PEER REVIEW HISTORY ARTICLE DETAILS VERSION 1 - REVIEW

Transcription

1 PEER REVIEW HISTORY BMJ Open publishes all reviews undertaken for accepted manuscripts. Reviewers are asked to complete a checklist review form ( and are provided with free text boxes to elaborate on their assessment. These free text comments are reproduced below. TITLE (PROVISIONAL) AUTHORS ARTICLE DETAILS Trends in diagnostic patterns and mortality in emergency ambulance service patients in a population-based cohort study from the North Denmark Region Christensen, Erika; Bendtsen, Mette; Larsen, Thomas; Jensen, Flemming; Lindskou, Tim; Holdgaard, Hans; Hansen, Poul; Johnsen, Søren; Christiansen, Christian VERSION 1 - REVIEW Vivienne Tippett Director of Research School of Clinical Science Queensland University of Technology Brisbane Australia 07-Oct-2016 This paper provides additional evidence for the developing body of knowledge related to the impact of socio-demographic change in societies globally - it will provide part of an important baseline against which to measure adjustments to emergency services operations, both pre-hospital and inpatient Janet Bray Monash University, Australia 28-Oct-2016 This well written study aimed to examine whether incidence, patients characteristics and mortality of ambulance uses (transported to hospital) has changed over time. To do this the investigators linked ambulance data for patients attended and transported by ambulance to 3 registries to obtain hospital diagnosis and mortality. The study is scientifically sound and the only major suggestions I have would be to strengthen the missing data aspect by conducting a sensitivity analysis of the best/worse case scenarios for those with missing mortality data. This is important given data is not missing randomly but was greater in the earlier period. Other minor suggestions below: 1. Abstract should state patient transported in aim (as per main paper). 2. Methods might be useful to list the data sources together so they can be easily identified. 3. Methods did you study exclude hospital transfers? 4. Results second sentence is confusing and suggest deleting. 5. Given the change in ICD10 codes related to the study were similar trends seen when 2013/2014 data is excluded?

2 6. The discussion is well written but the implications of change in exposure to the types of cases seen by EMS could also be discussed (e.g. see Dyson Circulation: Cardiovascular Quality and Outcome 2016:9: ). Annette Kjær Ersbøll University of Southern Denmark National Institute of Public Health Denmark 07-Feb-2017 Page 7, line 28: It is stated that patients with more than one call are included. How did authors handle multiple calls and contacts from the same patient? Page 7, line 39: Primary hospital diagnosis - do authors mean main diagnosis, or the first diagnosis? Further down in the description it is mentioned that primary diagnosis is the main diagnosis. Consider change wording to main diagnosis. Page 7, line 59 - page 8, line 2: Authors used Poisson regression to model the incidence rate of emergency ambulance service patients with year, age and sex. Did authors use all call for a patients or just the first call?? A patient (e.g. an old patient with high CCI) who has many call/transportations many times will have a large influence on parameter estimates Mårten Sandberg Air Amb. Dept., Oslo Univ. Hospital and Faculty of Medicine, Univ. of Oslo, Norway 22-Apr-2017 I appreciate the opportunity to review this manuscript by professor Christensen and coworkers. The authors have studied the increasing need for ambulance transport in a Danish region from 2007 to The results are interesting and will probably be applicable in similar systems also in other countries. The manuscript is well written and - as far as I can tell not being a statistician - the use of statistics is correct but may be wise of BMJ Open to consider expert evaluation regarding this. My major objection regardind the manuscript is actually that it is too detailed. I do not believe that a casual reader will be interested in the level of detail that is presented and I believe that the manuscript will profit from a reduction in the number of illustrations. Furthermore, it is quite disturbing that a regional administartive recommendation of applying specific ICD-10 codes was issued in 2013 resulting in a large increase in use of ICD-10 chapter 18 and 21 codes maling it more difficult to interprete the observed change in patient diagnoses in the study period. I think that the quality and applicability of the study would improve if the study period was changed to 2007 to

