Networks of General Practitioners in Italy: Any Impact? 1

2

Giovanni Fattore 3

Department of Institutional Analysis and Public Management, Bocconi University 4

Domenico Salvatore 5

Università Parthenope and Fondazione SDN 6

7

8

Word count: 6.535 9

10

2

ABSTRACT 11

BACKGROUND: At the end of the 1990' General Practitioners (GPs) in Italy were given 12

the opportunity to adopt network forms of organization with the aim of improving the quality of 13

their services. 14

PURPOSE: To provide a preliminary evaluation of the factors affecting GPs' choices to 15

join a network and the consequences of network membership on several measures of GP per-16

formances. 17

METHODOLOGY/APPROACH: Administrative data of a Local Health Authority in Cen-18

tral Italy were analyzed with statistical methods at the individual and at the dyadic level of 19

analysis. 20

FINDINGS: Homophily factors seem to drive a GP's choice of the network to join. Conse-21

quences of network membership on GP performances seem very limited. 22

PRACTICE IMPLICATIONS: When considering to foster the diffusion of network organ-23

izational forms in health care, creating a network structure as it has been done with Italian GPs is 24

not enough. Other features of the implementation phase, the work organization and human re-25

source management should also be considered 26

KEY WORDS: General Practitioner, Network Organization, Primary Care, Social Net-27

work Analysis, Italy 28

RUNNING HEAD: Networks in General Practice 29

30

3

Networks of General Practitioners in Italy: Any Impact? 31

INTRODUCTION 32

In organization theory, the idea that a network is a form of organization of activities just as hier-33

archy (i.e., a company or a public organization) or market has found legitimacy and popularity 34

during the first half of 1990 (Powell, 1990; Uzzi, 1996; 1997). The network form of governance 35

of activities which is characterized by “reciprocal patterns of communication and exchange” 36

(Powell, 1990) has been described as a hybrid form of organizations of activities offering some 37

advantages over both hierarchies and markets. During the same period, national and regional 38

governments have deeply reformed the Italian National Health Service (INHS), and by the end 39

of the decade a network form of organization of general practice was institutionalized in various 40

Regions (Cavallo et al. 2001; Mapelli and Lucioni, 2003). Those networks of general practitio-41

ners (henceforth GPs), termed “forme associative” in Italian, have been presented by politicians, 42

administrators and GP unions as the solution to many of the problems affecting the organization 43

of GPs in Italy as they create an organizational structure for professionals otherwise isolated and 44

facilitate knowledge exchange while keeping the administrative autonomy that Italian GPs had 45

always enjoyed. In this article we describe the structure of GPs networks in Italy and report evi-46

dence of the antecedents of their formation and of their consequences on GPs behaviour in one 47

Italian province. 48

NETWORK ORGANIZATIONS 49

In a network organization actors enjoy an independence similar to that they can enjoy in mar-50

kets. In networks actors such as GPs have incentives to increase efficiency and easy-to-observe 51

4

quality similar to those present in markets as well as the flexibility to experiment and adopt in-52

novations that may be useful in their specific context. The market context pressures GPs to in-53

vest in quality and to signal it to patients. In addition, when compared to market exchanges, 54

network arrangements frame repeated interactions and create greater communication between 55

GPs, that in turn facilitate knowledge sharing and the diffusion of innovations. In addition, better 56

circulation of information generates greater opportunities to evaluate difficult-to-observe quality 57

and increase the cost of behaving opportunistically. 58

In a network organization actors also enjoy stability of relationships similar to that present in 59

hierarchies. Actors can develop trust and are able to make organization-specific investments 60

such as specializing on some activities which would have lower value once the GP is independ-61

ent or moves to another organization. But when compared to hierarchy, networks allow greater 62

efficiency and flexibility in knowledge-intensive activities mainly because networks better con-63

trol opportunistic behaviour through norms of reciprocity and reputation concerns and therefore 64

avoid the need of direct supervision or of ex-ante specification of processes characterizing bu-65

reaucratic control. 66

At the individual level of analysis, social network theorists have focused on the behavioral, 67

perceptual, and attitudinal consequences (Inkpen and Tsang, 2005) of the social relationships in 68

which actors are embedded (Granovetter, 1985; Coleman, 1988). Those relationships are the ties 69

linking nodes of the networks and are also said to be the social capital of actors (Burt, 2000; 70

Salvatore, 2006). Ties influence actors behaviors through four mechanisms: network closure de-71

termines the social control typical of small and closed communities; brokerage determines the 72

power and informational advantage of actors bridging other actors otherwise disconnected; con-73

5

tagion determines the adoption of a characteristic of connected actors; and, prominence which 74

favors the adoption of a characteristic of an actor central to the network. 75

One variable that many social network theorists have been concerned with is the redundancy 76

of connections, which is the degree to which an actor's ego network is made by contacts that are 77

already connected to each other. Being connected to a highly redundant network implies having 78

access to fewer resources than someone connected to a non-redundant network with the same 79

amount of ties has. Non redundant contacts create social capital through the brokerage mecha-80

nism (Burt, 1992). 81

82

METHODS 83

Context 84

The INHS was established in 1978 and modelled after the British NHS. It is a mainly public sys-85

tem financed through general taxation (Fattore, 1999). From an organizational viewpoint the 86

