Pharmaceutical policies: effects of restrictions on reimbursement

Carolyn J Green; Malcolm Maclure; Patricia M Fortin; Craig R Ramsay; Morten Aaserud; Stan Bardal

doi:10.1002/14651858.CD008654

. 2010 Aug 8;2010(8):CD008654. doi: 10.1002/14651858.CD008654

Pharmaceutical policies: effects of restrictions on reimbursement

Carolyn J Green ^1,^✉, Malcolm Maclure ², Patricia M Fortin ³, Craig R Ramsay ⁴, Morten Aaserud ⁵, Stan Bardal ¹

PMCID: PMC6791298 PMID: 20687098

Abstract

Background

Public policy makers and benefit plan managers need to restrain rising pharmaceutical drug costs while preserving access and optimizing health benefits.

Objectives

To determine the effects of a pharmaceutical policy restricting the reimbursement of selected medications on drug use, health care utilization, health outcomes and costs (expenditures).

Search methods

We searched the 14 major bibliographic databases and websites (to January 2009).

Selection criteria

Included were studies of pharmaceutical policies that restrict coverage and reimbursement of selected drugs or drug classes, often using additional patient specific information related to health status or need. We included randomised controlled trials, non‐randomised controlled trials, interrupted time series (ITS) analyses, repeated measures studies and controlled before‐after studies set in large care systems or jurisdictions.

Data collection and analysis

Two authors independently extracted data and assessed study limitations. Quantitative re‐analysis of time series data was undertaken for studies with sufficient data.

Main results

We included 29 ITS analyses (12 were controlled) investigating policies targeting 11 drug classes for restriction. Participants were most often senior citizens or low income adult populations, or both, in publically subsidized or administered pharmaceutical benefit plans. Impact of policies varied by drug class and whether restrictions were implemented or relaxed. When policies targeted gastric‐acid suppressant and non‐steroidal anti‐inflammatory drug classes, decreased drug use and substantial savings on drugs occurred immediately and for up to two years afterwards, with no increase in the use of other health services (6 studies). Targeting second generation antipsychotic drugs increased treatment discontinuity and the use of other health services without reducing overall drug expenditures (2 studies). Relaxing restrictions for reimbursement of antihypertensives and statins increased appropriate use and decreased overall drug expenditures. Two studies which measured health outcomes directly were inconclusive.

Authors' conclusions

Implementing restrictions to coverage and reimbursement of selected medications can decrease third‐party drug spending without increasing the use of other health services (6 studies). Relaxing reimbursement rules for drugs used for secondary prevention can also remove barriers to access. Policy design, however, needs to be based on research quantifying the harm and benefit profiles of target and alternative drugs to avoid unwanted health system and health effects. Health impact evaluation should be conducted where drugs are not interchangeable. Impacts on health equity, relating to the fair and just distribution of health benefits in society (sustainable access to publically financed drug benefits for seniors and low income populations, for example), also require explicit measurement.

Keywords: Adult; Aged; Humans; Drug Costs; Health Care Costs; Health Care Costs/legislation & jurisprudence; Health Services; Health Services/statistics & numerical data; Insurance, Pharmaceutical Services; Insurance, Pharmaceutical Services/legislation & jurisprudence; Prescription Drugs; Prescription Drugs/economics; Prescription Drugs/supply & distribution; Process Assessment (Health Care); Reimbursement Mechanisms; Reimbursement Mechanisms/legislation & jurisprudence

Policies that restrict reimbursement on some drugs to ensure better use in health care

Large amounts of healthcare money is spent on medications, and these amounts are increasing. Spending more on medications could mean less money for hospitals, doctors and even other public services such as education or infrastructure. Misuse, overuse or underuse of medications may also result in poor health outcomes and a waste of money. Publically funded drug benefit plans look for ways to ensure better use of medications and to control costs without limiting health benefits. Policies that restrict reimbursement of specific prescription drugs ‐‐ often called 'prior' or 'special authorization' policies ‐‐ are one type of policy that may be used. Physicians generally apply on behalf of the patient and supply information verifying the patient's need before authorization is granted. These policies provide a safety valve when restrictions are applied by allowing for reimbursement when there is a need for the specific drug. If authorization is not obtained, an alternative and often cheaper drug with the same or similar benefit is reimbursed, or a patient may have the means to pay out‐of‐pocket expenses. Medications targeted for reduction in use are often newer, expensive drugs with cheaper, effective alternatives.

This review found 29 studies that evaluated policies that restrict reimbursement of specific prescriptions drugs. Where drugs have cheaper, effective alternatives and they target symptoms, this review found that reimbursement restriction polices can ensure better use of the medications with reduced costs and without an increase in the use of other health services, as would be expected if there were negative health effects of the restriction policies. Evaluation is required if alternative drugs are not effective substitutes. Removing restrictions for drugs that prevent complications of disease can result in an intended increase in their use as well as cost savings. When restrictions to reimbursement policies are designed using the best available evidence on the health impact of the medications, they support equitable access to the drugs that best support health by supporting the sustainability of publically subsidized drug plans.

A summary of this review for policy‐makers is availablehere

Summary of findings

Summary of findings for the main comparison.