3 2013. I have some comments and questions: Some places the authors write prehospital, other places prehospital. Please be consistent. It is unclear whether all ambulance transports are included independent of degree of emergency. Please clarify this. Page 7, line 20: It could be an idea to very briefly describe Charlson Comorbidity Index (CCI) for readers unfamiliar with the concept. Page 9, line 15: I have some problems interpreting the first paragraph of the Results section. In the first two sentences, it is stated that 201,996 ambulances were dispatched and that each of these transported a patient. However, in the next sentence it is stated that that 17,276 patients were not transported. Please rephrase to avoid ambiguity. The flow chart (figure 1) is superfluous since identical information has been presented in the text. Figure 1 can therefore be deleted. Would fig. 2 be easier to interprete if 2007 and 2014 were shown in the same figure? The differences in age distribution for the two years would be easier to see. There are some many numbers presented that it is quite difficult to see the broad picture. Table 1 is one example. Is it necessary to present number of patients for each and every ICD-10 chapter for each year? I think that this table can be included as supplemental material and that a simplified table showing the distribution for a typical year can be included. The relative distribution of diagnoses was stable during the study period, as the authors indicate themselves in the manuscript. The number of patients with other symptoms did however almost double from 2013 to 2014, but does that change reflect a change in the patient population or a change in the way the ICD-10 diagnoses were used (seecomment above)?

4 Page 11, line 14: The information presented here regarding a relative increase in the number of patients with a CCI 3 is interesting and sufficient. The entire table on page 11 can be deleted. Page 12, table 3: This table can be deleted (or be available as supplemental material) but keep figure 6 that is much more illustrative for the casual reader. But change the legend of figure 6 since it depicts day-to-day mortality (between 1 and 30 days) and not just 1- and 30-day mortality. Page 12, line 31: You show a dramatic decrease in 1-day mortality (and 30-dagys mortality) for patients with other symptoms. Is that patients with a ICD-10 XXVIII diagnosis? If so, the number of patients in this group almost doubled between 2013 and 2014, as mentioned earlier. Is the mortality-reduction between 2014 and 2007 real or does it reflect a changed coding behaviour? Addressed on page 13, line 50, but not really explained.was there a significant mortality reduction if you compare the 2013 numbers with 2007? As stated before, it would be easier to discuss the findings if the study period was changed to 2007 to Page 14, line 4: The authors claim that patients with missing civil registration number may be more or less severely ill or injured, and may introduce bias in either direction. In theory, that is correct, but isn t possible for the authors to investigate this because I assume they have all the necessary information about the patient, diagnoses and prehospital interventions, the only piece of information that is missing is the civil registration number? Is it or is it not possible for the authors to compare the two groups the ones with and the ones without civil registration number? It would be of interest to identify in what direction if any one can suspect bias. Page 14, line 33: The authors state that the ambulance response times did not decrease in the study period. If the authors mean that the response time did not change, say so. If the authors mean that the response time increased, again, please say so. Page 14, line 35: The largest decrease in 30-day mortality was found among patients with cardiovascular disease. However, the largest decrease in1-day mortality was found in patients with other symptoms. Please clarify this in the text.

5 Page 15, line 15: I believe that the authors of the publications the authors of the present manuscript refer to (ref ) question the need for the number of emergency dispatches to patients with all kinds of illnesses and injuries they do not question the medical necessity of emergency ambulances as such. Please rephrase. Page 16, line 2. Replace in hospitals Denmark with in hospitals in Denmark. References Please use the correct abbreviations for the journals you are referring. A large number of the references are incorrect. VERSION 1 AUTHOR RESPONSE Reviewer: 1 Reviewer Name: Vivienne Tippett Institution and Country: Director of Research, School of Clinical Science, Queensland University of Technology, Brisbane, Australia This paper provides additional evidence for the developing body of knowledge related to the impact of socio-demographic change in societies globally - it will provide part of an important baseline against which to measure adjustments to emergency services operations, both pre-hospital and inpatient We are grateful for the valuable comment, and this was exactly what we were aiming at. Within prehospital care, there are few population-based studies covering the entire EMS patient population. Reviewer: 2 Reviewer Name: Janet Bray Institution and Country: Monash University, Australia 1) This well written study aimed to examine whether incidence, patients characteristics and mortality of ambulance uses (transported to hospital) has changed over time. To do this the investigators linked ambulance data for patients attended and transported by ambulance to 3 registries to obtain hospital diagnosis and mortality. The study is scientifically sound and the only major suggestions I have would be to strengthen the missing data aspect by conducting a sensitivity analysis of the best/worse case scenarios for those with missing mortality data. This is important given data is not missing randomly but was greater in the earlier period. Thank you for your very useful comments, and we do agree that the missing data due to missing patient identity (no civil registration number) is the major weakness of our study. We followed your advice and made a worst-case-scenario sensitivity analyses on mortality data with respect to the first contact of each patient with known identity and including the patients with missing data based on the

