Nair, Kavita V. PhD; Park, Jinhee CPhil, MA, MS; Wolfe, Pamela MS; Saseen, Joseph J. PharmD, FCCP, BCPS; Allen, Richard Read MS; Ganguly, Rahul PhD
In response to the rising costs of health care, an environment of ever-increasing health care standards and an emphasis on patient empowerment, a new form of health insurance benefits has emerged, ie, consumer-driven health plans (CDHPs). These plans provide financial incentives for consumers to become more involved in health care purchasing decisions.1 An estimated 5.4 million patients were enrolled in CDHPs in 2007 and the number is expected to grow.2 The typical CDHP is a high deductible plan combined with either a Health Savings Account (HSA) or a Health Reimbursement Account (HRA). The HSA based CDHPs differ from HRA based CDHPs in many ways.3 Primarily, HSAs do not require employer involvement; they can be created and funded by an individual, and are fully portable so that unused funds remain with the beneficiary even if employment changes. HRAs can only be established by employers and they are funded solely through employer funds. Unused HRA funds can usually be rolled over to the next year for individuals to use and is the norm in most HRAs today. In the current market place, an estimated 3 million covered lives have HRAs, whereas HSAs cover less than that number with 1.5 million lives.4
Employers with 1000 or more workers are more likely (12%) than smaller firms (6%) to offer CDHPs2: CDHPs accounted for 17% of new policies in the small business market and 8% in the large business market in 2007.2 In both markets, 90% of CDHP members were previously in preferred provider organization (PPO) products.2
The focus of this paper is on HRA based CDHPs. Although plan details vary, the typical HRA consists of three components. The first dollar of coverage for medical services such as outpatient visits, emergency room (ER) visits, hospitalizations, and prescription drugs comes from a personal health spending account (PHA) funded by the employer. Once PHA funds are exhausted, employees are responsible for an out-of-pocket deductible. After this, coinsurance-based health care coverage applies. Preferred medications are usually available at a greater discounted rate than nonpreferred medications. Preventive care is usually covered 100%. Some employers allow their employees to roll over any unused PHA balances to the next year. Consumers also typically have access to advanced Internet-based decision tools for costing services.
As enrolment in CDHPs increases, debate continues about their impact on consumer behavior. Proponents argue that the financial incentives in CDHPs will reduce cost by discouraging unnecessary use of health care resources. However, opponents argue that shifting the costs to consumers through large deductibles could cause them to forgo necessary care,5 and that CDHPs could attract healthier and wealthier individuals, leaving sicker individuals in traditional plans.
Empirical research into the impact of health reimbursement arrangement (HRA)-based CDHPs has produced mixed results.6 Parente et al7 compared individuals who switched to a CDHP offered by a self-insured employer to individuals enrolled in either a PPO- or HMO-based health plan more than a 2-year period (2001–2002). CDHP enrollees had lower total health care expenditures than PPO enrollees but higher expenditures than HMO enrollees in both years. However, CDHP enrollees had the lowest pharmacy expenditures. All groups experienced increases in physician-related expenditures and visits. The most notable finding was that CDHP enrollees had the highest increase in hospital-based expenditures and hospitalizations. In an updated evaluation using an additional year of data (2003), Feldman et al8 also found that increases in hospital expenditures were highest for CDHP enrollees. The authors attributed this to the selection of high-priced hospitals by CDHP enrollees who, in most cases, exceeded their deductible amount during just one hospital stay and therefore faced limited cost sharing.
More recently, Wharam et al9 examined the impact of high deductible plans on ER visits and hospitalizations as a result of ER visits for members who switched from a traditional health maintenance plan to a high deductible plan. Switching to a high deductible plan was associated with a 10% reduction in ER visits, a 27% reduction in hospitalizations associated with ER visits and a 21% reduction in length of stay for hospitalizations.
