ 1997; Kippen, Strasser, and Joshi, 1997). ate alternative models. We analyzed data The few studies that have investigated from the first MMC-CAHPS sur vey, the relationship between satisfaction rat­ based on CAHPS 1.0, and from the sec­ ings and the presence of specific medical ond and third sur veys, based on CAHPS conditions have yielded inconsistent 2.0 (Clear y et al., 2000). results (Hall et al., 1990; Zapka et al., 1995; Kippen, Strasser, and Joshi, 1997). PREVIOUS RESEARCH Emotional distress and social-activity limi­ tations have been found to be negatively Given that there are few published stud­ associated with satisfaction ratings, ies of factors affecting health plan ratings, although work limitations or other limita­ we reviewed studies of hospital care, ambu­ tions due to emotional health status have latory medical services, and health plans. not (Marshall, Hays, and Mazel, 1996; Patient characteristics that have been iden­ Greenley, Young, and Schoenherr, 1982). tified as correlates of patient reports about Early investigations revealed that older their health care include (1) patient sociode­ patients are generally more satisfied than mographic characteristics, (2) overall per­ younger patients with their medical care ceived health status, (3) functional status, (Cleary and McNeil, 1988; Aharony and (4) diagnoses or conditions, (5) length of Strasser, 1993; Zapka et al., 1995), although relationship with provider or health plan findings are not consistent (Weiss, 1988; and prior use of services, (6) whether the Kane, Maciejewski, and Finch, 1997), and survey was completed by a proxy, and (7) some studies have found that this association institutional status (Cleary and McNeil, is not present in the oldest groups of patients 1988; Aharony and Strasser, 1993; Weiss, (Hall et al., 1990; Lee and Kasper, 1998). 1988; Hall et al., 1990; Cleary et al., 1992; Although some studies have found that Zapka et al., 1995; Kane, Maciejewski, and females were less satisfied than males Finch, 1997; Kippen, Strasser, and Joshi, (Cleary et al., 1992), the preponderance of 1997; Lee and Kasper, 1998). studies found that gender is not a significant Better self-reported health is consistent­ predictor of satisfaction (Clear y and McNeil, ly associated with higher ratings of health 1988; Aharony and Strasser, 1993; Weiss, care services by consumers and patients 1988; Hall et al., 1990; Zapka et al., 1995; (Cleary and McNeil, 1988; Aharony and Kane, Maciejewski, and Finch, 1997; Lee Strasser, 1993; Marshall, Hays, and Mazel, and Kasper, 1998). Studies of the association 1996; Hall, Milburn, and Epstein, 1993; between respondents’ race and ratings of Roberts, Pascoe, and Attkisson, 1983; Hall their medical care and health insurance et al., 1998). Current general health status plans have had inconsistent results (Weiss, tends to be the strongest predictor of 1988; Zapka et al., 1995; Kane, Maciejewski, patient or consumer satisfaction with health and Finch, 1997; Kippen, Strasser, and Joshi, care ser vices (Hall et al., 1990; Clear y et al., 1997). Evidence about the association 1992; Zapka et al., 1995; Kane, Maciejewski, between education levels and consumer rat­ and Finch, 1997; Kippen, Strasser, and ings of medical care and health plans is also 110 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 inconsistent (Weiss, 1988; Zapka et al., 1995; et al., 1989; Rothman et al., 1991; Magaziner Kane, Maciejewski, and Finch, 1997; et al., 1988; Sprangers and Aaronson, 1992; Kippen, Strasser, and Joshi, 1997). Magaziner et al., 1996; Rubenstein et al., 1984; Most investigations of the relationship Magaziner et al., 1997). Reports about the between income and consumer reports patient’s general health status tend to have about health care find that those with high­ the lowest levels of concordance (Magaziner er incomes provide modestly higher or et al., 1988). Some studies have found that similar ratings, compared with those with the association between proxy and patient lower incomes (Weiss, 1988; Hall et al., reports increased with higher education lev­ 1990; Clear y et al., 1992; Zapka et al., 1995; els for either respondent (Sprangers and Kane, Maciejewski, and Finch, 1997). A Aaronson, 1992; Hays et al., 1995) and with study of Medicare beneficiaries indicated increased contact between patient and proxy that more income is associated with higher respondent (Epstein et al., 1989; Sprangers levels of health care satisfaction (Lee and and Aaronson, 1992; Rubenstein et al., 1984). Kasper, 1998). An increased burden of caring for the patient Possession of additional health insur­ is associated with more negative ratings by ance coverage such as Medicaid or private proxies relative to those given by patients supplemental policies may be an impor tant (Epstein et al., 1989; Magaziner et al., 1988). predictor of health plan satisfaction. One Reports about more subjective aspects of study found that Medicare fee-for-service health appear to be influenced much more by beneficiaries with secondary insurance the proxy’s own level of psychological dis­ were more likely than others to be highly tress, age, and health status than by charac­ satisfied with their medical care (Lee and teristics of the patient (Rothman et al., 1991). Kasper, 1998). Groups with both Medicaid Few studies have investigated the effect and Medicare coverage were more likely of proxy ratings on reports about medical to be highly satisfied with their health care care and health plans (Epstein et al., 1989; than those with Medicare only. Lavizzo-Mourey, Zinn, and Taylor, 1992). Older individuals are often cognitively or One study comparing responses of 60 elder­ physically impaired and may be unable to ly patients with their proxies’ satisfaction complete a survey or unavailable to respond ratings found the association to be modest (Corder, Woodbury, and Manton, 1996). As a (Epstein et al., 1989). Proxies consistently result, sur veys are often completed by a close rated the patient’s care more negatively relative or caregiver. Proxies tend to rate the than did patients themselves. The majority patient’s health status lower than the patients (62 percent) of proxies in this study were rate themselves (Epstein et al., 1989; spouses. Another study found modest asso­ Rothman et al., 1991; Magaziner et al., 1988; ciations between proxy and patient ratings. Sprangers and Aaronson, 1992; Magaziner et However, in that study, surrogates general­ al., 1996; Rubenstein et al., 1984; Magaziner et ly rated the patient’s care more positively al., 1997). Subjective health-status dimen­ than did patients themselves (Lavizzo- sions tend to be more greatly underrated by Mourey, Zinn, and Taylor, 1992). proxies. Reports about observable physical or functional characteristics, such as the abil­ PRELIMINARY DATA AND ANALYSES ity to walk upstairs or dress, are much more consistent between patient and proxy than The data sets we used for preliminary less concrete or more private health dimen­ analyses were chosen, in part, to comple­ sions, such as emotional well-being (Epstein ment the existing literature. Only one pub- HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 111 lished study investigated the relationship for commercially insured groups. We limit­ between consumers’ ratings and sociode­ ed our analyses to data from Medicare ben­ mographic characteristics specifically for eficiaries (862 responses). Medicare beneficiaries. We selected data The NCQA data presented us with two sets that included recent consumer evalua­ unique opportunities to study the relation- tions of their health insurance plans, in ship between self-reported health-status addition to ratings of their medical care measures and health care ratings for (Table 1). Three of the five data sets Medicare beneficiaries. First, the data included a substantial number of Medicare included the Medical Outcomes Study beneficiaries from diverse geographic (MOS) Short Form 12 (SF-12) health sur­ regions of the United States. Two of the vey (Ware, Kosinski, and Keller, 1996), data sets (Washington and private employ­ which collects data about physical func­ er) included many questions that matched tioning and emotional well-being. items on the 1997 MMC-CAHPS . Each Additionally, the NCQA survey included data set included information on age, sex, unusually complete data (26 questions) education, general health status, and race. about chronic or disabling conditions that Information about chronic or disabling respondents had. conditions, physical functioning, emotional well-being, income, and proxy responses Medicare Current Beneficiar y Sur vey was included in many of the data sets. (MCBS) Minnesota Data The MCBS is administered each year to measure the economic and quality of life In 1995, the Minnesota Health Data effects of the Medicare program on Institute (MHDI) conducted a study of all enrollees. Each year approximately 16,000 health plans in Minnesota for commercially Medicare beneficiaries or their proxies are insured groups, Medicare, and medical inter viewed, of whom about 4,000 are new assistance programs. The study included to the panel. Although the MCBS focuses traditional indemnity and managed care on economic aspects of the enrollees’ expe­ plans. More than 17,000 surveys were col­ rience with Medicare, it also includes a lected for 46 different health plans. The number of questions about their experi­ MHDI used a survey that was similar to ences with their doctors and medical care. the National Committee on Quality The MCBS is unique in that it collects sat­ Assurance’s (NCQA’s) Annual Member isfaction information for a nationally repre­ Health Care Sur vey (AMHCS) (NCQA 1.0). sentative sample of Medicare beneficiaries. We used the 1996 MCBS, which includes a NCQA Sur vey supplemental sample of beneficiaries in managed care organizations. In 1996, NCQA required that health plans submitting results for the Health Plan Private Employer Sur vey Employer Data and Information Set (HEDIS ) report data for the AMHCS. A In fall 1998, a large private employer spon­ total of 43 health plans submitted results for sored an evaluation of the plans offered to more than 18,000 completed sur veys. The employees and retirees, using the CAHPS majority of data compiled by the NCQA was adult core survey. Survey responses were 112 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 113 Table 1 Data Set Components for Preliminary Analyses Project Minnesota NCQA Washington Private Employer MCBS ® ® Survey Used NCQA Annual NCQA Annual CAHPS Adult Core CAHPS Adult Core 1996 MCBS Member Health Care Member Health Care Version 1.0 Version 1.0 with Survey Hybrid Survey SF-12 Attached Method of Data Collection Mail Not Available Mail Mail Personal Interview Percent Responding Not Available 72 52 61 (Salary) Not Available 45 (Hourly) Dates of data collection April-June 1995 1996 Summer 1997 Fall 1997 1996 Number of Health Plans 46 14 20 11 Not Available Number of Respondents 17,591 862 8,310 4,678 16,411 Medicare Yes Yes No Yes, not able to identify. Yes Sample Description Survey of A mix of enrollees State employees. Large manufacturing Rotating panel commercial, ranging from chil­ company employees design to provide Medicare, and dren to Medicare and their dependents. information about Medicaid plan in clients. No plans Stratified by hourly and Medicare bene- the State to provide were exclusively salary positions. ficiaries. 1996 information to Medicare. survey included a consumers. supplemental sample of enrollees in managed care. Demographics Percent in Managed Care Not Available Not Available 92.9 82.2 22.7 Age Percent 65 or Over 97.9 94.8 1.6 21.0 85.5 Percent 18-64 2.1 5.2 98.4 79.0 14.5 Percent Male 34.0 27.0 39.8 72.6 44.1 Education Percent with Some High School or Less 27.7 26.7 1.2 9.0 39.8 Percent High School Graduate 35.4 31.7 16.4 29.0 33.0 Percent with Some College or More 36.9 41.6 82.2 62.0 27.2 Percent White Not Available 92.0 86.5 85.0 79.7 Limited to Medicare enrollees. At least part of the data. Shown for Medicare sample. NOTES: NCQA is National Committee on Quality Assurance. MCBS is Medicare Current Beneficiary Survey. CAHPS is Consumer Assessment of Health Plans. SOURCE: Data from the CAHPS , Agency for Healthcare Research and Quality; data analysis by the authors. collected from 4,678 current employees and analyses in this article. A more detailed retirees in 11 different health plans. Both description of the analyses and results is managed care and traditional indemnity available from the authors (Cioffi et al., plans were included in the evaluation. 