Stephen B. Soumerai, ScD

To determine the 2-year impact of high-deductible health plans (HDHPs) on highacuity, expensive medical care.

Retrospective pre-post, with propensity score—matched comparison group.

 We studied emergency department visits, hospitalizations, and related expenditures among 15,847 HMO members for 1 year before and up to 2 years after an employer-mandated switch to HDHPs, compared with 15,847 propensity-matched controls who remained in HMOs. Members were aged 1 to 64 years and insured between 2001 and 2008.

 Emergency department visits among HDHP members declined by 15.0% and 15.7% from baseline to the first and second follow-up years, respectively (95% confidence intervals −21.1% to −8.4% and −24.1% to −6.4%, respectively). Rates of nonemergent visits declined significantly in both years (−19.6% [−28.2% to−9.9%] and −18.1% [−29.8% to −4.4%], respectively), while intermediate-severity visits declined to a lesser degree (−13.4% [−23.0% to −2.5%] in the first and −10.9% [−24.4% to 5.1%] in the second follow-up year). Reductions in emergent visits were not detectable in either the first or second follow-up year (−9.7% [−26.9% to 11.5%] and −15.3% [−36.8% to 13.3%], respectively). Hospitalization rates decreased in the first follow-up year (−22.8% [−33.8% to −10.0%]), but hospitalization and cost reductions were not detectable by the second follow-up year (−11.8% [−27.9% to 7.9%] and 1.9% [−22.2% to 33.4%], respectively).

HDHP members experienced sustained reductions in emergency department visits over 2 years, but reductions in hospital utilization and costs were not apparent by the second year. Longer-term studies that assess deferred utilization and its effects are needed.

	This is the first study to examine the longer-term impact of high-deductible health plans on high-acuity, expensive medical care. Policy makers should consider closely monitoring enrollees for unintended consequences of cost sharing.

		 High-deductible health plans are expanding at unprecedented rates.

		 High-deductible health plan members who remained enrolled for up to 2 years had fewer, mostly nonemergent visits to the emergency department.

		 Initial large reductions in hospital utilization among high-deductible health plan members diminished by the second year.

 are increasingly turning to high-deductible health plans (HDHPs) to control rising health care costs. Compared with traditional employer-based plans, HDHPs have relatively low premiums but subject most services to annual deductibles of at least $1000.

; membership tripled between 2006 and 2010

 National health insurance reform could cause an “explosion” 

	The effects of HDHPs on healthcare utilization, costs, and health are controversial. Some suggest that HDHPs will reduce inappropriate medical services, control cost trends, and improve health outcomes.

 Others are concerned that HDHPs will decrease necessary healthcare, causing adverse health outcomes and increased longer-term costs.

	Prior studies including the RAND Health Insurance Experiment suggest that increased cost sharing reduces use of both appropriate and inappropriate health services, including hospitalizations,

 These effects may be associated with worse health outcomes.

In contrast, patients primarily reduce discretionary services when emergency department care is subject to cost sharing, and adverse outcomes have not been detected.

	A previous 1-year follow-up study of emergency department and hospital utilization after HMO members were switched into HDHPs found that HDHP members primarily reduced low-severity emergency department visits after an initial emergency department visit, and hospitalizations also declined.

 We hypothesized that care patterns among HDHP members in their second year of experience would be a better indicator curve” that occurs under HDHPs.

	Harvard Pilgrim Health Care insures approximately 1 million individuals in the Northeast. In April 2002, Harvard Pilgrim began offering HDHPs with $500 to $2000 individual annual deductibles. Members of family plans also have family deductibles equal to twice their individual deductible. Full coverage begins for individuals their individual deductible or for family members when the family’s combined expenses exceed the family deductible. Most institution-based services (eg, emergency department and hospital care), laboratory studies, and imaging procedures are subject to the deductible. Members have a $20 copayment for most outpatient visits (independent of reaching the deductible). When HDHP member spending is below the deductible, healthcare providers bill patients directly at health plan—negotiated rates including for costs of emergency department care. After members exceed the deductible spending level, they have a $100 copayment for emergency department care that is waived if they are hospitalized. Employers purchasing the HDHPs may opt to combine them with a Health Reimbursement Arrangement. This allows employers to place money into an account to reimburse employees for out-of-pocket health expenses.

	The traditional HMO members we included had varying in-network emergency department and outpatient copayments (between $30-$100 and $5-$25, respectively). Inpatient copayments ranged from $0 to $1000 with a median of $250.

We identified Massachusetts members aged 1 to 64 years insured by Harvard Pilgrim between April 1, 2001, and February 28, 2008, with at least 1 year of continuous enrollment in a traditional HMO plan followed by at least 6 months in the HDHP. We chose members whose employers offered only a single health plan and who remained with the same employer for the entire period. Members were therefore unable to self-select their health plan. We included members from employers who bought plans directly from Harvard Pilgrim as well as from employers who purchased Harvard Pilgrim plans from independent brokers (“association” plans). We excluded members from employers that offered health plans from other insurers. This left an initial HDHP cohort of 16,472 members. We identified an index date for each member (the HDHP switch date), a 12-month baseline period, and a 6- to 24-month follow-up period.

	We excluded 125 members with missing descriptive data including geocoded education and poverty levels, family versus individual insurance plan, employer category, baseline copayment levels, and whether the employer offered health insurance exclusively through Harvard Pilgrim.

To identify a control group, we first developed a pool of Harvard Pilgrim members from Massachusetts employers who were enrolled in traditional HMO plans during the same April 2001 to February 2008 eligibility period and whose employers did not offer the choice to enroll in an HDHP, any other Harvard Pilgrim benefit type plan, or plans from insurers other than Harvard Pilgrim for at least 18 months during their eligibility. We matched 8 members from this control pool to the HDHP member sample (before the exclusions above) based on contemporaneous enrollment, adult/child status, and whether the member was in an association plan. Each matched control member was then assigned the same index date and had the same baseline and follow-up periods as the matched HDHP member.

	From this initial 111,014-member control pool we excluded 881 members with missing descriptive data and 13,590 members (12.2%) whose employers added a small deductible ($50-$100) or increased emergency department copayments at the index date. In addition, we censored the experience of 5810 control members after their first follow-up year because their employers added a small deductible or increased emergency department copayments in the second follow-up year. After these exclusions, the control pool comprised 96,543 members.