Buntin et al10 conducted a comprehensive review of CDHPs in 2006 on selection bias, cost savings, and utilization. The authors concluded that there was modest favorable selection bias among early adopters of a CDHP. The authors found that switching from a traditional plan to a CDHP would result in 4% to 15% reduction in health care spending. The impact of CDHPs on utilization was mixed. Some studies that were reviewed found an increase in the use of preventive services and compliance with drug therapy, but one study reported that members were more likely to forgo care for less serious health problems. Finally, some studies found that members adopted cost saving behavior in CDHPs that could result in adverse consequences.
A recent survey by Dixon et al11 of enrollees in a high deductible plan, low deductible plan, and a PPO found that members in the high deductible plan were more likely to report not seeing a doctor when they thought they should have or took a lower dose of a prescription than was recommended compared with PPO members. In a separate study using the same population, Greene et al12 conducted an analysis of pharmacy claims and found that members in high deductible plans were two to three times more likely to discontinue antihypertensives and lipid lowering agents after enrolling in a CDHP compared with members in a PPO plan.
Anecdotal reports in press releases support the notion that CDHPs reduce utilization and costs. CDHPs offered by two health plans Humana and Aetna led to lower medical utilization and spending. Humana13 found that hospitalizations were reduced by 4% and office visits by 1%. Aetna14 saw a 11% reduction in primary care office visits, a 3% decrease in ER visits, a 14% decrease in outpatient visits, and a 5% decrease in hospitalizations.
Despite this growing evidence, what remains to be further explored is the impact of CDHPs on people with chronic diseases. One third of the working-age population has at least one chronic condition and faces higher health care costs.15 For example, in the 2003 Community Tracking Survey, 19% of individuals with chronic diseases reported out-of-pocket health care costs exceeding 5% of their family income. The impact of CDHPs on chronic diseases has been limited. One study compared the use of preventive and screening services for people with diabetes in a CDHP with those in a PPO plan and found no differences in the use of preventive care services such as HbgA1C tests and retinal screening.16 Another study of cancer screening rates found no differences between CDHP and HMO members in the use of breast, cervical, and colorectal cancer screening rates, but CDHP members were less likely to use the more expensive colonoscopy for which they had higher out-of-pocket spending and appeared to substitute this test for the less expensive and fully covered screening tests.17 In a recent study, Hibbard et al18 examined the impact of switching to a high or low deductible from a PPO plan on office visits for those with both chronic and acute diseases. The authors found that outpatient visits for both chronic and acute diseases decreased by 22% compared with PPO members in year 1 of enrollment but this decrease was not as substantial in year 2 of enrollment. However, individuals who enrolled in the high deductible CDHP showed a substantially lower utilization of office visits in both years of enrollment.
In a comparison of prescription drug use in a PPO three tier plan and high and low deductible CDHPs, Greene et al12 conducted an analysis of pharmacy claims of five drug classes: anti-depressants, asthma controllers, anti-ulcerants, anti-hypertensive, and lipid lowering agents. The authors found that members in high deductible plans were two to three times more likely to discontinue antihypertensives and lipid lowering agents after enrolling in a CDHP compared with members in a PPO three tier plan.13 Enrollment in a CDHP, however, did not reduce medication adherence.
We took a broader approach and sought to examine whether individuals with chronic diseases who were enrolled in CDHPs changed their use of outpatient visits, ER visits, and hospitalizations. We also examined whether individuals were more likely to be adherent with medications in a CDHP. Finally, we evaluated whether CDHPs control health care spending better than traditional options.
The employer group was one of the largest transportation carriers in North America, with 28,000 employees and dependents covered. It switched from offering a PPO plan with two options in 2004 to two CDHP options in 2005.
The PPO Plans.