1998). These data were unique because they included SF-12 health-status measures as Self-Reported Health Status well as CAHPS survey data. These data allowed us to explore the relationships In each of the preliminar y data sets, cur- between self-reported health measures rent general health status was the and consumer ratings that were exactly the strongest predictor of health care and same as those collected with the 1997 health plan satisfaction for both commer­ MMC-CAHPS . cially insured and Medicare enrollees. Individuals who rated their general health Washington State levels higher also gave higher ratings of their health care ser vices. An evaluation of 20 health plans offered A general health-status variable was to State employees by the Washington analyzed as a continuous or categorical State Health Care Authority obtained 8,310 variable in the MCBS, Washington, private responses. The study was conducted in employer, and NCQA data sets. The con­ summer 1997 using the CAHPS adult tinuous variable accounted for the same core survey, which includes a number of amount of variation as the categorical vari­ items that matched those in the 1997 able, because ratings of medical care and MMC-CAHPS . of health plans improved by about the same amount for each step on the general ANALYSES health-status response scale. We estimated linear models in which the Physical Functioning, Comorbidities, dependent variable is the response on a and Chronic Conditions survey item or set of items (composite) and the independent variables are case-mix Our preliminar y analyses generally indi­ adjusters. In the data sets in which sample cated that measures of physical-function­ size was adequate and the data distin­ ing limitations were not significant inde­ guished among multiple plans, we includ­ pendent predictors of care ratings. ed dummy variables for each of the plans. Analyses of data from Medicare beneficia­ When we control for plan ef fects, the case- ries in the NCQA data set also indicated mix coefficients represent within-plan that physical functioning was not a signifi­ ef fects of the adjuster variables. We tested cant predictor, after controlling for emo­ the predictive power of variables individu­ tional status, general health status, and ally and in combination. age. Analysis of the private employer data, which consists of retirees and current Results of Preliminar y Analyses employees, also revealed that work or activity limitations, physical functioning Because of the large number of analyses limitations, and limitations due to pain involved, we do not present details of the were infrequently related to health care empirical results from the preliminary ratings after controlling for age, general 114 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 health status, and emotional well-being. In pendent predictor of satisfaction levels, the Washington State data, a variable indi­ because it was strongly correlated with cating whether the survey respondent feeling calm and peaceful. needed help with personal care or meeting routine needs, or had a condition that inter­ Age fered with his or her independence, with work, or with school activities, was not a Older adults are generally more satisfied significant predictor for the respondent’s with their medical care and health plan ser­ ratings of overall medical care, specialists, vices. However, ratings do not increase or personal doctors. Having a limitation monotonically with age for those over age was significantly associated only with the 65. Among Medicare beneficiaries in respondent’s overall rating of health plan. Minnesota, older individuals are less satis­ We studied the association between fied than younger ones with their health medical care and health plan ratings, and plan, medical care, and access to care. having a chronic or disabling condition, There were no significant differences using the MCBS, NCQA, and private among age groups for appointment access employer data sets. A variable constructed and physician choice. from the NCQA data set indicated the num­ Results from Medicare enrollees in the ber of conditions (from a list of 26) an indi­ NCQA data file indicated that satisfaction vidual reported having. For the private increases with age until the 80-84 or 85-89 employer data set, we analyzed a variable age groupings, at which point it levels off that indicated whether respondents had a or declines. Evidence from the MCBS sug­ medical condition that had lasted for 3 or gests that satisfaction decreases with age, more months. Neither chronic nor dis­ with most of this effect resulting from abling conditions were significant predic­ lower satisfaction among the oldest tors of satisfaction outcomes in these data respondents (those over age 85). sets. In the MCBS, however, individuals There is little evidence that age affects who reported having any of four physical the relationship between ratings and the conditions tended to provide higher health other sociodemographic characteristics. care satisfaction ratings. No significant interaction effect was dis­ covered between age and health status. Emotional Well-Being The effects of education, income, and Hispanic background were related to age In the NCQA analyses, a general mea­ for those 85 years and over, but these inter- sure of emotional well-being was a signifi­ actions were inconsistent and only margin- cant predictor of health care ratings. ally significant. Feeling calm and peaceful was the most important emotional-status predictor of Sex higher levels of satisfaction in the private employer data analysis. Respondents with Sex was not significantly related to fewer work and social limitations due to health plan or medical care ratings. emotional distress gave higher satisfaction Analyses of employed adults and Medicare ratings, but these results were inconsis­ beneficiaries indicated that females were tent. Feeling energized was not an inde more satisfied than males, but the effects were only marginally significant. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 115 Race Income The relationships between race and rat­ Only the 1996 MCBS data set included ings of health plan or medical care were an income variable that allowed us to study inconsistent in the preliminary data sets. the relationship between the income levels Among retirees and current employees in of Medicare beneficiaries and ratings of the private employer data, race was a signifi­ medical care. Our results were consistent cant predictor only for ratings of specialists with findings from a study showing that but not for health plans, personal doctors, or elderly respondents with higher income overall health care. Black persons, Asians, levels tended to rate their medical care bet­ and Pacific Islanders tended to be more sat­ ter than other respondents (Lee and isfied than white persons in this study. Kasper, 1998). Increases in satisfaction at (Throughout this discussion, the term “white higher income levels were generally mod­ persons” refers to white people who are not est. Ef fects of secondar y sources of health Hispanic.) In the Washington State data, insurance coverage were not assessed in race had a significant relationship only with any of the preliminar y data sets. the specialist ratings: Hispanic persons and Asians were less satisfied than white people. Proxy Respondents Among Medicare beneficiaries in the NCQA data set, Hispanic people tended to We were able to assess the impact of give lower ratings than white persons of their response by a proxy only for the MCBS health plans and of getting approvals or refer­ data set. Proxy ratings of medical care are rals for care. No other differences were sig­ only marginally higher than those provid­ nificant. Analyses of the MCBS data set ed by the intended Medicare respondents. revealed somewhat different results. Black people tended to be less satisfied with both Context Variables their doctors and medical care than white people. Hispanic persons were more satisfied We studied several variables describing than white people with their doctors, but less the social context or community in which satisfied than white people with their care. the respondent lives, using ZIP Code level 1990 U.S. census data. We considered Education seven variables, each of which is measured as percentage of residents in the respon­ More educated Medicare beneficiaries dent’s ZIP Code who belong to the respec­ in the Minnesota and MCBS data sets tive group: Ethnicity (Black, Asian, rated their care higher than those who Hispanic), College-Educated, High-Status were less educated. There was no rela­ Occupation, Urban Resident, Public tionship, however, among Medicare Assistance Recipients (overall and among enrollees in the NCQA data set. In the pri­ those over the age of 65). vate employer and Washington data, those The NCQA analysis revealed that indi­ with more education tended to be less sat­ viduals who live in areas with a high per­ isfied with their medical care and health centage of Asians were more likely to insurance plan. An ordinal education vari­ report satisfaction with their plan and with able predicted ratings as well as a set of ability to get referrals. Respondents from categorical variables, because satisfaction areas where there is a high concentration levels changed roughly linearly. of black residents and from densely 116 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 ® populated urban areas report fewer prob­ MMC-CAHPS Data lems with their health care. These respon­ dents report greater satisfaction with their Instrument plan overall and also with the quality of care received. In the MCBS data, resi­ The MMC-CAHPS survey, fielded in dents of areas with large concentrations of 1997, included all items of the CAHPS 1.0 college-educated individuals and residents adult core instrument (Hays et al., 1999) of areas with many persons with high-sta­ and 28 additional MMC-specific items tus occupations tended to be more satisfied (Clear y, Zaslavsky, and Ciof fi, 2000). Of 85 with their doctors and rated their overall items, 4 elicit overall ratings and 34 elicit care higher. reports of respondent experiences. Other In the Washington State analysis, resi­ questions are used to determine the applic­ dents of more urbanized areas were more ability of par ticular repor t questions or ask satisfied with doctors, specialists, and about sociodemographic characteristics, health plans. Respondents from areas with health status, and health care utilization. high percentages of college graduates or The MMC-CAHPS sur vey fielded in 1998 persons with high-status occupations tend­ and 1999 included all items of the CAHPS ed to provide higher ratings for both their 2.0 adult core instrument and 41 additional doctors and their overall health care. MMC-specific items. The potential case- However, when we include both of these mix variables in the MMC-CAHPS ques­ highly correlated variables in our model, tionnaire are available from the author, only education had a significant positive including 10 variables from the sur vey and effect on the CAHPS scores (an effect 6 variables based on respondents’ ZIP that is opposite to that of individual level Codes. education in this par ticular analysis), while occupation has a positive ef fect on satisfac­ Sample tion with doctor. For each MMC-CAHPS survey, HCFA SUMMARY OF PRELIMINARY drew a stratified sample of Medicare bene­ ANALYSIS RESULTS ficiaries who had been enrolled in an eligi­ ble plan. Eligible plans included all health The effect of a few patient characteris­ plans with Medicare contracts in effect on tics, particularly health status and age, are or before January 1 in the year preceding consistent across multiple studies, while the survey and in business for 2 years. others have effects that are either weak or Contracts that covered large areas were inconsistent. Some of the inconsistencies divided into geographically defined repor t­ might be attributable to the diverse set­ ing units. A simple random sample of up to tings and populations studied. In particu­ 600 members was drawn from each plan or lar, population-based studies might con- reporting unit. found case-mix effects with selection of For each survey, we deleted cases sam­ some groups or patients into more or less pled from contracts that had ceased activity, favorable situations. In the next section, had only one beneficiary (two plans in the we report analyses of Medicare CAHPS second year) or had been terminated, and data that compare patients’ reports within beneficiaries that left their plan before the a single reimbursement system and a large survey was administered, as well as number of health plans. deceased and institutionalized beneficiaries. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 117 In the first survey, there were 89,802 unadjusted means (Zaslavsky, 1998). We valid sur veys from 119,267 eligible benefi­ first evaluated explanatory power (EP) ciaries, representing a response rate of 75 using a linear specification and then tested percent. In the second sur vey, there were the improvement in fit by replacing it with 123,000 valid sur veys from 152,144 eligible variables for each response categor y. beneficiaries, representing a response rate We also investigated the possibility that of 81 percent. In the third survey, there the effects of case-mix variables would were 166,072 valid surveys from 202,775 var y by region. We assigned all responses eligible beneficiaries, representing a from each plan to the single region in response rate of 82 percent. which the plan had the largest enrollment, as determined by the Medicare Managed Sur vey Procedures Care Market Penetration for All Medicare Plan Contractors Quarterly State/County/ Survey data collection took place from Plan data file. This allowed us to adjust Februar y to May 1998 for the first sur vey each plan using a single model and facili­ and from September to December for the tated comparisons among plans operating second (1998) and third (1999) surveys. in the same area. For most plans, 70 per- Although there were slight modifications cent or more of their enrolled population in survey protocols, the basic approach was within a single region. Because of the was comparable each year. The survey small number of plans and managed care firm mailed a preliminary notification let­ enrollees in several HCFA regions in the ter, followed by the survey. Non-respon­ first year of CAHPS data examined, we dents were sent a reminder postcard, and if combined regions 5, 7, and 8 for case-mix no sur vey was received, a duplicate sur vey modeling. For consistency across years, was sent. Interviewers contacted respon­ we retained that grouping for each year. dents by telephone to complete missing We tested interactions between region items and to followup for non-response, if a and a linear effect of age, education, and telephone number could be obtained. reported health status, in models predict­ ing the four CAHPS general ratings. To