Health plans and physician groups increasingly use sophisticated tools to predict individual patient outcomes. Such analytics will accelerate as US medicine enters the digital age. Promising applications of forecasting include better targeting of disease management as well as innovative patient care approaches such as personalized health insurance and clinical decision support systems. In addition, stakeholders will use predictions to advance their organizational agendas, and unintended consequences could arise. Forecasting-based interventions might have uncertain effectiveness, focus on cost savings rather than long-term health, or specifically exclude disadvantaged populations. Policy makers, health plans, and method developers should adopt strategies that address these concerns in order to maximize the benefit of healthcare forecasting on the long-term health of patients.

Predicting patient outcomes using healthcare forecastingâ€”frequently termed â€œpredictive modelingâ€â€”will accelerate as US medicine enters the digital age.

		The potential benefit of forecasting is facilitation of innovative patient care improvement strategies.

		Â Possible pitfalls include the exacerbation of health disparities, the propagation of interventions that have little evidence of effectiveness, lack of transparency, and a short-sighted focus.

		Health plans, method developers, and policy makers should adopt strategies that address these concerns in order to maximize the benefits of healthcare forecasting for patients.

The future is not what it used to be. Health plans and physician groups increasingly use sophisticated tools to predict individual patient outcomes,

and such analytics will rapidly expand as US medicine enters the digital age. Over the coming decade, the Affordable Care Act will spawn new healthcare organizations and existing ones will convert to electronic health records (EHRs). These stakeholders will dip into the burgeoning reservoir of health data to make predictions that advance their organizational agendas and financial well-being.

Despite widespread and increasing adoption,

 the use of “healthcare forecastingâ€ to predict outcomes of individual patients has received little attention in the health policy community. This commentary reviews potential unintended consequences of such profiling, outlines policies to maximize benefits, and suggests innovative uses.

Three related fields–healthcare forecasting, risk adjustment, and actuarial science–use similar statistical techniques to predict the future behavior of patients, but for different reasons. Healthcare forecasting generally implies predicting an individual's costs or healthcare utilization for interventional purposes such as proactive disease management, patient education, or surveillance to promote population health. Risk adjustment refers to determining comparability between patient populations based on current or predicted healthcare utilization or expenditures. It is most often used to compensate health plans or physician groups fairly based on their patients' morbidity. Actuaries use similar statistical methods to set insurance premiums and to determine coverage eligibility.

Previous literature has explored the uses and pitfalls of actuarial techniques and risk adjustment

; this article instead focuses on the use of healthcare forecasting to intervene (eg, using disease management) with individual patients at risk of high costs or suboptimal outcomes.

HEALTHCARE FORECASTING: BACKGROUND AND TRENDS

Healthcare forecasting of this type–often called “predictive modeling”–is used by all major health insurers, integrated delivery systems, and many other healthcare organizations (

At present, these stakeholders are primarily known to use predictive modeling for disease management–identifying and intervening with individuals likely to incur high costs.

For example, statistical tools might detect the diabetic patients with the highest probability of hospitalization in the following year based on age, coexisting chronic illnesses, medication adherence, and past patterns of care. Less common predicted end points include achievement of specific medical outcomes, process quality indicators, or patient behavior changes.

The variables in statistical prediction models usually come from computerized medical and pharmacy insurance claims and other administrative sources. Data are less commonly drawn from surveys, laboratory results, or EHRs.

Healthcare reform is likely to increase the use of healthcare forecasting. The Affordable Care Act includes multiple provisions intended to accelerate quality improvement.

 Medicare will expand performance-based physician compensation by 2014 and initiate pilot projects that bundle hospitalization and peri-hospitalization payments.

 Physician groups will therefore have greater incentives to “know the futureâ€ (ie, determine how their patient populations will affect relevant outcomes and payments). The legislation also creates accountable care organizations that can receive substantial compensation based on attaining risk-adjusted cost and quality-of-care targets.

These groups will have incentives to anticipate and prevent high-cost events and nonadherence to evidence-based care.

Contributing to Adverse Selection and Increased Disparities

Some predictive model vendors openly acknowledge that their forecasting tools can be used to avoid high-risk patients or to identify those that will remain healthy.

Moreover, because forecasts are used to include patients in interventions, they can also be viewed as excluding the persons not identified. For example, using forecasting or other structured case-finding methods, some health plans explicitly exclude patients with mental health diagnoses, addictions, and language barriers from disease management, because these factors might “predictâ€ a minimal impact of interventions.

More broadly, even in an environment where persons with preexisting conditions are guaranteed coverage and where performance monitoring and payment are risk adjusted, both insurers and physicians can benefit financially from selecting healthier patients to serve.

Propagating Disease Management With Limited Effectiveness

There is little evidence regarding how or whether forecasting improves healthcare value.

 This is due to both the modest level of research and what is termed the “impactibilityâ€ problem.

 That is, even if prediction algorithms accurately identify at-risk patients, intervening to achieve desired outcomesis often inhibited by limitations of current disease management approaches or the general state of medical science.

Private companies often develop and sell statistical prediction algorithms and therefore have strong incentives to keep “recipesâ€ proprietary. Furthermore, standards do not exist for governing key forecasting realms such as appropriate patients and outcomes to target, risk factors that should be considered, appropriate structures of forecasting equations, and acceptable levels of accuracy. This lack of transparency is reminiscent of the 1990s, when insurers often treated criteria for medical necessity coverage as proprietary, likely contributing to the managed care backlash.

