Infertility Treatment, ART and IUI Procedures and Delivery

Infertility Treatment, ART and IUI Procedures and Delivery Outcomes: How Important is Selection? Abstract: High infertility levels are of great concern in the American society and there has been a recent surge in the use of Assisted Reproductive Technologies (ART) and other infertility procedures such as Intrauterine Insemination (IUI). Women that use these treatments come from a different health risk pool compared to women who are able to conceive naturally, thus creating a selection problem. It is unclear whether less favorable maternal and neonatal outcomes stem from infertility treatment type itself or preexisting health conditions. Using a database on medical claims made by people working mostly in large firms, we first demonstrate that state mandates that require firms to offer health plans with coverage for infertility treatment are associated with increased ART/IUI usage. We then use this state mandate information as instruments for the ART treatment decision to correct for the selection problem. The mother’s pre-existing age and health condition are the most important predictors of poor outcomes such as complications during pregnancy, miscarriage, abortion and ectopic pregnancy rate and neo-natal health, while infertility treatments such as ART and IUI are at best only weakly associated. Acknowledgements: The Authors would like to acknowledge Verisk Health (formerly DxCG, Inc.) for funding and data support and Kevin Lang for the helpful comments and suggestions. 1 1 Introduction Medical treatments that help otherwise infertile couples to get pregnant are highly successful and widely used these days. These treatments range from ovulation-inducing fertility drugs to invasive procedures such as Intrauterine Insemination (IUI) and Assisted Reproduction Technology (ART). While highly successful, there are concerns regarding the safety of these drugs and procedures and their effect on maternal and neonatal outcomes. Specifically, it has been observed that these treatments are associated with higher rates of complications during pregnancy, increased likelihood of a miscarriage, abortion or ectopic pregnancy, and neonates with a poorer health status. A vast literature in the medical field recognizes the fact that while it’s true that women undergoing these treatments tend to have poorer delivery outcomes on average compared to those reproducing spontaneously; it’s also true that women undergoing these treatments come from a different health risk pool compared to their peers. Thus, some part of the worse outcomes is the result of the mother’s pre-existing health conditions, and it is difficult to determine how much is due to the ART treatments themselves. In this paper we examine this issue using instrumental variable approach in order to determine how much of the observed difference is due to the mother’s characteristics and how much is due to the procedures themselves. 2 Many couples face the problem of not being able to start a family naturally due to some form of infertility. The 2002 National Survey of Family Growth reported that of the 62 million women of reproductive age in US, nearly 2% or 1.2 million had infertility related medical appointment within the past 12 months. Another 10% had an infertility-related medical visit at some point in the past. Additionally, 7% of married couples in which the woman was of reproductive age (2.1 million couples) reported that they had not used contraception for 12 months and the woman had not become pregnant. Infertility is often defined broadly as not being able to get pregnant after trying for 12 months and infertility treatments includes services like medical tests to diagnose infertility, medical advice and treatments to help a woman become pregnant, and services other than routine prenatal care to prevent miscarriage. Fertility enhancement strategies such as Intrauterine Insemination (IUI), in which male sperm are injected into the mother, have been available for a long time and are still widely used. Prescription fertility drugs, which affect a woman’s ovulation and fertility, have been used since at least the 1970’s. Since 1981 in the United States Assisted Reproductive Technology (ART) is also commonly used to help women become pregnant, most commonly through the transfer of fertilized human eggs into a woman’s uterus (in vitro fertilization or IVF). Both infertility itself and the use of infertility drugs and infertility treatments may increase the risk of adverse obstetrical and perinatal outcomes. Thomson et.al, (2005), without distinguishing selection from causal effects, suggests a two-fold increased risk of preeclampsia, placental abruption, caesarean section, and vacuum extraction, and a fivefold increased risk of placenta previa in women requiring greater than one year to 3 spontaneously conceive singleton pregnancies, compared with women conceiving without delay. Similarly, annual reports published by Center for Disease Control and Prevention (CDC) on the number of ART cycles performed the previous year