In 2004, the PPO plan had two options, a high and a low deductible. Deductibles ranged from $300 (for employee only) to $900 (for family) for the low-deductible option and $500 (for employee only) to $1500 (for family) for the high-deductible option. After the deductible amounts were met, the plan paid 80% for all in-network covered services including ER visits, hospitalizations, outpatient visits, and laboratory-based services and 60% for all out-of-network services. The prescription benefit had three tiers: $10 copay for generics, the greater of $20 or 20% for preferred brands, and the greater of $30 or 30% for nonpreferred brands. Mail order cost sharing was a flat copay of $15 for generics, $30 for preferred brands or $45 for nonpreferred brands. Total out-of-pocket maximum ranged from $1800 (employee only) to $2500 (for family) for the low deductible option and $5400 (employee only) to $ 7500 (for family) for the high deductible option.
The CDHPs offered in 2005 were both HRA based CDHPs that shared the same features but differed in the amount of the PHA offered by the employer as shown in Table 1. The employer funded the PHAs for both CDHPs. In the more generous CDHP option (referred to as the “higher PHA option”), the employer funded a PHA ranging from $750 to $1500 depending on the employee's status. For example, an individual employee received an annual PHA of $750, for an employee with a child or spouse the PHA was $1125, and for an employee with family the PHA was $1500. The corresponding deductible amounts were $1500 for individual employee, $2250 for an employee with a child or spouse, and $3000 for an employee with a family. Employees were responsible for paying the difference between the PHA and the deductible amount until they exhausted the deductible and this amount was referred to as their “member responsibility.” Therefore, the out-of-pocket costs after their PHA funds exhausted were $750 for an individual employee, $1125 for an employee with a child or spouse, and $1500 for an employee with family.
The less generous CDHP option (referred to as the “lower PHA option”) differed from the higher PHA option in which the member responsibility stayed the same but members had lower amounts of PHA funding. In this option, the employer funded a PHA ranging from $200 to $400 depending on the employee's status. For example, an individual employee received an annual PHA of $200, for an employee with a child or spouse the PHA was $300, and for an employee with family the PHA was $400. The corresponding deductible amounts were $950 for an individual employee, $1425 for an employee with a child or spouse, and $1900 for an employee with a family. Member responsibility in this option was similar to that described under the higher PHA option.
All medical expenses, with the exception of preventive care were subject to the deductible. Once the deductibles were met by the employees, health care coverage was 80% for in-network services and prescriptions and 60% for out-of-network services. The maximum amount that employees would pay after the deductible was met for in-network services ranged from $1050 (employee only) to $2700 (for family) for the higher PHA option and from $1650 (employee only) to $4500 (employee only) for the lower PHA option. The maximum amount that employees would pay after the deductible was met for out-of-network services ranged from $1800 (employee only) to $4200 (employee with family) for the higher PHA option and from $2400 (employee only) to $6000 (employee with family) for the lower PHA option. These out-of-pocket maximum amounts for both CDHP options included the member responsibility and coinsurance amounts that members were responsible for after the deductible has been met.
As we examined a national employer, monthly premium amounts varied by state and by the job description (for example drivers, shop, and administrative employees). However, monthly premium amounts were greater for the higher PHA option compared to the lower PHA option.
Examination of the migration patterns between these options indicated that the majority of individuals (63%) switched from the low-deductible PPO plan to CDHP higher PHA option; one quarter of the individuals (25.5%) switched from the high-deductible PPO plan to CDHP higher PHA option; 8% switched from the high-deductible PPO plan to CDHP lower PHA option; and 3.5% from the low-deductible PPO plan to lower PHA option.
The study sample consisted of individuals and their adult dependents with one or more specific chronic diseases who had a PPO plan option in 2004 and switched to a CDHP option in 2005 and were continuously eligible for 24 months. Pharmacy and medical claims data and eligibility information from the two plans were merged. Individuals with at least one ICD-9 code in the medical claims data for outpatient visits, ER visits, hospitalizations, or other services for allergic rhinitis, asthma, arthritis, diabetes, depression, dyslipidemia, gastroesophageal reflux disease (GERD), or hypertension were selected.
Because the employer offered only CDHPs in 2005, an external comparison group with similar patient characteristics was generated from a large commercially available database with pharmacy and medical claims and enrolment data from more than 35 health plans covering over 40 million American patients.
