|Home | About | Journals | Submit | Contact Us | Français|
Depression imposes enormous burdens on the elderly. Despite this, rates of initiation of and adherence to recommended pharmacotherapy are frequently low in this population. Although initiatives such as the Medicare Modernization Act (MMA) in have improved seniors' access to antidepressants, there are concerns that the patient cost-sharing incorporated in the MMA may have unintended consequences if it reduces essential drug use. Age-related pharmacokinetic and pharmacodynamic changes could make seniors particularly vulnerable to antidepressant regimens used inappropriately to save costs, increasing their risks of morbidity, hospitalizations, and nursing home placements.
Two sequential large-scale “natural experiments” in British Columbia provide a unique opportunity to evaluate the effect of cost sharing on outcomes and mental health service use among seniors. In January 2002 the province introduced a $25 Canadian copay ($10 for low-income seniors). In May 2003 this copay policy was replaced by a second policy consisting of an income-based deductible, 25% coinsurance once the deductible was met, and full coverage once an out-of-pocket ceiling was met. The transition between the two policies is analogous to what many U.S. seniors experience when they transition from private insurance requiring copays to Medicare Part D requiring deductibles and coinsurance.
To evaluate whether declines in antidepressant initiation after the introduction of two drug cost-sharing policies in British Columbia were associated with increased use of physician services, hospitalizations, and nursing home admissions among all British Columbia residents aged 65+.
Records of physician service use, inpatient hospitalizations, and residential care admissions were obtained from administrative databases. Population-level patterns over time were plotted, and effects of implementing the cost-sharing policies examined in segmented linear regression models.
Neither policy affected the rates of visits to physicians or psychiatrists for depression, hospitalizations with a depression diagnosis, or long-term care admissions.
The cost-sharing policies studied may have contained non-essential antidepressant use without substantially increasing mental health service utilization. However, it is possible that the policies had effects that we were unable to detect, such as increasing rates of visits to social workers or psychologists or forcing patients to reduce other spending. Further, the sequential implementation of the policy changes, makes it difficult to estimate the effect of a direct change from full coverage to a coinsurance/income-based deductible policy.
It may be possible to design policies to contain non-essential antidepressant use without substantially increasing other service utilization or adverse events. However, because undertreatment remains a serious problem among depressed elderly, well-designed prescription drug policies should be coupled with interventions to address under-treatment.
Depression imposes enormous burdens on the elderly(1-3). Despite this, their depression care—consisting almost entirely of antidepressant pharmacotherapy(4, 5)— is frequently problematic(6, 7). Many depressed older persons never initiate appropriate antidepressant regimens and of those who do, fewer than half fill prescriptions for even a month(8-13).
Although initiatives such as the Medicare Modernization Act (MMA) have improved seniors' access to antidepressants, there are competing concerns that these could lead to large, unsustainable expenditures for these drugs(14). Antidepressants are among the most widely prescribed classes of medications in most health care systems(15, 16). Costs of psychotropic drugs have increased 17% annually, far outpacing spending increases on other mental health care and other medications. Because the proportion of spending on prescription drugs is twice as high in mental health care as in general health care, there are particular pressures to control psychotropic costs(17).
For these reasons, prescription benefit programs such as Medicare Part D have attempted to contain costs through copayments, coinsurance, income-based deductibles, and combinations of these(18). Copayments require a fixed amount to be paid for each prescription while coinsurance requires payment of a proportion of the medication price. To prevent coinsurance policies from being unfair to sicker patients requiring more medications(19), most coinsurance policies have an annual deductible or out-of-pocket ceiling with costs above the ceiling being reimbursed. Deductible amounts have also been linked to income in the prior year to allow patients with higher incomes to pay more for medications.
