The Relationship Between Managed Care Insurance and Use of Lower-Mortality Hospitals for CABG Surgery FREE

Managed care health plans offer a package of health care benefits to their enrollees on a prepaid basis, assuming financial risk for the covered services. Plans manage this risk in many ways. For instance, they may selectively contract with a restricted set of clinicians and hospitals. For acute care, prices and other contract terms are negotiated in advance, and health plans may require patients to use hospitals under contract with the plan or to pay more if they choose a hospital outside the health plan network.

Financial risk provides a strong incentive for health plans to select low-priced hospitals. However, health plans should also consider quality of care when contracting with hospitals, especially if explicit data on quality are available. Health plans that ignore publicly available quality data are at risk of adverse publicity about their contracting decisions. A recent study by Escarce et al¹ found that managed care patients in California were more likely than insured non–managed care patients to use hospitals with lower-than-expected mortality rates for coronary artery bypass graft (CABG) surgery, but they found no difference between these 2 types of patient groups in Florida. Data on CABG surgery mortality were not publicly available in either setting, making it unlikely that health plans explicitly considered mortality rates in their CABG surgery contracting decisions.

In contrast, New York was the first and one of the few states to provide information on CABG surgery, including risk-adjusted mortality rates, to the public,² allowing health plans, at least in theory, to explicitly consider quality of care when contracting with hospitals to provide CABG surgery. To explore the relationship between insurance type and patterns of hospital use, we compared the probabilities that patients with managed care insurance and patients with fee-for-service (FFS) insurance would undergo CABG surgery at lower-mortality hospitals in New York State, where quality-of-care information is available.

We examined CABG surgery discharges in New York State for the years 1993 through 1996, using records from New York State annual hospital discharge databases.^{3 - 6} These legislatively mandated databases included patient age, race, ethnicity, insurance type, ZIP code, hospital of admission, and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes for all patients aged 18 years or older who underwent CABG surgery in these years. We selected New York State residents who underwent CABG surgery (ICD-9-CM codes 36.1, 36.10-36.17) at 1 of 31 CABG surgery hospitals in New York State and had 1 of the following primary health insurance types: private FFS, private managed care, Medicare FFS, or Medicare managed care. Single codes represented each insurance type except private FFS, which included codes 06 (Blue Cross), 08 (commercial insurance company), and 15 (self-insured company). Private managed care was defined as code 11 (health maintenance organization); no codes existed for preferred provider organizations or other variations of private managed insurance. We excluded patients who underwent CABG surgery combined with any valve procedure or left ventricular aneurysm resection. Because they represent such a heterogeneous group, patients younger than 65 years who were enrolled in either Medicare FFS or Medicare managed care programs were also excluded.

Rather than ranking hospitals separately for each year, we ranked hospitals into lower- and higher-mortality groups based on volume-weighted average adjusted mortality rates to simplify the analysis and improve interpretability of results. These rates, published by the New York Department of Health,^{7 - 10} are adjusted for multiple demographic and medical risk factors, including age, sex (for years 1995-1996 only), body surface area, hemodynamic status, medical comorbidities, severity of atherosclerotic process, measures of ventricular function, and history of open-heart surgery.

Based on these averages, we divided hospitals into lower- and higher-mortality groups using a mortality rate cut point that allocated about half of the patients to each group. The lower-mortality hospital group included 14 hospitals, constituting 49.8% of the total number of patients, with a mean adjusted mortality rate of 2.1% (range, 1.2%-2.5%) and a mean annual case volume of 561 (range, 189-1528). The higher-mortality hospital group included 17 hospitals, with a mean adjusted mortality rate of 3.2% (range, 2.5%-5.1%) and a mean annual case volume of 466 (range, 63-1019).

This method of ranking hospitals appeared to be fairly stable from year to year. The published hospital adjusted mortality rates averaged for the first 2 years of the study correlated highly with those for the last 2 years (Pearson r=0.53; P=.002). Rankings derived from these rates correlated significantly as well (Spearman r=0.37; P=.04). Thus, at least in theory, insurers could use this type of rating to predict hospital mortality rates in future years.