We then created propensity score models that predicted the likelihood of enrolling in an HDHP versus remaining in a traditional HMO after the baseline year. Propensity score matching is a well-established method for generating a control group with a similar likelihood of being exposed to a given “intervention” (in this case, shifting to HDHP coverage) based on measured characteristics when individuals have not been randomly allocated into study groups.

	In preparation for propensity score matching, we determined the variables that predicted employer choice into an HDHP in the overall cohort. We found that all variables we tracked (age; sex; Adjusted Clinical Groups morbidity

; family vs individual plan; census-block education and poverty level; employer type; baseline outpatient, emergency department, and hospital copayment levels; year of index date; and baseline member and health plan expenditures) predicted the choice of an HDHP at a statistically significant level.

	We matched HDHP members to HMO members using 1 to 1 caliper matching without replacement.

Caliper matching selects the control with the propensity score closest to that of the case, but only if the control’s score is within a predefined distance (caliper) of the case.

Therefore, cases might not be matched to any controls; those cases would be dropped from the sample to avoid inappropriate matches. Caliper matching “without replacement” does not allow selecting the same control as a match more than once.

 We used a caliper width equal to 0.6 of the pooled standard deviation of the propensity score, which has been found to eliminate the majority of bias due to measured confounding variables.

	Our final study group included 15,847 HDHP members and their matched controls.

	We identified emergency department claims from Harvard Pilgrim’s claims database and used a validated

modification of the emergency department visit classification algorithm of Billings and colleages

 to categorize visit severity as nonemergent, intermediate, or emergent. This algorithm and a method of developing clinically meaningful visit clusters using Clinical Classifications Software

 have been described in a previous report.

To determine emergency department costs, we summed annual expenditures paid by members and Harvard Pilgrim for members’ emergency department use.

	We identified hospitalizations based on health insurance claims, and we calculated length of hospital stay. We excluded hospitalizations with the same admission and discharge date because these represent day procedures or surgeries such as colonoscopy or hernia repair. Multiple studies have used hospitalization rates or length of stay as proxy measures of patient morbidity.

 We calculated mean length of stay and summed annual hospitalization days per member. To assess whether members deferred needed ambulatory care, we measured ambulatory care—sensitive hospitalizations using the Agency for Healthcare Research and Quality Prevention Quality Indicators algorithm.

International Classification of Diseases, 9th Revision, Clinical Modification

) codes indicating potentially preventable hospitalizations for 14 conditions.

 Prevention Quality Indicators have been used in multiple academic studies.

	We calculated annual hospital expenditures by summing the amounts paid by members and Harvard Pilgrim for all claims during members’ hospitalization dates after excluding emergency department and outpatient claims. We assigned hospitalizations spanning a study year to the previous year.

	To estimate comorbidity, we applied the Adjusted Clinical Group algorithm to each member’s 12-month baseline period. The Adjusted Clinical Group score is based on age, sex, and 

 diagnostic codes derived from claims. It calculates a morbidity weight standardized across a reference population of adults and children.

 Average morbidity is 1.0 and persons with more comorbidities have higher scores.

	To derive proxy measures of socioeconomic status, we linked members’ residential addresses to their 2000 US Census block group

and created previously established variables measuring education and poverty status.

We used a principal components approach to calculate a neighborhood socioeconomic status index based on census block levels of poverty and high school education. This approach creates a single variable from correlated variables

and has been validated as a measure of socioeconomic status.

We calculated baseline year total expenditures for each member by summing the amounts paid by Harvard Pilgrim and members for all claims. Other covariates included age, sex, employer category (based on number of employees and whether members were in association plans), and whether members were in individual or family plans.

	We compared baseline characteristics of our study groups using 

 tests. To determine whether longer follow-up duration and differential dropout might bias results, we used 

 tests to compare baseline emergency department visits, hospitalizations, and total costs for HDHP and control members with at least 18 months of follow-up. We used a difference-in-differences analytic framework to examine changes in the outcomes of interest in the HDHP group compared with the control group from baseline to the first and second follow-up years. Difference-in-differences analyses subtract the pre-post difference in outcomes in the intervention group from the pre-post difference in outcomes in the control group.

	In the overall cohort, we analyzed annual rates of emergency department visits as well as those classified as nonemergent, intermediate severity, and emergent; the 5 most common clinical clusters (combined) of classified emergency department visits; annual overall and ambulatory care—sensitive hospitalizations; and annual hospitalization days among the groups in the baseline, first, and second follow-up year. We also examined total annual expenditures for emergency department visits and hospitalizations in the overall cohort. We calculated lengths of hospital stay among hospitalized members.

	We then used Poisson regression to model the independent association between HDHP status and these outcomes after controlling for the aforementioned covariates. Poisson regression is used to generate results adjusted for covariates when analyzing count outcomes that have skewed distributions. We included the index date in models to adjust for secular utilization trends. Because some members had partial- year follow-up, we also adjusted for follow-up duration. Two-part modeling can be used to model the 2-step process of deciding whether to use healthcare and how much to use, but we used 1-part general linear modeling because of key advantages when examining healthcare utilization, including predictions that are not biased by heteroscedasticity and coefficients that are easier to interpret.

 Therefore, all statistical models used generalized estimating equations

to adjust for clustering of events within individuals between the baseline and follow-up years 1 or 2. Analyses were performed using SAS version 9.2 software (SAS Institute Inc, Cary, North Carolina).

	The study was approved by the Harvard Pilgrim Health Care Institutional Review Board.

Articles

High-deductible health plan members had sustained reductions in emergency department visits over 2 years but initial reductions in hospital utilization and costs were not sustained.

High-Deductible Insurance: Two-Year Emergency Department and Hospital Use

To evaluate the impact of switching from an HMO to a high-deductible health plan on the costs and utilization of maternity care.

 We compared 229 women who delivered babies before or after their employers mandated a switch from HMO coverage to a high-deductible health plan, with a control group of 2180 matched women who delivered babies while their employers remained in an HMO plan. Administrative claims from a large Massachusetts-based health insurance program were used in a difference-indifferences regression analysis.