Patients enrolled in a PPO plan both in 2004 and 2005 were selected from this commercial database to produce the comparison cohort. Propensity analysis was used to create this comparison group to minimize the differences in patient characteristics between the study sample and the comparison group.19 The probability of belonging to the CDHP group was estimated with logistic regression based on age, age squared, gender, a risk score, interaction of age and age squared with hypertension, and the presence of the eight chronic diseases of interest in 2005. Where an exact match for the probability of group membership could not be found, all PPO individuals in the commercial database within a 0.00,001 probability neighborhood of the CDHP individual were identified; the one at the smallest absolute distance was selected and, failing a unique instance of minimum distance, one subject from the PPO comparison group was chosen at random. Once the comparison cohort was defined, variables were compared in the baseline year (2004) to ensure the similarity of the two groups. The groups were not statistically different on any measure characteristic, based on χ2 tests for independent proportions, or, in the case of age and risk score, two-sample t tests.
Medical utilization was measured using the number of disease-specific ER visits, hospitalizations, outpatient visits and laboratory or diagnostic visits for each member with corresponding ICD-9 codes as a primary diagnosis. Expenditure-based measures included the mean medical and pharmacy expenditures per member per year. Pharmacy utilization variables included total number of disease-specific prescriptions and overall medication adherence by disease states. Medication adherence was defined as a binary variable based on the medication possession ratio (MPR).20 MPR is measured from the first to the last prescription, with the denominator being the duration from index to the exhaustion of the last prescription and the numerator being the days supplied over that period from first to last prescription.20 MPR was computed for a individual using drugs within one or more key drug classes used to treat a specific disease, and a weighted average was used to calculate the overall disease-specific MPR. For example, if a hypertensive individual was taking a beta-blocker and a diuretic, the MPR was calculated for their beta-blocker and diuretic use separately and a weighted average used to calculate their overall disease specific MPR for hypertension depending on the total days supply for the beta-blocker and diuretic, respectively. A similar approach was used to compute an overall MPR for each medication an individual was taking. Individuals were considered adherent to their medications if their overall MPR levels for a chronic disease state were greater than 80%.20
The generalized estimating equations method for repeated measures was used for all modeling to accommodate the correlation due to serial observations on each subject.21 With the exception of cost data, equations for all models have the general form:
Equation (Uncited)Image Tools
g, link function
X, the design matrix (a matrix of the explanatory variables)
β, a vector of parameters to be estimated
Equation (Uncited)Image Tools
Y = the outcome measure.
The GENMOD procedure in SAS estimates the parameters of the model numerically through an iterative fitting process. The algorithm for fitting the specified model using generalized estimating equations is not in general a likelihood-based method of estimation. The empirical, or robust, estimator of the covariance matrix of β has the property of being a consistent estimator of the covariance matrix of β, even if the working correlation matrix is misspecified. The within-subject correlation is treated as a nuisance parameter. The estimated parameters are interpretable as marginal or population averaged.
ER visits and hospitalizations were highly skewed with over 80% zero values, so this was transformed to a binary response for analysis by logistic regression: 0, none; 1, one or more. Medication adherence, defined as adherent if the MPR was >80% and not adherent otherwise, was also examined using logistic regression. Number of prescriptions and number of outpatient visits are counting processes and were evaluated with Poisson regression. The number of prescriptions was characterized by more zeroes than might reasonably be expected in a Poisson distribution; thus, we used a zero inflated Poisson model to evaluate this dependent variable.
The cost variables were characterized by some proportion of the population having zero costs and extreme right skew in the non-zero data. The members' differences between pre and post periods, however, had a symmetric distribution and were used for inference on changes in costs, both between and within cohorts. Adjusted differences were estimated by least squares regression of the differences on a subset of independent variables.
Independent variables included age, gender, the risk score represented by the RxRisk score,22 indicators for time period (pre and post period) by group, and indicator variables for each of the eight chronic diseases. All analyses were done using SAS v 9.1 (SAS Institute, Cary, NC) with the exception of the zero inflated Poisson model, which was estimated in STATA (STATA Corporation, College Station, TX).