Because of the rapid turnover of insurance coverage and current rules allowing some high-risk individuals to be denied coverage, most health plans and employers have tended to forecast short-term, high-cost outcomes such as rehospitalization or next year's utilization. But predictive techniques could also identify patients who need care that has longer-term benefits; for example, identifying persons especially in need of interventions such as cancer screening, obesity prevention, or cholesterol control.

Policy makers, tool developers, and health plans should consider strategies to enhance the benefits and minimize potential negative consequences of healthcare forecasting. Such strategies include the following 4 approaches.

Increased Transparency and Standardization

Health insurers and physician groups should share key information about their use of healthcare forecasting such as data inputs, types of algorithms used, and predicted outcomes. This sharing would enable better understanding of the impact of predictions on their enrolled populations. Such data collection could be voluntary and added to existing quality reporting that occurs through the Centers for Medicare & Medicaid Services and the National Committee for Quality Assurance. As suggested by a recent article critiquing hospital readmission forecasting,

 the scientific community should establish minimum accuracy benchmarks for models predicting various established outcomes. Additional standards should be developed for model development and deployment.

In the future, key research priorities should include examining the impact of predictive models on quality, the contribution of new data sources on forecasting accuracy, and the degree to which forecast-based interventions contribute to disparities between targeted and excluded patients. Better understanding is needed regarding optimum predictive algorithms such as automated approaches that detect patterns and correlations (data mining and artificial intelligence) versus algorithms that incorporate investigator hypotheses and iterative feedback (structured and learning models). To improve the effectiveness of disease management, continuing research on how forecasting can best identify patients amenable to feasible interventions will be essential.

Incentives for Long-Term Health Maintenance

A full discussion of policies that encourage employers, physicians, and health plans to cultivate long-term accountability toward patients goes beyond the scope of this commentary. However, such policies could encourage stakeholders to expand the application of innovative forecasting tools that promote health among patients likely to experience adverse health events well into the future, rather than just those at short-term “risk.”

If negative effects as described above are in fact documented, increased external oversight or self-regulation by stakeholders might be needed. For example, regulations could address how long-term health should be added to forecasts. Although the Affordable Care Act proscribes bias against patients with high predicted morbidity, potential abuses such as dropping of vulnerable patients should be monitored.

The increasing sophistication of predictive modeling and the changing health policy environment present opportunities for innovative forecasting applications that could enhance healthcare value and equity.

Personalized Health Insurance Design to Reduce Disparities

Articles

WEB EXCLUSIVE

Healthcare forecasting could facilitate personalized insurance benefit design and clinical decision support systems, although increased socioeconomic disparities could result from selective disease management programs.

The Promise and Peril of Healthcare Forecasting

 To assess medication adherence rates of patients utilizing an online prescription management account compared with nonusers.

 A retrospective analysis was conducted using de-identified pharmacy claims data from a pharmacy benefit manager covering the period from April 1, 2009, to March 31, 2011. Patients who were continuously eligible throughout the study period and that had at least 1 prescription fill for any of the 8 therapeutic groups examined in the study were selected for inclusion.

 Adherence was assessed by measuring the proportion of days covered (PDC). Propensity score matching was utilized to minimize differences in age, gender, chronic condition score, copay, household income, and urban locality between the users and nonusers groups. Results were reported for all therapeutic groups combined, as well as by individual therapeutic group.

 Overall, patients utilizing the online account had a significantly higher weighted average PDC (73.19% vs 61.64%, P <.0001). Users also had a higher average PDC for each individual therapeutic group, although the beta-blocker group was not statistically significant. The percentage of patients achieving an average PDC of >80% was also found to be greater in the users group across each therapeutic group and overall.

Patients who utilized an online prescription management account had higher rates of medication adherence as compared with nonusers. Additional studies are needed to assess which specific components of the prescription management account have the biggest impact on adherence.

We performed a retrospective analysis to evaluate the impact of an online prescription management account on patient adherence to medication for select therapeutic groups.

		Â Patients who utilized the account had higher rates of medication adherence.

		Â Online accounts which allow patients to manage their prescription refills and enroll in services such as automatic refills may help increase medication adherence.

		Â Payers and pharmacies should promote the use of online accounts for prescription management.

Medication is generally a key component in managing patients' chronic health conditions. However, in order for the treatment to be effective, patients need to properly follow the prescribing orders and obtain refills in a timely matter. Patients who do not adhere to their chronic medication therapies as indicated run the risk of detrimentally impacting their overall health and disease progression. Several studies associate medication nonadherence with increased hospitalizations and even mortality.

 Specifically, for diabetic patients, adherence rates of less than 80% have been linked to increased blood pressure, glycated hemoglobin (A1C) levels, and low-density lipoprotein levels.

Nonadherent diabetics are also more likely to be stricken with amputations, neuropathy, and myocardial infarctions.

Nonadherence can also have an impact on medical costs for patients and payers. Patients with adherence rates of greater than 80% have been shown to have lower medical costs for diabetes, hypertension, and hyperlipidemia compared with nonadherent patients.

 For diabetes and hypertension in particular, the savings are estimated at more than $3500 per person per year.

Despite the complications and financial burden associated with inadequate medication adherence, an estimated 50% of patients do not follow their regimen as prescribed.

 Given the negative outcomes and the magnitude of nonadherence, innovative approaches for managing medication therapies are constantly being sought. Researchers have been utilizing technology, such as text messaging, to improve patient adherence to medication.