INHS divides the Italian territory in 180 Local Health Authorities (LHAs) each responsible for 87

the provision of health services to its population. 88

GPs are primary care physicians working for LHAs as independent contractors and act as 89

gatekeepers of higher levels of care. In 2005 there were 47,022 GPs in Italy costing about 5 bil-90

lion Euros that is the 6% of public healthcare expenditure (Agenzia Nazionale per i Servizi Sani-91

tari Regionali, 2007). Traditionally, Italian GPs work in solo practices without any auxiliary 92

staff and any formal linkages with other GPs. Their status of independent but exclusive contrac-93

tors with a LHA is coupled with a financing system based on capitation: GPs are paid on the ba-94

6

sis of the number of patients on their list (Longo, 2007). The freedom of patients to choose the 95

list they are on and to switch GP in every moment induces a form of competition among GPs to 96

keep or acquire patients 97

Since the INHS inception in 1978, various initiatives have slowly but constantly shifted re-98

sources and responsibilities to non-hospital settings. The number of hospital beds in Italy de-99

creased by 28% from 1995 to 2005 (Lecci and Maestri, 2007). This process of de-hospitalization 100

of care has increased the workload and the responsibilities of GPs, as they are a pivotal part of 101

community care, given their gatekeeping role and their direct knowledge of patients. Nonethe-102

less, Italian health reforms have not changed the role nor the organization of the GPs’ activities. 103

The diffusion of network forms of organization eventually reached the organization of Ital-104

ian general practitioners through their definition in the national contract agreement and repre-105

sented the main, if not the only, major organizational innovation of GPs’ activity since WWII. 106

GP networks were and still are presented as the solution to many of the problems affecting the 107

organization of primary care in Italy as they create an organizational structure and facilitate 108

knowledge exchange, while keeping GPs administrative autonomy. 109

General practitioners networks 110

The contract according to which GPs provide services to citizens enrolled in their list and fi-111

nanced by the INHS is negotiated every three years by the GPs unions with a governmental 112

agency and represents the main document regulating the relationship of GPs with the INHS. Ad-113

ditionally, integrative contracts may be signed by unions and organizations and local govern-114

ments at the regional or local level. Through the national contract, organizational innovation, 115

such as the GPs networks, were introduced into the system. The first national contract mention-116

7

ing the idea that organizational forms creating some collaboration among GPs could be negoti-117

ated at the local level was signed in 1996, but only when the contract was renewed in 2000 the 118

regulation of GP networks was detailed and networks an important element in GP regulation. 119

In the current form of the national contract, GPs willing to undertake some form of collabo-120

ration in the provision of health services to their patients can choose among three forms. They 121

are named association (“medicina in associazione”), net (“medicina in rete”), and group (“medi-122

cina di gruppo”) and each determines a different strength of the collaboration among GPs (table 123

1). By simply joining one of these typified forms GPs have the right to additional compensation 124

from the INHS which tops a 18% increase over the base GP salary. None of these GP networks 125

are legal entities and every contractual relationship is among the INHS and the individual GP 126

member of the network. In any of these three forms GPs have to coordinate their office hours to 127

be open till 7pm during weekdays and to commit to share guidelines and meet to discuss and 128

improve their work. Each patient stays listed with one of the GP of the network, but when the 129

GP is unavailable he/she may be visited by any other GP in the network. The freedom of patients 130

to leave their GP and to enrol in the list of another in the same network is limited in order to 131

mitigate the fear of GPs that, by collaborating could lose patients in favour of colleagues. In the 132

case of network of intermediate intensity, the net, GPs also have to share an electronic database 133

of their patients. While in associations or nets GPs keep working in their own office, the most 134

intense form requires that GPs also share the clinic and, therefore, are in the position of sharing 135

investments in medical equipment or employ nursing or administrative staff. 136

------------------------------------------------------------------- 137

INSERT TABLE 1 ABOUT HERE 138

8

------------------------------------------------------------------- 139

The basic idea behind these networks is that organizational and professional development in 140

primary care requires collaboration between GPs and sharing of resources and knowledge, but 141

also that networks can improve service delivery (office hours, continuity of care, wider range of 142

services). As of 2004, 59% of Italian GPs had entered into a type of network, 22% of which 143

joined a group, the more intense organizational form (Ministero della Salute, 2004). GP net-144

works are, indeed, network of peers, which each individual GP is free to choose whether to join 145

with the colleagues he/she prefers. Contrarily to experiences in other countries with primary care 146

organizations, GPs network in Italy are organizational arrangements for which every member is 147

a GP and specialists are not involved. Although most groups (the more intense and less adopted 148

form) employ a part-time nurse or an administrative assistant, associations, nets, and groups are 149

better understood as uni-professional networks. Members of each network are free to choose 150

their governance structure which, given that these organizational forms are small networks of 151

peers which GPs can leave whenever they prefer, is very likely that any decision is taken by 152

unanimous agreement. Moreover, sharing of guidelines and meeting to discuss and improve their 153

work, two of the main requirements of any of GP networks are vaguely defined and very diffi-154

cult to verify by the LHA. 155

DATA 156

Data used in this article was provided by the Piacenza LHA in Emila-Romagna, a Regional in 157

the central part of the country. This LHA is responsible for an area with approximately 280.000 158

inhabitants and comprises a town of 100.000. The administrative information systems of the 159

