. 2014 Oct 20;14:469. doi: 10.1186/1472-6963-14-469

Table 2.

Key characteristics of the eligible studies

Authors	Population	Drugs	Methods	Variables
Álvárez and Hernández 2005 [16]	32 healthcare centres, 313321 inhabitants, Spain	50 new drugs	multiple linear regressions	Practice characteristics: size (number of doctors and number of patients), location (urban or rural), number of years functioning as a primary care centre, pharmaceutical expense per secured patients, number of consultations per doctor
Álvárez and Hernández 2005 [16]	32 healthcare centres, 313321 inhabitants, Spain	50 new drugs	multiple linear regressions	Patient characteristics: proportion of patients in pension
Behan et al. 2005 [17]	126991 inhabitants, 134 full-time equivalent GPs, Australia	2 new drugs (celecoxib and rofecoxib)	comparison of means (Student’s t-test)	Practice characteristics: location (urban or rural)
Bourke and Roper 2012 [18]	616 GPs and all their prescriptions, Ireland	6 new drugs, from 6 therapeutic classes	survival analysis	Prescriber characteristics: age, portfolio width, savings made from meeting prescribing targets, GP being an early adopter in at least one of the other five study drugs
				Practice characteristics: employee composition* (practice with nurse, practice with secretary), practice with in-house dispensary, location (urban or rural)
				Drug/market characteristics: time-variant percentage of GPs who have adopted the study drug
Coleman et al. 1966 [7]	125 GPs (prescriptions and interviews) and 103 SPs (interviews), four small cities in Illinois, US	1 new drug (tetracycline, a broad-spectrum antibiotic)	elementary statistics	Prescriber characteristics: age, prescribing volume in the therapeutic class of the new drug, speciality, number of contacts with drug representatives, number of professional journals read, number of pharmaceutical house organs read, number of speciality meetings attended, number of non-speciality meetings attended, number of hospital meetings attended, number of county medical society meetings attended, perceived scientific orientation, social position in advisor, discussion, and friendship networks
				Practice characteristics: type (solo or group/partnership)
				Note: The survey questionnaire resulted in a very large number of variables—only those most frequently discussed in the relevant literature are reported here
Corrigan and Glass 2005 [19]	4216 doctors, US	38 new compounds	analysis of covariance (ANCOVA) model	Prescriber characteristics: gender, age, board certification, hospital affiliation, type of doctor (trialist or control), prescribing volume
Dybdahl et al. 2004 [20]	191 practices, 470000 inhabitants, Denmark	14 new drugs, grouped in 4 categories	Pearson’s correlation coefficient	Practice characteristics: type (solo or group/partnership)
Dybdahl et al. 2004 [20]	191 practices, 470000 inhabitants, Denmark	14 new drugs, grouped in 4 categories	Pearson’s correlation coefficient	Drug characteristics: prescription cost*
Dybdahl et al. 2005 [21]	191 practices, 470000 inhabitants, Denmark	14 new drugs, grouped in 4 categories	multiple linear regressions	Practice characteristics: prescribing volume within the therapeutic class of the new drug, number of patients eligible for the new drug, prescribing volume of all other drugs, number of all other patients
Dybdahl et al. 2011 [22]	68 GPs, Denmark	2 new drug groups (COX-2 and AT-II)	univariate and multivariate linear regressions	Prescriber characteristics: perceived scientific orientation, perceived need for continuing medical education (CME), current CME activities*, previous hospital employment
Florentinus el al 2007 [23]	86 GPs, 13997 patients, the Netherlands	5 new drugs, from 5 therapeutic classes	logistic multilevel model	Prescriber characteristics: quality of pharmacotherapy audit meetings (PTAMs), PTAM composition (number of pharmacists and pharmacies, total number of participants, number of GPs)
García et al. 2000 [24]	74 GPs and SPs (paediatrics), Spain	28 new drugs, 10 with therapeutic novelty and 18 without	univariate and multivariate linear regressions	Prescriber characteristics: age, gender, speciality, type of contract (permanent or temporary), number of workplaces, drug expenditure captured by the deviation from the district’s mean
García et al. 2000 [24]	74 GPs and SPs (paediatrics), Spain	28 new drugs, 10 with therapeutic novelty and 18 without	univariate and multivariate linear regressions	Practice characteristics: management (reformed or non-reformed), region
Garjón et al. 2012 [25]	1248 doctors, Spain	8 new drugs, suitable for both primary and secondary care	survival analysis	Prescriber characteristics: speciality
Garjón et al. 2012 [25]	1248 doctors, Spain	8 new drugs, suitable for both primary and secondary care	survival analysis	Drug characteristics: therapeutic novelty