 Providing patients with convenient options for prescription fulfillment and management may help improve adherence.

As the Internet has shaped American society, with nearly 70% of households reporting having Internet access at home in 2009,

Internetbased applications and programs may be a way to guide patients toward refilling their medications at proper intervals. Americans are increasingly comfortable with online shopping and the prescription management account allows them to access their prescription history and refill prescriptions at the touch of a button. To our knowledge, no one has assessed medication adherence rates of patients utilizing online prescription management accounts. We hypothesize that patients who use Internet-based services to fill their prescriptions will have greater rates of adherence.

We performed a retrospective study utilizing de-identified pharmacy claims and eligibility data from a pharmacy benefit manager (PBM), covering the period from April 1, 2009, to March 31, 2011.

This analysis was limited to patients who were continuously eligible during the study period and had at least 1 prescription fill for any of the top 8 chronic therapeutic groups during the identification period of April 1, 2009, through March 31, 2010. Patients were then followed for 12 months from the date of their initial (index) prescription. Therapeutic groups were determined by the first 2 digits of the generic product identifier (GPI) based on Medi-Span. The top 8 therapeutic groups were as follows: antidepressants (58), antidiabetics (27), antihyperlipidemics (39), antihypertensives (36), beta blockers (33), diuretics (37), thyroid agents (28), and ulcer drugs (49). All maintenance drugs (as described by First DataBank) within each therapeutic group were included, regardless of dosage form. In the event a patient filled a prescription for more than 1 therapeutic group during the identification period, he or she was included in each of the respective therapeutic groups. For example, a patient filling an antidepressant and a beta blocker would be represented in both of these groups.

Patients meeting the inclusion criteria who were registered for and utilizing an online prescription management account with the PBM and who were matched were placed in the “usersâ€ group. Utilization consisted of a patient accessing the online account for at least 1 of the following for at least 6 months of the follow-up period: to check prescription coverage, locate a pharmacy, submit a refill request, or request automatic refills. Patients who were not registered for an online account with the PBM, but were matched, were placed into the “nonusersâ€ group. All members of the PBM are given the option of enrolling in the online prescription management account. The online account provides patients with access to their prescription fill history and the opportunity to obtain automatic refills, process refills electronically, and view prescription coverage. The account also gives patients the opportunity to create a customized support page which provides information on health conditions chosen by the patient.

Propensity scoring was used to obtain 1:1 matches, using a “greedyâ€ 5â†’1 matching algorithm.

This algorithm first matched users to non-users on 5 digits of the propensity score. This was repeated for those who did not match using 4 digits of the propensity score, continued down to a 1-digit match. Matches were found for all patients. Propensity scores were obtained by fitting covariates, which included age, gender, chronic condition score, copay rank, household income level quartile, and urban locality into a logistic regression model. Chronic condition score was obtained by converting chronic condition counts to a rank (0-5) scale (0 = no chronic conditions; 1 = 1 chronic condition; 2 = 2-5 chronic conditions; 3 = 6-10 chronic conditions; 4 = 11-15 chronic conditions; and 5 = 

16 chronic conditions). Chronic condition counts were the number of unique drug-inferred conditions identified during the identification period (April 1, 2009, through March 31, 2010). These drug-inferred chronic conditions were established by mapping prescriptions to the Medi-Span Drug Indications Database.

This database provides information on common medical conditions treated by prescribed medications and is updated monthly to reflect the most current drug-disease mapping relations. Copay rank was calculated by converting copay per therapy day into a categorical variable and then quantified as low (0), medium (1), high (2), or very high (3). Copay per therapy day was calculated at the patient level and was based on all drugs within the 8 therapeutic groups included in the study. Therefore, if a patient filled prescriptions for drugs within 2 or more therapeutic groups, he or she had 1 overall copay per therapy day computed. Household income level was determined by linking patient zip code to the Internal Revenue Service Statistics of Income Tax Stats Tax Year 2008 Zip Code Data (http://www.irs.gov/taxstats/indtaxstats/ article/0,,id=242739,00.html). Household income level was then scaled into quartiles (0 = low income, 1 = medium, 2 = high, and 3 = very high income). Urban locality was determined by linking Rural-Urban Commuting Area Codes from the WWAMI Rural Health Research Center (http://depts.washington.edu/uwruca/) to the patient's zip code.

Medication adherence was measured using proportion of days covered (PDC). PDC was calculated as the sum of the days covered divided by 365 (the number of follow-up days), where days covered is based on the fill date and days supply, as indicated in the prescription claim. In the event a patient had a prescription fill for more than 1 drug within the same therapeutic group, causing overlapping fills, overlapping days are included once.

Chi-square test was used to assess for differences between groups for categorical variables and Student's 

 test was used to assess for differences between groups for continuous variables. A 

 value of <.05 was considered statistically significant. All statistical analyses were performed using SAS statistical software, version 9.2 (SAS Institute Inc, Cary, North Carolina). Results presented are for propensity score—matched groups.

, the users and nonusers groups each consisted of 3560 unique patients. Population characteristics were found to be similar between the 2 groups.