The intended consequence of such cost sharing policies is to reduce non-essential drug use and payers' expenditures, thereby ensuring the fiscal viability of medication assistance programs for seniors. Unfortunately, cost sharing policies can also theoretically drive down essential drug use and have unintended consequences, such as increasing utilization of other non-drug mental health services and general medical care(20, 21). Age-related pharmacokinetic and pharmacodynamic changes could make seniors particularly vulnerable to regimens used inappropriately to save costs, increasing their risks of morbidity, hospitalizations, and nursing home placements(7, 22). Because of these possibilities, the net economic savings to insurers from cost-sharing plans might be less than anticipated(23, 24). Therefore, it is critical to study not only whether cost-sharing policies can help contain drug utilization and costs but also whether they do so at the expense of increasing these unintended outcomes.
This current study evaluated if two sequential large-scale “natural experiments” in cost sharing had unintended consequences on seniors using antidepressants. Prior to 2002, seniors in British Columbia received full coverage for prescription ingredient costs and paid only a $7 pharmacy dispensing fee when filling prescriptions. In January 2002 the province introduced a $25 Canadian copay ($10 for low-income seniors). In May 2003 this copay policy was replaced by a second policy consisting of an income-based deductible, 25% coinsurance once the deductible was met, and full coverage once an out-of-pocket ceiling was met. The transition between the two policies is analogous to what many U.S. seniors experience when they transition from private insurance requiring copays to Medicare Part D requiring deductibles and coinsurance. In a previous analysis(25), we observed that the copayment policy was associated with a drop in frequency of antidepressant initiation, equivalent to 0.38 fewer patients initiating treatment per 1000 per month among the general population and 12.3 per 1000 per month among patients with a prior antidepressant diagnosis. Substitution with the income-based deductible/coinsurance policy was further associated with reductions in the rate of growth of antidepressant initiation and dispensing. The rate of growth in initiation declined by 1.2 starts per 1000 seniors, and overall dispensing was reduced by 14 patient-days per 1000 seniors per month. In the current analysis, we investigated whether elderly patients may have dealt with reduced antidepressant use by substituting other services such as psychiatrist and other physician visits or experienced adverse events such as hospitalizations and nursing home admissions.
Our study population included all seniors (age 65+) living in British Columbia (BC), Canada. Population statistics were obtained from the BC statistics office. Records of physician service use, inpatient hospitalizations, residential care admissions, and deaths for this population were obtained from administrative databases maintained by the British Columbia Ministry of Health. These databases contain diagnostic codes and dates of service, admission, or death. Up to 25 diagnoses for hospital discharges and one diagnosis for each medical service are recorded, with good specificity and completeness(26). Prescription records were obtained from the PharmaNet database, which contains records of all prescriptions dispensed at community pharmacies in BC, regardless of payer, since 1996.
We included all seniors in our study rather than restricting the population to those taking antidepressants prior to the cost-sharing policy in order to capture any adverse outcomes occurring among patients failing to initiate appropriate treatment due to cost concerns.
Our outcomes of interest were rates of physician visits, psychiatrist visits, and hospitalizations for depression over time. We identified physician office visits with an International Classification of Diseases (ICD-9) diagnosis code of depression (293.83, 296.2×, 296.3×, 296.90, 298.0×, 300.4×, 309.0×, 309.1×, 309.28, 311) as well as the subset of these visits that occurred with a psychiatrist, based on the billing physician's recorded specialty. Hospitalizations with a diagnosis depression were identified based on the ICD-9 diagnosis codes listed above. In order to determine whether the BC policy changes affected mental health service use differentially from overall health service use, we also examined trends in physician office visits, psychiatrist visits, and hospitalizations for any diagnosis. Because the policy changes that occurred in January 2002 and May 2003 imposed cost-sharing on all community dwelling seniors, we also examined rates of nursing home admission to determine whether these policies resulted in an increase in nursing home admission rates. For each month from January 1, 1997 to December 31, 2004, we estimated the rate of each event as the number of events divided by the BC senior population.
Outcome rates were plotted over time. Segmented linear regression was used to identify changes in slope or level at the time of the policy changes(27). Our multivariate linear regression models of aggregated monthly rates included a constant term, a linear time trend, a binary indicator for the period after the introduction of the co-pay policy, a linear time trend for time since the co-pay policy, a binary indicator for the period after the introduction of the income-based deductible policy, and a linear trend for time since the income based deductible policy. A Durbin-Watson test was used to determine autocorrelation, and where appropriate, an autoregressive error process was specified. Statistical significance of policy effects was determined from two-sided t-tests.