Mean out-of-pocket maternity care costs for high-deductible group members increased from $356 for women who delivered before the insurance transition (n = 86) to $942 for women who delivered after the transition (n = 143), compared with a change from $262 (n = 711) to $282 (n = 1569) for HMO members, a relative increase of 106% (P <.001) for high-deductible members. Delivery after transition to a high-deductible plan was not associated with changes in the odds of receiving early prenatal care (odds ratio [OR], 1.02; 95% confidence interval [CI], 0.32-3.19), recommended prenatal visits (OR, 1.64; 95% CI, 0.89-3.02), or postpartum care (OR, 0.74; 95% CI, 0.42-1.32).

 Switching from an HMO to a high-deductible plan with exemptions for routine care increased out-of-pocket member costs for maternity care, but had no apparent adverse impacts on receipt of recommended prenatal and postpartum care.

	Transition from an HMO to a high-deductible health plan with deductible exemptions for routine preventive maternity services was associated with:

		Substantial increases in out-of-pocket expenses for maternity and delivery-related care (although information on changes in employer account contributions and employee premium contributions was not available).

		A statistically insignificant (but large) reduction in overall costs.

		No apparent adverse impacts on recommended prenatal and postpartum care.

Childbirth is the leading reason for hospitalization in the United States.

 For both private insurers and Medicaid, hospital maternity and newborn charges exceed those for any other condition.

 In US hospitals, vaginal and caesarean deliveries cost approximately $8000 and $11,000 respectively.

High-deductible health plans (HDHPs) have expanded rapidly in recent years, with 27% of workers having a deductible of at least $1000.

 Between 2006 and 2010, enrollment in HDHPs tripled to 10.0 million members,

 and analysts predict an “explosion” in growth due to the recent health reform legislation.

Although HDHPs are intended to decrease escalating health costs and discretionary care, they have generated controversy.

 High cost sharing can reduce appropriate  healthcare utilization, including hospitalizations, preventive services, and essential medicines.

 Therefore, HDHPs often fully cover preventive services, including prenatal care.

 No previous studies have measured the impact of HDHPs on the costs and use of maternity care.

Our study examined this question among women insured by a large Massachusetts-based health insurance program (Harvard Pilgrim Health Care) that predominantly insures members through HMO plans and began offering HDHPs in April 2002. We anticipated that HDHP membership would increase out-of-pocket maternity expenditures, shift costs to members, and possibly reduce total maternity care expenditures, consistent with the cost-containment goal of consumer-directed healthcare.

Although use of exempted services might be expected to remain stable, we hypothesized 3 mechanisms that could decrease utilization. (1) Cost sharing for nonexempt services (such as laboratory tests) ordered during prenatal visits could discourage subsequent visits. (2) The intricacy of maternity benefit design in HDHPs (eg, deductible exemptions based on clinical conditions) combined with uncertainty regarding costs of anticipated services could cause confusion and reduced care even for exempt services.

 (3) Experiencing a previous large deductible payment (eg, an emergency department visit) might induce “sticker shock” and broadly reduce utilization.

The objective of this study was to evaluate the impact of transition from an HMO to an HDHP on the costs and utilization of maternity and delivery care services within the 6 months before and 3 months following delivery.

The Office for Sponsored Programs (Harvard Pilgrim Health Care Institute) reviewed this study and granted institutional review board approval.

We used a pre—post study design with matched control group and a difference-in-differences analysis to measure changes in study outcomes.

Harvard Pilgrim Health Care is a health plan serving approximately 1 million individuals in New England. On April 1, 2002, Harvard Pilgrim began offering HDHPs with annual deductible amounts of $500 to $2000 for individuals and $1000 to $4000 for families. Members of family plans also had individual deductibles equal to half of the family deductible. Of Harvard Pilgrim’s commercially insured members, approximately 70% have HMO plans, 20% have preferred provider organization (PPO) plans, and 10% have HDHPs, compared with corresponding national rates of 58%, 19%, and 13%.

 Although PPO plans are more common than the HMO plans we study, cost-sharing requirements for services such as preventive tests, hospitalizations, and specialist visits are quite similar4; the primary differences between HMO and PPO plans are the more limited provider network, the need for specialist referrals, and the inclusion of high deductibles in approximately 40% of PPO plans.

Insurance coverage for maternity care in Harvard Pilgrim HMO plans includes full coverage of prenatal and postpartum care, with no cost sharing beyond office visit and  hospitalization copayments. Outpatient visit copayments among HMO members in our study ranged from $5 to $25 (median $15) while hospitalization copayments ranged from $0 to $1000 (median $250).

In the HDHPs we studied, many maternity care services are exempt from the deductible, having either first dollar coverage or low copayments. First dollar coverage applies to routine prenatal and postpartum visits, fetal ultrasounds (sonograms),  routine urinalysis, Papanicolaou tests, and screenings for sexually transmitted infections. All hospital delivery charges and many outpatient procedures are subject to the deductible, but are covered in full after the deductible has been met. Copayments of $20 apply to non—maternity care outpatient visits, urgent care visits, and specialist consultations. These copayments apply regardless of whether the member has exceeded the deductible spending level. However, the HDHPs have an out-of-pocket maximum (including copayments and deductibles) of $2000 to $4000 for individuals and $4000 to $8000 for families. After a member reaches the out-of-pocket maximum, all services are covered in full. Provider networks for women in HDHPs and HMO plans are identical.

The HDHPs we studied were not eligible to be paired with health savings accounts (HSAs); all were eligible to be combined with health reimbursement accounts (HRAs). Although we could not account for HRA purchases from other companies, only 3% of the HDHPs we studied had HRAs through Harvard Pilgrim. Nationally, fewer than half of HDHPs with deductibles over $1000 are paired with HSA or HRA accounts,

 so our analysis is relevant for the predominant type of HDHP.

 we created a cohort of health plan members enrolled through employers who exclusively offered insurance through Harvard Pilgrim Health Care. We defined 2 cohorts: an HDHP group and an HMO control group. The HDHP group included members enrolled in traditional HMO plans during a 1-year baseline period prior to an employer-mandated switch to an HDHP. The date of this switch was defined as the index date. For each HDHP member, we identified 8 contemporaneous members who  were continuously enrolled in traditional HMO plans during the same time period. We assigned HMO controls the same index date as their HDHP counterparts, and the distribution of index dates was similar in the HDHP and HMO control groups in the final study population. We selected only HMO members whose employers did not offer an option to enroll in an HDHP or any other plan types. Therefore, no study members were able to self-select their health insurance plan or benefits structure.