Patients utilizing the online account had a significantly higher weighted average PDC compared with that of the nonusers (73.19% vs 61.64%, 

 <.0001). Users also had a higher average PDC for each individual therapeutic group, although the beta-blocker group was not statistically significant. Results can be seen in 

 shows results for the percentage of patients within each PDC range (<50%, 50%-79%, or

80%). Overall, the percentage of patients achieving an average PDC of 

80% was nearly 13 percentage points higher in the users group. The users group also had a higher percentage of patients achieving a PDC of 80% or higher across each of the individual therapeutic groups. A smaller percentage of users had an average PDC of <50% compared with the nonusers (25.30% vs 37.09%).

Our results indicate that patients who have utilized an online account have significantly greater medication adherence across 7 out of 8 therapeutic groups studied, as well as overall, compared with nonusers. Additionally, the study showed that the users group had a higher proportion of patients achieving the 80% adherence mark, which can be crucial for chronic illness management.

 Having an online prescription management account offers patients several advantages, such as the ability to order refills online and receive notification when a prescription needs to be refilled. All of the services and benefits offered through an online account may help to increase a patient's adherence to his or her medication.

The online PBM account studied in this paper allowed patients to easily manage and refill their prescriptions, as well as view prescription history. This opportunity to easily refill a prescription allows patients to minimize travel and wait time. It also eliminates the need for knowing the prescription number, as is generally required when filling through automated telephonic systems. Upon signing in, the online account displays an alert on the main Web page if a patient has any prescriptions ready to be refilled. The ease of accessing prescription records and obtaining refills, and the convenience of being notified that a prescription is in need of being refilled, may help account users obtain their refills in a timely manner, thereby ensuring they are in possession of an adequate supply of medication, thus helping to avoid potential gaps in therapy.

The pharmacy benefit account also gives patients the option to have their prescriptions automatically refilled. Automatically dispensing prescriptions allows patients to obtain their refills without the risk of having gaps in therapy and without having to remember to do so. This type of service can combat one of the biggest barriers to adherence–forgetfulness.

The ease and convenience of using Internet-based services, as well as providing patients with multiple options for refilling their prescriptions, may help patients stay current on their prescribed drug regimens. Given the apparent effect on adherence, more payers and pharmacies should continue to promote the use of online prescription management accounts.

As society becomes more and more technologically advanced, new opportunities for Internet-based applications aimed at improving medication adherence are likely to be introduced. Mobile/smart phone applications are one example. These applications can likely improve medication adherence through the ease and convenience of ordering refills. Future studies should assess the impact of such programs, as well as examine the specific component(s) of these programs that impact adherence the most.

Online prescription management accounts may help promote medication adherence, as utilizing patients had a higher proportion of days covered than nonusers.

Impact of an Online Prescription Management Account on Medication Adherence

To identify the optimal target of a future intervention to improve physician decision making in trauma triage.

Comparison of incremental costeffectiveness ratios (ICERs) of current practice with hypothetical interventions targeting either physicians’ decisional thresholds (attitudes toward transferring patients to trauma centers) or perceptual sensitivity (ability to identify patients who meet transfer guidelines).

Taking the societal perspective, we constructed a Markov decision model, drawing estimates of triage patterns, mortality, utilities, and costs from the literature. We assumed that an intervention to change the decisional threshold would reduce undertriage but also increase overtriage more than an intervention to change perceptual sensitivity. We performed a series of 1-way sensitivity analyses and studied the most influential variables in a Monte Carlo simulation.

 The ICER of an intervention to change perceptual sensitivity was $62,799 per qualityadjusted life-year (QALY) gained compared with current practice. The ICER of an intervention to change the decisional threshold was $104,975/ QALY gained compared with an intervention to change perceptual sensitivity. These findings were most sensitive to the relative cost of hospitalizing patients with moderate to severe injuries and their relative risk of dying at non—trauma centers. In probabilistic sensitivity analyses, at a willingness-to-pay threshold of $100,000/QALY gained, there was a 62% likelihood that an intervention to change perceptual sensitivity was the most cost-effective alternative.

 Even a minor investment in changing decision making in trauma triage could greatly improve quality of care. The optimal intervention depends on the characteristics of the individual trauma systems.

Investing in an intervention to improve the decision making of individual physicians in trauma triage would yield significant value even if the intervention were expensive.

		 States such as Pennsylvania could invest up to $20 million to ensure triage according to the guidelines, and the combination of the intervention and care at a trauma center would still cost less than $100,000 per quality-adjusted life-year gained.

		 Quality improvement interventions should be tailored to the specific characteristics of each trauma system instead of using a one-size-fits-all strategy.

More than 30 million people receive trauma care every year in the United States,

 consuming approximately $400 billion annually.

 To optimize the allocation of resources, the American College of Surgeons advocates the regionalization of patients. 

 Regionalization requires that physicians at non—trauma centers transfer patients with moderate to severe injuries to specialty trauma centers while admitting those with minor injuries.

 Despite an unprecedented effort at the federal, state, and local levels to increase adherence to clinical practice guidelines supporting regionalization, only one-third of patients with moderate to severe injuries taken initially to non—trauma centers are actually transferred to trauma centers.

 Additional quality improvement interventions are needed.

Physician adherence to clinical practice triage guidelines can be thought of as the product of both decisional thresholds and perceptual sensitivity.

Decisional thresholds reflect physicians’ preferences for transferring or keeping patients based on norms, attitudes, and incentives. Perceptual sensitivity reflects physicians’ ability to discriminate between patients who do and do not meet clinical practice guidelines for transfer based on their knowledge of the guidelines and their intuitive judgments (heuristics).