Outcome: drug use ‐ CITS and ITS studies

Study ID* (CITS)	Outcome / drug or drug class	Immediate after transition period (95% CI)	Immediate after transition period (95% CI)	Short Term (6 mo, 1 yr) (95% CI)	Long Term (12mo, 24 mo) (95% CI)
		Absolute change in level	Relative change in level	Relative change in level	Relative change in level
Delate 2005	Per patient per month PPI and H2RA drug claims	‐0.02 PPI 0.01 H2RA	‐ 92% PPI 98% H2RA
Fischer 2007 /CITS	Proportion of renin‐ angiotensln ‐aldosterone system (RAAS)‐blocking defined daily doses (DDDs) accounted for by ARBs		Any PA policy ‐ 0.4% (NS) Slope effect 0.0% (NS) States requiring ace inhibitor trial first ‐1.6% (p=0.026) * one time decrease		States requiring ace inhibitor trial first ‐1.3 % per quarter to 18 months (p<0.001)
Fischer 2004 /CITS	Proportion of nonsteroidal anti inflammatory drugs (NSAID) DDDs accounted for by COX‐2 Inhibitors			‐ 11.1% ( ‐5.7%, ‐16.5%) 6 mo post relative to 6 mo pre	‐1.6% (0.0%,3.1%) per quarter to 18 months (p=0.03)
Hartung 2004 /CITS	Days' supply per person‐year celecoxib	‐ 0.54 (from 1.07 to 0.53)	‐ 58.9% (‐50.0%, ‐67.9%)	decrease slope (P < 0.001).
Hartung 2006	Market share in aggregate of prefered drugs in the PPIs, NSAIDS, long‐acting opioids and statin drug classes			"Dispense as Written" period 8‐9 months +28 "Soft PA period " ‐ education 5 months +42.9 Voluntary 9 months ‐17.4
Keith 1994	Mean number of dosage units H2RAs dispensed per month			11 months ‐ 35% (from 69,212 to 44,751 units /mo) 0.7% slope
	Percentage of patients taking other drugs known to produce significant drug interactions			‐ 7.5 % (from 14 to 6.5%)
Kephart 2005 /CITS	Monthly use of wet nebulization therapy		decrease p<0.001	slower decline p<0.001	24 months Dec 99 to 01 Decline from 100% to 36% in heavy users Decline from 67% to 20% in all users
Law 2008 /CITS	Market share level of nonpreferred 2nd generation antipsychotics		West Virginia: ‐ 3.5% (‐5.7%, ‐1.3% p=0.003) Texas: ‐2.6% (0% , ‐5.2% p=0.55)	West Virginia: ‐1.3% (–1.8%, –.8%, p<.001) per quarter Texas: decrease in trend NS	24 months ‐13.8% (‐9.4%, ‐18.2% p<0.001)
MacCara 2001	Fluoroquinolone claims			‐ 80.2%
Marshall 2002	Mean no. defined daily doses (DDD) of all PPIs per 100,000 seniors per month	‐22,959 (from 84,531 to 62,708)	‐15.6% (during 6 mo transition following announcement)	‐26% first year	13 to 41 months 9% over baseline
Marshall 2006	Fluoroquinolone prescriptions per week				To 15 months post policy absolute decrease of 1905 (p<0.0001)
Schneeweiss 2004	Use of respiratory drugs during two months follow up			At 6 months nebulised drugs ‐ 7% p < 0.001 nebulised in combination with inhaled drugs ‐ 25% (p < 0.001)
Schneeweiss 2006	Monthly dispensed daily doses per 10,000 residents restricted and reimbursed PPIs			8 months post implementation Total PPI utilization No level change (p=0.82) ‐ 383 slope (p‐0.08) Restricted PPI ‐ 14,850 ( ±1100) Reimbursed PPI rabeprazole 19,300 ( ± 2200) (p < .0001) with 45% switching rate
Smalley 1995 /CITS REPORTED			‐ 26% (‐21, ‐31%)		2 year period compared with baseline ‐19% (‐13, ‐25)
Smalley 1995 /CITS REANALYSIS**	Days of NSAID use per Person‐Year	‐4.35	‐ 23%	6 months ‐ 21% (‐23%, ‐19%)**	At 23 months ‐ 21% (‐22, 19%)**
Soumerai 2008 /CITS	Atypical antipsychotic use			10 month No significant difference
	Relaxation of restrictions to reimbursement
Bjerrum 2001	Incidence prevalence rates of lipid lowering drug per 100,000 per month	14		1st year 0.4% increase prevalence
Fretheim 2007	Proportion of thiazide prescriptions among all prescriptions for patients started on treatment for uncomplicated hypertension			16.5% (9.9%, 24.8%) ‐ 0.05% slope change (NS) per month (0.65% to 0.35%).
Jackevicious 2008	Monthly rate of clopidogrel use within 30 days of hospital discharge following MI recieving stents		53%
Sakshaug 2007	1 yr prevalence of statin use			At 1 yr 6.8% increase women 8.1% increase men
van Driel 2008	Monthly reimbursed DDD for total H2As and PPIs				from 1997 to 2005 222%

OUTCOME	NO OF STUDIES	RELATIVE EFFECT: MEDIAN (RANGE)	QUALITY	COMMENTS
Target drug use, immediately after introduction	7	‐26% (‐.04, ‐92%)	1 study met all 7 criteria, 4 met 6 and 2 met 5	All studies were set in publically funded drug benefit plans: 5 US Medicaid and 2 Canadian provincial plans. One study was reanalysed.
Target drug use, at 2 years after introduction	4	‐17% (‐9%,‐70%)	2 studies met all 7 criteria, 2 met 6	All studies were set in publically funded drug benefit plans: 2 US Medicaid and 2 Canadian provincial plans
Expenditures on target drug or drug class at 6 months after introduction	3	‐ 57% (‐85%, ‐36%)	2 studies met all 7 criteria, 1 met 6	Studies suitable for reanalysis provided the best estimates of relative effect. All studies were set in publically funded drug benefit plans: 2 Canadian provincial plans and 1 US Medicaid
Expenditures on target drug at 2 years after introduction	2	‐ 49% (‐79%, ‐18%)	1 studies met all 7 criteria, 1 met 6	Studies suitable for reanalysis provided the best estimates of relative effect. Both studies were set in Canadian provincial publically funded drug benefit plans