LHA contains a large data-set useful to investigate both antecedents and consequences of par-160

9

ticipation in GP networks. In 2006, the year analyzed in this study, there were 1.23 GPs every 161

1,000 residents in the LHA; 69% of GPs had entered into a type of network and 16% of them 162

joined a group. Some of the analyses reported in this paper were performed at the individual 163

level of analysis (i.e. the individual GP) and some were performed at the dyad-level of analysis 164

(i.e., the unit of observation is the pair of GPs). 165

------------------------------------------------------------------- 166

INSERT TABLE 2 ABOUT HERE 167

------------------------------------------------------------------- 168

Participation in a network 169

Based on administrative data, we created affiliation matrices linking each GP to all possible 170

networks. The LHA keeps record of which network each GP has joined for its administrative 171

purposes. In the following analysis we used this binary variable both as independent and de-172

pendent variables. For the analysis conducted at the dyadic level of analysis this variable be-173

comes “participation to the same network” and is coded 1 if both GPs are members of the same 174

network and 0 if they are members of different networks or are not members of any network. 175

Indeed, a connection between two GPs exists when they are members of the same network. This 176

type of data is named in social network analysis “affiliation networks” (Wasserman and Faust, 177

2004).14

In the LHA studied, the network created by co-membership in the same GP network has 178

a density equal to 0.012 and each GP, or each node in social network terminology, has an aver-179

age degree of 2.6696. 180

181

1

0

GP Performance dimensions 182

GPs activity is complex and is very hard to find any single variable which may represent satis-183

factorily their performance. Using the data collected by the LHA, we report the analyses at the 184

individual and at the dyadic level on the consequences of participation in GP networks using ten 185

dependent variables. Nine of them capture specific aspects of their performance, and one is a 186

performance index that summarizes them. The first variable, “Pharma”, is per-capita pharmaceu-187

tical expenditure of the patients enrolled with a GP. With the exception of some particular cate-188

gories of drugs, in Italy patients need GP prescriptions to get drugs under the INHS. GPs are 189

therefore gatekeepers of the pharmaceutical expenditure of their patients and, in order to keep 190

INHS expenditure under control, they are under pressure to keep it as low as possible. The sec-191

ond, “Home Care” measures how many patients of the GP are visited regularly at home by the 192

GP; home care is specifically encouraged by LHAs as it is assumed to reduce inappropriate hos-193

pitalization and increase the quality of care for the patients. For the same reason, GPs may coor-194

dinate a group of other professionals (for example a nurse, a therapist and a specialist) who regu-195

larly visit the patient; the number of patients benefiting from this activity for each GP is meas-196

ured by the variable “Multi-professional home care”. The variable “Preventable re-admissions” 197

measures the number of patients who are re-admitted to hospital for conditions that could have 198

managed effectively by the GP. This measure detects a lack of adequate continuity of care after 199

the patient’s discharge. “Printed prescription” is the percentage of prescriptions that has been 200

printed by the GP using a computer and (thus not handwritten). Printing prescription is encour-201

aged by the LHA because it guarantees that the GP keeps an electronic archive and eases part of 202

the subsequent administrative work. “Screening Paptest” is the fraction of targeted patients of a 203

GP target by the paptest screening program who has joined the program; “Screening (average)” 204

1

1

is the average participation rate of the patients of a GP to the three screening programs working 205

in the LHA during 2006: pap-test, mammogram, and colon-rectal. GPs are assumed to play a 206

role in the health promotion of their patients and to inform them about the advantages of the 207

screening programs. “White Codes day” is the number of times patients of the GP have been in 208

an emergency department during the day and have been assigned a white code (lowest level of 209

priority). “White Codes Total” is the number of times patients of the GP have been in an emer-210

gency department and been assigned a white code during any time of the day. GPs are assumed 211

to be able to manage most of the non-urgent health issues and white codes can be partially attri-212

buted to the quality of GPs’ work. To try to have a single, although arbitrary measure of the per-213

formance we ranked each GP along these nine performance dimensions and thus we created a 214

tenth variable “Performance Index” which is simply the average rank of a GP along the nine 215

other dependent variables. 216

All these variables have been computed on a per-patient basis, that is dividing the dependent 217

variable by the number of patients enrolled in each GP list. For the analysis at the dyadic level 218

each variable becomes the difference in the value of the variable of the first and the second GP 219

in the dyad. 220

Control Variables 221

The control variables used in the models include gender, the years since the medical degree, a 222

dummy variable coded 1 if the GP has any kind of specialization other than the medical degree, 223

the number of patients that are in the GP’s list; a Risk Adjustment Index depending on the age of 224

the enlisted patients and which is used by the LHA to weight the number of patients in its inter-225

nal statistics; and the three dummy variables to identify the district where the GP operates. Dis-226