Demographic Characteristics for Both Groups at Baseline
The demographic characteristics of the primary and comparison cohorts at baseline (2004) are shown in Table 2. A total of 4397 individuals having at least one of the chronic diseases of interest switched from a PPO plan in 2004 to a CDHP in 2005. The mean age was 43 years and 54% of the population was men in each group. Over 40% of the CDHP members chose family coverage, with 38% having family sizes of two members. In 2004, 66.6% of the individuals were in the low-deductible PPO plan; in 2005, 88.6% of the individuals were in higher PHA CDHP option. The prevalence of disease states was highest for hypertension (23.7%), followed by dyslipidemia (15.7%), allergic rhinitis (10.5%), diabetes (10.2%), asthma (6.1%), GERD (5.6%), arthritis (4.3%), and depression (4.5%). The CDHP and PPO comparison groups were compared in the baseline year (2004) for age, gender, the RxRisk score, and the prevalence of disease states, and no significant differences were found for any of these characteristics.
Adjusted Pre and Post differences in Medical Utilization Between the CDHP and Matched Comparison Group
Table 3 shows that decreases in outpatient visits (35.7%) and laboratory and diagnostic visits (34.3%) for the CDHP group were much greater compared with the PPO comparison group in the post period (21.9% and 19.9%, respectively). There was a 31% reduction in the total number of prescriptions for the CDHP group compared with a 3.9% increase in the PPO comparison group. Differences between groups were significant for all measures (P < 0.0001).
Table 4 outlines ER visits and hospitalizations. The CDHP group had 90% lower odds for at least one ER visit (odds ratio [OR] = 0.10, 95% CI: 0.07 to 0.14) whereas the PPO comparison group had 41% lower odds (OR = 0.59, 95% CI: 0.44 to 0.78) for at least one ER visit in the post period. The reduction in the likelihood of any ER visits in the post period was significantly greater for the CDHP group (P < 0.0001). The CDHP group had 65% lower odds (OR = 0.35, 95% CI: 0.29 to 0.41) to have at least one hospitalization in the post period, whereas the PPO comparison group had only 32% lower odds (OR = 0.68, 95% CI: 0.58 to 0.80). Again, the reduction in the post period for the CDHP group was significantly greater than for the PPO comparison group (P < 0.0001).
Impact on Medication Adherence
Table 5 shows the OR for being adherent to medication therapy for each chronic disease state in the post period compared with the pre period for the CDHP group and the PPO comparison group and the results of a test for differences between the groups. Overall, the CDHP cohort was less likely to be adherent with their medications in the post period compared with the pre period (OR = 0.69, 95% CI = 0.64 to 0.74) whereas individuals in the PPO comparison cohort had no change (OR = 1.0). The difference between the two groups was statistically significant (P < 0.0001).
For the CDHP group, there was a statistically significant reduction in the odds of being adherent for the following disease states in the post period: asthma (55%), allergic rhinitis (29%), diabetes (45%), depression (27%), dyslipidemia (32%), GERD (43%), and hypertension (27%). For the PPO comparison group, the only statistically significant reduction in the odds of adherence in the post period was for depression (32%). Differences in the ORs between the two groups were statistically significant for all but three of the eight comparisons: arthritis, allergic rhinitis, and depression.
Adjusted Pre and Post Differences in Expenditures Between the CDHP and Matched Comparison Group
Table 6 outlines the differences in medical, pharmacy, and total expenditures between the CDHP and PPO comparison groups. Mean medical expenditure decreased by 26.9% ($276) per member in the CDHP group in the post period compared with 20.8% ($346) in the PPO comparison group (P = 0.69). Pharmacy expenditures decreased in the CDHP group (28.4%) whereas the PPO comparison group had an increase of 3.8% (P < 0.0001). Per member, total health care expenditures (combined medical and pharmacy expenditures) decreased by 27.5% ($447) in the CDHP group compared to 14.7% in the PPO comparison group ($325). This difference was not statistically significant.