•	Prior authorization policies are an acceptable component of pharmaceutical plans that seek to influence prescribing behaviour so that the right drug is used for the right problem. Designed appropriately, it would appear that PA policies can support sustainability of drug benefit plans and thereby preserve access to necessary drugs for low income populations
•	When drugs are interchangeable, restrictions to reimbursement policies decrease prescription drug utilization and costs for the targeted drugs often without increases in the utilization of other health services however this proxy for health outcomes may not be sensitive enough to detect all important impacts
•	Results may not be generalizable beyond the drugs that are most easily interchangeable, like NSAIDs and PPIs. Selection of drugs for restriction or removal should be based on the clinical attributes of the drug class and properly designed interrupted time series evaluation is important to measure impact
•	Application for exemption from restrictions to reimbursement provides a 'safety valve' allowing access to restricted drugs under some circumstances thereby maintaining a range of therapeutic options and facilitating acceptance by physicians and patients
•	Relaxation of or exemption from reimbursement restriction policies for drugs that are useful for secondary prevention and under‐used drugs increase appropriate use and produce more modest savings
•	High quality studies with measures of health outcomes are few and, where available, are inconclusive highlighting the requirement that restrictions be solidly based on clinical trial and and pharmacosurveillance studies

•	The adoption of restriction to reimbursement policies for less interchangeable drugs and drugs for non‐symptomatic conditions should be accompanied by impact evaluations
•	More rigorous study designs including randomization with implementation delays, health outcome evaluation, reporting of baseline characteristics, measurement over longer time periods and in a greater range of settings and drug classes as well as in combination with other pharmaceutical policies are appropriate

Methods	Equivalent to ITS with monthly incidence rates calculated using an epidemiology approach
Participants	County of Funen, Denmark, n=470,000 inhabitants
Interventions	Release of restriction to prescribing lipid lowering drugs (LLD) to encourage use for secondary prevention. After policy date there was no need for special authorization to recieve reimbursement for LLD with known coronary heart disease (CHD), inherited hyperlipidimia, high cholesterol and concurrent important risk factors for developing CHD.
Outcomes	Incidence and annual prevalence rate of LLD use
Notes	Included as tracks lipid lowering drugs usage 5 years prior to implementation and 12 months after policy implementation
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Odense University Pharmacoepidemiological Database covers whole population of the county with an age sex distribution corresponding to the Danish population.
Selective reporting (reporting bias)	Low risk	Incidence and prevalence rates for all years calculated
Other bias	Low risk	No other bias detected
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Low risk	Reviewers note that published educational guidelines on the prevention of CHD were part of intervention. The authors describe this as a mixed intervention combining guidelines with changes in reimbursement however an integral part of most implementations is providing prescribers with guidelines so this could be interpreted as a difference in intensity. The authors also report that ' local medical journals had an increased focus on the subject at the time of intervention.'
The shape of intervention pre‐specified?	Unclear risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	ITS in 6 month intervals
Participants	Newfoundland and Labrador drug subsidy program for seniors ≥ 65 years: 1996 n=117,637; 1997 n=11403, 1998 109797; (population decrease of 6%)
Interventions	Special authorization policy implemented July 1, 1996 for PPIs famotidine and nizatadine
Outcomes	Drug use and drug costs
Notes	Follow up 6 months after policy implementation
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Low risk	Drop of 6% in plan enrollment noted but unlikely to overturn study results
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of health care utilization
Intervention independent of other changes	Low risk	Reviewers note that implementation intensity was increased through the distribution of algorithms for the management dyspepsia and gastroesophogeal reflux distributed to all physicians and pharmacists
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	"analysis of DHCS (Department of Health and Community Services) database" (p 820) Sources and methods of data collection were the same before and after the intervention

Methods	ITS with monthly data points and no transition period
Participants	Fee‐for‐service (FFS) plan members in Missouri Medicaid program n=42,262
Interventions	Automated PA program for COX‐2 inhibitors
Outcomes	Drug costs ‐ figure
Notes	Baseline period was 12 months immediately preceding intervention and post intervention period was 12 months after date of implementation. Controlled Before After (CBA) analysis did not meet selection criteria and therefore drug use data was not included.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All use and cost outcomes from methods section reported in findings
Other bias	Low risk	No bias detected that pertained to the ITS data presented in figure
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of health care utilization
Intervention independent of other changes	Low risk	Known and unknown changes in the environment are unlikely to be responsible for sudden change and magnitude of effect size
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time (time series data provided).
Intervention unlikely to affect data collection?	High risk	Intervention aimed at better implementation through better data collection at the point of sale source.