 Decisional thresholds and perceptual sensitivity are distinct cognitive processes. Therefore, modification requires targeted interventions, which have different costs and benefits. Current efforts have included both top-down approaches aimed at shifting physicians’ decisional thresholds through a mixture of regulatory and organizational incentives

 and a bottom-up approach aimed at increasing physician perceptual sensitivity through Advanced Trauma Life Support (ATLS) training and certification.

To set priorities for quality improvement in trauma triage, we need to know whether it is more cost-effective to invest additional resources in changing physicians’ decisional thresholds or their perceptual sensitivity. The objective of our study was to compare the cost-effectiveness of current practice with that of 2 distinct hypothetical interventions to improve physician adherence to clinical practice guidelines: one that shifted decisional thresholds versus one that increased perceptual sensitivity.

Signal detection theory, a well-established behavioral science method, identifies 2 separate aspects of every dichotomous decision: decisional threshold and perceptual sensitivity (

). Interventions designed to impact these distinct cognitive processes require different designs and have different anticipated costs and benefits. An intervention to change decisional thresholds, such as a pay-for-performance initiative, would increase sensitivity. Although relatively inexpensive and easy to implement, shifting physicians’ decisional thresholds to make them more willing to transfer patients with moderate to severe injuries would reduce system efficiency, by increasing the number of patients with minor injuries transferred to trauma centers. On the other hand, an intervention to change perceptual sensitivity, such as a program to modify physician heuristics, would increase specificity. This kind of intervention would be relatively more expensive and difficult to implement, but would increase system efficiency.

) to compare the clinical and economic outcomes of a 40-year-old trauma patient taken initially to a non—trauma center under current conditions of national compliance with trauma triage guidelines compared with compliance resulting from 2 alternative hypothetical interventions—one targeting physicians’ decisional threshold and one targeting their perceptual sensitivity. As recommended by the Panel on Cost-Effectiveness in Health and Medicine, we took the societal perspective when we considered direct medical and nonmedical costs of obtaining care, and used a 1-month cycle length and a lifetime time horizon in the analysis. We discounted future costs and benefits at an annual rate of 3%.

 We constructed our model using TreeAge Pro 2009 software (TreeAge, Williamstown, Massachusetts).

All clinical parameters, utilities, and costs used in the base case and sensitivity analyses are summarized in 

 and described below. We drew probabilities of minor and moderate to severe trauma, of undertriage and overtriage, and of mortality conditional upon triage from the literature.

 In a retrospective analysis of Pennsylvania discharge data from 2001 to 2005, we found that 25% of patients presenting to non—trauma centers after trauma had a moderate to severe injury, defined as either an Injury Severity Score greater than 15 or as an injury categorized by the American College of Surgeons Committee on Trauma as “life-threatening” or “critical” (eg, open long bone fracture). Of these patients, 30% were transferred to a trauma center. Of patients with minor injuries, defined as the absence of a moderate to severe injury, 15% were transferred to a trauma center.

We drew age-adjusted probabilities of all-cause mortality from national health statistics.

 We assumed that a minor injury did not affect that probability, but that a moderate to severe injury increased the probability of dying in the first year after the traumatic event.

 We assumed that up to 40% of patients with a minor injury and 100% of patients with a moderate to severe injury experienced some period of disability, defined as an inability to return to work. If a patient remained disabled for a full year, we assumed that they remained disabled for the rest of their lifetime.

We calculated the number of quality-adjusted life-years (QALYs) associated with treatment of patients at trauma centers and non—trauma centers by multiplying the time spent in each health state and its associated utility.

We equated time spent acutely injured to the number of days spent hospitalized. We used a computer program to calculate an Injury Severity Score based on trauma codes in the 

International Classification of Diseases, 9th Revision, Clinical Modification

 We estimated the mean number of days hospitalized after minor and moderate to severe injuries by using national average length-of-stay (LOS) data for each primary

 Based on our analysis of Pennsylvania discharge data, we estimated that LOS would be equal to the average at non—trauma centers and 0.9 times the average at trauma centers for patients with minor injuries. We estimated that LOS would be 0.9 times the average at non– trauma centers and 1.4 times the average at trauma centers for patients with moderate to severe injuries. We estimated time spent disabled based on published estimates of rates of return to work after moderate to severe injuries and minor injuries.

Similarly, we estimated utilities associated with health states from published data on functional outcomes after moderate to severe and minor trauma.

Since treatment at trauma centers does not affect functional outcomes, we assumed that treatment at trauma centers and non—trauma centers would not affect the utilities associated with health states.

Costs for the treatment of minor and moderate to severe injuries included hospitalization, acute disability (inpatient and outpatient rehabilitation, follow-up care, and informal care), chronic disability (inpatient care, follow-up care, and informal care), and long-term healthcare costs. For moderate to severe injuries, we estimated the cost of hospitalizations, acute disability, and chronic disability from the literature.

 For minor injuries, we used 2008 national cost data for each 

 code to identify the average cost of hospitalizations.

 We estimated that care at a non—trauma center would be 0.75 times the national average for each injury and that care at a trauma center would be 1.3 times the national average.

 We assumed that acute disability costs included 10 days in a rehabilitation facility or nursing home, 3 outpatient physical therapy sessions, and 2 physician clinic visits. We assumed that chronic disability costs would include 4 outpatient physical therapy sessions and 1 physician clinic visit a month.