1

2

tricts are the organizational unit of the LHA to which GP management is delegated, they are re-227

sponsible for the same services on different geographical areas. The district which has the same 228

borders of the main city is the reference one in the analysis. 229

FINDINGS 230

We used these data to study both the factors affecting the choice to enter a GP network and the 231

consequences of being a member. In both cases we performed the analysis at the individual level 232

and at the dyad level. By “dyad” is meant each of all the possible pairs of GPs in the sample. At 233

the dyadic level of analysis values of the variables for each dyadic observation has been calcu-234

lated as the difference of the value of the first GP and the value of the second in the case of con-235

tinuous variables or by the presence of a characteristic for both GPs in the case of binary vari-236

ables. 237

Factors affecting the choice to enter a GP network 238

Table 3 shows the results of a logistic regression for which the dependant variable was the par-239

ticipation of a GP in any of the three GP networks. Among the GPs of the studied LHA, the 240

number of patients has a small but highly significant effect and being female makes more likely 241

to join a network (p <0.1). 242

----------------------------------------------------------------- 243

INSERT TABLE 3 ABOUT HERE 244

------------------------------------------------------------------- 245

1

3

In table 4 we present an analysis at the dyadic level which investigates factors predicting the 246

presence of a connection between two GPs. The analysis uses an exponential random graph 247

model (ERGM), also called the p* class of models, that is a probability model for networks on a 248

given set of actors allowing generalization beyond the restrictive dyadic independence assump-249

tion (Robins et al, 2007a; 2007b). The analysis has been performed using the software SIENA 250

(Snijders, 2006) .17

The dependant variable is the co-membership of two GPs in the same col-251

laboration initiative (the level of analysis is, thus, dyadic) and the independent variables are: be-252

ing of the same gender, the difference of the year since graduation form medical school of the 253

two GPs in the dyad, the fact that the two GPs have both a specialization or they both do not 254

have one, and the similarity in the number of patients enrolled with each GP. Two variables are 255

significant: the similarity of the year of graduation and the similarity in the number of patients 256

enrolled. These results may mean that GPs tend to join a GP network with colleagues of their 257

generation and with colleagues that have a similar number of patients (therefore a similar use of 258

common resources). 259

------------------------------------------------------------------- 260

INSERT TABLE 4 ABOUT HERE 261

------------------------------------------------------------------- 262

263

Consequences of participation to GP networks 264

To study the consequences of participating in any GP network, we used an ordinary least square 265

(OLS) regression. Table 4 reports 10 regression models, where the same set of regressors are 266

1

4

tested and 10 different dependent variables are used (9 specific measures and an overall per-267

formance index). The two variables of interest are “in any network” representing the participa-268

tion of the GP to any of the three forms of network and “in group” representing the participation 269

to the network implying the most intense collaboration. As shown in table 4 being in a network 270

has a positive and significant effect only on the percentage of prescriptions printed using a com-271

puter and being in a group has a significant effect only in reducing the number of patients going 272

to an hospital emergency department for a condition classified as a white code by the hospital. 273

------------------------------------------------------------------- 274

INSERT TABLE 5 ABOUT HERE 275

------------------------------------------------------------------- 276

To analyze the consequences of participation to a GP network at the dyadic level we ran a mul-277

tiple regression technique called QAP (Krackhardt, 1988). This technique is able to regress a 278

valued matrix on other matrixes and does not assume independence of observations. MR-QAP 279

were performed to assess the similarity in each of the ten performance dimensions of each of all 280

the possible dyads of GPs in the sample (table 5) using the UCINET software package, version 281

6.172 (Borgatti et al, 2002). Being in the same network significantly makes two GPs similar in 282

the percentage of prescriptions they print with a computer, suggesting that some form of conta-283

gion or of social control goes on inside networks. Being in the same network also makes two 284

GPs significantly more different in the number of the patients they visit at home, maybe suggest-285

ing that some form of division of labour goes on in networks. Considering all the dimensions of 286

GP performance for which we had data, overall, the standardized effect on the similarity of per-287

formance of being part of the same network are very small and generally non-significant. 288

1

5

------------------------------------------------------------------- 289

INSERT TABLE 6 ABOUT HERE 290

------------------------------------------------------------------- 291

292

DISCUSSION 293

We reported results on the factors affecting participation and on the consequences of participa-294

tion of Italian GPs to three forms of network organizations that have been institutionalised by 295

their national contracts. We found that when GPs are left free to chose the colleagues with 296

whom to form a network, as it is the case in Italy, they tend to join networks of colleagues who 297

have similar age and number of patients in their rosters. We also found that participation to the 298

GP network has a very weak effects on most of the performance indicators we had available. 299

Most of the performance indicators do not seem to improve for GPs in networks nor they are 300

more similar among GPs who are in the same network. An exception is for the number of pre-301

scriptions GPs print with a computer rather than handwrite, which is an important dimension of 302

GP performance as it eases the burden of subsequent administrative work but may not be as fun-303

damental as other dimensions more directly related to patient care. The evidence reported sug-304

gests that the GP networks of the type created and promoted in Italy at the end 1990s have had 305

modest consequences on the ability of GPs to care for patients. Specifically, informal sharing of 306

knowledge may not be present or it may not have an impact on the performance of GPs meas-307

ured in most of the dimensions of performance. Theoretical contributions to the network form of 308

1

6

organization and on social networks allow to identify four characteristics of the GP network that 309

are regulated by the Italian GPs’ national contract which may have contributed to these results. 310