From the HDHP and HMO groups, we identified women who had given birth using 

International Classification of Diseases, Ninth Revision

 diagnosis codes and diagnosis-related group payment codes. We further divided the study cohorts based on whether each woman delivered a baby prior to or following the index date. The final study population included 2409 women who delivered babies between October 1, 2001, and November 30, 2007. This population included 229 women in the HDHP group (86 delivered prior to the index date and 143 after) and 2180 women in the HMO control group (711 delivered prior to the index date and 1569 after). Among women in the HDHP group, 95 had a $500 deductible, 92 had a $1000 deductible, and 34 had a $2000 deductible (values unavailable for 8 women). The study used data from 6 months prior to and 3 months following delivery (the second and third trimesters of pregnancy and early postpartum period).

Study outcomes included costs, quality, and utilization, categorized as delivery related (for care provided during the delivery hospitalization) and maternity related (for all types of maternity care, including delivery-related care). Cost measures included out-of-pocket and total expenditures. Out-of-pocket costs comprised all member expenditures toward copayments and deductibles. Total costs were calculated as the sum of the expenditures by the insurer (payer costs) and out-of-pocket expenditures by the member.

We calculated maternity and delivery-related costs using a previously established set of diagnosis, procedure, and payment codes that included prenatal and postpartum visits,  obstetrical anesthesia, introduction and repair services, vaginal or caesarean delivery, radiology procedures, pathology and laboratory expenses, and in utero procedures.3 We distinguished delivery services from non—delivery-related maternity care both by the specific service and procedure codes and by the timing of occurrence during the delivery hospitalization. Measures of maternity care quality included 3 measures based on Healthcare Effectiveness Data and Information Set (HEDIS) indicators (early prenatal care, appropriate frequency of prenatal care, and timely postpartum care).25 Early prenatal care was defined as having at least 1 prenatal visit within 42 days of the second trimester. Appropriate frequency of prenatal care comprised at least 7 prenatal visits in the 190 days prior to delivery, as per consensus recommendations.26 We calculated this measure only among women who delivered full-term babies (>37 weeks of gestation) due to lack of information on gestational age at delivery for preterm infants. Timely postpartum care was achieved by having a visit between 21 and 56 days after delivery. General maternity care utilization measures included the total number of prenatal visits and sonograms, and the length of stay for the delivery hospitalization.

Other variables were age at delivery, month of delivery, type of delivery (vaginal or caesarean), gestational diabetes, high-risk pregnancy, preterm delivery, chronic disease score, socioeconomic status, and employer size (number of employees). Clinical conditions were defined using payment or diagnosis codes. As a measure of comorbidity and  care-seeking behavior, we computed the chronic disease score, which predicts future healthcare spending based on pharmacy claims in the baseline year.27 To develop a measure of socioeconomic status, we linked members’ residential addresses to their 2000 US Census block group and created measures of neighborhood education levels and poverty status based on previously established standards.28 Low socioeconomic status was indicated by living in a census block group with either more than 25% of adults having less than a high school education or more than 10% of households living below the poverty level.

We used regression models in a difference-in-differences analysis to assess the independent effect of switching to an HDHP on costs, quality, and utilization outcomes, after controlling for the covariates listed above. This analysis used 3 types of regression models: log-linear for costs, logistic for dichotomous quality measures, and linear for utilization measures (number of visits or days). We also conducted several sensitivity analyses to test the stability of results. Analyses excluding socioeconomic status (which is measured at the census block level rather than the individual level) and adjusting for inflation (using the medical component of the Consumer Price Index) had no impact on the direction or significance of the original findings.

All statistical analyses used 2-tailed tests and a levels of .05, and were conducted using SAS version 9.1 (SAS Institute Inc, Cary, NC).

Study groups were similar, with no statistically significant differences detected (

). The average age at delivery was approximately 33 years. Between 15% and 18% lived in low socioeconomic status neighborhoods. More than 60% worked for small employers. Approximately one-third of women had a caesarean delivery, and between 20% and 30% experienced complications of high-risk pregnancy. Preterm deliveries accounted for fewer than 10% of births, and rates of gestational diabetes were low. Chronic disease scores were consistent across groups and indicate a generally healthy population.

presents costs, quality, and utilization outcomes in the study groups as well as regression-based estimates of the impact of switching from an HMO to an HDHP on each of these outcomes. Regression coefficients for cost outcomes may be interpreted as the estimated percentage change in cost. Mean total maternity care costs for HDHP members  were $9761 for women who delivered prior to the insurance transition and $9997 for those who delivered after the transition, compared with a change from $9731 to $10,434 for HMO members, a 12% (

 = .114) relative reduction in the HDHP group, after adjusting for covariates. This occurred despite a significant increase in the underlying cost trend over time (8%, 

). Further, the HDHP group experienced a 15% relative decline in delivery costs 

In contrast to overall costs, out-of-pocket spending increased following the transition from an HMO to an HDHP. Average out-of-pocket maternity care expenditures among HDHP members rose from $356 before the HDHP transition to $942 for those who delivered after, compared with nearly constant values ($262 to $282) for women who remained inHMO  plans, a relative increase of 106% (

 <.001). Mean out-of-pocket costs for delivery care rose from $258 to $547 among HDHP members, compared with $188 to $218 among HMO members, a 45% relative increase (

 <.001). Average insurer expenditures decreased for women who delivered babies after HDHP enrollment, falling from $9405 to $9055, while rising from $9469 to $10,152 for the control group, a 21% relative decrease (

 = .005). The relative decrease in insurer expenditures for delivery care was similar (22%, 

Achievement of quality standards for early and frequent prenatal care as well as timely postpartum care did not vary significantly across study groups (Table 2). Nearly all women received early prenatal care, and the majority accessed recommended prenatal and postpartum care. Likewise, maternity care utilization was similar across groups, with an average of 9 to 10 prenatal visits and 7 to 8 sonograms (mode = 4; 25% of women had >10 sonograms) and an average delivery hospitalization stay of about 4 days (3 days for vaginal deliveries; 5 days for caesarean deliveries).