MMC Analyses evaluate whether it was necessary to cre­ ate an interaction term for each age cate­ The statistical criteria for usefulness of a gor y, we assessed alternative models, com­ variable for case-mix adjustment include paring a model in which a regional interac­ both its predictive power in the pooled tion was estimated for each categor y with a within-plan regression model and the model containing a regional interaction degree of between-plan variability in the with the linear age ef fect. Similar analyses variable, relative to its within-plan variabili­ were performed for the education and ty. In the analyses presented here, we health-status interactions. combined information about predictive We analyzed all 3 years of MMC­ power and between-plan variability to CAHPS data. To simplify the presenta­ obtain an overall summar y of the impact of tion, we show only year three results in the the variable on adjustment: the ratio of the tables. Tables containing all 3 years are variance of the adjustments for the new available from the authors. variable to the between-plan variance of the 118 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 MMC Results In Model 1 in Table 2, we first controlled for age and health and then entered each of We examined the associations between the other potential adjuster variables sepa­ adjuster variables and CAHPS ratings and rately to determine their EP. Controlling the explanator y power of potential adjuster for health status and age, individuals with variables to select a final adjustment more education rated their health plans, model. In Table 2, we present two sets of medical care, personal doctors, and spe­ analyses. In Model 1, we present the cialists lower than those with less educa­ explanator y power of variables controlling tion. This relationship was consistent for only for age and health. In Model 2, we all models that were tested, even when all present the explanatory variables after possible predictors were included in the controlling for a set of core variables. model. The between-plan variance for edu­ Later, we discuss the rationale for using cation is large compared with that for age these two approaches and the results. We and health-status variables. It has the discuss both the predictive strength and largest or second largest EP on all of the explanator y power of variables but present four global ratings for all 3 years. For all data only on explanator y power for all vari­ the composites, education had the largest ables. The coef ficients for the core models EP in at least 1 of the years and for Plan are available from the authors. Paper work, it was the most impor tant in all Age and general health status were cho­ 3 years (data not shown). Based on these sen for inclusion in the core model because results, we decided to include education as the literature and preliminar y analyses indi­ part of our base case-mix model. cated that they are consistently the The MMC-CAHPS survey asked strongest predictors of satisfaction and whether beneficiaries received help filling because they were the case-mix adjusters out the survey and what type of help they in the standard model for the core CAHPS received. The two proxy variables were project. In the MMC data, there was a pos­ PROXY (helped with the sur vey in any way) itive relationship between age and the and ANYPROXY (somebody answered the CAHPS ratings, even when other demo- sur vey for the subject). Both variables were graphic, health status, and contextual vari­ significant predictors of most of the ratings ables were entered into the equation. The and composites, but both had small EPs younger group (under 65, essentially all because their contribution to the predictive disabled) and the group age 65 to 69 years power of the model was relatively small, and tended to give the lowest ratings. The frac­ the proportion of individuals receiving help tions in the extreme age groups, under age did not vary much across health plans. 65 and over age 80, varied greatly between Nevertheless, adjusting for proxy respons­ plans. This suggests that age would have es may be important because of common some impact on case-mix adjustment. concerns that the inability of some benefi­ Health status was consistently the ciaries to complete a survey by themselves strongest positive predictor of consumers’ will compromise the validity of the survey ratings for all measures tested. On the other results. Thus, we included this variable in hand, there was less between-plan variance the case-mix model despite its limited for health status than for some other vari­ impact on adjustments of scores. ables. Nonetheless, because of its predictive Model 2 in Table 2 controls for age, power, health status is an important variable health, education, and proxy responses, in the case-mix model (Table 2). and tests the explanator y power of all other HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 119 120 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Table 2 Explanatory Power of Each Potential Adjuster Ratings of Health Plan Medical Care Personal Doctor Specialists Adjusters Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Age 0.431 — 0.241 — 0.144 — 0.091 — Health 0.274 — 0.295 — 0.140 — 0.104 — Education 0.745 — 0.453 — 0.339 — 0.194 — Proxy 0.035 — 0.024 — 0.024 — 0.014 — Proxy Answer 0.003 — 0.003 — 0.004 — 0.002 — Male 0.005 0.002 0.004 0.001 0.005 0.003 0.002 0.001 Race White 0.030 0.007 0.070 0.045 0.207 0.163 0.000 0.000 Black 0.030 0.009 0.076 0.047 0.148 0.107 0.005 0.001 Hispanic 0.014 0.005 0.007 0.004 0.017 0.013 0.002 0.001 Asian 0.012 0.004 0.009 0.006 0.005 0.002 0.016 0.011 Native American 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Other 0.004 0.003 0.002 0.001 0.001 0.002 0.004 0.003 Medical Condition 0.001 0.003 0.011 0.015 0.012 0.015 0.024 0.026 ADLs 0.012 0.005 0.005 0.002 0.002 0.000 0.003 0.001 IADLs 0.022 0.010 0.011 0.005 0.002 0.000 0.004 0.002 Independent 0.026 0.014 0.029 0.020 0.009 0.004 0.011 0.008 ZIP Code Variables Percent College Degree 0.468 0.075 0.279 0.054 0.187 0.025 0.074 0.003 Percent Urban 0.000 0.014 0.042 0.011 0.006 0.029 0.038 0.069 Percent Black 0.051 0.013 0.122 0.072 0.171 0.110 0.005 0.000 Percent Hispanic 0.306 0.120 0.209 0.104 0.353 0.195 0.148 0.116 Percent Asian 0.047 0.007 0.040 0.012 0.003 0.023 0.010 0.021 Percent Over Age 65 Receiving Public Assistance 0.162 0.024 0.308 0.158 0.287 0.129 0.050 0.016 NOTES: Explanatory power=(variance of adjustments for a variable)/(variance of plan means). All values are multiplied by 1,000 for legibility. In Model 1, the base model for calculation of age and health explanatory power is a null model (with only plan effects). For all other variables, Model 1 includes age and health, with plan differences absorbed as well. In Model 2, the base model for all variables includes age, health, education, and proxy. ADLs is activities of daily living. IADLs is instrumental activities of daily living. SOURCE: Data from the Consumer Assessment of Health Plans , Agency for Healthcare Research and Quality; data analysis by the authors. ® potential adjusters. These analyses indi­ MMC-CAHPS respondents were asked cated that ZIP Code Hispanic, public assis­ three questions about having a health tance (senior), and/or self-reported Asian problem that (1) caused them to need help race had some marginal explanatory with personal care needs, such as eating, power. ZIP Code education, which dressing, or getting around the house, (2) appeared potentially important in tests caused them to need help with routine with the first base model, was not impor­ needs, such as everyday household tant after controlling for individual educa­ chores, doing necessary business, shop- tion. The influence of each is much less ping, or getting around for other purposes, than that of age, health, and beneficiary and (3) seriously inter fered with their inde­ education, however. Also, the influence of pendence, participation in the community, these variables was not consistent across or quality of life. Two of the physical-func­ all dependent variables or the three sur­ tioning indicators were related to ratings, veys. Of the variables tested, ZIP Code even after controlling for general health Hispanic appeared to be the most impor­ status. Respondents with a physical limita­ tant. The inconsistency in these results tion that interfered with independence, and the age of the census data on which participation in the community, or quality these ZIP Code variables are based would of life rated their health plans, medical argue against including these variables in care, specialists, and personal doctors the MMC-CAHPS case-mix adjustment lower. Respondents that needed help with model. However, we examined regional personal care were more likely to give interactions and a model that includes ZIP lower ratings of the health plans and med­ Code Hispanic as a potential model option ical care overall and marginally lower rat­ in subsequent analyses. ings of specialists. Needing help with rou­ Respondents reporting more medical tine needs such as household chores or conditions provided higher ratings of their shopping was not a significant predictor of health plan, medical care overall, special­ ratings due to its high correlation with ists, and personal doctors. This counterin­ needing help with personal care needs. tuitive finding may indicate that it is not the Although physical-functioning indicators mere presence of disease that leads to were significant predictors for the lower satisfaction ratings, but the level of Medicare population, their predictive severity and disabling effect that accompa­ power was modest compared with self- nies the disease. In addition, individuals reported general health status, and they who use health care services more fre­ varied little across plans. Therefore, quently might be more knowledgeable including these variables in the model about their condition and be more likely to would have little effect on the outcomes. report a condition on the survey. Higher Males reported lower ratings than use of ser vices may also indicate increased females of their health plans and personal satisfaction with the services received. doctors in year one and lower scores on all However, the mean number of chronic con­ ratings in years two and three. However, ditions did not var y much across Medicare even when sex was a significant indicator, plans. Therefore, adjustment for the preva­ its predictive power was small and it had lence of medical conditions would not have the smallest variation between health much impact on health plan ratings. plans. Therefore, it had very little impact in the case-mix adjustment. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 121 The relationships between race and health care ratings in year one for respon­ health plan or medical care ratings were dents from areas in which many residents not consistent. Asian Medicare beneficia­ were black or Hispanic. In years two and ries were the only group to consistently three, persons from areas with many rate aspects of their health plans, medical Hispanic residents had higher ratings of care, personal doctors, and specialists plans and doctors, and in year two for care lower than white beneficiaries. Black and as well. Respondents from an urban area Hispanic persons rated their health plans gave slightly lower health care ratings in marginally higher than white persons but years two and three but higher specialist did not differ significantly for ratings of ratings in year two. Although the effects of their medical care overall. Black persons the racial/ethnic and poverty contextual were significantly more positive than white variables are interesting, we are reluctant persons about their personal doctors, while to use them now because the effects are Hispanic people were marginally more pos­ inconsistent and for the same reasons as for itive. Hispanic persons also provided mar­ the individual racial/ethnic variables. ginally lower ratings of their specialists For each of the individual-level variables— compared with white persons. Native age, health status, and education—we cal­ Americans provided marginally lower rat­ culated F-tests that compared the model ings of their health plans than did white with the variable entered in the linear (one- persons. We did not recommend using coefficient) specification to the model with race and/or ethnicity variables in a national the variable entered as a set of dummy vari­ case-mix model because of the lack of con­ ables. For the age and education variables, sistency in their effects. We were also con­ the test clearly rejected the simpler (lin­ cerned that their effects might depend on ear) specification for each of the four rating local associations of cultural and socioeco­ scales (data not shown). The effect of age nomic characteristics with race and ethnic­ showed a clear trend for most levels and ity that might var y from region to region. outcome variables, in which ratings All six of the contextual variables, which increased with age. On the other hand, the describe the ZIP Code area in which a per- steps in mean satisfaction were not equal son lives, had large between-plan differ­ for each increase in age categor y; instead, ences. This is understandable because satisfaction appeared to level off in the these variables represent averages over older categories. Similarly, the trend in the areas, and plans also tend to operate within individual education variable was toward areas. The ratio of within- to between-plan lower satisfaction with more education, but variances for these variables are larger the steps were not equal for each increase than those for almost all of the individual- in education categor y. Therefore, the cate­ level