 We used the 2008 Medicare physician fee schedule to estimate the cost per day for a stay in a rehabilitation facility, the cost per physical therapy session, and the cost per clinic visit.

 We used the 2008 Centers for Medicare & Medicaid Services estimates of age-adjusted personal healthcare spending per capita to adjust for the healthcare costs of patients 1 year after injury.

Assumptions About the Hypothetical Interventions

We made 3 assumptions about the hypothetical interventions. First, based on the literature on quality improvement interventions, we assumed that the decisional threshold intervention would increase the transfer of patients with moderate to severe injuries by 40%.

 We estimated that the minimum and maximum improvements in transfer rates were 10% and 65%.

 However, changing decisional thresholds would increase tolerance for false-positive decisions (ie, the transfer of patients with minor injuries). We assumed the rate of transfer of patients with minor injuries would be half of the increase in the rate of transfer of patients with moderate to severe injuries.

Second, we assumed that the complexity of changing individual physician perceptual sensitivity would make the intervention half as likely to change rates of the transfer of patients with moderate to severe injuries as an intervention that targeted decisional thresholds.

not affect the transfer of patients with minor injuries.

Third, we assumed a fixed budget for implementation of a quality improvement initiative. Based on practice patterns in Pennsylvania, physicians at non—trauma centers treat 5 to 10 trauma patients a year

 and spend approximately $200 per year to participate in ATLS (approximately $40 per patient). We arbitrarily chose an additional $40 per patient as the budget, which would be equivalent to doubling the cost of participating in ATLS to fund the intervention.

We compared the clinical and economic consequences of an intervention that modified physicians’ decision threshold with those of an intervention that modified physicians’ perceptual sensitivity and compared both interventions with current practice patterns by using the incremental cost-effectiveness ratio (ICER), defined as the extra cost imposed by the intervention divided by its extra clinical benefit measured in QALYs. We performed a series of 1-way analyses for all variables to assess the effect of varying the baseline estimates on cost-effectiveness. To explore the relationship between the relative cost of the interventions and their effect on undertriage and overtriage, we varied 3 parameters (the relative cost of the interventions, the relative effectiveness of the intervention at reducing undertriage, and the relative cost of hospitalization at a trauma center compared with a non—trauma center) in a further sensitivity analysis.

We studied variables that influenced the ICER by more than 10% in the 1-way analyses in a multivariable sensitivity analysis using Monte Carlo simulation. We varied these parameters simultaneously over triangular probability distributions in each of 10,000 Monte Carlo iterations. We computed the percentage of Monte Carlo iterations for which a given strategy was more cost-effective for a willingness-to-pay ceiling of $100,000 per QALY, a commonly used threshold for cost-effectiveness.

To set priorities for quality improvement in trauma triage, we compared the cost-effectiveness of current practice with interventions to improve physician adherence to clinical guidelines.

Optimal Approach to Improving Trauma Triage Decisions: A Cost-Effectiveness Analysis

To verify implementation and use of TELEMACO (TELEMedicina Ai piccoli COmunilombardi; http://www.telemaco.regione.lombardia.it/), which provides specialized continuity of care with innovative healthcare services in remote areas of the Lombardy region of Italy; to design a network in the territory for sharing of continuityof- care programs; and to allow the relevant health authorities to collect cost data to establish a model for sustainable pricing for implementing these services.

TELEMACO provides home-based telemanagement services for patients with chronic heart failure and chronic obstructive pulmonary disease (COPD), as well as second-opinion teleconsultations in cardiology, dermatology, diabetology, and pulmonology for general practitioners and second-opinion teleconsultations on digital images in cases of traumatic brain injury and stroke. A total of 2 service centers, 10 cardiology and pneumology departments, 30 specialists, 176 general practitioners, 40 nurses, 2 emergency departments, and 2 consultant hospitals were involved.

A total of 166 patients with chronic heart failure and 474 patients with COPD were enrolled. There were 4830, 51, and 44 second-opinion teleconsultations for cardiologic, dermatologic, and diabetic conditions, respectively. There were 147 second-opinion teleconsultations on digital images, 68 for stroke, and 79 for traumatic brain injury. Implementation of TELEMACO introduced innovations in working methods and provided evidence to the health authorities for allocating funds for such services.

 TELEMACO provided evidence that there is a growing need for home management of patients using telemedicine, a common and efficacious approach that can ensure care continuity, especially in chronic diseases.

		 The TELEMACO project enabled structured management of patients with chronic heart failure and chronic obstructive pulmonary disease who lived in remote areas of the Lombardy region of Italy.

		 TELEMACO demonstrated the potential of telemedicine to support general practitioners in management of cardiologic conditions.

		 TELEMACO also allowed development of innovative medical and nursing skills that are becoming part of the routine medical care practice.

		 Overall evaluation of services showed that more than 95% of patients had a degree of satisfaction with the project.

Use of information and communication technology has rapidly increased worldwide and has contributed significantly to the clinical and health delivery sectors.

Information and communication technology provides medical care to patients who have poor access to hospitals while at the same time ensuring continuity of care and optimal use of the available health resources.

Telemedicine (TM) is widely implemented in rural communities, where residents face factors that can create disparities and difficulties in provisions of healthcare.

 The primary benefit of TM in these areas lies in facilitating remote access to healthcare and reducing cost of access, thus promoting better outcomes.

Moreover, an aging population, a group in whom the incidence of chronic or degenerative diseases tends to increase, can also benefit from TM since TM can provide more efficient, prompt, and appropriate diagnostic and therapeutic services.