As expected, pregnancy and delivery complications, high chronic disease scores, and caesarean and preterm deliveries were generally associated with higher costs and utilization (

This study is the first to characterize the impact of HDHPs on maternity care services. Our findings demonstrate that transitioning from an HMO to an HDHP was associated with increased out-of-pocket costs but stable utilization and quality for maternity and delivery care. Out-of-pocket costs represented a small percentage of total maternity costs for both HMO (4%) and HDHP (11%) members and were in the same range as national averages (6% of total costs).

Nevertheless, HDHP members faced out-of-pocket costs 3 times higher than HMO members but received recommended prenatal and postpartum care at similar rates.

An overall reduction in costs associated with better consumer choice is one of the goals of HDHPs and other consumer-driven healthcare models.

 We found a 12% (nonsignificant) decrease in the average costs of maternity-related care following transition to an HDHP. This trend might result from reductions in expensive individualized services or elective tests and procedures (such as prenatal genetic counseling or early screening for chromosomal abnormalities). Transitioning from an HMO to an HDHP was not associated with adverse impacts on basic prenatal and postpartum quality measures, lending support to deductible exemptions for such services. Further, there was no  measurable effect on prenatal visits, sonograms, or hospitalization following delivery. While a highly informed and “rational” consumer would not forgo any such services even if facing full cost sharing (because the expected delivery hospitalization essentially caps spending at the deductible level), immediate-term outof- pocket costs might remain important potential barriers to HDHP members, especially among lower-income women.

High-deductible benefit design was associated with increased out-of-pocket costs for maternity care, but had no adverse impacts on receipt of recommended prenatal and postpartum care.

High-Deductible Health Plans and Costs and Utilization of Maternity Care

Transition from an HMO to a high-deductible health plan with deductible exemptions for routine preventive maternity services was associated with:  n Substantial increases in out-of-pocket expenses for maternity and delivery-related care (although information on changes in employer account contributions and employee premium contributions was not available).  n A statistically insignificant (but large) reduction in overall costs.  n No apparent adverse impacts on recommended prenatal and postpartum care.

	Childbirth is the leading reason for hospitalization in the United States.

	Although HDHPs are intended to decrease escalating health costs and discretionary care, they have generated controversy.

	Our study examined this question among women insured by a large Massachusetts-based health insurance program (Harvard Pilgrim Health Care) that predominantly insures members through HMO plans and began offering HDHPs in April 2002. We anticipated that HDHP membership would increase out-of-pocket maternity expenditures, shift costs to members, and possibly reduce total maternity care expenditures, consistent with the cost-containment goal of consumer-directed healthcare.

	Although use of exempted services might be expected to remain stable, we hypothesized 3 mechanisms that could decrease utilization. (1) Cost sharing for nonexempt services (such as laboratory tests) ordered during prenatal visits could discourage subsequent visits. (2) The intricacy of maternity benefit design in HDHPs (eg, deductible exemptions based on clinical conditions) combined with uncertainty regarding costs of anticipated services could cause confusion and reduced care even for exempt services.

	The objective of this study was to evaluate the impact of transition from an HMO to an HDHP on the costs and utilization of maternity and delivery care services within the 6 months before and 3 months following delivery.

	The Office for Sponsored Programs (Harvard Pilgrim Health Care Institute) reviewed this study and granted institutional review board approval.

	We used a pre—post study design with matched control group and a difference-in-differences analysis to measure changes in study outcomes.

	Harvard Pilgrim Health Care is a health plan serving approximately 1 million individuals in New England. On April 1, 2002, Harvard Pilgrim began offering HDHPs with annual deductible amounts of $500 to $2000 for individuals and $1000 to $4000 for families. Members of family plans also had individual deductibles equal to half of the family deductible. Of Harvard Pilgrim’s commercially insured members, approximately 70% have HMO plans, 20% have preferred provider organization (PPO) plans, and 10% have HDHPs, compared with corresponding national rates of 58%, 19%, and 13%.

	Insurance coverage for maternity care in Harvard Pilgrim HMO plans includes full coverage of prenatal and postpartum care, with no cost sharing beyond office visit and  hospitalization copayments. Outpatient visit copayments among HMO members in our study ranged from $5 to $25 (median $15) while hospitalization copayments ranged from $0 to $1000 (median $250).

	In the HDHPs we studied, many maternity care services are exempt from the deductible, having either first dollar coverage or low copayments. First dollar coverage applies to routine prenatal and postpartum visits, fetal ultrasounds (sonograms),  routine urinalysis, Papanicolaou tests, and screenings for sexually transmitted infections. All hospital delivery charges and many outpatient procedures are subject to the deductible, but are covered in full after the deductible has been met. Copayments of $20 apply to non—maternity care outpatient visits, urgent care visits, and specialist consultations. These copayments apply regardless of whether the member has exceeded the deductible spending level. However, the HDHPs have an out-of-pocket maximum (including copayments and deductibles) of $2000 to $4000 for individuals and $4000 to $8000 for families. After a member reaches the out-of-pocket maximum, all services are covered in full. Provider networks for women in HDHPs and HMO plans are identical.

	The HDHPs we studied were not eligible to be paired with health savings accounts (HSAs); all were eligible to be combined with health reimbursement accounts (HRAs). Although we could not account for HRA purchases from other companies, only 3% of the HDHPs we studied had HRAs through Harvard Pilgrim. Nationally, fewer than half of HDHPs with deductibles over $1000 are paired with HSA or HRA accounts,

	From the HDHP and HMO groups, we identified women who had given birth using 

	Study outcomes included costs, quality, and utilization, categorized as delivery related (for care provided during the delivery hospitalization) and maternity related (for all types of maternity care, including delivery-related care). Cost measures included out-of-pocket and total expenditures. Out-of-pocket costs comprised all member expenditures toward copayments and deductibles. Total costs were calculated as the sum of the expenditures by the insurer (payer costs) and out-of-pocket expenditures by the member.