Methods	ITS calculated on a per month per member basis
Participants	Plan members in medicaid program n=1,142,866 to n=1,324,643 during 24 month study period
Interventions	PPI PA program
Outcomes	Drug use, health services utilization, drug costs
Notes	Data collected 12 months prior and 12 months post‐implementation. Time series consisted of 24 PMPM (per member per month) for PPI and H2RA expenditures
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the Medicaid administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Unclear risk	It is not clear whether the restriction of the analysis to charges that were at lease partially reimbursed to the extent to which they were reimbursed introduced a selective reporting bias
Other bias	High risk	New claims for PPIs were not separated from renewals so the before‐after comparison is confounded by the indication. New claims would be triggered by sickness just before. The ratio of new to continuing users might differ between the 3 groups. People with Helicobacter infection were granted 1‐month of PPI. PPI users were more likely to have a gastro intestinal (GI) diagnosis and GI screening that H2RA users or non‐users
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of health care utilization
Intervention independent of other changes	Low risk	Known and unknown changes in the environment are unlikely to be responsible for sudden change and magnitude of effect size
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	Controlled ITS measured per member per month
Participants	Intervention: Georgia N=5178 Medicaid members n=984,843; control: Mississippi members N=2218 (no restriction) Medicaid members n= 434,782
Interventions	To document effect of step therapy PA policy for atypical antipsychotics
Outcomes	Health services utilization, drug costs
Notes	Jan 1‐1996‐Dec 31 1997 modeled 10 months pre‐policy, 11 months policy period (long transition period), 3 months post‐policy
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Unclear risk	Unclear as it is difficult to evaluate model assumptions Dementia was included as a condition of interest (along with schizophrenia, bipolar, depression and other mental health conditions) even though antipsychotics are not recommended in dementia, because they are commonly used to control symptoms.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Unclear risk	"Our study examined a period of adoption for atypical antipsychotic medications. The atypical antipsychotic olanzapine, for example, was approved by the US Food and Drug Administration during the prepolicy period shortly before the policy was enacted."
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	Controlled ITS with quarterly data points
Participants	50 Medicaid programs in 22 states
Interventions	PA policy for COX‐2 inhibitors
Outcomes	Drug use and drug costs
Notes	Policies classified based on 5 cinical risk factors for gastrointestinal adverse effects: More strict ≥4 gastrointestinal complicatons, and patients have a history of GI complications from NSAIDS, Strict ≥4 gastrointestinal, all others defined as less restrictive
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Unclear risk	"States with prior‐authorization programs may have differed systematically from states without such programs in ways not captured by our data, since states with these programs had slightly lower rates of coxib use before the programs were implemented. Patterns of NSAID prescribing for Medicaid beneficiaries appear to have similar variability, even independent of the effects of prior‐authorization programs, and should be a target of subsequent studies."
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Low risk	Known and unknown changes in the environment are unlikely to be responsible for sudden change and magnitude of effect size
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	Controlled ITS with monthly data ponts
Participants	All patients in 3 groups of general practice sites in and around Oslo Norway n=61 practices:1. practices randomized to guideline intervention, 2. practices in the control group, 3. practices that did not participate in the trials
Interventions	Thiazides made the only reimbursable drug for non‐complicated hypertension. Documenting a reason for prescribing non thiazides in the medical record was the only requirement to prescribe an alternate drug.
Outcomes	Patient health outcomes, drug use, drug costs
Notes	Incorported a transition period from Dec 2003 to Feb 2004 with 11 monthly measurements before and after the intervention period
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data from 3 practices excluded due to technical difficulties
Selective reporting (reporting bias)	Low risk	Change from protocol documented and reasonable
Other bias	Unclear risk	We do not know how representative the study practices are of general practices in Norway. Almost 40% of the invited practices did not agree to participate in our study, and this may have introduced some degree of bias.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Low risk	Changes in policy and market environment (changes in drug prices) a consequence of the policy and not bias.
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Study ID* (CITS)	Outcome / drug or drug class	Immediate after transition period (95% CI)	Immediate after transition period (95% CI)	Short Term (6 mo, 1 yr) (95% CI)	Long Term (12mo, 24 mo) (95% CI)	Long Term (>24 Months)
		Absolute change in level	Relative change in level	Relative change in level	Relative change in level	Relative change in level
Bursey 2000	PPI restriction Expenditures to treat upper gastrointestinal disorders in 6 mo intervals	$153	‐19.6%	At 6 months ‐ 85% (‐92.6%, ‐77.9%) At 12‐months ‐ 84% (‐89.7%, ‐79.3%)	At 24 mo ‐ 79% (‐82.8%, ‐75.2%)
Delate 2005 /ITS	PPIs Per member per month		PPIs decreased 90.9% H2RAs Increased 223.2%	PPI decreased from $44.1 million to $13.5 million H2A drug expenditures increased from $6.0 million to $13.5 million	Absolute decrease from $3.44/PMPM to $1.74/PMPM Net expenditure decrease of $23.4 million
Fischer 2007 /CITS	Angiotensin receptor blockers (ARBS)	For PDL: 0.4% p=0.049 For ACE trial: ‐1.0% p=0.003	ACE trial: ‐0.7 p<0.001 Slope Effect PDL: 0.3% p<0.001
Fischer 2004 /CITS	COX‐2 Inhibitors	‐ $10.28 ($7.56, $13.00) p<0.001	18%
Grootendorst 2005 /ITS	NSAIDS, analgesic drugs				Cummulative effect per year: Type 1 RP: ‐$1,035,340 (95% CI ‐$1,505,318, ‐565,362 p<0.001) Type 2 RP: ‐$4,007,322 (‐$4,378,332, ‐$3,36, 312 p<0.001)	Over all months: Type 1 RP: ‐$7,506,21 (95% CI ‐10,900,00, ‐$4,098,872) Type 2 RP (95% CI ‐$22,700,000 (95% CI ‐$24,800,000, ‐$20,600,000 p<0.001)
Hartung 2004 /CITS	COX‐2 Inhibitors				Savings attributable to the PA policy were projected to be approximately $10,402 (linear model) and $4999 (logarithmic model) per month.	Mean projected savings attributed to the PA policy was $2.87/Person Year (PY) (linear) $1.40/PY (logarithmic) .
Hartung 2006/ ITS	PPIs, long acting opioids, NSAIDS, statins	Aggregate: DAW Exception: ‐$0.18 (‐$0.08, ‐$0.02 p<0.05) Soft PA: $0.28 ($0.11, $0.44 p<0.05) Voluntary ‐$0.10 (‐$0.26, $0.06)	Aggregate: DAW Exception: ‐9.1% (95 % CI ‐13.8%, ‐4.3% p<0.05) Soft PA: ‐17.7% (95% CI ‐25.4%, ‐10.0% p<0.05) Voluntary 5.5% (95% CI‐‐1.1%, 12.1%)			Estimated savings during entire period: DAW policy: $1,727,392 (95% CI $976,102, $2,478,682) Soft PA policy: $2,223,300 (95% CI $1,816,027, $2,854,353)
Keith 1994 /ITS	H2RAs				Adjusted net savings of: $275,920.
Law 2008/ CITS	2nd generation antipsychotics	No significant change				Costs rose 7.95 to $9.84 over study period, Texas: $9.19 to $10.49
MacCara 2001 /ITS	Fluoroquinolone				$605,890. (23.7%) decrease
Marshall 2006 /ITS	Fluoroquinolones		Fluoroquinolones: $105,707 less per week, p<0.001)
Motheral 2004	NSAIDS, PPIs, SSRIs	In all 3 therapy classes, an immediate decrease of $0.93 PMPM costs p<0.01. Savings of 19% of net costs relative to mean monthly preperiod expenditures	In month following step therapy: NSAIDS: decrease of $0.29/PMPM p<0.001 SSRI: no significant change PPIs: decrease of $0.48 net drug cost p<0.05
Roughead 2006 /CITS	COX‐2 Inhibitors					Average cost per NSAID prescription: Unrestricted access: $59.00 Late policy adopting: $46.00 Early policy: $40.00
Schneeweiss 2004 / CITS	Nebulized respiratory drugs	‐ $24 PMPM (‐$19, $29)
Schneeweiss 2006 /ITS	PPIs			Reduction of $3.2 per senior Estimated savings of $2.9 million
Smalley 1995 /CITS REPORTED	Non‐generic NSAIDS per person per year	‐ $14.63	‐ 65% (‐ 60%, ‐71%)	Slight upward trend of $0.17 per month ($0.02, $0.32)		‐ 53% (‐48%, ‐57%)
Smalley 1995 /CITS REANALYSIS		‐$13.45**	‐ 58.5%**	‐57.2%** (‐59.7%, 54.7%)	‐ 56.1%** (58.0%, ‐54.1%)
Relaxation or exemption from restriction
Fretheim 2007 /ITS	Antihypertensives				Savings of U.S. $0.72 million, or U.S. $0.16/inhabitant
van Driel 2008 /ITS	H2As and PPIs					Public expenditure for acid suppressants increased from a total of € 7.5 million in 1997 to €12. million in 2005

