Back to History

Fields Changed

Registration

Field Before After
Last Published June 26, 2018 08:14 AM November 04, 2024 08:46 AM
Study Withdrawn No
Data Collection Complete Yes
Final Sample Size: Number of Clusters (Unit of Randomization) Part 1: 112,859 individuals. Part 2: 6,059 clusters (6,906 individuals clustered by residential address)
Was attrition correlated with treatment status? No
Final Sample Size: Total Number of Observations Part 1: 112,859 individuals. (137 individuals died or left the region before we extracted address information (for administrative reasons, address data was extracted after the randomization date. One individual chose to opt out from the study after randomization. In accordance with the recommendation from the regional ethical board, we gave all individuals the option to not be a part of the study, by announcing the project in two local newspapers in August 2015 (i.e., after the interventions). This is a standard procedure when using register data in Sweden. Note that the advertisements did not mention neither the information campaign nor the experimental set up. Part 2: 6,906 individuals. Among these we have complete information from the health authority's registers for all but one individual.
Final Sample Size (or Number of Clusters) by Treatment Arms In Part 1: 6,906 individuals (6,059 clusters)
Is there a restricted access data set available on request? Yes
Is data available for public use? No
Intervention (Public) There are three parts to the experiments: Part 1: In this part, the regional health authority in the region of Skåne (located in the south-west of Sweden) - Region Skåne - provides 1 percent of the population that are over 18 years old with information about the primary care center that they are listed at and its three geographically nearest competitors. The information is sent by mail, and contains information about, among other things, opening hours, quality ratings, and special competences. Within the 1 percent group, there are two treatments: one group receives only information about primary care centers (T1), the other receives this information along with a pre-paid form that notifies the new and old provider about the change (T2). Part 2: The procedure for this part is similar to the one described for Part 1, it is basically only the target group that differs. In Part 3, information will be sent specifically to individuals who have recently moved to Skåne. There may be such individuals also in the sample for Part 1, but they are a very small subset. There are three parts to the experiments: Part 1: In this part, the regional health authority in the region of Skåne (located in the south-west of Sweden) - Region Skåne - provides 1 percent of the population that are over 18 years old with information about the primary care center that they are listed at and its three geographically nearest competitors. The information is sent by mail, and contains information about, among other things, opening hours, quality ratings, and special competences. Within the 1 percent group, there are two treatments: one group receives only information about primary care centers (T1), the other receives this information along with a pre-paid form that notifies the new and old provider about the change (T2). Part 2: The procedure for this part is similar to the one described for Part 1, it is basically only the target group that differs. In Part 2, information will be sent specifically to individuals who have recently moved to Skåne. There may be such individuals also in the sample for Part 1, but they are a very small subset.
Experimental Design (Public) The premise for Part 1 is that the regional health authority has to treat all primary care centers equally, in order to be neutral in terms of competition on this market. Using the random number generator within Stata (which we use for all randomizations), we therefore first draw 11 percent of listed individuals over 18 from each of the 150 primary care centers. Then, we randomly select 9.0909091 percent of these 11 percent, which constitute the treatment group. In total, this procedure implies that 1 percent of the total population over 18 will be treated. The remaining 10 percent constitutes the control group. Among the treatment group, we then randomly assign 25 percent to the group that only receives information (T1), and 75 percent to the group that receives information and a pre-paid change form (T2). Several details about how we will conduct the randomization for Part 2 and 3 are presently not known. We will use a cluster randomized trial for Part 3, but a detailed description of the procedure will have to be added at a later date. The premise for Part 1 is that the regional health authority has to treat all primary care centers equally, in order to be neutral in terms of competition on this market. Using the random