variables. Therefore, the ZIP Code gorical effects were more accurate repre­ variables typically had an impact on case- sentations of age and education effects mix adjustment when they were signifi­ than linear variables. For health status, the cantly related to CAHPS scores. linear trend toward lower ratings with Respondents from areas containing more worse health status (coded by higher num­ educated residents were more likely to pro- bers on the health-status response scale) vide slightly lower ratings for health plans appears to be an adequate description of in all 3 years and for specialists in year one. the relationship. There were marginally positive effects on 122 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Regional Interactions expected if more or less favorably rated plans had been randomly distributed Analyses of interaction effects showed across regions. The regional interaction that there were strong regional interac­ effect is tested against the plan interaction tions for health status, education, and ZIP ef fect. With this test, health status, educa­ Code percent Hispanic response for at tion, and proxy response have significant least four of the nine outcome measures in differences across a region, while the the year three data. Health status had sig­ region-specific ZIP Code Hispanic effect is nificant regional interactions for six of the no longer significant. This suggests that variables and was one of the strongest pre­ there is substantial variation in the dictors of the ratings. To evaluate the sta­ Hispanic coefficient from plan to plan, so bility of these interaction effects across although the average coefficient differs multiple years of CAHPS Medicare analy­ across regions, it does not dif fer more than sis, we also used a model incorporating it would if plans had been randomly data from years two and three of the assigned to regions. CAHPS sur vey. We did not use data from The absolute and relative magnitude of year one because of differences in the for- the regional effects varies substantially mat and questions of the survey in that from year to year. Nevertheless, we sug­ year. Fitting a model with 2 years of data, gest that it is useful to include interaction we allowed for slopes on case-mix terms for health and education by region; adjusters to vary (by including both the the two variables that appeared to have the region-interaction effect and a region-by- most consistent interregional variability. year interaction). We estimated separate models for regional interactions for age, Impact of Case-Mix Adjustment education, health status, proxy response, and ZIP Code percent Hispanic. Each of To assess the effects of adjustment on these models included an additional inter- the ratings of plans, relative to the unad­ action term allowing these regional inter- justed ratings, we compared adjusted rat­ action slopes to var y by year. In all cases, ings with unadjusted ratings, using several we found no evidence of change in the measures of the dif ferences. The results of interaction effects across the years. In par­ the impact analyses were comparable for ticular, for education and health, the ratio the 3 years of data. Considering the ratios of the overall effect to the interaction with of adjustment to unadjusted standard devi­ year was large, indicating that the regional ations for each variable, the largest impact interactions were stable over the 2 years of adjustments is on “getting care you compared (and therefore likely to repre­ need” and the smallest is for “ease of get­ sent consistent patterns rather than ran­ ting referrals.” The standard deviation of dom variations). plan means is only slightly smaller for the We also calculated F-tests of the signifi­ various adjusted means than for the unad­ cance of regional-interaction effects in an justed means. ANOVA model, treating plan effects as the The largest adjustments upward are random error term. This tests whether the comparable to one standard deviation of effect of our case-mix adjusters varies by the plan means for most measures. The region (i.e., an interaction of each case-mix largest adjustments downward are usually adjuster with region) more than would be much smaller, half as big or less. This HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 123 suggests that there are a few plans with SUMMARY AND CONCLUSIONS unusually adverse case mix, from the standpoint of the ef fect of case-mix on con­ Previous studies as well as the analyses sumer assessments. presented here support the continued use Comparison of the ratio of the standard of perceived health status and age in deviation of adjustments to the standard CAHPS case-mix adjustment models. deviation of unadjusted means across Although education does not explain a regions suggests that the impact of adjust­ large proportion of the variance in the ment may be somewhat larger in some dependent variables assessed, there is regions than in others. Generally, the ratio more interplan variability in education than is above average in the Pacific, New in age or health, and as a consequence, England, and Upper Midwest Regions, education predicts more interplan variabil­ below average in New York and New ity than either health status or age in some Jersey, Mid-Atlantic, and South Atlantic, models. and mixed in Northwest and Southwest. Response by a proxy is not an impor tant To quantify the ef fect of case-mix adjust­ predictor of responses, either for individual- ment on the ranking of plans, we calculated level analyses or for assessments of inter- the Kendall Tau correlation coefficient plan variability. We suggest including the between the adjusted and unadjusted plan proxy variable primarily because of con­ ratings. This measure is related to the frac­ cerns about the potential effects of cogni­ tion of pairs of plans that switched ordering tive impairment on reports about plan as a consequence of case-mix adjustment, experiences in this population and the like­ where the denominator is the total number lihood that proxy respondents describe of pairs of plans. (The Kendall Tau statistic experiences with the health plan different­ stretches this quantity to a scale from -1 to ly than enrollees would. Thus, the +1, to make it comparable to other correla­ Medicare adjustment model now includes tion coefficients.) health status, age, education, and a vari­ The Kendall Tau statistics for overall rat­ able indicating whether a proxy answered ing of plans in the 3 years were 0.92, 0.89, the sur vey. We also recommend including 0.91, indicating that the percentages of interaction terms for health and education pairs of plans whose ordering would be by region because they are the two vari­ changed using that adjustment model were ables that appeared to have the most con­ 3.9, 5.5, and 4.5 percent, respectively. sistent interregional variability. Generally, where the ratio of the standard In general, the case-mix adjustments are deviation of the adjustment divided by the not large and do not greatly change the pic­ standard deviation of the adjusted mean is ture of which plans are high- or low-rated. larger, the Kendall Tau is smaller and the It is noteworthy, however, that the largest fraction of pairs that would be switched is adjustments are quite substantial, so there larger. Nonetheless, the unadjusted and are at least a few plans for which, under our adjusted means give between-plan compar­ models, an important part of their mea­ isons that are in agreement, most of the sured satisfaction can be attributed to case time, in ever y region. mix rather than to actual plan per formance. 124 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Hays, R.D., Shaul, J.A., Williams, V.S.L., et al.: REFERENCES Psychometric Properties of the CAHPS 1.0 Sur vey Measures. Medical Care 37(3) Supp:MS22-MS31, Aharony, L., and Strasser, S.: Patient Satisfaction: What We Know About and What We Still Need to Kane, R.L., Maciejewski, M., and Finch, M.: The Explore. Medical Care Review 50(1):49-79, Spring Relationship of Patient Satisfaction with Care and Clinical Outcomes. Medical Care 35(7):714-730, 1997. Ciof fi, M.J., Clear y, P.D., Ding, L., et al.: Analysis of Kippen, L.S., Strasser, S., and Joshi, M. : Improving Case-Mix Strategies and Recommendations for the Quality of the NCQA Annual Member Medicare Managed Care CAHPS . HCFA Report. Satisfaction Survey Version 1.0. American Journal December 1998. of Managed Care 3 (5):719-730, 1997. Cleary, P.D. and McNeil, B.J.: Patient Satisfaction Lavizzo-Mourey, R., Zinn, J., and Taylor, L.: Ability as an Indicator of Quality Care. Inquiry 25(1):25-36, of Surrogates to Represent Satisfaction of Nursing Spring 1988. Home Residents with Quality of Care. Journal of the Cleary, P.D., Zaslavsky, A.M., and Cioffi, M.: Sex American Geriatrics Society 40(1):39-47, 1992. Differences in Assessments of the Quality of Lee, Y., and Kasper, J.D.: Assessments of Medical Medicare Managed Care. Women’s Health Issues Care by Elderly People: General Satisfaction and 10(2):70-79, 2000. Physician Quality. Health Services Research Cleary, P.D., Edgman-Levitan, S., McMullen, W., 32(6):741-757, Februar y 1998. and Delbanco, T.L.: The Relationship Between Magaziner, J., Itkin, Z.S., Gr uber-Baldini, A.L., et al.: Reported Problems and Patient Summary Proxy Repor ting in Five Areas of Functional Status: Evaluations of Hospital Care. Quality Review Comparison with Self-Reports and Obser vations of Bulletin 18(2):53-59, Februar y 1992. Performance. American Journal of Epidemiology Corder, L.S., Woodbury, M.A., and Manton, K.G.: 146(5):418-428, September 1997. Proxy Response Patterns Among the Aged: Effects Magaziner, J., Simonsick, E.M., Kashner, T.M., and on Estimates of Health Status and Medical Care Hebel, J.R.: Patient-Proxy Response Comparability Utilization from the 1982-1984 Long-Term Care on Measures of Patient Health Status and Sur veys. Journal of Clinical Epidemiology 49(2):173- Functional Status. Journal of Clinical Epidemiology 182, 1996. 41(11):1065-1974, 1988. Epstein, A.M., Hall, J.A., Tognetti, J., et al.: Using Magaziner, J., Pear, B.S., Hebel, J.R., and Gruber- Proxies to Evaluate Quality of Life: Can They Baldini, A.: Use of Proxies to Measure Health and Provide Valid Information About Patients’ Health Functional Status in Epidemiologic Studies of Status and Satisfaction with Medical Care. Medical Community-Dwelling Women Aged 65 Years and Care 27(3):S91-S98, March 1989. Older. American Journal of Epidemiology 143:283- Greenley, J.R., Young, T.B., and Schoenherr, R.A.: 292, 1996. Psychological Distress and Patient Satisfaction. Marshall, G.N.., Hays, R.D., and Mazel, R.: Health Medical Care 20(4):373-385, April 1982. Status and Satisfaction with Health Care: Results Hall, J.A., Feldstein, M., Fretwell, M.D., et al.: from the Medical Outcomes Study. Journal of Older Patients’ Health Status and Satisfaction with Consulting and Clinical Psychology 64(2):380-390, Medical Care in an HMO Population. Medical Care 28(3):261-269, March 1990. Roberts, R.E., Pascoe, G.C., and Attkisson, C.C.: Hall, J.A., Roter, D.L., Milburn, M.A., and Daltroy, Relationship of Service Satisfaction to Life L.H.: Why Are Sicker Patients Less Satisfied with Satisfaction and Perceived Well-Being. Evaluation Their Medical Care? Tests of Two Explanatory and Program Planning 6(3):373-383, 1983. Models. Health Psychology 17(1):70-75, 1998. Rothman, M.L., Hedrick, S.C., Bulcroft, K.A., et al.: Hall, J.A., Milburn, M.A., and Epstein, A.M.: A The Validity of Proxy Generated Scores as Causal Model of Health Status and Satisfaction with Measures of Patient Health Status. Medical Care Medical Care. Medical Care 31(1):84-94, January 29(2):115-124, Februar y 1991. Rubenstien, L.Z., Schairer, C., Wieland, D., and Hays, R.D., Vickrey, B.G., Perrine, H.K., et al.: Kane, R.: Systematic Biases in Functional Status Agreement Between Self-Reports and Proxy Assessment of Elderly Adults: Effects of Different Reports of Quality of Life in Epilepsy Patients. Data Sources. Journal of Gerontology 39(6):686- Quality of Life Research 4(2):159-168, 1995. 691, 1984. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 125 Sprangers, M.A.G., and Aaronson, N.K.: The Role Zapka, J.G., Palmer, R.H., Hargraves, J.L., et al.: of Health Care Providers and Significant Others in Relationships of Patient Satisfaction with Evaluating the Quality of Life of Patients with Experience of System Performance and Health Chronic Disease: A Review. Journal of Clinical Status. Journal of Ambulatory Care Management Epidemiology 45(7):743-760, 1992. 18(1):73-83, 1995. Ware, J.E., Kosinski, M., and Keller, S.D.: A 12-Item Zaslavsky, A.M.: Issues in Case-Mix Adjustment of Short-Form Health Sur vey: Construction of Scales Measures of the Quality of Health Plans. and Preliminary Tests of Reliability and Validity. Proceedings, Government and Social Statistics Medical Care 34(3):220-233, 1996. Sections. American Statistical Association. Alexandria, VA. 1998. Weinberger, M., (ed.): Consumer Assessment of Health Plans Study (CAHPS ). Medical Care Reprint Requests: Paul D. Clear y, Ph.D., Department of Health 37(3)Supp, 1999. Care Policy, 180 Longwood Avenue, Harvard Medical School, Weiss, G.L.: Patient Satisfaction with Primary Boston, MA 02115-5899. E-mail: clear [email protected] vard.edu Medical Care: Evaluation of Sociodemographic and Predispositional Factors. Medical Care 26(4):383- 392, April 1988. 126 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Health Care Financing Review Pubmed Central

Adjusting Performance Measures to Ensure Equitable Plan Comparisons

Zaslavsky, Alan M.; Zaborski, Lawrence B.; Ding, Lin; Shaul, James A.; Cioffi, Matthew J.; Cleary, Paul D.
Health Care Financing Review , Volume 22 (3) – Sep 1, 167
Download PDF
18 pages

Loading next page...