Study ID* (CITS)	Outcome / drug or drug class	Immediate after transition period (95% CI)	Immediate after transition period (95% CI)	Short Term (6 mo, 1 yr) (95% CI)	Long Term (12mo, 24 mo) (95% CI)
		Absolute change in level	Relative change in level	Relative change in level	Relative change in level
Delate 2005/ITS	PPI restriction Expenditure and use of ambulatory services and inpatient care (including long‐term care) events and expenditures for the management of GI‐related and all health conditions			At 6 mo PA request granted PPI users were more likely to have had at least 1 diagnosis for a GI condition and GI‐related screening during baseline (thus were not comparable to groups denied PPI reimbursement ‐‐H2RA users and nonusers)
Farley 2008/CITS	Atypicial antipsychotic restriction Expenditures (Medicaid payments) for outpatient, inpatient and long term care services				To 14 months post policy implementation Oupatient services: + 18% increase (1.1%, 1.2%) p<0.001 Inpatient services: + 0.96 (0.78 to 1.20) NS Long term care services: + 1.243 (0.95 to 1.62) NS
Hartung 2004/CITS	Celecoxib restriction Medical services claims: office visits, emergency department (ED) encounters, hospitalizations Medical services claims (as above) for musculoskeletal conditions and GI ulceration	All enrollees Absolute number of GI‐related events per month was 'small and sporadic' Secondary analysis 'mean of only 4.3 GI‐related events' ‐too few for analysis	Immediate to 5 months Primary analysis (All claims) Medical claims: +18.0 (2.2, 33.9) NS at predetermined level of 0.01 + slope change, p<0.001 Musculoskeletal‐related encounters: NS change in utilization + slope change, p<0.001 GI related encounters: Decline (unspecified) p = 0.003 No other notable changes Secondary analysis (previous NSAID, COX‐2 use) ED use: ‐ 7.3% (‐20.8%, 6.3%) ‐ slope change (NS) Other medical services use: No significant change
Hazlet 2002/ CITS	PPI Restriction (Relative ratio of extrapolated trend period 1 versus monthly period 2 observation) Office visits and transactions		Month 0 0.95 (0.91, 0.99) Month 1 0.93 (0.89, 0.97)	Month 5 0.86 (0.81, 0.92) Month 11 0.71 (0.65, 0.79)
	ER visits		Month 0 1.06 (0.87, 1.28) Month 1 1.08 (0.89, 1.30)	Month 5 1.17 (0.88,1.55) Month 11 1.49 (0.97, 2.28)
	Hospital admissions		Month 0 0.85 (0.67, 1.07) Month 1 0.88 (0.70, 1.10)	Month 5 1.03 (0.73, 1.46) Month 11 1.07 (0.6, 1.79)
	Length of stay		Month 0 0.75 (0.50, 1.14) Month 1 0.76 (0.51, 1.13)	Month 5 0.79 (0.44, 1.42) Month 11 0.81 (0.3,1.92)
Keith 1994	H2RA restriction Number of GI studies (n/month)			7 months post implementation ‐ 5.7%
Kephart 2005/CITS	Wet Nebulization therapy restriction Visits to a fee‐for‐service general practitioner by month				To 23 months post policy announcement No change overall Wet nebulization group: decrease in general practitioner visits (p< 0.001) relative to controls Heavy wet nebulization cohort: no change
	Rates of hospital admission for a respiratory condition by month				Wet nebulization cohort: decrease in hospital admissions for respiratory conditions relative to controls (p=0.02) Heavy wet nebulization cohort: no change
Schneeweiss 2004	Restriction nebulized respiratory therapy Contacts with doctors per 100 patients per month			Followed to 6 months 2.6 (−5.0 to 10) p = 0.10
	Admissions to emergency departments per 100 patients per month			0.4 (−0.1 to 0.9) p = 0.10
	All admissions to hospital per 100 patients per month			0.7 (0.0 to 1.3) p = 0.08
Schneeweiss 2006/ITS	PPIs restriction Hospitalization for GI hemorrhage per 10,000 residents per month	Immediate Level + 0.15, (‐0.17 to 0.47) p=0.35 NS Slope ‐0.02 (‐0.08 to 0.04)
	Hospitalization for complicated peptic ulcer disease (PUD) per 10,000 residents per month	Immediate Level ‐ 0.64, (‐1.54, 0.26), p= 0.16 NS Slope 0.05, (‐0.10 to 0.21), (p= 0.48)
	Physician visits for gastroesophageal reflux (GERD), PUD, or gastritis (per 10,000 residents per month (rate) and spending)	Immediate +2.61, p= 0.59, NS Slope ‐ 0.20 (‐1.94 to 1.54) (p= 0.81) NS 3 to 5 months post policy implementation Level + 11, p= 0.01)
	Spending on physician visits for gastroesophageal reflux (GERD), PUD, or gastritis (per 10,000 residents per month	3 to 5 months post policy implementation Spending +Can $21,000, p= 0.20, NS
Smalley 1995/CITS REANALYSIS	NSAIDs PA Expenditures for Outpatient Services per Person‐Year ($)	Reanalysis** $9.02	Reanalysis** ‐ 10.6% (SE 5.6)	Reanalysis at 6 months** 12.3% (6.0%, 18.7%) Reanalysis at 12 months 11.7 (6.7%, 16.8%)	From study report at 24 months 3% (‐13%, 18%) (NS)
Smalley 1995/CITS REPORTED	Expenditures for Inpatient Admissions per Person‐Year ($)				From study report over 24 months ‐14% (‐8%, 36%)
Smalley 1995/CITS REPORTED	High dose NSAID users: Expenditures for Outpatient services per Person‐Year ($)				From study report over 24 months ‐2% (‐15%, 19%)