and their success rates show that women who have ART are more susceptible to pregnancy complications, miscarriages, ectopic pregnancies and bad neo-natal health outcomes. The major weakness of their analysis is that they cannot distinguish whether women using ART come from a different health risk pool than women who are able to conceive spontaneously or whether using ART itself exposes women and babies to worse outcomes compared to women who conceive spontaneously. Correcting for selection effects is further compounded by the diverse array of infertility treatments available, including less invasive treatments like drug therapies and IUI and more expensive, invasive and painful procedures like ART. The approach we use is straightforward. We start with insurance claims information from a very large sample of 3.6 million women of ages 21-54, from 2000 to 2004, and classify them according to their health conditions, pregnancy status, and neonatal outcomes. We separately look at drug and medical claims to identify procedures associated with various types of fertility treatment, and describe these patters using univariate and graphical techniques. Separately, we review state level data sources to identify states that have implemented mandates broadening or restricting coverage of against various levels of fertility treatment, and use this to create state and year dummies for fertility treatments. After first showing that these state-year mandates are significant predictors of rates of fertility treatment, we use these variables as instruments for reexamining the question of 4 whether fertility treatments are associated with elevated risks of bad outcomes for both mothers and babies. 2 Related Literatures Earlier studies in this area have generally been inconclusive about the effects of infertility treatments on perinatal outcomes. This occurs either because they have very little data or because they use inferior estimation strategies to what we propose. Most studies do not use proper controls in the sense that they do not compare ART patients with similar infertile couples that conceived spontaneously. Matching controls is difficult to achieve because many couples never go through the testing needed to determine that they face infertility challenges. The Society of Obstetricians and Gynecologists of Canada (SOGC) published a guideline in 2006 based on comprehensive survey of literature relating to ART and perinatal outcomes from 1995 to 2006 (Allen and Wilson, 2006). The first recommendation of this guideline strengthens our point on the selection problem and is worth noting: “Spontaneous pregnancies in untreated infertile women may be at a higher risk for obstetrical complications and perinatal mortality than spontaneous pregnancies in fertile women. Further research is required to clarify the contribution of infertility itself to adverse obstetrical and perinatal outcomes” Several papers from the medical literature discuss maternal and neonatal outcomes following ART. Schieve, et al. (2003) use data on ART patients in US clinics between 1996 and 1998 and show that ART does not pose a risk for spontaneous abortions among 5 pregnancies conceived using ART. The results are based on a logistic regression model that estimates the likelihood of spontaneous abortions for various treatments. In another paper (Schieve, et al. (2004)) perinatal outcomes were studied to conclude that singletons born after ART remain at increased risk for adverse perinatal outcomes like low birth weight. Winter, et al. (2002) use a much smaller sample of 1196 clinical pregnancies from three reproductive centers in Australia and perform a multivariate logistic regression analyses to conclude that the risk of early pregnancy loss following ART is the same as that in general population. In another study, Clayton, et al. (2006) conclude that ectopic risk among ART pregnancies varied according to ART procedure type, reproductive health characteristics of the woman carrying the pregnancy, and estimated embryo implantation potential. Hansen, et al. (2004) conducted a systematic review of 25 papers published by 2003 that addressed the association between birth defects among children conceived using ART and ART treatment type. They also did a fixed-effects multivariate analysis to obtain pooled odds ratio for 7 most appropriate studies and found a statistically significant 30-40% increased risk of birth defects associated with ART. However, they themselves note that part of this increased risk could be due to the underlying causes of infertility in couples seeking treatment rather than the treatment itself. Few papers have documented outcomes following procedures like the IUI, which is viewed as being much safer than ART, and even fewer studies have contrasted outcome rate for ART and IUI procedures, even though they are often used as substitutes. A study in Belgium (Sutter, et al. (2005)) compared 126 IVF patients with 126 IUI 6 patients to conclude that pregnancy outcomes after IUI and IVF are not different from each other. They also conclude that since there