	We calculated maternity and delivery-related costs using a previously established set of diagnosis, procedure, and payment codes that included prenatal and postpartum visits,  obstetrical anesthesia, introduction and repair services, vaginal or caesarean delivery, radiology procedures, pathology and laboratory expenses, and in utero procedures.3 We distinguished delivery services from non—delivery-related maternity care both by the specific service and procedure codes and by the timing of occurrence during the delivery hospitalization. Measures of maternity care quality included 3 measures based on Healthcare Effectiveness Data and Information Set (HEDIS) indicators (early prenatal care, appropriate frequency of prenatal care, and timely postpartum care).25 Early prenatal care was defined as having at least 1 prenatal visit within 42 days of the second trimester. Appropriate frequency of prenatal care comprised at least 7 prenatal visits in the 190 days prior to delivery, as per consensus recommendations.26 We calculated this measure only among women who delivered full-term babies (>37 weeks of gestation) due to lack of information on gestational age at delivery for preterm infants. Timely postpartum care was achieved by having a visit between 21 and 56 days after delivery. General maternity care utilization measures included the total number of prenatal visits and sonograms, and the length of stay for the delivery hospitalization.

	Other variables were age at delivery, month of delivery, type of delivery (vaginal or caesarean), gestational diabetes, high-risk pregnancy, preterm delivery, chronic disease score, socioeconomic status, and employer size (number of employees). Clinical conditions were defined using payment or diagnosis codes. As a measure of comorbidity and  care-seeking behavior, we computed the chronic disease score, which predicts future healthcare spending based on pharmacy claims in the baseline year.27 To develop a measure of socioeconomic status, we linked members’ residential addresses to their 2000 US Census block group and created measures of neighborhood education levels and poverty status based on previously established standards.28 Low socioeconomic status was indicated by living in a census block group with either more than 25% of adults having less than a high school education or more than 10% of households living below the poverty level.

	We used regression models in a difference-in-differences analysis to assess the independent effect of switching to an HDHP on costs, quality, and utilization outcomes, after controlling for the covariates listed above. This analysis used 3 types of regression models: log-linear for costs, logistic for dichotomous quality measures, and linear for utilization measures (number of visits or days). We also conducted several sensitivity analyses to test the stability of results. Analyses excluding socioeconomic status (which is measured at the census block level rather than the individual level) and adjusting for inflation (using the medical component of the Consumer Price Index) had no impact on the direction or significance of the original findings.

	All statistical analyses used 2-tailed tests and a levels of .05, and were conducted using SAS version 9.1 (SAS Institute Inc, Cary, NC).

	Study groups were similar, with no statistically significant differences detected (

	In contrast to overall costs, out-of-pocket spending increased following the transition from an HMO to an HDHP. Average out-of-pocket maternity care expenditures among HDHP members rose from $356 before the HDHP transition to $942 for those who delivered after, compared with nearly constant values ($262 to $282) for women who remained inHMO  plans, a relative increase of 106% (

	Achievement of quality standards for early and frequent prenatal care as well as timely postpartum care did not vary significantly across study groups (Table 2). Nearly all women received early prenatal care, and the majority accessed recommended prenatal and postpartum care. Likewise, maternity care utilization was similar across groups, with an average of 9 to 10 prenatal visits and 7 to 8 sonograms (mode = 4; 25% of women had >10 sonograms) and an average delivery hospitalization stay of about 4 days (3 days for vaginal deliveries; 5 days for caesarean deliveries).

	As expected, pregnancy and delivery complications, high chronic disease scores, and caesarean and preterm deliveries were generally associated with higher costs and utilization (

	This study is the first to characterize the impact of HDHPs on maternity care services. Our findings demonstrate that transitioning from an HMO to an HDHP was associated with increased out-of-pocket costs but stable utilization and quality for maternity and delivery care. Out-of-pocket costs represented a small percentage of total maternity costs for both HMO (4%) and HDHP (11%) members and were in the same range as national averages (6% of total costs).

	An overall reduction in costs associated with better consumer choice is one of the goals of HDHPs and other consumer-driven healthcare models.

	Our findings have important implications for state and federal policy. Recently enacted healthcare reform legislation will establish state-sponsored health insurance exchanges, and policy makers must set minimum coverage criteria for included health plans. Given that previous studies have detected adverse impacts of cost sharing on the poor, further research is needed to investigate how HDHPs affect maternity care among lower-income employed women, who might increasingly be enrolled in HDHPs because of the Patient Protection and Affordable Care Act’s individual mandate for health insurance coverage and taxation of more generous “Cadillac” health plans.

	This study has a number of important limitations. Data on the prevalence of HRAs, employer contributions to such accounts, and employee premium payments were not available, limiting our ability to examine overall cost impacts on healthrelated expenditures. However, as above, we expect a very limited prevalence of HRAs among the small employers included in our cohort. Although it would be more broadly generalizable to include information on PPO members, we were not able to do so in this study. The primary differences compared with PPOs are that HMOs generally have a more limited provider network, lower potential out-of-pocket maternity costs, and higher acn tuarial values.

 These  differences are important when interpreting study results in the national context, and future studies should include PPOs with and without high deductibles. Womenin our study were insured primarily through small employers. HDHP benefit structures and cost-sharing requirements may differ in the small, individual, and large employer group markets.

To examine how enrollment in highdeductible health plans (HDHPs) affects use of well-child visits relative to traditional plans, when preventive care is exempt from the deductible.

We selected children aged <18 years enrolled in a large Massachusetts health plan through employers offering only 1 type of plan. Children were in traditional plans for a 12-month baseline period between 2001 and 2004, then were either switched by a decision of the parent’s employer to an HDHP or kept in the traditional plan (controls) for a 12-month follow-up period. Preventive and other office visits were exempt from the deductible and subject to copayments, as in traditional plans. The primary outcome was whether the child received well-child visits recommended for the 12-month period. Using generalized linear mixed models, we compared the change in receipt of recommended well-child visits between baseline and follow-up for the HDHP group relative to controls.

We identified 1598 children who were switched to HDHPs and 10,093 controls. Between baseline and follow-up, the mean proportion of recommended well-child visits received by HDHP children decreased slightly from 0.846 to 0.841, and from 0.861 to 0.855 for controls. In adjusted models, there was no significant difference in the change in probability that recommended well-child visits were received by HDHP children compared with controls (P = .69).