Methods	ITS monthly data points
Participants	Senior beneficiaries in provincial drug benefit plan, British Columbia, Canada
Interventions	Type 2 reference pricing for first line restricted non‐steroidal anti‐inflammatory drugs (NSAIDs) (different NSAIDs considered interchangeable)
Outcomes	Drug costs, drug use
Notes	Pharmacare data aggregated from Feb 1993 to June 2001
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Low risk	No evidence of other risk of bias including seasonality.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Unclear risk	A moderate upward trend in use of opiates that was almost a mirror image of the decline in NSAID use from 1998 to 2001.
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention however the introduction of PharmaNet at the same time as RP (in Sep 95) caused better capture of drug claims because it eliminated the need for seniors to mail in their drug claims.

Methods	Controlled ITS with monthly data points
Participants	Intervention group Medicaid managed care organization (MCO): 1999: n=74,866 patients, 200: n=91,816; control group (fee for service): 1999:n=110,076, 2000: n=153,784. Control group: members of the same MCO with past NSAID or COX2 use and continuous enrollment
Interventions	PA policy for celecoxib
Outcomes	Drug use, health servics utilization, drug costs
Notes	No notification to physicians and payments as to when the policy implemented.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Unclear risk	All outcomes listed in the methods section are reported in the results sections
Other bias	Unclear risk	Control population consisted of patients who were generally older and likely to be more severely ill which would tend to increase drug use
Knowledge of allocated interventions adequately prevented during the study	Low risk	Data objective: All Medicaid claims for patients in CareOregon and in a Medicaid fee‐for‐service program were reviewed.
Intervention independent of other changes	Low risk	Known and unknown changes in the environment are unlikely to be responsible for sudden change and magnitude of effect size
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	ITS with monthly data points
Participants	N=6161 who received a PCI with coronary stenting during hospitalization in Ontario
Interventions	Change in a PA policy to a limited use policy allowing for use of clopidogrel for approved indicatons according to pre‐specified prescribing codes
Outcomes	Health: composite rate of readmission for myocardial infarction, death, repeat percutaneous coronary intervention (PCI) and coronary artery bypass graft surgery (CABG) within 1 year of discharge, rates of death from any cause and major bleeding, drug use
Notes	Baseline characteristics differed: during the limited use period, patients had lower rates of aspirin use and higher rates of statin use (increase from 63% to 80% P<0.001)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Low risk	No evidence of other risk of bias including seasonality.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	High risk	"There were some differences in baseline characteristics between patients in the prior‐authorization period and those in the limited‐use period. In particular, during the limited‐use period, patients had lower rates of aspirin use and higher rates of statin use"
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	ITS with weekly data points
Participants	Ontario Drug Benefit (ODB) benficiaries
Interventions	Reimbursement of 2 fluoroquinolone antibiotics, ciprofloxacin and ofloxacin limited to beneficiaries for whom second line antibiotics were indicated
Outcomes	Drug use, drug costs
Notes	Data from Jan 1999 to Sep 2002, policy implemented Mar 4 2001
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	First, antibiotic groups were specified based on the active ingredient, irrespective of manufacturer, dosage form, or strength, and were not adjusted to account for formulation or dosing regimen using defined daily doses. Measurement of drug use in this study was measured by the number of prescription claims, because this is the approach used by the Ontario Ministry of Health in its annual update on the ODB programme
Selective reporting (reporting bias)	Low risk	All outcomes reported: The model results for antibiotic expenditures generally reflect the results for antibiotic use. There was a statistically significant decrease in expenditures for ciprofloxacin and the fluoroquinolone group and an increase for Trimethoprim‐sulfamethoxazole (TMP/SMX) and nitrofurantoin, but no change in overall antibiotic expenditure.
Other bias	Low risk	Although antibiotic use and expenditures were expected to change during the month the policy was implemented, the data were tested for breaks at other ates. Failure to account for the presence of other breakpoints might bias the estimates of the policy implementation indicator variable.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes objective: Our objective was to determine the impact of this policy on the volume and cost of antibiotic prescribing.
Intervention independent of other changes	Unclear risk	Second, it is possible that the changes in antibiotic use and expenditure were due to factors other than the implementation of the LU policy....The lack of cost savings could be explained by higher utilization of more costly drugs, more repeat prescriptions, or the use of drugs requiring longer duration of therapy.
Intervention unlikely to affect data collection?	Low risk	Data objective: weekly data on the prescribing volumes and expenditures (excluding dispensing fees) on individual antibiotics reimbursed by the ODB were analysed from January 1999 to September 2002.