 
/lp/pubmed-central/adjusting-performance-measures-to-ensure-equitable-plan-comparisons-ky9B739dHf

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Pubmed Central
ISSN
0195-8631
eISSN
1554-9887
Publisher site
See Article on Publisher Site

Abstract

Adjusting Performance Measures to Ensure Equitable Plan Comparisons Alan M. Zaslavsky Ph.D., Lawrence B. Zaborski, M.S., M.A., Lin Ding, Ph.D., James A. Shaul, M.H.A., Matthew J. Ciof fi, B.S., and Paul D. Clear y, Ph.D. When comparing health plans on scores ance arrangements and health care plans from the Medicare Managed Care (Weinberger, 1999). In 1997, HCFA fund­ Consumer Assessment of Health Plans ed the CAHPS consortium to develop a (MMC-CAHPS ) survey, the results should version of CAHPS suitable for assessment be adjusted for patient characteristics, not of the experiences of Medicare beneficia­ under the control of health plans, that ries in managed care. HCFA now uses that might affect survey results. We developed survey to assess Medicare managed care an adjustment model that uses self-repor ted plans annually. measures of health status, age, education, Several methodological problems com­ and whether someone helped the respondent plicate the measurement and reporting of with the questionnaire. The associations of health care data, par ticularly when repor ts health and education with survey responses draw comparisons among health plans, as differed by HCFA administrative region. is the case in the MMC-CAHPS project. Consequently, we recommend that the case- Among the challenges is the need to adjust mix model include regional interactions. appropriately for patient characteristics Analyses of the impact of adjustment show such as patients’ health and sociodemo­ that the adjustments were usually small but graphic characteristics, which are not not negligible. under the control of plans and which may af fect CAHPS scores. INTRODUCTION There are at least two reasons why it might be desirable to adjust plan CAHPS scores. In 1995, the Agency for Healthcare First, there are certain processes that one Research and Quality (then called the would expect to vary according to the char­ Agency for Health Care Policy and acteristics of patients. For example, one Research) initiated a cooperative agree­ CAHPS question is “…how much of a prob­ ment with RAND, Harvard, and Research lem did you have finding or understanding Triangle Institute to conduct the CAHPS the information…from your health plan?” study. The goals of the CAHPS project Although it is desirable to communicate included developing a standardized sur vey clearly with all patients, it probably is harder that could be used to assess the experience to do so with patients who have less educa­ of consumers in different types of insur- tion than with other patients. Second, certain personal patient characteristics might influ­ The authors are with the Har vard Medical School. Research for ence the response to questions, even if the this article was supported by the Health Care Financing process of care is the same for all patients. Administration (HCFA) under Contract Number 500-95-0057- TO#9 with the Barents Group of KPMG Consulting, Inc. in af fil­ To develop a case-mix adjustment model iation with Har vard Medical School, the MEDSTAT Group, and Westat. The views expressed in this article are those of the for the MMC-CAHPS data, we first authors and do not necessarily reflect the views of Barents reviewed published studies. Next, we ana­ Group of KPMG Consulting, Inc., Har vard Medical School, the MEDSDAT Group, Wesat, or HCFA. lyzed five data sets from sur veys of health HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 109 maintenance organization (HMO) popula­ Joshi, 1997; Lee and Kasper, 1998). tions in dif ferent par ts of the United States Perceived improvement in health also has a and identified potentially important case- strong positive association with health care mix variables (Cioffi et al., 1998). Finally, ratings (Kane, Maciejewski, and Finch, we analyzed MMC-CAHPS data to evalu­ 1997; Kippen, Strasser, and Joshi, 1997). ate alternative models. We analyzed data The few studies that have investigated from the first MMC-CAHPS sur vey, the relationship between satisfaction rat­ based on CAHPS 1.0, and from the sec­ ings and the presence of specific medical ond and third sur veys, based on CAHPS conditions have yielded inconsistent 2.0 (Clear y et al., 2000). results (Hall et al., 1990; Zapka et al., 1995; Kippen, Strasser, and Joshi, 1997). PREVIOUS RESEARCH Emotional distress and social-activity limi­ tations have been found to be negatively Given that there are few published stud­ associated with satisfaction ratings, ies of factors affecting health plan ratings, although work limitations or other limita­ we reviewed studies of hospital care, ambu­ tions due to emotional health status have latory medical services, and health plans. not (Marshall, Hays, and Mazel, 1996; Patient characteristics that have been iden­ Greenley, Young, and Schoenherr, 1982). tified as correlates of patient reports about Early investigations revealed that older their health care include (1) patient sociode­ patients are generally more satisfied than mographic characteristics, (2) overall per­ younger patients with their medical care ceived health status, (3) functional status, (Cleary and McNeil, 1988; Aharony and (4) diagnoses or conditions, (5) length of Strasser, 1993; Zapka et al., 1995), although relationship with provider or health plan findings are not consistent (Weiss, 1988; and prior use of services, (6) whether the Kane, Maciejewski, and Finch, 1997), and survey was completed by a proxy, and (7) some studies have found that this association institutional status (Cleary and McNeil, is not present in the oldest groups of patients 1988; Aharony and Strasser, 1993; Weiss, (Hall et al., 1990; Lee and Kasper, 1998). 1988; Hall et al., 1990; Cleary et al., 1992; Although some studies have found that Zapka et al., 1995; Kane, Maciejewski, and females were less satisfied than males Finch, 1997; Kippen, Strasser, and Joshi, (Cleary et al., 1992), the preponderance of 1997; Lee and Kasper, 1998). studies found that gender is not a significant Better self-reported health is consistent­ predictor of satisfaction (Clear y and McNeil, ly associated with higher ratings of health 1988; Aharony and Strasser, 1993; Weiss, care services by consumers and patients 1988; Hall et al., 1990; Zapka et al., 1995; (Cleary and McNeil, 1988; Aharony and Kane, Maciejewski, and Finch, 1997; Lee Strasser, 1993; Marshall, Hays, and Mazel, and Kasper, 1998). Studies of the association 1996; Hall, Milburn, and Epstein, 1993; between respondents’ race and ratings of Roberts, Pascoe, and Attkisson, 1983; Hall their medical care and health insurance et al., 1998). Current general health status plans have had inconsistent results (Weiss, tends to be the strongest predictor of 1988; Zapka et al., 1995; Kane, Maciejewski, patient or consumer satisfaction with health and Finch, 1997; Kippen, Strasser, and Joshi, care ser vices (Hall et al., 1990; Clear y et al., 1997). Evidence about the association 1992; Zapka et al., 1995; Kane, Maciejewski, between education levels and consumer rat­ and Finch, 1997; Kippen, Strasser, and ings of medical care and health plans is also 110 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 inconsistent (Weiss, 1988; Zapka et al., 1995; et al., 1989; Rothman et al., 1991; Magaziner Kane, Maciejewski, and Finch, 1997; et al., 1988; Sprangers and Aaronson, 1992; Kippen, Strasser, and Joshi, 1997). Magaziner et al., 1996; Rubenstein et al., 1984; Most investigations of the relationship Magaziner et al., 1997). Reports about the between income and consumer reports patient’s general health status tend to have about health care find that those with high­ the lowest levels of concordance (Magaziner er incomes provide modestly higher or et al., 1988). Some studies have found that similar ratings, compared with those with the association between proxy and patient lower incomes (Weiss, 1988; Hall et al., reports increased with higher education lev­ 1990; Clear y et al., 1992; Zapka et al., 1995; els for either respondent (Sprangers and Kane, Maciejewski, and Finch, 1997). A Aaronson, 1992; Hays et al., 1995) and with study of Medicare beneficiaries indicated increased contact between patient and proxy that more income is associated with higher respondent (Epstein et al., 1989; Sprangers levels of health care satisfaction (Lee and and Aaronson, 1992; Rubenstein et al., 1984). Kasper, 1998). An increased burden of caring for the patient Possession of additional health insur­ is associated with more negative ratings by ance coverage such as Medicaid or private proxies relative to those given by patients supplemental policies may be an impor tant (Epstein et al., 1989; Magaziner et al., 1988). predictor of health plan satisfaction. One Reports about more subjective aspects of study found that Medicare fee-for-service health appear to be influenced much more by beneficiaries with secondary insurance the proxy’s own level of psychological dis­ were more likely than others to be highly tress, age, and health status than by charac­ satisfied with their medical care (Lee and teristics of the patient (Rothman et al., 1991). Kasper, 1998). Groups with both Medicaid Few studies have investigated the effect and Medicare coverage were more likely of proxy ratings on reports about medical to be highly satisfied with their health care care and health plans (Epstein et al., 1989; than those with Medicare only. Lavizzo-Mourey, Zinn, and Taylor, 1992). Older individuals are often cognitively or One study comparing responses of 60 elder­ physically impaired and may be unable to ly patients with their proxies’ satisfaction complete a survey or unavailable to respond ratings found the association to be modest (Corder, Woodbury, and Manton, 1996). As a (Epstein et al., 1989). Proxies consistently result, sur veys are often completed by a close rated the patient’s care more negatively relative or caregiver. Proxies tend to rate the than did patients themselves. The majority patient’s health status lower than the patients (62 percent) of proxies in this study were rate themselves (Epstein et al., 1989; spouses. Another study found modest asso­ Rothman et al., 1991; Magaziner et al., 1988; ciations between proxy and patient ratings. Sprangers and Aaronson, 1992; Magaziner et However, in that study, surrogates general­ al., 1996; Rubenstein et al., 1984; Magaziner et ly rated the patient’s care more positively al., 1997). Subjective health-status dimen­ than did patients themselves (Lavizzo- sions tend to be more greatly underrated by Mourey, Zinn, and Taylor, 1992). proxies. Reports about observable physical or functional characteristics, such as the abil­ PRELIMINARY DATA AND ANALYSES ity to walk upstairs or dress, are much more consistent between patient and proxy than The data sets we used for preliminary less concrete or more private health dimen­ analyses were chosen, in part, to comple­ sions, such as emotional well-being (Epstein ment the existing literature. Only one pub- HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 111 lished study investigated the relationship for commercially insured groups. We limit­ between consumers’ ratings and sociode­ ed our analyses to data from Medicare ben­ mographic characteristics specifically for eficiaries (862 responses). Medicare beneficiaries. We selected data The NCQA data presented us with two sets that included recent consumer evalua­ unique opportunities to study the relation- tions of their health insurance plans, in ship between self-reported health-status addition to ratings of their medical care measures and health care ratings for (Table 1). Three of the five data sets Medicare beneficiaries. First, the data included a substantial number of Medicare included the Medical Outcomes Study beneficiaries from diverse geographic (MOS) Short Form 12 (SF-12) health sur­ regions of the United States. Two of the vey (Ware, Kosinski, and Keller, 1996), data sets (Washington and private employ­ which collects data about physical func­ er) included many questions that matched tioning and emotional well-being. items on the 1997 MMC-CAHPS . Each Additionally, the NCQA survey included data set included information on age, sex, unusually complete data (26 questions) education, general health status, and race. about chronic or disabling conditions that Information about chronic or disabling respondents had. conditions, physical functioning, emotional well-being, income, and proxy responses Medicare Current Beneficiar y Sur vey was included in many of the data sets. (MCBS) Minnesota Data The MCBS is administered each year to measure the economic and quality of life In 1995, the Minnesota Health Data effects of the Medicare program on Institute (MHDI) conducted a study of all enrollees. Each year approximately 16,000 health plans in Minnesota for commercially Medicare beneficiaries or their proxies are insured groups, Medicare, and medical inter viewed, of whom about 4,000 are new assistance programs. The study included to the panel. Although the MCBS focuses traditional indemnity and managed care on economic aspects of the enrollees’ expe­ plans. More than 17,000 surveys were col­ rience with Medicare, it also includes a lected for 46 different health plans. The number of questions about their experi­ MHDI used a survey that was similar to ences with their doctors and medical care. the National Committee on Quality The MCBS is unique in that it collects sat­ Assurance’s (NCQA’s) Annual Member isfaction information for a nationally repre­ Health Care Sur vey (AMHCS) (NCQA 1.0). sentative sample of Medicare beneficiaries. We used the 1996 MCBS, which includes a NCQA Sur vey supplemental sample of beneficiaries in managed care organizations. In 1996, NCQA required that health plans submitting results for the Health Plan Private Employer Sur vey Employer Data and Information Set (HEDIS ) report data for the AMHCS. A In fall 1998, a large private employer spon­ total of 43 health plans submitted results for sored an evaluation of the plans offered to more than 18,000 completed sur veys. The employees and retirees, using the CAHPS majority of data compiled by the NCQA was adult core survey. Survey responses were 112 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 113 Table 1 Data Set Components for Preliminary Analyses Project Minnesota NCQA Washington Private Employer MCBS ® ® Survey Used NCQA Annual NCQA Annual CAHPS Adult Core CAHPS Adult Core 1996 MCBS Member Health Care Member Health Care Version 1.0 Version 1.0 with Survey Hybrid Survey SF-12 Attached Method of Data Collection Mail Not Available Mail Mail Personal Interview Percent Responding Not Available 72 52 61 (Salary) Not Available 45 (Hourly) Dates of data collection April-June 1995 1996 Summer 1997 Fall 1997 1996 Number of Health Plans 46 14 20 11 Not Available Number of Respondents 17,591 862 8,310 4,678 16,411 Medicare Yes Yes No Yes, not able to identify. Yes Sample Description Survey of A mix of enrollees State employees. Large manufacturing Rotating panel commercial, ranging from chil­ company employees design to provide Medicare, and dren to Medicare and their dependents. information about Medicaid plan in clients. No plans Stratified by hourly and Medicare bene- the State to provide were exclusively salary positions. ficiaries. 