The British Columbia senior population grew from 493,000 in January 1997 to 590,000 in December 2005. During 2001, the midpoint of the study period, 71,390 seniors received at least one antidepressant prescription. As seen in table 1, these seniors had a mean age of 75.5 and were predominantly female (69.4%). Approximately 9% had seen a psychiatrist in the past year, while less than 1% had been hospitalized for a mental health diagnosis.
Among the entire BC senior population, smoothed monthly rates of physician visits for depression increased from 14.26 per 1,000 seniors in January 1997 to 14.98 in December 2005 (see Figure 1). Visits for depression accounted for approximately 1.4% of all physician visits during this period. Neither the introduction of the copayment policy in January 2002 nor its replacement with the coinsurance/income-based deductible policy in May 2003 had a significant effect on the rate of physician visits for depression. Comparing trends of physician visits for depression to trends of overall physician visits, we observed no differential effect of the policy changes on visits for depression (see Table 2).
Visits to psychiatrists accounted for approximately one third of all physician visits for depression. The number of psychiatrist visits for depression increased from 4.73 per 1,000 seniors in January 1997 to 5.48 in December 2005 (see Figure 2). Overall monthly rates of psychiatrist visits for any diagnosis increased from 9.07 to 12.44 during the same period. Neither the rate of psychiatrist visits for depression nor the rate of overall psychiatrist visits was significantly affected by implementation of the copayment or coinsurance/income-based deductible policies, and we found no evidence of a differential effect across diagnoses.
British Columbia seniors were hospitalized in January 1997 at a rate of 25.77 per 1,000 seniors. Rates of hospitalization declined significantly during the study period, falling to 18.55 per 1,000 seniors in December 2005. Approximately 5% of all hospitalizations had depression coded as a primary or secondary diagnosis. Hospitalizations with depression as a diagnosis also decreased significantly during the study period, from 1.22 per 1,000 seniors in January 1997 to 1.00 in December 2005. Neither policy change had a significant effect on either the overall hospitalization rate or the rate of hospitalizations specifically for depression.
Rates of long-term care admissions tended to fall from 1.89 per 1,000 seniors in January 1997 to 1.5 in December 2005. The rate of long-term care admissions was not significantly affected by implementation of either the copayment or coinsurance/income-based deductible policy.
In this study of a large general population of seniors, a decline in antidepressant initiation that occurred after implementing a copay policy(25) was not associated with increased use of psychiatrist and other physician services or increased hospitalizations for depression and nursing home admissions. Likewise, reductions in the growth of antidepressant dispensing and initiation after substituting a coinsurance/deductible policy were not associated with increases in these service utilization or adverse event outcomes.
Studies that do not find statistically significant changes merit special attention to their statistical power to detect a clinically meaningful change(28). The results of our analysis indicate that we can exclude a copay policy effect on physician visits for depression greater than an increase of 0.92 visits per 1,000 seniors (507 visits per month among all seniors) and a coinsurance effect of an increase greater than 0.8 visits per 1,000 seniors (446 visits among all seniors) with a confidence level of 95%. Similarly our results exclude with a confidence level of 95% an increase of more than 0.12 hospitalizations with a depression diagnosis / 1,000 seniors (65 hospitalization per month among all seniors) under the copay policy and an increase of more than 0.09 / 1,000 seniors (50 hospitalizations per month among all seniors) under the coinsurance policy(29). It is difficult to predict what size effect the previously observed reductions in antidepressant initiation and use might have on physician visits and hospitalizations. For example, monthly rates of antidepressant use decreased by 14 patient-days per 1000 seniors per month under the co-insurance policy. This could correspond to 0.5 fewer patients per 1000 receiving any treatment or a number of patients becoming less adherent. In addition, while there is evidence to suggest that non-persistence may be associated with higher medical costs, these data are relatively limited(30, 31, 32). However, it is possible that small changes occurred and were simply too small to be detected in the current study.