The outcome of interest was the probability that a patient underwent a CABG operation at 1 of the lower-mortality hospitals. The independent variable of interest was insurance type. In addition, models controlled for age, sex, urgent/emergent admission, and the presence of medical comorbidities, including diabetes, chronic renal failure, congestive heart failure, systemic hypertension, and chronic obstructive pulmonary disease. Age was represented as a categorical variable with 4 values: younger than 55 years, 55 to 64 years, 65 to 75 years, and older than 75 years. Other independent variables, determined using the 1990 US census,¹¹ were median family income for the patient's ZIP code of residence and designation of the patient's residential area as urban. Median family income for ZIP code was represented as a categorical variable in quartiles.

Because the location of a patient's residence may be correlated with both insurance status and the choice of hospital,^{12 - 15} our analysis also used 2 factors to account for the impact of distance on the use of lower-mortality centers. These factors were the distance to the nearest lower-mortality center and the distance to the nearest higher-mortality center. Both were based on the distances from the geographic centroids¹⁶ of the hospital and patient residential ZIP codes because street addresses were not included in the available data. Such "straight-line" methods of estimating geographic accessibility through geographic longitude and latitude data have been shown to correlate closely with estimated travel times¹⁷ using actual driving routes.¹⁴ These distance factors were log-transformed prior to analysis based on the observation in our data set and in other similar studies¹⁸ that such transformation resulted in a closer fit to the relationship between distance and use of a lower-mortality hospital than did actual distance and several other transformations investigated.

We identified univariate predictors of use of lower-mortality centers using 2 × 2 χ² tests for categorical variables. In the case of categorical variables with more than 2 possible values, each value was compared with a reference value (eg, the youngest age group). The Wilcoxon rank sum test was used to determine differences among insurance types for all continuous variables, none of which were normally distributed.

We then used a stepwise logistic regression model to estimate the multivariate-adjusted odds of use of lower-mortality centers, considering each of the independent variables listed herein, including the 2 distance factors. Because the rate of the outcome (proportion of patients in lower-mortality hospitals) exceeded 10%, we transformed odds ratios (ORs) to approximate relative risks (RRs) according to the method described by Zhang and Yu.¹⁹

Two hospitals did not report patient race or ethnicity for the study period, so these variables were not entered into the models. However, we did investigate the relationships among insurance type, hospital used, and race/ethnicity by repeating the analysis separately for patients coded as non-Hispanic white and those coded as nonwhite or Hispanic. We also performed supplementary analyses by stratifying patients according to age, urgency of admission, and residence in New York City and by varying the threshold for inclusion in the lower-mortality hospital group. In addition, we examined the distribution of patients among individual hospitals in 4 delimited regions of the state, including Buffalo, Rochester, Long Island, and New York City.

Among 58,902 adults admitted for CABG surgery between 1993 and 1996, most patients were male (71%), non-Hispanic white (87%), and Medicare FFS beneficiaries (54%). Total inpatient mortality was 2.8%. Hypertension (52%) and diabetes (27%) were the most commonly coded comorbidities. Urgent/emergent admissions constituted 51% of all cases. Table 1 summarizes the demographic characteristics according to insurance group.

Unadjusted analyses (Table 2) revealed that patients with private managed care, Medicare FFS, and Medicare managed care insurance were less likely than patients with private FFS insurance to undergo CABG surgery at lower-mortality centers. Other factors associated with less frequent use of lower-mortality centers included female sex, nonwhite race/ethnicity, lower median family income for ZIP code, urgent/emergent admission, diabetes, and chronic renal failure. In contrast, patients with congestive heart failure, hypertension, and chronic obstructive pulmonary disease were more likely to use lower-mortality hospitals. On average, patients undergoing surgery at a lower-mortality hospital lived closer to a lower-mortality center (15.4 vs 24.6 miles; P<.001) and patients undergoing surgery at a higher-mortality hospital lived closer to a higher-mortality center (38.9 vs 13.9 miles; P<.001).