First, the contract focused on creating a structure for knowledge exchange but the presence of 311

a structure does not imply that knowledge actually flows in the network. The creation of an in-312

terpersonal network, indeed, does not imply that resources, knowledge or others, actually flow 313

through the connections that have been created. Studies on social networks are often heralded as 314

the tool to uncover the informal dynamics behind organizational charts (Krackhardt and Hanson, 315

1993), but, being empirically difficult to gather data on actual flow of resources in network, 316

most empirical studies have focused on network structure assuming that structure is a good 317

proxy for the flow of resources. Many scholars have also posited that structure has consequences 318

on its own; in management the work of Ronald Burt (1987, 1992, 2001) typifies this approach. 319

A concern with structure can also be found in most public management reforms in Italy as struc-320

ture is considered the more objective part of the organization and it is therefore the one bureauc-321

racies are most comfortable with. But the experience of GP networks in Italy suggests that the 322

creation of a network structure through the concession of monetary incentives to “get together” 323

and join a collaboration initiative does not guarantee that sharing of resources that rhetoric on 324

network promises. 325

Second, the redundancy of the connections both in terms of redundancy of ties and of similar-326

ity of knowledge possessed by the people connected by those ties reduces the probability that 327

new knowledge enters the network. Italian GP networks are essentially uni-professional organ-328

izational forms and, therefore, the diversity of knowledge resources is not leveraged. Moreover, 329

the spontaneous birth of social networks is widely believed to be influenced by homophily (i.e., 330

1

7

the tendency to create relationships with someone similar on relevant respects) (McPherson et al, 331

2001). Homophily also drives the formation of GP networks in this setting as suggested by table 332

4. GPs in the sample tend to join networks of GPs who graduated from medical school in the 333

same period who are also likely to have a similar education and knowledge, therefore, creating 334

redundancy in the type of information a GP is able to access. 335

Third, all network forms analysed here are organizational forms where parts are undifferenti-336

ated. Each GP in the network performs the same type of activities to patients who are in the 337

great majority of the cases always seeing the same GP. Therefore GP networks in Italy do not 338

leverage the advantages of specialization that network forms of organization may create. More-339

over the lack of strong interdependencies reduces the possibility of knowledge exchanges 340

through observation as it would happen if GPs systematically see the patients that are also seen 341

by their colleagues. 342

Finally, the relationship of the INHS with its GPs has always mainly focused on monetary 343

compensation with little attention paid to other forms of remuneration and incentives. The atti-344

tude of the INHS towards GPs has always been transactional rather than relational. A relational 345

psychological contract is a work relationship that is understood by the employee and the em-346

ployer not only as monetary as a long-term or open-ended employment arrangements based 347

upon mutual trust and loyalty (Rousseau, 1995). A transactional attitude, typically reflected in 348

contracts and explicit rules, may induce GPs to perceive their obligations as simply fulfilling the 349

clear expectations about the organizational standards. An approach more relational, instead, may 350

induce GP to perceive their obligations as working together to achieve the complex goals for 351

which GP networks were introduced and to invest in the relation per se. 352

1

8

The introduction of GP networks has focused on the incentives to motivate GPs to enter in 353

such networks, without sufficient attention to the context necessary to let networking have im-354

pact on the way professionals work. The government succeeded in having GPs accepting to join 355

these collaborative initiatives, but the evidence we collect do not support the claim that these 356

network modify how professionals take care of their patients. Probably, they are structural ar-357

rangements that do not change the pattern of communication among members. 358

PRACTICE IMPLICATIONS 359

These findings suggest that managerial strategies that try to increase networking among employ-360

ees should not focus solely on the creation of structure, as it has often been done in the INHS, 361

but should rather focus on the creation of a need to access others’ resources. Italian GPs in the 362

same network are not interdependent, and therefore, being part of the same network does not 363

imply interaction nor exchange. Moreover, in order to take advantage of the informational ad-364

vantages of network form of organizations, members should be selected in order to have hetero-365

geneous competencies so that each actor can find the knowledge needed in a specific situation 366

by activating the more expert colleague in the network on the issue at stake. Another important 367

lesson that can be learned from the introduction of GP networks in Italy is in the way the adop-368

tion of complex organizational structures should be promoted among GPs. Focusing only on 369

monetary incentives induces GP to comply with the contractual requirement which, although 370

superficial, is observable and thus enforceable. Since it is too difficult to enforce a contract re-371

quiring knowledge exchange among GPs and is also difficult to design an incentive structure to 372

reward successful knowledge exchange in medicine, the type of exchange should be more rela-373

tional in nature. 374

1

9

REFERENCES 375

Agenzia Nazionale per i Servizi Sanitari Regionali (2007). Evoluzione della spesa sani-376

taria 2001-2005. Available at: www.assr.it/Spesa/evoluzione_spesa_01-05.pdf 377

Borgatti SP, Everett MG, Freeman LC. (2002). Ucinet for Windows: Software for Social 378

Network Analysis. 379

Brunello C. (2001). Le forme di associazionismo medico: contenuti assistenziali e fattori 380

produttivi di sostegno. In: Il budget e la medicina generale, Longo F. and Vendramini, E 381