is no reason to believe that the IUI technique in itself leads to an increased obstetric or neonatal risk, the worse pregnancy outcome after IVF as compared with spontaneous conceptions is due to the specific mother characteristics rather than to the use of IVF itself. There has been some research on the effects of the presence of insurance mandates on increased use of infertility treatments. Buckles (2005) use the presence of mandates and multiple birth rates in a state as a proxy for increased ART usage in that state. She concludes that fertility decisions are affected by the availability of infertility treatments. Schmidt (2007) uses a difference-in-differences approach to conclude that the mandates significantly increase first birth rates for women over 35, and these results are robust to a number of specification tests. 3 Motivations 3.1 Background on Assisted Reproductive Technologies (ART) Many couples who want to start a family face with the disease of infertility. In US, about 12% of women of childbearing age have received an infertility service (ASRM, 2007). Depending on the diagnosis, a gamut of treatment options is available to these women to medically assist them in their childbearing endeavor. These treatments range from ovulation inducing drugs to intrauterine insemination to in vitro fertilization. For this 7 paper we divide up these treatments into two broad categories that differ in their expense, intrusiveness, and success rates: Assisted Reproductive Technologies (ART), and other fertility treatments (FERT) which includes all fertility treatments (including drug therapies) other than ART. According to the Center for Disease Control (CDC), ART includes all fertility treatments in which both eggs and sperm are handled. In general, ART procedures involve surgically removing eggs from a woman’s ovaries, combining them with sperm in the laboratory, and returning them to the woman’s body or donating them to another woman. They do not include treatments in which only sperm are handled (i.e., Intrauterine—or artificial—insemination (IUI)) or procedures in which a woman takes drugs only to stimulate egg production without the intention of having eggs retrieved. The types of ART include the following:  IVF (in vitro fertilization) involves extracting a woman’s eggs, fertilizing the eggs in the laboratory, and then transferring the resulting embryos in to the woman’s uterus through the cervix. For some IVF procedures, fertilization involves a specialized technique known as intracytoplasmic sperm injection (ICSI). In ICSI a single sperm is injected directly into the woman’s egg.  GIFT (gamete intrafallopian transfer) involves using a fiber-optic instrument called a laparoscope to guide the transfer of unfertilized eggs and sperm into the woman’s fallopian tubes through small incisions in her abdomen 8  ZIFT (zygote intrafallopian transfer) involves fertilizing a woman’s eggs in the laboratory and then using a laparoscope to guide the transfer of the fertilize eggs into her fallopian tubes. In addition, ART often is categorized according to whether the procedure used a woman’s own eggs or eggs from a donor and according to whether the embryos used were newly fertilized or previously fertilized, frozen and then thawed. Because an ART procedure includes several steps, it is typically referred to as a cycle of treatment. 3.2 FERT: Fertility treatments other than ART Typically a patient diagnosed with infertility will first go through a series of less invasive treatments such as drug therapy and/or Intra-Uterine Insemination before moving on to ART. For example a woman may begin with therapy in which she takes hormonal drugs along with timed intercourse in order to achieve a pregnancy. If this treatment fails, or if she is identified with other medical problems or if there is a presence of male factor infertility the couple may be given an IUI treatment. In this treatment washed sperm is injected directly into the woman’s cervix in order to give the sperm a better chance to mate with the eggs. Sometimes a tubal surgery that opens fallopian tubes may be required on the woman to help facilitate natural conception. Since these procedures do not involve egg retrievals, they fall outside the purview of assisted reproductive treatments. However, it’s very important to study the effectiveness and outcome rates following these treatments as success rates in these treatments affect selection into and perhaps the success of ART. Cost considerations and patient preferences may also affect the sequencing of treatments. While some couples paying out of pocket may bypass these 9 FERT treatments and move straight on to IVF in order to increase their chances of speedy success, other couples who are covered under insurance may be forced by health plans to take a minimum number of such treatments before moving on to IVF. Although one cannot be sure of the sequence in which these treatments are administered to a patient, the most common progression of treatments is from less invasive and costly to more invasive and costly. Henceforth we refer to all the fertility procedure or drugs associated with diagnoses that do not fall in the category of ART as ‘FERT’. We explain later on the creation of these variables in our sample and conduct our analyses while jointly controlling for both FERT and ART. 3.3 National Summary of ART success rates for 2004 Tables 3 and 4 give some comparisons of the national data to our sample. About 74% of ART cycles carried out in 2004 used fresh non-donor eggs or embryos. The average age of women using ART services was 36. 