6 assumption that: 1) mortality among the patients with missing civil registration number in 2007 was set to be like the mortality in 2014 for those with civil registration number (thus simulating lower mortality in 2007) And 2) mortality among the patients with missing civil registration number in 2014 was set to be like mortality in 2007 for those with civil registration number (thus simulating higher mortality in 2014) The sensitivity analyses showed that 1-day mortality in 2007 and 2014 was 2.17 (95% CI: 1.95 to 2.40) and 1.49 (95% CI: 1.31 to 1.68) (p<0.001) and 30-day mortality in 2007 and 2014 was 4.67 (95% CI: 4.36 to 5.00) and 4.15 (95% CI: 3.85 to 4.46). (p=0.01). Thus, it did not change our conclusion. 2) Other minor suggestions below: Abstract should state patient transported in aim (as per main paper). We corrected this (p 3): This study aims to examine time trends in diagnoses and mortality among patients transported with emergency ambulance to hospital. 3) Methods might be useful to list the data sources together so they can be easily identified. We have listed all data sources in a separate paragraph (p 6): Data sources: We used the following data sources. The ambulance dispatch logistic system provided us with data on dispatched emergency ambulances, patient identity (including age and sex), time of ambulance transports, and destination hospital. Statistics Denmark provided data on the background population divided into calendar year, age, and sex. [11] The regional patient administrative system gave us information on hospital contacts, including contact to emergency departments, outpatient clinic and hospital admittance, and main diagnoses coded according to the International Classification of Diseases, 10th edition (ICD-10). [12] Danish hospitals routinely collect and report ICD-10 diagnoses to the regional patient administrative registry. The regional registries report data to the Danish National Patient Registry, which has been validated for research as an overall sound data source. [6,13] From the Danish Civil Registration System, we retrieved data on the vital status on patients who were residents in the North Denmark Region during the study period. 4) Methods did you study exclude hospital transfers? We only included calls from the 112, thus excluding calls from hospitals for inter-hospital transfers. However, as described in the Methods, among the patients calling 112, we identified all patients where the first hospital diagnosis was the non-specific factors influencing health status and contact with health services (ICD-10 chapter 21). For these patients, we also included information on subsequent hospital transfers during the same hospital stay (typically to higher level of specialization). In case of the patients were registered with a more specific diagnosis after transfer, we registered this as the main diagnosis. Please, see the text p 7: In cases where, the main hospital diagnosis was factors influencing health status and contact with health services (ICD-10 chapter 21, abbreviated other factors ), which includes tentative observation for diagnoses, we searched for more specific ICD-10 diagnoses during the same hospital stay; this also included individuals transferred to other departments or other hospitals during the stay. If we found an additional and more specific diagnosis, this was used in the analysis instead of ICD-10 chapter 21, other factors. 5) Results second sentence is confusing and suggest deleting.