(eds). Milano: McGrawHill. 382

Burt RS. (1987). Social Contagion and Innovation: Cohesion Versus Structural Equiva-383

lence. American Journal of Sociology.92(6):1287―1335. 384

Burt RS. (1992). Structural holes : the social structure of competition. Cambridge, Mass.: 385

Harvard University Press:313―. 386

Burt RS. (2001). Structural Holes versus Network Closure as Social Capital. In: Burt RS, 387

Lin N, Cook KS, Social capital : theory and research. New York: Aldine de Gruy-388

ter;:333―. 389

Burt R. (2000). The Network Structure of Social Capital. In: Sutton R and Staw BM, 390

Research in Organizational Behavior.Vol 22. Greenwich CT: JAI press. 345―423. 391

Burt RS.(1992) Structural Holes: The Social Structure of Competition. Cambridge: MA: 392

Harvard University Press. 393

2

0

Cavallo M.C., Gerzeli S., Vendramini E. (2001) “Organizzazione e gestione delle Cure 394

Primarie” Milano - Mc Graw-Hill. 395

Coleman JS. (1988) Social Capital in the Creation of Human Capital. American Journal 396

of Sociology. 94(Supplement: Organizations and Institutions: Sociological and Economic 397

Approaches to the Analysis of Social Structure):S95―S120. 398

Granovetter M. (1985) Economic Action and Social Structure: The Problem of Embed-399

dedness. American Journal of Sociology. 91(3):481―510. 400

Fattore G. (1999) Clarifying the scope of Italian NHS coverage. Is it feasible? Is it desir-401

able? Health Policy. 50(1-2):123-142. 402

Inkpen AC, Tsang EWK. (2005) Social Capital, Networks, and Knowledge Transfer. 403

Academy of Management Review.30:146―165. 404

Krackhardt D. (1988). Predicting with Networks: A Multiple Regression Approach to 405

Analyzing Dyadic Data. Social Networks. 10:359-381. 406

Krackhardt D, Hanson JR. (1993) Informal networks: the company behind the chart. 407

Harvard Business Review. 71(4):104-11. 408

Lecci F., Maestri B. La struttura del Servizio Sanitario Nazionale . In: Rapporto OASI 409

2007. L’aziendalizzazione della sanità in Italia , Anessi-Pessina, E. and Cantù, E (eds) . 410

Milano: Egea. Available at: 411

http://portale.unibocconi.it/wps/wcm/connect/Centro_CERGASit/Sanit%C3%A0+in+Ital412

ia%3A+dati+e+analisi/Rapporto+OASI/Per+anno/Rapporto+OASI+2007/. 413

2

1

Longo, F. (2007). Implementing managerial innovations in primary care: can we rank 414

change drivers in complex adaptive organizations? Health Care Management Review, 415

32(3), 213-25. 416

Mapelli, V., & Lucioni, C. (2003). Spending on pharmaceuticals in Italy: macro con-417

straints with local autonomy. Value in Health: The Journal of the International Society 418

for Pharmacoeconomics and Outcomes Research, 6 Suppl 1, S31-45. 419

McPherson M, Smith-Lovin L, Cook JM. (2001) Birds of a Feather: Homophily in Social 420

Networks. Annual Review of Sociology. 27(1):415-444. 421

Ministero della Salute (2004). Ricognizione ed analisi dei modelli organizzativi innovati-422

vi per l’erogazione di cure primarie. Available at: 423

http://www.ministerosalute.it/imgs/C_17_pubblicazioni_472_allegato.pdf [Last accessed 424

September 26th

, 2008]. 425

Pellegrini L. (2005) Appropriatezza nell'erogazione dei livelli essenziali di assistenza fra 426

compatibilità economica e garanzie per i cittadini. Available at: http://www.assr.it. 427

Powell W. (1990). Neither Market nor Hierarchy: Network Forms of Organization. Re-428

search in Organizational Behavior. 12:295-336. 429

Robins G, Pattison P, Kalish Y, Lusher D. (2007a). An introduction to exponential ran-430

dom graph (p*) models for social networks. Social Networks. 29(2):173-191. 431

Robins G, Snijders T, Wang P, Handcockc M, Pattison P. (2007). Recent developments 432

in exponential random graph (p*) models for social networks. Social Networks. 433

2

2

2007;29(2):192-215. 434

Rousseau DM. (1995). Psychological Contracts in Organizations: Understanding Writ-435

ten and Unwritten Agreements. Sage. 436

Salvatore D. (2006). Physician Social Capital: Its Sources, Configuration, and Useful-437

ness. Health Care Management Review. 31(3):213-222. 438

Snijders TAB, Steglich CEG, Schweinberger M, Huisman M. (2006). Manual for SIE-439

NA, version 3. Groningen, The Netherlands: University of Groningen. 440

Uzzi, B. (1996). The Sources and Consequences of Embeddedness for the Economic Per-441

formance of Organizations: The Network Effect. American Sociological Review, 61, 442

674-698. 443

Uzzi B.(1997). Social structure and competition in interfirm networks: The paradox of 444

embeddedness. Administrative Science Quarterly. 1997;42(1):35-67. 445

Wasserman S, Faust K. Social Network Analysis: Methods and Applications. Cambridge 446