Glass 2003 [26]	1876 doctors, US	new drugs for the outpatient treatment of 8 disorders or diseases	comparison of means (Fischer’s least significant difference method)	Prescriber characteristics: investigator type (phase IIIb investigator, phase IV investigator, non-investigator)
Glass 2004 [27]	2108 clinical trial investigators, US	72 new compounds	multiple linear regressions	Prescriber characteristics: gender, age, board certification, hospital affiliation, per cent of working hours spent in a hospital setting, clinical research experience (total number of clinical studies, number of clinical studies conducted for the sponsoring company, number of clinical studies conducted in the therapeutic class of the new drug, number of clinical studies with the compound), relative grant amount received by investigator, pre-launch prescribing volume, pre-launch prescribing volume in the therapeutic class of the new drug, pre-launch prescribing of the sponsoring company’s products as a percentage of total prescribing (company loyalty), market share of the sponsoring company 3 and 6 months post launch as percentage of total prescribing volume
Glass 2004 [27]	2108 clinical trial investigators, US	72 new compounds	multiple linear regressions	Drug/market characteristics: therapeutic novelty, number of drugs tested for the same indication, participation of a contract research organisation, market share 1 year post launch
Glass and Rosenthal 2004 [28]	3646 doctors, US	32 new drugs	binomial logistic regression	Prescriber characteristics: gender, age, board certification, speciality, clinical investigation experience, hospital affiliation, pre-launch prescribing volume, pre- launch prescribing volume in the same therapeutic class as the new drug, pre-launch company loyalty
Glass and Rosenthal 2004 [28]	3646 doctors, US	32 new drugs	binomial logistic regression	Drug characteristics: therapeutic novelty, marketing budget of the pharmaceutical company assigned for the new drug
Glass and Rosenthal 2005 [29]	2287 clinical trial investigators, US	38 new drugs	ordinary least squares (OLS) and binomial logistic regression	Prescriber characteristics: gender, age, board certification, speciality, hospital affiliation, percentage of working hours spent in a hospital setting, clinical research experience (total number of clinical trials, total number of clinical studies conducted, total number of clinical studies conducted for the sponsoring company, total number of clinical studies conducted in the therapeutic class of the new drug, total number of clinical studies conducted with the compound), total pre-launch prescribing volume, pre-launch prescribing volume in the therapeutic class of the new drug, pre-launch prescribing volume of sponsoring company’s products, pre-launch prescribing volume of sponsoring company’s products as a percentage of total prescribing (company loyalty)
Glass and Rosenthal 2005 [29]	2287 clinical trial investigators, US	38 new drugs		Drug characteristics: market share of the new drug 12 months post launch, marketing budget of the pharmaceutical company assigned for the new drug
Glass and Dalton 2006 [30]	484 phase IV clinical trial investigators, US	new drugs for the outpatient treatment of 8 disorders or diseases	binomial logistic regression	Prescriber characteristics: gender, age, board certification, speciality, practice type (office, hospital, academic medical centre, other), per cent of working hours spent in a hospital setting, clinical research experience (number of clinical trials conducted, number of clinical studies conducted for the sponsoring company, number of clinical studies conducted in the same therapeutic class as the new drug, number of clinical studies conducted with the compound), relative grant amount received, pre-study prescribing volume, pre-study prescribing volume in the same therapeutic class as the new drug, pre-study prescribing volume of the new drug, pre-study prescribing volume of sponsoring company’s products, pre-study prescribing volume of sponsoring company’s products as a percentage of total prescribing (company loyalty)
Glass and Dalton 2006 [30]	484 phase IV clinical trial investigators, US		binomial logistic regression	Drug characteristics: market share of the new drug in its therapeutic class 12 months post launch, revenue of the pharmaceutical company
Greving et al. 2006 [11]	70 GPs, 9470 hypertensive patients, the Netherlands	1 new drug (angiotensin II receptor blocker (ARB))	multilevel logistic regressions	Prescriber characteristics: use of commercial information, use of medical journals, CME, use of other professional information, use of a prescribing decision support system, personal involvement in PTAMs
				Practice characteristics: type (solo or group/partnership), location (urban or rural)
				Patient characteristics: age, gender, number and type of comorbidities, referrals to internist/cardiologist
				Drug characteristics: perceived benefits (survey data)