Table 2 also shows coefficients and 95% confidence intervals (CIs) for categorical covariates in the final multivariate model. Also included in the final model were the 2 distance variables, discussed herein. Compared with patients with private FFS insurance, patients with private managed care insurance were less likely to undergo CABG surgery at a lower-mortality hospital (RR, 0.77; 95% CI, 0.74-0.81; P<.001), as were patients with Medicare managed care insurance (RR, 0.61; 95% CI, 0.54-0.70; P<.001). After controlling for other factors, patients with Medicare FFS insurance used lower-mortality hospitals at a rate similar to those with private FFS insurance (RR, 0.95; 95% CI, 0.91-0.98; P=.004). As in the univariate analysis, patients who were female, lived in lower ZIP code median family income groups, were admitted urgently or emergently, or had a code recorded for either diabetes or chronic renal failure were less likely to be admitted to a lower-mortality center, although the magnitude of the difference was small in some cases. In contrast, those older than 75 years and those who had a code recorded for heart failure, hypertension, or chronic obstructive pulmonary disease had a greater probability of admission to a lower-mortality center. Increasing distance from the nearest lower-mortality center reduced the odds of admission to a lower-mortality hospital (for a 0.1-log increase in distance, OR, 0.76; 95% CI, 0.76-0.77; P<.001), but increasing distance to the nearest higher-mortality hospital increased the odds of admission to a lower-mortality center (for a 0.1-log increase in distance, OR, 1.45; 95% CI, 1.44-1.46; P<.001).

When we restricted the sample to patients living within 25 miles of both a lower-mortality and higher-mortality hospital, the results were similar. Compared with patients with private FFS insurance, patients with private managed care insurance were less likely to undergo CABG surgery at a lower-mortality hospital (RR, 0.62; 95% CI, 0.58-0.66; P<.001), as were patients with Medicare managed care insurance (RR, 0.57; 95% CI, 0.48-0.66; P<.001) and, to a smaller degree, Medicare FFS insurance (RR, 0.92; 95% CI, 0.88-0.96; P<.001).

Table 3 shows the sensitivity of these results to changes in the cut point for inclusion in the lower-mortality group at between 38% and 68%, which was as close as the data would allow to division at the upper and lower tertiles. Regardless of the threshold, patients with private managed care or Medicare managed care insurance were significantly less likely than patients with private FFS insurance to use a lower-mortality center. Medicare FFS estimates of use were slightly lower than those for private FFS at all cut points as well.

In subgroups stratified by region (New York City vs upstate New York and Long Island), admission type (nonurgent vs urgent or emergent), age, and patient race/ethnicity, patients with private managed care and Medicare managed care insurance were significantly less likely to use lower-mortality centers than patients with private FFS insurance (Table 3). Patients with Medicare FFS insurance had rates of use much closer to those with private FFS insurance, with nonsignificant differences in several patient subgroups (New York City, nonurgent admissions, and non-Hispanic white race/ethnicity).

When we examined 4 delimited areas (Buffalo, Rochester, New York City, and Long Island), patients with private managed care and, particularly, Medicare managed care insurance were virtually excluded from many of the hospitals in a given area (Table 4). For example, of 3 hospitals in the Buffalo area, 99% of patients in Medicare managed care plans were admitted to a single (higher-mortality) hospital that otherwise accounted for only 59% of that region's caseload. Those admitted to hospitals in Long Island were restricted to the 3 hospitals with the highest adjusted mortality rates among 5 hospitals. Medicare managed care patients admitted to hospitals in New York City were admitted to 8 of 14 hospitals, and 46% were admitted to a single higher-mortality hospital that otherwise accounted for only 11% of the region's caseload.

In all, 4 (13%) of 31 hospitals admitted no CABG surgery patients with private managed care insurance. Fourteen (45%) admitted no patients with Medicare managed care plans. Restricting the sample to patients admitted to the 17 hospitals that admitted at least 1 patient with Medicare managed care insurance did not significantly change the results of multivariate analysis. Compared with patients with private FFS insurance, the corrected odds of being admitted to a lower-mortality center were still lower for patients with private managed care (RR, 0.66; 95% CI, 0.62-0.71; P<.001) and Medicare managed care (RR, 0.64; 95% CI, 0.56-0.73; P<.001) insurance, but to a lesser extent for Medicare FFS insurance (RR, 0.92; 95% CI, 0.87-0.97; P=.002).