To our knowledge, this is the first study to examine the impact of CDHPs on pharmacy and medical utilization and expenditures for individuals with chronic diseases. We examined a continuum from asymptomatic diseases such as hypertension, dyslipidemia, and diabetes to symptomatic diseases such as depression, asthma, arthritis, allergic rhinitis, and GERD. We also examined the impact of CDHPs on utilization and expenditures for an employee population that was offered only CDHP options. This unique design allowed us to examine the true impact of CDHPs on consumer behavior without the confounding effects of favorable selection of healthier or better paid employees into a CDHP, as previously reported.8. Our key findings are as follows. First, there were reductions in all components of medical utilization after switching to a CDHP.
There was an appreciable decrease in the number of outpatient visits and laboratory and diagnostic services for those switching to a CDHP. After switching to a CDHP, individuals experienced a 35.7% decrease in outpatient visits and 20% decrease in laboratory and diagnostic services. This is a surprising finding as preventive care was covered 100% (with no cost sharing for members or deductions from the PHA account) in both CDHP options. Examples of covered preventive services included routine outpatient visits, wellness examinations for children, cholesterol screenings, and mammograms. This could cause problems for employers. Many chronic disease sufferers require outpatient visits and laboratory services to manage their disease, for example, to optimize drug therapy. Some diseases we examined (diabetes and dyslipidemia) have guidelines that require monitoring of clinical markers, such as HbgAIC and low-density lipoprotein levels, through laboratory and diagnostic visits. A reduction in these services, when preventative care was covered at 100%, could indicate underutilization of necessary medical resources. We need a further investigation into the types of outpatient services that were reduced to determine whether they were for routine management of the disease or complications, to get more insights into whether these reductions were appropriate. The likelihood of ER visits and hospitalizations was also lower after switching to a CDHP. This finding is in contrast to that of Parente and Feldman, who reported an increase in hospitalizations in all 3 years of enrolment in a CDHP.7,8 However, our results are somewhat consistent with the findings of Wharam et al who found a reduction in ER-related hospitalizations and of those for Humana and Aetna, who found reductions in any hospitalization.9,13,14 We must note that although ER visits and hospitalizations represent key drivers of health care expenditures, we believe these components may be less sensitive to changes in health plan benefit design, especially in the short term. Most individuals do not anticipate the need for these services and so are not necessarily influenced by their coverage when choosing a health plan. The research to support this finding is mixed. Although researchers using data from the RAND Health Insurance Experiment found that insurance coverage does not influence hospitalizations,23 other researchers have found that increased cost sharing for ER visits was associated with reduced utilization. However, there was no evidence that increased cost sharing for ER visits led to any adverse outcomes such as higher mortality rates, or an increase in hospitalizations and ICU admissions.24,25 In addition, the reduction in ER visits that was observed in this previous research was for conditions considered less likely to be an emergency.25 Hence, we cannot conclusively attribute the change in utilization of ER visits and hospitalizations to CDHP design, especially as we did not examine the nature of ER visits and hospitalizations that were reduced, ie, whether they were for necessary, urgent or avoidable care.
To see if these reductions in medical utilization were influenced by CDHP design, we examined patterns of PHA and deductible usage for the CDHP group (results not shown). The majority of the study population (63%) shifted from the low deductible PPO option to the higher PHA CDHP option on January 1, 2005. Over half of employees (55%) in the higher PHA option were still faced with paying their deductible amount at the end of 2005. Faced with this financial burden individuals may have decided to reduce their use of services. In addition, the CDHP options we examined were also less generous plans with lower-than-average PHA amounts ($750 to 1500 for the higher PHA option and $200 to 400 for the lower PHA option) and high member responsibility ($750 to $1500 for both PHA options) than what was commonly available in the market.3 Consequently, higher deductibles with lower PHA spending amounts may have contributed to some of the reduction in medical utilization we observed.