7 The sentence has been deleted. 6) Given the change in ICD10 codes related to the study were similar trends seen when 2013/2014 data is excluded? Thank you, for pointing out, that this could be explained better. We described in the Results, that our main results was that there was no major changes in the diagnoses. The only exceptions were a gradual relative decrease in proportion of injured patients (ICD-10 chapter 19) throughout the entire study period, and a sudden increase ICD-10 chapter 18 diagnoses together with a change in the relative proportions of the two non-specific diagnostic chapters (ICD-10 chapter 18 and 21) between 2013 and We clarified this in the abstract (Results p 3). The only exceptions were a gradual relative decrease in proportion of injured patients (ICD-10 chapter 19) throughout the entire study period, and a sudden increase ICD-10 chapter 18 diagnoses together with a change in the relative proportions of the two non-specific diagnostic chapters (ICD-10 chapter 18 and 21) between 2013 and We clarified this in the abstract results p 3): The proportion of injuries gradually declined, non-specific diagnoses increased, especially the last year. And in Results (p 11): We noticed a shift from 2013 towards a sudden large increase in patients with the diagnosis of other symptoms (ICD-10 chapter 18) and at the same time fewer with the diagnosis other factors (ICD-10 chapter 21). Together the proportion of these two non-specific diagnoses increased from 28.8% in 2007 to 37.6% in The diagnostic pattern varied according to age. We found an increase among all age groups in number of patients diagnosed with other symptoms (ICD-10 chapter 18), but otherwise the relative distribution was similar in 2007 and 2014 (Figure 3). 7) The discussion is well written but the implications of change in exposure to the types of cases seen by EMS could also be discussed (e.g. see Dyson Circulation: Cardiovascular Quality and Outcome 2016:9: ). Thank you, for raising the important issue. We have added a comment in the discussion on this (p 17): A change in the pattern of diseases means a change in exposure to the types of cases seen by EMS, which can have implications for their routine in handling less frequent, critical emergencies. This may influence outcome, as shown in the study of Dyson et al of the association of exposure and survival after cardiac arrest. Reviewer: 3 Reviewer Name: Annette Kjær Ersbøll Institution and Country: University of Southern Denmark, National Institute of Public Health, Denmark 1) Page 7, line 28: It is stated that patients with more than one call are included. How did authors handle multiple calls and contacts from the same patient? Thank you for raising this important issue. We have clarified the issue further in the statistical section (p 8): As each ambulance takes care of one patient, the fundamental data unit in all analyses was a dispatched ambulance transporting a patient to the hospital after a 112 call (defined as an emergency ambulance service patient), including patients transferred several times to hospital after an emergency call during the eight year study period. And similarly in the Results (p 9) This study thus included emergency ambulance service patients transported to hospital with an ambulance, including patients with several 112 calls during the study period.

8 Secondly, we re-analyzed with cluster-robust variance estimation to overcome the problem of observations not being independent, due to the repeated callers the same patient, participating several times. This only changed little on the estimates of the confidence intervals and did not alter the conclusions. Thirdly, in response to your comments we have added the mortality analysis including only patient with their first hospital contact after calling for an ambulance as appendix 2. As described in the initial manuscript, this did not lead to substantially changes in mortality trends. Finally, our main purpose of the study was to describe the increasing number of emergency ambulances and the disease and comorbidity pattern seen from the EMS point of view, and therefore we kept the results concerning all patients, including patients calling several times for an ambulance as our main result. We developed the discussion (strengths and limitations) on this aspect (p 14): Patients with multiple 112 calls and transported to hospital by ambulance several times were included, because our main purpose of the study was to describe the increasing number of emergency ambulances and the disease and comorbidity pattern seen from the EMS point of view. The ambulance professionals meet each patient as a new patient, and only in very few cases, the same ambulance crew will attend the same patient several times. 2) Page 7, line 39: Primary hospital diagnosis - do authors mean main diagnosis, or the first diagnosis? Further down in the description it is mentioned that primary diagnosis is the main diagnosis. Consider change wording to main diagnosis. This has been corrected as requested. 3) Page 7, line 59 - page 8, line 2: Authors used Poisson regression to model the incidence rate of emergency ambulance service patients with year, age and sex. Did authors use all call for a patients or just the first call?? A patient (e.g. an old patient with high CCI) who has many call/transportations many times will have a large influence on parameter estimates Please see the response to comment no. 1. Reviewer: 4 Reviewer Name: Mårten Sandberg Institution and Country: Air Amb. Dept., Oslo Univ. Hospital and Faculty of Medicine, Univ. of Oslo, Norway 1) I appreciate the opportunity to review this manuscript by professor Christensen and coworkers. The authors have studied the increasing need for ambulance transport in a Danish region from 2007 to The results are interesting and will probably be applicable in similar systems also in other countries. The manuscript is well written and - as far as I can tell not being a statistician - the use of statistics is correct but may be wise of BMJ Open to consider expert evaluation regarding this. 2) My major objection regarding the manuscript is actually that it is too detailed. I do not believe that a casual reader will be interested in the level of detail that is presented and I believe that the manuscript will profit from a reduction in the number of illustrations. Thank you for a very useful comment. We followed your advice. Please see our responses below. 3) Furthermore, it is quite disturbing that a regional administrative recommendation of