1996 information to Medicare. survey included a consumers. supplemental sample of enrollees in managed care. Demographics Percent in Managed Care Not Available Not Available 92.9 82.2 22.7 Age Percent 65 or Over 97.9 94.8 1.6 21.0 85.5 Percent 18-64 2.1 5.2 98.4 79.0 14.5 Percent Male 34.0 27.0 39.8 72.6 44.1 Education Percent with Some High School or Less 27.7 26.7 1.2 9.0 39.8 Percent High School Graduate 35.4 31.7 16.4 29.0 33.0 Percent with Some College or More 36.9 41.6 82.2 62.0 27.2 Percent White Not Available 92.0 86.5 85.0 79.7 Limited to Medicare enrollees. At least part of the data. Shown for Medicare sample. NOTES: NCQA is National Committee on Quality Assurance. MCBS is Medicare Current Beneficiary Survey. CAHPS is Consumer Assessment of Health Plans. SOURCE: Data from the CAHPS , Agency for Healthcare Research and Quality; data analysis by the authors. collected from 4,678 current employees and analyses in this article. A more detailed retirees in 11 different health plans. Both description of the analyses and results is managed care and traditional indemnity available from the authors (Cioffi et al., plans were included in the evaluation. 1998). These data were unique because they included SF-12 health-status measures as Self-Reported Health Status well as CAHPS survey data. These data allowed us to explore the relationships In each of the preliminar y data sets, cur- between self-reported health measures rent general health status was the and consumer ratings that were exactly the strongest predictor of health care and same as those collected with the 1997 health plan satisfaction for both commer­ MMC-CAHPS . cially insured and Medicare enrollees. Individuals who rated their general health Washington State levels higher also gave higher ratings of their health care ser vices. An evaluation of 20 health plans offered A general health-status variable was to State employees by the Washington analyzed as a continuous or categorical State Health Care Authority obtained 8,310 variable in the MCBS, Washington, private responses. The study was conducted in employer, and NCQA data sets. The con­ summer 1997 using the CAHPS adult tinuous variable accounted for the same core survey, which includes a number of amount of variation as the categorical vari­ items that matched those in the 1997 able, because ratings of medical care and MMC-CAHPS . of health plans improved by about the same amount for each step on the general ANALYSES health-status response scale. We estimated linear models in which the Physical Functioning, Comorbidities, dependent variable is the response on a and Chronic Conditions survey item or set of items (composite) and the independent variables are case-mix Our preliminar y analyses generally indi­ adjusters. In the data sets in which sample cated that measures of physical-function­ size was adequate and the data distin­ ing limitations were not significant inde­ guished among multiple plans, we includ­ pendent predictors of care ratings. ed dummy variables for each of the plans. Analyses of data from Medicare beneficia­ When we control for plan ef fects, the case- ries in the NCQA data set also indicated mix coefficients represent within-plan that physical functioning was not a signifi­ ef fects of the adjuster variables. We tested cant predictor, after controlling for emo­ the predictive power of variables individu­ tional status, general health status, and ally and in combination. age. Analysis of the private employer data, which consists of retirees and current Results of Preliminar y Analyses employees, also revealed that work or activity limitations, physical functioning Because of the large number of analyses limitations, and limitations due to pain involved, we do not present details of the were infrequently related to health care empirical results from the preliminary ratings after controlling for age, general 114 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 health status, and emotional well-being. In pendent predictor of satisfaction levels, the Washington State data, a variable indi­ because it was strongly correlated with cating whether the survey respondent feeling calm and peaceful. needed help with personal care or meeting routine needs, or had a condition that inter­ Age fered with his or her independence, with work, or with school activities, was not a Older adults are generally more satisfied significant predictor for the respondent’s with their medical care and health plan ser­ ratings of overall medical care, specialists, vices. However, ratings do not increase or personal doctors. Having a limitation monotonically with age for those over age was significantly associated only with the 65. Among Medicare beneficiaries in respondent’s overall rating of health plan. Minnesota, older individuals are less satis­ We studied the association between fied than younger ones with their health medical care and health plan ratings, and plan, medical care, and access to care. having a chronic or disabling condition, There were no significant differences using the MCBS, NCQA, and private among age groups for appointment access employer data sets. A variable constructed and physician choice. from the NCQA data set indicated the num­ Results from Medicare enrollees in the ber of conditions (from a list of 26) an indi­ NCQA data file indicated that satisfaction vidual reported having. For the private increases with age until the 80-84 or 85-89 employer data set, we analyzed a variable age groupings, at which point it levels off that indicated whether respondents had a or declines. Evidence from the MCBS sug­ medical condition that had lasted for 3 or gests that satisfaction decreases with age, more months. Neither chronic nor dis­ with most of this effect resulting from abling conditions were significant predic­ lower satisfaction among the oldest tors of satisfaction outcomes in these data respondents (those over age 85). sets. In the MCBS, however, individuals There is little evidence that age affects who reported having any of four physical the relationship between ratings and the conditions tended to provide higher health other sociodemographic characteristics. care satisfaction ratings. No significant interaction effect was dis­ covered between age and health status. Emotional Well-Being The effects of education, income, and Hispanic background were related to age In the NCQA analyses, a general mea­ for those 85 years and over, but these inter- sure of emotional well-being was a signifi­ actions were inconsistent and only margin- cant predictor of health care ratings. ally significant. Feeling calm and peaceful was the most important emotional-status predictor of Sex higher levels of satisfaction in the private employer data analysis. Respondents with Sex was not significantly related to fewer work and social limitations due to health plan or medical care ratings. emotional distress gave higher satisfaction Analyses of employed adults and Medicare ratings, but these results were inconsis­ beneficiaries indicated that females were tent. Feeling energized was not an inde more satisfied than males, but the effects were only marginally significant. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 115 Race Income The relationships between race and rat­ Only the 1996 MCBS data set included ings of health plan or medical care were an income variable that allowed us to study inconsistent in the preliminary data sets. the relationship between the income levels Among retirees and current employees in of Medicare beneficiaries and ratings of the private employer data, race was a signifi­ medical care. Our results were consistent cant predictor only for ratings of specialists with findings from a study showing that but not for health plans, personal doctors, or elderly respondents with higher income overall health care. Black persons, Asians, levels tended to rate their medical care bet­ and Pacific Islanders tended to be more sat­ ter than other respondents (Lee and isfied than white persons in this study. Kasper, 1998). Increases in satisfaction at (Throughout this discussion, the term “white higher income levels were generally mod­ persons” refers to white people who are not est. Ef fects of secondar y sources of health Hispanic.) In the Washington State data, insurance coverage were not assessed in race had a significant relationship only with any of the preliminar y data sets. the specialist ratings: Hispanic persons and Asians were less satisfied than white people. Proxy Respondents Among Medicare beneficiaries in the NCQA data set, Hispanic people tended to We were able to assess the impact of give lower ratings than white persons of their response by a proxy only for the MCBS health plans and of getting approvals or refer­ data set. Proxy ratings of medical care are rals for care. No other differences were sig­ only marginally higher than those provid­ nificant. Analyses of the MCBS data set ed by the intended Medicare respondents. revealed somewhat different results. Black people tended to be less satisfied with both Context Variables their doctors and medical care than white people. Hispanic persons were more satisfied We studied several variables describing than white people with their doctors, but less the social context or community in which satisfied than white people with their care. the respondent lives, using ZIP Code level 1990 U.S. census data. We considered Education seven variables, each of which is measured as percentage of residents in the respon­ More educated Medicare beneficiaries dent’s ZIP Code who belong to the respec­ in the Minnesota and MCBS data sets tive group: Ethnicity (Black, Asian, rated their care higher than those who Hispanic), College-Educated, High-Status were less educated. There was no rela­ Occupation, Urban Resident, Public tionship, however, among Medicare Assistance Recipients (overall and among enrollees in the NCQA data set. In the pri­ those over the age of 65). vate employer and Washington data, those The NCQA analysis revealed that indi­ with more education tended to be less sat­ viduals who live in areas with a high per­ isfied with their medical care and health centage of Asians were more likely to insurance plan. An ordinal education vari­ report satisfaction with their plan and with able predicted ratings as well as a set of ability to get referrals. Respondents from categorical variables, because satisfaction areas where there is a high concentration levels changed roughly linearly. of black residents and from densely 116 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 ® populated urban areas report fewer prob­ MMC-CAHPS Data lems with their health care. These respon­ dents report greater satisfaction with their Instrument plan overall and also with the quality of care received. In the MCBS data, resi­ The MMC-CAHPS survey, fielded in dents of areas with large concentrations of 1997, included all items of the CAHPS 1.0 college-educated individuals and residents adult core instrument (Hays et al., 1999) of areas with many persons with high-sta­ and 28 additional MMC-specific items tus occupations tended to be more satisfied (Clear y, Zaslavsky, and Ciof fi, 2000). Of 85 with their doctors and rated their overall items, 4 elicit overall ratings and 34 elicit care higher. reports of respondent experiences. Other In the Washington State analysis, resi­ questions are used to determine the applic­ dents of more urbanized areas were more ability of par ticular repor t questions or ask satisfied with doctors, specialists, and about sociodemographic characteristics, health plans. Respondents from areas with health status, and health care utilization. high percentages of college graduates or The MMC-CAHPS sur vey fielded in 1998 persons with high-status occupations tend­ and 1999 included all items of the CAHPS ed to provide higher ratings for both their 2.0 adult core instrument and 41 additional doctors and their overall health care. MMC-specific items. The potential case- However, when we include both of these mix variables in the MMC-CAHPS ques­ highly correlated variables in our model, tionnaire are available from the author, only education had a significant positive including 10 variables from the sur vey and effect on the CAHPS scores (an effect 6 variables based on respondents’ ZIP that is opposite to that of individual level Codes. education in this par ticular analysis), while occupation has a positive ef fect on satisfac­ Sample tion with doctor. For each MMC-CAHPS survey, HCFA SUMMARY OF PRELIMINARY drew a stratified sample of Medicare bene­ ANALYSIS RESULTS ficiaries who had been enrolled in an eligi­ ble plan. Eligible plans included all health The effect of a few patient characteris­ plans with Medicare contracts in effect on tics, particularly health status and age, are or before January 1 in the year preceding consistent across multiple studies, while the survey and in business for 2 years. others have effects that are either weak or Contracts that covered large areas were inconsistent. Some of the inconsistencies divided into geographically defined repor t­ might be attributable to the diverse set­ ing units. A simple random sample of up to tings and populations studied. In particu­ 600 members was drawn from each plan or lar, population-based studies might con- reporting unit. found case-mix effects with selection of For each survey, we deleted cases sam­ some groups or patients into more or less pled from contracts that had ceased activity, favorable situations. In the next section, had only one beneficiary (two plans in the we report analyses of Medicare CAHPS second year) or had been terminated, and data that compare patients’ reports within beneficiaries that left their plan before the a single reimbursement system and a large survey was administered, as well as number of health plans. deceased and institutionalized beneficiaries. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 117 In the first survey, there were 89,802 unadjusted means (Zaslavsky, 1998). We valid sur veys from 119,267 eligible benefi­ first evaluated explanatory power (EP) ciaries, representing a response rate of 75 using a linear specification and then tested percent. In the second sur vey, there were the improvement in fit by replacing it with 123,000 valid sur veys from 152,144 eligible variables for each response categor y. beneficiaries, representing a response rate We also investigated the possibility that of 81 percent. In the third survey, there the effects of case-mix variables would were 166,072 valid surveys from 202,775 var y by region. We assigned all responses eligible beneficiaries, representing a from each plan to the single region in response rate of 82 percent. which the plan had the largest enrollment, as determined by the Medicare Managed Sur vey Procedures Care Market Penetration for All Medicare Plan Contractors Quarterly State/County/ Survey data collection took place from Plan data file. This allowed us to adjust Februar y to May 1998 for the first sur vey each plan using a single model and facili­ and from September to December for the tated comparisons among plans operating second (1998) and third (1999) surveys. in the same area. For most plans, 70 per- Although there were slight modifications cent or more of their enrolled population in survey protocols, the basic approach was within a single region. Because of the was comparable each year. The survey small number of plans and managed care firm mailed a preliminary notification let­ enrollees in several HCFA regions in the ter, followed by the survey. Non-respon­ first year of CAHPS data examined, we dents were sent a reminder postcard, and if combined regions 5, 7, and 8 for case-mix no sur vey was received, a duplicate sur vey modeling. For consistency across years, was sent. Interviewers contacted respon­ we retained that grouping for each year. dents by telephone to complete missing We tested interactions between region items and to followup for non-response, if a and a linear effect of age, education, and telephone number could be obtained. reported health status, in models predict­ ing the four CAHPS general ratings. To MMC Analyses evaluate whether it was necessary to cre­ ate an interaction term for each age cate­ The statistical criteria for usefulness of a gor y, we assessed alternative models, com­ variable for case-mix adjustment include paring a model in which a regional interac­ both its predictive power in the pooled tion was estimated for each categor y with a within-plan regression model and the model containing a regional interaction degree of between-plan variability in the with the linear age ef fect. Similar analyses variable, relative to its within-plan variabili­ were performed for the education and ty. In the analyses presented here, we health-status interactions. combined information about predictive We analyzed all 3 years of MMC­ power and between-plan variability to CAHPS data. To simplify the presenta­ obtain an overall summar y of the impact of tion, we show only year three results in the the variable on adjustment: the ratio of the tables. Tables containing all 3 years are variance of the adjustments for the new available from the authors. variable to the between-plan variance of the 118 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 MMC Results In Model 1 in Table 2, we first controlled for age and health and then entered each of We examined the associations between the other potential adjuster variables sepa­ adjuster variables and CAHPS ratings and rately to determine their EP. Controlling the explanator y power of potential adjuster for health status and age, individuals with variables to select a final adjustment more education rated their health plans, model. In Table 2, we present two sets of medical care, personal doctors, and spe­ analyses. In Model 1, we present the cialists lower than those with less educa­ explanator y power of variables controlling tion. This relationship was consistent for only for age and health. In Model 2, we all models that were tested, even when all present the explanatory variables after possible predictors were included in the controlling for a set of core variables. model. The between-plan variance for edu­ Later, we discuss the rationale for using cation is large compared with that for age these two approaches and the results. We and health-status variables. It has the discuss both the predictive strength and largest or second largest EP on all of the explanator y power of variables but present four global ratings for all 3 years. For all data only on explanator y power for all vari­ the composites, education had the largest ables. The coef ficients for the core models EP in at least 1 of the years and for Plan are available from the authors. Paper work, it was the most impor tant in all Age and general health status were cho­ 3 years (data not shown). Based on these sen for inclusion in the core model because results, we decided to include education as the literature and preliminar y analyses indi­ part of our base case-mix model. cated that they are consistently the The MMC-CAHPS survey asked strongest predictors of satisfaction and whether beneficiaries received help filling because they were the case-mix adjusters out the survey and what type of help they in the standard model for the core CAHPS received. The two proxy variables were project. In the MMC data, there was a pos­ PROXY (helped with the sur vey in any way) itive relationship between age and the and ANYPROXY (somebody answered the CAHPS ratings, even when other demo- sur vey for the subject). Both variables were graphic, health status, and contextual vari­ significant predictors of most of the ratings ables were entered into the equation. The and composites, but both had small EPs younger group (under 65, essentially all because their contribution to the predictive disabled) and the group age 65 to 69 years power of the model was relatively small, and tended to give the lowest ratings. The frac­ the proportion of individuals receiving help tions in the extreme age groups, under age did not vary much across health plans. 65 and over age 80, varied greatly between Nevertheless, adjusting for proxy respons­ plans. This suggests that age would have es may be important because of common some impact on case-mix adjustment. concerns that the inability of some benefi­ Health status was consistently the ciaries to complete a survey by themselves strongest positive predictor of consumers’ will compromise the validity of the survey ratings for all measures tested. On the other results. Thus, we included this variable in hand, there was less between-plan variance the case-mix model despite its limited for health status than for some other vari­ impact on adjustments of scores. ables. Nonetheless, because of its predictive Model 2 in Table 2 controls for age, power, health status is an important variable health, education, and proxy responses, in the case-mix model (Table 2). and tests the explanator y power of all other HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 119 120 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Table 2 Explanatory Power of Each Potential Adjuster Ratings of Health Plan Medical Care Personal Doctor Specialists Adjusters Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Age 0.431 — 0.241 — 0.144 — 0.091 — Health 0.274 — 0.295 — 0.140 — 0.104 — Education 0.745 — 0.453 — 0.339 — 0.194 — Proxy 0.035 — 0.024 — 0.024 — 0.014 — Proxy Answer 0.003 — 0.003 — 0.004 — 0.002 — Male 0.005 0.002 0.004 0.001 0.005 0.003 0.002 0.001 Race White 0.030 0.007 0.070 0.045 0.207 0.163 0.000 0.000 Black 0.030 0.009 0.076 0.047 0.148 0.107 0.005 0.001 Hispanic 0.014 0.005 0.007 0.004 0.017 0.013 0.002 0.001 Asian 0.012 0.004 0.009 0.006 0.005 0.002 0.016 0.011 Native American 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Other 0.004 0.003 0.002 0.001 0.001 0.002 0.004 0.003 Medical Condition 0.001 0.003 0.011 0.015 0.012 0.015 0.024 0.026 ADLs 0.012 0.005 0.005 0.002 0.002 0.000 0.003 0.001 IADLs 0.022 0.010 0.011 0.005 0.002 0.000 0.004 0.002 Independent 0.026 0.014 0.029 0.020 0.009 0.004 0.011 0.008 ZIP Code Variables Percent College Degree 0.468 0.075 0.279 0.054 0.187 0.025 0.074 0.003 Percent Urban 0.000 0.014 0.042 0.011 0.006 0.029 0.038 0.069 Percent Black 0.051 0.013 0.122 0.072 0.171 0.110 0.005 0.000 Percent Hispanic 0.306 0.120 0.209 0.104 0.353 0.195 0.148 0.116 Percent Asian 0.047 0.007 0.040 0.012 0.003 0.023 0.010 0.021 Percent Over Age 65 Receiving Public Assistance 0.162 0.024 0.308 0.158 0.287 0.129 0.050 0.016 NOTES: Explanatory power=(variance of adjustments for a variable)/(variance of plan means). All values are multiplied by 1,000 for legibility. In Model 1, the base model for calculation of age and health explanatory power is a null model (with only plan effects). For all other variables, Model 1 includes age and health, with plan differences absorbed as well. In Model 2, the base model for all variables includes age, health, education, and proxy. ADLs is activities of daily living. IADLs is instrumental activities of daily living. SOURCE: Data from the Consumer Assessment of Health Plans , Agency for Healthcare Research and Quality; data analysis by the authors. ® potential adjusters. These analyses indi­ MMC-CAHPS respondents were asked cated that ZIP Code Hispanic, public assis­ three questions about having a health tance (senior), and/or self-reported Asian problem that (1) caused them to need help race had some marginal explanatory with personal care needs, such as eating, power. ZIP Code education, which dressing, or getting around the house, (2) appeared potentially important in tests caused them to need help with routine with the first base model, was not impor­ needs, such as everyday household tant after controlling for individual educa­ chores, doing necessary business, shop- tion. The influence of each is much less ping, or getting around for other purposes, than that of age, health, and beneficiary and (3) seriously inter fered with their inde­ education, however. Also, the influence of pendence, participation in the community, these variables was not consistent across or quality of life. Two of the physical-func­ all dependent variables or the three sur­ tioning indicators were related to ratings, veys. Of the variables tested, ZIP Code even after controlling for general health Hispanic appeared to be the most impor­ status. Respondents with a physical limita­ tant. The inconsistency in these results tion that interfered with independence, and the age of the census data on which participation in the community, or quality these ZIP Code variables are based would of life rated their health plans, medical argue against including these variables in care, specialists, and personal doctors the MMC-CAHPS case-mix adjustment lower. Respondents that needed help with model. However, we examined regional personal care were more likely to give interactions and a model that includes ZIP lower ratings of the health plans and med­ Code Hispanic as a potential model option ical care overall and marginally lower rat­ in subsequent analyses. ings of specialists. Needing help with rou­ Respondents reporting more medical tine needs such as household chores or conditions provided higher ratings of their shopping was not a significant predictor of health plan, medical care overall, special­ ratings due to its high correlation with ists, and personal doctors. This counterin­ needing help with personal care needs. tuitive finding may indicate that it is not the Although physical-functioning indicators mere presence of disease that leads to were significant predictors for the lower satisfaction ratings, but the level of Medicare population, their predictive severity and disabling effect that accompa­ power was modest compared with self- nies the disease. In addition, individuals reported general health status, and they who use health care services more fre­ varied little across plans. Therefore, quently might be more knowledgeable including these variables in the model about their condition and be more likely to would have little effect on the outcomes. report a condition on the survey. Higher Males reported lower ratings than use of ser vices may also indicate increased females of their health plans and personal satisfaction with the services received. doctors in year one and lower scores on all However, the mean number of chronic con­ ratings in years two and three. However, ditions did not var y much across Medicare even when sex was a significant indicator, plans. Therefore, adjustment for the preva­ its predictive power was small and it had lence of medical conditions would not have the smallest variation between health much impact on health plan ratings. plans. Therefore, it had very little impact in the case-mix adjustment. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 121 The relationships between race and health care ratings in year one for respon­ health plan or medical care ratings were dents from areas in which many residents not consistent. Asian Medicare beneficia­ were black or Hispanic. In years two and ries were the only group to consistently three, persons from areas with many rate aspects of their health plans, medical Hispanic residents had higher ratings of care, personal doctors, and specialists plans and doctors, and in year two for care lower than white beneficiaries. Black and as well. Respondents from an urban area Hispanic persons rated their health plans gave slightly lower health care ratings in marginally higher than white persons but years two and three but higher specialist did not differ significantly for ratings of ratings in year two. Although the effects of their medical care overall. Black persons the racial/ethnic and poverty contextual were significantly more positive than white variables are interesting, we are reluctant persons about their personal doctors, while to use them now because the effects are Hispanic people were marginally more pos­ inconsistent and for the same reasons as for itive. Hispanic persons also provided mar­ the individual racial/ethnic variables. ginally lower ratings of their specialists For each of the individual-level variables— compared with white persons. Native age, health status, and education—we cal­ Americans provided marginally lower rat­ culated F-tests that compared the model ings of their health plans than did white with the variable entered in the linear (one- persons. We did not recommend using coefficient) specification to the model with race and/or ethnicity variables in a national the variable entered as a set of dummy vari­ case-mix model because of the lack of con­ ables. For the age and education variables, sistency in their effects. We were also con­ the test clearly rejected the simpler (lin­ cerned that their effects might depend on ear) specification for each of the four rating local associations of cultural and socioeco­ scales (data not shown). The effect of age nomic characteristics with race and ethnic­ showed a clear trend for most levels and ity that might var y from region to region. outcome variables, in which ratings All six of the contextual variables, which increased with age. On the other hand, the describe the ZIP Code area in which a per- steps in mean satisfaction were not equal son lives, had large between-plan differ­ for each increase in age categor y; instead, ences. This is understandable because satisfaction appeared to level off in the these variables represent averages over older categories. Similarly, the trend in the areas, and plans also tend to operate within individual education variable was toward areas. The ratio of within- to between-plan lower satisfaction with more education, but variances for these variables are larger the steps were not equal for each increase than those for almost all of the individual- in education categor y. Therefore, the cate­ level variables. Therefore, the ZIP Code gorical effects were more accurate repre­ variables typically had an impact on case- sentations of age and education effects mix adjustment when they were signifi­ than linear variables. For health status, the cantly related to CAHPS scores. linear trend toward lower ratings with Respondents from areas containing more worse health status (coded by higher num­ educated residents were more likely to pro- bers on the health-status response scale) vide slightly lower ratings for health plans appears to be an adequate description of in all 3 years and for specialists in year one. the relationship. There were marginally positive effects on 122 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Regional Interactions expected if more or less favorably rated plans had been randomly distributed Analyses of interaction effects showed across regions. The regional interaction that there were strong regional interac­ effect is tested against the plan interaction tions for health status, education, and ZIP ef fect. With this test, health status, educa­ Code percent Hispanic response for at tion, and proxy response have significant least four of the nine outcome measures in differences across a region, while the the year three data. Health status had sig­ region-specific ZIP Code Hispanic effect is nificant regional interactions for six of the no longer significant. This suggests that variables and was one of the strongest pre­ there is substantial variation in the dictors of the ratings. To evaluate the sta­ Hispanic coefficient from plan to plan, so bility of these interaction effects across although the average coefficient differs multiple years of CAHPS Medicare analy­ across regions, it does not dif fer more than sis, we also used a model incorporating it would if plans had been randomly data from years two and three of the assigned to regions. CAHPS sur vey. We did not use data from The absolute and relative magnitude of year one because of differences in the for- the regional effects varies substantially mat and questions of the survey in that from year to year. Nevertheless, we sug­ year. Fitting a model with 2 years of data, gest that it is useful to include interaction we allowed for slopes on case-mix terms for health and education by region; adjusters to vary (by including both the the two variables that appeared to have the region-interaction effect and a region-by- most consistent interregional variability. year interaction). We estimated separate models for regional interactions for age, Impact of Case-Mix Adjustment education, health status, proxy response, and ZIP Code percent Hispanic. Each of To assess the effects of adjustment on these models included an additional inter- the ratings of plans, relative to the unad­ action term allowing these regional inter- justed ratings, we compared adjusted rat­ action slopes to var y by year. In all cases, ings with unadjusted ratings, using several we found no evidence of change in the measures of the dif ferences. The results of interaction effects across the years. In par­ the impact analyses were comparable for ticular, for education and health, the ratio the 3 years of data. Considering the ratios of the overall effect to the interaction with of adjustment to unadjusted standard devi­ year was large, indicating that the regional ations for each variable, the largest impact interactions were stable over the 2 years of adjustments is on “getting care you compared (and therefore likely to repre­ need” and the smallest is for “ease of get­ sent consistent patterns rather than ran­ ting referrals.” The standard deviation of dom variations). plan means is only slightly smaller for the We also calculated F-tests of the signifi­ various adjusted means than for the unad­ cance of regional-interaction effects in an justed means. ANOVA model, treating plan effects as the The largest adjustments upward are random error term. This tests whether the comparable to one standard deviation of effect of our case-mix adjusters varies by the plan means for most measures. The region (i.e., an interaction of each case-mix largest adjustments downward are usually adjuster with region) more than would be much smaller, half as big or less. This HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 123 suggests that there are a few plans with SUMMARY AND CONCLUSIONS unusually adverse case mix, from the standpoint of the ef fect of case-mix on con­ Previous studies as well as the analyses sumer assessments. presented here support the continued use Comparison of the ratio of the standard of perceived health status and age in deviation of adjustments to the standard CAHPS case-mix adjustment models. deviation of unadjusted means across Although education does not explain a regions suggests that the impact of adjust­ large proportion of the variance in the ment may be somewhat larger in some dependent variables assessed, there is regions than in others. Generally, the ratio more interplan variability in education than is above average in the Pacific, New in age or health, and as a consequence, England, and Upper Midwest Regions, education predicts more interplan variabil­ below average in New York and New ity than either health status or age in some Jersey, Mid-Atlantic, and South Atlantic, models. and mixed in Northwest and Southwest. Response by a proxy is not an impor tant To quantify the ef fect of case-mix adjust­ predictor of responses, either for individual- ment on the ranking of plans, we calculated level analyses or for assessments of inter- the Kendall Tau correlation coefficient plan variability. We suggest including the between the adjusted and unadjusted plan proxy variable primarily because of con­ ratings. This measure is related to the frac­ cerns about the potential effects of cogni­ tion of pairs of plans that switched ordering tive impairment on reports about plan as a consequence of case-mix adjustment, experiences in this population and the like­ where the denominator is the total number lihood that proxy respondents describe of pairs of plans. (The Kendall Tau statistic experiences with the health plan different­ stretches this quantity to a scale from -1 to ly than enrollees would. Thus, the +1, to make it comparable to other correla­ Medicare adjustment model now includes tion coefficients.) health status, age, education, and a vari­ The Kendall Tau statistics for overall rat­ able indicating whether a proxy answered ing of plans in the 3 years were 0.92, 0.89, the sur vey. We also recommend including 0.91, indicating that the percentages of interaction terms for health and education pairs of plans whose ordering would be by region because they are the two vari­ changed using that adjustment model were ables that appeared to have the most con­ 3.9, 5.5, and 4.5 percent, respectively. sistent interregional variability. Generally, where the ratio of the standard In general, the case-mix adjustments are deviation of the adjustment divided by the not large and do not greatly change the pic­ standard deviation of the adjusted mean is ture of which plans are high- or low-rated. larger, the Kendall Tau is smaller and the It is noteworthy, however, that the largest fraction of pairs that would be switched is adjustments are quite substantial, so there larger. Nonetheless, the unadjusted and are at least a few plans for which, under our adjusted means give between-plan compar­ models, an important part of their mea­ isons that are in agreement, most of the sured satisfaction can be attributed to case time, in ever y region. mix rather than to actual plan per formance. 124 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 Hays, R.D., Shaul, J.A., Williams, V.S.L., et al.: REFERENCES Psychometric Properties of the CAHPS 1.0 Sur vey Measures. Medical Care 37(3) Supp:MS22-MS31, Aharony, L., and Strasser, S.: Patient Satisfaction: What We Know About and What We Still Need to Kane, R.L., Maciejewski, M., and Finch, M.: The Explore. Medical Care Review 50(1):49-79, Spring Relationship of Patient Satisfaction with Care and Clinical Outcomes. Medical Care 35(7):714-730, 1997. Ciof fi, M.J., Clear y, P.D., Ding, L., et al.: Analysis of Kippen, L.S., Strasser, S., and Joshi, M. : Improving Case-Mix Strategies and Recommendations for the Quality of the NCQA Annual Member Medicare Managed Care CAHPS . HCFA Report. Satisfaction Survey Version 1.0. American Journal December 1998. of Managed Care 3 (5):719-730, 1997. Cleary, P.D. and McNeil, B.J.: Patient Satisfaction Lavizzo-Mourey, R., Zinn, J., and Taylor, L.: Ability as an Indicator of Quality Care. Inquiry 25(1):25-36, of Surrogates to Represent Satisfaction of Nursing Spring 1988. Home Residents with Quality of Care. Journal of the Cleary, P.D., Zaslavsky, A.M., and Cioffi, M.: Sex American Geriatrics Society 40(1):39-47, 1992. Differences in Assessments of the Quality of Lee, Y., and Kasper, J.D.: Assessments of Medical Medicare Managed Care. Women’s Health Issues Care by Elderly People: General Satisfaction and 10(2):70-79, 2000. Physician Quality. Health Services Research Cleary, P.D., Edgman-Levitan, S., McMullen, W., 32(6):741-757, Februar y 1998. and Delbanco, T.L.: The Relationship Between Magaziner, J., Itkin, Z.S., Gr uber-Baldini, A.L., et al.: Reported Problems and Patient Summary Proxy Repor ting in Five Areas of Functional Status: Evaluations of Hospital Care. Quality Review Comparison with Self-Reports and Obser vations of Bulletin 18(2):53-59, Februar y 1992. Performance. American Journal of Epidemiology Corder, L.S., Woodbury, M.A., and Manton, K.G.: 146(5):418-428, September 1997. Proxy Response Patterns Among the Aged: Effects Magaziner, J., Simonsick, E.M., Kashner, T.M., and on Estimates of Health Status and Medical Care Hebel, J.R.: Patient-Proxy Response Comparability Utilization from the 1982-1984 Long-Term Care on Measures of Patient Health Status and Sur veys. Journal of Clinical Epidemiology 49(2):173- Functional Status. Journal of Clinical Epidemiology 182, 1996. 41(11):1065-1974, 1988. Epstein, A.M., Hall, J.A., Tognetti, J., et al.: Using Magaziner, J., Pear, B.S., Hebel, J.R., and Gruber- Proxies to Evaluate Quality of Life: Can They Baldini, A.: Use of Proxies to Measure Health and Provide Valid Information About Patients’ Health Functional Status in Epidemiologic Studies of Status and Satisfaction with Medical Care. Medical Community-Dwelling Women Aged 65 Years and Care 27(3):S91-S98, March 1989. Older. American Journal of Epidemiology 143:283- Greenley, J.R., Young, T.B., and Schoenherr, R.A.: 292, 1996. Psychological Distress and Patient Satisfaction. Marshall, G.N.., Hays, R.D., and Mazel, R.: Health Medical Care 20(4):373-385, April 1982. Status and Satisfaction with Health Care: Results Hall, J.A., Feldstein, M., Fretwell, M.D., et al.: from the Medical Outcomes Study. Journal of Older Patients’ Health Status and Satisfaction with Consulting and Clinical Psychology 64(2):380-390, Medical Care in an HMO Population. Medical Care 28(3):261-269, March 1990. Roberts, R.E., Pascoe, G.C., and Attkisson, C.C.: Hall, J.A., Roter, D.L., Milburn, M.A., and Daltroy, Relationship of Service Satisfaction to Life L.H.: Why Are Sicker Patients Less Satisfied with Satisfaction and Perceived Well-Being. Evaluation Their Medical Care? Tests of Two Explanatory and Program Planning 6(3):373-383, 1983. Models. Health Psychology 17(1):70-75, 1998. Rothman, M.L., Hedrick, S.C., Bulcroft, K.A., et al.: Hall, J.A., Milburn, M.A., and Epstein, A.M.: A The Validity of Proxy Generated Scores as Causal Model of Health Status and Satisfaction with Measures of Patient Health Status. Medical Care Medical Care. Medical Care 31(1):84-94, January 29(2):115-124, Februar y 1991. Rubenstien, L.Z., Schairer, C., Wieland, D., and Hays, R.D., Vickrey, B.G., Perrine, H.K., et al.: Kane, R.: Systematic Biases in Functional Status Agreement Between Self-Reports and Proxy Assessment of Elderly Adults: Effects of Different Reports of Quality of Life in Epilepsy Patients. Data Sources. Journal of Gerontology 39(6):686- Quality of Life Research 4(2):159-168, 1995. 691, 1984. HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3 125 Sprangers, M.A.G., and Aaronson, N.K.: The Role Zapka, J.G., Palmer, R.H., Hargraves, J.L., et al.: of Health Care Providers and Significant Others in Relationships of Patient Satisfaction with Evaluating the Quality of Life of Patients with Experience of System Performance and Health Chronic Disease: A Review. Journal of Clinical Status. Journal of Ambulatory Care Management Epidemiology 45(7):743-760, 1992. 18(1):73-83, 1995. Ware, J.E., Kosinski, M., and Keller, S.D.: A 12-Item Zaslavsky, A.M.: Issues in Case-Mix Adjustment of Short-Form Health Sur vey: Construction of Scales Measures of the Quality of Health Plans. and Preliminary Tests of Reliability and Validity. Proceedings, Government and Social Statistics Medical Care 34(3):220-233, 1996. Sections. American Statistical Association. Alexandria, VA. 1998. Weinberger, M., (ed.): Consumer Assessment of Health Plans Study (CAHPS ). Medical Care Reprint Requests: Paul D. Clear y, Ph.D., Department of Health 37(3)Supp, 1999. Care Policy, 180 Longwood Avenue, Harvard Medical School, Weiss, G.L.: Patient Satisfaction with Primary Boston, MA 02115-5899. E-mail: clear [email protected] vard.edu Medical Care: Evaluation of Sociodemographic and Predispositional Factors. Medical Care 26(4):383- 392, April 1988. 126 HEALTH CARE FINANCING REVIEW/Spring 2001/Volume 22, Number 3

Journal

Health Care Financing Review – Pubmed Central

Published: Sep 1, 167

There are no references for this article.