Our findings confirm the hypothesis that increasing travel distance from population centroid of resident municipality to a casualty clinic is highly correlated with lower rates of face-to face consultations with a doctor, and with all other kinds of contacts with the casualty clinic. Only telephone consultations with a doctor increased slightly with increasing distance. The effect was most pronounced for non-urgent cases, but even acute events (red responses) were reduced by 22 per cent for people living 50 kilometres away. The proportion of first contacts that resulted in a face-to-face consultation with a doctor also decreased considerably with increasing distance. The trend was the same for all ages and both genders, but males and the oldest patients comprised a larger share of contacts with increasing distance. In contrast, teenagers living in the most remote municipalities had lower contact rates than expected. Multivariate analysis showed that availability of primary care doctors and education level were the only confounding factors that contributed to the variance in face-to-face consultation rate, although to a limited extent. We failed to identify any demographic or socioeconomic factors that influenced the rate of all first contacts with the out-of-hours service.
The data used in this study have many properties that make it well suited to study how geography affects the utilisation of out-of-hours services. The data set contains many cases and many relevant variables. It includes almost all activity year-round for five years in a large area with a wide range of distances. Additional information on potential demographic and socioeconomic confounding factors was obtained from reliable official sources.
However, great care should be taken when discussing causal relationships in studies like this, in order to avoid the Ecological fallacy [13]. Potential confounders were not analysed on individual level, but on data aggregated for year and municipality. There is a risk that important information might have disappeared in the aggregation process. It cannot be completely ruled out that some of our findings can be explained by unknown confounders. On the other hand, the strength of the associations observed, both quantitatively and by statistical significance, suggests a true causal relationship between distance and utilisation of out-of-hours services.
Another limitation is some inaccuracy of the Watchtower data. The information was most often recorded per telephone, and the patients were not identified in our database. Multiple contacts with the same patient on the same day were possible. The priority grade was based on initial triage performed by nurse upon initial contact [8], and did not necessarily reflect the actual need of all patients. However, we find it unlikely that these errors correlate with distance and thus cause recording bias.
The distance from the population centroid to the casualty clinic is a good proximity measure for all municipalities. A notable exception is Arendal city itself. Because the location of the clinic and the population centroid of Arendal municipality almost coincide, the estimated average travel distance is smaller than the actual one. There has been some discussion about the appropriateness of different distance measures. We performed post-hoc analyses on travel time and on distances from town halls in each municipality. No changes were observed when analysing alternative distance measures (see Additional files 6 and 7). Potential limiting traffic factors such as ferries, toll roads, mountain passes or extreme poor road quality are almost non-existent in Arendal out-of-hours district. The only two toll booths are both located at the outer borders, so that very few patients had to pass them to get to the casualty clinic.
The results of this study indicate that travel distance to the casualty clinic is a major barrier for the use of the out-of-hours service in this district. Both over-utilisation near the casualty clinic and under-utilisation in remote municipalities probably contribute to the large differences in contact and consultation rates. The vanishingly small impact of the investigated demographic and socioeconomic factors indicates that the travel costs in terms of time and money is the key factor when a patient decides whether to contact the out-of-hours service or not.
It should be of particular concern that also the rate of red responses and cases needing doctor assisted ambulance action also decreased with distance. We do not know how many of the most seriously ill patients that are taken directly to the hospital bypassing the casualty clinic, and whether this happens more often in rural areas, e.g. by ambulance helicopters. Rates of hip fractures and cardiovascular disease treated in hospital did not contribute to the variance of red response rate in a post-hoc multivariate analysis. The observed lower rate of the most acute cases in the remote municipalities raises serious safety questions. In a Scottish cohort study, the mortality risk from myocardial infarction increased twofold for patients living more than 14.5 kilometres away from hospital, non-admittance was a major contributing factor [14].
The proportion of the oldest patients increase with increasing distance, reflecting higher needs in this group, while the decrease in contact rate among teenagers may represent underutilisation of the out-of-hours service.
It would be interesting to know whether our results can predict the contact and consultation rate in primary emergency healthcare elsewhere. The effects on main outcomes are in the same magnitude as previous findings. In a study on the effects on distances to hospital in the UK, the inpatient acute emergency episode rate (unadjusted) was reduced by 42% for people living 40 kilometres away from the hospital [15]. In our model, moving 40 kilometres away from the casualty clinic reduce the consultation rate by 47%. A study from Northern Ireland showed that people living 34 kilometres away had an odds ratio of face-to-face consultation of 0.50 compared with patients living close to an out-of-hours primary care centre [4]. The authors of this study chose to present the association as linear, but like in our study, a negative exponential function was found to fit the data best.
The emergency primary health care in Norway is organised differently than in most other countries. General practitioners constitute the mainstay of the out-of-hours services. Alternatives to the publicly organised scheme are almost non-existent. Norway is also sparsely populated compared to most other European countries. Arendal casualty clinic has been specifically chosen to be a part of the Watchtower project, because it is a typical Norwegian inter-municipal out-of-hours service.