Methods	Regression and random‐effects meta‐analyses of state level Medicaid ITS data with quarterly data points
Participants	48 states, n=21 states implemented PA policies for controlled‐release oxycodone
Interventions	Impact of no PA, strict PA and lenient PA
Outcomes	Drug use, drug costs
Notes	Novel method of analysing state level administrativ data ‐ evaluated using ITS critieria Analysis between 1996‐2005
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	For 10 states that did not respond to interview requests by phone and e‐mail, only website data were included.
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	High risk	Reimbursement figures do not include dispensing fee or reflect manufacturers’ rebates to states.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	High risk	"It is also likely that drug manufacturers’ marketing and other tactics in response to PA implementation contribute to the limited success of such access restriction efforts.
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	Prevalence study equivalent to ITS using monthly data points
Participants	Population of Norway (4.6 million)
Interventions	National restriction on statin prescribing to simvastatin generics
Outcomes	Drug use, drug costs
Notes	Data from Jan 2005 to June 2006, policy implemented June 2005
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	"...all dispensed prescriptions of statins to individuals in the total population in Norway. The NorPD contains information that makes it possible to follow each individual over time in order to study changes in pharmacological treatment"
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Low risk	No other bias detected
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Low risk	Changes in prescribing with announcement and price changes part of intervention
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	Sources and methods of data collection were the same before and after the intervention

Methods	Controlled ITS with repeated measures analysis with monthly data points
Participants	Senior and low income beneficiaries in provincial drug benefit plan, British Columbia, Canada N=53 pairs of clusters of doctors in British Columbia matched by size of location and number of patients. Policy implementation group n=449 patients affected by policy, policy exempt group n=386 patients. Observational cohort N=4624 patients (excluding n=386 in control goup)
Interventions	Restricted reimbursement for nebulizers, doctors had to request exemption
Outcomes	Drug use, health services utilization, durg costs
Notes	Control group contaminated by announcing policy change to all doctors (including control doctors) therefore cluster randomized controlled trial component of this study was excluded
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"We randomly assigned one cluster of each pair to either the policy implementation group or the policy exempted control group. Randomising isolated but matched clusters of doctors minimised the risk of contamination and reduced imbalances owing to chance."
Allocation concealment (selection bias)	High risk	"The non‐compliance resulted from a change that Pharmacare made to the protocol at the last minute, which can be avoided in future policy trials. Wanting to underscore the independence of the evaluation from the government, Pharmacare cautiously sent letters announcing the policy change six weeks in advance to all doctors in the province. Control doctors were not told of their exemption until they received a separate letter from the investigators two weeks later. Many control doctors either overlooked the second letter or decided to switch patients to inhaler drugs in anticipation of the new policy."
Incomplete outcome data (attrition bias) All outcomes	Low risk	All outcomes reported and adjusted
Selective reporting (reporting bias)	Low risk	All relevant outcomes in the methods section are reported in the results section
Other bias	Low risk	Considered by authors: "in the first month after the policy was implemented doctors reduced their office hours to meet budget caps at the end of the fiscal year, possibly more than in the previous year. The randomised analysis was protected from this bias since the concurrent control patients were equally as affected as the intervention patients.
Baseline outcome measurements similar	Low risk	"The historical control groups showed some seasonal variation, with higher use of respiratory drugs in the six months preceding 1 March compared with the subsequent six months."
Baseline characteristics similar	Low risk	"The baseline distributions of age, sex, drug use, comorbidity score, and use of healthcare did not differ between the randomised groups (table 1; all P values > 0.1). The observational cohorts had slightly more women than the randomised groups (62% v 59%). Otherwise the observational cohorts were comparable to each other."
Knowledge of allocated interventions adequately prevented during the study	Low risk	"We used de‐identified data from the databases of British Columbia’s Ministry of Health "
Study adequately protected against contamination?	High risk	Control doctors were notified of the policy change 2 weeks before recieving a letter informing them of their exemption.
Intervention independent of other changes	Low risk	We first adjusted for seasonality separately, using each of the two historical control cohorts.A significant interaction would mean that changes in end point trends because of the policy were different from the control cohorts, independent of seasonality.
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time
Intervention unlikely to affect data collection?	Low risk	"We used de‐identified data from the databases of British Columbia’s Ministry of Health to monitor use of and expenditure..."

Methods	Controlled ITS using segmented regression model with monthly data points
Participants	Tennessee Medicaid Program N=495,821 to 547,403, included regular NSAID users who received either non‐generic or generic NSAIDs for at least 274 days prior to policy.
Interventions	PA policy on NSAIDs
Outcomes	Health services utilization, Drug costs
Notes	Duration from baseline year: Oct 1, 1988 ‐Sep 30, 1989, post policy: Oct 1, 1989 ‐ Sept 30 1991, no transition period
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any difference in the proportion of missing data data in state Medicaid datasets pre to post intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the findings section
Other bias	Low risk
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	Low risk	Changes in the environment are unlikely to be responsible for magnitude of effect size.Note: In the second year after the policy change Medicaid prescriptions for non‐study drugs increased by 8%, an increase that corresponded to the implementation of the provisions of the Omnibus act of 1990.
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time.
Intervention unlikely to affect data collection?	Low risk	Objective data obtained from an administrative database with standard collection rules

Methods	Controlled cohort study ‐ ITS with monthly data points
Participants	Maine, USA, Medicaid and Medicare, newly treated patients with schizophrenia diagnosis n=4600
Interventions	PA policy on atypical antipsychotics
Outcomes	Drug use, drug costs, time to treatment discontinuation
Notes	Intervention state: Maine, control state: New Hampshire; Medicaid claims for 201‐2004. Policy period too short (policy discontinued) and study lacked power to detect treatment discontinuities.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results "Only Maine had enough patients to analyse utilization trends by preferred and nonpreferred categories"
Selective reporting (reporting bias)	Low risk	All outcomes listed in the methods section are reported in the results section
Other bias	Low risk	Treatment discontinuities common in schizophrenia. Maine had a significantly greater percentag of patients in the 18‐34 categories ‐ could confound the treatment discontinuities however these were not a study outcome.
Knowledge of allocated interventions adequately prevented during the study	Low risk	Outcomes are objective measures of healthcare utilization
Intervention independent of other changes	High risk	"The termination of Maine's PA policy after out study began resulted in limited follow‐up, reducing our statistical power to detect treatment discontinuities among newly treated patients"
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time.
Intervention unlikely to affect data collection?	Low risk	Objective data obtained from an administrative database with standard collection rules.

