Skip to main content

Table 4 Results of the conditional logit model

From: Public preferences regarding data linkage for research: a discrete choice experiment comparing Scotland and Sweden

Attribute and level

Estimate coefficient

(standard error)

Scotland

Sweden

Country comparisonb

Researchers:

University researchers

0.214*** (0.05)

0.168** (0.05)

0.047 (0.07)

University /health service staff

0.500*** (0.05)

0.312*** (0.05)

0.188** (0.07)

University/health service staff/government

0.445*** (0.05)

0.337*** (0.05)

0.108 (0.07)

University/health service staff/government/commercial2

Base levela

Data to be linked:

Primary care linked to other health records

0.918*** (0.05)

0.706*** (0.05)

0.212** (0.07)

Health records linked to social care/education records

0.664*** (0.05)

0.403*** (0.05)

0.261*** (0.07)

Health records linked to social care/education/employment/benefits records

0.407*** (0.05)

0.171*** (0.05)

0.236** (0.07)

Health records linked to social care/education/ employment/benefits records/private sector

Base levela

Purpose:

Direct benefits for the people whose information is used

0.322*** (0.04)

0.254*** (0.04)

0.068 (0.06)

Research conducted if it will have general public benefits

0.548*** (0.04)

0.430*** (0.04)

0.118* (0.06)

Research for any reason

Base levela

Profit-making:

Nobody profits

0.326*** (0.05)

0.171*** (0.05)

0.156* (0.07)

Profit shared with the public

0.579*** (0.05)

0.397*** (0.05)

0.182* (0.07)

Profit invested into public services

0.739*** (0.05)

0.506*** (0.05)

0.233** (0.07)

Profit goes to those doing the research

Base levela

Oversight:

Overseen by independent body

0.346*** (0.05)

0.420*** (0.05)

−0.074 (0.07)

Overseen by relevant public service

0.265*** (0.05)

0.457*** (0.05)

−0.192** (0.07)

Overseen by Government

0.066 (0.05)

0.289*** (0.05)

−0.223** (0.07)

Overseen by the organisation undertaking the research

Base levela

Constant

0.886*** (0.08)

0.620*** (0.08)

0.266* (0.12)

Number of observations

18,072

17,532

35,604

  1. *p < 0.05; **p < 0.01; ***p < 0.001
  2. a Each attribute used categorical levels, which were dummy coded relative to a base level (Table 1) that was deemed to be the ‘worst’
  3. b The country comparison model, estimated using pooled data using a condition logit model, included interaction terms between dummy variables that identified the respondent’s nationality (1 = Scottish) and each attribute level