The most concerning finding is that prescription utilization and medication adherence in patients with chronic conditions decreased significantly after switching to a CDHP. Overall prescription utilization decreased by 31%. The long-term negative impact of nonadherence on health-related outcomes has been well documented.26 The likelihood of medication adherence was lower for most of the disease states examined. Our results are supported by the work of Greene et al12 who also found reductions in medication adherence for antihypertensives and lipid lowering drugs that ranged from 8% to 20%.
One possible explanation for the substantial impact of CDHPs on medication adherence is the cost-sharing structure for prescriptions under these plans. In the pre period, both PPO plans had a mixed coinsurance-copay benefit, with $10 for generics, 20% for preferred brands, and 30% for nonpreferred brands medications. Although the PPO plan saw patients paying only a small percentage of the total costs of medication, CDHPs exposed members to the full cost of the medication when paying from their PHA accounts, which may have sensitized their prescription purchasing behavior. Once again, the less generous nature of the CDHP PHA amounts, particularly in option 2, and the accompanying high deductibles may have contributed to the decision to purchase fewer medications.
Finally, although CDHPs appear to lower medical utilizations significantly more than traditional PPOs, the trend did not necessarily translate into savings in medical expenditures. There was a significant reduction in medical expenditures in the post period in both groups but the magnitude of the reductions in expenditures were not statistically different between the two groups (−27% vs −21%, P = 0.69). Pharmacy expenditure was much lower in the CDHP group (−28%) compared with the PPO comparison group, which saw an increase (3.8%). For the CDHP group, the reduction in pharmacy expenditures may have resulted from underutilization of necessary medications. Reductions in total expenditures (the sum of the medical and pharmacy expenditures) in the post period were greater in the CDHP group than in the PPO comparison group. The difference again was not statistically significant (−28% vs −15%, P = 0.5). Our findings support the expectation that CDHPs can reduce health care expenditures, but whether the reduction in utilization is appropriate remains to be seen.
Health care utilization and medication adherence are reduced in the first year when CDHPs are offered as the only health plan by employers. The less generous nature of the CDHPs coupled with higher deductibles examined in this study may explain some of the reductions we observed. However, our results show that reductions in medical utilization in the CDHP group did not result in significant differences in health care savings when compared with the PPO group. More detail into the types of utilization that were reduced will provide further insight into whether these reductions were appropriate. Further research needs to be conducted to determine whether this reduction is a sign of underutilization with long-term consequences or efficient use of health care resources.
The study had several limitations. First, we used an external comparison group drawn from a nationally represented database. Even though, we matched the patients using propensity scoring, there may have been important differences between the two groups that were not observable and hence could not be measured.
Most notably, since our comparison group was drawn from a large group of several PPO plans, the specific cost sharing members faced in the comparison group were not available thereby limiting the interpretation of the comparative results.
Second, the employer group we examined offered less generous and less normative CDHP options. Thus, the experiences we describe may only be applicable to these CDHP offerings, which may occupy a limited role in today's marketplace. Parente, Feldman and Wharam examined more generous CDHP options, so our study of more restrictive options adds to the existing understanding of how CDHPs affect utilization and behavior.
Although our results showed a reduction in medical services for ER visits, hospitalizations, outpatient visits and laboratory or diagnostic services, we did not examine the nature of these reductions, whether they were for preventable, whether they were for necessary care, or whether they were for less serious conditions as demonstrated by previous research.24,25 An investigation into the types of reductions could provide more evidence about the utility of CDHPs in reducing appropriate or inappropriate care. Finally, we looked at utilization in the first year following a switch to a CDHP. Most of the chronic diseases we examined may require much longer follow-up periods to estimate the true impact of a CDHP on utilization and expenditures.
We thank Regina Levin and Nancy Hardie for their help in data management.
This study was supported by GlaxoSmithKline.
2.KFF Report: Kaiser Family Foundation. Employer Health Benefits, 2006 Annual Survey. Available at: www.kff.org
. Accessed March 24, 2008.