9 applying specific ICD-10 codes was issued in 2013 resulting in a large increase in use of ICD- 10 chapter 18 and 21 codes making it more difficult to interpret the observed change in patient diagnoses in the study period. I think that the quality and applicability of the study would improve if the study period was changed to 2007 to We acknowledge this limitation of using routinely collected health care data, but we prefer not to omit the year The recommendations and instructions for coding diagnoses do change over time, both nationwide and regional. Around new and larger emergency departments were established, across Denmark following recommendations from the National Board of Health for centralization of emergency hospitals. Importantly, we reported the diagnoses by calendar year so it is easy for the reader to see the change. We did however, clarify that our main result was that there was no major changes in the diagnoses during the period. The only exceptions were a gradual relative decrease in proportion of injured patients (ICD-10 chapter 19) throughout the entire study period, and a sudden increase ICD-10 chapter 18 diagnoses together with a change in the relative proportions of the two non-specific diagnostic chapters (ICD-10 chapter 18 and 21) between 2013 and We rephrased it (Results, p 11): The relative distribution of diagnoses was stable during the study period, with a few exceptions (Table 1 and Figure 2). We found a gradual decrease in injuries (ICD-10 chapter 19) throughout the period and an increase in non-specific diagnoses, i.e. other symptoms (ICD-10 chapter 18). Together with other factors (ICD-10 chapter 1) the latter two were assigned for a considerable number, corresponding to about two third, of patients throughout the entire study period. We noticed a shift from 2013 towards a sudden large increase in patients with the diagnosis of other symptoms (ICD-10 chapter 18) and at the same time fewer with the diagnosis other factors (ICD-10 chapter 21). Together the proportion of these two non-specific diagnoses increased from 28.8% in 2007 to 37.6% in ) I have some comments and questions: Some places the authors write prehospital, other places pre-hospital. Please be consistent. This has been corrected. 5) It is unclear whether all ambulance transports are included independent of degree of emergency. Please clarify this. Thank you, and yes, all ambulances dispatched after a call was included, independent of the level of urgency. This has been clarified on the paragraph, Participants (p 6): We included all ambulance runs after 112 call, no matter the level of urgency as assessed by the dispatch centre. 6) Page 7, line 20: It could be an idea to very briefly describe Charlson Comorbidity Index (CCI) for readers unfamiliar with the concept. We agree, thank you for pointing this out. We added a short explanation (p 7): CCI is a scoring system with weighting 19 defined comorbidities, specific medical chronic diseases, among others specific cardiac, cerebrovascular, liver and renal diseases, diabetes and cancer. CCI of zero corresponds to no comorbidity and CCI>3 to severe morbidity. 7) Page 9, line 15: I have some problems interpreting the first paragraph of the Results section. In the first two sentences, it is stated that 201,996 ambulances were dispatched and that each of these transported a patient. However, in the next sentence it is stated that that 17,276 patients were not transported. Please rephrase to avoid ambiguity. There was an ambulance sent to the scene to all the patients, including the 17,276 patients, but these