Methods	ITS with segmented regression analysis with monthly data points
Participants	Belgian National Health Insurance members
Interventions	Two policies issued: March 2001: H2As became available without restrictions; March 2003: two PPIs became available without resrictions
Outcomes	Utilization and costs
Notes	4 month transition period (2 months before and 2 months after) were included retrospectively
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Any undocumented difference in the proportion of missing data in the administrative datasets pre‐ and post‐intervention is unlikely to overturn study results
Selective reporting (reporting bias)	Low risk	All outcomes reported
Other bias	Low risk	No other source of bias detected
Intervention independent of other changes	High risk	"an evidence report that was distributed to all physicians and pharmacists in September 2004. Several events and interactions interfering with the policies have been discussed, but many other events may also have played a role."
The shape of intervention pre‐specified?	Low risk	The point of analysis is the point of intervention; ie, the date the policy intervention was implemented was used to delineate pre and post policy time periods with adequate data points to capture the shape of the pattern of intervention effect over time.
Intervention unlikely to affect data collection?	Low risk	Objective data obtained from an administrative database with standard collection rules.

PERMALINK

Pharmaceutical policies: effects of restrictions on reimbursement

Carolyn J Green

Malcolm Maclure

Patricia M Fortin

Craig R Ramsay

Morten Aaserud

Stan Bardal

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Policies that restrict reimbursement on some drugs to ensure better use in health care

Summary of findings

Summary of findings for the main comparison.

Summary of findings 2.

Summary of findings 3.

Background

Restriction to reimbursement policies

How the intervention might work

Implementation of restriction to reimbursement policies

Relaxation or exemption from restrictions

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Table 1.

Table 2.

Table 3.

Table 4.

Table 5.

Table 6.

Table 7.

Table 8.

Table 9.

Table 10.

Table 11.

Table 12.

Table 13.

Table 14.

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Dealing with missing data

Data synthesis

Subgroup analysis and investigation of heterogeneity

Results

Description of studies

Participants

Drug classes

Outcomes

Table 15.

Results of the search

Included studies

Study designs

Excluded studies

Risk of bias in included studies

Figure 1.

Figure 2.

Summary of appraisal

Effects of interventions

Effects of interventions

Drug use

Drug use effects by drug class

Drug expenditure

Relaxation or exemption from restriction to reimbursement policies

Study	Reason for exclusion
Ackman 2006	Fatally flawed study design: populations defined for comparison on the basis of prescription filling behaviour without knowledge of application for or denial of the medication targeted for restriction
Bloom 1985	Inadequate study design: uncontrolled before‐after (BA) study also excluded patients that discontinued treatment after policy implementation thereby producing an inadequate compartor
Gleason 2005	Inadequate study design: described as 'Prospective, pre‐ and post‐implementation cohort study with reference group.' Reference groups not suitable as comparators as they differed by prescription filling, use and PA authorization patterns. Quarterly drug usage and cost data with insufficient data points pre‐implementation.
Kahan 2006	Insufficient observation points before PA implementation
Kotzan 1993	Inadequate study design: uncontrolled BA study
Lu 2007	Inadequate study design: national trends in the use of DMARDs before and after the introduction of the PBS access criteria for biologicals. Study limitations include the absence of data on use of biologicals (the drug targeted by the restrictive policy) and the assumption that because the drugs studies were 'mainly' used as DMARDS that the national trend for all users of DMARDS would reflect changes in use patterns for the largest RA subpopulation
McCombs 2002	Inadequate study design: BA study without adequate control group: logistic regression analysis adjusted for differences between treated and untreated groups with major depressive disorder
Meissner 2007	Inadequate study design: uncontrolled BA study of medication switchers
Momani 2002	Inadequate study design: uncontrolled BA study
Phillips 1997	Inadequate study design: uncontrolled BA study ‐ simple comparison of baseline with after observation without consideration of time trend or bias arising from other environmental or policy changes
Sheely 2008	Inadequate study design: fatally flawed. Groups were compiled for comparison on the basis of prescription filling history without additional information on authorization process. Also patient deaths were attributed to failure to fill a clopidogrel prescription because of delays caused by the PA process whether or not the clopidogrel had been prescribed or authorization sought.
Siracuse 2007	Inadequate study design: denominator data not specified. The study design was labelled as cross‐sectional however data collected from administrative datasets on a quarterly basis so snapshot was not taken at a consistent point in time and individuals could have contributed data multiple times.
Virabhak 2005	Inadequate study design: regression modelling using data from 2 states with PA policies that differed in criteria used to determine reimbursement
Yokoyama 2007	Insufficient denominator data on PA applications for analysis within this study of step therapy versus PA only. "The use of rejected and paid claims data in this analysis, compared with PA data only, allows for identification of cost savings associated with ARB claims avoided or delayed as a result of the step‐therapy program. We identified those patients who encountered a rejected claim for an ARB and followed them for 12 months. While we do not know which patients attempted to get a PA or did not attempt to get a PA but were denied, we identified the antihypertensive therapy and associated costs in the 12‐month follow‐up period."