41% of women using ART were less than 35 years and about 20% were 40 years or more. Majority of the women younger than 35 used their own eggs while almost 50% of women 40 and older used donor eggs. About 34% of ART cycles resulted in a clinical pregnancy. 20.3% of ART cycles resulted in a single-fetus pregnancy, 11.3% resulted in a multiple-fetus pregnancy and 2.8% resulted in a miscarriage or ectopic pregnancy. Of all the women who achieved a clinical pregnancy, 82% had a live birth. Of these above mentioned live births, 60% were singleton pregnancies, 29% were twins and about 5% were triplets or more. A woman’s age is the most important factor affecting the chances of a live birth when her own eggs 10 are used. Success rates declined steadily after the mid 30s as fertility declined with age. The rates of miscarriages also increased steadily with age. 3.4 Mandate variable as an instrument for ART A key problem with implementing a simple regression studying the association of ART with each of the above mentioned variables is that ART is an endogenous variable. For example, women who have ART have higher rates of complications but it is also true that the women going for ART come from a different health risk pool than women getting pregnant unassisted. In general, national data shows that women going for ART are older in age and may have some other pre-existing health conditions that prevent them from natural conception. Thus ART affects the complications rate but existing complications lead to greater ART utilization. This circularity makes it important for us to look for instruments that are correlated with ART but not with any of our dependent variables. In this paper we use information on state mandates to treat this selection problem. 3.5 State Laws regarding Insurance coverage of infertility treatments including ART Fifteen state legislatures in US have passed laws mandating coverage for infertility treatments. However, there is a wide variation in coverage across these states. Table 1 explains this variation in greater detail. Some states require coverage of only select treatments while others offer more comprehensive coverage. There are two broad forms 11 of state mandates—one is a ‘mandate to offer’ which requires insurers to offer coverage of infertility treatments; however employers are not required to include such coverage in their benefit plans. The other is a ‘mandate to cover’ which requires health insurance companies to include coverage of infertility treatments as a benefit in every policy. Currently, only Texas and California have a mandate to offer, while the remaining thirteen states have mandate to cover laws. These states can be further divided into four categories based on the differences in coverage as follows: IVF Coverage only: a number of states’ mandates address only IVF. For example, Arkansas requires group health insurance companies—exempting HMOs—to cover the cost of IVF. However, there is a marital restriction on cost coverage and the woman is required to use her spouse’s sperm. Legislation in Maryland also covers only IVF with a limit of three IVF attempts per live birth and a lifetime maximum level of $100,000 while Hawaii covers only one IVF treatment with a number of preconditions that need to be satisfied. Insurance companies in Texas that offer infertility treatment benefits must provide the same amount of coverage for IVF as for any other pregnancy-related procedures. Furthermore, the patient should have a five year history of infertility, have tried other treatments and must use her spouse’s sperm. Exemption of IVF: In other states, legislation specifically excludes IVF. California, for example, requires insurance companies to offer coverage of infertility treatments, including diagnostic testing and medication. Even though California’s mandate does not 12 include IVF, they cover gamete intrafallopian transfer (GIFT), a treatment similar to IVF. Patients are eligible for treatment after one year of infertility or if their infertility is caused by a medically recognized condition. California’s mandate does not include age or marital status restrictions. Similarly, in New York insurance companies do not have to provide IVF, GIFT or ZIFT and the age restriction is from 21-44. They do need to cover diagnostic tests and infertility procedures such as a tubal surgery and infertility drugs. Hence mandates in these states would be expected to increase the rate of FERT but not ART. Comprehensive Coverage: the states of Massachusetts, Rhode Island and Illinois provide more comprehensive