Receipt of recommended well-child visits did not change for children switching from traditional plans to HDHPs that exempt preventive care from the deductible.

There is concern that increased costs in high-deductible health plans and confusion over which services are subject to the deductible will cause enrollees to underuse important services such as pediatric preventive care.

Receipt of recommended well-child visits did not significantly change for children switching to high-deductible plans compared with those remaining in traditional plans when preventive care was exempt from the deductible.

Use of well-child visits appears to be preserved in high-deductible plans when preventive care is exempt from the deductible.

Rising healthcare costs have led to greater cost-sharing requirements in health insurance plans. Among the newer cost-sharing strategies are high-deductible health plans (HDHPs). Such plans have lower premiums than traditional plans and attempt to control costs by requiring that enrollees assume more responsibility for healthcare costs through annual deductibles that often exceed $1000 per individual and $2000 per family. Adoption of  HDHPs has risen among employers and employees, including those with children.

 In 2007, 18% of privately insured children were enrolled in HDHPs.

Prevalence of HDHPs is likely to rise, as such plans have been part of health reform proposals to make lower-cost coverage available. Proponents of HDHPs believe that increased responsibility for healthcare costs will lead enrollees to reduce unnecessary care, use more cost-effective services, and increase preventive care use.

 Others worry that out-of-pocket costs in HDHPs may lead to underuse of needed care. More than 20 years ago, the RAND Health Insurance Experiment showed that adults and children used fewer services of all types, including preventive care, with increased cost-sharing and deductibles.

 Some recent data in adults show that use of preventive services in HDHPs is similar to use in traditional plans.

 However, other data suggest a reduction in preventive services such as screening colonoscopy.

 Current research on the effects of HDHPs on children’s use of healthcare is limited, although the response to cost-sharing may differ between children and adults.

Many HDHPs exempt pediatric preventive services such as well-child visits and immunizations from the deductible, although anywhere from 8% to 53% of enrollees in HDHPs have plans where preventive care is subject to the deductible.

 Exempting important services such as preventive care from high levels of cost-sharing, a strategy known as value-based insurance design, has been suggested as a means of preserving their use in the presence of increased cost-sharing.

 However, the complexity of HDHPs may cause confusion for enrollees over which services are or are not subject to the deductible, making them more likely to forego services they mistakenly believe are subject to the deductible.

 Enrollees in HDHPs that exempt preventive visits from the deductible also may be dissuaded from seeking them because of unexpected deductible costs for laboratory or other services that arise during preventive visits.

Our study sought to examine whether children in HDHPs that exempt preventive services from the deductible were as likely to receive recommended well-child visits as children in traditional plans.

This study was designed as a pre—post comparison between groups, comparing children who were switched into HDHPs with those who were kept in traditional health plans. The study used health plan enrollment and claims data from Harvard Pilgrim Health Care, a nonprofit New England health plan with more than 1 million members who receive care in a variety of organizational settings. This study focused on employer-sponsored plans in Massachusetts.

In March 2002, Harvard Pilgrim began offering HDHPs with family deductibles ranging from $1000 to $4000 per year. The benefit structure was similar across HDHPs studied. Services subject to the deductible included emergency department visits, diagnostic tests, hospitalizations, and therapeutic procedures (eg, physical therapy). Office visits (including well-child visits) were subject to a $20 copayment and were excluded from the deductible (ie, families paid only the copayment and not the full cost, even if the deductible limit had not been reached). Prescription drugs also were exempt from the deductible and subject to copayments. Preventive services such as immunizations, routine hemoglobin and lead levels, and tuberculosis screening were covered at no cost. A Health Reimbursement Arrangement (an account for out-of-pocket healthcare expenses) was available but offered by few employers.

In contrast, traditional HMO plans in Harvard Pilgrim lacked deductibles and had copayments for emergency department visits and office visits (including well-child visits), full coverage for preventive care and diagnostic tests, and limited cost-sharing for hospitalizations. Office visits and prescription drugs had copayments that were similar to those in the HDHPs, with most office visit copayments between $10 and $20.

This study was approved by the institutional review board of Harvard Pilgrim Health Care.

As part of a larger study of families in HDHPs,

 we identified families with at least 1 child aged 18 years or younger who were enrolled together in the same health plan account in a traditional Harvard Pilgrim HMO for a 12-month baseline period anytime between April 2001 and June 2004, and then were switched to an HDHP for a follow-up period of at least 12 months. Families were excluded if (1) a family member was >65 years of age or (2) a family member did not have continuous enrollment through the same employer for the 24-month period. We identified a control group of families with at least 1 child aged 18 years or younger who remained in a traditional Harvard Pilgrim HMO plan for at least 24 months. To reduce selection effects, we limited our study to families who were insured through employers offering only 1 type of health plan; that is, they did not offer a choice between Harvard Pilgrim and other health plans, or a choice between different types of Harvard Pilgrim plans (89% of families in this population). In this way, the choice to switch to an HDHP (or to remain in a traditional HMO) was made at the employer level, not at the employee level. National data suggest that enrollees without a choice of plans constitute half of the population in HDHPs.

We selected all eligible families that were switched by the employer to an HDHP. For each HDHP family, we selected 8 control families whose employer continued to offer only a traditional HMO plan; control families were matched only on the basis of contemporaneous enrollment periods and were otherwise randomly selected. The date the family switched to an HDHP was assigned as the index date separating the baseline and follow-up periods. Controls were assigned the same index date as their matched HDHP family. Some control families were subsequently excluded if a family member no longer met the age or continuous enrollment criteria for the 24-month period around the assigned index date. Children younger than age 18 years on the index date were selected from this population for the study sample.

The primary outcome of interest was whether recommended well-child visits were received by a child in a 12-month period (baseline or follow-up) based on an algorithm adapted from other studies.

 Well-child visits were identified from claims data as having either (1) one of the following 

 codes: 99381, 99382, 99383-99385, 99391, 99392, 99393-99395, or 99432; or (2) one of the following 

International Classification of Diseases, Ninth Revision 

) codes: V20.2, V70.0, V70.3, V70.5, V70.6, V70.8, or V70.9.23 Based on the child’s age and recommendations of the American Academy of Pediatrics,

we determined the minimum number of well-child visits that should be received over the prior 12 months (

). In calculating the minimum number of visits recommended, we allowed a 1-month grace period for receiving visits through the 24-month visit.