The results of earlier studies on a variety of drug-cost sharing policies have been mixed. Some cost-sharing policies have not been found to increase other health care utilization or adverse events(33-40); other policies limiting reimbursements for prescription drugs have been found to increase these unintended consequences(41, 42), in some cases dramatically(24). Understanding why specific cost-sharing mechanisms can have such different impacts on utilization, health, and economic outcomes is crucial for designing sound prescription drug policies in the future.
Several potential mechanisms might explain how the cost-sharing policies examined in this study could apparently decrease utilization and initiation of a therapeutic medication, without increasing other service use and unintended consequences(43). One possibility is that due to potentially high demand and long waiting lists for mental health services, increased needs may not have translated into more utilization especially of ambulatory services. However arguing against this possibility is our finding that adverse consequences requiring urgent attention such as hospitalizations or long-term care, also did not increase.
Another possibility, which is a desired outcome of benefit managers, is that the policies were able to selectively decrease non-essential antidepressant use but spare essential use. Likewise, the policies may have reduced use of poor quality or ineffective antidepressant regimens while leaving intact use of effective ones. Findings from several prior studies support the hypothesis that cost-sharing can reduce inappropriate medication use.(44) Finally, generous exemption rules for patients and prescribers that accompanied both policies may have helped them achieve their desired effects while minimizing unintended consequences.
These results should be interpreted with several potential limitations in mind. First, it is possible that the cost-sharing policies had unintended impacts on elderly antidepressant users that we were unable to detect. For example, the policies may have driven up the use of other mental health (e.g., psychotherapy visits to non-psychiatrists) or general medical services (e.g., emergency room visits). Likewise, elderly patients may have had to obtain their antidepressant medications by borrowing or by doing without other essential drugs, other forms of health care, and even non-health care necessities (e.g., food, utilities).
Because of the sequential implementation of the policy changes, it is difficult to estimate the effect of a direct change from full coverage to a coinsurance/income-based deductible policy. Although we compared trends during the income-based deductible policy with those during the baseline policy, patterns after the income-based deductible period may have been affected by the introduction of the copay policy. The other comparisons we made—of the coinsurance/income-based deductible policy compared with the copay policy—may underestimate the impact of a because some patients who would have experienced “sticker shock” after the income-based deductible policy went into effect may already have done so after the transition from the full-coverage policy to the copay policy.
A final limitation worth considering is our lack of a concurrent control group. Because the policy changes affected all seniors in British Columbia (with the exception of noncomparable populations, such as seniors residing in nursing homes), no appropriate concurrent control group exists. The validity of an uncontrolled time trend analysis such as ours rests on the assumption that no changes resulting from other interventions or changes in the population occurred at the same time as the interventions of interest.
With these potential limitations in mind, it appears that implementing the two forms of cost-sharing studied here—a copay policy followed by substitution with an income-based deductible/coinsurance policy—may have successfully contained non-essential antidepressant use without substantially increasing other service utilization or adverse events. However it is important to keep in mind that care of depression among the elderly continues to suffer from the dual problems of both non-essential use as well as undertreatment, with the latter being by far the greater public health and public policy concern(8-13). It is unlikely that cost-containment policies alone, no matter how clinically and fiscally sound, would increase utilization of treatments by those with unmet needs. Instead, model programs of screening, outreach, and enhanced care have been developed and proven successful at overcoming barriers to care and improving clinical outcomes among elderly persons with undertreated depression(45). Such programs, together with well-designed prescription drug policies, may ultimately be needed for extremely vulnerable older populations with mental disorders.
This work was supported by grant R01-MH069772 from the National Institute of Mental Health, grant R01-AG021950 from the National Institute on Aging, and grant 5-R01-HS010881-07 from the Agency for Healthcare Research and Quality.
The views expressed in this article do not necessarily represent the views of the National Institute of Mental Health, the National Institutes of Health, the Department of Health and Human Services, or the United States government.