10 patients were not transported from the scene to hospital. We deleted the second sentence in order to improve the readability. 8) The flow chart (figure 1) is superfluous since identical information has been presented in the text. Figure 1 can therefore be deleted. Figure 1 has been deleted. 9) Would fig. 2 be easier to interpret if 2007 and 2014 were shown in the same figure? The differences in age distribution for the two years would be easier to see. We have revised fig.2 as requested. 10) There are some many numbers presented that it is quite difficult to see the broad picture. Table 1 is one example. Is it necessary to present number of patients for each and every ICD- 10 chapter for each year? I think that this table can be included as supplemental material and that a simplified table showing the distribution for a typical year can be included. We thank for your advice. We however, believe that it is important to be able to get an overview over the years, but in the revised table 1 we only present the six most frequent diagnoses. The original complete table 1 is now found as appendix 1. 11) The relative distribution of diagnoses was stable during the study period, as the authors indicate themselves in the manuscript. The number of patients with other symptoms did however almost double from 2013 to 2014, but does that change reflect a change in the patient population or a change in the way the ICD-10 diagnoses were used (see comment above)? Please see response above to your comment no 3). 12) Page 11, line 14: The information presented here regarding a relative increase in the number of patients with a CCI 3 is interesting and sufficient. The entire table on page 11 can be deleted. The table has been deleted as requested. 13) Page 12, table 3: This table can be deleted (or be available as supplemental material) but keep figure 6 that is much more illustrative for the casual reader. But change the legend of figure 6 since it depicts day-to-day mortality (between 1 and 30 days) and not just 1- and 30- day mortality. This change has been implemented. 14) Page 12, line 31: You show a dramatic decrease in 1-day mortality (and 30-dagys mortality) for patients with other symptoms. Is that patients with a ICD-10 XXVIII diagnosis? If so, the number of patients in this group almost doubled between 2013 and 2014, as mentioned earlier. Yes, other symptoms refer to ICD10 chapter 18 (we added the chapter number in the text). The main reason was a nationwide revision of the Danish ICD-10 version as described in manuscript (Statistical analyses): Furthermore, a second sensitivity analysis was performed for overall mortality. This excluded diagnoses described in chapter 18 of the ICD-10 (symptoms and abnormal findings, not classified elsewhere), referred to as other symptoms, because the Danish National Board of

11 Health revised the Danish version of ICD-10 in 2012 and removed the subcategory of unknown dead and found dead (ICD-10, chapter 18, R.96- R99.9)." In the Results, we wrote (p 13): The second sensitivity analysis of mortality, which excluded the diagnoses of chapter 18 in the ICD- 10 showed no decrease in the unadjusted RR of the 30-day mortality, whereas the age- and sexadjusted RR showed a decrease in both the 1- and 30-day mortality; the latter is in line with the primary analyses including all ICD-10 diagnoses. So, no, the overall decrease in mortality was not explained by the large decrease in mortality among patients with ICD-10 chapter 18 diagnoses, when adjusted for age- and sex. 15) Is the mortality-reduction between 2014 and 2007 real or does it reflect a changed coding behaviour? Addressed on page 13, line 50, but not really explained.was there a significant mortality reduction if you compare the 2013 numbers with 2007? As stated before, it would be easier to discuss the findings if the study period was changed to 2007 to Please see response to comment no 14. The change in mortality is partly explained by the change from National Board of Health in the Danish version of ICD-10 with omission of unknown dead and found dead, which also influenced the observed overall difference in mortality, but we made a sensitivity analysis, excluding this chapter 18, and it did not change the conclusion. Furthermore, it should be noted that, the increase in number of patients, was not only among chapter 18 diagnoses, but almost ALL ICD10 diagnosis chapters increased during the years, please see table 1 and appendix 1. Part of the increase in number of patients can be explained by the increase in number of patients with civil registration number, but not entirely, as the increase in registration number was 13% compared to the 67% increase in total number of patients (please, see Results, p 9) 15) Page 14, line 4: The authors claim that patients with missing civil registration number may be more or less severely ill or injured, and may introduce bias in either direction. In theory, that is correct, but isn t possible for the authors to investigate this because I assume they have all the necessary information about the patient, diagnoses and prehospital interventions, the only piece of information that is missing is the civil registration number? Thank you for raising an important issue. Our main aim of the study was to investigate the final diagnoses made at the hospitals in emergency call 112 patients. In prehospital care the ambulance professional do not apply diagnoses, they only note main complaints, examinations and treatment in the patient record. In case of involvement of prehospital doctors, preliminary ICD-10 diagnoses may be applied. A doctor is involved about 20% of the patients, but the diagnoses will typically be tentative, not the final diagnoses. Furthermore, the proportion with civil registration number improved during the study period. 16) Is it or is it not possible for the authors to compare the two groups the ones with and the ones without civil registration number? It would be of interest to identify in what direction if any one can suspect bias. No, unfortunately, we cannot compare hospital diagnoses nor the mortality for patients without civil registration number, as this is the only link to both the patient administrative registry and the Danish Civil Registry (vital status, date of death). We cannot retrieve valid information on death from the Prehospital Patient Record, as only the doctors, not the ambulance professionals legally can pronounce a patient s death. We mention this in the discussion (p ) The civil registration number is the only link to both the registry of hospital diagnoses and the Danish Civil Registry (vital status, date of death), and it is not possible to retrieve valid information on death