 However, a child receiving a visit more than 1 month late could still end up receiving the recommended number of visits by the end of the 12-month period. For the yearly visits after the 24-month visit, we allowed the child to have a visit anytime over a 12-month period.

 As American Academy of Pediatrics recommendations do not include visits at age 7 or 9 years, we did not require a well-child visit for children aged 7.00 to 9.99 years. Based on the child’s age at the end of the baseline and follow-up periods, we calculated the number of well-child visits received and the number recommended for that 12-month period.

The primary predictor variable was whether the child was switched to an HDHP or remained in a traditional plan. Other covariates included clinical and neighborhood-level socioeconomic variables. We used geocoded addresses from enrollment files to obtain data on each family’s census block group from the 2000 United States Census. We defined a family’s neighborhood (census block group) as high poverty if 10% or more of residents had incomes below the federal poverty level,

 as low education if 25% or more of residents aged 25 years or older lacked a high school degree,

 and as predominantly black if 66% or more of residents were black.

Using Harvard Pilgrim data, the parent’s employer was categorized as small (<50 employees) or large (>50 employees). We used the Johns Hopkins Adjusted Clinical Group (ACG) System to measure morbidity for each child using age, sex, and 

 diagnosis codes from claims from the 12-month baseline period.

 This system assigns a morbidity weight standardized across a reference population of adults and children that is scaled around an average morbidity of 1.0, with higher scores indicating sicker patients.

 We identified chronic conditions during the baseline period using the Chronic Condition Checklist created by researchers at the Johns Hopkins School of Public Health. The Checklist uses 

codes from claims data to identify conditions expected to last more than 12 months and have a substantive impact on future health or functional status.

For bivariate analyses, we used χ2 tests to compare baseline child characteristics between the HDHP group and controls, and Fisher’s exact test in cases where frequencies were small. In unadjusted and adjusted analyses, we modeled the probability of receiving recommended well-child visits using logistic regression. Generalized linear mixed models were used to account for multiple recommended visits per child, for multiple children per family, and for subjects being measured in both baseline and follow-up periods. The model included study group (HDHP or control), study period (baseline or follow-up), and an interaction term between study group and study period. The interaction term reflects the difference between groups in the baseline—to–follow-up difference, or the difference in differences, and is the key element of interest. For adjusted analyses, covariates in the model were chosen a priori and included age; sex; number of children in the family; whether the child’s neighborhood was high poverty, low education, or predominantly black; morbidity weight; presence of a chronic condition; office visit copayment; and index year. To make the findings more interpretable, we used this model to calculate the predicted probabilities of receiving recommended well-child visits in the baseline and follow-up periods for children in the HDHP and control groups. To illustrate the range of predicted probabilities, we calculated predicted probabilities for children of different ages. Other covariates were held at their mean or modal values. We then used the model-derived predicted probabilities for the HDHP and control groups for the baseline and follow-up periods to calculate the difference between the HDHP and control groups in change from baseline to follow-up [(predicted probability HDHP

)]. If children were aged 7.00-9.99 years at the end of the baseline or follow-up periods, they were not included for that study period, as there is no recommended visit for 7- and 9-year-olds.

We identified 1598 children who were switched to HDHPs through 126 employers and 10,093 control children who remained in traditional plans through 945 employers. Employers ranged in size from fewer than 9 employees to more than 1000. 

 shows the characteristics of children in the HDHP and control groups in the baseline year when both groups were in traditional plans. Although few children lived in neighborhoods that were low education or predominantly black, children in the control group were significantly more likely than those in the HDHP group to be from such neighborhoods, a finding consistent with other data that HDHP enrollees are more likely to have higher education and less likely to be black.

 Children in the control group were also more likely to have above-average morbidity, but there was no significant difference in the percentage with a chronic condition. Children in the HDHP group were significantly more likely than control children to have parents who worked for small employers. There were no significant differences between the HDHP and control groups in other measured characteristics. In unadjusted analyses for children in the HDHP group, the mean proportion of recommended well-child visits received declined from 0.846 in the baseline year to 0.841 in the follow-up year (data not shown). For control children, the mean proportion of recommended well-child visits received declined from 0.861 to 0.855. Compared with controls, the change in proportion of recommended well-child visits received from baseline to follow-up for HDHP children was minor (only 0.001 higher) and not significant (P = .968 from the model without covariates).

In adjusted analyses, we did not see a significant change in the receipt of recommended well-child visits from baseline to follow-up for children switching to an HDHP compared with children staying in a traditional plan (P = .69 for the interaction term) (

). Of note, receipt of recommended well-child visits was significantly associated with younger age, living in a neighborhood that was not high poverty or low education, and having a chronic condition. Compared with changes from baseline to follow-up for controls, the predicted probability of receiving recommended well-child visits after switching to an HDHP was only slightly smaller (

). This finding was similar across age groups, ranging from 0.001 less for children younger than age 3 years to 0.008 less for children aged 13-18 years.

This study found that use of recommended well-child visits did not significantly change for children switching to an HDHP compared with those remaining in traditional plans. In these HDHPs, where well-child visits were not subject to the deductible, differences in receipt of recommended well-child visits compared with traditional plans were minimal.

Our study confirmed in a pediatric population the findings from studies in adult populations that show no change in preventive care use in HDHPs that exempt such services from the deductible.

 This is reassuring given other data indicating that preventive care and office visits may decrease in HDHPs even when these visits are exempt from the deductible, perhaps due to confusion about which services are subject to the deductible or attempts to avoid other associated costs where the deductible might apply.

When preventive care was exempt from the deductible, well-child visits did not change for children switching from traditional plans to high-deductible plans.


Stephen B. Soumerai, ScD

Articles

High-Deductible Insurance: Two-Year Emergency Department and Hospital Use

High-Deductible Health Plans and Costs and Utilization of Maternity Care

High-Deductible Health Plans and Costs and Utilization of Maternity Care

Use of Well-Child Visits in High-Deductible Health Plans

Improving Laboratory Monitoring of Medications: An Economic Analysis Alongside a Clinical Trial

Changing Medication Use in Managed Care: A Critical Review of the Available Evidence