To investigate the possible role of lag-time effects in these results, we repeated the analysis examining only patients admitted in the second half of the study (1995-1996), using public data from the first half of the study (1993-1994) to designate lower-mortality centers. Compared with patients with private FFS insurance, those with private managed care insurance were less likely to use a lower-mortality center (RR, 0.69; 95% CI, 0.65-0.73; P<.001), as were patients with Medicare managed care (RR, 0.49; 95% CI, 0.40-0.58; P<.001) and Medicare FFS (RR, 0.75; 95% CI, 0.69-0.80; P<.001) insurance.

Over the period of the present study, patients in New York State with managed care insurance were significantly less likely to undergo CABG surgery at a hospital with lower CABG mortality compared with patients with FFS insurance. This finding remained significant within categories of age, race, urgency, and region. The findings were not sensitive to the threshold used to define lower-mortality hospitals, nor did they appear to be a manifestation of a differential lag time in the recognition of changing hospital outcomes.

Some insight into the mechanisms of these differences in hospital use patterns can be gained by examining insurance-specific use on a hospital-by-hospital basis. Patients with managed care insurance and, particularly, managed Medicare insurance were often excluded from many lower-mortality hospitals entirely, implicating relatively powerful disincentives, such as use restrictions set by insurance companies, rather than differences in patient or referring physician preferences. Such restrictions could include removing a hospital from a plan's preferred provider list or requiring a significant patient copayment for the use of that hospital.

How do we reconcile our finding that New York State managed health plans appeared to ignore risk-adjusted mortality rates in contracting decisions with the finding that patients in California health plans appeared more likely to use lower-mortality hospitals?¹ The absence of public data on risk-adjusted mortality in California makes it unlikely that mortality rates played a significant, if any, role in health plans' decision making. In a commentary to the study by Escarce et al, Hannan²⁰ suggests that another reason for the difference between California and Florida may be that, because California has no certificate-of-need system, numerous low-volume hospitals with high mortality rates perform CABG surgery. Because low volumes make contracting unattractive, managed care plans in California avoid sending their patients to the highest-mortality hospitals. This indirect effect of CABG surgical volume would not be present in New York, where the certificate-of-need program dictates that all CABG surgery hospitals have high surgical volumes.

Our study was possible because of the New York Hospital Discharge Datasets (SPARCS), which are large, independently administered data sets maintained by the New York Department of Health. Administrative data sets have certain disadvantages relative to prospectively collected clinical data sets, particularly with respect to the consistency and precision with which variables used in risk adjustment are defined and collected.^{21 - 24} With respect to use of lower-mortality hospitals, however, illness severity is less relevant than other potential confounders, such as urgency of the admission, patient socioeconomic status, and the geographic relationship between the residence of the patient and area hospitals. In the present study, we were able to control for each of these variables, except that an ecological proxy, ZIP code median family income, was the only measure of socioeconomic status available. The use of this type of ecological variable has major drawbacks,^{25 - 26} but in the present study it allowed some adjustment for affluence, as it has in other studies on access to medical care.^{27 - 28} Despite the strong relationship of race/ethnicity to socioeconomic status,^{29 - 31} we did not include it as a covariate in our main analysis both because of significant missing data and because of limitations on the interpretation of racial and ethnic information.^{32 - 34} Nonetheless, in a supplementary analysis that stratified by race/ethnicity, the effects of insurance type on the use of lower-mortality centers was generally consistent with those noted in the nonstratified model.

It must be noted that the results of the present study may not generalize beyond New York State. Because the nature of managed care, the availability of hospital quality information, and the types of hospitals performing CABG surgery vary substantially from state to state, these relationships could differ in other parts of the country.

Despite these limitations, our findings suggest that explicit monitoring of the process and outcomes of care could play an important role in identifying problems with the quality of care for managed care enrollees. The role of managed care organizations in the reduced use of lower-mortality centers among their beneficiaries is likely to be complex and multifactorial. Plans may enter into relationships with hospitals largely on the basis of anticipated costs and may create incentives for primary care providers and patients to use lower-cost centers as well. It is also possible that lower-mortality centers may themselves be unwilling to contract with managed care organizations, if they expect better remuneration from other payers. Meanwhile, by limiting patient choices, managed care organizations may prevent patients and their advocates from taking full advantage of available information about hospital quality. This could inadvertently stifle incentives for hospitals to compete on the quality of care. Additional studies on the impact of quality information on health plans' contracting decisions will be important as price competition among health plans becomes more intense.