coverage. Illinois requires any policy that covers more than 25 people to provide coverage for diagnosis and treatment of infertility after one year of infertility. IVF, GIFT and ZIFT are covered only if less expensive treatments have failed. The same is true for New Jersey and Massachusetts. Rhode Island allows co-payments which do not exceed 20% but coverage is limited to married individuals. Connecticut also offers comprehensive coverage but with age and enrollment restrictions. Additionally, Connecticut is the only state to limit the number of embryos transferred per cycle. Preventative Services only: Montana, Ohio and West Virginia have laws that require HMOs to cover infertility services as part of a plan’s “preventative health care services”. The laws are jotted down in fairly broad terms and thus their scope is unclear. It is 13 unlikely that ARTs are covered since they do not prevent infertility, but, rather, are designed to remedy the problem of infertility. ERISA Preemption The Employee Retirement Income Security Act (ERISA) regulates employee benefit plans, such as employer-sponsored health benefits. ERISA distinguishes between insured and self-insured private health care plans and preempts the self-insured plans. These plans are not subject to state laws regulating insurance and are therefore attractive to many employers. In 2005, the majority of covered workers were in a plan that was completely or partially self-insured. This means that the majority of covered workers are unable to benefit from state laws mandating insurance coverage for infertility treatments. However, most firms may offer these benefits anyway as a means of competing with other peers. Some firms for instance will have one high premium plan that covers infertility treatment but is less generous otherwise. This will ensure that only those employees who really need these services will opt for such plan. Since the cost of these services is high enough, one can imagine that a firm can compete with other firms for employees on the basis of such benefits. 3.6 Why mandates should be good instruments The use and success of infertility treatments are steadily increasing. Between 1996 and 2003, the number of infertility treatments performed each year doubled. The costs of infertility treatments vary depending on the complexity of problem and therapy used. One 14 month of prescription medications to stimulate ovulation can cost anywhere from $40 to $3000, excluding the costs of monitoring the effects of these drugs and other medications often taken in conjunction with these drugs. A typical IUI cycle would cost about $5000 for medication, patient monitoring and the insemination procedure. ARTs are even more expensive with a typical cycle costing upwards of $10,000. In addition, patients who choose these treatments typically undergo numerous cycles before becoming pregnant— if at all—posing a substantial financial burden on such couples. Thus, insurance coverage for IUI and other less invasive and the more complex ART treatment is bound to affect decision making regarding utilization of these treatments. The presence of an infertility insurance mandate in a state should increase the utilization of these procedures in those states as more couples can internalize the cost by purchasing insurance. At the same time, the presence of a mandate does not affect health outcomes during pregnancy and after delivery. Thus, the presence or absence of a mandate seems to be a good instrument as it affects the rate at which infertility treatments are used but not the rate of their success. It must also be noted that whether or not a mandate will be effective would depend on how generous it is in specifying the minimum coverage. For example, even though states like Montana, Ohio, West Virginia, and Louisiana have mandates, it is unlikely that they will provide enough incentives for insurance companies to offer these treatments and hence are unlikely to affect FERT and ART utilization. 15 4 Data Description 4.1 MarketScan Database Overview5 Our analysis uses the commercial claims and encounters database on privately insured patients in Thompson’s MEDSTAT MarketScan database. This represents the inpatient and outpatient healthcare service use of individuals nationwide who are covered by the benefit plans of large employers, health plans, and government and public organizations. The MarketScan database links paid claims and encounter data to detailed patient information across sites and types of providers, and over time. The annual medical database includes private sector health data from approximately 100 payers. These data represent the medical experience of insured employees and their dependents for active employees, early retirees, COBRA continuees and Medicare-eligible retirees with employer-provided Medicare Supplemental plans. No Medicaid or Workers Compensation data are included. The medical and surgical claims consist of inpatient admissions and both inpatient and outpatient services. Outpatient prescription drug claims are also covered. 