12 from the Prehospital Patient Record, as only the doctors, not the paramedics legally can pronounce a patient s death. 17) Page 14, line 33: The authors state that the ambulance response times did not decrease in the study period. If the authors mean that the response time did not change, say so. If the authors mean that the response time increased, again, please say so. Thank you, this comment lead us to perform analyses of response times. The response time increased significantly from mean 9.60 minutes (95% CI: 9.50 to 9.71) in 2007 was to minutes (95% CI: to 10.71) in We added that in the results and discussion (p. 15): The ambulance response time actually increased from mean 9.60 minutes (95% CI: 9.50 to 9.71) in 2007 to minutes (95% CI: to 10.71) in ) Page 14, line 35: The largest decrease in 30-day mortality was found among patients with cardiovascular disease. However, the largest decrease in1-day mortality was found in patients with other symptoms. Please clarify this in the text. In relative percentages, this is correct, but the major explanation for the lower mortality among patients with other symptoms was the revision of the ICD-10 chapter 18 diagnoses. To illustrate this, we changed the order of discussion of ICD-10 chapter 18 and ICD-10 chapter 9 (cardiovascular) (p 15). The decline in the mortality among patients with other symptoms (ICD-10 chapter 18) was mainly explained by the nationwide revision of the Danish ICD-10 with removal of the subcategories of dead of unknown cause and found dead in chapter 18. However, this was not the explanation of the overall decrease in mortality, as the sensitivity analysis excluding ICD-10 chapter 18 confirmed the findings of decline in the relative risk of mortality, when adjusted for age-and sex. Apart from chapter 18, the largest decrease in mortality was found among patients with cardiovascular disease. 19) Page 15, line 15: I believe that the authors of the publications the authors of the present manuscript refer to (ref ) question the need for the number of emergency dispatches to patients with all kinds of illnesses and injuries they do not question the medical necessity of emergency ambulances as such. Please rephrase. Thank you, very much we rephrased this (p. 15) several studies question the medical necessity of the number emergency ambulances to patients with all kinds of illnesses and injuries 20) Page 16, line 2. Replace in hospitals Denmark with in hospitals in Denmark. This has been corrected. 21) References Please use the correct abbreviations for the journals you are referring. A large number of the references are incorrect. Thank you for bringing this to our attention, the reference style has been revised and corrected

13 VERSION 2 REVIEW VIVIENNE TIPPETT QUEENSLAND UNIVERSITY OF TECHNOLOGY BRISBAQNER AUSTRALIA 14-Jun-2017 Thank you for this interesting paper, your work supports findings from elsewhere in the world as you acknowledge. The use of linked data across the healthcare spectrum is to be commended and the only process likely to allow us to interrogate the actual impact of changes in healthcare demographics & demand for ambulance services effectively. There are some minor grammatical corrections to be made but I imagine these will be picked up in your final editing process. Dr Janet Bray Monash University 22-Jun-2017 I am satisfied with the response from authors. Mårten Sandberg Air Amb. Dept., Oslo Univ. Hospital, Oslo, Norway 03-Jun-2017 The authors have - to a large degree - made the changes suggested in the previous round. I still think that the tables can be simplified and that the results would be easier to interprete if the study period ended in However, the results are presented in an acceptable way in the present manuscript and I leave it to the discretion of the editor to decide if the figures and tables shall be changed as suggested by me in the first revision. BMJ Open: first published as /bmjopen on 21 August Downloaded from on 7 March 2019 by guest. Protected by copyright.