Our project, entitled TELEMACO (TELEMedicina Ai piccoli COmuni lombardi; http://www.telemaco.regione.lombardia.it/), is an assignment pursuant to regional Law 11/2004: “Measures to support the small municipalities of Lombardy aimed at counteracting the depopulation of the territory.” This law was developed to provide specialized continuity of care with innovative healthcare services in remote areas of the region.

The aims of the project were 3-fold: (1) to implement and use continuity-of-care services in the Lombardy region of Italy; (2) to design a network in the territory for sharing continuity-of-care programs for the management of chronic diseases; and (3) to allow the health authority of the Lombardy region to collect data to establish sustainable pricing at the regional health level for implementing TM.

Territories with socioeconomic and infrastructural difficulties were selected within the rural areas of the Lombardy region, in conjunction with local health authorities and hospitals. Three study programs were introduced, as detailed below.

Home-Based Telemanagement for Patients With Chronic Heart Failure or Chronic Obstructive Pulmonary Disease

The methods were extensively described in previous articles.

 In brief, the chronic heart failure (CHF) group consisted of patients with New York Heart Association (NYHA) class II-IV left ventricular systolic dysfunction with ejection fraction less than 40% or CHF with diastolic dysfunction, with at least 1 episode of hospitalization for CHF within previous 6 months. The chronic obstructive pulmonary disease (COPD) group consisted of patients with stage III-IV COPD according to Global Initiative for Chronic Obstructive Lung Disease guidelines who received long-term oxygen therapy at home for at least 3 months, or were treated in hospital, in the outpatient setting, or in the emergency department for respiratory failure. Patients who could not be discharged from the hospital and those with severe cognitive impairment or other diseases with poor prognosis were excluded.

All patients were screened for eligibility before hospital discharge and all provided informed consent.

Before discharge from the hospital all patients were given instruction on their respective disease conditions (CHF or COPD). Patients were given pointers for recognizing signs and symptoms of decompensation/worsening. Patients with CHF were provided with a portable 1-lead electrocardiogram (ECG) device (Card-Guard 2206). Patients with COPD received a pulse oximeter (Vitalaire, Italy) device.

The home-based telemanagement (HBT) service was organized with a 6-month follow-up, with the possibility of readmission to the service for an additional 6 months. It consisted of 

 done by patients. During the scheduled calls, the nurse carried out a scheduled standardized interview on the general clinical condition of the patients, including the names and the doses of prescribed drugs and patient compliance. During these calls the nurse reinforced the initial educational intervention and the strategies offered to improve the patient’s compliance. Occasional calls were made by patients in case of symptoms or signs of possible decompensation/worsening.

During either scheduled or occasional calls, CHF patients could transmit via landline or mobile phone the recording  from the 1-lead ECG to a service center, and talk to the nurse or doctor 24 hours a day, 7 days a week. The COPD patients were not required to transmit any data.

Both at the beginning and at the end of the program, on an outpatient basis, a clinical evaluation together with instrumental analyses such as an echocardiogram and a 6-minute walking test in CHF patients and hemogasanalysis in COPD patients were performed. Moreover, patients were asked to complete a qualityof- life questionnaire (ie, the Minnesota Living Heart Failure Questionnaire

for CHF and COPD patients, respectively). Satisfaction was assessed by a questionnaire prepared ad hoc for the project.

Second Opinion for General Practitioners in Cardiology, Dermatology, Diabetology, and Pulmonology

The method has been described in detail in a previous work.

 The service was provided to general practitioners (GPs) working in rural areas of the Lombardy region through a service center during daily in-office or home visits. For cardiac problems, GPs were equipped with a 12-lead ECG device (Card-Guard 7100, Rehovot, Israel) that could be interfaced to a landline or mobile telephone, thus transferring in real time the recorded ECG tracing back to the receiving station. 

 For dermatologic problems, the doctors used a highresolution digital camera. The GPs transmitted the images by electronic mail to the service center. Problems related to diabetology and/or pulmonology were addressed by a telephone contact only.

The service center was implemented for cardiac consultation 24 hours a day, 7 days a week. For the other specialties, the service center was implemented from Monday to Friday from 8:00 am to 8:00 pm. Cardiology, pulmonology, and diabetology teleconsultations were performed in real time, while dermatology teleconsultations were performed within 30 minutes. All conversations with health professionals were recorded.

Second Opinion on Digital Images for Traumatic Brain Injury and Stroke Between Rural Hospitals and Specialized Hospitals

Two centers, each referring to a consultant hospital, were activated. Those patients who arrived at the emergency department with signs or symptoms of suspected traumatic brain injury or stroke were enrolled.

After documenting the patient’s information and the severity scale adopted (National Institutes of Health Stroke Scale

) in the patient’s health record, the emergency department physician sent the images (computed tomography) to the specialists at the reference hospital (neuroradiologist, neurologist, or neurosurgeon) and asked for a teleconsultation. The consulting physician had access to the images and data, and could give his/her diagnostic and therapeutic indications. Expert opinion was available within 30 minutes of arrival of images.

The 2 service centers offered technological support for transmitting the relevant medical data and management of clinical data that were collected. Service centers offered telephone clinical support overnight and during holidays, managed by professional staff (expert nurses and specialists) connected via telework.

 In addition, service centers provided training courses to all professionals involved.