University Press; 1994. 447

448

Table 1: Features of the three forms of collaborative initiatives typified in the national labour 449

contract of GPs.(adapted from Brunello, 2001) 450

Associations Nets Groups

Number of members 3 to 10 3 to 10 3 to 8

Minimum number of

office hours during

weekdays

6 6 6

Shared electornic pa-

tient records

No Yes Yes

Shared ambulatory No No Yes

Additional income for

GP for each enrolled

patient

2.58 €

(7% increase)

4.70 €

(12% increase)

7 €

(18% increase)

Table 2: Means, standard deviations and correlations of the variables employed in the analyses 451

Mean Std.Dev. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1. Gender1

0.21

2.DistrictA 0.27 0.21

3. DistrictB 0.06 0.06 -0.15

4. DistrictC 0.19

-

0.05 -0.29 -0.12

5. Years since

graduation 25.10 6.85

-

0.36 -0.10 -0.03 0.02

6. Number of

patients 1111.22 463.01

-

0.33 0.01 -0.08 0.02 0.50

7. Risk Adj.

Index 1.8 0.26

-

0.09 0.05 0.45 -0.01 0.17 0.23

8. Specialized 0.65

-

0.03 0.06 -0.02 0.06 0.16 0.00 0.05

9.In any net-

work 0.67

-

0.06 -0.10 -0.03 -0.12 0.28 0.43 0.13 -0.04

10.Association 0.3 0.08 0.04 0.01 -0.20 0.06 0.09 0.05 0.02 0.46

11.Net 0.3

-

0.13 -0.31 0.00 0.15 0.21 0.29 0.08 -0.05 0.47 -0.43

12.Group 0.07

-

0.01 0.32 -0.07 -0.13 0.03 0.11 0.00 -0.03 0.19 -0.18 -0.18

13. Pharma 115.45 82.29 0.15 0.02 0.24 -0.02 -0.09 -0.11 0.34 0.07 0.06 -0.05 0.10 0.03

14. Home Care 7.23 6.93

-

0.04 0.08 0.18 0.27 0.05 0.18 0.20 0.06 -0.04 -0.08 0.06 -0.04 -0.01

15. Multi-

professional

home care 13.65 12.49

-

0.05 0.02 -0.01 -0.04 0.03 0.27 0.19 0.08 0.07 0.06 0.03 -0.03 0.02 0.22

16. Printed

Prescriptions 0.7 0.36

-

0.07 0.02 -0.13 -0.21 0.15 0.27 -0.10 0.01 0.43 0.09 0.28 0.14 0.00 0.02 -0.01

17. White

Codes day 7.09 5.04

-

0.01 0.13 -0.15 0.05 0.13 0.42 -0.10 0.00 0.11 0.01 0.12 -0.04 0.02 0.15 0.12 0.1

18. White

Codes Total 37.85 20.56

-

0.09 0.13 -0.15 0.15 0.26 0.60 -0.01 0.02 0.12 0.00 0.16 -0.09 0.06 0.16 0.13 0.12 0.79

19. Preventa-

ble re-

admissions 5.01 3.59

-

0.25 -0.10 0.11 -0.11 0.29 0.54 0.27 -0.08 0.20 0.05 0.20 -0.08 -0.05 0.06 0.31 0.07 0.2 0.36

2

5

Mean Std.Dev. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

20. Screening

Paptest 0.44 0.13 0.12 0.44 -0.16 -0.19 0.18 0.15 0.17 -0.02 0.11 0.04 0.00 0.12 0.07 -0.01 -0.02 0.03 0.12 0.13 0.03

21. Screening

(average) 0.55 0.16 0.05 0.21 -0.21 -0.21 0.24 0.35 0.11 0.06 0.10 -0.02 0.11 0.01 -0.09 -0.03 0.12 0.2 0.23 0.29 0.24 0.68

22. Perfor-

mance Index 106.75 23.15

-

0.04 -0.06 0.02 0.20 0.05 0.25 0.10 -0.09 0.01 0.01 0.05 -0.08 0.26 -0.24 -0.15 -0.24 0.47 0.53 0.40 -0.1 -0.17

1 Male =0; Female=1452

Table 3: Results of a Logistic Regression Model for the membership in a collaboration initiative 453

(-2 Log likelihood = 232.720; Cox & Snell R Square = 0.220; Nagelkerke R Square = 0.303). 454

Variable Beta (standard errors)

Gender 0.812 (0.440)*

Years since graduation 0.032 (0.023)

Specialized -0.453 (0.343)

Number of patients 0.002 (000)**

Constant -2,115 (0.614)**

* p< 0.10; ** p< 0.05. 455

456

457

458

459

Table 4: Results of a exponential random graph regression model for the presence of any con-460

nection between two actors (n=223). 461

Variable Estimates (standard errors)

Same Gender -0.0639 (0.1490)

Years since graduation Similarity 1.5726 (0.5109)**

Same Being specialized -0.0606 (0.1226)

Number of patients Similarity 1.7080 (0.3132)**

** p< 0.05 462

463

2

7

Table 5: Standardized regression coefficients (p-value) of linear regression models with individual GPs performance indicators as Depen-464

dent Variables 465

Pharma Home Care Multi-

professional

home care

Preventable

re-

admissions

Printed Pre-

scriptions

Screening

Paptest

Screening

(average)