4.2 Sample Selection and Variable Creation Using the commercial claims and encounters database, for each year, we first selected all women aged 21-54. After creating the state variable based on enrollee zip codes, we dropped all women who could not reliably be grouped into a particular state. We then 16 made a comprehensive list of all procedure codes, diagnosis codes and drug codes that indicated an infertility treatment. There was some fuzziness in identifying women with infertility claims based on codes. For example, a large number of women had submitted claims for the leading hormone drug used for ovulation induction, gonadotropin, but did not have any infertility diagnosis over the five year period. In such cases, we included only those women in our infertility group who were both taking gonadotropin and also had an infertility related diagnosis/procedure. For any woman who was identified as having an infertility issue, we retained information on all her claims—not just infertility related claims We then grouped the procedure, diagnosis and drug codes into eight categories: P0 = no diagnoses, procedure or drug indicating infertility. P1 = 1 if any diagnosis indicating infertility P2 = 1 if any drug claims that are specific to infertility i.e., a drug claim associated with an infertility diagnosis. P3 = 1 if any tests or procedures at all related to infertility treatment (almost all of our sample unless drug only people) P4 = 1 if any artificial insemination procedure indication P5 = 1 if any egg removal procedure regardless of any indication of a transfer to the mother (includes most incomplete procedure where egg was removed but not transferred either because the patient was an egg donor or because of cryopreservation) P6 = 1 of Any IVF/GIFT/ZIFT 17 P7 = 1 if Any micro technique used. P8 = 1 if any cryogenic procedure used With the exception of P1, the Pi variables were not created to be mutually exclusive. Rather, these are just binary flags about whether any of these events occurred during a given year. One approach to further analyze these procedure codes is to try to put them into mutually exclusive categories, even if women do not always do them in a given order. The numbering system we use here is one natural order that generally goes from less expensive and complex to more expensive and complex. It is true that many women may skip one or more of these steps. For example, a woman may start off with an ovulation induction in hope of getting an intrauterine insemination (IUI) but if during ultrasound testing she shows a potential risk of hyper-stimulation, her IUI cycle may be cancelled and she may be bumped up to an IVF cycle. Plus, they may occur in different years so that it looks like one or more have been skipped. Still, having some form of hierarchies may prove useful for further analyses such as finding which stages of infertility related process are most affected by mandates. We thus create the following hierarchical categories using the Pi variables. D0 = 1 if the women did not have any procedure, diagnosis or drug related to infertility— No ART or FERT women D1 = 1 if only a diagnosis of infertility with no procedures or drugs D2 = 1 if fertility related drugs associated with an infertility diagnosis but with no fertility procedures 18 D3 = 1 if any procedures with no higher level removal or transfers D4 = 1 if any artificial insemination procedure with no egg removal or transfer D5 = 1 if egg removal without any transfer into a mother D6 = 1 if any IVF without any micro technique D7 = 1 if any IVF plus micro technique used D8 = 1 if any cryogenic procedure used Once we set up these variables we regrouped them to form measures of infertility treatment. We grouped D1-D4 into a variable called ‘FERT’ which indicates whether the woman had infertility related diagnosis or procedure or drug associated with an infertility diagnosis but did not get any ART. We also created a variable called ‘ART’ which indicated the presence of any D5, D6, D7 or D8. This was a natural division since in general; all procedures, drugs and diagnosis in the FERT variable tend to be less invasive and less costly than anything that shows up in ART. In most cases, they also act as baseline procedures for evaluating whether or not the woman should be treated with ART. The health outcomes for new mothers and neo-natal are also deemed to be different for these two groups. Mother’s Risk Score: DxCG compiles risk scores for each enrollee in the MarketScan data. These risk scores were calculated using the DxCG risk adjustment classification system version 6.2 which uses diagnoses to characterize the medical conditions of each individual with 184 binary flags, and then generates predictions using other samples of expected costs. The risk 19 score used here was the concurrent risk score, predicting the cost of health services used during the same year as the delivery, except that we omitted the components of the risk score related to maternity conditions. Hence this risk score reflects the expected cost of non maternity related health care spending for each woman. 20