During the study, field training programs (clinical audit/training activities) were organized. Training of health professionals included face-to-face classes and audioconferences to provide theoretical and practical knowledge on the basics of TM, associated technologies, sending data and tracings, and filling in the personal health record of the patient. An overview of the application models was also provided, together with information on the overall organization of the system and service model, types of requests, and responses to the teleconsultation. In addition, education on the care of patients with cardiac or respiratory diseases was provided.

The outcome measures for the 3 project programs were the following:

Assessment of the extent of TELEMACO services.

 This issue was addressed by evaluating the number of patients enrolled in the HBT program. The diffusion of HBT services for CHF and COPD patients was assessed in a single area of the Lombardy region (Valle Camonica). For CHF, the number of patients enrolled in the HBT program was compared with the number of patients enrolled in the HBT in Valle Camonica. For evaluating the diffusion of HBT, the total number of patients hospitalized in Valle Camonica represented the total population that could potentially access the HBT service. For COPD, the prevalence of disease in the territory (3% of the population has COPD, calculated as shown at www.goldcopd.it) was compared with the percentage of the patients enrolled in the HBT program.

 This assessment was done by evaluating the organization of the integrated services across the Lombardy territories.

The TELEMACO project successfully used telemedicine to establish a healthcare continuity from hospital to territory in remote areas of the Lombardy region of Italy.

Healthcare Continuity From Hospital to Territory in Lombardy: TELEMACO Project

To determine how often antipsychotics are prescribed off-label to adults without schizophrenia or bipolar disorder who are enrolled in Medicaid, which pays for more than 70% of antipsychotic prescriptions in the United States.

Retrospective analysis of 2003 administrative data from 42 state Medicaid programs.

Continuously enrolled patients with at least 1 prescription for an antipsychotic medication were identified. For these patients, inpatient and outpatient records were checked for any diagnosis of schizophrenia or bipolar disorder; those without any such diagnoses were considered to have received these medications off-label. Offlabel antipsychotic use was compared across sociodemographic groups (age, race/ethnicity, sex). Multivariate logistic regression models identified characteristics associated with off-label use.

Of the 372,038 individuals given an antipsychotic medication, 214,113 (57.6%) received these agents for off-label disorders. Off-label use among patients receiving an antipsychotic was more prevalent among individuals under age 21 years (75.9%) and those 65 years and older (64.8%) than it was among those aged 21 to 64 years (49.0%). Rates of off-label use were relatively high for Hispanics (65.7%) and low for African Americans (52.3%) compared with whites (58.2%). Off-label use was most common among patients receiving risperidone and least common among patients receiving clozapine.

Off-label use of antipsychotic medications is common, particularly among the elderly and children/adolescents. Given that these drugs are expensive, have potentially severe side effects, and have limited evidence supporting their effectiveness off-label, they should perhaps be used with greater caution.

This study documented rates of off-label use of antipsychotic medications in 42 state Medicaid programs in 2003.

		 57.6% of patients given an antipsychotic had no diagnoses of schizophrenia or bipolar disorder during the year.

		Rates of off-label use were higher among children/adolescents and the elderly, among patients given risperidone, and among those diagnosed with depression.

		Given healthcare reform and widespread crisis in state revenues, state Medicaid programs will be under pressure to serve larger patient populations, increasing their fiscal stress. Reducing off-label antipsychotic use may generate savings with little impact on patient outcomes.

Antipsychotic medications have long been an important component of effective treatment for schizophrenia. They also make up a large and growing share of Medicaid prescription drug programs, which covered nearly 75% of all prescriptions for antipsychotic medications in the United States in 2002,

 although since 2006 the public share of financing for antipsychotic medications has been roughly equally divided between Medicaid and Medicare.

 In addition to schizophrenia, most second-generation antipsychotics have also been approved to treat bipolar disorder. More recently, the Food and Drug Administration (FDA) approved aripiprazole (in 2007) and olanzapine (in 2009) for the adjunctive treatment of major depressive disorder and risperidone (in 2006) and aripiprazole (in 2009) for irritability in children and adolescents with autism. However, once a drug has been approved by the FDA, clinicians are free to prescribe it as they see fit. Because there often is not the same level of high-quality clinical research demonstrating the safety and efficacy of these drugs for non—FDA-approved indications, the benefits of such off-label use are usually unclear. Given that these drugs are expensive and have serious side effects (including weight gain, diabetes mellitus, tardive dyskinesia, and extrapyramidal symptoms), their off-label use may represent significant risk and cost with undemonstrated clinical benefit.

Several previous studies have investigated off-label use of antipsychotic medications.

The majority of patients prescribed antipsychotic medications in state Medicaid programs are taking these agents for off-label indications.

March 2012

The Promise and Peril of Healthcare Forecasting

Impact of an Online Prescription Management Account on Medication Adherence

Optimal Approach to Improving Trauma Triage Decisions: A Cost-Effectiveness Analysis

Healthcare Continuity From Hospital to Territory in Lombardy: TELEMACO Project

Off-Label Use of Antipsychotic Medications in Medicaid

Clinical Pathways for Oncology: More Rigor Needed When Evaluating Models

Adherence and Dosing Frequency of Common Medications for Cardiovascular Patients

Reducing Long-Term Cost by Transforming Primary Care: Evidence From Geisinger's Medical Home Model

What Determines Successful Implementation of Inpatient Information Technology Systems?

IT-Enabled Systems Engineering Approach to Monitoring and Reducing ADEs