4.AHIP, Center for Policy and Research. Census Shows 4.5 Million People Covered by HSA/High Deductible Health Plans
. AHIP, Center for Policy and Research; January 2007. Available at: www.ahipresearch.com
. Accessed April 2007.
5. RAND. Consumer Directed Health Care: Early Evidence Shows Lower Costs, Mixed Effects on Quality of Care. RAND Research Highlights, RAND 2007. Available at: www.rand.org
. Accessed February 5, 2009.
6. Buntin MB, Damberg C, Haviland A, et al. Consumer directed health care: early evidence about effects on cost and quality. Health Aff. 2006;25:W516–W530.
7. Parente ET, Feldman R, Christenson JB. Evaluation of a consumer driven health plan on medical care expenditures and utilization. Health Serv Res. 2004;39:1091–1111.
8. Feldman R, Christenson JB, Parente ET. Consumer directed health plans: new evidence on cost and utilization. Paper presented at the fifth World Congress of the International Health Economics Association, Barcelona, Spain, July 10–13, 2005.
9. Wharam FJ, Landon BE, Galbraith AA, Kleinman KP, Soumerai SB, Degnan DR. Emergency department use and subsequent hospitalizations among members of a high deductible health plan. JAMA. 2007;297:1093–1102.
10. Buntin MB, Damberg C, Haviland C, et al. Consumer directed health care: early evidence about effects on cost and quality. Health Aff. 2006;25 w516–w530.
11. Dixon A, Greene J, Hibbard J. Do consumer directed health plans drive change in enrollees' healthcare behavior. Health Aff. 2008;27:1120–1131.
12. Greene J, Hibbard J, Murray JF, Teutsch SM, Berger ML. The impact of consumer directed health plans on prescription drug use. Health Aff. 2008;27:1111–1119.
14. Aetna. Aetna HealthFund® First-Year Results Validate Positive Impact of Health Care Consumerism
. Aetna; June 22, 2004. Available at: www.aetna.com/news/2004/pr_20040622.htm
. Accessed March 24, 2008.
15. Center for Health System Change. Rising Health Costs, Medical Debt and Chronic Conditions. Issue Brief No. 88. Center for Health System Change; September 2004.
16. Rowe JW, Stevenson TB, Downey RL, Newhouse JP. The effect of consumer directed health plans on the use of preventive and chronic illness services. Health Aff. 2007;27:113–120.
17. Wharam FJ, Galbraith AA, Kleinman KP, Soumerai SB, Degnan DR, Landon BE. Cancer screening before and after switching to a high deductible plan. Ann Intern Med. 2008;148:647–655.
18. Hibbard JH, Greene J, Tusler M. Does enrollment in a CDHP stimulate cost effective utilization? Med Care Res Rev. 2008;65:437–449.
19. D'Agostino. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17:2265–2281.
20. Fairman K, Motheral B. Evaluating medication adherence: which measure is right for your program? J Manag Care Pharm. 2000;6:499–504.
21. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13–22.
22. Fishman PA, Goodman MJ, Hornbrook MC, Meenan RT, Bachman DJ, O'Keeffe Rosetti MC. Risk adjustment using automated ambulatory pharmacy data: the RxRisk model. Med Care. 2003;41:84–99.
23. Gabel JR, Jensen GA, Hawkins S. Self-insurance in times of growing and retreating managed care. Health Aff (Millwood). 2003;22:202–210.
24. Hsu J, Price M, Brand R, et al. Cost sharing for emergency care and unfavorable clinical events: findings from the safety and financial ramifications of ED copayments study. Health Serv Res. 2006;41:1801–1820.
25. Selby JV, Fireman BH, Swain BW. Effect of a copayment on use of the emergency department in a health maintenance organization. N Engl J Med. 1996;334:635–641.
26. World Health Organization. Adherence to Long-Term Therapies: Evidence for Action. Geneva: World Health Organization; 2003.