JP Griffin and TD Griffin 1993 [31]	10 developed countries	drugs introduced in the last 5 years	descriptive statistics	Prescriber characteristics: nationality
Groves et al. 2010 [6]	925 doctors and all their prescriptions, Canada	4 new drugs (COX-2 inhibitors)	correlation analysis with t-tests	Prescriber characteristics: gender, age, birthplace, speciality, training location (domestic or overseas qualification), professional age, prescribing volume of four related drugs
Groves et al. 2010 [6]	925 doctors and all their prescriptions, Canada	4 new drugs (COX-2 inhibitors)	correlation analysis with t-tests	Practice characteristics: location (urban or rural)
Helin-Salmivaara et al. 2005 [32]	2558 doctors, 507262 prescriptions from the same therapeutic class, Finland	2 new drugs (celecoxib and rofecoxib)	general linear mixed model	Prescriber characteristics: gender, professional age, speciality
Helin-Salmivaara et al. 2005 [32]		2 new drugs (celecoxib and rofecoxib)	general linear mixed model	Patient characteristics: age, gender
Huskamp et al. 2013 [33]	30369 doctors, US	9 new drugs (second-generation antipsychotics)	Cox’s proportional hazard model	Prescriber characteristics: gender, age, speciality, hospital affiliation, prescribing volume in the same therapeutic class as the new drug, training location (domestic or overseas qualification, top- or non-top-25 medical school)*
				Practice characteristics: type (solo or group/partnership)
				Drug characteristics: degree of innovation (original formulation or reformulation)
Inman and Pearce 1993 [34]	3346 GPs, England	27 new drugs	descriptive statistics	Prescriber characteristics: gender, number of treatments, professional age, training location (domestic or overseas qualification)
Iyengar et al. 2011 [12]	185 doctors, US	1 new drug, third entry in the category (for treatment of viral infections)	discrete-time hazard model	Prescriber characteristics: speciality (GP or SP), number of patients, referral to other doctors, prescribing volume in the same therapeutic class as the new drug, self-reported leadership, detailing, indegree centrality (referral, discussion, and total network), outdegree centrality (referral, discussion, and total network), peer influence through adoption and usage
Iyengar et al. 2011 [12]	185 doctors, US		discrete-time hazard model	Practice characteristics: type (solo or group/partnership), type of hospital (university/teaching), location (city dummies)
Kozyrskyj et al. 2007 [35]	12 million patients and 2000 doctors, Canada	4 new drugs, from 4 therapeutic classes	polytomous logistic regression	Prescriber characteristics: age, gender, speciality, professional age, training location, hospital affiliation
				Practice characteristics: type (solo or group/partnership)
				Patient characteristics: age, gender, neighbourhood income quintile*, prescription reimbursement status, presence of chronic conditions
Lin et al. 2011 [36]	155 SPs (psychiatry) affiliated with 12 healthcare centres, Taiwan	1 new drug (antidepressant, in the selective norepinephrine reuptake inhibitor (SNRI) family)	Cox’s proportional hazard model	Prescriber characteristics: age, gender, speciality (in depressive disorder), hospital experience (number of current workplaces), outpatient services (number of visits), peers’ adoption ratio, opinion leaders’ adoption ratio, (dis)similarities with adopting peers’ age, gender, and tenure, SNRI proportion (past experience and preference for SNRI)
Lin et al. 2011 [36]			Cox’s proportional hazard model	Practice characteristics: size (number of SPs), ownership (public or private), location
Liu et al. 2011 [37]	41488 patients, 4429681 prescriptions, Taiwan	7 new drugs (oral hypo-glycemic agents, for treatment of diabetes)	logit model	Practice characteristics: market share (number of outpatient visits), ownership (public or private not-for-profit or private for-profit), accreditation level (clinic, academic medical centre, metropolitan hospital, or local community hospital)*, location
				Patient characteristics: age, gender, disease severity, number of prescriptions per visit
				Drug characteristics: prescribing volume in the same therapeutic class as the new drug
Liu and Gupta 2012 [13]	2129 doctors, US	1 new drug (for treatment of a chronic condition)	discrete-time hazard model	Prescriber characteristics: speciality, prescribing volume in the same therapeutic class as the new drug, number of detailing visits, number of medical meetings and events, number of patient requests made to doctors, number of peers in geographic proximity who had already adopted the new drug
				Drug characteristics: journal advertising, linear time trend
				Patient characteristics: median age of people in the community, average household income, health insurance index, percentage of white population