number generator within Stata (which we use for all randomizations), we therefore first draw 11 percent of listed individuals over 18 from each of the 150 primary care centers. Then, we randomly select 9.0909091 percent of these 11 percent, which constitute the treatment group. In total, this procedure implies that 1 percent of the total population over 18 will be treated. The remaining 10 percent constitutes the control group. Among the treatment group, we then randomly assign 25 percent to the group that only receives information (T1), and 75 percent to the group that receives information and a pre-paid change form (T2). In Part 2, we randomly select 6906 individuals (6059 clusters based on residential address) into treatment and control. The treated group recieves the same information and a pre-paid change form. Several details about how we will conduct the randomization for Part 3 are presently not known. For administrative reasons Part 3 is postponed.
Randomization Method Part 1: Using the random number generator within Stata (which we use for all randomizations), we therefore first draw 11 percent of listed individuals over 18 from each of the 150 primary care centers. Then, we randomly select 9.0909091 percent of these 11 percent, which constitute the treatment group. In total, this procedure implies that 1 percent of the total population over 18 will be treated. Part 1: Using the random number generator within Stata (which we use for all randomizations), we therefore first draw 11 percent of listed individuals over 18 from each of the 150 primary care centers. Then, we randomly select 9.0909091 percent of these 11 percent, which constitute the treatment group. In total, this procedure implies that 1 percent of the total population over 18 will be treated. Update: Part 2: We randomly assigned clusters of individuals (based on the residential adress) to treatment and control groups using random number generator within Stata. These individuals had that entered the primary care enrolment register held by Region Skåne between Feb 4 and May 11.
Randomization Unit The unit of randomization is for Part 1 individuals, Part 2 is address (i.e., household) (or potentially postal code) and Part 3 antenatal care centers, and Part 3 individuals. The unit of randomization is for Part 1 individuals, Part 2 is address (i.e., household) and Part 3 antenatal care centers.
Planned Number of Clusters Part 2: Number of addresses/households in the sample is currently unknown (postal codes >2000) Part 3: 80 antenatal care centers. Part 2: Number of addresses/households in the sample is currently unknown. (Update: 6,059 clusters) Part 3: 80 antenatal care centers.
Planned Number of Observations Part 1: Total sample is 112 861 individuals. Part 2 and Part 3: to be added. Part 1: Total sample is 112 861 individuals. Part 2: Total sample 6,906. Part 3: yet to be implemented.
Sample size (or number of clusters) by treatment arms Part 1: T1 = 2,559, T2 = 7,700, and the control group = 102,602. Part 2 and Part 3: to be added. Part 1: T1 = 2,559, T2 = 7,700, and the control group = 102,602. Part 2: T: 3,454 Control: 3,452 and Part 3: yet to be implemented.
Keyword(s) Health Health
Building on Existing Work No
Back to top

Papers

Field Before After
Paper Abstract Consumer choice policies may improve the matching of consumers and providers, and may spur competition over quality dimensions relevant to consumers. However, the gains from choice may fail to materialise in markets characterised by information frictions and switching costs. We use two large-scale randomised field experiments in primary health care to examine if individuals reconsider their provider choice when receiving leaflets with comparative information and pre-paid choice forms by postal mail. The first experiment targeted a representative subset of the 1.3 million residents in a Swedish region. The second targeted new residents in the same region, a group expected to have less prior information and lower switching costs than the general population. The propensity to switch providers increased after the interventions in both the population-representative sample (by 0.6–0.8 percentage points, 10–14%) and among new residents (2.3 percentage points, 26%). The results demonstrate that there are demand side frictions in the primary care market. Exploratory analyses indicate that the effects on switching were larger in urban markets and that the interventions had heterogeneous effects on the type of providers chosen, and on health care and drug consumption.
Paper Citation Anell, A., Dietrichson, J., Ellegård, L. M. & Kjellsson, G. (2021). Information, switching costs, and consumer choice: Evidence from two randomized field experiments in Swedish primary health care. Journal of Public Economics, 196, 104390
Paper URL https://doi.org/10.1016/j.jpubeco.2021.104390
Back to top