White Codes

day

White Codes

Total

Performance

Index

Intercept 0.00 (0.94) 0.09 (0.20) 0.01 (0.88) 0.00 (0.95) 0.02 (0.77) -0.02 (0.69) 0.00 (0.97) 0.00 (0.97) 0.01 (0.92) -0.01 (0.81)

Gender 0.11 (0.11) 0.07 (0.38) 0.04 (0.59) -0.01 (0.95) -0.01 (0.93) 0.15* (0.02) 0.22**

(0.00)

0.18* (0.01) 0.23**

(0.00)

0.11 (0.13)

DistrictA -0.02 (0.80) 0.09 (0.28) -0.17* (0.04) -0.17* (0.04) -0.02 (0.81) 0.39**

(0.00)

0.06 (0.39) 0.06 (0.48) 0.06 (0.38) -0.01 (0.89)

DistrictB 0.05 (0.54) 0.10 (0.26) -0.17* (0.05) -0.08 (0.30) -0.07 (0.34) -0.18**

(0.01)

-0.30**

(0.00)

-0.05 (0.51) -0.16* (0.03) -0.01 (0.89)

DistrictC 0.02 (0.81) 0.15 (0.06) -0.14 (0.07) -0.21**

(0.01)

-0.14* (0.04) -0.14* (0.02) -0.24**

(0.00)

0.06 (0.38) 0.18**

(0.01)

0.20** (0.00)

Years since

graduation

-0.03 (0.68) -0.07 (0.44) -0.02 (0.86) -0.01 (0.92) 0.01 (0.94) 0.14* (0.05) 0.16* (0.03) 0.04 (0.62) 0.05 (0.56) -0.09 (0.26)

Number of pa-

tients

-0.2* (0.02) -0.32**

(0.00)

-0.17 (0.08) -0.05 (0.56) 0.12 (0.15) 0.11 (0.15) 0.33**

(0.00)

-0.03 (0.71) -0.09 (0.3) -0.08 (0.33)

Risk Adj. Index 0.38**

(0.00)

0.11 (0.28) 0.30**

(0.00)

0.21* (0.01) -0.12 (0.09) 0.14* (0.03) 0.18* (0.01) -0.26**

(0.00)

-0.18* (0.02) -0.04 (0.65)

Specialized 0.07 (0.28) -0.01 (0.87) 0.11 (0.11) -0.07 (0.34) 0.05 (0.4) -0.08 (0.16) 0.04 (0.54) -0.03 (0.6) 0.00 (0.99) -0.08 (0.21)

In any network 0.11 (0.14) -0.15 (0.07) -0.13 (0.10) -0.09 (0.26) 0.37**

(0.00)

0.04 (0.56) -0.12 (0.09) 0.01 (0.94) -0.11 (0.12) -0.10 (0.19)

In group 0.04 (0.56) -0.04 (0.57) 0.03 (0.66) -0.06 (0.44) 0.04 (0.53) -0.07 (0.27) -0.08 (0.24) -0.08 (0.28) -0.13* (0.05) -0.05 (0.48)

R2 0.19 0.21 0.11 0.09 0.24 0.34 0.30 0.14 0.24 0.14

Adj R2 0.15 0.16 0.06 0.04 0.21 0.31 0.27 0.09 0.21 0.10

* p< 0.05; ** p< 0.01 466

2

8

Table 6: Standardized regression coefficients of quadratic assignment procedure (QAP) regression models with the similarity of any two 467

GPs performance indicators as DVs 468

Dependent variables

∆ ∆ ∆ ∆ ∆ ∆ ∆ ∆ ∆ ∆ Pharmaceut-

ical expendi-

ture

Home care

Multi-

professional

home care

Printed pre-

scriptions

Day-time

white codes

Total white

codes

Preventable

readmissions

Screening

paptest

Screening

average

Performance

index

Intercept 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Same Gender 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Grad YY diff -0.03 -0.07 0.00 -0.01 -0.01 -0.03 -0.05 0.16* 0.08 -0.11

Number pa-

tients diff

-0.09 -0.20** -0.14* 0.30** -0.13 -0.2** -0.02 0.08 0.30** -0.19**

Risk Adj diff -0.11* -0.13* 0.08 0.20** 0.01 -0.06 0.17** -0.19** 0.06 -0.11*

Both specia-

lized or not

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Same any dist 0.00 0.00* 0.00 0.00* 0.00 0.00 0.00 0.00 0.00 0.00

Both in any

network

0.00 0.00* 0.00 0.00* 0.00 0.00 0.00 0.00 0.00 0.00

Both in any

group

0.00 0.00* 0.00 0.00* 0.00 0.00 0.00 0.00 0.00 0.00

Both same

network

0.00 -0.01* -0.01 0.01* 0.00 0.00 0.00 0.01 0.00 0.01

R2 0.03 0.08 0.03 0.13 0.02 0.05 0.03 0.07 0.13 0.09

Adj. R2 0.03 0.08 0.03 0.13 0.02 0.05 0.03 0.07 0.13 0.09

* p< 0.05; ** p< 0.01 469