Manchanda et al. 2008 [38]	466 doctors, Manhattan (New York City), US	1 new drug (for treatment of a chronic condition)	discrete-time hazard model	Prescriber characteristics: contemporaneous effect of detailing, accumulated stock of detailing, accumulated stock of sampling, contagion measure (number of adopting doctors in geographic proximity)
Manchanda et al. 2008 [38]	466 doctors, Manhattan (New York City), US	1 new drug (for treatment of a chronic condition)	discrete-time hazard model	Drug characteristics: aggregate marketing expenditure (direct-to-consumer advertising (DTCA)), time since launch
Mark et al. 2002 [14]	187 doctors, 752 patients, prescriptions from medical records, US	4 new drugs (antipsychotics—(clozaril, risperidone, olanzapine, and quetiapine)	bivariate and multivariate probit regression analysis	Prescriber characteristics: age, gender, board certification, number of patients*, number of contacts with pharmaceutical representatives*, attendance on professional meetings*, preference for atypical initial treatments, preference for atypical treatments for nine different conditions**, percentage of patients influenced by individual medication costs
				Practice characteristics: location
				Patient characteristics: age, gender, race/ethnicity, education, marital status, insurance status, first onset of the disorder, diagnosis, symptoms, hospitalisation in the past 12 months
Mizik and Jacobson 2004 [39]	74075 doctors, US	1 new drug, within a well-established therapeutic area, and 2 older drugs	dynamic fixed effects distributed lag regression	Prescriber characteristics: speciality, detailing volume, sampling volume*, prescribing volume in the same therapeutic class as the new drug (competitor prescribing)
Ohlsson et al. 2009 [5]	73547 doctors, 32011 patients, Sweden	1 new drug (rosuvastatin, for treatment of high blood cholesterol)	generalised estimation equations and alternating logistic regression	Practice characteristics: ownership (public or private), proximity to SPs, location (urban or rural), size (prescribing volume)
Ohlsson et al. 2009 [5]	73547 doctors, 32011 patients, Sweden			Patient characteristics: age, gender, income, marital status, birthplace, length of residence in Sweden
Steffensen et al. 1999 [40]	319 GPs, 193876 prescriptions, Denmark	5 generically new compounds	multiple logistic regression	Prescriber characteristics: gender, age
Steffensen et al. 1999 [40]	319 GPs, 193876 prescriptions, Denmark	5 generically new compounds	multiple logistic regression	Practice characteristics: type (solo or group/partnership), size (number of patients), number of consultations per patient, number of telephone consultations per patient, number of home visits per patient, number of procedures performed per patient, number of laboratory tests performed per patient
Ruof et al. 2002 [41]	72 GPs, 28 SPs (neurology), Germany	1 new drug class (for treatment of Alzheimer’s disease)	Sperman’s rank correlation coefficient	Prescriber characteristics: speciality (GP or neurologist)
Ruof et al. 2002 [41]	72 GPs, 28 SPs (neurology), Germany	1 new drug class (for treatment of Alzheimer’s disease)	Sperman’s rank correlation coefficient	Drug characteristics (perceived): safety, efficacy, life quality improvement, nursing home admission delayed, budgetary impact
Tamblyn et al. 2003 [1]	1661 doctors, 669867 elderly patients, Canada	20 new drugs, from 6 therapeutic classes	multivariate logistic and conditional Poisson regressions	Prescriber characteristics: gender, speciality, professional age, training location*
				Practice characteristics: location (urban or rural), referral rate to SPs who prescribe drugs in the same therapeutic class as the new drug, size*
				Drug characteristics: detailing (minutes), advertising (pages)
				Patient characteristics: proportion of elderly patients in the practice population
Van den Bulte and Lilien 2001 [15]	121 GPs, four small cities in Illinois, US	1 new drug (tetracycline, a broad-spectrum antibiotic)	discrete-time hazard model	Prescriber characteristics: professional age, number of journals read, position, scientific orientation, status (number of nominations received as advisor or discussant—2 contagion variables to capture word of mouth operating over direct ties and 2 contagion variables to capture competition for status between structurally equivalent doctors
Van den Bulte and Lilien 2001 [15]	121 GPs, four small cities in Illinois, US	1 new drug (tetracycline, a broad-spectrum antibiotic)	discrete-time hazard model	Drug characteristics: seasonal effect, depreciation-adjusted stock of marketing effort by the first entrant, depreciation-adjusted stock of marketing effort by the two subsequent entrants

In alphabetical order, by first author. Variables in italics without asterisk: significant impact on new drug uptake in all specifications of the study. Variables in italics with asterisk: significant impact on new drug uptake in some specifications of the study.